128 77 2MB
English Pages 174 Year 2023
Applied Economics and Policy Studies
Xuezheng Qin Chee-Ruey Hsieh
Economic Analysis of Mental Health in China
Applied Economics and Policy Studies Series Editors Xuezheng Qin , School of Economics, Peking University, Beijing, China Chunhui Yuan, School of Economics and Management, Beijing University of Posts and Telecommunications, Beijing, China Xiaolong Li, Department of Postal Management, Beijing University of Posts and Telecommunications, Beijing, China
The Applied Economics and Policy Studies present latest theoretical and methodological discussions to bear on the scholarly works covering economic theories, econometric analyses, as well as multifaceted issues arising out of emerging concerns from different industries and debates surrounding latest policies. Situated at the forefront of the interdisciplinary fields of applied economics and policy studies, this book series seeks to bring together the scholarly insights centering on economic development, infrastructure development, macroeconomic policy, governance of welfare policy, policies and governance of emerging markets, and relevant subfields that trace to the discipline of applied economics, public policy, policy studies, and combined fields of the aforementioned. The book series of Applied Economics and Policy Studies is dedicated to the gathering of intellectual views by scholars and policymakers. The publications included are relevant for scholars, policymakers, and students of economics, policy studies, and otherwise interdisciplinary programs.
Xuezheng Qin · Chee-Ruey Hsieh
Economic Analysis of Mental Health in China
Xuezheng Qin School of Economics Peking University Beijing
Chee-Ruey Hsieh Independent Researcher Taichung
ISSN 2731-4006 ISSN 2731-4014 (electronic) Applied Economics and Policy Studies ISBN 978-981-99-4208-4 ISBN 978-981-99-4209-1 (eBook) https://doi.org/10.1007/978-981-99-4209-1 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023, corrected publication 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore
The original version of the book was revised: The book editors affiliation details has been updated. The correction to the book is available at https://doi.org/10.1007/978-981-99-4209-1_7
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Conceptual Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Contents of the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Other Causes of Rising Prevalence of Mental Illness in China . . . . . 1.5 Policy Updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1 1 3 5 7 12 15 17
2 The Prevalence of Depression and Depressive Symptoms Among Adults in China: Estimation Based on a National Household Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Data Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 Cross-tabulation Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.2 Regression Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Estimation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.1 Cross-tabulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.2 Regression Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Robustness Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6 Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19 19 21 24 24 24 25 25 31 34 39 41 42
3 Depression Hurts, Depression Costs: The Medical Spending Attributable to Depression and Depressive Symptoms in China . . . . . 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Background and Previous Research . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Data Source and Descriptive Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Estimation Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1 The Two-Part Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
45 45 47 49 55 55 vii
viii
Contents
3.4.2 The Four-Part Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.3 Estimating the Cost of Depression and Depressive Symptoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5.1 Regression Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5.2 Estimating the Medical Cost of Depression and Depressive Symptoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5.3 Robustness Checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6 Discussions and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56 57 59 59 63 66 73 74 75
4 The Hidden Costs of Mental Depression: Implications on Social Trust and Life Satisfaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 4.2 Research Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 4.2.1 The Rising Prevalence of Depression . . . . . . . . . . . . . . . . . . . 82 4.2.2 The Role of Trust in Economic Development . . . . . . . . . . . . . 84 4.2.3 Life Satisfaction and Well-Being . . . . . . . . . . . . . . . . . . . . . . . 86 4.3 Data and Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 4.3.1 Data Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 4.3.2 Sample Descriptive Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 89 4.4 Estimation Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 4.4.1 Baseline Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 4.4.2 Instrumental Variable (IV) Regression . . . . . . . . . . . . . . . . . . 95 4.4.3 Subsample Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 4.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 5 Relative Economic Status and the Mental Health Status Among Chinese Adults: Evidence from the China Health and Retirement Longitudinal Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Data and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1 Sample Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.2 Regression Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5 Discussions and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
113 113 114 116 116 118 120 120 123 129 133 136
Contents
ix
6 Understanding and Addressing the Treatment Gap in Mental Healthcare: Economic Perspectives and Evidence from China . . . . . . 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Conceptual Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Empirical Evidence from China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.1 Stigma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.2 Out-of-Pocket Costs of Mental Healthcare . . . . . . . . . . . . . . . 6.3.3 Mental Healthcare Resources . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.4 Diffusion of New Medical Knowledge and Technology in Mental Healthcare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4 Policy Options for Bridging the Treatment Gap in Mental Healthcare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4.1 Out of the Shadow: Information Campaign for the Awareness of Mental Illnesses . . . . . . . . . . . . . . . . . . . 6.4.2 Increasing the Public Investment in Mental Healthcare . . . . 6.4.3 Integrated People-Centered Health System . . . . . . . . . . . . . . . 6.4.4 E-health System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
158 159 160 161 162 162
Correction to: Economic Analysis of Mental Health in China . . . . . . . . . .
C1
141 141 143 145 145 147 148 152 157
Chapter 1
Introduction
1.1 Background The complex and dynamic relationship between health and wealth, the two primary dimensions of human development, has been widely recognized and is often called the health–wealth nexus. A noticeable example is the relationship between a country’s national income and population health, which are often measured by the gross domestic product (GDP) and the life expectancy at birth, respectively (Bloom and Canning 2000). On the one hand, people living in rich countries are expected to lead longer lives; on the other hand, good population health is an important driver to foster economic growth. Over the past several decades, a large body of research has estimated the effect of wealth on health as well as the impact of better health on economic growth (Weil 2014). These studies identified several important channels through which better health may positively contribute to economic growth: healthier workers are on average more productive and can produce more outputs giving the same size of labor force; longer life expectancy also leads to the increases in public and private education investment, higher savings rates, and labor force participation rates, which ultimately drives economic growth. The accumulation of empirical studies on this topic has not only increased our understanding on the contribution of population health on economic growth, but it also sheds light on the broader relationship between the healthcare system, health and wealth (McKee et al. 2009; Figueras and McKee 2012). Under this framework, the healthcare system is conceptualized as an important and integral part of the triangular relationship, and it has been seen as a productive socioeconomic sector rather than a drain on resources, as social and personal investment in healthcare can generate positive returns by promoting economic growth. However, most of these studies have focused on physical health, while the understanding on the role of mental health in this triad is limited. A plausible explanation for the paucity of economic research in mental health is that mental health has been described as an “invisible problem” in the international development (Mills 2018). © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 X. Qin and C.-R. Hsieh, Economic Analysis of Mental Health in China, Applied Economics and Policy Studies, https://doi.org/10.1007/978-981-99-4209-1_1
1
2
1 Introduction
Both clinical and household survey data show that the global prevalence of mental illness has been increasing over the past decades, and many international agencies such as the World Health Organization (WHO) and the World Bank have initiated calls to set mental health as a global development priority. As a result, mental health is now being included in the UN Sustainable Development Goals (SDGs). The inclusion of mental health in the SDGs highlights the importance to fill the research gap on the relationship between mental health and wealth. Compared to physical health, there are several characteristics of mental health that make its impact on the society as a whole more complex and ambiguous. First, the occurrence of mental illness is not limited to the elderly. Children, adolescent, and young adults are all vulnerable to various types of mental illness. As a result, the impact of mental illness on labor productivity and human capital investment pertains to all age groups and can be more substantial than the physical health. Second, certain types of mental illness such as depression can lead to suicide or other life-threatening events if the patients do not receive appropriate treatment, which in turn often imposes substantial loss and indirect costs to their family and the society. Third, unlike physical health conditions, the rate of diagnosis and treatment for mental illness are quite low due to the medical, social and economic barriers, indicating that the hidden costs of mental illness can be much larger than the direct treatment costs due to the underdiagnosis problem. Thus, multi-disciplinary studies are in urgent need for a better understanding of the unique complexities of the mental health problems. To fill this research gap, we have devoted efforts to explore the causes and consequences of rising prevalence of mental illnesses in China during the past several years. Figure 1.1 illustrates the ranking of the causes of loss in the Disease Adjusted Life Year (DALY) for the Chinese population in 1990 and 2019. As the figure shows, mental disorders (MDA) rose from ninth place in 1990 to the fifth place in 2019, indicating that mental health problems are gradually becoming a more dominant cause of health loss among the Chinese residents. The rapidly increasing prevalence and disease burden of mental illness generate formidable impact on China’s healthcare system and medical expenditure, calling for more awareness and responses from the policymakers, researchers, and the general public. To serve this purpose, the authors of this book conducted a series of academic and policy-oriented studies, attempting to explore the linkage between the population mental health and the macroeconomy in China and to propose a series of policies that help to address the under-diagnosis and under-treatment problems in China’s context. The aim of this book is to select and reorganize our previously published articles on this topic into a single collection so that we can draw a more comprehensive picture on the causes and consequences of mental health conditionals (particularly mental depression) in China. In this introductory chapter, we first provide a conceptual framework to analyze the economics of mental health in China. This framework serves as a roadmap to understand the importance of our previous efforts for making the research in mental health more visible and relevant to policy challenges. We then review several important papers that fit in with our research framework. These papers address the causes of mental illnesses that are closely related to the process of economic development in China, including internal migration, urbanization, and air pollution. Such papers
1.2 Conceptual Framework
3
Fig. 1.1 Causes of loss in the disease adjusted life years (DALY) among Chinese residents (rankings in 1990 and 2019)
contribute to closing the gap between the theoretical framework and the empirical evidence on the topic we address in this book. In addition, we also update the recent policy reforms and implementation progress in China. These policy updates highlight what have been done to address the mental health challenges during the past decades and what still remains to be done in the future. Based on the review of existing academic studies and policy progress, we further suggest the avenues for future research.
1.2 Conceptual Framework As shown in the pioneer work developed by a group of WHO experts, there is a triangular relationship between the health system, health, and wealth (McKee et al. 2009). This triangle is illustrated in Fig. 1.2, which yields two important policy implications. First, the health system, health, and wealth have a strong two-way link in that each can have a direct impact on the other two, and the causation of such impacts in each pair can go either way. Second, the triangular relationship can lead to a virtuous (or vicious) circle of mutually reinforcing effects in that the increase (or decrease) in the investment in health system can benefit (or undermine) both the health and wealth of a nation. We adapt this conceptual framework in this book and specifically focus on the role of mental health in the health system–health–wealth triad.
4
1 Introduction
Fig. 1.2 Triangular relationship between health system, mental health, and wealth
The upper part of Fig. 1.2 illustrates that the relationship between mental health and wealth is bidirectional. On the one hand, the rising prevalence of mental illness may lead to a slowdown of economic growth through several channels, including the increase in healthcare costs and the decrease in the labor supply, worker productivity, and human capital investment. On the other hand, the accumulation of wealth in the form of higher economic growth rates may yield both positive and negative impacts on the population’s mental health. For example, the increase in the percapita income may make people happier on average and less likely to suffer from mental illness, while the enlarged income inequality and other social problems that accompany the economic growth (e.g. urbanization and environmental pollution) may have detrimental effects on mental health, making people more likely to suffer from depression, anxiety and other types of mental illness. The left side of Fig. 1.2 illustrates the bidirectional relationship between mental health and the health system. WHO defines a health system as “all activities whose primary purpose is to promote, restore, or maintain health” (WHO 2000). From the perspective of economics, two essential activities are the key to make the system productive in promoting, restoring and maintaining mental health. One is the flow of money, or the financing activities, that secures the funds for the mental health sector. The other is the flow of medical goods and services, or the delivery of mental healthcare including the preventive, outpatient and hospital care, the pharmaceutical products, and medical devices. The bidirectional relationship also means that the healthcare system provides important inputs, including labor and capital, into the production of mental health, while the mental health outcomes in turn feed back into the mental health system to shape its policy framework and priorities on the resource allocation.
1.3 Contents of the Book
5
The right side of Fig. 1.2 illustrates the bidirectional relationship between wealth and the mental health system. On the one hand, a country’s economic performance is the key determinant of the fiscal capacity for allocating resources to the health system. Such choices as how much and what kinds of resources to commit to the health sector have a large impact on the performance of the country’s mental health services, such as their accessibility, affordability and the structure of mental healthcare financing. In certain countries such as UK and Denmark, the public sector devotes a substantial share of its resources to financing healthcare. By contrast, in other countries such as India and Mexico, the government spends much smaller shares on healthcare, leaving households to rely on private sources to finance their personal health services. Although public sectors in high-income countries generally allocate larger shares of resources to healthcare, there is appreciable variation in the methods of financing: some countries rely on general tax revenues as the major source of healthcare financing, while other countries use earmarked taxes (e.g. social security taxes, payroll taxes) for healthcare. The variation in healthcare financing methods across countries in turn have important implications for the efficiency and equity of the whole economy. On the other hand, the size of the health sector as measured by the national healthcare expenditure often accounts for 8–10% of GDP in high-income countries, indicating that the health system also makes a significant contribution to economic growth and the accumulation of national wealth. Specifically, the health sector is often made up of labor-intensive industries that serve as major sources of job creation, contributing to sustained employment growth in the economy. In addition, the health sector is also characterized as R&D intensive, which plays a significant role in promoting the advancement in medical technology. Such investment and activities in research and development in turn serves as the engine for innovation and economic growth. Furthermore, how the health system collects fund to finance healthcare and how it delivers services to the patients have important impacts on the efficiency and equity of the economy. For example, it has long been recognized that financing healthcare through general taxes helps to mitigate the income inequality within a country as compared to the method of out-of-pocket financing.
1.3 Contents of the Book In Chap. 2, we provide an overview on the prevalence and determinants of mental depression in China, with the special focus on how the prevalence varies with different demographic, socioeconomic, and geographic characteristics. Our estimates are based on the household survey data from the 2012 China Family Panel Studies (CFPS), which cover 40,000 respondents from 25 Chinese provinces. CFPS questionnaire is the first to contain the standard 20 questions of the Center for Epidemiologic Studies Depression Scale (CES-D), which is originally developed by Radloff (1977) and still one of the most widely used self-evaluation tests on the respondent’s mental health condition. Our findings indicate that about 4.08% of adults in China suffer
6
1 Introduction
from depression and 37.86% frequently experience depressive symptoms. Moreover, we find that the above prevalence rates are distributed unevenly across different geographical regions and subpopulations. After controlling for other factors, our analyses highlight four subpopulation groups that are vulnerable to mental health problems, including the elderly, the rural residents, the poor, and the low-educated people. Given the significant socioeconomic gradient in mental health, we hypothesize that the inequalities in income and education among regions and subpopulations may be an important contributing factor to the mental depression and depressive symptoms among the Chinese adult population. In Chap. 3, we provide the first nationally representative estimate of the medical costs attributable to depression and depressive symptoms among the adult population in China. On the basis of the same household survey data that we used in Chap. 2, we are able to quantify the impacts of depression and depressive symptoms on the personal healthcare costs of the survey respondents. Such estimation methods have the advantage of overcoming the potential estimation bias in traditional studies based on the patient survey data, which arises from two phenomenon commonly seen among the mental health patients: one is the under-diagnosis and under-treatment for mental health conditions, and the other is the co-existence of mental health conditions and other chronic physical conditions. Specifically, we adopt the two-part and fourpart models to characterize the heterogeneous cost impacts of mental depression for medical users and non-users. Our results indicate that about 6.9% of the total personal medical expenditure is attributable to depression, and an additional 7.8% of the expenditure is attributable to the depressive symptoms. Putting together, our empirical study shows that about 14.7% of the total personal medical expenditures in China are attributed to depression and depressive symptoms. We also find that the induced costs are not evenly distributed across regions and subpopulations, with women, the rural residents, the elderly people, and the low-educated groups paying a higher share of medical spending due to depression and depressive symptoms. In Chap. 4, we estimate the impacts of mental health on social trust and life satisfaction, respectively, using the same dataset mentioned above. We find that individuals who have a higher tendency for depression or depressive symptoms are less likely to trust other people, and they also have significantly lower life satisfaction than their counterparts who are mentally healthier. Our measures of trust include general trust and particularistic trust toward specific groups ranging from parents, neighbors, doctors, cadre, strangers, and foreigners. The measures of life satisfaction include the satisfaction of one’s family and one’s life, self-evaluation of one’s own and family’s social status, and the degree of confidence to one’s future. Given that trust is an essential component of social capital, which in turn is an important input to foster economic growth in general and innovation in particular, the reduction in trust induced by the increasing prevalence of depression imposes a significant cost to the society in terms of poor economic performance. Similarly, as life satisfaction has been widely recognized as an important measure of well-being, our study also highlights that the increase in the prevalence of depression leads to a reduction in the well-being that individuals can enjoy. An important implication of this study is that
1.4 Other Causes of Rising Prevalence of Mental Illness in China
7
the burden of mental health conditions is not limited to their direct healthcare costs but also includes the consequences on the social and economic well-being. In Chap. 5, we empirically investigate the question of whether income inequality plays an important role in accounting for the rising prevalence of mental illness in China. Specifically, we use the survey data from the 2013 China Health and Retirement Longitudinal Study to explore the relationship between one’s relative economic status and mental health among the middle-aged and elderly in China. In measuring the relative economic status, we use five reference groups for comparison, including one’s relatives, colleagues, schoolmates, neighbors, and people in the same city or county. Overall, the results provide strong evidence to support the argument that individuals with relatively lower economic status are more likely to be mentally depressed. We also find that the impact of relative economic status on the respondent’s mental health is stronger for the comparison against acquaintances than to the regional averages, and it is stronger for the unfavorable comparisons than the favorable ones, suggesting that such impacts are asymmetric. These results provide strong evidence to support the argument that wealth and its distribution have strong influences on mental health. As China enjoyed rapid economic growth over the past four decades, income and wealth inequality has become a significant side effect during this process, which in turn becomes a strong socioeconomic factor to account for the rising prevalence of mental illnesses in China. In Chap. 6, we address the potential causes of the treatment gap in mental health, which refers to the substantial unmet mental healthcare needs and propose several policy options for bridging the treatment gap in the context of China’s healthcare and socioeconomic framework. We first hypothesize that there are four hurdles to explain why so many people with mental illness go untreated, including the non-monetary cost imposed by social stigma, the out-of-pocket medical cost arising from the insufficient public funding, the time cost due to the poor availability of mental healthcare resources, and the slow diffusion of new medical knowledge and technology in the mental healthcare sector. Based on the available survey data such as CFPS and other government statistical reports, we find evidence to support our hypothesis on the four major hurdles in accessing mental healthcare. We then discuss four policy options for bridging the treatment gap in mental healthcare, including raising the awareness of mental illness through information campaigns, increasing the public investment in mental healthcare sector, developing an integrated people-centered primary healthcare system, and strengthening the e-health system.
1.4 Other Causes of Rising Prevalence of Mental Illness in China The chapters included in this book reflect the authors’ research efforts to increase the understanding of the causes and consequences of rising prevalence of mental illness in China. However, our works are far from complete. Several recent studies
8
1 Introduction
that fit in well with our conceptual framework also contribute to filling the gap in this line of literature. In this section, we provide a brief review of their major findings. Specifically, we focus on studies that explore how the population mental health is influenced by the three major by-products of rapid economic growth in China, including the labor migration from rural-to-urban areas, fast urbanization, and environmental pollution. It has been widely recognized that domestic migration, particularly the rural-tourban migration in search of better employment opportunities and higher wages, is one of the engines of rapid economic growth in China. Although this large-scale internal migration helps to provide the much-needed low-cost labor for the manufacturing firms in urban areas (particularly those in the coastal export-oriented industries) and hence is beneficial to the overall economic development, these migration activities may also have an adverse impact on the mental health of certain groups of people, including the left-behind children (LBC), the left-behind elderly parents, and the migrant workers themselves. Shi et al. (2016) use two follow-up surveys to examine the effects of parental migration on the mental health of the left-behind children. They choose two prefectures from the Gansu and Shaanxi provinces as their study sites and conduct surveys in 252 elementary schools. Specifically, they interview 19,934 students who are in the fourth and fifth grades. They measure the students’ mental health status by three internationally recognized psychological scales, including a comprehensive mental health test, a social anxiety scale for children, and a self-esteem scale questionnaire. Based on the comparison of parental migration status, this study is able to divide the sample students into two treatment groups and two comparison groups, which yields two important findings: first, parental out-migration has a negative effect on the mental health of LBC, including the increase in the level of anxiety and the decrease in self-esteem; second, there is no significant effect on the mental health of LBC when parents return home. Putting together, the study concludes that the negative impact of parental migration on children’s mental health cannot be offset directly by parental return, indicating the harm on the mental health of LBC may be long lasting and cannot be corrected in the short run. Scheffel and Zhang (2019) empirically investigate how the migration of an adult child affects the emotional health of elderly parents. They address this issue by using the data from China Health and Retirement Longitudinal Study, which is the same dataset we used in Chap. 5. In addition to the CES-D scores for depressive symptoms, this study also uses the elderly’s feeling of happiness and loneliness as measures for their emotional health. The results show that children’s migration is associated with lower levels of happiness as well as significantly higher levels of loneliness and depressive symptoms among the elderly parents. Specifically, they find that children’s migration reduces parents’ level of happiness by 6.6 percentage points and leads to a 3.4 percentage points higher probability of feeling loneliness. In addition, the CES-D score is raised by 0.71 points if an adult child migrates, which in turn pushes the average scores close to the clinical cut-off level of depressive symptoms.
1.4 Other Causes of Rising Prevalence of Mental Illness in China
9
Internal migration may also have adverse impacts on the mental health status of the migrant workers themselves, as they are usually engaged in the 3D (dirty, dangerous and demanding) work. In addition, due to the restriction of household registration system, the rural migrant workers are not always entitled to urban public resources such as healthcare services. Despite the above disadvantages, migrant workers in China still have chances to mitigate the risks of mental health problems through adaptation or self-investment. For example, the self-selection effect refers to the situation where only healthy individuals may migrate from rural-to-urban areas for employment. Another channel is the social networks built by the migrant workers in the cities, which in turn helps to reduce the negative impacts of migration on mental health. Based on the survey data from rural-to-urban migration in China, two recent studies provide evidence to show how the self-selection effect and the migrant social networks affect mental health outcomes among the rural-to-urban migrant workers, respectively. Specifically, Ma et al. (2020) test the healthy migrant phenomenon by comparing migrant workers with urban residents. They measure the mental health of the two groups by the General Health Questionnaire (GHQ) 12, which contains 12 questions with a 4-point Likert-type scale (ranging from 0 to 3, with 0 indicating most healthy and 3 indicating least healthy). This score is widely used as an index to measure the common types of mental disorders. After controlling for socioeconomic variables in regression analyses, they find evidence in support of the healthy migrant phenomenon; i.e. the migrant workers tend to have better mental health than urban residents. Another study by Meng and Xue (2020) further examines the impact of social networks on the mental health of rural migrants, using the same survey data and the same measures of mental health status. After controlling for the endogeneity of the social networks, which are measured by the number of contacted people living in the urban area, this study finds that social networks help relieve the mental health problems of rural migrants. Specifically, the results show that each additional friend in the social network can reduce the migrant’s mental health Likert score by 0.089– 0.154 points. This means that a standard deviation increase in the social network measurement can reduce the mental health Likert score by 0.47 standard deviation. Mechanism analysis further shows that the two major channels through which social network takes its effects is by boosting migrant workers’ confidence and reducing their anxiety. In addition to the large-scale internal migration, another major phenomenon associated with the rapid economic growth in China is fast urbanization. Similar to the experience of Western countries, the relationship between rapid urbanization and population health is complex and may include both positive and negative impacts. On the one hand, urbanization may offer better access to healthcare services as well as higher income opportunities that enable individuals to invest in their health. On the other hand, the physical and mental health of urban residents may be threatened by the unhealthy living environment such as air pollution, over-crowded living space, stressful workload, and sedentary lifestyle. Thus, the net effect of urbanization on health is an empirical question and depends on the relative magnitude of the positive
10
1 Introduction
and negative impacts. The following papers try to disentangle such relationship using various dataset. Van de Poel et al. (2012) measure the urbanization in China by constructing an index of urbanicity from a broad set of community characteristics, including population size, the proportion of workforce engaged in agriculture, proximity to health and educational facilities and the presence of paved roads, shops, restaurants, etc. They then define urbanization in terms of movements across the distribution of this index. Based on the survey data obtained from the China Health and Nutrition Survey (CHNS), their study finds evidence of urban health penalty; i.e. urbanization raises the probability of reporting poor health. Due to the lack of empirical measure for mental health in CHNS, this study is not able to address the question on whether the urban health penalty include the adverse impact on one’s mental health. Chen et al. (2014) fill this gap in research by focusing on the relationship between mental health and a more comprehensive measurement of urbanization. They argued that urbanization is a multi-faceted process, and hence, they used three dimensions to measure the process of urbanization, including the static level of urbanization at a point in time, the speed of urbanization over time and whether the speed is in acceleration or deceleration dynamically. They obtained data for these measures from the Defense Meteorological Satellite Program Operational Linescan System (DMSP-OLS) night-time light data. They measure the mental health by the shortform CES-D obtained from the 2011 Migration and Quality of Life Survey. Their results show that all the three measures of urbanization at the county level are significantly and positively associated with the CES-D scores of local residents, indicating that all dimensions of urbanization have adverse effects on mental health. Specifically, individuals are more likely to have a greater depressive distress when living in a county with a higher level of urbanization, experiencing faster urbanization rates, and with acceleration of the urbanization speed. Another risk factor linking the rapid economic growth and the mental health conditions is environmental pollution, particularly air pollution. It has been widely recognized in the literature that air pollution impairs physical health, especially for infant and child health (Chay and Greenstone 2003; Currie and Neidell 2005; Coneus and Spiess 2012). However, few studies have paid attention to the potential impacts of air pollution on mental health in China. In recent years, two studies have filled this research gap. Zhang et al. (2017) match the CES-D score obtained from China Family Panel Studies (CFPS) with the city-level air quality data based on the respondent’s geographic locations and the dates of the interview. Air quality is measured by the air pollution index (API), which reflects the local concentrations of several air pollutants. The results show that higher API significantly raises the rate of depressive symptoms experienced by the respondent. In addition, the marginal effects of pollution increase significantly with the dose of exposure. Based on a different dataset (three waves of survey data from the China Health and Retirement Longitudinal Study) and a different measure of pollution (PM2.5), Ao et al. (2021) also yield similar conclusions that air pollution imposes significant adverse effects on one’s mental health. This study measures the mental health status using the short-form CES-D score which ranges from 0 to 30 and uses the cut-off
1.4 Other Causes of Rising Prevalence of Mental Illness in China
11
value of 10 to measure the presence of depressive symptoms. The city-month level PM2.5 data are retrieved from the satellite-based information released by the National Aeronautics and Space Administration (NASA). The findings based on the merged datasets show that the concentration of PM2.5 has a significantly positive effect on the respondent’s CES-D score as well as her probability of having depressive symptoms, indicating that air pollution has noticeable adverse effects on the mental health for the middle-aged and elderly population in China. Putting all the studies that we reviewed in this section together, it suggests that economic growth generates several important factors that drives the rising prevalence of mental illness in China, including internal migration, environmental pollution and rapid urbanization. These forces work together to produce profound changes in people’s lifestyles, putting strong pressure on the human psyche and hence imposing detrimental effects on mental health. This raises a concern on whether there is a mental health Kuznets curve in that the prevalence rate of mental illness is positively associated with income per capita in the early stage of economic growth and then this association turns negative in the latter stage of growth. During the past decades, numerous studies have documented the evidence to support the existence of an environmental Kuznets curve in China, but few have shown evidence for the above hypothesis on the mental health Kuznets curve, which can be an avenue for future research. In 2009, China launched a pension program for rural residents known as the New Rural Pension Scheme (NRPS). Since then, more than 400 million Chinese have enrolled in the NRPS. Based on the survey data obtained from 2012 China Family Panel Studies (CFPS), Chen et al. (2019) examine the effect of pension enrollment on the rural resident’s mental health, as measured by CES-D and self-reported depressive symptoms. This study capitalizes on the geographic variation in pension program implementation to overcome the endogeneity of pension enrollment and finds that pension receipt decreases the CES-D score and depression symptoms by 0.70 and 0.35 standard deviations, respectively. On average, receiving pension reduces the prevalence of depressive symptoms by 25.4%. Overall, this study provides strong evidence that pension enrollment is an effective mechanism to improve the mental health of the Chinese population. An important implication of the above study is that it is possible to incur the Kuznets curve in the mental health sector through appropriate policy reforms, especially when economic growth increases the government funding capacity and put the government in a better position to implement such policies toward the overall improvement of population mental health. In the following section, we will further review the recent progress of the government policies in the mental health sector in China, and discuss whether the policy progress is sufficient and strong enough to support the mental health Kuznets curve hypothesis in that further growth in percapita income may lead to a decrease in the prevalence of mental illness in China’s population.
12
1 Introduction
1.5 Policy Updates China in the recent years has improved laws and regulations related to mental disorders, strengthening the public mental healthcare system at legislative and policy levels, and paying more attention to mental health in reforming its health insurance, public education, and social support systems. Through these joint efforts between the public and private sectors, China’s society is increasingly focusing on improving the prevention, treatment, and assistance for mental disorders, and promoting better mental health for the whole population. From the perspective of legislation and regulation, China adopted the Mental Health Law in 2012, which for the first time stipulates in legal form the prevention, diagnosis, and treatment of mental disorders, as well as the rehabilitation, supporting measures, and legal responsibilities of persons with mental disorders. In 2015, the State Council approved the National Mental Health Work Plan (2015–2020), which requires that by 2020, the country will have made significant improvements in seven areas including an integrated mental health management and coordination mechanism, mental healthcare service networks, training of mental health specialists, treatment and monitoring of persons with severe mental disorders, prevention and control of common mental health problems, establishing a rehabilitation and social support system, and increasing the public awareness of mental health. In 2016, the Communist Party of China (CPC) Central committee and the State Council released the “Healthy China 2030” Blueprint, which further made “promoting mental health” an important goal of healthy living for all people in China. This outline further proposed to “strengthen the building of mental health service system and standardize its management”, as well as to improve the efficacy of mental health services provided to persons with mental disorders, vulnerable groups susceptible to mental problems, and all the other residents. In 2019, China adopted the Basic Medical Health and Health Promotion Law, which also sets the principles on the work of mental health, clearly stating that mental health is an integral part of the population health, and the building of a comprehensive mental health service system is an important part of China’s medical healthcare system. In 2021, the National People’s Congress approved the 14th Five-Year Plan for National Economic and Social Development and Long-Range Objectives for 2035, which continues to stress “the prevention and treatment of mental disorders”, “improving the system of psychological health and mental hygiene”, and “perfecting the social psychological service system and psychological crisis intervention mechanism”. The plan also continues to implement the goals of “strengthening the registration, service management, medical treatment and assistance of people with severe mental disorders”, “carrying out pilot study on early screening and intervention of commonly seen mental health disorders”, and “enhancing mental health services”, etc. In the dimension of mental healthcare provision, China has in recent years continued to strength the supply of medical services, improve the mental healthcare delivery system, and enhance the service quality of mental healthcare centers
1.5 Policy Updates
13
and institutions. First, the newly introduced principles and standards for the treatment of mental disorders serve as important guidelines for the clinical diagnosis and current treatment of mental disorders. In 2013, the former National Health and Family Planning Commission published the Guiding Principles for the Treatment of Mental Disorders (2013 Edition) and the Work Rules on the Psychological Treatment (2013 Edition) to further standardize the treatment of mental disorders and psychological problems. These series of principles and treatment guidelines elaborate on the classifications of psychotherapy and provide specific operational techniques for 13 types of therapies. In 2020, the National Health Commission published the Work Rules on Diagnosis and Treatment of Mental Disorders (2020 Edition), which further refines the specific treatment principles and approaches for various types of mental disorders. Compared with the 2013 edition, the 2020 edition describes the diagnosis and treatment specifications of each disorder type in more details that reflect updated clinical practices. Second, China has continued to strengthen the construction of a multi-level mental healthcare delivery system. In 2018, the National Health Commission formulated the Pilot Model for the Establishment of Social Mental Health System in 49 cities across the country to try to establish a comprehensive social psychological service network: pilot areas are required to equip their community health centers with a counseling office for local mental healthcare needs and to set up mental health outpatient departments in all psychiatric hospitals and 40% of Grade II general hospitals. In 2019, the Healthy China Program (2019–2030), developed by the National Health Commission and promoted by the Promotion Committee of the Healthy China Initiative, also encourages psychological counseling professionals to set up social psychological counseling services and also encourages local governments to expand health insurance coverage to be able to purchase mental healthcare service from those counseling providers. In 2020, the National Health Commission drew up the Opinions on Strengthening and Improving Psychiatric Specialized Medical Services to increase the supply of mental healthcare resources into the construction of the national health service system during the 14th Five-Year Plan period (2021–2025) and to put forward specific requirements on the number and level of mental health departments in different types of hospitals and cities. Third, China continues to invest various resources to support the training of mental health professionals and strengthen the mental health workforce. In 2014, the Ministry of Education promulgated the Opinions on Collaboration Between Medical Education and Healthcare and Deepening the Reform of Clinical Professional Training to encourage universities to open psychiatry majors and strengthen “the training of psychiatry and other urgent and shortage talents”. In 2016, the former National Health and Family Planning Commission formulated the Guiding Opinions on Strengthening the Mental Health Services, which clearly calls for “strengthening the training of mental health professionals”, “promoting the orderly development of mental health service professionals”, and “improving the incentive mechanism for mental health service professionals”. In 2017, the Ministry of Civil Affairs, the Ministry of Finance, the former National Health and Family Planning Commission, and China Disabled Persons Federation jointly issued the Opinions on Accelerating
14
1 Introduction
the Development of Community Rehabilitation Services for Mental Disorders, which proposed to widely establish the integrated service teams that include psychiatrists, social workers, nurses, psychological counselors, public health professionals, and community rehabilitators to meet the multi-dimensional mental healthcare needs. The above policy measures contributed to a sustained growth in mental healthcare workforce in China. By the end of 2021, there were 5,936 mental healthcare institutions nationwide, a 205% increase from 2010; there were more than 50,000 registered psychiatrists practicing nationwide, a 144% increase from a decade ago. In addition to mental healthcare provision, China also continued to deepen its health insurance reforms in recent years, which has provided more coverage for persons with mental disorders under medical insurance policies. For inpatient care, China continues to promote the reform of medical insurance payment methods to reduce the heavy burden of patients with chronic mental disorders. In 2020, the State Council of China released the Opinions on Deepening the Reform of the Medical Insurance System, which clearly proposed that long-term hospitalization costs for chronic mental illnesses will be paid on a bed-day basis. For outpatient care, China in 2021 created a list of chronic and special diseases for medical coverage that includes mental illnesses. Prior to this time, various regional reforms were in fact conducted to add severe mental illnesses into the special outpatient diseases that are covered by the basic medical insurance. For example, in Beijing, “severe mental illness” was included as outpatient special diseases in the city’s basic medical coverage in 2020, which encompasses such mental illnesses as schizophrenia, bipolar disorder, schizoaffective mental disorder, paranoid mental disorder, epilepsy-induced mental disorder, and severe mental retardation. For another example, the Dongguan city of Guangdong province added five severe mental illnesses (schizo-affective disorder, persistent delusional disorder, bipolar disorder, epilepsy-induced mental disorder, and mental retardation with psychotic disorder) into the range of outpatient basic medical coverage as early as 2015. In addition, the reimbursement rate for mental illnesses in China’s medical insurance has been increased. Take the urban employee basic medical insurance (UEBMI) in Beijing as an example, when “severe mental illness” was included in the outpatient basic medical coverage in 2020, the program also reduced the reimbursement deductible for mental illness from 1800 RMB to 1300 yuan, increased the reimbursement ceiling from 20,000 RMB to 500,000 yuan, and increased the reimbursement rate for community and tertiary hospital visits from 90 and 70%, respectively, to 95 and 85%, respectively. Last but not least, China has made great efforts in recent years to strengthen the social support system for mental health and implement mental health promotion campaigns. For example, the Ministry of Education issued requirements in 2012 to integrate mental health coursework throughout the teaching process of primary and middle schools, to establish psychological counseling rooms on campus, and to connect with parents to jointly create a mentally healthy environment inside and outside schools. For colleges and universities, the Ministry of Education formulated guidelines in 2018 on four main tasks including promoting mental health education, carrying out mental health promotion activities, strengthening counseling services, and enhancing preventive interventions of mental problems. The
1.6 Conclusions
15
Healthy China Program (2019–2030) also requires educational institutions at all levels to teach healthy behaviors and lifestyles to promote the mental health of students. Additionally, China also relies on business enterprises and social organizations to provide targeted mental health education services to various groups. For example, The Healthy China Program (2019–2030) requires employers to integrate mental health education into the ideological and political work of employees and encourage companies’ human resource department to hire mental health service personnel or purchase corresponding mental health services. Social organizations such as the Disabled Persons’ Federation, senior citizen schools, women’s homes are also required to promote mental health knowledge and provide psychological support services in accordance to the special mental healthcare need of the vulnerable groups. Finally, China has continued to carry out mental health promotion campaigns. Health departments at all levels continued to carry out mental health promotion and education activities, such as utilizing World Mental Health Day to carry out awarenessraising programs to popularize mental health knowledge to the public. During the COVID-19 pandemic, efforts have been made to relieve the psychological impact and stress caused by COVID-19, particularly for individuals under quarantine, the cured patients, and the frontline healthcare workers.
1.6 Conclusions The rising prevalence of mental disorders in China reflects the complex relationship between human health and economic development. This complexity is illustrated in the conceptual framework of the triangular relationship between mental health, healthcare system and wealth (Fig. 1.2). Through the whole process of economic development, the increase in income and the accumulation of wealth are both positively and negatively contributing to the change in population mental health, and this process is further interacting with a country’s healthcare system reforms, health financing arrangement, urbanization process, environmental degradation, and lifestyle changes. The global outbreak of the COVID-19 pandemic and the revolutionary development of digital health technology is further contributing to this complex triad and having a profound impact of the global mental health profile. In this introductory chapter, we have briefly summarized the recent status of mental disorders and the policy responses in China and reviewed a series of academic works that explore the causes and consequences of mental health problems in the largest developing country in the world. These studies are multi-disciplinary in nature, and uses quantitative and qualitative methodologies to address the complex issue of rising prevalence of mental disorders and its interaction with the socioeconomic development in China. Among all the factors, economic inequalities, social stigma, the availability and affordability of mental healthcare resources, and the diffusion of medical knowledge and technology are shown to play a key role in driving the rates of diagnosis and treatment of in the mental healthcare sector, while massive labor
16
1 Introduction
migration, rapid urbanization, and environmental pollution are some of the primary factors that drive the increasing prevalence of mental disorders. In addition to the topics covered by this book, there are a few emerging new issues related to mental health in China that are still under-explored. We list them here and hope that they will be soon addressed by future research. First, it is important to address the question on whether there is an empirical basis for a mental health Kuznets’ curve in China. As we found in our previous studies, the increasing prevalence of mental illness in China concentrated in certain groups of population, including the elderly, the rural residents, the poor, and the low-educated people. This suggests that the income-related inequalities tend to be correlated with the disparities in health outcomes, creating an income-health gradient across the population. Our previous study also showed that health inequality (measured by the maldistribution of healthcare resources) is likely to increase at the early stage of economic development in China, while it tends to decrease once the fruits of economic development spread across the entire population (Qin and Hsieh 2014; Costa-Font et al. 2018). This raises the question of whether the same Kuznets curve pattern will be found in the mental health sector; i.e. the disparity in mental health outcomes will decrease with the continued economic growth in China. Our updates on the recent policy reforms have shown that the government has increased the investment in mental health resources through a series of policy efforts, including the expansion of health insurance coverage for mental illnesses and the public health programs that targeted on improving the mental health of vulnerable populations. We speculate that these policy efforts in turn may become the driving forces that lead to an inverted U-shape in the association between the mental health inequalities and economic development. Although we are optimism about the policy effectiveness, more research needs to be done to provide empirical evidence. Second, another complementary mechanism that leads to a decrease in health inequality may come from the recent application of digital health interventions for the treatment of mental health conditions, including virtual reality, telemedicine and computer-assisted therapy. As noted, the higher time cost of seeking care associated with the maldistribution of mental healthcare resources in China has been widely recognized as one of the culprits of under-treatment in mental health conditions. One of the advantages of digital healthcare such as telemedicine is to remove the distance barrier in seeking for care and hence may be beneficial to mitigate the health inequality in metal health sector. Although there is ample evidence in high-income countries on the effectiveness of digital health intervention in delivering mental health treatment, the evidence is still limited in China (Philippe et al. 2022). Given the fast growth in digital technology adoption within the mental health sector, an important avenue of future research is to increase our understanding on the effectiveness of such interventions and its implications on the mental health disparities and health policy reforms. Finally, one of the important driving forces that shifted much of the mental healthcare delivery to digital healthcare platforms is the global COVID-19 pandemic. As the health risks of the infectious disease prevented many people from visiting
References
17
healthcare professionals in person, telemedicine and other digital health interventions have been widely adopted in China during this time. However, recent studies have documented the evidence that COVID-19 and the related government interventions (such as mandated quarantines and lockdowns) have caused worsening mental health, which may be due to the fear and uncertainty about becoming infected or the anxiety about the severe mobility restrictions (Serrano-Alarcón et al. 2022; Altindag et al. 2022). This suggests that the COVID-19 pandemic has imposed complicated impacts on both the incidence and treatment of mental health conditions. It takes time to know how these impacts would affect the overall health and socioeconomic outcomes and whether these impacts may persist in the long term. As a result, the dynamic impact of COVID-19 on the mental health of China’s population is another important and interesting avenue for future research.
References Altindag O, Erten B, Keskin P (2022) Mental health costs of lockdowns: evidence from age-specific curfews in Turkey. Am Econ J Appl Econ 14(2):320–343 Ao C-K, Dong Y, Kuo P-F (2021) Industrialization, indoor and ambient air quality, and elderly mental health. China Econ Rev 69:101676 Bloom DE, Canning D (2000) The health and wealth of nations. Science 287(5456):1207–1209 Chay K, Greenstone M (2003) The impact of air pollution on infant mortality: evidence form geographic variation in pollution shocks induced by a recession. Quart J Econ 118(3):1121–1167 Chen J, Chen S, Landry PF, Davis DS (2014) How dynamics of urbanization affect physical and mental health in urban China. China Q 220:988–1011 Chen X, Wang T, Busch SH (2019) Does money relieve depression? Evidence from social pension expansion in China. Soc Sci Med 220:411–420 Coneus K, Spiess CK (2012) Pollution exposures and child health: evidence for infants and toddlers in Germany. J Health Econ 31(1):180–196 Currie J, Neidell M (2005) Air pollution and infant health: what can we learn from California’s recent experience? Quart J Econ 120(3):1003–1030 Costa-Font J, Hernandez-Quevedo C, Sato A (2018) A health ‘Kuznets’ curve’? Cross-sectional and longitudinal evidence on concentration indices’. Soc Indic Res 136:439–452 Figueras J, McKee M (eds) (2012) Health systems, health, wealth and societal well-being: assessing the case for investing in health system. Open University Press Ma C, Qu Z, Xu Z (2020) Internal migration and mental health: an examination of the healthy migration phenomenon in China. Popul Res Policy Rev 39(3):493–517 McKee M, Suhrcke M, Nolte E et al (2009) Health system, health, and wealth: a European perspective. Lancet 373:349–351 Meng X, Xue S (2020) Social networks and mental health outcomes: Chinese rural-urban migrant experience. J Popul Econ 33(1):155–195 Mills C (2018) From “invisible problem” to global priority: the inclusion of mental health in the sustainable development goals. Dev Chang 49(3):843–866 Philippe TJ, Sikder N, Jackson A, Koblanski ME, Liow E, Pilarinos A, Vasarhelyi K (2022) Digital health interventions for delivery of mental health care: systematic and comprehensive metareview. JMIR Mental Health 9(5):e35159 Qin X, Hsieh CR (2014) Economic growth and the geographic maldistribution of health care resources: evidence from China, 1949–2010. China Econ Rev 31:228–246
18
1 Introduction
Radloff LS (1977) The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol 1(3):385–401 Scheffel J, Zhang Y (2019) How does internal migration affect the emotional health of elderly parents left-behind? J Popul Econ 32(3):953–980 Shi Y, Bai Y, Shen Y, Kenny K, Rozelle S (2016) Effects of parental migration on mental health of left-behind children: evidence from northwestern China. Chin World Econ 24(3):105–122 Serrano-Alarcón M, Kentikelenis A, Mckee M, Stuckler D (2022) Impact of COVID-19 lockdowns on mental health: evidence from a quasi-natural experiment in England and Scotland. Health Econ 31(2):284–296 Van De Poel E, O’Donnell O, Van Doorslaer E (2012) Is there a health penalty of China’s rapid urbanization? Health Econ 21:367–385 Weil DN (2014) Health and economic growth. Handb Econ Growth 2B:623–682 World Health Organization (2000) The world health report 2000: health systems: improving performance. World Health Organization Zhang X, Zhang X, Chen X (2017) Happiness in the air: How does a dirty sky affect mental health and subjective well-being? J Environ Econ Manag 85:81–94
Chapter 2
The Prevalence of Depression and Depressive Symptoms Among Adults in China: Estimation Based on a National Household Survey
2.1 Introduction The global prevalence of depression and depressive symptoms has been increasing over the past decades, and depression has been recognized as a leading cause of disability and a major contributor to disease burdens worldwide (WHO 2012). With its fast economic growth, China’s population is exposed to a high-stress lifestyle, which may result in a significant increase in mental health problems. According to Fan et al. (2013), over 100 million Chinese experience different kinds of mental disorders during a year, and these mental diseases account for over 20% of the total burden of disease in China, making mental disorders a major public health concern. Zhou (2010) suggests that mental health problems are severe in China, and that the prevalence of mental diseases is increasing in at least three groups of Chinese people, including the youth, migrant workers (and their families) and the elderly. These concerns need to be addressed, as depression not only disrupts physical wellbeing but also interferes with human capital accumulation and worker productivity, leading to high direct medical costs and indirect economic costs associated with the psychiatric disorder. The mental health of the Chinese people has drawn increasing attention of the central government in China. Since the new millennium, the government introduced a series of policies, regulations, and laws in an effort to build a comprehensive system for improving mental health. The evolution of China’s mental health policies is summarized in Table 2.1, which reflects an overall strengthening of the mental healthcare system, more detailed regulations on mental health service delivery, and
The content of this chapter is published in Qin X, Wang S, Hsieh CR. The prevalence of depression and depressive symptoms among adults in China: estimation based on a National Household Survey. China Economic Review, 2018, 51: 271–282. Copyright Elsevier (2018), reproduction license granted.
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 X. Qin and C.-R. Hsieh, Economic Analysis of Mental Health in China, Applied Economics and Policy Studies, https://doi.org/10.1007/978-981-99-4209-1_2
19
20
2 The Prevalence of Depression and Depressive Symptoms Among Adults …
the separation of mental diseases from other medical conditions in hospital management. As a milestone in the process, the Mental Health Law of China launched in 2013 formalizes the legal protection and treatment of people with mental disorders. Meanwhile, many academic studies have assessed the prevalence of mental disorders among different age groups and geographic areas. For example, using individual and group interviews, Tian (2012) focuses on the mental health problems among those aged between 15 and 21 and concludes that a trusting relationship with family members tends to reduce the occurrence of youth mental disorders in China. Based on an epidemiological survey conducted in four provinces (Shandon, Table 2.1 Evolution of mental health policies in China since 2002 Year
Date
Major policy document and issuing authority
2002
April 10
National Mental Health Plan (2002–2010): Ministry of Health, Ministry of Civil Affairs, Ministry of Public Security, China Disabled Person’s Federation
2004
August
Proposal on Further Strengthening the Mental Health System in China: Ministry of Health, Ministry of Education, Ministry of Public Security, Ministry of Civil Affairs, Ministry of Justice, Ministry of Finance, China Disabled Person’s Federation
2006
November 14
Inter-ministerial Joint Conference on Mental Health System: Ministry of Health. Ministry of Education, Ministry of Public Security, Ministry of Civil Affairs, Ministry of Justice, Ministry of Finance, China Disabled Person’s Federation
2008
January 15
Guiding Compendium on Development of National Mental Health System (2008–2015): Ministry of Health
2011
March 17
Basic Outpatient Guidelines for Clinical Psychology at Medical Institutions (Trial): Ministry of Health
2012
April 5
Specifications of Management and Intervention for Psychoses (Version 2012): Ministry of Health
June 6
National Mental Health Plan (2012–2015) (Draft for Consultation): Ministry of Health
June 25
Proposal on Further Strengthening the Social and Psychological Assistance to Natural Disasters: China National Commission for Disaster Reduction
October 26
Mental Health Law of the People’s Republic of China (To be Enforced in 2013)
July 29
Measures for the Management of Psychoses Reporting (Trial): National Health and Family Planning Commission (former Ministry of Health)
May 3
Specification on Establishing the Mental Health Outpatient Departments within General Hospitals: National Health and Family Planning Commission
2013
Notes This table outlines the major mental health policies and the issuing government authorities in China after 2002; the policy evolution is based on the content of the related policy documents, which to some extent reflects the so-called documental politics in China; the information is obtained from the official website of the various government agencies in China
2.2 Data Source
21
Zhejiang, Qinghai and Gansu) on 63,004 adults, Phillips et al. (2009) report that the adjusted one-month prevalence rate of mental disorder—measured by the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV)—is 17.5% during 2001–2005. Based on a community survey of people aged 16 and above in Xi’an city, Chen (2012) uses similar measures to estimate that the lifetime prevalence of mental disorder is 21.01%. Wang (2014) conducts a survey of 6000 randomly selected individuals aged between 35 and 74 in Qingdao city and finds that the prevalence rate of depression as measured by the Zung Self-Rating Depression Scale (SDS) is 12.12%. Various city- and region-level estimates are also reported by other studies. For example, Zhu et al. (2010) report a prevalence rate of 21.66% for the Hangzhou–Jiaxing region; Cui et al. (2009) find a rate of 7.4% for Baoding city; Zhao et al. (2009) estimate that the prevalence of depression in Guangzhou city is 15.76%. The above research suggests that mental disorders have become a widespread health concern for the Chinese population, with uneven distribution among different demographic, socioeconomic, and geographic groups. However, given that most of these studies are based on moderate sample sizes (with less than 10,000 respondents), limited sampling regions (data are often collected in a few number of cities or provinces), and specific age ranges (youth or elderly), there is a need for a more comprehensive study based on nationally representative data. This paper aims to fill this gap by providing a systematic and updated overview of the mental health status (using depression and depressive symptoms as an example) of the adult population in China. Based on the 2012 China Family Panel Studies (CFPS) survey that covers 40,000 respondents from 25 Chinese provinces, we use cross-tabulation analysis and ordered probit model to estimate the national prevalence and correlates of depressive symptoms and depression among the Chinese adults aged 18 and above. Our results suggest that the overall prevalence of depressive symptoms is 37.86% and that of severe depression is 4.08% among all Chinese adults. We also conduct comparative analysis among different subsamples and estimate the correlation between mental health status and the demographic, socioeconomic, and geographic characteristics of the sampled individuals.
2.2 Data Source CFPS is a nationally representative longitudinal survey designed and implemented by the Institute of Social Science (ISSS) of Peking University. The survey so far involves five waves (2008, 2009, 2010, 2011, 2012), each covering about 15,000 households and about 57,000 individuals of all age groups in 25 provinces of China (the excluded provincial-level regions include Hong Kong, Macao, Xinjiang, Qinghai, Inner Mongolia, Ningxia, and Hainan).1 To collect comprehensive information of the covered respondents and regions, CFPS contains three modules: the individual1
The 25 provinces covered by CFPS jointly account for 95% of the Chinese population, making CFPS a nationally representative survey in China.
22
2 The Prevalence of Depression and Depressive Symptoms Among Adults …
and household-level surveys collect detailed information on the respondents’ demographic, socioeconomic and health-related characteristics, while the communitylevel survey collects information on the infrastructure, demographic profiles, social services, and economic conditions of the rural villages or urban communities that the respondents live in. CFPS follows a multi-stage stratified sampling method. The 25 provinces are first divided into six strata: five “key” provinces (Liaoning, Henan, Gansu, Guangdong, Shanghai) are each assigned to an independent stratum and the other twenty “non-key” provinces consist a sixth stratum.2 For each of the “key” provinces, 1600 households are selected by an oversampling method, while the twenty “non-key” provinces contain 8000 households in total; thus, they jointly contribute to the target sample size of 16,000 households. The households are selected according to the probability proportionate to size (PPS) sampling method with implicit stratification, where administrative districts (provinces, cities and counties) and socioeconomic status (mainly local GDP per capita) serve as the main stratification variables. We use the 2012 CFPS national sample for our study, which is the most up-todate wave available. The 2012 CFPS questionnaire is the first to contain the standard 20 questions of the Center for Epidemiologic Studies Depression Scale (CES-D), which is originally developed by Radloff (1977) and still one of the most widely used self-evaluation tests on the respondent’s mental health condition. Out of the 20 questions in the CES-D, 16 measure negative feelings and 4 measure positive feelings. Respondents are requested to choose from four frequency values (“rarely”, “little”, “occasionally”, and “most”) to indicate how often they have those feelings. The choice for each question corresponds to an index value of 0, 1, 2, 3 for negative feelings and 3, 2, 1, 0 for positive feelings, which in total makes up an overall CES-D score ranging from 0 to 60. According to Radloff (1977, 1991), a score of 16–28 indicates that the person suffers from depressive symptoms, and a score higher than 28 indicates severe depression.3 Some researchers suggest alternative classification schemes: Somrongthong (2013) suggests a threshold value for depressive symptoms of 22; Chen et al. (2009) suggests a more detailed segmentation with healthy (0–17), possible depression (18– 23), probable depression (24–28), and severe depression (above 28) as possible categories. We adopt the original classification standard in Radloff (1977, 1991) for our main analysis and use the alternative classification methods for robustness check. The central outcome variable in our analysis is mental health status (MHS), which falls into three categories according to the value of CES-D scores: mentally 2
According to the CFPS official explanation, the five “key” provinces are oversampled compared to the other “non-key” provinces in order to study the social and economic characteristics of these major regions in China, while the 25 sample provinces are jointly representative of the country with provincial weight assignment. 3 Note that the “depression” indicated by CES-D is different from the clinical definition of major depression disorder (MDD) and major depression episodes (MDE). MDD and MDE are usually diagnosed with 12-month occurrence of depressive events according to the DSM-IV criteria, while the depression indicated in our study by CES-D is diagnosed with one-week occurrence of depressive events.
2.2 Data Source
23
healthy (0–15), experiencing depressive symptoms (16–28), and suffering from severe depression (above 28), which correspond to the values of 0, 1 and 2 of MHS respectively. We restrict our sample to adult respondents aged between 18 and 99 and further drop the observations with missing information on the key variables such as gender, age, and CES-D scores. Table 2.2 reports the sample summary statistics. The final sample contains 29,345 respondents, with 48.86% being male. The sample average age is 46.91. In terms of regional distribution, there are more respondents in the eastern region (44.42%) in comparison to the central (29.96%) and western (25.62%) parts of the country, which is consistent with the overall population distribution in China. Due to the large number of people living in the rural areas, the rural respondents account for the majority share (69.47%) of the sample. Most (84.63%) of the respondents are married, while the rest are categorized into Table 2.2 Sample summary statistics of select variables Variable
Definition
Mean
MHS
Mental health status (0 = mentally healthy, 1 = depressive 0.471 symptoms, 2 = depression)
0.584
Male
Gender (1 = male)
0.500
0.489
Standard Deviation
Age
Age in years
46.908 15.622
Central
Live in Central China (1 = yes)
0.300
0.458
West
Live in Western China (1 = yes)
0.256
0.437
Urban
Urban residents (1 = yes)
0.305
0.461
Primary
Primary school (1 = yes)
0.216
0.411
Middle
Middle school (1 = yes)
0.268
0.443
High
High school (1 = yes)
0.132
0.338
Junior
Junior college (1 = yes)
0.048
0.215
University
4-year university (1 = yes)
0.027
0.163
Graduate
Master or Doctoral (1 = yes)
0.002
0.041
Married
Married (1 = yes)
0.846
0.361
Divorced
Divorced (1 = yes)
0.014
0.117
Widowed
Widowed (1 = yes)
Unemployed Unemployed (1 = yes)
0.058
0.234
0.032
0.176
Retired
Retired (1 = yes)
0.134
0.340
OLF-1
Out of labor force due to disability or diseases (1 = yes)
0.031
0.173
OLF-2
Out of labor force due to other reasons (1 = yes)
0.077
0.267
Income
Annual personal income (thousand yuan)
11.594 31.518
Observation
Sample size
29,345
Note (1) Data Source: China Family Panel Studies (2012). (2) The statistics reported are sample mean and standard deviation of main variables
24
2 The Prevalence of Depression and Depressive Symptoms Among Adults …
single (8.16%), divorced (1.39%) and widowed (5.82%). For socioeconomic characteristics, less than 20% of the sample receives education beyond the middle school (9-year compulsory education) level, and the illiterate/semi-illiterate group4 accounts for 30.60% of the whole sample. In addition, the majority (72.61%) of respondents are employed, while 3.18% are “unemployed”, 13.37% are “retired”, and about 10% are out of the labor force due to various reasons. The average annual personal income of our sample is 11,594.36 yuan. The mean value of MHS is 0.471 on a scale of 0 to 2, indicating that our sample respondents are on average mentally healthy.
2.3 Method 2.3.1 Cross-tabulation Analysis Our first set of analysis uses cross-tabulations to obtain the overall and groupwise prevalence of depression and depressive symptoms by gender, age, geographic regions (eastern/central/western China), residential status (urban/rural), education attainment (illiterate/primary school/middle school/high school/junior college/4year university/graduate school), marital status (single/married/divorced/widowed), employment status (employed/unemployed/retired/out of labor force due to disability/out of labor force due to other reasons), and income levels (measured in 2012 yuan). In the cross-tabulation, we report for each group the number and share of respondents who are mentally healthy, frequently experiencing depressive symptoms and suffering from depression. We also report the mean, standard deviation, minimum and maximum values of the CES-D scores of each group. In order to obtain nationally representative estimates, the reported group prevalence rates (percentage values) and the mean CES-D scores are all adjusted by the official sampling weights of the 2012 CFPS data which reflect the sampling probability of each observation. We also report the 95% confidence intervals for all the prevalence rate estimates and use the Chi-square tests to assess the statistical significance of the group differences in the CES-D scores with the associated p-values reported.
2.3.2 Regression Analysis Following prior studies such as Chatterji and Markowitz (2012), our second set of analyses uses the ordered probit model5 to characterize the statistical association 4
According to the official CFPS questionnaire, the definition of the “illiterate/semi-illiterate group” is “people who know less than 1500 Chinese characters”. 5 We tested the “parallel regression” (or proportional odds) assumption of the ordered probit model using the approximate Likelihood Ratio test and the Wald test. Although the test results reject the
2.4 Estimation Results
25
between the probability of depression/depressive symptoms and the individual demographic, socioeconomic, and geographic characteristics. The model can be specified as follows: Hi∗ = X i β + u i ,
(2.1)
Pr(Hi = 0|X) = Pr(Hi∗ ≤ ω1 |X) = F(ω1 − X i β),
(2.2)
Pr(Hi = 1|X) = Pr(ω1 ≤ Hi∗ ≤ ω2 |X) = F(ω2 − X i β) − F(ω1 − X i β), (2.3) Pr(Hi = 2|X) = Pr(Hi∗ ≥ ω2 |X) = 1 − F(ω2 − X i β),
(2.4)
where Hi denotes individual i’s MHS category, taking a value of 0, 1 or 2. X i denotes a set of individual characteristics including age, gender, marital status, education, employment status, geographic region, residential status, and income. Hi∗ is a latent variable measuring the continuous mental health status of individual, i which is assumed to hold a linear relationship with X i . The realization of Hi depends on the interval that Hi∗ falls into, with the thresholds of intervals determined by ω1 and ω2 and the corresponding probabilities determined by the cumulative distribution function [F(.)] of the random error term [u i ]. Following the literature convention, we assume u i follows the standard normal distribution and estimate the above model by the maximum likelihood estimation (MLE). It should be noted that the coefficient estimates in the above model only indicate the impacts of individual characteristics on Hi∗ , i.e. the tendency of experiencing depression or depressive symptoms. To obtain the (nonlinear) marginal effects of these characteristics on the probabilities of the observed MHS categories, we use the sample average of individual marginal effects calculated by the finite-difference method.
2.4 Estimation Results 2.4.1 Cross-tabulation Results Table 2.3 presents the cross-tabulation results. Among the 29,345 respondents in our sample, about 58.06% are mentally healthy, while 37.86% experience depressive symptoms and 4.08% suffer from depression. The sample CES-D has a weighted mean of 15.53, which is slightly lower than the threshold between the mentally null hypothesis of parallel regressions, the inherent drawbacks of the alternative multi-nominal models suggest that ordered probit model is still preferred. The Bayesian information criterion (BIC) test results also provide strong support for the ordered probit model over the generalized ordered probit model, providing validation for our model selection.
15,008 100
Female
8564 100
9290 100
4941 100
1114 100
31–45 years
46–60 years
61–75 years
Above75 years
8793 100
7518 100
Central China
Western China
Illiterate/ Semi−literate
9007 100
8959 100
Urban
Education
20,386 100
Rural
Residency
13,034 100
Eastern China
Region
5436 100
18–30 years
Age
14,337 100
29,345 100
4152
5915
10,949
3194
5428
8242
606
2785
5270
4918
3285
7724
9140
16,864
N
% (95% CI)
Depressive Symptoms (16 ≤ CES-D ≤ 28) N
45.14 (43.74−46.54)
66.10 (64.68−67.53)
53.84 (52.92−54.75)
46.63 (44.87−48.37)
61.07 (59.76−62.38)
61.85 (60.70−63.01)
50.31 (45.97−54.65)
56.02 (54.09−57.95)
58.02 (56.69−59.35)
58.52 (57.09−59.94)
60.22 (58.35−62.09)
51.81 (52.93−65.69)
64.62 (63.54−65.70)
572
312
461
98
319
503
333
92
894
451
4039 45.87 (44.45−47.28)
2804 31.49 (30.09−32.89)
816
240
8332 41.21 (40.30−42.12) 1105
3752 47.98 (46.24−49.72)
3053 35.31 (34.02−36.59)
4331 34.42 (33.29−35.56)
410 40.04 (35.76−44.31)
1837 37.35 (35.47−39.23)
3517 37.03 (35.73−38.32)
3313 38.23 (36.83−39.63)
2059 38.35 (36.49−40.21)
6390 42.81 (33.74−43.91)
4746 32.68 (31.62−41.70)
8.99 (8.11−9.87)
2.41 (1.98−2.83)
4.95 (4.54−5.37)
5.39 (4.63−6.15)
3.62 (3.14−4.09)
3.72 (3.26−4.19)
9.65 (6.91−12.39)
6.63 (5.62−7.64)
4.95 (4.37−5.54)
3.25 (2.74−3.76)
1.43 (1.01−1.85)
5.38 (3.07−5.88)
2.70 (2.34−4.89)
4.08 (3.76−4.39)
% (95% CI)
Depression (CES-D > 28)
58.06 (57.28−58.85) 11,136 37.86 (37.09−38.63) 1345
% (95% CI)
N
N
%
Mentally Healthy (0 ≤ CES-D ≤ 15)
Whole Sample
Male
Gender
Total
Variable
Table 2.3 Cross-tabulation of mental health status and individual characteristics
17.55
14.13
16.00
16.94
14.88
14.88
16.85
15.97
15.47
15.18
14.78
16.29
14.38
15.53
Mean
7.63
6.02
6.80
7.01
6.31
6.32
7.71
7.23
6.95
6.33
5.55
6.94
6.14
6.63
S.D.
Group-wise CES-D P
0.00
(continued)
1153.1
416.60 0.00
982.67 0.00
217.58 0.00
492.42 0.00
χ2
Chi-square Test
26 2 The Prevalence of Depression and Depressive Symptoms Among Adults …
409 100
1708 100
Divorced
Widowed
OLF-2 (other)
OLF-1 (disabled)
Retired
Unemployed
Employed
2272 100
909 100
3924 100
934 100
21,306 100
24,834 100
Married
Employment
2394 100
Single
Marital Status
50 100
1421 100
Junior college
804 100
3863 100
High school
Graduate
7872 100
University
6328 100
Middle school
3655
1338
303
2315
541
12,367
732
207
14,553
1372
38
578
955
2473
5013
59.56 (56.66–62.47)
37.30 (32.77–41.83)
57.10 (54.80–59.41)
55.63 (51.21–60.05)
58.99 (58.08–59.89)
43.87 (40.55–47.20)
47.66 (40.64–54.68)
59.04 (58.20−59.89)
59.29 (56.07−62.02)
66.30 (46.56−86.03)
69.81 (65.18−74.44)
67.56 (64.00−71.11)
65.99 (63.92−68.04)
63.07 (61.60−64.53)
57.77 (56.12−59.42)
% (95% CI)
N
N
%
Mentally Healthy (0 ≤ CES-D ≤ 15)
Whole Sample
Primary school
Variable
Table 2.3 (continued)
N
68
0
4
18
83
174
250
N
178 46.58 (39.48–53.67)
871 37.82 (34.93–40.70)
452 47.29 (42.81–51.78)
1389 36.68 (34.43–38.92)
363 40.90 (36.50–45.31)
8061 37.53 (36.65–38.42)
767 43.70 (40.40–46.99)
24
63
154
220
30
878
209
16.94
15.18
14.98
14.12
13.75
13.84
14.10
14.45
15.32
Mean
15.73
15.32
15.18
2.62 (1.80–3.43)
14.90
15.41 (12.43–18.39) 19.70
6.22 (0.50–0.74)
3.46 (1.84–5.08)
3.48 (3.15–3.81)
6.05
8.44
7.02
6.12
6.47
8.21
6.86
6.51
5.97
4.66
4.95
5.18
5.78
5.86
6.27
S.D.
Group-wise CES-D
12.43 (10.06–14.79) 18.19
5.76 (2.53–8.99)
3.72 (3.40−4.05)
2.16 (1.45−2.88)
0.00 (0.00−0.00)
0.69 (0.00−0.14)
1.43 (0.44−2.41)
2.08 (1.50−2.66)
1.74 (1.38−2.10)
3.61 (3.01−4.20)
% (95% CI)
Depression (CES-D > 28)
9237 37.23 (36.40−38.06) 1044
954 38.54 (35.84−41.25)
12 33.70 (13.97−53.44)
222 29.49 (24.89−34.09)
448 31.01 (27.51−34.51)
1307 31.93 (29.90−33.97)
2685 35.19 (33.73−36.64)
2423 38.62 (36.99−40.25)
% (95% CI)
Depressive Symptoms (16 ≤ CES-D ≤ 28) P
(continued)
478.00 0.00
584.65 0.00
χ2
Chi-square Test
2.4 Estimation Results 27
8730 100
3245 100
5–30
Above 30
2308
5494
9062
70.43 (68.14–72.71)
62.67 (61.23–64.10)
53.06 (52.04–54.07)
N
897 28.26 (26.01–30.50)
2987 34.65 (33.24–36.05)
7252 41.56 (40.55–42.56)
% (95% CI)
Depressive Symptoms (16 ≤ CES-D ≤ 28)
40
249
1056
N
1.32 (0.01–1.94)
2.68 (2.20–3.16)
5.39 (4.93–5.84)
% (95% CI)
Depression (CES-D > 28)
13.49
14.56
16.16
Mean
5.23
5.98
7.00
S.D.
Group-wise CES-D P
645.93 0.00
χ2
Chi-square Test
Note (1) Data source: China Family Panel Studies (2012). (2) The statistics reported are the group-wise CES-D scores and prevalence rates of depressive symptoms and severe depression by personal characteristics of the respondents, followed by the 95% confidence intervals of the prevalence rate estimates in parenthesis. (3) Column title “N” represents the number of respondents in each sample group; column title “%” represents the nationally representative prevalence rates in each sample group (adjusted with the official sampling weights of CFPS). (4) Mean, SD, min, and max represent the weighted group mean (based on CFPS sampling weights), standard deviation, the minimum and the maximum of CES-D scores in each group. (5) χ2 and P represent the Chi-square statistics and p-values associated with the Chi-square test on whether the variables are significantly correlated with the mental health group categorization
17,370 100
% (95% CI)
N
N
%
Mentally Healthy (0 ≤ CES-D ≤ 15)
Whole Sample
0–5
Income (k)
Variable
Table 2.3 (continued)
28 2 The Prevalence of Depression and Depressive Symptoms Among Adults …
2.4 Estimation Results
29
healthy and depressive symptom categories. Under the assumption that the CFPS sample is nationally representative and the cut-off points set by Radloff (1977, 1991) are suitable for the diagnosis of mental disorders among the Chinese population, the above statistics can be translated to national estimates based on China’s 2012 census data: with an adult population size of 1.04 billion, it is estimated that 393.74 million adults (37.86%) are experiencing depressive symptoms, among whom 202.16 million are men and 191.58 million are women; additionally, there are 42.43 million adults (4.08%) suffering from severe depression, among whom 21.78 million are men and 20.65 million are women.6 The gender difference in the prevalence rates is striking. Not only are there fewer females in the mentally healthy group in comparison to the male (51.81% vs. 64.62%), but significantly more female respondents are found in the depressive symptom group (42.81% vs. 32.68%) and the severe depression group (5.38% vs. 2.70%). The mean CES-D score for women is also higher than men (16.29 vs. 14.38), indicating that women are on average more depressive in China. We also find that age is correlated with people’s likelihood of being depressed. From the young adult group (aged 18–30) to the oldest group (75+), the probability of being mentally healthy drops from 60.22 to 50.31%, while the prevalence of depression increases from 1.43 to 9.65%, indicating that older adults are more likely to be depressed. Also, the standard deviation of CES-D scores increases from 5.55 to 7.71 as age grows, indicating that the divergence of mental health status increases along with aging. This possibly reflects the social phenomenon of “empty-nest elderly” in China: a large number of elderly parents are living by themselves due to China’s onechild policy and the large-scale labor migration, contributing to the high prevalence of depression and other mental disorders among the left-behind. In previous research, Gao et al. (2014) and Ma et al. (2012) suggest that the “empty-nest syndrome” and the lack of social support are the main causes of depression among the elderly. On the other hand, we also find young adults are at higher risks of depressive symptoms (38.35%) compared to the more senior adults, which is consistent with the findings in other developing countries such as Asante et al. (2015). The regional disparity in mental health status is noteworthy. In western China, only 46.63% respondents are mentally healthy, while the rate is 61.85 and 61.07% in the eastern and central regions. The relatively high prevalence of depression in western China (5.39%) is also in sharp contrast with eastern and central China (3.72 and 3.62%). Moreover, our data indicate that the western region tends to have higher prevalence rate of depressive symptoms (47.98%) relative to that of central (35.31%) and eastern regions (34.42%). The mentally healthy status, on the other hand, is more likely to be seen in the eastern and central regions (with prevalence rates above 61%) and less prevalent in the western region (with a prevalence rate of 46.63%). The urban population seems to suffer less from depression according to our analysis. The mentally healthy rate reaches 66.10% for the urban sample, which is 8.04%
6
The numbers of men and women are calculated using the sex-ratio of 2012 in China (105.51: 100) from the China Statistical Year Book.
30
2 The Prevalence of Depression and Depressive Symptoms Among Adults …
higher than the national average (58.06%). In comparison, the prevalence of depression in rural areas is 4.95%, which is about twice as high as the urban rate (2.41%), indicating that the mental disorder in rural China is likely to be a major source of concern for the health policymakers. This finding is also supported by prior studies such as Zhou et al. (2014) and Phillips et al. (2009), who report that the prevalence of depressive conditions is higher among the rural population in China. The sample probability of being mentally healthy generally increases with education levels: it is 45.14% for the illiterate/semiliterate group and 69.81% for the university graduates. However, there is a slight but noticeable drop beyond the university level; i.e. the probability decreases from 69.81% (university) to 66.30% (master/ doctoral degrees), indicating that the respondents who receive graduate-level education are at higher risks of depressive symptoms compared with the undergraduate degree earners. Thus, we hypothesize that education may be nonlinearly correlated with depression risks, and will formally test this relationship in the regression analysis. In addition, the standard deviation of the CES-D scores decreases (from 7.63 to 4.66) with the education attainment, suggesting that education may contribute to a more concentrated distribution of mental health status in the adult population. For marital status, we find a significant and negative association between divorce/ widowhood and people’s mental health: for example, the widowed group has the lowest rate of being mentally healthy (43.87%) and the highest prevalence of severe depression (12.43%) among all the marital subsamples. This suggests that marriage dissolution due to divorce or losing partners may present great risks of depression, a finding also supported by Weissman et al. (1996) who conclude based on a 10country study that those who are separated or divorced have significantly higher rates of major depression than married people. Being unemployed is associated with higher risks of depressive symptoms (40.90%) in comparison to the employed (37.53%), indicating that those who are unable to find jobs are likely to suffer from frustration, nervousness and stress. Being out of the labor force can be a result of several reasons: retirement is not correlated with deteriorated mental health status, but being out of the labor force due to disability or disease-related reasons is positively correlated with a higher prevalence of depression (15.41%); the other reasons that lead to the out-of-labor-force status tend to be negatively associated with the depressive symptoms, possibly because these reasons include schooling, caring for family members and being wealthy enough not to work. In another socioeconomic dimension, we categorize our sample into three groups according to the individual’s annual income, and the results indicate that higher income is positively correlated with people’s mental health: among those with annual income above 30,000 yuan, 70.43% are mentally healthy, which is much higher than those with less income (53.06% for the below 5000 yuan group, 62.67% for the 5000– 30,000 yuan group); the prevalence of depressive symptoms decreases from 41.56 to 28.26% as income grows, and the prevalence of depression in the high-income group (1.32%) is also significantly lower than the low-income group (5.39%). It is also worth noting that the Chi-square tests in Table 2.3 indicate significant discriminant power of individual characteristics on the mental health status (χ2 ≥
2.4 Estimation Results
31
217.58, p < 0.01). In the following section, we will further examine their statistical correlation through regression analyses.
2.4.2 Regression Results The marginal effect estimates in the benchmark regressions are presented in Table 2.4, with columns (1)–(3) corresponding to being mentally healthy (mhs = 0), having depressive symptoms (mhs = 1) and suffering from depression (mhs = 2). In addition to the variables reported in Table 2.2, we also include a set of province dummies in the regressions to control for the unobserved heterogeneity among geographical regions (such as the economic, cultural or institutional factors) that are directly correlated with individual mental health status. We first examine the impacts of demographic characteristics. The estimation results reflect noticeable gender disparity in mental health status. In comparison to women, men are 11.0% more likely to be mentally healthy, and are 8.9% less likely to suffer from depressive symptoms. This gender disparity is also demonstrated by comparable international studies such as Weissman et al. (1993), who find that the rate of major depression is higher for women in most countries. One additional year in age tends to make the respondents 0.5% less likely to be mentally healthy, 0.4% more likely to have depressive symptoms, and 0.1% more likely to have depression. Though quantitatively small, age tends to have an inverted U-shape effect on the probability of having depressive symptoms and depression. The nonlinear pattern may result from impaired recognition of mood and motivational symptoms among the elderly compared to the middle-age or near-elderly people (Hegeman et al. 2015). The results indicate that the diagnosis and prevention of mental health conditions is more difficult for the elderly population, thus more policy attention should be devoted to this group. Married people are 7.6% more likely to be mentally healthy and 6.0% less likely to suffer from depressive symptoms compared to the single people. On the other hand, those who are divorced and widowed see a 4.2 and 3.8% increase in the probability of having depressive symptoms and are associated with 5.4 and 4.9% less chance of being mentally healthy in comparison to the single group. The above estimates indicate that marriage might play a positive role in improving people’s mental health, which may be explained by the less emotional stress experienced among married people (Kessler and Essex 1982).7 The negative correlation between marriage dissolution (due to divorce or loss of a partner) and mental health status is also supported by other empirical studies (e.g. Breslau et al. 2011; Keller et al. 2007; Kendler et al. 2002).
7
Beacha et al. (1998) and Jacobson et al. (1991) even suggest using marital therapy to treat depression.
32
2 The Prevalence of Depression and Depressive Symptoms Among Adults …
Table 2.4 Benchmark regression result (ordered probit model) Variable
(1)
(2)
(3)
mhs = 0
mhs = 1
mhs = 2
0.110***
− 0.089***
− 0.021
(0.006)
(0.005)
(0.000)
− 0.005***
0.004***
0.001***
(0.001)
(0.001)
(0.000)
4.08·10–5 ***
− 3.29·10–5 ***
− 7.87·10–6 ***
(0.000)
(0.000)
(0.000)
− 0.203***
0.155***
0.048
(0.038)
(0.027)
(0.000)
West
0.775
− 0.582
− 0.193
(0.019)
(0.007)
(0.017)
Urban
0.039***
− 0.031***
− 0.007
(0.015)
(0.012)
(0.000)
0.068***
− 0.056***
− 0.012
(0.008)
(0.007)
(0.000)
Middle
0.109***
− 0.090***
− 0.019
(0.009)
(0.008)
(0.000)
High
0.097***
− 0.081***
− 0.016
(0.012)
(0.010)
(0.000)
0.108***
− 0.091***
− 0.017
(0.016)
(0.014)
(0.000)
University
0.141***
− 0.120***
− 0.021
(0.020)
(0.017)
(0.000)
Graduate
0.146**
− 0.125**
− 0.021
(0.061)
(0.055)
(0.000)
0.076***
− 0.060***
− 0.017
(0.013)
(0.010)
(0.000)
Divorced
− 0.054**
0.042**
0.012
(0.026)
(0.020)
(0.000)
Widowed
− 0.049***
0.038***
0.010
(0.018)
(0.014)
(0.000)
− 0.026
0.020
0.005
(0.019)
(0.015)
(0.004)
Retired
0.036***
− 0.029***
− 0.007
(0.013)
(0.011)
(0.000)
OLF-1
− 0.251***
0.169***
0.082***
Male Age Age2 Central
Primary
Junior
Married
Unemployed
(continued)
2.4 Estimation Results
33
Table 2.4 (continued) Variable
(1)
(2)
(3)
mhs = 0
mhs = 1
mhs = 2
(0.017)
(0.009)
(0.008)
0.020*
− 0.016*
− 0.004
(0.012)
(0.010)
(0.000)
0.001***
− 0.001***
− 0.000***
(0.000)
(0.000)
(0.000)
Incomes2
− 2.46 · 10–13 ***
1.98 · 10–13 ***
4.75 · 10–14 ***
(0.000)
(0.000)
(0.000)
Province dummies
Yes
Yes
Yes
OLF-2 Income
Note (1) Data source: China Family Panel Studies (2012). The statistics reported are the marginal effects of independent variables, with the clustered robust standard errors (clustered at the community level) in parentheses. (2) ***, ** and * denote statistical significance at 1, 5 and 10% levels, respectively. (3) The sample size for the benchmark ordered probit model is 29,345
Compared with eastern China, respondents in the central region are 20.3% less likely to be mentally healthy and 15.5% more likely to experience depressive symptoms. However, living in western China has no significant correlation with people’s mental health, which suggests that the seemingly strong geographical disadvantage of the western residents (as indicated by Table 2.3) is no longer significant after the other personal and environmental characteristics are controlled. One possible explanation is that economic development tends to cluster in eastern and central China, leaving the western regions in disadvantage in economic growth; thus, the respondents’ personal income, which is highly correlated with the regional per-capita GDP and individual mental health status, may mediate (dilute) the influence of the geographical variable. Compared with the rural residents, respondents living in the urban areas have 3.1% lower likelihood of having depressive symptoms and are 3.9% more likely to be mentally healthy, which may be explained by the higher quality of life and the relatively more abundant mental healthcare resources in urban China (Qin and Hsieh 2014). Prior studies such as Probst et al. (2006) and McCall-Hosenfeld et al. (2014) also suggest that depression is more prevalent in the rural areas in US, which is largely caused by the poor accessibility to community mental health facilities and professional psychiatrists in these areas. Among the socioeconomic factors, education is positively related to people’s mental health status: compared to the illiterate/semi-literate group, those with primary, secondary and tertiary education are 6.8–14.6% more likely to be mentally healthy and 5.6–12.5% less likely to have depressive symptoms. The educationdepression gradient may be caused by the better living and social environment and the stress management capabilities acquired through the accumulation of human capital. These findings are consistent with Miech and Shanahan (2000) and Bjellanda (2008), who suggest that low education levels are significantly associated with
34
2 The Prevalence of Depression and Depressive Symptoms Among Adults …
both anxiety and depression and high educational attainment has a protective effect against these mental disorders. Compared to the employed individuals, those who are out of the labor force due to disability or diseases are 25.1% less likely to be mentally healthy and 8.2% more likely to be depressed; this indicates that the unintentional loss of a job due to physical reasons might coincide with worse mental health status. On the other hand, we find that being retired or out of the labor force due to non-medical reasons tend to be positively correlated with people’s mental health. More specifically, they are associated with 3.6 and 2.0% higher likelihood of being mentally healthy and 2.9 and 1.6% less risks of depressive symptoms, respectively; this is possibly attributable to the less work-related stress and more leisure time enjoyed by the retirees and nonworking adults. Contrary to the cross-tabulation results, being unemployed does not seem to be associated with deterioration in mental health status when other variables are controlled. Lastly, Table 2.4 suggests a nonlinear correlation between income and mental health status: an increase in income is positively correlated with people’s mental health and negatively associated with their possibility of having depressive symptoms and depression, though the marginal effect is diminishing with incremental income growth. Previous research such as Easterlin (2001) and Dluhosch et al. (2014) also suggests that there exists an income-happiness nexus, and the correlation between income and mental health is not linear. One plausible explanation is that more income can increase people’s life satisfaction through entertainment and other consumptions. However, as noted by Hall and Jones (2007), the marginal utility per dollar of nonhealthcare consumption declines as people become more affluent. Moreover, the additional earnings may also come with the work-related stress and even the anxiety of wealth management. As a result, the income effect on happiness may diminish very fast at high-income levels.
2.5 Robustness Tests In addition to the benchmark model, we also conduct sensitivity tests with alternative sample selection criteria and model specifications to check the robustness of our findings. The results of these robustness tests are presented in Table 2.5. For comparison purposes, we show the coefficient estimates (instead of marginal effects) of the ordered probit model under different specifications [including the benchmark specification as reported in Column (1)], which represent the correlation between explanatory variables and the tendency of depression (i.e. coefficient β in Eq. 2.1). First, we convert the education dummy variables into a continuous variable representing the respondents’ years of formal schooling to test if different measures of socioeconomic status may change the regression results. The use of a continuous education measurement (and its squared term) also enables us to formally test the potential nonlinear association between education and people’s mental health. The estimation results of this model are presented under “Robustness Test 1”, which
Junior
High
Middle
Primary
Urban
West
Central
Age2
Age
Male
Variable
(0.039)
(0.280)
− 0.099***
(0.038)
(0.224)
− 0.099***
(0.043)
(0.044)
(0.041)
(0.031) − 0.295***
− 0.268***
− 0.255***
(0.031)
(0.024)
(0.025)
− 0.286***
− 0.212***
− 0.288***
− 0.285***
(0.059)
− 0.184***
(0.034)
− 0.269***
(0.033)
(0.023)
(0.024)
− 0.165***
(0.043)
− 0.091***
(0.089)
0.868
(0.106)
0.452***
(0.000)
− 0.000***
(0.004)
0.015***
− 0.180***
(0.249)
− 3.635
(4) Robust test 3
− 0.175***
(0.038)
− 0.102***
− 3.820
− 3.740
(0.122)
(0.097)
(0.000) 0.369***
(0.098)
(0.000)
(0.000)
− 0.000***
(0.004)
0.016***
(0.015)
− 0.293***
0.515***
− 0.000***
− 0.000***
(3) Robust test 2
0.516***
(0.004)
(0.015)
0.012***
(0.015)
− 0.283***
(0.004)
− 0.283***
Benchmark model
0.012***
(2)
Robust test 1
(1)
Table 2.5 Robustness test results (5)
(0.057)
− 0.353***
(0.042)
− 0.281***
(0.034)
− 0.293***
(0.033)
− 0.188***
(0.042)
− 0.100**
(0.265)
− 3.918
(0.155)
0.557***
(0.000)
− 0.000
(0.005)
0.008*
Robust test 4
(6)
(0.065)
− 0.324***
(0.038)
− 0.239***
(0.029)
− 0.272***
(0.028)
− 0.170***
(0.405)
− 3.756
(0.342)
0.672**
(0.000)
− 0.000***
(0.004)
0.019***
(0.019)
− 0.281***
Robust test 5
(7)
(continued)
(0.062)
− 0.293***
(0.054)
− 0.268***
(0.050)
− 0.275***
(0.051)
− 0.145***
(0.117)
0.706***
(0.096)
0.354***
(0.000)
0.000
(0.006)
− 0.006
(0.026)
− 0.301***
Robust test 6
2.5 Robustness Tests 35
Income2
Income
OLF-2
OLF-1
Retired
Unemployed
Widowed
Divorced
Married
Graduate
University
Variable
(3)
(0.034)
(0.034)
0.000***
0.000***
(0.001)
(0.001)
(0.033)
(0.031)
− 0.002***
(0.031)
− 0.002***
− 0.040
− 0.052*
− 0.052*
0.000***
(0.001)
− 0.002***
(0.045)
(0.046)
0.640***
0.645***
(0.046)
(0.034)
0.644***
(0.034)
(0.034)
− 0.092***
(0.048)
(0.049)
− 0.093***
(0.048)
− 0.092***
0.063
(0.044)
0.111**
(0.064)
0.149**
0.067
(0.046)
(0.046)
0.065
0.121***
0.124***
(0.065)
− 0.219***
− 0.195***
− 0.193***
(0.065)
(0.173)
(0.183) (0.033)
− 0.373*
− 0.398**
0.136**
(0.060)
0.137**
− 0.403***
Robust test 2
(0.057)
Benchmark model
− 0.381***
(2)
Robust test 1
(1)
Table 2.5 (continued) (4)
0.000
(0.001)
− 0.005***
(0.033)
− 0.076**
(0.059)
0.528***
(0.039)
− 0.131***
(0.064)
0.078
(0.063)
0.190***
(0.094)
0.186*
(0.051)
− 0.109**
(0.240)
− 0.157
(0.078)
− 0.295***
Robust test 3
(5)
0.000*
(0.001)
− 0.001*
(0.069)
− 0.019
(0.069)
0.792***
(0.051)
− 0.047
(0.069)
0.055
(0.072)
0.131*
(0.090)
0.110
(0.043)
− 0.242***
(0.310)
− 0.612**
(0.077)
− 0.434***
Robust test 4
(6)
(7)
(0.072)
0.000**
− 0.000
(continued)
(0.001)
− 0.001**
(0.047)
− 0.044
(0.080)
0.706***
(0.054)
− 0.126**
(0.001)
− 0.002*
(0.042)
− 0.051
(0.056)
0.634***
(0.044)
0.001
(0.062)
(0.078) 0.190***
− 0.038
0.239***
(0.092)
0.267***
(0.063)
− 0.073
(0.201)
− 0.362*
(0.077)
− 0.430***
Robust test 6
(0.057)
0.057
(0.098)
0.082
(0.040)
− 0.243***
(0.438)
− 0.740*
(0.102)
− 0.344***
Robust test 5
36 2 The Prevalence of Depression and Depressive Symptoms Among Adults …
29,345
Yes
(0.000)
0.001***
(0.005)
(3)
29,345
Yes
(0.000)
Robust test 2
(4)
15,008
Yes
(0.000)
Robust test 3
Note ***, ** and * denote statistical significance at 1%, 5% and 10% levels, respectively
Yes
29,345
(0.000)
Observations
(0.000)
Benchmark model
− 0.040***
(2)
Robust test 1
(1)
Province
Eduyear2
Eduyear
Variable
Table 2.5 (continued) (5)
14,337
Yes
(0.000)
Robust test 4
(6)
20,386
Yes
(0.000)
Robust test 5
(7)
8959
Yes
(0.000)
Robust test 6
2.5 Robustness Tests 37
38
2 The Prevalence of Depression and Depressive Symptoms Among Adults …
suggest that there exists a U-shaped correlation between education and the tendency of depression. The other coefficient estimates are quantitatively similar to those in the benchmark ordered probit model; thus, they validate our main findings in Sect. 2.4. Second, we change the cut-off points of CES-D scores to examine if different thresholds of mental health categories will influence our estimates. As previous studies suggest, the CES-D classification of mental health status may vary by countries. Thus, we follow Chen et al. (2009) and use the CES-D scores of 17, 23, and 28 as the threshold values of being mentally healthy (0–17), possible depression (18–23), probable depression (24–28), and severe depression (above 28), as these CES-D categories may provide more accurate estimates on the depression prevalence among the Chinese population. The associated regression results are presented under “Robustness Test 2”, which are shown to be consistent with the benchmark model. Moreover, we conduct subsample analyses on the male and female respondents separately to explore the gender heterogeneity in the determinants of mental health status. Both prior studies and our cross-tabulation analysis reflect noticeable gender differences in mental health, which in turn suggests that the prevalence of depressive symptoms and depression may be driven by different risk factors between men and women. The subsample estimation results are reported under “Robustness Test 3” (for the female sample) and “Robustness Test 4” (for the male sample), which indicate considerable parameter heterogeneity across genders. First, in terms of improving mental health and decreasing depression risks, marriage might play a more productive role for men while income might be more important for women. Second, divorce significantly correlates with female’s tendency of having depressive symptoms and depression, while it does not have significant association with men’s mental health status. Third, being retired or out of the labor force for non-healthrelated reasons might play a positive role for women’s mental health, while these employment statuses are not significantly correlated with men’s mental health status. Lastly, given the large disparity in mental health status between the urban and rural samples (as shown by Table 2.3), we explore whether the regression coefficients are heterogeneous between the two sectors. The subsample estimation results are reported under “Robustness Test 5” (for the rural sample) and “Robustness Test 6” (for the urban sample), which show substantial urban–rural differences in the statistical association between MHS and individual characteristics. For example, age has a significant and nonlinear correlation with the probabilities of depression and depressive symptoms among the rural residents, while this correlation is not statistically significant among the urban respondents. In addition, certain traumatic experiences, such as marital disruption, the loss of a spouse or a job, tend to have stronger association with the propensity of depression and depressive symptoms in urban areas, indicating that the urban residents may have more difficulty overcoming these emotional shocks (possibly due to the lack of family or social support). These urban–rural differences in the mental health determinants may have contributed to the overall disparity in the prevalence of depressive symptoms and depression in the two sectors.
2.6 Discussions
39
2.6 Discussions Our empirical analyses yield two important findings. First, the prevalence of depressive symptoms and depression is high, indicating that the prevention and treatment of depression are important tasks of future healthcare reforms in China. Second, the prevalence rates of depressive symptoms and depression are highly concentrated in certain subpopulations. After controlling for other factors, our analyses highlight four subpopulation groups who are vulnerable to mental health problems, including the elderly, rural residents, the poor, and the low-educated people. This finding suggests that multiple risk factors are at play to account for the overall high prevalence of depressive symptoms and depression in China. Although our data lack the ability to make causal claims, it is important to further discuss the potential mechanisms that may cause a high correlation between these risk factors and the prevalence rates. In the following, we also discuss the potential consequences of the high prevalence rates in these four subpopulation groups. In both univariate and multi-variate analyses, we find that age is strongly correlated with the prevalence rates. Although this is consistent with the stylized fact that the depreciation rate of health capital increase with age, it is important to note that two important socioeconomic factors in China may accelerate the mental health problems among the elderly population. First, due to the labor mobility and the onechild policy, many elderly people may be forced to live alone or live with spouse only. The decline of multiple-generation families in China implies that the family support from the younger generation(s) is shrinking, which in turn is an important mechanism to account for the high depression rate among the elderly. The problem of “empty-nest elderly” is particularly severe in rural China and is shown to be correlated with elevated suicide rates among the rural population over age 65 (Wang et al. 2014). Second, compared to physical diseases, the most distinctive feature of mental illness is stigma, which has been widely recognized as the main barrier to the mental healthcare utilization (Bharadwaj et al. 2015). Although the scientific evidence on the existence and consequences of stigma is still lacking in China, it is reasonable to speculate that stigma is another mechanism that leads to the lack of effective prevention for mental disorders and hence a higher prevalence rate of depression among the elderly population. It has been widely known that China now faces a series of challenges arising from population aging. Based on the results of our study, it is reasonable to expect that one of the challenging issues in this field is the rapid increase in the prevalence of mental health problems. Many studies have accumulated the evidence that mental disorders such as depression are often co-existent with other NCDs such as diabetes and hypertensions (Patel and Chatterji 2015). The co-morbidity of depression and other NCDs will become the double burden of rising healthcare costs in the future. One potential way to relieve the threat is to transform the current health system in China from hospital-centered care to person-centered care, which in turn has been widely recognized as an important step toward an efficient health system (Yip and Hsiao 2014). One of the major ingredients of the person-centered care is to integrate
40
2 The Prevalence of Depression and Depressive Symptoms Among Adults …
the treatment of mental health problems and other NCDs in the primary care platforms (Patel and Chatterji 2015). Our results also highlight a stylized fact that rural residents are more likely to have depressive symptoms and depression than the urban residents after controlling all other variables. There are two plausible explanations for this outcome. Frist, China imposes a strong regulation on household registration system, which in turn affects the eligibility of many social programs including public education and health insurance between the rural and urban residents. As a result, the rural residents are put into a disadvantaged position to access the high-quality healthcare and education. For example, the per-capita funds for health insurance system offered to rural residents was 61 Yuan in 2003, which was about one seventh of that offered to the urban employed people (Meng et al. 2015). Second, as in other countries, China’s healthcare resources are also geographically maldistributed (Qin and Hsieh 2014), and the recent development of opening China’s healthcare market to private capital may even enlarge the inequality of healthcare resources between the rural and urban areas (Yip and Hsiao 2015). In summary, the disparity between the rural and urban areas in many dimensions, especially in access to education and healthcare, has put rural residents to disadvantage for their future investment on human capital, which in turn becomes a major source of depressive symptoms and depression. The third stylized fact we found is that the poor is more likely to have depression than the rich. Although this is consistent with the conventional wisdom on the nexus of wealth and health, it is important to highlight two institutional features in China that may have played a crucial role in causing this outcome. First, China currently still lacks a comprehensive and strong social security system, indicating that many low-income people may not have sufficient protection on their income security in the future, and numerous studies show that the insecurity of future income is a major source of depression. Second, the rapid economic growth in the past few decades was accompanied by an increase in income inequality in China. According to the relative income hypothesis, the distribution of income is also an important determinant of mental health (Jones and Wildman 2008). Thus, the rising inequality in income (and other related socioeconomic outcomes) may also play an important role in accounting for the high prevalence of mental health problems among the poor. Lastly, we find a stylized fact that poorly educated people are more likely to have depressive symptoms and depression. Although this is consistent with the evidence on the education-health gradient, it is also important to note the underlying mechanism given that the low-educated people form the largest group in our study sample (30% is illiterate and semi-illiterate). Cutler and Lleras-Muney (2010) identified several important mechanisms that potentially account for the positive relationship between education and a series of health behaviors that are important inputs for the production of good health. These mechanisms include (1) education as command over resources; (2) specific knowledge; (3) cognitive ability; (4) taste; (5) personality; and (6) social integration. Based on several datasets obtained from US and UK, Cutler and Lleras-Muney (2010) conclude that specific knowledge and cognitive ability working together can account for about 42% of the education-health gradient. Although we are lacking data to provide a similar test in China, we speculate that this conclusion
2.7 Conclusions
41
is valid in China’s context; i.e. given the large share of low-educated population, specific health knowledge and cognitive ability in processing health information will play an important role in shaping good mental health. This speculation suggests that an important channel for the government to improve the population mental health is through health promotion and health education, which in turn reinforces the notion on the integration of mental health and other NCDs in the primary care system.
2.7 Conclusions In this paper, we provide an overview of the prevalence of depressive symptoms and depression among the adult population in China and explore how the prevalence varies with different demographic, socioeconomic and geographic characteristics using both cross-tabulations and the ordered probit regressions. Based on the nationally representative CFPS 2012 dataset and the internationally comparable CES-D metrics, our findings indicate that the overall mental health status in China is worrisome, with about 4.08% of adults suffering from depression and 37.86% frequently experiencing depressive symptoms. If not properly addressed, the particularly high prevalence of depressive symptoms will translate to major threat to China’s public health system as these minor mental health conditions can develop into severe depression over time. Moreover, we find that the above prevalence rates are distributed unevenly across geographical regions and subpopulations. For example, women and older people are more likely to be depressed, indicating the mental health disadvantage of these vulnerable groups; better socioeconomic status (such as education and income) is shown to be positively correlated with mental health with diminishing marginal effects; the prevalence rates of depressive symptoms also tend to be higher in the rural areas and among the central and western provinces. Our results have three important policy implications. First, the estimated high prevalence of depressive symptoms and depression suggests that the potential medical costs of treating these conditions and their co-morbidities may also be large. Our analyses thus call for higher social awareness of mental health problems in China, and they point to the importance of further government involvement in the treatment and prevention of depression, possibly through legal, regulation, and policy changes. In particular, a well-established mental health management system should be constructed with the alliance of government agencies, public and private hospitals, psychological consultation entities, and other healthcare institutions. Second, given the significant socioeconomic gradient in mental health, we hypothesize that the inequalities in income and education among regions and subpopulations may be an important contributing factor to the mental depression and depressive symptoms among Chinese adults (particularly in its economically less developed regions). Thus, any policy design toward the reduction of such disparities may serve as an effective channel to improve the mental health status for the Chinese population. Such policies may include, for example, rebalancing the economic growth
42
2 The Prevalence of Depression and Depressive Symptoms Among Adults …
model toward equity and social welfare, accelerating the economic and educational investment in the under-developed areas, etc. Third, the relatively higher prevalence of depression in the inland (central and western) and rural regions indicates greater demand for mental healthcare in these areas. However, recent studies find significant geographic maldistribution of healthcare resources in China and that the majority of physicians and healthcare facilities are concentrated in the urban and eastern regions (Qin and Hsieh 2014). This suggests that there may exist a significant gap between the supply and demand of mental healthcare in China (particularly its inland and rural regions). A potentially effective solution is to strengthen the primary care system, as the integrated primary care-based delivery system is widely recognized as a more cost-effective system to provide quality healthcare in prevention and disease management than the hospitalcentered system (see discussions in Sect. 2.6). In particular, more primary mental health facilities should be established in the underserved inland and rural areas, where higher depression rates are found in our estimation. There are several limitations to our study. Firstly, the data were collected using selfreport measures, which may overestimate the strength of the relationship between mental health and certain variables (Lindell and Whitney 2001). Secondly, the CESD questionnaire covers a relatively short period of time—just one week before the survey, thus the indicators of mental health status may suffer from measurement errors for people who experience recent shocks (e.g. the death of a family member) or people whose mood fluctuate frequently. Thirdly, due to the existence of unobserved determinants of depression, more sophisticated statistical or econometric models may be needed to identify the causal relationship between depression and some key explanatory variables, which in turn may be an important direction for future research.
References Asante KO, Andoh-Arthur J (2015) Prevalence and determinants of depressive symptoms among University Students in Ghana. J Affect Disord (forthcoming). https://doi.org/10.1016/j.jad.2014. 09.025 Beacha S, Finchamb FD, Katzc J (1998) Marital therapy in the treatment of depression: toward a third generation of therapy and research. Clin Psychol Rev 18(6):635–661 Bharadwaj P, Pai MM, Suziedelyte A (2015) Mental health stigma. National bureau of economic research working paper 21240, Cambridge, MA Bjellanda I, Krokstadb S, Mykletunc A, Dahld AA et al (2008) Does a higher educational level protect against anxiety and depression? The HUNT study. Soc Sci Med 66(6):1334–1345 Breslau J, Miller E, Jin R, Sampson N, Alonso J, Andrade L, Kessler RC (2011) A multinational study of mental disorders, marriage, and divorce. Acta Psychiatric Scandinavica 124:474–486 Chatterji P, Markowitz S (2012) Family leave after childbirth and the mental health of new mothers. J Ment Health Policy Econ 15:61–76 Chen Z, Yang X, Li X (2009) Psychometric features of CES-D in Chinese Adolescents. Chin J Clin Psychol 27:38–42 (in Chinese)
References
43
Chen X (2012) Epidemiological survey of mental health in community population in Xi’an, China. PhD Dissertation, Fourth Military Medical University (in Chinese) Cui C, Chen H, Liu C et al (2009) Epidemiological survey of depression in Hebei Province. J Clin Psychiatry 19(2):94–96 (in Chinese) Cutler DM, Lleras-Muney A (2010) Understanding differences in health behaviors by education. J Health Econ 29:1–28 Dluhosch B, Horgos D, Zimmermann KW (2014) Social choice and social unemployment-income cleavages: new insights from happiness research. J Happiness Study 15:1513–1537 Easterlin RA (2001) Income and happiness: towards a unified theory. Econ J 111(473):465–484 Fan P, Pei J, Hou Z et al (2013) The discussion of current mental health status and mental health management strategy in China. J Pract Med Tech 20(8):911–912 (in Chinese) Gao Y, Wei Y, Shen Y et al (2014) China’s empty nest elderly need better care. J Am Geriatrics Soc 62(9):1821–1822 Hall RE, Jones CI (2007) The value of life and the rise in health spending. Quart J Econ 122(1):39–72 Hegeman JM, de Waal MW, Comijs HC et al (2015) Depression in later life: a more somatic presentation? J Affect Disord 170:196–202 Jacobson NS, Dobson K, Alan EF et al (1991) Marital therapy as a treatment for depression. J Consult Clin Psychol 59(4):547–557 Jones AA, Wildman J (2008) Health, income and relative deprivation: evidence from the BHPS. J Health Econ 27:308–324 Keller MC, Neale MC, Kendler KS (2007) Association of different adverse life events with distinct patterns of depressive symptoms. Am J Psychiatry 164:1521–1529 Kendler KS, Gardner CO, Prescott CA (2002) Toward a comprehensive developmental model for major depression in women. Am J Psychiatry 159:1133–1145 Kessler RC, Essex M (1982) Marital status and depression: the importance of coping resources. Soc Forces 61(2):484–507 Lindell MK, Whitney DJ (2001a) Accounting for common method variance in cross-sectional research designs. J Appl Psychol 86(1):114–121 Lindell MK, Whitney DJ (2001b) Accounting for common method variance in cross-sectional research designs. J Appl Psychol 86(1):114–121 Ma Y, Fu H, Wang JJ et al (2012) Study on the prevalence and risk factors of depressive symptoms among ‘empty-nest’ and ‘non-empty-nest’ elderly in four Provinces and Cities in China. Chin J Epidemiol 33:478–482 (in Chinese) McCall-Hosenfeld JS, Mukherjee SL (2014) The prevalence and correlates of lifetime psychiatric disorders and trauma exposures in urban and rural settings: results from the national comorbidity survey replication (NCS-R). PLoS One 9(11) Meng QY, Fang H, Liu X, Yuan B, Xu J (2015) Consolidating the social health insurance schemes in China: towards an equitable and efficient health system. Lancet 386(10002):1484–1492 Miech RA, Shanahan MJ (2000) Socioeconomic status and depression over the life course. J Health Soc Behav 41(2):162–176 Patel V, Chatterji S (2015) Integrating mental health in care for noncommunicable diseases: an imperative for person-centered care. Health Aff 34(9):1498–1505 Phillips MR, Zhang J, Shi Q et al (2009) Prevalence, treatment, and associated disability of mental disorders in four provinces in China during 2001–05: an epidemiological survey. Lancet 373(9680):2041–2053. https://doi.org/10.1016/S0140-6736(09)60660-7 Probst JC, Laditka SB, Moore CG, Harun N, Powell MP, Baxley EG (2006) Rural-urban differences in depression prevalence: implications for family medicine. Fam Med 38(9):653–660 Qin X, Hsieh C (2014) Economic growth and the geographic maldistribution of health care resources: evidence from China, 1949–2010. China Econ Rev 31:228–246 Radloff LS (1977) The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol 1(3):385–401 Radloff LS (1991) The use of the center for epidemiologic studies depression scale in adolescents and young adults. J Youth Adolesc 20(2):149–166
44
2 The Prevalence of Depression and Depressive Symptoms Among Adults …
Somrongthong R, Wongchalee S, Laosee O (2013) Depression among adolescents: a study in a Bangkok Slum Community. Nordic College Caring Sci 27:327–334 Tian X (2012) Research of social work intervention on the adolescent mental health problems. Master Thesis, Shandong University (in Chinese) Wang X (2014) Prevalence of depression in Qingdao and related factors analysis. Master Thesis, Qingdao University (in Chinese) Wang CW, Chan C, Yip PSF (2014) Suicide in China from 2002–2011: an update. Soc Psychiatry Psychiatr Epidemiol 49(2):211–219 Weissman MM, Bland R, Joyce PR et al (1993) Sex differences in rates of depression: cross-national perspectives. J Affect Disord 29(2–3):77–84 Weissman MM, Bland RC, Canino GJ et al (1996) Cross-national epidemiology of major depression and bipolar disorder. JAMA 276(4):293–299 World Health Organization (2012) Global burden of mental disorders and the need for a comprehensive, coordinated response from the health and social sectors at the country level. Document A65/10 Yip W, Hsiao W (2014) Harnessing the privatization of China’s fragmented health-care delivery. The Lancet 384:805–818 Yip W, Hsiao WC (2015) What drove the cycles of Chinese health system reforms? Health Syst Reform 1(1):52–61 Zhou Y (2010) Mental health social work in America and its significance to China. J Sichuan Univ (soc Sci Edn) 3:127–132 (in Chinese) Zhao Z, Huang Y, Li J et al (2009) An epidemiological survey of mental disorders in Guangzhou area. Chin J Nerv Mental Dis 35(9):530–534 (in Chinese) Zhou X, Bi B, Zheng L et al (2014) The prevalence and risk factors for depression symptoms in a rural Chinese sample population. Plos One 9(6). https://doi.org/10.1371/journal.pone.0099692 Zhu D, Xu S, Yang Q et al (2010) Prevalence of anxiety, depression and influencing factors among urban residents of Zhejiang Province. China J Publ Health 26(4):429–432 (in Chinese)
Chapter 3
Depression Hurts, Depression Costs: The Medical Spending Attributable to Depression and Depressive Symptoms in China
3.1 Introduction It has been widely recognized that mental disorders such as depression have become a fast-growing cause of disability and of the global burden of diseases (Patel et al. 2016). Around the world, more than 650 million people live with diagnosable mental disorders, which in turn imposes considerable costs on the individuals, their families, and the society as a whole (WHO 2013a). Depression is the most common type of mental disorder, with an estimated 350 million people affected, indicating that on average about 1 in every 20 people in the world are affected by depression according to the world’s current population size (WHO 2016). Several studies have documented that the impact of depression not only lies in its detrimental effect on health (such as disability and mortality) but also on economic productivity (e.g. Greenberg et al. 2015; Chisholm et al. 2016). As a result, the rising prevalence of depression is not just a public health concern but also a significant concern for economic development and social welfare. A recent policy recommendation initiated by the WHO and the World Bank Group even suggested to include mental health in the global development priority agenda (World Bank Group and WHO 2016). After enjoyed fast economic growth for several decades, China now also faces the challenge of increasing prevalence of depression, which has become one of the leading causes of disability-adjusted life years in this country (Phillips et al. 2009; Yang et al. 2013; Qin et al. 2016). Although the number of people with mental illness has increased in recent years, less research has been done on the mental health status in China compared to the developed countries, and little has been known about the impacts of depression on the overall healthcare costs. The purpose of this paper is to The content of this chapter is published in Hsieh C R, Qin X. Depression hurts, depression costs: the medical spending attributable to depression and depressive symptoms in China. Health Economics, 2018, 27(3): 525–544. Copyright John Wiley and Sons (2018), reproduction license granted.
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 X. Qin and C.-R. Hsieh, Economic Analysis of Mental Health in China, Applied Economics and Policy Studies, https://doi.org/10.1007/978-981-99-4209-1_3
45
46
3 Depression Hurts, Depression Costs: The Medical Spending …
fill this gap by estimating the medical cost attributable to depression and depressive symptoms among the adult population in China. An important challenge for this task is the under-diagnosis and under-treatment of mental depression. Unlike physical health conditions, mental health problems are more difficult to be detected and properly treated because of the added barriers to the access of mental healthcare services, especially in developing countries like China. Some of these barriers (such as stigma) are due to social factors,1 while other barriers are caused by country-specific institutional arrangement. For example, the available mental healthcare manpower in China is insufficient and unevenly distributed across regions (Liu et al. 2011), partially due to the government control of medical education and accreditation. The qualified personnel tend to be concentrated in urban-based specialty psychiatric hospitals, indicating that the mental health services are quite limited in the rural areas (Philips et al. 2009). Another example of the institutional barriers in China is the knowledge gap between the frontier of new medical technology and the local practice standards: due to the Essential Drug Policy and the regulated insurance reimbursement schedule, there may be a long delay in the launch of new mental healthcare drugs or treatment procedures in China2 ; as a result, physicians may not be able to prescribe what proves to be the most effective treatment regimes. The above psychological, social, and institutional hurdles in the access to mental healthcare in turn cause a delay in treatment or a “cost shift” toward the non-mental healthcare spending, as mental disorders such as depression often co-occur with other NCDs such as diabetes and hypertension (Patel and Chatterji 2015), and many patients with mental illness may choose to seek general healthcare instead of mental healthcare. Thus, empirical estimation on the depression-induced medical costs can be biased if such estimation is based solely on the data of mental healthcare utilization. In this paper, we try to address this issue by (i) adopting the two-part and four-part models that overcome the problem of no-use or under-use of mental health services, and (ii) using a national household survey dataset that records more comprehensive information on personal medical spending than the utilization-based datasets. To the best of our knowledge, this paper is among the first in the health economic literature to use the hurdle models (such as the two-part and four-part models) to 1
Bharadwaj et al. (2015) provide strong evidence of stigma associated with mental illness: they find that more than one-third of survey respondents under-report mental health conditions, while the respondents are less likely to under-report other physical illnesses such as diabetes or hypertension. 2 Between 2008 and 2012, there were 12 global new molecular entities available for treating mental illness. In 2013, only one of these 12 new drugs is available in China, indicating a long delay in the launch of new drugs (IMS Institute 2014). For a more comprehensive study on the launch delay of new drugs in India, see Berndt and Cockburn (2014). One of the plausible explanations for the launch delay in China as well as other low- and middle-income countries is the lack of insurance coverage for these new drugs. In addition, the three social health insurance schemes in China are heterogeneous in terms of funding sources and benefit packages, which in turn leads to a disparity in access to new drugs across different insurance programs. For example, in 2013, the per-capita fund for the rural new cooperative medical scheme (NCMS) was only 61 USD, just around 15% of the per-capita fund of urban employee basic medical insurance (UEBMI) scheme (Meng et al. 2015).
3.2 Background and Previous Research
47
characterize the cost impact of mental depression, and it provides the first nationally representative estimates on the medical costs induced by depression and depressive symptoms among adults in China. Our results indicate that depression and depressive symptoms have significant impacts on the individual expected medical expenditure. Specifically, we find that about 6.9% of total personal medical expenditure is attributable to depression. In addition, about 7.8% of total medical expenditure is attributable to the depressive symptoms. Putting together, our empirical study shows that about 14.7% of total personal medical expenditures in China are attributed to depression and depressive symptoms, which is almost three times as large as the impact of obesity and overweight on healthcare costs obtained from a previous study by Qin and Pan (2016). The significant impact of depression and depressive symptoms on healthcare costs indicate that reforming China’s mental healthcare system to cope with the increase in the disease burden is an urgent need.
3.2 Background and Previous Research The trend on the rising prevalence of depression has stimulated many studies to estimate the economic consequences of the mental health condition, which usually adopt a similar estimation approach as that for other NCDs. One of the important characteristics that distinguish mental disorders such as depression from the physical illnesses is its high likelihood of under-diagnosis and under-treatment. The World Health Organization (WHO) estimates that only 15–24% of people with severe mental disorders receive medical treatment in low- and middle-income countries. Although the treatment rate in high-income countries is higher, it is also in the range of 50–65% (WHO 2013b), indicating that at least one-third to one half of people with mental disorders go untreated. The consistent pattern of low treatment rate for mental disorders across countries highlights the importance to separate the estimation of depression-induced medical costs into two strands of research. The first strand includes the studies using patient-level data, which provides a conditional estimate on the depression-induced costs focusing on a sample of patients who received medical treatment. Given the pervasive evidence of under-treatment in mental disorders, this type of studies may suffer from the problem of underestimating the cost impact of mental depression. The second category of studies uses the population-level data, which provides an unconditional estimate on the medical costs based on a broader survey of general population including those with mental disorders but never receive medical treatment. The latter methodology hence has the advantage of avoiding the underestimation bias for the cost impact. Most existing studies adopted the first approach, using a set of sampling criteria to recruit patients from the outpatient settings in mental health clinics. Two types of costs are associated with depression: the direct cost, or the outpatient and inpatient medical cost for the treatment of depression and its complications; the indirect cost, or the opportunity cost of depression, which includes the morbidity costs caused
48
3 Depression Hurts, Depression Costs: The Medical Spending …
by absenteeism (missed work days due to depression) and presenteeism (reduced productivity while at work due to depression). Given the heterogeneity in their estimation methods and data sources, the results of these studies are difficult to compare directly. However, the existing studies did find several consistent patterns in the cost impact of mental depression. First, the estimated cost of depression is quite high and increases over time. For example, Andlin-Sobocki et al. (2005) estimated the cost of depression in 28 European countries to be e118 billion in 2005, accounting for 13% of the total healthcare expenditure of these countries. Greenberg et al. (2015) report that the economic burden of depression in the United States was $210.5 billion in 2010, an increase of 21.5% as compared to the estimated figure in 2005. Second, the estimated cost of depression is positively correlated with the disease severity. Based on data obtained from a retrospective, multicenter, non-interventional study in Switzerland, Tomonaga et al. (2013) report that the mean total direct costs per person per year, mainly due to hospitalization costs, were e3561 for mild, e9744 for moderate and e16,240 for severe depression; and the mean indirect costs per person per year, mainly due to workday losses, were e8730 for mild, e12,675 for moderate and e16,669 for severe depression. Third, the direct cost often accounts for a small portion of the total economic burden of depression, indicating that the major economic burden of depression arises from the indirect costs such as the labor productivity loss. This pattern is also quite consistent across countries: using data obtained from the United States, Greenberg et al. (2003) find that 31% of the estimated economic burden of depression was attributable to direct medical costs, 7% was suicide-related mortality costs, and 62% was workplace costs; Sobocki et al. (2007) also report a very similar result by using data from Sweden; based on data obtained from Spain, Salvador-Carulla et al. (2011) find that about 21% of the disease burden corresponded to direct costs, and 79% to indirect costs mainly due to productivity losses; in China, Hu et al. (2007) find that the direct cost accounts for about 16% of the total cost of depression, while the indirect costs contribute about 84%. As noted, the stylized facts in the epidemiological studies indicate that many people with mental disorders go undiagnosed (Trivedi et al. 2004; Bor 2015). This highlights the advantage of using population-level data to estimate the cost impact of depression. To our knowledge, there are virtually no studies applying this approach to provide an unconditional estimate of the healthcare costs attributable to depression at the population level. However, this approach is widely used to estimate the cost impact attributable to some other chronic diseases such as obesity. For example, based on the hurdle models and national medical expenditure survey data, Finkelstein et al. (2009) find that obesity accounts for an increasing share of annual medical spending in the United States, from 6.5% in 1998 to 9.1% in 2006. Their evidence suggests that the major driver of the cost growth is the increase in obesity prevalence instead of the per-capita cost inflation. By taking the endogeneity of weight into account, Cawley and Meyerhoefer (2012) show that obesity raises annual personal medical costs by $2741 (in 2005 dollars), which is equivalent to 20.6% of the US national healthcare expenditure. Based on a similar research method and data obtained from Spain,
3.3 Data Source and Descriptive Analysis
49
Mora et al. (2015) find that the obesity-induced cost varies with the severity of the condition: being severely obese is associated with 26% increase in medical costs, and this share reduces to 16 and 8.5% for moderate obesity and overweight respectively. Using longitudinal data obtained from China, Qin and Pan (2016) find that 5.29% of the total personal medical expenditures are attributable to obesity and overweight, which is equivalent to about 2.46% of China’s national health expenditure. In China, less than one-tenth of individuals with mental disorders have ever received any types of mental health services (Philips et al. 2009). This extremely low treatment rate suggests that the medical costs of depression will be underestimated if we only use the patient-level data obtained from mental healthcare utilization. As a result, the survey data that report all medical spending of an individual in a given year provide an advantage over the existing studies, as they can take into account the cost-shifting effect as well as the co-morbidity effects between mental health and physical health problems. Using the internationally comparable methods and survey instruments, we will estimate the impact of depression and depressive symptoms on the respondents’ reported annual healthcare costs and compute the expected personal medical spending attributable to these mental health conditions in China. There are several advantages of using Chinese data to provide an unconditional estimate of healthcare costs attribute to depressive symptoms and depression. First, although the under-diagnosis and under-treatment of mental illness such as depression are very common in all countries, the diagnosis and treatment rate is particularly low in China, which in turn highlights the importance and the value-added of using population-based data to quantify the impact of increasing prevalence of mental illness on healthcare costs. Second, China has experienced both rapid economic growth and fast epidemiological transition (from communicable to noncommunicable diseases) in recent years; thus, how China copes with the rising challenges of increasing mental illness has important implications for many countries with the same development experience. Although our paper does not directly address the future path of reform in the mental health sector, our quantitative analysis sheds new light for the reform directions for both China and many low- and middle-income countries that face similar healthcare and economic challenges.
3.3 Data Source and Descriptive Analysis China Family Panel Studies (CFPS) is a nationally representative longitudinal survey designed and implemented by the Institute of Social Science Surveys (ISSS) of Peking University. It was conducted in 25 Chinese provinces (these provinces jointly cover 95% of the Chinese population) in five years (2008, 2009, 2010, 2011, 2012). In each wave, the CFPS survey samples about 15,000 households nationwide using the multistage probability proportional to size (PPS) sampling method and interviews all member of the family in each sample household. The questionnaire gathers individual-, family, and community-level information on the demographic and socioeconomic variables, as well as information on the respondents’ health
50
3 Depression Hurts, Depression Costs: The Medical Spending …
outcomes. In the 2012 CFPS survey, a full 20-question version of the Center for Epidemiologic Studies Depression (CES-D) questionnaire (Radloff 1977) is included to assess the respondents’ mental health status. The CES-D questionnaire is one of the most frequently used self-assessment tools for depression and depressive symptoms. An advantage of using this survey-based instrument is that the questions contained in CES-D are non-intrusive and related to everyday feelings,3 which makes it easier for the respondents to answer, leading to better detection of their depressive symptoms compared to other clinical instruments. This in turn may help to alleviate the underreporting problem commonly experienced among the mental illness patients (Bharadwaj et al. 2015). The CES-D questionnaire contains four subscales: somatic-retarded activity, interpersonal relations, depressed affect, and positive affect. The former three measure negative emotions, while the latter measures positive ones. Respondents are asked to rate how often they experienced the specified emotions in the past week, with the options varying from 0 to 3 for each question (0 = rarely, 1 = little, 2 = occasionally, 3 = often). The CES-D score can thus be calculated based on the responses as follows: CES − D =
i
+
Scorei,somatic +
k
j
Scorek,depressed +
Score j,interpersonal
(4 − Scorel,positive ),
(3.1)
l
where Scorei,somatic , Score j,interpersonal , Scorek,depressed , and Scorel,positive represent the score for the i-th question on the somatic-retarded activity, the j-th question on interpersonal relations, the k-th question on the depressed affect, and the l-th question on the positive affect, respectively. Thus, the overall CES-D score ranges from 0 to 60, with a higher score indicating more frequent occurrence of depressive symptoms and higher likelihood of depression. According to Radloff (1977, 1991), the values of 16 and 28 approximately correspond to the 80th and 95th percentile of the CES-D distribution in the US-based Community Mental Health Assessment (CMHA) survey; thus, these two thresholds were commonly used to define the mental health conditions of depression and depressive symptoms among the US population. Following this approach and considering the country difference in the mental health conditions between US and China, we set our threshold values based on the 80th and 95th percentile of the CES-D distribution in the national sample of CFPS 2012, which correspond to the CES-D scores of 20 and 28, i.e. a CES-D of 20–27 indicates that the person suffers from depressive symptoms, and a score of 28 or higher indicates depression.4 Based on the above CES-D classification, the key explanatory variable in our study is the 3
Examples of the CES-D questions include: “How often do you feel that everything I did was an effort?”; “How often do you feel not like eating (your appetite is poor)?”. 4 The “depression” identified by CES-D is different from the clinical definition, which is based on psychiatric diagnostic criteria, e.g. the DSM-IV or the ICD-10. Clinical depression, including major depression disorder (MDD) and major depression episodes (MDE), is usually diagnosed with
3.3 Data Source and Descriptive Analysis
51
respondent’s mental health status (Mhs), which takes on three possible values: 0 indicates being mentally healthy, 1 indicates depressive symptoms, and 2 indicates depression. The central outcome variable is the respondent’s total annual medical expenditure as reported in the CFPS survey, which includes the inpatient and outpatient medical costs on the treatment of injuries and diseases (including mental health diseases) paid by the individuals and the insurance providers. Based on the 2012 CFPS dataset, we drop the observations younger than 16 and older than 99, as well as the ones with missing information on key variables such as gender, age and the CES-D scores. In addition, we drop the observations with zero personal income (defined as the per-capita annual household net income) for the reason that their medical costs are likely to be paid and decided by their family members.5 The final study sample consists of 30,568 observations, among whom 81.2% are mentally healthy, 13.5% experience depressive symptoms, and 5.3% suffer from depression.6 Table 3.1 provides the sample summary statistics of key variables, with column (1)–(4) representing the full sample and the subsamples in the three mental health categories (mentally healthy, depressive symptoms, depression), respectively. The table shows that the annual medical expenditure is strongly related to the status of individual mental health. On average, a mentally healthy person spends 1561 Yuan on medical services per year, while for those who suffer from depressive symptoms, the costs increase to 2628 Yuan. Furthermore, the mean medical expenses of the group with depression are nearly 4200 Yuan, almost three times higher than that of the mentally healthy group. There are three plausible explanations for this outcome. First, the treatment and control of depression and depressive symptoms, such as the use of antidepressants and cognitive-behavioral therapy, typically incur substantial medical costs (March et al. 2004). Second, depression not only causes a reduction in psychological well-being but also damages people’s physical health and hence create a co-morbidity pattern in which depression often co-exists with other NCDs that require costly medical treatments (Penninx et al. 1999). Third, the existence of the social and institutional barriers in access to mental health services often causes a delay in treatment or a cost shift to non-mental healthcare. Putting together, people suffering from depression and depressive symptoms are also likely to struggle under the financial burden of medical care, and the following sections intend to numerically estimate the medical cost induced by these conditions through econometric models. In our econometric analysis, the control variables include the respondent’s demographic and socioeconomic characteristics, such as gender ( female), age in years 12-month occurrence of depressive events, while the “depression” in this paper is diagnosed with one-week occurrence of depressive events. 5 From the original sample of 35,720 observations, 3 observations are dropped due to age restriction, 4484 are dropped due to their missing information on key variables, and 665 are dropped due to zero income. 6 Due to the discrete nature of the CES-D scores, the distributional percentiles corresponding to the CES-D cut-off values (20 and 28) differ slightly from the proposed percentiles (80th and 95th). In the Robustness Check section, we also use the alternative CES-D thresholds of 16 and 28 as originally proposed by Radloff (1977, 1991) for sensitivity check purpose.
52
3 Depression Hurts, Depression Costs: The Medical Spending …
Table 3.1 Sample summary statistics for key variables Variable CESD
Definition CES-D score
Full sample
Mhs = 0
Mhs = 1
Mhs = 2
(1)
(2)
(3)
(4)
12.92
9.97
22.68*
33.36*
(7.96)
(4.88)
(2.17)
(5.31)
1560.89
2628.27*
4199.53*
Expenditure
Annual medical expenditure
1843.75 (6984)
(6258)
(8450)
(1144)
Female
Gender (1 = female)
0.51
0.49
0.61*
0.65*
(0.50)
(0.50)
(0.49)
(0.48)
45.22
44.35
47.53*
52.76*
(16.60)
(16.56)
(16.3)
(15.3)
0.80
0.81
0.79
0.76
(0.40)
(0.40)
(0.41)
(0.43)
Age
Age in years
Marriage status Married
Married (1 = yes)
Single
Never married (1 = yes)
0.13
0.14
0.11
0.07*
(0.34)
(0.35)
(0.31)
(0.26)
Divorced/widowed
Divorced or widowed (1 = yes)
0.07
0.06
0.10*
0.17*
(0.25)
(0.23)
(0.3)
(0.38)
Number of family members
4.34
4.34
4.39
4.28
(1.87)
(1.86)
(1.93)
(1.91)
Urban residents (1 = yes)
0.31
0.33
0.23*
0.20*
(0.46)
(0.47)
(0.42)
(0.4)
Primary
Primary school or below (1 = yes)
0.50
0.46
0.63*
0.76*
(0.50)
(0.5)
(0.48)
(0.43)
Middle
Middle school (1 = yes)
0.28
0.30
0.23*
0.15*
(0.45)
(0.46)
(0.42)
(0.36)
High school (1 = yes)
0.14
0.15
0.10*
0.07
(0.35)
(0.36)
(0.3)
(0.25)
Familysize Urban Education
High College
College or above (1 = yes)
0.08
0.09
0.04*
0.02*
(0.27)
(0.28)
(0.21)
(0.13)
Income
Per-capita annual household net income (yuan)
13,323
14,009
10,778*
9260*
(22,803)
(24,031)
(17,776)
(10,775)
0.25
0.27*
0.33*
(0.43)
(0.44)
(0.47)
Employment status Nowork
Not working (1 = yes) 0.25 (0.44)
Farm
Household farming (1 = yes)
0.24
0.22
0.28*
0.34*
(0.42)
(0.42)
(0.45)
(0.48) (continued)
3.3 Data Source and Descriptive Analysis
53
Table 3.1 (continued) Variable Government
Definition
Full sample
Mhs = 0
Mhs = 1
Mhs = 2
(1)
(2)
(3)
(4)
Government employee 0.08 (1 = yes) (0.26)
0.08
0.05*
0.03*
(0.28)
(0.21)
(0.18)
Collective
Collective firm employee (1 = yes)
0.02
0.02
0.01*
0.01*
(0.13)
(0.13)
(0.11)
(0.11)
Private
Private firm employee (1 = yes)
0.28
0.30
0.25*
0.15*
(0.45)
(0.46)
(0.43)
(0.36)
Other employment type (1 = yes)
0.13
0.13
0.14*
0.13
(0.34)
(0.34)
(0.35)
(0.34)
Sample size
30,568
24,833
4128
1607
Other Observation
Note Data Resource: China Family Panel Studies (2012). The reported statistics are the sample mean with standard deviation in parentheses. Columns (2)–(4) correspond to the mentally healthy group (Mhs = 0), depressive symptom group (Mhs = 1) and depression group (Mhs = 2), which are categorized using the CES-D score (depressive symptoms = CES-D between 20 and 27; depression = CES-D of 28 or higher). Asterisks (*) in column (3) and (4) denote statistically significant differences between the depressive symptom group/depression group and the mentally healthy group (at 5% confidence level)
(age), education attainment (“primary school or below”, “middle school”, “high school” and “college or above”), residential status (urban), employment types (“not working”, “household farming”, “government employee”, “employed by collective enterprise”, “employed by private enterprise”, “other types of employment”), marital status (married), household size ( familysize) and annual personal income (income). According to column (1) of Table 3.1, 51% of the respondents in our sample are female, and 31% live in the urban areas. 80% of the sample are married (including common-law marriage), with a sample average familysize of 4.3 people. The average age of the full sample is 45.2, and the average annual personal income is 13,323 Yuan. For education attainment, 50% of the full sample do not have formal schooling or only complete primary school education; 28 and 14% of the sample have middle school and high school education, respectively; only 8% of the sample have tertiary education. In terms of employment status, 25% of the respondents are not currently working (including retirees and students), while the rest of the sample is categorized into household farmers (24%), government employees (8%), collective enterprise employees (2%), private enterprise employees (28%), and other types of employment (13%). For healthcare utilization, 77.8% of the sample have incurred medical spending in the previous year, and 8.9% have inpatient spending. Among the medical users, the average annual medical cost is 2371.35 Yuan, and the average inpatient spending is 11,643.19 Yuan among the inpatient users. A comparison among the three mental health subsamples indicates considerable group differences in several key variables. For example, women are more likely
54
3 Depression Hurts, Depression Costs: The Medical Spending …
to be depressive than men: 61 and 65% of the observations experiencing depressive symptoms and severe depression are female, which are higher than the female percentage in the full sample. According to the previous research (Nolen-Hoeksema 2001; Simon 2002; Tsang et al. 2008), both biological (or hormone), factors and social factors (e.g. sexual discrimination) may account for women’s higher vulnerability to depression. The average ages of the three mental health groups are 44, 48, and 53, respectively, which demonstrates that the elderly are at higher risks of severe depression than the younger people. The conclusion is consistent with other studies that find higher depression prevalence rates among older people in both developed and developing countries (Van Itallie 2005; Tsang et al. 2008). One plausible reason is that the socioeconomic status of the elderly is relatively low due to their poor health status, lack of social support, and the absence of financial support from adult children [see detailed discussions on the “empty nest” syndrome among China’s elderly by Liu and Guo (2007) as well as Xie et al. (2010)]. Another conclusion drawn from the subsample comparison is that depression is more prevalent in rural than urban areas. The urban population accounts for 33% of the respondents who are mentally healthy, but the percentage of urban residents decreases to 23 and 20% in the depressive symptom and the severe depression groups, which is also consistent with the literature (Ma et al. 2009; Philips et al. 2009). Compared to their rural peers, urban residents in China have easier access to education and healthcare, and hence, they generally enjoy higher income and better quality of life, which may in turn contribute to their better mental health status. Socioeconomic status is highly correlated with people’s tendency of depression, suggesting the strong socioeconomic gradient in mental health. For example, the severe depression group has the highest representation of the lowest education attainment (76%), much higher than the depressive symptom group (63%), and mentally healthy group (46%). On the other hand, the proportions of people completing middle school, high school, and college (or above) are 30, 15, and 9% in the mentally healthy group, while the proportions are 23, 10, and 4% in the depressive symptom group, and they are 15, 7, and 2% in the severe depression groups, respectively. Mirowsky and Ross (1998, 2003) show that education plays a critical role in helping people accumulate human capital and develop personal control over life events, which may explain why better-educated people are less vulnerable to depression. There is considerable income inequality among the three groups: the mentally healthy respondents earn an average of 14,009 Yuan per year, while the mean annual income of the depressive symptom group and depression group are only 10,778 Yuan and 9260 Yuan, respectively. The results are consistent with the theoretical and empirical studies on the promoting effect of income on happiness (Zimmerman and Katon 2005), which suggest that individuals in low-income status may not be able to afford quality consumptions and comprehensive health services, leading to potential anxiety and disappointment that cause depression. Additionally, low-income people are also more likely to be exposed to violence and unstable living environment, which are risk factors for depression (Fitzpatrick 1993).
3.4 Estimation Method
55
3.4 Estimation Method 3.4.1 The Two-Part Model We use a baseline two-part model (2PM) to characterize the determinants of an individual’s annual medical expenditure, which can be expressed as follows: Pr(yi > 0|X i ) = G(θ1 D1i + θ2 D2i + β X i + u i ),
(3.2)
yi = exp(δ1 D1i + δ2 D2i + γ Z i ) + ei , for yi > 0,
(3.3)
where the outcome variable yi is the total medical expenditure incurred in the previous year by individual i. The key explanatory variables are the two dummies—D1i and D2i with the former indicating whether the person experiences depressive symptoms (Mhs = 1) and the latter indicating whether the person suffers from depression (Mhs = 2). The parameters θ1 , θ2 , ϕ, and δ2 are the coefficients of interest, which represent the effects of depressive symptoms and depression on the probability of medical usage and the conditional medical expenditure among medical users, respectively. X i is a vector of individual characteristics including gender, age, education, residential type, employment status, marriage status, household size, personal income, and province dummies. Consistent with the literature convention, vector Z i contains the same set of variables as X i . The above 2PM assumes that the individual medical spending is determined by two separate decision-making processes: Eq. (3.2) is the “participation equation” and captures the systematic difference between medical users and non-users; Eq. (3.3) is the “intensity equation” and characterizes the determination mechanism of the amount of medical cost among medical users. Following the suggestion of prior studies (Jones 2000; Manning and Mullahy 2001), we estimate Eq. (3.2) with the Logit model (specifying G(·) as the cumulative distribution function of the logistic distribution) and estimate Eq. (3.3) with the Gamma GLM model (generalized linear model with a Gamma distribution for ei ). The model specification is justified by the modified Park test,7 which shows that the conditional variance function of the medical expenditure distribution is consistent with the Gamma-class model. In addition, the result of the Hosmer–Lemeshow test8 also confirms that our choice of log link function is consistent with the data generating process. 7
The modified Park test is used to identify the potential distribution of the dependent variable in GLM. The coefficient is 1.755, suggesting that variance is proportional to square of mean, which means that the assumption of Gamma distribution as the right variance function is broadly appropriate for the data. 8 The Hosmer–Lemeshow test intends to verify the link function in GLM by regressing the prediction errors on the deciles of the predicted expenditure. Under the log-link assumption, the p-value is 0.702, which cannot reject the null hypothesis that the decile coefficients are jointly non-significant and suggests that the regression model is fit.
56
3 Depression Hurts, Depression Costs: The Medical Spending …
3.4.2 The Four-Part Model As an extension of the baseline 2PM, we follow Finkelstein et al. (2003) and use the four-part model (4PM) to characterize the medical spending of inpatient and outpatient users separately: Pr(yi > 0|X i ) = G(θ1 D1i + θ2 D2i + β X i + u i ),
(3.4)
Pr(gi > 0|yi > 0, X i ) = G(ω1 D1i + ω2 D2i + σ Z i + ri ),
(3.5)
yi = exp(δ1 D1i + δ2 D2i + γ Z i ) + ei , for gi ≤ 0 and yi > 0,
(3.6)
yi = exp(ϕ1 D1i + ϕ2 D2i + μZ i ) + vi , for gi > 0, yi > 0,
(3.7)
where gi is the inpatient expenditure incurred in the previous year by individual i. Compared to the baseline 2PM, 4PM adds another “participation” equation (Eq. 3.5) and an “intensity” equation (Eq. 3.7) to the model: Eq. (3.5) is a Logit regression that denotes whether a person with positive medical expenditure incurs any inpatient spending; Eq. (3.7) is a Gamma GLM regression on the determinants of medical expenditure based on the sample with positive inpatient spending. Prior literature suggests that the determination mechanism of medical expenditure can be different between the medical users with and without inpatient utilization. First, the type of treatment differs with the severity of depression. Druss and Rosenheck (1999) find that the treatment for depressive symptoms includes both inpatient and outpatient services, but for people facing severe symptoms of depression, the inpatient treatment or hospitalization become one of the best solutions to the illness. Second, compared to the outpatient treatments, the inpatient treatments are more likely to be prescribed for chronic pains and physical diseases, which are strongly correlated with depression (Moussavi et al. 2007). To conclude, depression would cause an increase in both the inpatient and outpatient costs, especially the inpatient costs. Thus, the above 4PM assumes a three-stage determination process of an individual’s medical spending: Eq. (3.4) estimates the systematic difference between medical users and non-users; Eq. (3.5) captures the difference between medical users with and without inpatient utilization; Eqs. (3.6) and (3.7) then characterize the total medical spending among users of outpatient services only and those of inpatient services, respectively. Accordingly, the parameters θ1 and θ2 indicate the impacts of depressive symptoms and depression on the probability of incurring positive medical spending, and the parameters ω1 and ω2 reflect the impacts on the probability of incurring inpatient spending among the medical users. The influences of depressive symptoms and depression on the amount of medical spending are represented by parameters δ1 and δ2 (on the outpatient users) and ϕ1 and ϕ2 (on the inpatient users).
3.4 Estimation Method
57
3.4.3 Estimating the Cost of Depression and Depressive Symptoms Following the method used by Finkelstein et al. (2003, 2009), Wang et al. (2011), and Cawley and Meyerhoefer (2012) on estimating the medical cost induced by overweight and obesity, we can estimate the expected medical spending induced by depressive symptoms and depression through the following three-step approach based on the coefficients in 2PM and 4PM [similar methods are discussed in Buntin and Zaslavsky (2004), Deb et al. (2006) and Trogdon et al. (2008)]. First, we calculate the predicted medical cost of each sample individual using the fitted values in the “participation” equation(s) and “intensity” equation(s). Using 2PM as an example, the predicted individual medical spending can be specified as:
E(yi |D1i , D2i , X i , Z i ) = Pr(yi > 0|D1i , D2i , X i ) × E(yi |yi > 0, D1i , D2i , Z i ), (3.8) where the first term on the right-hand side is the predicted probability of having positive medical expenditure based on Eq. (3.2), and the second term is the expected medical costs of medical users according to Eq. (3.3). Since Eq. (3.3) is specified as Gamma GLM, the link function directly characterizes how the expectation of yi is related to the regressors, avoiding the complication in a log-linked OLS model where a log dependent variable needs to be consistently retransformed back to its original scale (Buntin and Zaslavsky 2004). The sample average of E(yi ) thus becomes the expected medical spending of the population. In case of 4PM, the predicted individual medical spending can be written as:
E(yi |Di , X i , Z i ) = Pr(yi > 0|Di , X i ) × Pr(gi > 0|yi > 0, Di , X i ) × E(yi |gi > 0, yi > 0, Di , Z i ) + 1−Pr(gi > 0|yi > 0, Di , X i ) × E(yi |gi ≤ 0, yi > 0, Di , Z i )] .
(3.9) In the second step, we calculate the counter-factual medical spending of an individual by setting his or her mental health indicators (D1i and D2i ) to 0, while holding the other control variables at the original values. This counter-factual prediction can be specified as follows in the 2PM setting:
E(yi |D1i = 0, X i , Z i ) = Pr(yi > 0|D1i = 0, X i ) × E(yi |yi > 0, D1i = 0, Z i ), (3.10) E(yi |D2i = 0, X i , Z i ) = Pr(yi > 0|D2i = 0, X i ) × E(yi |yi > 0, D2i = 0, Z i ). (3.11)
58
3 Depression Hurts, Depression Costs: The Medical Spending …
Thus, for individuals with depressive symptoms or depression, the above counterfactual spending is their expected medical cost if they were to become mentally healthy. The sample average of such counter-factual individual spending is thus the expected medical cost of a mentally healthy population with the same baseline characteristics except for the mental health status. Similarly, the counter-factual medical spending in a 4PM can be written as:
E(yi |D1i = 0, X i , Z i ) = Pr(yi > 0|D1i = 0, X i ) × Pr(gi > 0|yi > 0, D1i = 0, X i )
× E(yi |gi > 0, yi > 0, D1i = 0, Z i ) + 1 − Pr(gi > 0|yi > 0, D1i = 0, X i ) ×E(yi |gi ≤ 0, yi > 0, D1i = 0, Z i )] ,
(3.12)
E(yi |D2i = 0, X i , Z i ) = Pr(yi > 0|D2i = 0, X i ) × Pr(gi > 0|yi > 0, D2i = 0, X i )
× E(yi |gi > 0, yi > 0, D2i = 0, Z i ) + 1 − Pr(gi > 0|yi > 0, D2i = 0, X i ) ×E(yi |gi ≤ 0, yi > 0, D2i = 0, Z i )] .
(3.13)
In the third step, we calculate the expected personal medical costs attributable to depressive symptoms and depression, represented by E 1 and E 2 respectively, by taking the differences between the three expected medical costs (one from step 1 and the other two from step 2). E 1 = E(yi |D1i , D2i , X i , Z i ) − E(yi |D1i = 0, X i , Z i ),
(3.14)
E 2 = E(yi |D1i , D2i , X i , Z i ) − E(yi |D2i = 0, X i , Z i ).
(3.15)
These cost estimates can be expressed in monetary values or as a percentage of the total expected medical expenditure, and their statistical significance can also be obtained using the t test on the difference between the two expected medical costs.
3.5 Empirical Results
59
3.5 Empirical Results 3.5.1 Regression Results Table 3.2 reports the main results for the baseline 2PM, which contain information on the variable marginal effects for both the “participation” equation and “intensity” equation, with standard errors clustered at the county level.9 The baseline model shows that the mental health status has a statistically significant impact on both the probability of using healthcare services and the amount of medical spending among the users of healthcare services. Specifically, the results indicate that individuals with depressive symptoms are 8.8% more likely to have nonzero medical expenditure and will spend 1029.78 Yuan more on healthcare services in a year. For individuals with depression, the impacts of mental health status on healthcare costs are even stronger: they are 11% more likely to use healthcare services and will spend 1836.52 Yuan more on healthcare. These results reinforce the findings obtained from previous studies that the cost impacts of mental illness such as depression are high. The coefficients of other control variables are generally consistent with the existing studies on the demand for healthcare. To be specific, we find that females are more likely to use healthcare services but spend less money on healthcare conditional on utilization. In addition, both the probability and the amount of healthcare spending increase with age. Income also has a significantly positive impact on the use of healthcare and the healthcare costs, indicating that healthcare services are normal goods in the sense that the demand for healthcare increases with income. Although marital status does not significantly influence the probability of healthcare usage, evidence shows that the single or divorced/widowed individuals tend to spend less on healthcare compared to their married counterparts. We also find that working status has a significant impact on the demand for healthcare: compared to individuals who are not currently working, individuals who are active in the labor market are less likely to use healthcare services and spend less if they use healthcare. There are two possible explanations for this result. First, the labor market participants may face a higher time price in seeking healthcare as compared to non-participants. Since the full price of using healthcare service consists of both monetary cost and time cost, individuals with higher time price will have less demand for healthcare, holding other things constant (Sloan and Hsieh 2017, p. 93). Second, the participation of labor market may serve as a proxy for being in good health, and hence, labor market participants may use less healthcare services as compared to non-participants. However, we find that education, residential status, and familysize do not have significant impacts on either the probability of using healthcare services or the amount of spending among healthcare users. The insignificant result may be contributed 9
CFPS follows a multistage stratified sampling method, and the primary sampling unit (PSU) is either an administrative district (in urban areas) or a county (in rural areas). Thus, following the literature convention, standard errors are clustered at the county level in all regressions [see Xie and Hu (2014) for more details].
60 Table 3.2 Regression results for the two-part model
3 Depression Hurts, Depression Costs: The Medical Spending …
Variable
Participation
Intensity
(1)
(2)
Mhs = 1
0.088***
1029.778***
(0.0101)
(124.2407)
Mhs = 2
0.110***
1836.518***
(0.0171)
(195.9880)
0.053***
− 263.787***
Female
(0.0051)
(100.6466)
Age
0.004***
35.998***
(0.0003)
(4.1991)
Single
0.004
− 1057.578***
(0.0091)
(140.0699)
Divorced/widowed
− 0.011
− 731.291***
(0.0109)
(115.2224)
− 0.004*
− 40.335
Familysize
(0.0022)
(28.0694)
Urban
− 0.011
173.464
(0.0137)
(146.8622)
Middle
0.001
− 46.192
(0.0087)
(121.3465)
− 0.006
188.364
High
(0.0107)
(181.1377)
College
0.017
− 302.750
(0.0114)
(192.4721)
Log (income)
0.012***
152.486***
(0.0034)
(46.6227)
0.003
− 1469.186***
Farm
(0.0118)
(157.8444)
Government
− 0.016
− 1208.706***
(0.0127)
(230.6926)
Collective
− 0.024
− 1020.709*
(0.0204)
(525.6120)
− 0.004
− 1191.097***
Private
(0.0097)
(170.6624)
Other
− 0.042***
− 845.372***
(0.0142)
(188.1412)
Province dummy
Yes
Yes (continued)
3.5 Empirical Results Table 3.2 (continued)
61
Variable Sample size
Participation
Intensity
(1)
(2)
30,568
23,767
Note The reported statistics are the marginal effects of the explanatory variables with the county-level clustered standard errors shown in parentheses. *, **, *** denote statistical significance at 10, 5, 1% levels, respectively
by the two offsetting effects working in opposite directions. For example, on the one hand, high-educated individuals tend to have better awareness of their health problems and hence are more likely to have higher healthcare demand. On the other hand, these individuals are more efficient in the production of their own health, and hence, they will use less healthcare inputs to achieve the same health improvement as compared to low-educated people (Sloan and Hsieh 2017, p. 54). Similarly, rural residents may need to pay a higher time price in seeking healthcare than urban residents due to the lower availability of healthcare resources, and thus, rural residents may use less healthcare services than urban residents at one point in time. However, there is also evidence indicating that rural residents are in disadvantage to manage their health problems, so they tend to have poorer awareness and treatment for NCDs (Lei et al. 2012). The poor health management in turn forces the rural residents to use more healthcare services in the long run to restore their health status, such as the use of avoidable inpatient services. Table 3.3 reports the regression results for the four-part model. The results are generally consistent with the two-part model with the additional information on the impact of mental health status on the probability of using inpatient services and the amount of spending on hospital care. Specifically, we find that individuals with depressive symptoms are 4% more likely to have nonzero inpatient expenditure and will spend 1768.32 Yuan more on hospital care. For individuals with depression, the impact of their mental health status on their inpatient utilization is stronger too: they are 6.9% more likely to use inpatient services and will spend 3773.92 Yuan more on hospital care. The estimated results on other control variables are similar to those reported in Table 3.2 with a few exceptions. For example, compared to the male group, the conditional healthcare expenditure for the female respondents is higher in the outpatient setting but lower in the inpatient setting, and this result suggests that the previous 2PM-based findings on the gender difference in medical expenditure are mainly driven by the inpatient spending in the data.
62
3 Depression Hurts, Depression Costs: The Medical Spending …
Table 3.3 Regression results for the four-part model Variable
Participation1
Participation2
Intensity 1
Intensity 2
(1)
(2)
(3)
(4)
0.088***
0.040***
457.181***
1768.317**
(0.0101)
(0.0051)
(60.5261)
(768.1409)
Mhs = 2
0.110***
0.069***
882.560***
3773.924***
(0.0171)
(0.0085)
(91.7069)
(1215.5623)
Female
0.053***
0.0070
95.085**
− 3953.329***
(0.0051)
(0.0044)
(43.2199)
(805.0371)
0.004***
0.001***
18.218***
22.7350
(0.0003)
(0.0002)
(2.2165)
(25.3398)
Single
0.0040
− 0.064***
− 241.929**
− 3033.633**
(0.0091)
(0.0072)
(105.2402)
(1405.1736)
Divorced/widowed
− 0.011
− 0.018***
− 141.357**
− 3915.030***
(0.0109)
(0.0061)
(60.3838)
(664.2030)
− 0.004*
− 0.001
− 7.644
− 322.126*
(0.0022)
(0.0013)
(12.8620)
(181.7481)
Urban
− 0.011
0.005
89.617
722.885
(0.0137)
(0.0071)
(61.4090)
(1002.6685)
Middle
0.0010
0.0000
(61.7280)
1088.8710
(0.0087)
(0.0062)
(46.5010)
(876.3226)
− 0.006
0.006
96.897
586.524
(0.0107)
(0.0075)
(93.7847)
(1232.6610)
College
0.017
0.009
− 77.149
− 1906.685
(0.0114)
(0.0117)
(84.5203)
(1367.5486)
Log (income)
0.012***
0.0020
60.405***
812.648***
(0.0034)
(0.0020)
(23.4463)
(282.0944)
0.0030
− 0.065***
− 487.127***
− 4197.002***
(0.0118)
(0.0078)
(71.1060)
(874.8755)
Government
− 0.016
− 0.061***
− 424.623***
− 3428.957**
(0.0127)
(0.0099)
(86.3087)
(1729.5091)
Collective
− 0.024
− 0.073***
− 271.94
− 1737.986
(0.0204)
(0.0165)
(197.8315)
(4560.7086)
− 0.004
− 0.069***
− 323.904***
− 3542.696***
(0.0097)
(0.0075)
(86.3577)
(987.8870)
− 0.042***
− 0.051***
− 186.488**
− 2298.614*
Mhs = 1
Age
Familysize
High
Farm
Private Other
(continued)
3.5 Empirical Results
63
Table 3.3 (continued) Variable
Participation1
Participation2
Intensity 1
Intensity 2
(1)
(2)
(3)
(4)
(0.0142)
(0.0088)
(84.8986)
(1284.2939)
Province dummy
Yes
Yes
Yes
Yes
Sample size
30,568
23,767
21,054
2713
Note The reported statistics are the marginal effects of the explanatory variables with the countylevel clustered standard errors shown in parentheses. *, **, *** denote statistical significance at 10, 5, and 1% levels, respectively
3.5.2 Estimating the Medical Cost of Depression and Depressive Symptoms Based on the regression results of the two-part model, Table 3.4 presents the estimated personal medical cost attributed to depressive symptoms and depression for the full sample as well as the subsamples of different regions, genders, age groups, and education levels. Following the three-step method described in Sect. 3.4.3, we report for each sample: (1) the predicted individual medical spending based on Eq. (3.8); (2) the counter-factual medical expenditure when depressive symptoms/depression are set to healthy mental status; (3) the expected medical costs attributed to depressive symptoms and depression, which are expressed in level (Yuan) and percentage terms; (4) the t-statistics and p-values associated with the t-tests on the significance of cost estimates. The upper part of Table 3.4 shows that the annual expected medical cost attributed to depressive symptoms is predicted to be 142.42 Yuan, or 7.8% of the total expected personal medical expenditure in a year. For the depression-induced medical cost, the estimate is 126.38 Yuan per annum, or 6.9% of total personal medical expenditure (lower part of Table 3.4). Putting together, we conclude that about 14.7% of the personal medical expenditure among Chinese adults can be attributed to depressive symptoms and depression.10 In comparison to a recent study on the impact of overweight and obesity on healthcare costs (Qin and Pan 2016), our results indicate that both depressive symptoms and depression are very costly to individuals and the society as a whole. A comparison among subsamples suggests that the personal medical expenditure attributable to depressive symptoms and depression are not evenly distributed across regions and subpopulations, with the female, the rural residents, the poorly educated, and the elderly people bearing a higher percentage of medical costs due to depression. For example, the share of total personal medical expenditure attributed to depression is larger in the rural areas (8.1%) than that in the urban areas (4.9%), and 10
For clarification, the estimated medical costs attributable to depression and depressive symptoms are the population-based expected costs, which factor in both the probability and the conditional costs of incurring these mental health conditions. Thus, the expected cost of depression can be lower than that of the depressive symptoms due to its much lower prevalence rate within the population.
Category
8666
4231
2342
Middle
High
College
Age
12,560
Middle aged
15,604
11,447
Female
Young
14,964
Urban
Male
9364
Rural
Region
Gender
21,204
Baseline
All sample
30,568
15,329
Primary
6561
Middle aged
Elderly
11,447
12,560
Young
15,604
(B) Depression
Education
Age
14,964
Female
Urban
Male
9364
Rural
Gender
21,204
Baseline
Region
30,568
Sample size
All sample
(A) depressive symptoms
Sample
1856.24
1001.06
1885.16
1786.69
2303.23
1631.04
1836.95
1341.1
1807.17
1542.17
2087.58
3258.41
1856.24
1001.06
1885.16
1786.69
2303.23
1631.04
1836.95
Expected expenditure
1722.98
966.48
1727.42
1693
2190.79
1498.5
1710.57
1280.92
1697.67
1437.79
1902
3006.19
1708.6
927.3
1720.67
1667.28
2161.27
1488.42
1694.53
Count-factual expenditure
133.26
34.58
157.73
93.68
112.45
132.53
126.38
60.18
109.5
104.38
185.58
252.22
147.64
73.76
164.49
119.41
141.97
142.62
142.42
Expected cost of depressive symptoms (C)
Table 3.4 Estimated personal medical costs attributable to depressive symptoms and depression based on 2PM
7.18
3.45
8.37
5.24
4.88
8.13
6.88
4.49
6.06
6.77
8.89
7.74
7.95
7.37
8.73
6.68
6.16
8.74
7.75
% cost of depressive symptoms (D)
25.99
15.69
29.00
20.77
15.70
32.91
35.58
12.58
17.81
26.82
47.87
31.18
41.51
34.86
46.74
35.72
27.54
53.43
58.53
t-statistics
(continued)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
p-value
64 3 Depression Hurts, Depression Costs: The Medical Spending …
Category
8666
4231
2342
Middle
High
College
6561
15,329
Primary
Elderly
Sample size
1341.1
1807.17
1542.17
2087.58
3258.41
Expected expenditure
1314.51
1740.75
1480.17
1893.01
2985.03
Count-factual expenditure
26.58
66.42
62.01
194.57
273.38
Expected cost of depressive symptoms (C)
1.98
3.68
4.02
9.32
8.39
% cost of depressive symptoms (D)
4.66
9.14
13.27
31.65
21.80
t-statistics
0
0
0
0
0
p-value
Note (1) All subsample results are based on the two-part model. (2) Counter-factual expenditure in Panel (A) is the expected medical expenditure when the people suffering from depressive symptoms are set to be mentally healthy, holding other personal characteristics at the actual levels. Similarly, the counter-factual expenditure in Panel (B) is the expected medical expenditure when people with depression are set to be mentally healthy. (3) The t-statistics and p-values are associated with the t test on the significance of the cost estimates (the difference between the predicted expenditure and counter-factual expenditure). (4) Age groups are based on the following age categorization: Young = between 16 and 40; Middle-aged = between 40 and 60 (include 40); Elderly = aged above 60 (include 60)
Education
(A) depressive symptoms
Sample
Table 3.4 (continued)
3.5 Empirical Results 65
66
3 Depression Hurts, Depression Costs: The Medical Spending …
the cost impact of depressive symptoms has a similar pattern (8.7% vs. 6.2%). This result indicates that although mental health problems are major contributors to the increasing healthcare costs in both urban and rural China, the rural residents shoulder a bigger burden as they pay a higher share for severe depression, and they also face larger barriers than their urban counterparts in seeking mental healthcare due to the low availability of mental health resources in the rural areas. Similarly, we also find that the depression-induced medical expenditure is higher for females and less educated people. For example, the expected medical cost attributable to depression (depressive symptoms) is 8.4% (8.7%) of total spending for female, which is much higher than male 5.2% (6.7%); likewise, the estimated cost of depression (depressive symptoms) is 9.3% (8.9%) among people with “primary school or below” education levels, while it is in the range of 2.0–4.0% (4.5–6.8%) for people with higher levels of educational attainment. The results suggest that these disadvantaged groups may have more difficulty in overcoming the accessibility hurdles of mental healthcare. For example, female and the less educated people may be subject to more social stigma when depressed; they may also be constrained by larger information gaps in seeking effective medical treatment for mental illness. A comparison among age groups indicates that the elderly people (over age 60) suffer the most from the depression-induced spending (about 8.4% of total medical spending), while the middle-aged individuals (aged between 40 and 60) are most affected by the medical cost related to depressive symptoms (accounting for 8.0% of total spending). A plausible explanation is that the elderly people in China (especially those living in the rural areas) tend to live away from their children and likely to suffer from the “empty nest” syndrome with insufficient social and emotional support, while the middle-aged group is more prone to work- and family-related stress, resulting in higher prevalence of mental health problems among these subpopulations. Table 3.5 presents the estimated personal medical cost attributed to depressive symptoms and depression based on the regression results of the four-part model. The results and subsample patterns are similar to those reported in Table 3.4, indicating that our medical cost estimates and previous conclusions are not sensitive to alternative model specifications.
3.5.3 Robustness Checks In this section, we test the robustness of our main results by setting different cut-off scores of CES-D and using alternative model specifications for estimation. First, our main analysis uses the CFPS-based thresholds (20 and 28), which correspond to the 80th and 95th percentiles of the CES-D distribution in our study sample. In this section, we use the original CES-D classification thresholds (16 and 28) from Radloff (1977, 1991) in defining the mental health categories. In other words, a CESD of 16–27 indicates depressive symptoms, while a score of 28 or higher indicates depression. The results associated with the new CES-D classification standard are reported in the first two columns of Table 3.6. These results are similar to those
Category
4231
2342
High
College
14,964
15,604
Male
Female
9364
Urban
Region
Gender
21,204
Baseline
Rural
All
30,568
8666
Middle
6561
15,329
Elderly
Primary
12,560
Middle
15,604
11,447
Female
Young
(B) Depression
Education
Age
14,964
Urban
Male
9364
Rural
Region
Gender
21,204
Baseline
30,568
Sample size
All
(A) depressive symptoms
Sample
1882.34
1785.26
2343.88
1610.01
1834.82
1398.61
1816.86
1607.57
2034.89
3186.71
1861.49
1030.7
1882.34
1785.26
2343.88
1610.01
1834.82
Expected expenditure
1729.19
1688.75
2231.08
1479.01
1709.39
1336.35
1709.39
1499.56
1861.55
2956.38
1716.77
955.15
1725.96
1668.08
2204.27
1473.89
1697.63
Count-factual expenditure
153.15
96.51
112.81
131
125.42
62.26
107.47
108.01
173.34
230.32
144.72
75.55
156.37
117.18
139.62
136.12
137.19
Expected cost of depressive symptoms
8.14
5.41
4.81
8.14
6.84
4.45
5.92
6.72
8.52
7.23
7.77
7.33
8.31
6.56
5.96
8.45
7.48
% cost of depressive symptoms
Table 3.5 Estimated personal medical costs attributable to depressive symptoms and depression based on 4PM
29.77
20.94
16.02
33.51
36.19
12.60
18.11
26.92
48.40
31.06
41.27
34.64
47.54
35.90
27.95
54.12
59.12
t-statistics
(continued)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
p-value
3.5 Empirical Results 67
4231
2342
High
College
1398.61
1816.86
1607.57
2034.89
3186.71
1861.49
1030.7
Expected expenditure
1370.33
1747.72
1540.44
1846.13
2930.39
1724.34
993.16
Count-factual expenditure
28.28
69.13
67.12
188.76
256.31
137.15
37.54
Expected cost of depressive symptoms
2.02
3.80
4.18
9.28
8.04
7.37
3.64
% cost of depressive symptoms
4.80
9.16
13.52
32.28
21.84
11.23
15.67
t-statistics
0
0
0
0
0
0
0
p-value
Note (1) All subsample results are based on the four-part model. (2) Counter-factual expenditure in Panel (A) is the expected medical expenditure when the people suffering from depressive symptoms are set to be mentally healthy, holding other personal characteristics at the actual levels. Similarly, the counter-factual expenditure in Panel (B) is the expected medical expenditure when people with depression are set to be mentally healthy. (3) The t-statistics and p-values are associated with the t test on the significance of the cost estimates (the difference between the predicted expenditure and counter-factual expenditure). (4) Age groups are based on the following age categorization: Young = between 16 and 40; Middle-aged = between 40 and 60 (include 40); Elderly = aged above 60 (include 60)
8666
Middle
6561
15,329
Elderly
Primary
12,560
Middle
Education
11,447
Young
Age
Sample size
Category
Sample
Table 3.5 (continued)
68 3 Depression Hurts, Depression Costs: The Medical Spending …
High
Middle
Urban
Familysize
Divorced/widowed
Single
Age
Female
Mhs = 2
Mhs = 1
Variable
0.053***
− 235.332** (97.2799)
0.051***
(0.0051)
(0.0062) 0.001
− 32.012 (27.7526) 178.624 (141.1751) − 2.894 (120.6344) 164.938 (173.0485)
− 0.004*
(0.0022)
− 0.010
(0.0137)
0.002
(0.0087)
− 0.004
(0.0106)
− 0.013
(0.0077)
− 0.006
(0.0059)
− 0.012*
(0.0014)
− 0.004***
(0.0109)
− 730.200*** (117.2782)
− 0.013
(0.0108)
(0.0084)
0.003
− 1033.960*** (144.7266)
0.003
(0.0092)
0.004*** (0.0002)
36.626*** (4.3740)
0.004***
(0.0003)
(0.0048)
(0.0125)
2600.123*** (332.9048)
0.109***
0.105***
(0.0075)
0.085***
(3)
Participation
Heckman model
(0.0133)
1087.669*** (123.3422)
0.077***
(2)
(1)
(0.0091)
Intensity
Participation
Alternative threshold
Table 3.6 Regression results for robustness check
(170.7450)
228.124
(129.7684)
193.525
(135.3414)
262.419*
(29.3085)
− 55.898*
(205.0186)
− 1027.534***
(197.7162)
− 988.934***
(4.5534)
31.880***
(106.3357)
− 386.944***
(218.3777)
2,626.256***
(145.8760)
1,144.711***
(4)
Intensity
(0.0080)
0.003
(0.0062)
0.008
(0.0062)
− 0.007
(0.0014)
− 0.003**
(0.0123)
− 0.026**
(0.0085)
− 0.002
(0.0002)
0.003***
(0.0055)
0.042***
(0.0928)
0.294***
(0.0735)
0.185**
(5)
Participation
IV regression Intensity
(177.7)
184.4
(136.3)
147.5
(137.8)
234.6*
(29.67)
(continued)
− 61.31**
(217.4)
− 893.9***
(200.9)
− 970.4***
(4.617)
34.62***
(119.0)
− 350.9***
(1577)
− 112.5
(1,465)
2003
(6)
3.5 Empirical Results 69
0.017*
− 253.250 (196.7381)
0.019*
(0.0112)
23,767
− 0.015
30,568
Yes
(0.0085)
− 0.041***
(0.0071)
− 0.003
(0.0182)
− 0.024
(0.0106)
30,568
Yes
(181.5054)
− 1428.392***
(157.9692)
− 1864.707***
(411.1295)
− 1892.504***
(236.0008)
− 1883.977***
(156.4790)
− 2028.600***
(46.4503)
174.378***
(234.3299)
− 121.796
(4)
Intensity
30,452
Yes
(0.0085)
− 0.045***
(0.0073)
− 0.005
(0.0186)
− 0.030
(0.0107)
− 0.017
(0.0076)
0.001
(0.0022)
0.015***
(0.0100)
0.026**
(5)
Participation
IV regression Intensity
23,678
Yes
(184.5)
− 1450***
(163.2)
− 1905***
(415.9)
− 1888***
(238.0)
− 1897***
(157.9)
− 2017***
(49.04)
173.9***
(241.2)
− 164.7
(6)
Note (1) The reported statistics are the marginal effects of the explanatory variables with the robust standard errors shown in parentheses. *, **, *** denote statistical significance at 10, 5, 1% levels, respectively. (2) Column (1) and (2) report the 2PM regression results based on alternative CES-D standard (depressive symptoms = CES-D between 16 and 27; depression = CES-D of 28 or higher). (3) Column (3) and (4) report the regression results of the Heckman selection model. The p-value for the LR test on H0: rho = 0 is 0.607, which implies that the correlation between the random errors of the participation and intensity equations is weak. (4) Column (5) and (6) report the regression results based on the IV-treated two-part model (IV-2PM), in which the instrumental variables are the community-level prevalence rates of depressive symptom and depression as well as their interaction term, where a community refers to an urban neighborhood or a rural village in which the respondent lives in. The p-value for the Hansen’s J test on the exclusion restrictions of the IV model is 0.2391
30,568
Sample size
(183.4866)
(0.0146) Yes
− 817.415***
− 0.043***
Yes
− 1163.862*** (169.5387)
− 0.005
(0.0096)
− 979.105* (541.0354)
− 0.027
(0.0204)
− 1197.322*** (234.6496)
− 0.017
(0.0127)
0.004
(149.1812)
(0.0075)
− 1481.633***
0.001
(0.0118)
0.012*** (0.0021)
159.926*** (44.8232)
0.013***
(0.0035)
(0.0100)
(3)
(2)
(1)
Participation
Heckman model Intensity
Participation
Alternative threshold
Province dummy
Other
Private
Collective
Government
Farm
Log (income)
College
Variable
Table 3.6 (continued)
70 3 Depression Hurts, Depression Costs: The Medical Spending …
3.5 Empirical Results
71
reported in Table 3.2, indicating that our basic results are not sensitive to the cut-off points of the CES-D classification. Second, our baseline hurdle models (including 2PM and 4PM) are based on the assumption that the error terms in the participation and intensity equations are not correlated, and thus, the two equations can be estimated separately. However, the sample selection problem can also be caused by some unobserved common factors, justifying the use of the Heckman selection model. For sensitivity check purpose, we hereby adopt the Heckman model and reestimate the participation and intensity equations through a joint maximum likelihood approach. The results are provided in Column (3)–(4) of Table 3.6. As shown, the estimated impacts of depression and depressive symptoms in both the participation and intensity equations are similar to those given by 2PM, suggesting the main results in Sects. 3.5.1 and 3.5.2 are robust. Furthermore, the estimated covariance between the random errors of the two equations is not significantly different from zero, indicating that the sample selection correction is not needed, which in turn supports the use of our baseline hurdle models. Third, we use the instrumental variable (IV) method to address the potential endogeneity problem of the depression indicators in our main regressions. The mental health status of an individual can be endogenous due to the following two reasons: (1) unobserved factors such as lifestyles can lead to depression and increased medical expenditure simultaneously; (2) depression can be caused by higher medical spending due to financial concerns, thus the two variables are subject to “reverse causality”. Such endogeneity can bias the coefficient estimates, either upward or downward, in the 2PM and 4PM regressions. As a solution, we use the prevalence rates of depression and depressive symptoms in the respondent’s residential community as well as the interaction term of these two variables as IVs to address the potential endogeneity of the two mental health indicators. The percentage prevalence rates are defined as the number of people with depression/depressive symptoms (using our original CES-D cut-offs of 20 and 28) by the total number of respondents (excluding the individual him/herself) within the community, where a community refers to an urban neighborhood or a rural village as defined by the CFPS site identifiers.11 The reasons why we choose these IVs are as follows. First, medical literature provides strong evidence on the geographic clustering of mental health problems such as depression due to the contextual effect12 (Aneshensel and Sucoff 1996; Chaix et al. 2006). This in turn is contributed by structural factors (such as unemployment and income inequality) and experiential factors (such as vandalism, criminality and the lack of social cohesion) in a common socioeconomic environment that may lead to the prevalence of mental and emotional impairment among local residents (Fox 1990; Liem and Liem 1978; Mechanic 1972). This phenomenon can be further explained by the following two epidemiological pathways: psychopathological outcomes may 11
Observations whose residential community contains only one sample individual are dropped; thus, the sample size associated with the IV-2PM regressions is reduced to 30,452. 12 In medical literature, a contextual effect is an aspect of cognitive psychology that describes the influence of environmental factors on one’s perception of a stimulus. Morgan (2005) provides strong evidence of the contextual effect among people in the same residential communities.
72
3 Depression Hurts, Depression Costs: The Medical Spending …
result from the daily stress of living in a place where social order is less apparent and social incivilities occur (Ross 2000; Silver et al. 2002; Ross et al. 2000; Wandersman and Nation 1998; Ewart and Suchday 2002); the difficulties of sustaining supportive social contacts in an unequal or disorganized social environment may present additional psychological stress on the local residents (Sampson et al. 1997; Geis and Ross 1998; Lindström et al. 2003; Elliott 2000). The strong correlation between the community-level prevalence and the individual-level mental health status is also evidenced in our data: the F-statistics associated with the Stock and Yogo test are 147.39 and 293.68 in the first-stage regressions (reported in Appendix Table 3.7), and the Cragg-Donald Wald F-statistic for the weak identification test is 51.371, both showing that the IVs are not likely to be weak (Stock et al. 2002). Second, the areabased prevalence rates should (arguably) not directly correlate with the individual medical expenditure without affecting the individual’s mental health status. This is because the medical expenditure and insurance reimbursement are not shared or cross-subsidized on the community level in China, and individuals within the same neighborhood are only responsible for paying their own medical bills. This exclusion restriction condition is statistically verified by Hansen’s J test in our over-identified IV model: the p-value for the exclusion restriction hypothesis is 0.2391, suggesting that the IVs are not directly correlated with the unexplained portion of the individual medical expenditure. Third, our IV approach is supported by many prior studies that also use area-based measures to instrument individual-level behaviors (Currie and Cole 1993; Goldman et al. 2001; Bhattacharya and Bundorf 2009; Grabowski and Hirth 2003; Lo Sasso and Buchmueller 2004; Morris 2007; Lei and Lin 2009; Pan et al. 2013; Qin and Pan 2016). The results of the IV-2PM regressions are reported in Column (5)–(6) of Table 3.6. After controlling for the endogeneity of mental health status, our basic results remain the same in the sense that both depression and depressive symptoms are associated with higher probability and higher conditional values of medical spending. However, some of the estimated coefficients (especially in the intensity equation) are not statistically significant, which may be due to the small number of the compliers13 who incur nonzero medical spending in our sample. With regard to other explanatory variables, the estimated coefficients reported in the IV-2PM model are generally consistent with the baseline regression reported in Table 3.2. Overall, these additional results indicate that the basic findings in our study on the positive association between depression and medical spending is valid and consistent.
13
In our context, the compliers are the individuals who change their medical spending due to the changes in IVs.
3.6 Discussions and Conclusions
73
3.6 Discussions and Conclusions During the past decades, China’s rapid economic growth has been accompanied by rapid changes in lifestyle and an increasing prevalence of NCDs such as mental disorders. Previous studies show that the prevalence rate of depression, estimated with CES-D, is high and unevenly distributed across regions and subpopulations (Qin et al. 2016). However, few studies have paid attention to the impact of depression on healthcare costs. This paper provides the first nationally representative estimate on the medical costs induced by depression and depressive symptoms in China, the largest developing country in the world; in addition, it contributes to the health economic literature by expanding the use of hurdle models (such as 2PM and 4PM) to the burden of disease estimation in the mental health area. Our population-based estimation methodology is justified by the stylized fact that psychological, social and institutional barriers often prevent or delay the mental health patients from seeking appropriate care, leading to severe under-diagnosis and under-treatment of mental health conditions. Furthermore, due to the co-existence of mental conditions and other chronic physical conditions (such as hypertension and diabetes), the cost impact of depression is also driven by the cost-shifting effect from mental healthcare to general healthcare and the co-morbidity effect between mental conditions and other NCDs. The existence of these two effects highlights the advantage of using survey data to quantify the impact of depression and depressive symptoms on healthcare costs, which in turn highlights two important findings. First, our regression results indicate that the mental health conditions significantly increase both the probability and the amount of medical spending by the Chinese adults, and the impact is significant for both the outpatient and inpatient spending, and robust under different model specifications. The counter-factual analysis (based on 2PM) suggests that depressive symptoms and depression are associated with 7.8 and 6.9% higher expected medical costs. Second, the induced costs are not evenly distributed across regions and subpopulations, with women, the rural residents, the elderly people, and the low-educated groups paying a higher share of medical spending due to depression and depressive symptoms. This suggests that these disadvantaged groups may have more difficulty in overcoming the social and institutional barriers in accessing mental healthcare in China. The above conclusions shed light on the urgent need for reforming the current mental health system in China, and further government involvement is required to improve the treatment and prevention of the mental health conditions. An important priority of the reform is to move away from a hospital-centered health system toward a patient-centered system, in which patients with mental illnesses and other NCDs are incentivized to be treated at the community level.14 Given that our estimated 14
As China has made significant progress in achieving universal coverage in recent years, the challenge of healthcare reform shifts from the financing system to the delivery system. A significant impact of expanding insurance coverage is that increasing percentage of population choose hospitals instead of primary care institutions for seeking healthcare, as local clinics are often seen as of poor quality.
74
3 Depression Hurts, Depression Costs: The Medical Spending …
medical costs of depression and depressive symptoms are almost three times as large as the cost impact of obesity and overweight (Qin and Pan 2016), which is another public health concern and increasingly catches the public attention, China will have to battle against the escalating disease burden induced by these NCDs in the coming years, and a hospital-centered health system would be ill-suited for this task. As a result, the legal, regulatory, and policy changes are needed to strengthen the primary mental healthcare system. For example, training more qualified mental health physicians and establishing more primary mental healthcare facilities are both in urgent need to close the fundamental gap between the supply and demand of mental healthcare in China. Favorable financing and payment schemes can also be designed to reduce the monetary hurdle that prevents the mental illness patients from accessing primary mental health services. In addition, legislation efforts can also be made to reduce the social stigma on people with mental conditions, which can be a formidable non-monetary barrier in seeking mental healthcare. Given the uneven distribution of depression-induced medical costs, our results also suggest that more policy attention should be devoted to the underserved areas and the disadvantaged groups such as women, rural residents, and the low-educated people, which in turn may be an effective way to improve the overall mental health status for the country that hosts the world’s one-fifth population.
Appendix See Appendix Table 3.7.
Table 3.7 First stage regression results for IV-2PM
Variable
mhs = 1 (1)
(2)
IV: mhs1_rate
0.0495**
0.526***
(0.0208)
(0.0327)
IV: mhs2_rate
0.355***
0.483***
(0.0725)
(0.0865)
1.402***
− 1.141***
IV: mhs1_rate * mhs2_rate
mhs = 2
(0.368)
(0.433)
Female
0.0528***
0.0257***
(0.00398)
(0.00257)
Age
0.00103***
0.00102***
(0.000168)
(0.000111)
0.00998**
0.0121*
(0.00434)
(0.00724)
Single
(continued)
References Table 3.7 (continued)
75
Variable
mhs = 1
mhs = 2
(1)
(2)
Divorced/widowed
0.0539***
0.0284***
(0.00764)
(0.00935)
Familysize
− 0.00389***
− 0.00240***
(0.00116)
(0.000757)
− 0.00227
0.000834
Urban
(0.00494)
(0.00307)
Middle
− 0.0216***
− 0.0259***
(0.00290)
(0.00484)
High
− 0.0211***
− 0.0319***
(0.00354)
(0.00603)
College
− 0.0218***
−0.0311***
(0.00395)
(0.00758)
−0.00999***
−0.00425***
Log (income)
(0.00185)
(0.00125)
Farm
0.00299
−0.00407
(0.00460)
(0.00648)
Government
−0.00468
−0.00114
(0.00474)
(0.00780)
0.00588
0.00923
Collective
(0.00835)
(0.0136)
Private
−0.00725*
0.0133**
(0.00374)
(0.00598)
Other
0.00194
0.0140**
(0.00465)
(0.00698)
Province dummy
Yes
Yes
Stock and Yogo F-statistics
147.39
293.68
Sample size
30,452
30,452
References Andlin-Sobocki P, Jonsson B, Wittchen H-U, Olesen J (2005) Cost of disorders of the brain in Europe. Eur J Neurol 12(Suppl. 1):1–27 Aneshensel CS, Sucoff CA (1996) The neighborhood context of adolescent mental health. J Health Soc Behav 37(4):293 Berndt ER, Cockburn IM (2014) The hidden cost of low prices: limited access to new drugs in India. Health Aff (millwood) 33(9):1567–1575. https://doi.org/10.1377/hlthaff.2013.1307 Bharadwaj P, Pai MM, Suziedelyte A (2015) Mental health stigma. National Bureau of Economic Research Working Paper 21240, Cambridge
76
3 Depression Hurts, Depression Costs: The Medical Spending …
Bhattacharya J, Bundorf MK (2009) The incidence of the healthcare costs of obesity. J Health Econ 28(3):649–658 Bor JS (2015) Among the elderly, many mental illnesses go undiagnosed. Health Aff 34(5):727–731 Buntin MB, Zaslavsky AM (2004) Too much ado about two-part models and transformation? Comparing methods of modeling medicare expenditures. J Health Econ 23(3):525–542 Cawley J, Meyerhoefer C (2012) The medical care costs of obesity: an instrumental variables approach. J Health Econ 31(1):219–230 Chaix B et al (2006) Spatial clustering of mental disorders and associated characteristics of the neighbourhood context in Malmö, Sweden, in 2001. J Epidemiol Community Health 60(5):427– 435 Currie J, Cole N (1993) Welfare and child health: the link between AFDC participation and birth weight. Am Econ Rev 83(4):971–985 Deb P, Munkin MK, Trivedi PK (2006) Bayesian analysis of the two-part model with endogeneity: application to health care expenditure. J Appl Economet 21(7):1081–1099 Dan C, Sweeny K, Sheehan P et al (2016) Scaling-up treatment of depression and anxiety: a global return on investment analysis. Lancet Psychiatry 3(5):415–424 Druss BG, Rosenheck RA (1999) Patterns of health care costs associated with depression and substance abuse in a national sample. Psychiatr Serv 50(2):214–218 Elliott M (2000) The stress process in neighborhood context. Health Place 6(4):287–299 Ewart CK, Suchday S (2002) Discovering how urban poverty and violence affect health: development and validation of a neighborhood stress index. Health Psychol Offic J Div Health Psychol Am Psychol Assoc 21(3):254–262 Finkelstein EA, Trogdon JG, Cohen JW, Dietz W (2009) Annual medical spending attributable to obesity: payer-and service-specific estimates. Health Aff 28(5):822–831 Finkelstein EA, Fiebelkorn IC, Wang G (2003) National medical spending attributable to overweight and obesity: how much, and who’s paying? Health Affairs W3 Fitzpatrick KM (1993) Exposure to violence and presence of depression among low-income, African-American youth. J Consult Clin Psychol 61(3):528–531 Fox JW (1990) Social class, mental illness, and social mobility: the social selection-drift hypothesis for serious mental illness. J Health Soc Behav 31(4):344–353 Geis KJ, Ross CE (1998) A new look at urban alienation: the effect of neighborhood disorder on perceived powerlessness. Soc Psychol Q 61(3):232–246 Goldman DP, Bhattacharya J, Mccaffrey DF, Duan N, Leibowitz AA, Joyce GF, Morton SC (2001) Effect of insurance on mortality in an HIV-positive population in care. J Am Stat Assoc 96(455):883–894 Grabowski DC, Hirth RA (2003) Competitive spillovers across non-profit and for-profit nursing homes. J Health Econ 22(1):1–22 Greenberg PE, Kessler RC, Birnbaum HG, Leong SA, Lowe SW, Berglund PA et al (2003) The economic burden of depression in the United States: how did it change between 1990 and 2000? J Clin Psychiatry 64(12):1465–1475 Greenberg PE, Fournier A-A, Sisitsky T et al (2015a) The economic burden of adults with major depressive disorder in the United States (2005 and 2010). J Clin Psychiatry 76(2):155–162 Hu T-W, He Y, Zhang M, Chen N (2007) Economic costs of depression in China. Soc Psychiatry Psychiatr Epidemiol 42(2):110–116 IMS Institute (2014) Global outlook for medicines through 2018. IMS Institute, Parsippany, New Jersey Jones AM (2000) Health econometrics. Handb Health Econ 1:265–344 Lei X, Lin W (2009) The new rural cooperative medical scheme in rural China: does more coverage mean more service and better health? Health Econ 18(S2):S25–S46 Lei X, Nina Y, Zhao Y (2012) Socioeconomic status and chronic disease: the case of hypertension in China. China Econ Rev 23(1):105–121 Liem R, Liem J (1978) Social class and mental illness reconsidered: the role of economic stress and social support. J Health Soc Behav 19(2):139–156
References
77
Lindström M, Merlo J, Ostergren PO (2003) Social capital and sense of insecurity in the neighbourhood: a population-based multilevel analysis in Malmö, Sweden. Soc Sci Med 56(5):1111–1120 Liu LJ, Guo Q (2007) Loneliness and health-related quality of life for the empty nest elderly in the rural area of a mountainous county in China. Qual Life Res 16(8):1275–1280 Liu J, Ma H, He YL et al (2011) Mental health system in China: history, recent service reform and future challenges. World Psychiatry 10(3):210–216 Lo Sasso A, Buchmueller TC (2004) The effect of the state children’s health insurance program on health insurance coverage. J Health Econ 23(5):1059–1082 Ma X, Xiang YT, Cai ZJ et al (2009) Prevalence and socio-demographic correlates of major depressive episode in rural and urban areas of Beijing, China. J Affect Disord 115(3):323–330 Manning WG, Mullahy J (2001) Estimating log models: to transform or not to transform? J Health Econ 20(4):461–494 March J, Silva S, Petrycki S et al (2004) Fluoxetine, cognitive-behavioral therapy, and their combination for adolescents with depression: treatment for adolescents with depression study (TADS) randomized controlled trial. JAMA 292(7):807–820 Mechanic D (1972) Social class and schizophrenia: some requirements for a plausible theory of social influence. Soc Forces 50(3):305–309 Meng QY, Fang H, Liu X, Yuan B, Xu J (2015) Consolidating the social health insurance schemes in China: towards an equitable and efficient health system. Lancet 386(10002):1484–1492 Mirowsky J, Ross CE (2003) Education, social status, and health. Transaction Publishers Mirowsky J, Ross CE (1998) Education, personal control, lifestyle and health: a human capital hypothesis. Res Aging 20(4):415–449 Mora T, Gil J, Sicras-Mainar A (2015) The influence of obesity and overweight on medical costs: a panel data perspective. Eur J Health Econ 16(2):161–173 Morgan G (2005) Highlighting the importance of “popular epidemiology.” J Epidemiol Community Health 59(4):254–257 Morris S (2007) The impact of obesity on employment. Labour Econ 14(3):413–433 Moussavi S, Chatterji S, Verdes E, Tandon A, Patel V, Ustun B (2007) Depression, chronic diseases, and decrements in health: results from the world health surveys. The Lancet 370(8–14):851–858 Nolen-Hoeksema S (2001) Gender differences in depression. Curr Dir Psychol Sci 10(5):173–176 Pan J, Qin X, Liu G (2013) The impact of body size on urban employment: evidence from China. China Econ Rev 27(4):249–263 Patel V, Chatterji S (2015) Integrating mental health in care for noncommunicable diseases: an imperative for person-centered care. Health Aff 34(9):1498–1505 Patel V, Dan C, Parikh R et al (2016) Addressing the burden of mental, neurological, and substance use disorders: key messages from disease control priorities. Lancet 387(10028):1672–1685 Penninx BW, Leveille S, Ferrucci L, Van Eijk JT, Guralnik JM (1999) Exploring the effect of depression on physical disability: longitudinal evidence from the established population for epidemiologic studies of the elderly. Am J Public Health 89(9):1346–1352 Phillips MR, Zhang J, Shi Q et al (2009a) Prevalence, treatment, and associated disability of mental disorders in four provinces in China during 2001–05: an epidemiological survey. Lancet 373(9680):2041–2053 Qin X, Pan J (2016) The medical cost attributable to obesity and overweight in China: estimation based on longitudinal surveys. Health Econ 25(10):1291–1311 Qin XZ, Wang SY, Hsieh C (2016) The prevalence of depression and depressive symptoms among adults in China: estimation based on a national household survey. China Econ Rev, in Press. https://doi.org/10.1016/j.chieco.2016.04.001 Radloff LS (1977) The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol 1(3):385–401 Radloff LS (1991) The use of the center for epidemiologic studies depression scale in adolescents and young adults. J Youth Adolesc 20(2):149–166
78
3 Depression Hurts, Depression Costs: The Medical Spending …
Ross CE (2000) Neighborhood disadvantage and adult depression. J Health Soc Behav 41(2):177– 187 Ross CE, Reynolds JR, Geis KJ (2000) The contingent meaning of neighborhood stability for residents’ psychological well-being. Am Sociol Rev 65(4):581–597 Salvador-Carulla L, Bendeck M, Fern´andez A, Alberti C, Sabes-Figuera R, Molina C, Knapp M (2011) Costs of depression in Catalonia (Spain). J Affect Disord 132(1–2):130–138 Sampson RJ, Raudenbush SW, Earls F (1997) Neighborhoods and violent crime: a multilevel study of collective efficacy. Science 277(5328):918–924 Silver E, Mulvey EP, Swanson JW (2002) Neighborhood structural characteristics and mental disorder: Faris and Dunham revisited. Soc Sci Med 55(8):1457–1470 Simon RW (2002) Revisiting the relationships among gender, marital status, and mental health. Am J Sociol 107(4):1065–1096 Sloan F, Hsieh CR (2017) Health economics, 2nd edn. The MIT Press, Cambridge Sobocki P, Ekman M, Agren H et al (2007) Resource use and costs associated with patients treated for depression in primary care. Eur J Health Econ 8(1):67–76 Stock JH, Wright JH, Yogo M (2002) A survey of weak instruments and weak identification in generalized method of moments. J Bus Econ Stat 20(4):518–529 Tomonaga Y, Haettenschwiler J, Hatzinger M, Holsboer-Trachsler E, Rufer M, Hepp U, Szucs TD (2013) The Economic burden of depression in Switzerland. Pharmacoeconomics 31(3):237–250 Trivedi DN, Lawrence LW, Blake SG et al (2004) Study of the economic burden of depression. J Pharm Finan Econ Policy 13(4):51–64 Trogdon JG, Finkelstein EA, Hoerger TJ (2008) Use of econometric models to estimate expenditure shares. Health Serv Res 43(4):1442–1452 Tsang A, Von Korff M, Lee S et al (2008) Common chronic pain conditions in developed and developing countries: gender and age differences and comorbidity with depression-anxiety disorders. J Pain 9(10):883–891 Van Itallie TB (2005) Subsyndromal depression in the elderly: underdiagnosed and undertreated. Metabolism 54(5):39–44 Wandersman A, Nation M (1998) Urban neighborhoods and mental health. Psychological contributions to understanding toxicity, resilience, and interventions. Am Psychol 53(6):647–56 Wang YC, McPherson K, Marsh T, Gortmaker SL, Brown M (2011) Health and economic burden of the projected obesity trends in the USA and the UK. Lancet 378(9793):815–825 WHO (2013a) Investing in mental health: evidence for action. World Health Organization, Geneva WHO (2013b) Mental health action plan 2013–2030. World Health Organization, Geneva WHO (2016) Mental disorders: fact sheet. Accessed on January 2017. Available from: http://www. who.int/mediacentre/factsheets/fs396/en/ Xiaoyan L, Yin N, Zhao Y (2012) Socioeconomic status and chronic disease: the case of hypertension in China. China Econ Rev 23(1):105–121 Xie Y, Hu J (2014) An introduction to the China family panel studies (CFPS). Chin Sociol Rev 47(1):3–29 Xie LQ, Zhang JP, Peng F, Jiao NN (2010) Prevalence and related influencing factors of depressive symptoms for empty-nest elderly living in the rural area of YongZhou, China. Arch Gerontol Geriatr 50(1):24–29 Yang G, Wang Y, Zeng Y et al (2013) Rapid health transition in China, 1990–2010: findings from the global burden of disease study 2010. Lancet 381(9882):1987–2015 Zimmerman FJ, Katon W (2005) Socioeconomic status, depression disparities, and financial strain: what lies behind the income depression relationship? Health Econ 14(12):1197–1215
Chapter 4
The Hidden Costs of Mental Depression: Implications on Social Trust and Life Satisfaction
4.1 Introduction In many countries, population health and economic output have simultaneously increased substantially over the previous century. As a result, many studies have paid attention to the relationship between health and the economy, with the special focus on an attempt to identify the direction of causal impacts and the channels of such impacts. The majority of research in this line focused on the casual effect of improvement in population health on economic growth (Weil 2014). The existing studies have identified that health contributes to the economic growth through four major channels: (1) higher productivity (Bhargava et al. 2001; Bloom et al. 2004; Well 2007); (2) higher labor supply (Bloom et al. 2000); (3) improved skills because of greater education and training (Soares 2005); and (4) increased savings available for investment in physical and intellectual capital (Bloom et al. 2003). Overall, these studies focused on the impact of health on the proximate factors of growth, stressing the role of labor supply, and the accumulation of human and physical capital. There is virtually no study exploring the potential impact of health on deeper factors affecting economic growth, such as social trust. In addition, in addressing the relationship between health and the economy, most of previous studies only paid attention on the physical health. Similar studies in the context of mental health are rare. The prevalence of mental illness such as depression has increased rapidly in recent decades, which in turn has generated scholarly interest around the world to investigate the economic burden of the disease. For example, Greenberg et al. (2015a, b) report that the economic burden of depression in the United States was estimated at $210.5 billion in 2010. Hsieh and Qin (2018) estimate that the annual costs attributable to depression and depressive symptoms in China are RMB 126 and 142 per person, The content of this chapter is published in Hsieh et al. (2019). Copyright John Wiley and Sons (2019), reproduction license granted.
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 X. Qin and C.-R. Hsieh, Economic Analysis of Mental Health in China, Applied Economics and Policy Studies, https://doi.org/10.1007/978-981-99-4209-1_4
79
80
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
which account for 6.9 and 7.8% of total personal expected medical spending, respectively. In this line of research, the existing literature typically considers two types of costs associated with depression: (1) the direct cost, which includes the outpatient and inpatient medical costs for the treatment of depression and its complications; and (2) the indirect cost, or the opportunity cost of being depressive, which includes the morbidity costs caused by absenteeism (missed work days due to depression), presenteeism (reduced productivity while at work due to depression), and the mortality costs defined as the product of the number of deaths due to depression and the average expected future earnings. However, the real cost of depression to the individuals and society as a whole goes well beyond the traditional boundary of disease burden estimation. These social costs, despite their great implication to the individual quality of life and the overall economic development, have not been widely recognized in the literature and no attempts have been made to quantify such burden. The purpose of this paper is to provide a conceptual analysis and empirical evidence on the importance of two hidden costs associated with depression: lower social trust and lower life satisfaction. Specifically, we estimate the impacts of a mental depression indicator (CES-D20) on a series of variables measuring the individuals’ tendency of trusting other people and their life satisfaction, using data from the 2012 China Family Panel Studies. Our study contributes to the growing body of research on the link between health and wealth in general and the effect of health on economic growth in particular with two paradigm shifts: from physical to mental health and from the proximate factors to deeper factors of economic growth. In addition, we extend the link between health and the economy from the economic variables such as GDP per capita to the non-economic variables such as subjective evaluation of well-being. Specifically, we attempt to bring three lines of research together: (1) the rising prevalence of mental health problems in the developing countries; (2) the role of trust in the economic development; and (3) the determinants of life satisfaction and well-being. Although many studies have accumulated evidence on the rising prevalence and disease burden of mental health such as depression, the policy action lags behind the research. On average, high-income countries at most spend only about 5% of their healthcare resources on mental health in spite of a relatively large share of disease burden arising from this disease. This share is even smaller than 1% in low- and middle-income countries. As a result, many international agencies have initiated the call to set mental health as a global development priority (World Bank Group and WHO 2016). They proposed for increasing investment on mental healthcare as an important strategy to close the gap of inadequate funding. Our research echoes this initiative by examining the costs of mental health in a more general framework that considers the link between health, wealth, and well-being, with a special focus on the impact on a fundamental source of economic growth: trust. In recent years, trust has been recognized as one of the most fundamental culture values, which in turn determines many economic choices and hence further affects the speed of development and the wealth of nation (Algan and Cahuc 2014; Alesina and Giuliano 2015). Thus, trust has been classified as one of the deeper factors that affect economic growth and development (Spolaore and Wacziarg 2013). The
4.1 Introduction
81
importance of trust has induced many researchers to pay attention on several questions such as how to measure trust and what are the major determinants of trust. For example, Luo (2005) proposed two types of trust: particularistic trust and general trust. The former one refers to the trust to specific individuals such as neighbors and doctors, while the latter refers to the general propensity to trust others. Alesina and Ferrara (2002) specifically examine the determinants of general trust. They find that both individual experiences (such as suffering from a major negative event) and community characteristics (such as living in a racially mixed community) have strong impacts on how much people trust each other. Although this study mentioned the potential role of individual health outcomes in the formation of trust, it did not measure in an explicit way on the relationship between the mental health status and the trust levels. Similar to the growth in the research literature on trust, life satisfaction has also received increasing attention from both researchers and policymakers, partly because of the widely recognized limitation of the traditional well-being measures (such as GDP per capita) and the strong desire of seeking for empirical alternatives (Frey and Stutzer 2002; Deaton 2008). Being an increasingly popularized concept, life satisfaction and its determinants have attracted substantial research efforts in recent years, which echoes the persistent interest in the economic literature to identify the important drivers of economic growth. The existing studies have shown that income, health, job/daily activities, and family/social contacts are the four important dimensions that shape the variation in life satisfaction among individuals in different countries (Kapteyn et al. 2009). However, the relative importance of each specific factor may vary across individuals and countries. As the rising prevalence of mental health problems has become a global public health concern, it bears important implications to study the impact of depression on life satisfaction, which in turn may shed new light on the fundamental sources of economic development. Our results indicate that individuals who have a higher tendency of suffering from depression or depressive symptoms are less likely to trust other people, and they also have significantly lower life satisfaction than their counterparts with better mental health. Given that trust is an important component of social capital, which in turn is a crucial input to foster economic growth in general and innovation in particular, the reduction in trust as induced by mental depression may impose a significant cost to the society in the form of weakened productivity and economic performance. This is a real cost that the society has to pay for the rising trend of depression and other mental health problems, but the empirical literature has devoted little research attention to quantify the magnitude of such costs. Similarly, our study also highlights another less-researched cost of depression: the increasing prevalence of depression leads to a reduction in the individual well-being in the form of lower life satisfaction. All these costs are real, but did not receive sufficient attention in the previous research, and thus we refer to them as hidden costs. The contribution of our research is to shed light on the existence of these hidden costs and to estimate their quantitative magnitude using China, the world’s largest developing country with the most rapidly increasing prevalence of depression, as an example.
82
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
The rest of the paper is organized as follows. Section 4.2 provides the research background by briefly reviewing the existing evidence on three lines of research, namely mental health, trust, and life satisfaction, and describes a conceptual framework on the linkage of these three dimensions. Section 4.3 describes the data and econometric models. Section 4.4 shows the main results of our empirical analysis. The last section concludes the paper and discusses the implications of the findings.
4.2 Research Background 4.2.1 The Rising Prevalence of Depression Mental disorders in general as well as depression and anxiety disorders in particular are becoming more prevalent worldwide. For example, a WHO report indicates that the size of the world’s population suffering from depression and/or anxiety increased from 416 million in 1990 to 615 million in 2013, suggesting that near 10% of the global population is affected (World Bank Group and WHO 2016). Consequently, many studies have pointed out that the disease burden of depression and anxiety disorders is growing rapidly and is likely to have a substantial social and economic impact (Layard 2013). A specific example is that mental illnesses account for nearly one-quarter of all years lived with disability (YLD) in China (Yang et al. 2013a, b). This study also finds that among the top 20 causes of YLD in China, seven of them are related to mental disorders, including major depressive disorder, alcohol use disorders, schizophrenia, anxiety disorders, bipolar disorder, dysthymia, and drug use disorders. In comparison to other non-communicable diseases (NCDs) with high disease burden and prevalence rate (such as hypertension and diabetes), the diagnosis and treatment for mental illnesses such as depression have two unique characteristics. First, the rate of treatment for mental disorders is quite low, indicating that there is a significant level of under-treatment. For example, a recent study in the US finds that only about one-third of adults with screen-positive depression receive medical treatment (Olfson et al. 2016). The treatment rate is even lower in the low- and middle-income countries, indicating that the under-treatment of mental illness is a worldwide phenomenon (Layard 2013). A study in China suggests that less than one-tenth of individuals with mental illness have ever received any type of mental health services (Philips et al. 2009). Second, in contrast to the high disease burden of mental illness in the world, the healthcare resources allocated to the treatment of mental illnesses are relatively low compared to general healthcare in both high- and middle-income countries. High-income countries on average spend about 5–14% of their healthcare expenditure on mental healthcare (Frank 2011), while in low-income countries this ratio is as low as 1% (World Bank Group and WHO 2016). Figure 4.1 provides a conceptual framework that illustrates how under-funding and under-treatment work together to create a vicious circle in the mental health
4.2 Research Background
83
sector. It has been widely recognized that stigma is an important reason to explain the low treatment rate among individuals with mental illness. However, the underfunding problem in the mental health sector also creates several access barriers to prevent individuals with mental illnesses to receive appropriate healthcare. First, the low share of healthcare funds allocated to mental healthcare often forces the government to impose high cost-sharing policies or less generous coverage for the insurance of mental healthcare. As a result, individuals with mental illness often need to pay a higher out-of-pocket expenditure in seeking medical treatment as compared to general healthcare. As noted in Frank and McGuire (2000), the demand for mental health services is more price elastic than that for general healthcare, indicating that the higher out-of-pocket costs are very likely to deter the use of mental healthcare. Second, the under-funding in the mental health sector also translates to the overall insufficiency and geographic misdistribution of healthcare resources for the appropriate delivery of mental health services, which in turn serves as the “availability barrier” for mental illness patients as they need to spend a higher time cost (in the form of long waiting or long-distance traveling) in seeking care, especially in comparison to the non-mental healthcare patients. Third, the under-funding in the mental health sector also reduces the speed of technology adoption in local practices and hence creates an additional treatment gap, leading to further reduction in the potential effectiveness of medical treatment. In summary, under-funding in the mental health sector creates several access barriers that cause under-treatment, which in turn is also a culprit to cause the underfunding, and hence a vicious circle takes shape. Frank (2011) identifies several reasons to explain the persistent trend of under-funds in the mental health sector across countries. One obvious reason arises from the budget rigidity in the public sector as many countries rely on the fixed budget to finance mental health sector. In a Access barriers to Mental Health Care 1. High cost-sharing 2. Low availability 3. Slow technology diffusion
Undertreat
Stigma
Underfund
1. Misperception of the benefits 2. Misperception of the productivity 3. Budget rigidity 4. Weak political power
Fig. 4.1 The vicious circle between under-treatment and under-funding in the mental health sector
84
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
typical fiscal arrangement, the spending in the previous periods usually has a strong impact on the size of budget in the current period. However, the existence of social stigma prevents the individuals with mental illness to form a strong interest group to persuade the decision makers for a higher share of healthcare budget, and the low treatment rate in mental illness may create a misperception on the benefits of effective treatment and the productivity of mental healthcare spending among the policymakers, which in turn plays an important role in shaping the public budget allocated to this sector. Thus, there is an urgent need to increase the awareness and understanding on the costs and benefits of the treatment for mental disorders, which in turn could be the key to break the vicious circle in the mental health sector. Our study contributes to this effort by increasing the understanding on the social benefits of depression treatment from the perspectives of social trust and life satisfaction, which are largely ignored in the previous investigations that primarily focus on the private medical benefits. The importance of spelling out these social consequences of mental illness is implied by the literature on the impact of trust on economic development as well as on the determinants of life satisfaction and well-being, which is summarized in the following.
4.2.2 The Role of Trust in Economic Development In recent years, trust has received a great deal of attention in economics literature. Empirically, trust can be measured with surveys. The major data sources that have been widely used in this line of research include the World Values Survey (WVS) and the General Social Survey (GSS). These surveys measure trust by asking the following standard question: “Generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?” Based on the answer to this question, researchers often construct a trust indicator which equals 1 if the respondent answers “most people can be trusted” and 0 otherwise (Knack and Keefer 1997; Alesina and Ferrara 2002; Delhey and Newton 2005; Delhey et al. 2011; Algan and Cahuc 2014). The advantage of using a simple binary variable to measure the general trust in most people is that the results can be compared across countries. A stylized fact obtained from the existing studies of general trust is that there are substantial variations in trust levels across countries. For example, based on the average responses to the trust question in various surveys obtained from 111 countries, Algan and Cahuc (2014) report that the average trust levels (measured by the percentage of samples expressing trust in most people) range from 3.8 in Trinidad and Tobago to 68.1 in Norway. The variations in trust levels across countries in turn have attracted many studies to investigate the determinants of trust on the country level (e.g. Knack and Keefer 1997; Delhey and Newton 2005). These studies yielded several important and consistent findings. First, income inequality is associated with low trust. Second, ethnic and linguistic divisions are also associated with low trust. Third, good governance, in terms of formal institutions for protecting
4.2 Research Background
85
property and contract rights, is positively associated with high trust. Similarly, some other research efforts have been devoted to investigate the determinants of trust on the individual level, and they also find a consistent pattern (e.g. Glaeser et al. 2000; Alesina and Ferrara 2002). First, high-income and well-educated individuals tend to have a higher trust in other people than the poor and low-educated people. Second, the community characteristics are important determinants of trust: individuals who live in racially mixed communities are less likely to report that “most people can be trusted”, indicating that ethnic heterogeneity has a negative impact on trust. One concern in the empirical measure of general trust is that people may interpret “most people” in different ways. Thus, relying on one simple question of general trust may not be sufficient to capture all the relevant contents of trust. For example, one puzzle found in the World Values Survey is that the trust level in China is very high, ranking number 4 among the 111 countries (Algan and Cahuc 2014), despite the severe income inequality and social conflicts. Delhey et al. (2011) suggest that people may have different radius of relationship in mind when they answer the standard question in the general trust survey. Thus, they explore the questions on the determinants of trust by adding more information on specific trust, which measures the inner-group and outer-group trust. The inner-group trust refers to the trust in family, neighborhood and the people that the respondent know personally. By contrast, the outer-group trust refers to the trust in people that respondents meet for the first time and people of another religion or nationality. By adding the information obtained from the inner-group and outer-group trust surveys Delhey et al. (2011) developed a radius-adjusted trust score that takes into account the variation of innerand outer-group trust across countries. Their results indicate that the radius of “most people” is narrower in Confucian countries such as China and South Korea and wider in Western high-income countries. After this adjustment, the ranking of trust level for China slides down roughly 10 places among 51 countries in their study samples. This study indicates that the radius of trust matters in the international comparison. Another line of research on trust is to investigate the impact of trust on economic performance. Aghion et al. (2010) provide evidence to support the argument that countries with low trust levels have strong public demand for regulation, which in turn discourages the spontaneous formation of trust. As a result, low trust and regulation interact together and create a vicious circle. This finding supports the argument in an experimental study that the control imposed by the principal is often perceived as a signal of distrust by the agent, which in turn leads to a reduction in the agent’s performance (Falk and Kosfeld 2006). This line of research indicates that low trust will impose a hidden cost to the society in terms of low economic performance. In summary, the growing body of research on trust has increased our knowledge on its determinants and impacts, which in turn has put trust onto the center stage in mainstream economics (Algan and Cahuc 2014). In spite of a growing body of research on trust, few studies have paid attention to the potential link between rising trend in mental health and trust, which is a gap that our study attempts to fill.
86
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
4.2.3 Life Satisfaction and Well-Being Since the late 1990s, the measurement of subjective well-being such as life satisfaction and happiness has been widely studied in economics literature. The accumulated evidence provides consistent support to the notion that subjective well-being is a good proxy of individual utility (Frey and Stutzer 2002; Kahneman and Krueger 2006). Empirically, individual happiness and life satisfaction can be captured by surveys. Specifically, the standard question of measuring life satisfaction adopted in many surveys such as WVS and GSS is the following: “All things considered, how satisfied are you with your life as a whole these days?” This question is often assessed on a five-point or ten-point scale from “very dissatisfied” to “very satisfied”, which in turn provides a quantitative measure that allows the researchers to capture human well-being directly and to compare it across individuals and countries. The previous studies have drawn several consistent conclusions on the determinants of subjective well-being. The most cited factor that accounts for the variations in life satisfaction is income (Deaton 2008). Based on the cross-section comparisons, both individual-level and country-level data show that richer people and richer countries, on average, report better life satisfaction levels and higher subjective well-being compared to poorer people and countries. In other words, income is positively correlated with life satisfaction at a given point in time. Overtime, however, subjective well-being, either measured by country- or individual-level data, does not increase significantly or even decreases slightly despite a considerable growth in per-capita income. For example, between 1958 and 1991, there was a sixfold increase in real per-capita income in Japan, but the average life satisfaction almost remained constant during this period (Frey and Stutzer 2002). Similarly, the individual-level data also show that there was a slight decrease in the reported life satisfaction in China between 1994 and 2005 although the real income per capita increased by a factor of 2.5 during this period (Kahneman and Krueger 2006). There are two plausible explanations for the inconsistent pattern on the income-life satisfaction relationship between cross-section and time-series data, also dubbed the “income paradox”. First, the rank or relative position in the income distribution of the population or of one’s peer group may play a more important role than the absolute income levels in accounting for the variation of life satisfaction across individuals, an argument also known as the relative income or social comparison hypothesis. For example, Huang et al. (2016) test this hypothesis using data obtained from Chinese Household Income Project and find that relative income is negatively associated with the happiness score. This suggests that an individual’s absolute income is not as meaningful to life satisfaction as the individual’s relative income. Second, although life satisfaction and income are positively correlated in a crosssection dataset, the correlation is relatively low, around 0.20, indicating that only a small portion of the difference in life satisfaction among persons can be attributed to the difference in income. Thus, subjective well-being is not just a matter of income, and other non-income factors may be even more important in explaining the determinants of subjective well-being (Frey and Stutzer 2002; Kahneman and Krueger 2006;
4.3 Data and Method
87
Oswald and Wu 2011). Among the non-income factors, Schnitzlein and Wunder (2016) emphasize the importance of family effects in shaping the subjective wellbeing. Specifically, they find that around 30% to 60% of the inequality in permanent well-being can be attributed to family background. Other non-income factors that have been widely studied include unemployment and institution. For example, the existing literature shows that unemployment, either measured on the individual or country level, has a significantly negative impact on the reported subjective wellbeing. Similarly, institutions that foster the direct participation in public decisionmaking such as referenda and decentralization have significantly positive impact on subjective well-being (Frey and Stutzer 2002). Overall, past research has identified several important factors that influence the life satisfaction and happiness on both the individual and country levels, including income, family background, unemployment, and institutional factors. By contrast, few studies have paid attention to the potential impact of rising prevalence in mental disorders on the subjective well-being, especially for the low- and middle-income countries such as China.
4.3 Data and Method 4.3.1 Data Source China Family Panel Studies (CFPS) is a nationally representative longitudinal survey designed and implemented by the Institute of Social Science Surveys (ISSS) of Peking University. This survey was conducted in 25 Chinese provinces (these provinces jointly cover 95% of the Chinese population) in five years (2008, 2009, 2010, 2011, 2012). In each wave, the CFPS survey samples about 15,000 households nationwide using the multi-stage probability proportional to size (PPS) sampling method, and all family members in each sample household are included. The questionnaire collects individual-, family-, and community-level information on the demographic, socioeconomic, and health-related variables. In the 2012 CFPS survey, a full 20-question version of the CES-D (Center for Epidemiologic Studies Depression) questionnaire (Radloff 1977) was administered to assess the respondents’ mental health status.1 The CES-D questionnaire is one of the most frequently used self-assessment tools for depression and depressive symptoms. An advantage of using this survey-based instrument is that the questions contained in CES-D are non-intrusive and related to everyday feelings,2 which makes it easier for the respondents to answer, leading 1
Although the CFPS dataset is a panel survey consisting of five waves, only the 2012 wave contains the full 20-question version of the CES-D, and thus, the study sample of this paper is only from the 2012 wave. 2 Examples of the CES-D questions include: “How often do you feel that everything I did was an effort?”; “How often do you feel not like eating (your appetite is poor)?”.
88
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
to better detection of their depressive symptoms compared to some other clinical instruments. This in turn may help to alleviate the underreporting problem commonly experienced among the mental illness patients (Bharadwaj et al. 2015). The CESD20 questionnaire contains four subscales: somatic-retarded activity, interpersonal relations, depressed affect, and positive affect. The former three measure negative emotions, while the latter measures positive ones. Respondents are asked to rate how often they experienced the specified emotions in the past week, with the options varying from 0 to 3 for each question (0 = rarely, 1 = little, 2 = occasionally, 3 = often). The CES-D score can thus be calculated based on the responses as follows: CES-D = +
k
Scorei,somatic +
i
Scorek.depressed +
Score j,interpersonal
j
(4 − Scorel,positive )
(4.1)
l
where Scorei,somatic , Score j,interpersonal , Scorek,depressed , and Scorel,positive represent the score for the i-th question on the somatic-retarded activity, the j-th question on interpersonal relations, the k-th question on the depressed affect, and the l-th question on the positive affect, respectively. Thus, the overall CES-D score ranges from 0 to 60, with a higher score indicating more frequent occurrence of depressive symptoms and higher likelihood of depression. We use the CES-D score in our main analysis to measure an individual’s tendency of mental depression. The CFPS questionnaire also contains a module to measure the respondent’s tendency to trust other people. This module includes seven questions, pertaining to the trustworthiness of most people in general (denoted as trust_dummy) and the degree of trust by the respondent on the following people in particular (ranked by the degree of closeness in the interpersonal relationship)—parents, neighbors, medical doctors, cadres (government officials), and strangers (denoted as trust_parent, trust_ neighbor, trust_doctor, trust_cadre, trust_stranger, respectively). These trust-related questions are closely analogous to those used in the World Values Surveys, among which trust_dummy for the general trust in other people is a binary (dummy) variable indicating a “yes or no” answer, and the other variables for the particularistic trust are ordinal scores varying between 0 and 10, with a higher score indicating more trust. Similarly, CFPS also surveys people’s satisfaction on their lives, using the following five questions: “How much are you satisfied with your family? How would you rank the social status of your family in the local area? How much are you satisfied with your life? How would you rank the social status of yourself in the local area? How confident are you on your future?” The answers to the above questions are ordinal scores (ranging from 1 to 5) and denoted as satis_family, ses_family, satis_self, ses_self, confi_self , respectively, with higher scores indicating stronger life satisfaction or confidence. We restrict our sample to adult respondents aged between 16 and 99 and further drop the observations with missing information on the key variables such as gender,
4.3 Data and Method
89
age, and CES-D scores. Our final study sample contains 30,858 observations, covering China’s 25 provinces with about 45% respondents from urban areas and 55% from rural regions.
4.3.2 Sample Descriptive Analysis Table 4.1 presents the sample summary statistics. In addition to the key variables (CES-D scores, trust, and satisfaction-related variables) introduced above, we also control the respondent’s demographic and socioeconomic characteristics, and such variables include gender, age in years, marital status (married, single, divorced, widowed), education levels (primary school or below, middle school, high school, college or above), work status (employed, unemployed, retired, out of labor force due to disability or diseases, out of labor force due to other reasons), personal annual income (in 1000 Yuan), etc. To control the regional influences on the respondents’ trust and life satisfaction, we also control for the urban/rural status as well as the provincial dummies of their residential locations. According to Table 4.1, about 51% of the respondents in our sample are female, and 45% live in the urban areas. About 80% of the respondents are married, and 14% are single; those who are divorced or widowed account for 1 and 5% of the full sample, respectively. The average age of our sample respondents is 45. In the socioeconomic dimension, the average annual personal income is 10,990 Yuan. As for the educational attainment, 50.5% of the respondents received primary school or below education, people who acquired middle and high school education account for about 28.3% and 13.8% of the full sample, respectively; only 7.5% of the sample received college or above education. For work status, 69.2% of the respondents are employed, while the rest are not currently working due to various reasons (e.g. 4.2% are unemployed, 16.2% are retired, and 10.3% are out of labor force due to disability, diseases, or other reasons). As mentioned above, the CES-D score ranges from 0 to 60, with a higher score indicating more depressive symptoms. The average CES-D score in our sample is 12.9, with a standard deviation of 7.95, suggesting that the respondents are on average mentally healthy. Radloff (1977) suggests that the threshold CES-D values of 16 and 28 can be used to categorize a person’s mental health status; i.e. a CES-D score between 16 and 28 suggests that the person has depressive symptoms, and a CES-D above 28 suggests that the person suffers from depression. According to this standard, around 27% of our sample respondents have depressive symptoms, and around 5% suffer from depression. In terms of trust-related variables, Table 4.1 shows about 54.3% of respondents think that most people are trustworthy, indicating a relatively high level of general
90
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
Table 4.1 Sample summary statistics of key variables Variable name
Variable definition
Mean/Std. deviation
cesd
CES-D score (0–60)
12.93
iv_comm
Average CES-D score in respondent’s community
12.94
(7.950) (3.627) iv_f
CES-D score of respondent’s father
12.11
iv_m
CES-D score of respondent’s mother
14.54
trust_dummy
Think most people are trustworthy (1 = yes)
0.543
(7.611) (8.567) (0.498) trust_parent
Level of trust on parents (0–10)
9.095 (1.685)
trust_neighbor Level of trust on neighbors (0–10)
6.373
trust_doctor
6.611
(2.217) Level of trust on doctors (0–10)
(2.255) trust_cadre
Level of trust on cadres (0–10)
4.880
trust_stranger
Level of trust on strangers (0–10)
2.187
satis_family
Satisfaction of one’s family (1–5)
3.476
(2.465) (2.135) (1.046) ses_family
Social status of one’s family in local area (1–5)
2.841
satis_self
Satisfaction of one’s life (1–5)
3.319
ses_self
Social status of oneself (1–5)
2.676
(0.957) (1.053) (1.020) confi_self
Degree of confidence to one’s future (1–5)
3.670
Female
0 = male, 1 = female
0.510
Age
Age in years (16–99)
45.28
(1.114) (0.500) (16.60) Married
Married (1 = yes)
0.796 (0.403) (continued)
4.3 Data and Method
91
Table 4.1 (continued) Variable name
Variable definition
Mean/Std. deviation
Urban
Live in urban areas (1 = yes)
0.451
Pincome
Personal annual income (in 1000 Yuan)
10.99
(0.498) (30.89) Education level Primary
Primary school or below (1 = yes)
0.505 (0.500)
Middle
Middle school (1 = yes)
0.283
High
High school (1 = yes)
0.138
College
College or above (1 = yes)
0.0748
(0.450) (0.345) (0.263) Work status Employed
Currently employed (1 = yes)
0.692 (0.462)
Unemployed
Currently unemployed (1 = yes)
0.0422
Retired
Retired from labor force (1 = yes)
0.162
OLF 1
Out of labor force due to disability or diseases (1 = yes) 0.0250
(0.201) (0.369) (0.156) OLF 2
Out of labor force due to other reasons (1 = yes)
0.0784
Observations
Sample size
30,858
(0.269) Note (1) Data Resource: China Family Panel Studies (2012). The reported statistics are the sample mean with standard deviation in parentheses. (2) For trust-related variables, higher trust scores represent higher level of trust towards certain groups. Similarly, for satisfaction-related variables, higher satisfaction score represents higher level of satisfaction towards life or higher social status. (4) The sample size associated with variable “iv_f” with non-missing values is 5698, and the sample size for variable “iv_m” with non-missing values is 6504
trust.3 But when looking at the degrees of trust toward particular social groups, the trust score (0–10) varies significantly. The sample average trust score toward parents is 9.1, suggesting high level of trust among immediate family members in China. The average trust score is 6.4 for neighbors, 6.6 for medical doctors, 4.9 for government 3
According to the World Values Survey (WVS), China is among the high-trust societies in the cross-country comparison in terms of the degree of general trust (Delhey and Newton 2005; Delhey et al. 2011).
92
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
cadre, and 2.2 for strangers, which suggests that the degree of trust declines with the distance in social connection. For illustration purpose, the sample distribution of these trust-related variables is shown in the top portion of Fig. 4.2. The other set of outcome variables pertains to one’s satisfaction toward life and family. In this dimension, Table 4.1 shows that the sample respondents are fairly satisfied, giving the average scores of 3.3 and 3.5 for the satisfaction on their lives and families (out of the possible values of 1–5). They also rank their socioeconomic status as “intermediate” on average, with the scores of 2.7 and 2.8 (on a scale of 1–5) for the assessment of themselves and their families, respectively. The average respondents also feel fairly confident about their future with the average score being 3.7 out of 5, although the standard deviation is relatively large (1.1). The bottom half of Fig. 4.2 illustrates the sample distribution of the above life satisfaction variables.
4.4 Estimation Method 4.4.1 Baseline Regression Given the discrete and sequential nature of the dependent variables, we use the discrete choice models, specifically probit and ordered probit models, to evaluate the impact of mental depression on the individual’s degree of trust and life satisfaction. We use the binary variable trust_dummy to measure people’s general trust on others. The probit model is used to estimate this impact: Hi∗ = θ · cesdi + X i β + u i Pr(Hi = 0|cesd, X ) = Pr(Hi∗ ≤ ω|cesd, X ) = F(ω − θ · cesdi − X i β)
(4.2) (4.3)
Pr(Hi = 1|cesd, X ) = Pr(Hi∗ > ω|cesd, X ) = 1 − F(ω − θ · cesdi − X i β) (4.4) where Hi denotes the respondent’s answer of “whether most people are trustworthy”. In this model, we assume that Hi is determined by the continuous latent variable Hi∗ that represents the respondent’s unobserved tendency of trust. cesdi represents the mental health status as measured by the CES-D score of respondent i, with its coefficient θ being the key parameter of interest for this study. X i is a vector of other individual characteristics such as age, gender, race, work status, etc., listed in Table 4.1. Variable Hi∗ holds linear relationship with the explanatory variables (cesd and X), the realization of Hi depends on the region in which Hi∗ falls (whether above or below the threshold ω), with the corresponding probability determined by F(.), the cumulative distribution function of the standard normal distribution. The
4.4 Estimation Method
Fig. 4.2 Sample distribution of dependent variables
93
94
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
maximum likelihood estimation will be conducted on the above probit model, which gives consistent estimates for θ and β. For the variables on particularistic trust that are ordinal in nature with more than two possible values (trust_parent, trust_neighbor, trust_doctor, trust_cadre, and trust_stranger), we use the ordered probit model, which takes the following form: Ti∗j = θ j · cesdi + X i β j + u i j
(4.5)
Pr(Ti j = 0|cesd, X ) = Pr(Ti∗j ≤ ω1 j |cesd, X ) = F(ω1 j − θ j · cesdi − X i β j ) (4.6) Pr(Ti j = t|cesd, X ) = Pr(ωt j ≤ Ti∗j ≤ ω(t+1) j |cesd, X ) t = 1, . . . 9 = F(ω(t+1) j − θ j · cesdi − X i β j ) − F(ωt j − θ j · cesdi − X i β j ) (4.7) Pr(Ti j = 10|cesd, X ) = Pr(Ti∗j > ω10 j |cesd, X ) = 1 − F(ω10 j − θ j · cesdi − X i β j ) (4.8)
where Ti j denotes the answer of question j of respondent i, ranging from 0 to 10 and taking on integer values. We assume Ti j is determined by the continuous latent variable Ti∗j that represents the respondent’s true trust level toward particular groups of people. Since Ti∗j holds linear relationship with the explanatory variables (cesd and X), the realization of Ti j depends on the intervals in which Ti∗j falls, with the corresponding probability determined by F(.), the cumulative distribution function of the standard normal distribution. The maximum likelihood estimation based on the above specification gives consistent estimates for θ j and β j . Similarly, for variables on life satisfaction (satis_family, ses_family, satis_self, ses_self, confi_self ), the model is specified as follows: ∗ = θk · cesdi + X i βk + u ik Sik
(4.9)
∗ Pr(Sik = 1|cesd, X ) = Pr(Sik ≤ ω1 j |cesd, X ) = F(ω1 j − θk · cesdi − X i βk ) (4.10) ∗ Pr(Sik = t|cesd, X ) = Pr(ω(t−1)k ≤ Sik ≤ ωtk |cesd, X ) t = 2, 3, 4 = F(ωtk − θk · cesdi − X i βk ) − F(ω(t−1)k − θk · cesdi − X i βk ) (4.11) ∗ Pr(Sik = 10|cesd, X ) = Pr(Sik > ω4k |cesd, X ) = 1 − F(ω4k − θk · cesdi − X i βk )
(4.12)
where Sik denotes the answer by respondent i for the life satisfaction-related question k, which ranges from 1 to 5 and takes on integer values. Using the maximum
4.4 Estimation Method
95
likelihood method with the standard normal distributional assumption on u ik , we can obtain the consistent estimates for θk and βk .
4.4.2 Instrumental Variable (IV) Regression The above models implicitly assume that an individual’s depression level is exogenous. However, the mental health status measure may suffer from endogeneity problem because of the following reasons: (1) unobserved factors such as lifestyles and ideology can affect both the degree of depression and one’s trust on others as well as life satisfaction; (2) higher degree of trust and life satisfaction can contribute to better interpersonal relationship, which in turn benefits one’s mental health. To address the above endogeneity concern due to variable omission or reverse causality, we use the CES-D scores of the respondent’s biological parents as the instrumental variables (IV), which include the CES-D scores of the respondent’s father (iv_f ) and mother (iv_m).4 The reasons why we choose these IVs are as follows. First, the parental CES-D scores should be directly correlated with the individual’s depression due to the heritability of depression. Prior studies such as McGue and Christensen (1997) estimate that such heritability ranges from 30 to 40%, which means that more than 30% of individuals with a family history of depression develop depression in their life. Second, parental CES-D scores should (arguably) not directly correlate with the individuals’ own attitude toward trust and life satisfaction without affecting the individual’s depression. For cross-validation purpose, we also follow the prior literature (such as Hsieh and Qin 2018) and use the community average CES-D scores as an alternative IV, where the calculation of the community average is based on all the sample observations in the 2012 CFPS dataset excluding the respondent himself/herself, and a community refers to an urban neighborhood or a rural village as defined by the CFPS site identifiers. The justification of this community-based instrumental variable is that medical literature provides strong evidence on the geographic clustering of mental health problems such as depression due to the contextual effect (Aneshensel and Sucoff 1996; Chaix et al. 2006), and that the area-based mental health measures should (arguably) not directly correlate with the individual’s social trust and life satisfaction without affecting the individual’s mental health status. After employing the IVs, the aforementioned probit and ordered probit model will be estimated using the two-stage maximum likelihood method as suggested by Wooldridge.5 Table 4.8 in the Appendix reports the first stage regressions and the statistical tests on the validity of each IV. The F-statistics associated with the Stock and Yogo tests on the statistical significance of the IVs are 393.1, 537.8, and 3960.6 for cesd_f , cesd_m, and cesd_comm, respectively, indicating strong correlation of 4
The IV regressions are based on the sample of respondents whose biological relatives’ information is available. 5 Wooldridge (2014).
96
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
CES-D scores between parents and their offspring, and between individual respondents and their community averages, which in turn suggests that the IVs are not likely to be weak.6
4.4.3 Subsample Regression To further explore the potential heterogeneity of the impacts of mental depression on social trust and life satisfaction, we also conduct the subsample analysis on different population groups in our data. The groups are divided by the following criteria: gender (male vs. female), age (young vs. middle age vs. elderly), marital status (married vs. unmarried), residential region (rural vs. urban), and education level (primary school or below vs. middle school vs. high school vs. college or above). For the age groups, “young” includes individuals aged between 16 and 39, “middle age” includes individuals aged between 40 and 59, and “elderly” includes individuals aged 60 or above. For the marital status, the “unmarried” group includes individuals who are single, divorced, or widowed. Similar to the baseline full sample regressions, the subsample regressions are based on the probit or ordered probit models, with the same set of control variables (the variable used as the group classification criterion is excluded).
4.5 Results Table 4.2 reports the results of the baseline regressions on a series of trust variables ranging from general trust to particularistic trusts in five different relationships. The key explanatory variable is CES-D score which measure the individual propensity to have depression. The results show that there is a statistically significant and negative relationship between the CES-D score and the trust variables, indicating that the more depressive individuals are less likely to trust other people. We also calculate the marginal effects of changes in CES-D, and the results show that one standard deviation increase of the CES-D score will decrease the probability of trusting other people in general by 5.7%. For particularistic trust, we find that the coefficient estimates of the CES-D score decrease with the distance of the relationship. For the inner circle such as the relationship with parents and neighbors, the coefficient estimates are in the range of 0.019 to 0.021. The estimated coefficients of CES-D decrease to about 0.015 for the trust on doctors and cadres, which represents the intermediate
6
For robustness check purpose, we also used all combination pair of the three IVs in the regressions and carried out the Sargan test for the over-identification restrictions for these over-identified IV models, and the comparatively high p-values for the Sargan tests suggest that these IVs are not likely to correlate with the error terms, providing supporting evidence on the validity of the IVs.
Unemployed
College
High
Middle
Urban
Married
Age2
Age
Female
cesd
Variables
−0.0326 −0.113*** −0.035
−0.0338
−0.0477
−0.0371
0.309***
−0.0236
−0.0241
0.650***
0.190***
0.380***
−0.018
−0.0186
−0.0161
−0.0165 0.0907***
0.0345**
−0.004
0.229***
−0.0224
−0.00297
−0.0087
−0.00287
−0.0103***
−0.0229
−0.00937***
−0.0152
−0.00689
−0.0148
−0.0265*
−3.02E−05
−0.00564
−0.000999
−3.15E−05
−0.000976
−0.0188***
2.19E−06
−0.0190***
Trust_dummy
0.000142***
(2) Parent
(1)
Table 4.2 Baseline regressions on the determinants of trust (3)
(4)
−0.0291
−0.106***
−0.0238
0.192***
−0.0186
0.0715***
−0.0149
0.0390***
−0.0132
−0.104***
−0.0185
−0.0593***
−2.59E−05
(5)
−0.029
−0.0501*
−0.03
−0.0693**
−0.024
−0.0432*
−0.144*** −0.023
−0.0183
−0.0949***
−0.0802*** −0.0184
−0.0149
−0.0149
−0.0132 −0.0466***
−0.0133
−0.205***
−0.0187
−0.0902***
−2.58E−05
0.000108***
−0.00243
−0.000685
−0.0121
0.0544***
−0.000861
−0.0158***
Cadre
−0.00854
−0.163***
−0.0186
−0.00894
−2.58E−05
−0.00242 2.42E−05
−0.00242
−0.00175
−0.0121
0.0445***
−0.000858
−0.0153***
Doctor
−3.54E−05
0.00771***
−0.0122
−0.0714***
−0.000868
−0.0209***
Neighbor
(6)
−0.0309
0.00193
−0.0261
0.472***
−0.0196
0.181***
−0.0156
(continued)
0.0629***
−0.0139
−0.0370***
−0.0189
−0.0587***
−2.65E−05
9.47e−05***
−0.00247
−0.00732***
−0.0127
−0.161***
−0.000862
−0.00455***
Stranger
4.5 Results 97
(3)
Yes 30,653
30,770
Yes
−0.000236 30,744
Yes
−0.000165
−0.0227 −0.000275*
−0.0227 −0.000116
−0.0248
−0.0381*
0.0626**
(5)
30,680
Yes
−0.000218
−0.000458**
−0.0223
−0.0516**
−0.025
0.169***
−0.0422
0.0598
Cadre
(6)
30,689
Yes
−0.000358
0.000617*
−0.024
0.0295
−0.0263
0.249***
−0.0439
0.04
Stranger
Note (1) Data Resource: China Family Panel Studies (2012). (2) All results are based on the probit or ordered probit regressions. The reported statistics are the coefficient estimates of the explanatory variables with the clustered robust standard errors shown in the parentheses. *, **, *** denote statistical significance at 10%, 5%, 1% levels, respectively
Yes
30,858
Observations
−0.000386
−0.000403
−0.0267 5.13E−05
−0.0278
0.000893**
0.0451*
0.0927***
−0.0314
−0.0331
−0.0419
0.112***
−0.0245
0.142***
−0.141***
0.209***
(4) Doctor
−0.0254
−0.0418
−0.0502
0.110***
0.153***
Neighbor
−0.0486
Trust_dummy
0.126***
(2) Parent
(1)
Province
Pincome
Retired
OLF_2
OLF_1
Variables
Table 4.2 (continued)
98 4 The Hidden Costs of Mental Depression: Implications on Social Trust …
4.5 Results
99
range in an individual’s social connection. For people in the outer relationship circles such as strangers, the coefficient estimate is very small (about 0.004). With regard to other control variables, the results also yield several interesting findings. First, we find that income and education are positively correlated with the general trust. This is consistent with the findings by Alesina and Ferrara (2002) and suggests that people in higher socioeconomic status are more prone to trusting other people. A plausible explanation for this result is that education is an important channel to create individual social capital by raising social skills. However, the effect of education on the particularistic trust is not uniform across specific individuals: bettereducated people are more likely to trust parents, neighbors, and strangers, but they are less likely to trust doctors and cadres. A plausible explanation for this discrepancy is that the high education groups in China may be better informed about the corruption scandals (such as bribery-taking behaviors) of the doctors and government officials, and thus they are more likely to hold a prejudice against these professionals. Second, we find that both gender and age have significant effects on general trust and particularistic trust, which are also consistent with the findings in previous studies (Glaeser et al. 2000; Alesina and Ferrara 2002). Women are less likely trusting, which may be because women participate less in social activities and are more likely to encounter the experience of social discrimination. Our estimates show that the relationship between age and trust is a U shape, with the turning point at age 36 (calculated from column 1), indicating that older cohorts are more like to trust other people after they reach the middle age. This suggests that trust is a learning process. Third, we found that some socioeconomic variables, such as marital status and residential place, do not have significant impacts on general trust, but they have heterogeneous effects on particularistic trust. For example, compared to rural residents, urban residents are more likely to trust their parents, but are less likely to build a trusting relationship with other non-family member, including neighbors, doctors, cadres, and strangers. As China experienced rapid urbanization in recent years, a higher proportion of urban residents are newcomers in town. Given the previous evidence that city size has a negative effect on the amount of trust (Glaeser 2000), the geographic mobility may put urban residents in disadvantage in forming a trusting relationship with other non-family members. Fourth, compared to people who are currently employed, those not in the labor force are more likely to have a general trust on other people. However, they may have less particularistic trust on specific individuals. For example, the unemployed are less likely to trust parents, neighbors, doctors, and cadres. Table 4.3 reports the results of the baseline regressions on a series of life satisfaction variables ranging from the satisfaction of one’s own life and one’s family to the degree of confidence to one’s future. The estimated coefficients reveal that the CES-D score has a significantly negative impact on life satisfaction, indicating that a high propensity for mental depression is associated with poorer subjective well-being. This finding is consistent with the previous finding that mental illness is a key determinant of unhappiness (Layard 2013), and it provides strong evidence for the nexus between health and happiness. More precisely, we find that higher CES-D scores lead to lower satisfaction on one’s family and one’s life as well as a
100
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
Table 4.3 Baseline regressions on the determinants of life satisfaction Variables
(1)
(2)
(3)
(4)
(5)
satis_family
ses_family
satis_self
ses_self
confi_self
−0.0426***
−0.0248***
−0.0437***
−0.0215***
−0.0443***
−0.000929
−0.000936
−0.000934
−0.000918
−0.000951
Female
0.134***
0.103***
0.165***
0.0591***
0.0465***
−0.0127
−0.0128
−0.0126
−0.0128
−0.0128
Age
−0.0208***
−0.0106***
−0.0242***
−0.00419
−0.0290***
−0.00256
−0.00261
−0.00259
−0.00259
−0.00256
0.000275***
0.000165***
0.000335***
0.000127***
0.000193***
−2.73E−05
−2.81E−05
−2.77E−05
−2.78E−05
−2.73E−05
Married
0.0847***
0.114***
0.0945***
0.106***
0.147***
−0.0197
−0.02
−0.0196
−0.0197
−0.0196
Urban
0.0296**
−0.131***
−0.0186
−0.168***
−0.0296**
−0.0137
−0.0138
−0.0137
−0.0139
−0.0139
−0.00694
0.0796***
−0.0132
0.0112
0.0677***
−0.0155
−0.0157
−0.0156
−0.0157
−0.0158
High
−0.0276
0.101***
−0.0456**
0.0506***
0.0529***
−0.0195
−0.0193
−0.0191
−0.0195
−0.0193
College
0.0158
0.185***
0.0431*
0.193***
0.0205
−0.0249
−0.0239
−0.0245
−0.0248
−0.0245
0.0255
−0.0141
0.0309
0.00765
−0.0653**
−0.0321
−0.0323
−0.0324
−0.032
−0.0321
OLF_1
−0.0743*
−0.150***
−0.107**
−0.195***
−0.208***
−0.045
−0.0465
−0.0454
−0.0461
−0.0456
OLF_2
0.189***
0.173***
0.266***
0.208***
0.0456*
−0.0263
−0.0257
−0.0262
−0.0263
−0.0264
0.0106
0.0209
0.0325
0.00703
−0.00216
−0.0234
−0.0247
−0.0237
−0.0245
−0.0238
Pincome
0.000753***
0.000795***
0.000555***
0.000634
0.000968***
−0.000286
−0.00027
−0.000199
−0.000414
−0.000307
Province
Yes
Yes
Yes
Yes
Yes
Observations
30,791
30,676
30,793
30,623
30,678
cesd
Age2
Middle
Unemployed
Retired
Note (1) Data Resource: China Family Panel Studies (2012). (2) All results are based on the Ordered Probit regressions. The reported statistics are the coefficient estimates of the explanatory variables with the clustered robust standard errors shown in the parentheses. *, **, *** denote statistical significance at 10%, 5%, 1% levels, respectively.
4.5 Results
101
lower confidence to one’s future. An interesting finding is that the marginal impact of CES-D scores on these three variables is almost identical, indicating that the negative impact of depression on life satisfaction is robust across different empirical measurements. In addition, we find that the CES-D score is negatively associated with the perception of social status (compared to other people in the same local areas) for both the respondents per se and their families. The marginal effects of CES-D on these two variables are also similar, thus providing cross-validation on the findings of the two outcome variables. A comparison between the two sets of life satisfaction measures suggests that the impacts of mental depression are larger for self-perceived confidence [Column (1), (3), and (5)] than for socioeconomic status [Column (2) and (4)], as the latter is more objectively measured. With regard to other control variables, the results are more homogenous across alternative measures of life satisfaction as compared to their estimated impacts on the trust variables. Specifically, we find that females have a higher life satisfaction than males, which can be explained by two plausible reasons: first, females are in an advantageous position to keep a good social network; second, women face a lower social pressure than men, especially in the labor market activities. Age has a nonlinear effect on life satisfaction, indicating a U-shaped relationship between age and subjective well-being: people tend to feel less satisfied about themselves and their families as they get older, but these perceptions start to improve after a certain age range, suggesting that the middle-aged individuals are more likely to have a lower life satisfaction as compared to the young and elderly adults. Not surprisingly, we also find that marriage is associated with higher life satisfaction. For the socioeconomic variables, we find income has a significantly positive effect on several indicators of life satisfaction, indicating that income still plays an important role in shaping individuals subjective well-being, especially with the backdrop of rapid income growth in China for the past three decades. In addition, we find education has a nonlinear effect on life satisfaction: compared to individuals with primary school education or below, secondary education (middle and high school) does not seem to make people more satisfied with their live, but higher education (college or above) does. The work and residential status also have certain impacts on life satisfaction, but the effects are heterogeneous across different variables. Specifically, individuals who are out of the labor force due to disability or diseases have lower life satisfaction compared to the currently employed. By contrast, individuals who are out of labor force due to other reasons (such as schooling) have higher life satisfaction. People living in the urban areas are in general more satisfied with their family as compared to rural residents. However, urban residents have a lower perception on their social status in the local area and show a low confidence about their future life. This may reflect the fact that urban residents face a higher competitive pressure for survival than the rural residents. Table 4.4 reports the IV regression results on the trust variables by taking into account of the endogeneity of CES-D score. Column (1) summarizes the coefficient estimates of cesd in the aforementioned mentioned baseline regressions (Table 4.2). In Column (2)–(4), we report the IV-based regression estimates of cesd using father’s CES-D score, mother’s CES-D score, and the community average CES-D score
102
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
Table 4.4 IV Regressions on the impact of CES-D on trust Dep. Variable
(1)
(2)
(3)
(4)
Baseline
iv_f
iv_m
iv_comm
−0.0188***
−0.0182***
−0.0184***
−0.0156***
(0.000999)
(0.00106)
(0.00107)
(0.00292)
trust_parent
−0.0190***
−0.0186***
−0.0189***
−0.0310***
(0.000976)
(0.00101)
(0.00102)
(0.00279)
trust_neighbor
−0.0209***
−0.0207***
−0.0206***
−0.0147***
(0.000868)
(0.000865)
(0.000874)
(0.00237)
−0.0153***
−0.0145***
−0.0144***
−0.00957***
(0.000858)
(0.000862)
(0.000873)
(0.00236)
trust_cadre
−0.0158***
−0.0149***
−0.0148***
−0.00278
(0.000861)
(0.000858)
(0.000868)
(0.00236)
trust_stranger
−0.00455***
−0.00443***
−0.00438***
−0.000350
(0.000862)
(0.000894)
(0.000905)
(0.00243)
trust_dummy
trust_doctor
Note (1) The reported results are based on the IV ordered probit model, which is implemented by the two-stage maximum likelihood estimation. (2) The reported statistics are the coefficient estimates of the explanatory variables with the clustered robust standard errors shown in the parentheses. *, **, *** denote statistical significance at 10%, 5%, 1% levels, respectively. (3) Column iv_f and iv_m respectively uses the CES-D scores of the respondent’s father and mother as the instrumental variable. Similarly, column iv_comm uses the community average CES-D score (excluding the respondent him/herself) as the instrument variable
as instrumental variables, respectively. These IV regression results show that the impact of CES-D score on all trust variables is similar to the baseline regression results, indicating that our basic results are robust to the control of endogeneity in the regressions. More precisely, the estimated results show a consistent pattern that an increase in the propensity of depression leads to a reduction in both the general trust and particularistic trust, indicating that depression causes a lower trust not just for the general people but also for specific individuals. The only exceptions are the impact on the trust of government cadres and strangers in Column (4), where the respondent’s CES-D score shows a quantitatively small and statistically insignificant influence. This in turn suggests that mental health status may present a weaker effect on the outer-group trust compared to inner-group trust, a finding consistent with the baseline regression results. Following the same identification strategy, Table 4.5 reports the IV regression results on a series of variables on life satisfaction that take account of the endogeneity of the CES-D score. The results are similar to those reported in the baseline regression models Table 4.3. In addition, the IV results are quite consistent across alternative instrumental variables. These results indicate that depression leads to a lower life satisfaction and they are robust across different specifications of instrumental variables. To address the potential concern that the IV regression results and the baseline regression results may not be directly comparable due to their different
4.5 Results
103
Table 4.5 IV Regressions on the impact of CES-D on life satisfaction Dep. Variable
(1)
(2)
(3)
(4)
baseline
iv_f
iv_m
iv_comm
−0.0426***
−0.0443***
−0.0448***
−0.0423***
−0.000929
−0.000905
−0.000915
−0.00245
ses_family
−0.0248***
−0.0258***
−0.0261***
−0.0247***
−0.000936
−0.000917
−0.000928
−0.00247
satis_self
−0.0437***
−0.0449***
−0.0447***
−0.0381***
−0.000934
−0.00091
−0.000922
−0.00247
−0.0215***
−0.0219***
−0.0221***
−0.0246***
−0.000918
−0.000908
−0.000919
−0.00245
−0.0443***
−0.0451***
−0.0452***
−0.0337***
−0.000951
−0.000918
−0.00093
−0.00252
satis_family
ses_self confi_self
Note (1) The reported results are based on the IV Ordered Probit model, which is implemented by the 2-stage maximum likelihood estimation. (2) The reported statistics are the coefficient estimates of the explanatory variables with the clustered robust standard errors shown in the parentheses. *, **, *** denote statistical significance at 10%, 5%, 1% levels, respectively. (3) Column iv_f and iv_m respectively uses the CES-D scores of the respondent’s father and mother as the instrumental variable. Similarly, column iv_comm uses the community average CES-D score (excluding the respondent him/herself) as the instrument variable
sample sizes, we also conduct the baseline regressions using the smaller samples that are identical to the IV-based regressions, and these results are reported in Appendix Table 4.9. It suggests that these baseline regressions based on smaller samples still give consistent results as the baseline regressions based on the full sample. Tables 4.6 and 4.7 report the baseline regression results based on the subsample analyses, which provide the basis to investigate whether the estimated coefficients of CES-D score on the trust and life satisfaction variables are heterogeneous across different population groups. The results show that the impacts of CES-D score on various trust variables are not uniform across subpopulations. More precisely, the negative impact of depression on trust is larger for males than that for females and this pattern is consistent across all alternative measures of trust, indicating that the hidden cost of depression in the form of lower trust to other people or to specific individuals are higher for men than for women. Similarly, the magnitude of the detrimental effect of CES-D score on trust also varies across age groups. Depression in general has a stronger impact on the young group (age 16 to 40) than the elderly (age 60 and above). With regard to regional variation, the results show that mental depression has a stronger effect on the propensity to trust other people among urban residents compared to rural residents, and this pattern is consistent for both the general trust and the particularistic trust. However, the disparity in the estimated coefficients seems small between the married and the unmarried groups. We also find the existence of an education gradient in the estimated impact of depression propensity. Specifically, people with college degree or above tend to
Education
Region
Marriage
Age
Gender
Primary school or below
Urban
Rural
Married
Unmarried
Elderly
Middle age
Young
Female
Male
Variable
−0.0254*** −0.00148
−0.0212*** −0.00141 −0.0176*** −0.00125 −0.0256***
−0.0200*** −0.0015 −0.0177*** −0.00132 −0.0235***
(3)
Trust dummy
−0.0223*** −0.00174
−0.00141 −0.0171*** −0.00177 −0.0190***
−0.00149 −0.0153*** −0.00192 −0.0183***
−0.00147
−0.00161
−0.00122
−0.00147
−0.0205***
−0.0275*** −0.0160***
−0.00123
−0.00131 −0.00154
−0.0179***
−0.0135***
−0.0109***
−0.0239***
−0.00105
−0.00112
−0.0013
−0.0146***
−0.00106
−0.0190***
−0.000901
−0.0203***
−0.00205 −0.0191***
−0.00216 −0.0188***
−0.00157
−0.0182***
−0.0012
−0.0195***
−0.0018 −0.0175***
−0.00185 −0.0179***
−0.00107
−0.0184***
−0.00121
−0.0240***
Trust neighbor
(2) Trust parent
(1)
Table 4.6 Subsample regressions on the impact of CES-D on trust (4)
−0.0013
−0.0105***
−0.00122
−0.0166***
−0.00106
−0.0147***
−0.000898
−0.0155***
−0.00173
−0.0146***
−0.00156
−0.0123***
−0.0012
−0.0144***
−0.00147
−0.0202***
−0.00106
−0.0117***
−0.0012
−0.0201***
trust doctor
(5)
−0.0013
−0.0105***
−0.00122
−0.0195***
−0.00106
−0.0145***
−0.000898
−0.0159***
−0.00173
−0.0151***
−0.00157
−0.0130***
−0.0012
−0.0154***
−0.00147
−0.0185***
−0.00106
−0.0135***
−0.0012
−0.0188***
Trust cadre
(6)
(continued)
−0.00135
−0.00139
−0.00127
−0.00853***
−0.0011
−0.00288***
−0.000934
−0.00459***
−0.00179
−0.00376**
−0.00163
−0.00189
−0.00125
−0.00362***
−0.00152
−0.00776***
−0.00111
−0.00423***
−0.00124
−0.00501***
trust stranger
104 4 The Hidden Costs of Mental Depression: Implications on Social Trust …
(2) Trust parent −0.0191*** −0.00207 −0.0183*** −0.00196 −0.0236*** −0.0029
(1) Trust dummy −0.0213*** −0.00219 −0.0223*** −0.00204 −0.0312*** −0.00297
(3)
−0.00237
−0.0320***
−0.00165
−0.0247***
−0.00176
−0.0215***
Trust neighbor
(4)
−0.00234
−0.0200***
−0.00163
−0.0179***
−0.00175
−0.0152***
trust doctor
(5)
−0.00236
−0.0241***
−0.00164
−0.0193***
−0.00175
−0.0158***
Trust cadre
(6)
−0.00243
−0.0116***
−0.0017
−0.00839***
−0.00183
−0.00330*
trust stranger
Note (1) The reported results are based on the probit or ordered probit regressions for each subsample. (2) The reported statistics are the coefficient estimates of the explanatory variables with the clustered robust standard errors shown in the parentheses. *, **, *** denote statistical significance at 10%, 5%, 1% levels, respectively. (3) Group young includes individuals aged between 16 and 40; group middle age includes individuals aged between 40 and 60 (including 40); group elderly includes individuals aged above 60 (including 60); group unmarried includes individuals who are single, divorced, or widowed
College or above
High school
Middle school
Variable
Table 4.6 (continued)
4.5 Results 105
106
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
Table 4.7 Subsample regressions on the impact of CES-D on life satisfaction
Gender
Variable
(1) satis_ family
Male
−0.0447*** −0.0273*** −0.0460*** −0.0255*** −0.0471***
Female
−0.0409*** −0.0229*** −0.0418*** −0.0185*** −0.0422***
Young
−0.0440*** −0.0268*** −0.0475*** −0.0216*** −0.0427***
(0.00127) (0.00113) Age
(0.00156)
(0.00112) (0.00157)
(0.00128) (0.00113) (0.00156)
(0.00127) (0.00112) (0.00156)
(5) confi_self (0.00129) (0.00114) (0.00158)
−0.0427*** −0.0220*** −0.0413*** −0.0216*** −0.0448***
Elderly
−0.0392*** −0.0254*** −0.0418*** −0.0192*** −0.0446***
(0.00127)
(0.00127) (0.00165)
(0.00127) (0.00166)
(0.00126) (0.00164)
(0.00128) (0.00167)
Unmarried −0.0392*** −0.0241*** −0.0449*** −0.0215*** −0.0460*** (0.00183)
(0.00184)
(0.00185)
(0.00183)
(0.00186)
Married
−0.0434*** −0.0249*** −0.0434*** −0.0216*** −0.0440***
Rural
−0.0393*** −0.0227*** −0.0400*** −0.0203*** −0.0396***
Urban
−0.0474*** −0.0274*** −0.0491*** −0.0233*** −0.0511***
(0.000951) Region
(0.00127)
(3) satis_self (4) ses_self
Middle age
(0.00165) Marriage
(2) ses_ family
(0.00112) (0.00130) Education Primary school or below
(0.000950) (0.00112) (0.00130)
(0.000952) (0.00112) (0.00131)
(0.000944) (0.00111) (0.00129)
(0.000959) (0.00113) (0.00132)
−0.0388*** −0.0210*** −0.0389*** −0.0172*** −0.0369*** (0.00137)
(0.00136)
(0.00137)
(0.00136)
(0.00138)
Middle school
−0.0411*** −0.0256*** −0.0432*** −0.0221*** −0.0440***
High school
−0.0458*** −0.0275*** −0.0470*** −0.0249*** −0.0478***
College or above
−0.0482*** −0.0281*** −0.0493*** −0.0281*** −0.0527***
(0.00185) (0.00174) (0.00251)
(0.00185) (0.00175) (0.00251)
(0.00186) (0.00174) (0.00252)
(0.00184) (0.00173) (0.00249)
(0.00187) (0.00176) (0.00254)
Note (1) The reported results are based on the ordered probit regressions for each subsample. (2) The reported statistics are the coefficient estimates of the explanatory variables with the clustered robust standard errors shown in the parentheses. *, **, *** denote statistical significance at 10%, 5%, 1% levels, respectively. (3) Group young includes individuals aged between 16 and 40; group middle age includes individuals aged between 40 and 60 (including 40); group elderly includes individuals aged above 60 (including 60); group unmarried includes individuals who are single, divorced, or widowed
4.6 Conclusions
107
decrease their trust to other people in a more substantial way than less educated people when their CES-D scores are high. This education gradient is consistent across alternative measures of trust, indicating that a higher cost of depression in the form of lowering trust (both general trust and particularistic trust) occurs to the individuals with better educational attainment. The results reported in Table 4.7 also show a clear pattern on the heterogeneous impacts of mental depression on life satisfaction across different subpopulation groups. An interesting finding is that the life satisfaction regressions demonstrate a similar pattern of the heterogeneous effects as observed in the trust-related regressions. More specifically, males and the younger-aged groups face a higher hidden cost of depression in terms of reduced life satisfaction, as compared to females and the older-aged groups, respectively. Similarly, compared to the rural residents, people living in the urban areas feel less satisfied with their life and family when they suffer from depression or depressive symptom. The education gradient also exists in the negative impact of CES-D score on life satisfaction: better-educated people tend to suffer more from this negative impact than people with lower education levels.
4.6 Conclusions This paper contributes to the growing body of literature in three lines of research: rising prevalence of depression as a source of increasing global disease burden, trust as a rooted factor to promote economic growth, and life satisfaction as a measure of subjective well-being. Putting together, we highlight a conceptual framework that the rising prevalence of depression may impose two hidden costs on individuals and the society as a whole in that depression causes a reduction in trust and life satisfaction. Based on the data obtained from 2012 CFPS, our study provides evidence on the existence of these hidden costs with three major empirical findings: First, the estimated results show that the propensity of depression measured by the CES-D score has significant and negative impact on a series of trust variables, including the empirical measures on general trust and particularistic trust toward specific groups ranging from parents, neighbors, doctors, cadre, and strangers. Similarly, the CES-D score also has a significant negative impact on a series of variables measuring life satisfaction, including the satisfaction of one’s family and one’s life, self-evaluation of own and family’s social status, and the degree of confidence to one’s future. Second, these negative impacts are not just a correlation, but can also be interpreted as a causal relationship. By employing the instrumental variable regressions, our study finds that the negative impact still holds and remains statistically significant after considering the endogeneity of CES-D score. Third, the estimated coefficients of depression on trust and life satisfaction are not uniform across subpopulation groups. More precisely, we find that the detrimental effects of mental depression are stronger among the following four subgroups: males, young-aged people, urban residents, and the well-educated individuals.
108
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
Given that trust has been characterized as one of the components of social capital, which in turn plays an important role to foster economic growth in general and innovation in particular, the decrease in trust caused by the rising prevalence of depression globally has an important consequence on the wellbeing for both individuals and the society as a whole. Similarly, life satisfaction has been recognized as an important component to measure the subjective well-being. A decrease in life satisfaction caused by the rising prevalence of depression also imposes a significant cost to individuals as well as to the society. An important implication of our study is that the burden of mental health conditions is not limited to their direct health consequences, but the impact on social and economic well-being is also substantial. As a result, the long-term costs of mental health problems and the value of investment in mental health resources will need to be reassessed when designing the mental health policies, particularly in the fast-growing developing countries like China.
Appendix See Tables 4.8 and 4.9. Table 4.8 Regression: first-stage results (Dep. Var. = cesd) cesd_f
(1)
(2)
(3)
iv_f
iv_m
iv_comm
0.224*** (0.0128)
cesd_m
0.219*** (0.0105)
cesd_comm
0.839*** (0.0139)
Female Age
1.428***
1.494***
(0.184)
(0.170)
2.034*** (0.0832)
0.183***
0.112*
0.136***
(0.0684)
(0.0591)
(0.0172)
−0.00224**
−0.00135
−0.000917***
(0.000983)
(0.000823)
(0.000187)
−1.159***
−1.085***
−1.974***
(0.234)
(0.225)
(0.136)
Urban
−0.113
−0.263
0.154*
(0.183)
(0.170)
(0.0904)
Middle
−1.141***
−1.258***
−1.270***
(0.227)
(0.209)
(0.0990)
Age2 Married
(continued)
Appendix
109
Table 4.8 (continued)
(1)
(2)
(3)
iv_f
iv_m
iv_comm
High
−1.652***
−1.799***
−1.411***
(0.261)
(0.241)
(0.126)
College
−1.960***
−1.922***
−1.662***
(0.298)
(0.283)
(0.168)
0.597
0.713*
0.360*
(0.395)
(0.378)
(0.189)
OLF_1
7.928***
10.00***
5.744***
(1.536)
(1.377)
(0.351)
OLF_2
−0.703***
−0.794***
−1.207***
(0.264)
(0.240)
(0.163)
Retired
0.173
0.0691
−0.181
(1.343)
(1.379)
(0.162)
0.00174
0.000264
−0.00469*
(0.00433)
(0.00184)
(0.00274)
Yes
Yes
Yes
F-statistics
393.08
537.84
3960.6
Observations
5698
6504
30,858
Unemployed
Pincome Province
Note (1) The reported statistics are the coefficient estimates of the explanatory variables (cesd) in the first stage maximum likelihood estimation associated with the IV-based models. Clustered robust standard errors are n in the parentheses. *, **, *** denote statistical significance at 10%, 5%, 1% levels, respectively. (2) Column iv_f and iv_m respectively uses the CES-D scores of the respondent’s father and mother as the instrumental variable. Similarly, column iv_comm uses the community average CES-D score (excluding the respondent him/herself) as the instrument variable. (3) The reported F-statistics are associated with the Stock and Yogo tests on the statistical significance of the IVs in the first-stage regressions
Table 4.9 Baseline regression results on the IV-based samples
Dep. Variable
(1)
(2)
(3)
Full sample
iv_f sample
iv_m sample
−0.0188***
−0.0218***
−0.0221***
−0.000999
−0.00274
−0.00253
−0.0190***
−0.0264***
−0.0235***
−0.000976
−0.00272
−0.00252
trust_neighbor
−0.0209***
−0.0248***
−0.0246***
−0.000868
−0.00236
−0.00215
trust_doctor
−0.0153***
−0.0219***
−0.0211***
trust_dummy trust_parent
(continued)
110
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
Table 4.9 (continued)
Dep. Variable
(1)
(2)
(3)
Full sample
iv_f sample
iv_m sample
−0.000858
−0.00239
−0.00216
−0.0158***
−0.0227***
−0.0213***
−0.000861
−0.00239
−0.00218
trust_stranger
−0.00455***
−0.00584**
−0.00550***
−0.000862
−0.00228
−0.0021
satis_family
−0.0426***
−0.0372***
−0.0377***
−0.000929
−0.00254
−0.00235
−0.0248***
−0.0249***
−0.0237***
−0.000936
−0.00246
−0.00234
−0.0437***
−0.0423***
−0.0445***
−0.000934
−0.00253
−0.00239
ses_self
−0.0215***
−0.0221***
−0.0221***
−0.000918
−0.00253
−0.00235
confi_self
−0.0443***
−0.0427***
−0.0448***
−0.000951
−0.00267
−0.0024
30,858
5698
6504
trust_cadre
ses_family satis_self
Obs
Note (1) All results are based on the probit or ordered probit regressions. The reported statistics are the coefficient estimates of the main explanatory variable (cesd) in the baseline regressions with alternative samples. Clustered robust standard errors are shown in the parentheses. *, **, *** denote statistical significance at 10%, 5%, 1% levels, respectively. (2) Column “full sample” represent the baseline regression estimates associated with the full study sample, column “iv_f sample” and “iv_m sample” represent the baseline regression estimates associated with the samples used for IV-based regressions with iv_f and iv_m as instrumental variables, respectively. It does not display “iv_comm”, as the IV regressions using “iv_comm” are based on the same sample as the baseline regressions (with a sample size of 30,858)
References Aghion P, Algan Y, Cahuc P, Shleifer A (2010) Regulation and distrust. Q J Econ 125(3):1015–1049 Alesina A, Giuliano P (2015) Culture and institutions. IZA Discussion Paper No. 9246 Alesina A, La Ferrara E (2002) Who trusts others? J Public Econ 85(2):207–234 Algan Y, Cahuc P (2014) Trust, growth, and well-being: New evidence and policy implications. In: Handbook of economic growth, vol 2. Elsevier, pp 49–120 Aneshensel CS, Sucoff CA (1996) The neighborhood context of adolescent mental health. J Health Soc Behav 37(4):293 Bharadwaj P, Pai MM, Suziedelyte A (2015) Mental health stigma. National Bureau of Economic Research Working Paper 21240, Cambridge, MA Bhargava A, Jamison DT, Lau LJ, Murray CJ (2001) Modeling the effects of health on economic growth. J Health Econ 20(3):423–440
References
111
Bloom DE, Canning D, Malaney PN (2000) Population dynamics and economic growth in Asia. Popul Dev Rev 26(supplement):257–290 Bloom DE, Canning D, Graham B (2003) Longevity and life-cycle savings. Scand J Econ 105(3):319–338 Bloom DE, Canning D, Sevilla J (2004) The effect of health on economic growth: a production function approach. World Dev 32(1):1–13 Chaix B et al (2006) Spatial clustering of mental disorders and associated characteristics of the neighbourhood context in Malmö, Sweden, in 2001. J Epidemiol Community Health 60(5):427– 435 Deaton A (2008) Income, health, and well-being around the world: evidence from the Gallup World Poll. J Econom Perspect 22(2):53–72 Delhey J, Newton K (2005) Predicting cross-national levels of social trust: global pattern or Nordic exceptionalism? Eur Sociol Rev 21(4):311–327 Delhey J, Newton K, Welzel C (2011) How general is trust in “most people”? Solving the radius of trust problem. Am Sociol Rev 76(5):786–807 Falk A, Kosfeld M (2006) The hidden costs of control. Am Econ Rev 96(5):1611–1630 Frank RG (2011) Economics and mental health: an international perspective. In: Glied S, Smith PC (eds) The Oxford handbook of health economics. Oxford University Press, pp 232–256 Frank RG, McGuire TG (2000) Economics and mental health. Handb Health Econ 1:893–954 Frey BS, Stutzer A (2002) What can economists learn from happiness research? J Econ Lit 40(2):402–435 Glaeser EL, Laibson DI, Scheinkman JA, Soutter CL (2000) Measuring trust. Q J Econ 115(3):811– 846 Greenberg PE, Fournier AA, Sisitsky T, Pike CT, Kessler RC (2015a) The economic burden of adults with major depressive disorder in the United States (2005 and 2010). J Clin Psychiatry 76(2):155–162 Greenberg PE, Fournier A-A, Sisitsky T et al (2015b) The economic burden of adults with major depressive disorder in the United States (2005 and 2010). J Clin Psychiatry 76(2):155–162 Hsieh CR, Liu S, Qin X (2019) The hidden costs of mental depression: implications on social trust and life satisfaction. Manchester School 87(2):259–296 Hsieh C, Qin XZ (2018) Depression hurts, depression costs: the medical spending attributable to depression and depressive symptoms in China. Health Econ 27(3):525–544 Huang J, Wu S, Deng S (2016) Relative income, relative assets, and happiness in urban China. Soc Indic Res 126(3):971–985 Kahneman D, Krueger AB (2006) Developments in the measurement of subjective well-being. J Econ Perspect 20(1):3–24 Kapteyn A, Smith JP, van Soest A (2009) Comparing life satisfaction. In: Working paper WR623-1, RAND population research center Knack S, Keefer P (1997) Does social capital have an economic payoff? A cross-country investigation. Quart J Econ 112:1251–1288 Layard R (2013) Mental health: the new frontier for labor economics. IZA J Labor Pol 2(2):1–16 Luo J-D (2005) Particularistic trust and general trust: a network analysis in Chinese organizations. Manag Organ Rev 1(3):437–458 McGue M, Christensen K (1997) Genetic and environmental contributions to depression symptomatology: evidence from Danish twins 75 years of age and older. J Abnorm Psychol 106(3):439 Olfson M, Blanco C, Marcus SC (2016) Treatment of adult depression in the United States. JAMA Intern Med 176(10):1482–1491 Oswald AJ, Wu S (2011) Well-being across America. Rev Econom Statist 93(6):1118–1134 Radloff LS (1977) The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol 1(3):385–401 Schnitzlein DD, Wunder C (2016) Are we architects of ouw own happiness? The importance of family background for well-being. BE J Econ Anal Pol 16(1):125–149
112
4 The Hidden Costs of Mental Depression: Implications on Social Trust …
Soares RR (2005) Mortality reductions, educational attainment, and fertility choice. Am Econ Rev 95(3):580–601 Spolaore E, Wacziarg R (2013) How deep are the roots of economic development? J Econ Lit 51(2):1–45 Weil DN (2014) Health and economic growth. Handbook Econom Growth 2B:623–682 Well DN (2007) Accounting for the effect of health on economic growth. Q J Econ 122(3):1265– 1306 WHO (2016) Mental disorders: fact sheet. [Accessed on 2017 January 2017]; Available from: http:/ /www.who.int/mediacentre/factsheets/fs396/en/ Wooldridge JM (2014) Quasi-maximum likelihood estimation and testing for nonlinear models with endogenous explanatory variables. J Econ 182(1):226–234 Yang G, Wang Y, Zeng Y et al (2013a) Rapid health transition in China, 1990–2010: findings from the global burden of disease study 2010. Lancet 381(9882):1987–2015 Yang G, Wang Y, Zeng YX et al (2013b) Rapid health transition in China, 1990–2010: findings from the global burden of disease study 2010. Lancet 381:1987–2015
Chapter 5
Relative Economic Status and the Mental Health Status Among Chinese Adults: Evidence from the China Health and Retirement Longitudinal Study
5.1 Introduction Since its market-oriented reforms in the early 1980s, the Chinese economy has been growing rapidly for more than three decades. The GDP per capita increased from 382 RMB in 1978 to 46,629 in 2014. Despite the recent economic slowdown, China’s percapita income growth has consistently been above 8% since the early 1980s, which is unprecedented in human history. However, despite the rapid growth of GDP per capita, the happiness index of the Chinese people did not increase at the same rate. In fact, the proportion of people who think they are “very happy” even decreased from 28% in 1990 to 12% in 2000 (Brockmann et al. 2009). Meanwhile, the prevalence rate of mental depression has increased rapidly in recent decades, reaching 4.1% for all Chinese adults in 2012 (Qin et al. 2016a, b). This suggests that the relationship between economic well-being and mental well-being can be very complicated, and this paper aims to study this relationship using China as an example. Mental disorders (such as depression) are important yet often neglected noncommunicable diseases and have gradually become a challenging public health issue in China (Phillips 2004; Hsieh and Qin 2018). According to Phillips et al. (2009), the one-month adjusted prevalence of any mental disorder is 17.5% in China. Meanwhile, mental illness is a strong risk factor for disabilities and suicide (Andrews and Titov 2007; Chesney et al. 2014); approximately half of the suicide victims had one or more mental illness (Zhang et al. 2010), and nearly eight million adults suffered psychiatric disability caused by mental disorders (Li et al. 2011). According to WHO (2012), mental health condition has become a major contributor to disease burden in China. Hence, it bears important academic value and policy relevance to explore these discrepancies by identifying factors that contribute to the increasing prevalence of mental health conditions despite the economic growth, given China accounts for The content of this chapter is published in Zhou et al. (2020). Copyright John Wiley and Sons (2020), reproduction license granted. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 X. Qin and C.-R. Hsieh, Economic Analysis of Mental Health in China, Applied Economics and Policy Studies, https://doi.org/10.1007/978-981-99-4209-1_5
113
114
5 Relative Economic Status and the Mental Health Status Among Chinese …
one-fifth of the world’s population and the welfare improvement of its population contributes greatly to the human development around the globe. The main focus of this paper is to explore the relationship between relative economic status and mental health among Chinese adults. Specifically, we use the Center for Epidemiologic Studies Depression Scale (CES-D) instrument to measure people’s tendency of suffering depressive symptoms, which has the advantage of directly reflecting people’s mental well-being. The CES-D is a widely used selfevaluation tool for the mental health condition, with reference values for clinical applications as well. We define “relative economic status” against different reference groups (i.e. relatives, colleagues, schoolmates, neighbors/villagers, and people in the same city/county), so that we can identify the actual socioeconomic comparisons that effectively impact people’s mental well-being. Our results strongly support the idea that people’s mental health conditions are significantly correlated with their relative economic status within certain social groups. In particular, relatives, colleagues, and schoolmates are more relevant reference groups, and the economic comparison against these groups generates larger influence on people’s mental health compared to people in the same geographic locations (such as people living in the same region). Furthermore, we find that the impacts of relative economic status are asymmetric in the sense that underperforming one’s reference group exerts a much more substantial effect on one’s mental health than outperforming the reference group. The subsample evidence further indicates that the effect of relative economic status is larger among the economically disadvantaged population such as the low-income group.
5.2 Literature Review The association between economic well-being and mental well-being (e.g. happiness, life satisfaction, and mental health) has long interested researchers. A variety of country-level studies have shown that the rich are happier than the poor within the same country, but raising income levels of all individuals does not increase the happiness of all (Easterlin 1974; Compbell et al. 1976; Frank 1990; Schor 1991). Easterlin (1995) suggests that the happiness of all may remain unchanged when the absolute income of all increases but the relative income remains constant. It reflects the importance of relative economic status for individuals in a society for their mental well-being. Some studies find micro-level evidences in support of the hypothesis that relative economic status does matter to individuals’ mental well-being in different countries, e.g. in the United Kingdom the United States (McBride 2001), European countries (Clark and Senik 2010), and in Asia (Zhang and Cai 2011; Oshio and Kobayashi 2011). For example, in Germany, Ferrer-i-Carbonell (2005) shows that individuals are happier when their incomes are higher than the reference income, where the reference income is defined as the average income of people with the same education level, at the same age, and residing in the same region. Dumludag (2013) evaluates the relative
5.2 Literature Review
115
impact of reference groups on life satisfaction in Turkey and finds that comparisons have significant impacts on life satisfaction. Knight et al. (2009) conclude that relative income within the village is important determinant for happiness of rural residents in China. Akay et al. (2012) evaluate the well-being of migrants in China, finding that mental well-being is negatively affected by the income of other migrants and workers of home regions. However, researchers have devoted less attention to examine the importance of economic well-being on mental well-being from the perspective of mental health (e.g. depression and suicides). Some studies have shown that low socioeconomic status (SES) is related to higher prevalence and incidence of depression (Lorant et al. 2003; Koster et al. 2006). Recently, Pickett and Wilkinson (2015) conduct a literature review to assess the causal relationship between economic inequality and health (including violence), and the evidence strongly suggests that economic inequality affects population health and well-being. Muramatsu (2003) examines whether county-level economic inequality (e.g. the Gini coefficient) is associated with depression among older Americans. This study shows that people living in counties with higher economic inequality are more depressed. Similar results are also obtained by Chiavegatto Filho et al. (2013) in Brazil. Zimmerman and Katon (2005) use the National Longitudinal Survey of Youth 1979 cohort to examine the robustness of depression-income relationship among adults in the United States. They argue that current employment status and financial strain, rather than income are robust predictors of depression. Daly et al. (2013) study the relative status and mental well-being from the perspective of US suicides and find that individual suicide risk rises with the reference group’s income. Lei et al. (2014) firstly examine the association between depressive symptoms and SES measured by education and percapita expenditure among Chinese adults. They find that depressive symptoms are associated with one’s educational level and per-capita expenditure. In summary, the existing literature hasn’t reached to a consensus on the effects of absolute and relative economic status on mental well-being. Several limitations in the current literatures might explain this inconsistency. First, most studies define relative economic status indirectly, i.e. the reference groups for comparisons are chosen by researchers instead of the respondents themselves, such as the mean income of the studied group (Blanchflower and Oswald 2004; Ferrer-i-Carbonell 2005; Oshio et al. 2011). The measured relative economic status thus may not represent the real position of the respondent’s subjective comparison with others, leading to substantial measurement error in the relationship between relative economic status and mental well-being. As pointed out by Senik (2009), the reference group chosen by one’s own trajectory is most powerful in predicting one’s well-being. Second, the traditional measures of relative status may suffer from endogeneity, creating bias in its estimated impact on mental well-being. For example, unobserved personal characteristics such as temper and personality may impact both relative economic status and mental well-being; additionally, happier people may also have higher productivities and thus higher relative economic status within their comparable social groups. Endogeneity caused by the above factors of unobserved heterogeneity and reverse causality could bias estimations both upward and downward.
116
5 Relative Economic Status and the Mental Health Status Among Chinese …
The main contribution of this study to the literature lies in the following aspects. Firstly, previous studies use self-reported happiness or life satisfaction measures to measure mental well-being, which in most cases are determined in part by the respondent’s current mood and memory (Kahneman and Krueger 2006) and are often subject to social desirability bias. We use an internationally comparable indicator to measure the individual’s mental well-being, namely the CES-D instrument. Secondly, in measuring people’s relative economic status, we use five reference groups for comparison, which are directly reported by the respondents during the survey. This direct measurement approach allows us to more accurately pinpoint the economic comparison and its impact on people’s mental health, compared to the traditional approach of reference-group selection by the researchers. Thirdly, our study is based on an updated nationally representative household survey in China and thus provides the latest evidence on relative economic status-mental health gradient for the most populous country in the world. The conclusions drawn from the results may offer meaningful policy implications for both China and other developing countries that face similar challenges of a widening income gap and deteriorating population mental health status.
5.3 Data and Methods 5.3.1 Data This study is based on data from the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative survey conducted by the National School of Development at Peking University. The target population of CHARLS is adults over the age of 45 and their spouses in both urban and rural households in China. The survey, aiming to serve as a multi-purposed social science and health survey of the elderly in China, is based on the design similar to that of the Health and Retirement Study (HRS) in the United States. The baseline survey was carried out in 2011 and 2012. Respondents of the survey were randomly selected, using a multi-stage cluster sampling design with three levels of sampling frames including county/city, village/ community, and households. At the first stage, 150 counties/cities were selected from over 2000 county-level units in 30 provincial-level administrative units in mainland China (Tibet is excluded from the survey) with probability proportional to size (PPS) method. Within each county/city, three rural villages/urban communities were selected with PPS. Within each village/community, 80 households were selected randomly using a specialized Geographic Information Systems program. Within each household, one randomly selected adult over the age of 45, called the main family respondent, and his/her spouse was asked to complete the questionnaire. CHARLS, with a total sample size of about ten thousand households, collected detailed information on individuals and families, which can be used for multidisciplinary research in economics, sociology, and demography (Zhao et al. 2013, 2014).
5.3 Data and Methods
117
The follow-up round survey was carried out in 2013 and 2014. Variables on the relative economic status are firstly added to the questionnaire. We focus on the main family respondents in the 2013/2014 survey who are aged 45 or above (10,493 observations). After excluding observations with missing information on the control variables (2,208 observations) and all five relative income variables (1871 observations), our final sample includes 6414 observations (see Appendix Table 5.6 with comparison of characteristics for those not included in the sample with those in the sample). The variable of focus is the perceived relative economic status compared to different groups. The comparison groups include relatives, colleagues, schoolmates, neighbors or villagers (residents living in the same village), and people living in the same city or county, respectively. Take the comparison with schoolmates as an example, respondents were asked, “Compared to the average living standard of your schoolmates who are at the same level of education with you, how would you rate your standard of living?” After making such a description, the person then assesses his/ her economic status with “much better”, “a little better”, “about the same”, “a little worse”, “much worse”. We categorize the former two answers as “better”, and the latter two as “worse”. These five comparison groups cover major dimensions in the respondent’s social connections as relatives have kindred relations with the respondents; colleagues are those with similar professional background; schoolmates have similar education background with the respondents; neighbors (or villagers) are those who share similar social network with the respondents; and people in the same city/ county have geographic proximity, and are influenced by similar regional cultures and social and economic conditions. Most studies on the impact of relative status use indirect measures chosen by a third party, usually the researcher, to identify the reference groups for comparison (Ferrer-i-Carbonell 2005; Firebaugh and Schroeder 2009; Knight et al. 2009). Since these are not the actual comparison groups identified by the respondents themselves, the results may not reflect the real impact of relative status (Knight et al. 2009; Senik 2009). This study improves upon the literature by measuring the relative status based on actual comparison groups as reported by the respondents and thus provides direct evidence on the impact of economic comparison on people’s mental health. The main outcome variable of interest is mental health, measured by the ten questions version of the Center for Epidemiological Studies Depression (CES-D) instrument. CES-D is originally developed by Radloff (1977), contains 20 items, and is widely used to assess depressive symptoms. Andresen et al. (1994) extract ten items from 20 and developed the simplified CES-D. Out of the ten items in the CESD, eight measure negative feelings (e.g. “I feel lonely”) and two measure positive feelings (e.g. “I am confident in the future”). The respondents recall their moods in the past seven days and tick one response to indicate the frequency of these feelings. Lei et al. (2014) and Ren et al. (2014) provide empirical validation of the simplified CES-D on the Chinese middle-aged and elderly population using the CHARLS data. As suggested by Radloff (1977), the responses to the questions on negative feelings are assigned to an integer value of 0 (rarely or none of the time), 1 (some or a little of the time), 2 (occasionally or a moderate amount of the time), 3 (most or all of the
118
5 Relative Economic Status and the Mental Health Status Among Chinese …
time), and the responses to the questions on positive feelings are recoded as 3 (rarely or none of the time), 2 (some or a little of the time), 1 (occasionally or a moderate amount of the time), 0 (most or all of the time). Based on the above scoring system, the CES-D score can thus be calculated by adding up the scores of items (Andresen et al. 1994; Hsieh and Qin 2018). The overall score ranges from 0 to 30, with higher scores indicating more severe depressive symptoms.
5.3.2 Methods 5.3.2.1
Correlation Between Relative Economic Status and Mental Health
The dependent variable in our main regression is the CES-D score. We use ordinary least square (OLS) model to estimate the effect of relative economic status on depressive symptoms. The equation can be expressed as follows: Depressioni = α +
k
βk ∗ Relative_EcoStatusik + Z i + Pi ς + u i
(5.1)
Depressioni represents the CES-D score of individual i. We treat the score as a continuous variable in all regression models. Relative_EcoStatusik denotes a set of dummies on the relative economic status of respondent i. We divide relative economic status into three groups: “as the same” (k = 1), “better” (k = 2), and “worse” (k = 3). We treat “as the same” (k = 1) as the base group. Z i is a set of control variables, including personal characteristics (age, sex, hukou, marital status, education level, working status, and health status), and family characteristics (family size, and family annual income per capita). “hukou” is a legally required household residential registration status that identifies all individuals as either urban or rural. Since the self-rated health may suffer from endogeneity, we use the presence of any chronic diseases or physical disability to measure the physical health of the respondent. Pi is a set of provincial dummy variables. ui is the random error term. The parameter β ik , the key coefficient of interest, identifies the relationship of depression and relative economic status k. For example, when the comparison group is the respondent’s relative, β i2 (β i3 ) refers to the impact of having better (worse) economic status than the relative on CES-D score. To avoid potential within-cluster correlation in the error term, all standard errors are clustered at the community level.
5.3.2.2
Instrumental Variable (IV) Methods
The above model implicitly assumes that relative economic status is exogenous. However, due to the aforementioned reasons, the estimation may suffer from endogeneity bias due to unobserved factors impacting both relative economic status and
5.3 Data and Methods
119
mental health as well as the reverse causality between the independent and dependent variables. To explicitly address this issue, we employ an IV model in our estimation for robustness check. The CHARLS asked both the respondents and their spouses separately about their perceived relative economic status (each comparing to the five reference groups). We use the relative economic status of the respondent’s spouse as the IV. For the IV estimation, the sample is thus limited to the married people, with a sample size of 5461. Family members tend to be affected by similar factors or interact with each other. Many prior studies employ the information on other family members as the IV for the respondent’s information to reduce the impact of endogeneity. For example, Yin et al. (2015) study the effect of financial knowledge on entrepreneurial activity. They argue that parents’ knowledge is highly correlated with the respondent’s knowledge and will not directly impact the decision making of the respondent. Therefore, they use parents’ knowledge as an IV for respondent’s financial knowledge, and the specification passes the weak IV test. Yin and Gan (2014) estimate the impact of smoking and drinking on income. They use smoking and drinking behavior of other family members as the IV for the respondent’s smoking and drinking behavior. In our study, on the one hand, husband and wife share part of family wealth and social capitals, thus their relative economic status would be highly correlated. On the other hand, the spouse’s relative economic status is not likely to have a direct impact on the respondents’ mental health after addressing other variables. This is because the spouse’s relative economic status is mainly decided by his/her comparison with the reference groups (based on their judgment of each other’s economic status), and the spouse’s relative economic status would not be significantly affected by the respondent’s depressive symptoms. Furthermore, to confirm the validity of our IV choice, we carry out the IV tests as suggested in the literature. As shown in Appendix Table 5.7, both the under-identification test and the weak identification test indicate that the excluded instruments are not weak, which in turn provides supporting evidence on the validity of the IV. However, the current IV may have its own limitation because the spouses’ perceived relative economic status may impact the respondents’ mental health status. For example, the spouses might nag about his/her low perceived relative economic status, which may cause the respondents’ more depressive. To partially address this problem, we try to control the possible pathways that the spouse’s mental health might impact the respondent’s mental health through the spouse’s perception. For this purpose, we add perception variables of the spouse in the IV regression models which are measured by the depressive symptoms and life satisfaction. The results are consistent with those of the original IV models.
120
5 Relative Economic Status and the Mental Health Status Among Chinese …
5.4 Empirical Results 5.4.1 Sample Description Table 5.1 reports the sample descriptive statistics. The average CES-D score of the sample population is 7.858, lower than the cutoff point (10 and above) for high depressive symptoms (Andreasen et al. 1994). In particular, 31.8% of the sample respondents have CES-D score of 10 or above, which is considerably higher than the statistics reported for other Asian populations. For example, Witoelar et al. (2012) report that in Indonesia, only 5% of men and 9.4% of women aged 45 and over have CES-D score of 10 or above. This result thus confirms the severity of mental depression in China. The average age of the sample respondents is about 61 and 79.2% of them are married. 77.8% of the respondents have rural hukou. The education level of the respondents is generally low: only 12.1% of respondents have an educational attainment of senior high school/technical secondary school or above. In terms of health conditions, 67.8% of respondents have at least one chronic disease, and 18.4% of respondents suffered physical disability, which is likely due to the older age profile of the sample. With regard to income, the average family annual income per capita is about 6173 RMB. This study includes five different comparison types of relative economic status. According to the descriptive statistics, about 50% of the respondents consider their economic status as the same as their relatives, schoolmates, colleagues, and neighbors in the same village, and 34–42% of the sample consider their economic status worse than the above reference groups. Less than 10% of the sample say that they are better off than the reference groups. For the reference group of people living in the same city or county, more than half of the respondents consider their economic status to be worse than the local average, which implies that Chinese people tend to be “humble” and underestimate their socioeconomic ranks. Figure 5.1 shows the depressive symptoms among people with different relative economic status. It shows that individuals who report their economic status to be lower than the average are more likely to have depressive symptoms as measured by CES-D. Moreover, those who have lower relative economic status see greater impacts on depressive symptoms than those whose relative economic status is higher than the reference groups, which indicates a possible asymmetric effect of relative economic status on mental health. Furthermore, the impact of comparison to different reference groups varies. For example, comparison made to the economic status of relatives and neighbors seem to present the largest impacts on respondent’s depressive symptoms, while comparison to the local residents shows the smallest impact.
5.4 Empirical Results
121
Table 5.1 Descriptive statistics on key variables Variable
N
The relative economic status compared to relatives
6330
Mean* SD
Min Max
Better
0.087
0.281 0
1
About the same
0.542
0.498 0
1
Worse
0.371
0.483 0
1
The relative economic status compared to schoolmates
4581
Better
0.083
0.276 0
1
About the same
0.495
0.500 0
1
0.421
0.494 0
1
Better
0.071
0.256 0
1
About the same
0.579
0.494 0
1
Worse
0.350
0.477 0
1
Worse The relative economic status compared to colleagues
3892
The relative economic status compared to neighbors or 6251 the villagers Better
0.079
0.270 0
1
About the same
0.576
0.494 0
1
Worse
0.344
0.475 0
1
The relative economic status compared to people in the 5928 city or county Better
0.040
0.196 0
1
About the same
0.276
0.447 0
1
Worse Depression symptoms score (CES-D score)
0.684
0.465 0
1
7.858
5.825 0
30
Age (Year)
6414 61.064 9.694 45
95
Male (Yes = 1)
6414 0.463
0.499 0
1
Agricultural hukou# (Yes = 1)
6414 0.778
0.416 0
1
Married (Yes = 1)
6414 0.792
0.406 0
1
Education level:
6414
Primary school and below
0.677
0.468 0
1
Middle school
0.203
0.402 0
1
0.121
0.326 0
1
Senior high school/technical secondary school and above Working status:
6414
Unemployed
0.244
0.430 0
1
Government and Gov. affiliated institutions
0.029
0.168 0
1
Firm
0.036
0.188 0
1
Self-employed
0.534
0.499 0
1 (continued)
122
5 Relative Economic Status and the Mental Health Status Among Chinese …
Table 5.1 (continued) Variable
N
Retired
Mean* SD 0.112
Min Max
0.315 0
1
0.045
0.207 0
1
Chronic disease (Yes = 1)
6414 0.678
0.467 0
1
Physical disability (Yes = 1)
6414 0.184
0.388 0
1
Others
Family size
6414 3.246
1.745 1
15
Family annual income per capita (yuan)
6414 6173
9981
1.69e+05
0
Note Mean* for continuous variables, and percentage for discrete variables. hukou# is a special vocabulary in China mainland, and it means registered permanent residence 10
CES-D score(relative economic status compared to relatives)
9 CES-D score (relative economic status compared to schoolmates)
CES-D score 8
CES-D score (relative economic status compared to colleagues)
7
CES-D score (relative economic status compared to neighbors or villagers)
6
5 Better
About the same
Worse
CES-D score (relative economic status compared to people living in the same city or county)
The relative economic status compared to the reference group
Fig. 5.1 Relative economic status and depressive symptoms. Note The horizontal x-axis represents relative economic status compared with five different reference groups, and the vertical Y-axis represents CES-D score of people having better (or about the same, worse) relative economic status compared with the reference group. Five lines show the CES-D score of people whose relative economic status better (or about the same, worse) compared with five different reference groups, respectively. For example, the dotted line shows that the CES-D score is 5.9, 6.9, and 9.8 for people whose relative economic status better, as the same as, and worse than relatives, respectively
5.4 Empirical Results
123
5.4.2 Regression Results 5.4.2.1
The Effect of Relative Economic Status on Individual Mental Health
Table 5.2 reports the regression results of the baseline model. We observe an inverse U-shaped relationship between age and depressive symptoms, which is consistent with Senik (2009), who finds a U-shaped relationship between age and life satisfaction.1 Women have higher CES-D score than men, which is consistent with most comparable studies in China. As suggested by Li et al. (2014), the poorer mental health status by the Chinese female population is attributable to their higher risk of onset, longer life expectancy, higher widowhood rate, lower social status, economic income, and degree of education. Married people have lower tendency of being depressive, which support the hypothesis that marriage is an important source of happiness, and it provides protection from many mental health problems. As expected, rural hukou holders have more depressive symptoms than the urban hukou holders, possibly due to their poorer access to healthcare and other social benefits. There is a negative association between education and depressive symptoms: better-educated people in China tend to be more mentally healthy, indicating the education-health gradient. Unemployed people are more likely to have higher CES-D score than the employed people and the voluntary retirees. Although depressive symptoms decrease with the annual household income per capita, the effect is not significant at 10% level.2 Table 5.2 also shows that comparison among social groups plays an important role in determining an individuals’ mental health. The results for the five different reference groups show a consistent pattern: those who have worse (better) relative economic status are more likely to have higher (lower) levels of depressive symptoms. Moreover, there is an asymmetric effect in comparisons: those with a lower perceived relative economic status than their reference groups see a larger impact on their mental health (in quantitative magnitude) than those who report a higher perceived relative economic status (Refer to the test coefficient on “worse”- coefficient on “better” = 0 in Table 5.2). A plausible reason is that those with lower perceived relative economic status tend to experience stress as they cannot derive social status from their wealth, and this in turn may make them suffer from depressive symptoms. It should also be noted that the coefficients of absolute income variable (family annual income per capita) become statistically insignificant after addressing the relative economic status measures, which suggests that the influence of relative economic status may 1
The numbers of observations are different for the five different relative status measures in Table 5.2. For robustness check, we deleted observations with missing information on any one of the five relative economic status variables. We also include observations with missing information of relative status measures. Results are reported in Appendix Table 5.8, which remain similar to those in Table 5.2. 2 We also use individual income instead of family per capita income to measure absolute economic status, and results remain the same.
124
5 Relative Economic Status and the Mental Health Status Among Chinese …
Table 5.2 Regression results (OLS) Variable
Dependent variable: CES-D score (1)
(2)
(3)
(4)
(5)
(6)
The relative economic status compared to relatives (base group: “About the same”) −0.827***
Better
(0.233) Worse
2.503*** (0.158)
The relative economic status compared to schoolmates (base group: “About the same”) Better
−0.723*
Worse
2.130***
(0.284) (0.174) The relative economic status compared to colleagues (base group: “About the same”) Better
−1.171***
Worse
2.172***
(0.276) (0.188) The relative economic status compared to neighbors or the villagers (base group: “About the same”) Better
−1.145***
Worse
2.384***
(0.219) (0.158) The relative economic status compared to people in the city or county (base group: “About the same”) Better
−0.731*
Worse
1.676***
(0.346) (0.151) Age Age square
0.346***
0.309***
0.374***
0.344**
0.312***
0.374***
(0.089)
(0.088)
(0.104)
(0.109)
(0.089)
(0.090)
−0.003*** −0.003*** −0.003*** −0.003**
−0.003*** −0.003***
(0.001)
(0.001)
(0.001)
(0.001)
(0.001)
(0.001)
Male (Yes = 1) −1.240*** −1.212*** −1.356*** −1.323*** −1.184*** −1.209*** (0.138)
(0.135)
(0.160)
(0.170)
(0.139)
(0.143) (continued)
5.4 Empirical Results
125
Table 5.2 (continued) Variable Agricultural hukou (Yes = 1)
Dependent variable: CES-D score (1)
(2)
(3)
(4)
(5)
(6)
1.305***
1.261***
1.358***
1.076***
1.299***
1.221***
(0.203)
(0.201)
(0.212)
(0.224)
(0.204)
(0.203)
Married (Yes = −1.215*** −1.005*** −1.492*** −1.330*** −1.051*** −1.170*** 1) (0.196) (0.191) (0.242) (0.255) (0.193) (0.204) Education level (base group: “Primary school and below”) Middle school Senior high school and above
−0.558**
−0.498**
−0.591**
−0.567** (0.213)
−0.382*
−0.555**
(0.185)
(0.179)
(0.190)
(0.179)
(0.186)
−0.700**
−0.597**
−0.759*** −0.762**
−0.427
−0.509*
(0.227)
(0.227)
(0.228)
(0.251)
(0.234)
(0.238)
Working status (base group: “Unemployed”) Government and Gov. affiliated institutions
−1.122**
−0.693
−0.617
−0.810
−0.505
−0.978*
(0.397)
(0.378)
(0.433)
(0.425)
(0.399)
(0.410)
Firm
−1.214*** −1.041**
−0.897*
−1.119**
−1.009**
−1.281***
(0.343)
(0.427)
(0.412)
(0.367)
(0.356) −0.764***
(0.359)
Self−employed −0.739*** −0.684*** −0.672** (0.201)
(0.194)
−0.806**
−0.621**
(0.228)
(0.249)
(0.197)
−0.851**
−1.329*** −0.797**
(0.204)
Retired
−1.124*** −0.858**
−0.798**
(0.279)
(0.279)
(0.303)
(0.324)
(0.281)
(0.295)
Others
−1.109**
−0.929**
−0.882*
−1.065**
−0.986**
−1.168**
(0.355)
(0.353)
(0.387)
(0.402)
(0.355)
(0.361)
Chronic disease 1.842*** (Yes = 1) (0.147)
1.718***
1.742***
1.506***
1.756***
1.795***
(0.141)
(0.169)
(0.183)
(0.146)
(0.152)
Physical disability (Yes = 1)
1.029***
0.888***
0.890***
1.040***
0.838***
0.952***
(0.204)
(0.199)
(0.219)
(0.232)
(0.196)
(0.208)
Family size
−0.119**
−0.090*
−0.030
−0.028
−0.092*
−0.131**
(0.046)
(0.044)
(0.056)
(0.058)
(0.045)
(0.047)
Log (Family annual income per capita)
−0.028
−0.016
−0.031
−0.045
−0.018
−0.019
(0.021)
(0.020)
(0.024)
(0.025)
(0.021)
(0.022)
Province dummies
Yes
Yes
Yes
Yes
Yes
Yes (continued)
126
5 Relative Economic Status and the Mental Health Status Among Chinese …
Table 5.2 (continued) Variable
Dependent variable: CES-D score (1)
(2)
(3)
(4)
(5)
(6)
−1.289
−2.413
−3.258
−2.443
−3.785
−5.068
(2.757)
(2.743)
(3.238)
(3.479)
(2.777)
(2.803)
N
6414
6330
4581
3892
6251
5928
R2
0.122
0.169
0.168
0.168
0.164
0.144
0.000
0.000
0.007
0.000
0.023
Constant
Test coefficient on “worse”coefficient on “better” = 0 (Prob > F)
Note “Yes” indicates province characteristics are controlled in the regression. The sample sizes for control variables are the same in the above six columns, but the sample sizes for the five variables of relative income are different. The results are similar if we use the same sample for the five variables of relative income. Clustered standard errors in parentheses. *p < 0.05. **p < 0.01. ***p < 0.001
outweigh the absolute wealth on a person’s mental health, and thus it is meaningful to distinguish between relative and absolute wealth status for our study purpose.
5.4.2.2
Subsample Analysis
As suggested by previous literature, lower socioeconomic status (SES) is generally associated with worse health, including mental health (Everson et al. 2002; Lorant et al. 2003; Lei et al. 2014). According to Koster et al. (2006), psychosocial factors explained on average 16% of the SES differences in the prevalence of depression. Possible explanations include poorer access to mental healthcare resources, less job stability, and poorer management of stress (Meltzer et al. 2010). However, there are few studies exploring directly the heterogeneity in the association between relative SES and mental health among different subpopulations. The most relevant is by Clark and Senik (2010) who estimated the intensity (how much) and direction (to whom) of wealth comparisons, and find that the self-employed and the rich were less likely to compare to others, and are happier than average when they do. In this section, we aim to extend the above literature by exploring the heterogeneous impact of relative economic status on mental health status among different SES groups (classified by income levels and working status). Table 5.3 shows that the impact of relative economic status is different among different income groups. A decrease in relative economic status appears to have a larger effect on mental health for those in lower-income groups than higher-income groups. The impact of relative economic status on mental health among the middleincome group is smaller than that among the lower-income group but with no significant difference between these two groups. The above results are consistent with the findings by Clark and Senik (2010) based on data from European countries,
5.4 Empirical Results
127
showing that wealth comparisons are associated with lower levels of subjective wellbeing with a larger effect for people with lower incomes. Additionally, doing better off than the reference group plays an important role on improving mental health for the higher-income group, while for the lower-income group it has little effect. This distinction further highlights the variation in the asymmetric gradient between relative economic status and mental health. We also test the gender heterogeneity in the determinants of mental health, and the results shown in Table 5.4 indicate that the asymmetric effects of relative economic status on depressive symptoms are similar among men and women (refer to SUR test in Table 5.4). The impact of comparison to local average economic status is the smallest for both men and women. Overall, these subsample analyses suggest that the role of different reference groups also varies among different population groups in affecting their mental health, and using only one reference group to measure the relative economic status level may cause estimation bias.
5.4.2.3
Robustness Check: IV Regression Results
In this part, we try to identify the causal relationship of relative economic status and mental health using an IV model for respondents with spouses. The IV, the spouse’s self-rated relative economic status, has been shown to meet the validity requirement (see Appendix Table 5.7). The second-stage regression results are shown in Table 5.5, which shows that while the impact of family absolute income remains statistically non-significant, the impact of relative economic status becomes stronger compared to the baseline regressions in Table 5.2. Moreover, except the results of using relatives as the reference group, the asymmetric effects associated with the other four reference groups are stronger than those in Table 5.2. Specifically, using colleagues as the reference group, when comparisons are favorable (outperforming one’s colleagues), there is no significant impact on mental health; however, when comparisons are unfavorable (underperforming one’s colleagues), the negative impact becomes more significant in statistical sense. It reflects that the negative impact of relative economic status comparisons has a larger influence than the positive impact. When comparing their economic status with the reference group, people tend to be more depressed when they consider themselves worse off than their peers, but it does not give them an equally significant mental relief when they find out that they are doing better off than their peers. In conclusion, after addressing the endogeneity of income, the effect of relative economic status on mental depression becomes stronger with more obvious asymmetry. Additionally, it is clear that the impact of comparison to colleagues and schoolmates is the largest among the five reference groups, followed by the comparison with relatives and neighbors in the same village, while comparing to the local average economic status of the same county/city has the smallest effect. Since the local average income level was often used in prior studies in forming the (hypothetical) reference groups, the above results suggest that such an approach may underestimate
128
5 Relative Economic Status and the Mental Health Status Among Chinese …
Table 5.3 Subsample regression results for different income groups Variable (1)
(2)
(3)
The The The SUR test SUR test lower-income middle-income higher-income between (1) & between (1) & group group group (2) Prob > Chi2 (3) Prob > Chi2 The relative economic status compared to relatives (base group: “About the same”) Better Worse N
−0.257
−0.959*
−1.050***
(0.456)
(0.452)
(0.302)
2.635***
2.476***
2.288***
(0.269)
(0.283)
(0.269)
2108
2111
2111
0.277
0.137
0.681
0.345
The relative economic status compared to schoolmates (base group: “About the same”) Better Worse N
−0.432
−0.468
−1.037*
(0.541)
(0.551)
(0.402)
2.457***
2.267***
1.698***
(0.327)
(0.323)
(0.264)
1435
1435
1711
0.962
0.357
0.664
0.064*
The relative economic status compared to colleagues (base group: “About the same”) Better Worse N
−0.710
−1.335*
−1.172**
(0.634)
(0.633)
(0.383)
2.562***
2.486***
1.533***
(0.379)
(0.360)
(0.295)
1183
1161
1548
0.472
0.543
0.881
0.030**
The relative economic status compared to neighbors or the villagers (base group: “About the same”) Better
−0.808
−1.239*
−1.201***
(0.424)
(0.492)
(0.335)
Worse
2.621***
2.563***
1.820***
(0.269)
(0.265)
(0.275)
N
2073
2119
2059
0.485
0.476
0.876
0.031**
The relative economic status compared to people in the city or county (base group: “About the same”) Better Worse
−0.193
−0.289
−1.121*
(0.669)
(0.693)
(0.451)
1.982***
1.607***
1.351***
(0.278)
(0.311)
(0.231)
0.921
0.229
0.371
0.089* (continued)
5.5 Discussions and Conclusions
129
Table 5.3 (continued) Variable (1)
(2)
(3)
The The The SUR test SUR test lower-income middle-income higher-income between (1) & between (1) & group group group (2) Prob > Chi2 (3) Prob > Chi2 N
1942
2010
1976
Note Only coefficient estimates of the relative economic status terms are reported, all the results are conditional on the characteristics, including age (and age square), gender, registered permanent residence (hukou), married, education level, working status, chronic disease, physical disability, family size, and province dummies. Clustered standard errors are in parentheses. *p < 0.05. **p < 0.01. ***p < 0.001. Seemingly Unrelated Regression (SUR) test is used to identify whether the coefficients vary between groups (H0: There is no significant difference of the coefficients between groups)
the impact of relative economic status. The differential roles of reference groups also indicate that individuals are more likely to compare with cohorts who have similar socioeconomic characteristics to themselves (e.g. similar age, working experience, and education background) and who they often associate with, and this is consistent with the findings of McBride (2001). To ensure that the distinction in regression results between the baseline model and the IV model is not driven by the difference in the samples. We re-estimate the OLS model on the married sample, i.e. using the same sample as the IV regressions. The results are shown in Appendix Table 5.9, which are similar to those reported in Table 5.2. Hence, excluding the sample of single adults for the IV regressions are not likely to affect our findings.
5.5 Discussions and Conclusions Based on a nationally representative sample of middle-aged and elderly adults, this study analyzes the relationship between relative economic status and mental health in China. We find that wealth comparison presents a significant impact on mental health. Firstly, when comparing themselves to the five different reference groups, lower relative economic status is associated with poorer mental health (i.e. higher tendency for depressive symptoms). Secondly, the impacts of comparisons are asymmetric: favorable comparisons are associated with quantitatively smaller effect on mental health than unfavorable comparisons. In other words, the mental cost of underperforming one’s reference group is disproportionally larger than the mental benefit of outperforming it. Thirdly, the impact of comparison differs across reference groups: the larger impacts are seen from the comparison against acquaintances (such as relatives, schoolmates, and colleagues), while the impact of comparison to regional average is the smallest though it was commonly used in literature. Fourthly, the subsample analyses suggest that the asymmetric effect of relative economic status on depressive
130
5 Relative Economic Status and the Mental Health Status Among Chinese …
Table 5.4 Subsample regression results for different genders Variable
Males
Females
SUR test Prob > Chi2
The relative economic status compared to relatives (base group: “About the same”) −0.943***
−0.741*
(0.301)
(0.358)
Worse
2.284***
2.650***
(0.203)
(0.215)
N
2930
3400
Better
0.658 0.182
The relative economic status compared to schoolmates (base group: “About the same”) −0.892**
−0.488
(0.306)
(0.494)
Worse
2.104***
2.103***
(0.215)
(0.275)
N
2367
2214
Better
0.477 0.996
The relative economic status compared to colleagues (base group: “About the same”) −0.950**
−1.276*
(0.324)
(0.495)
Worse
2.200***
2.136***
(0.248)
(0.279)
N
2009
1883
Better
0.587 0.858
The relative economic status compared to neighbors or the villagers (base group: “About the same”) Better
−0.902** (0.275)
(0.359)
Worse
2.320***
2.447***
(0.215)
(0.223)
2891
3360
N
−1.418***
0.258 0.678
The relative economic status compared to people in the city or county (base group: “About the same”) −1.018*
−0.502
(0.395)
(0.626)
Worse
1.538***
1.820***
(0.199)
(0.222)
N
2762
3166
Better
0.485 0.331
Note Other control variables include age (and age square), gender, registered permanent residence (hukou), married, education level, working status, chronic disease, physical disability, family size, and province dummies. Clustered standard errors are in parentheses. *p < 0.05. **p < 0.01. ***p < 0.001. Seemingly Unrelated Regression (SUR) test is used to identify whether the coefficients vary between groups (H0: There is no significant difference of the coefficients between groups)
5.5 Discussions and Conclusions
131
Table 5.5 Robustness check: regression results using IV methods Variable
Dependent variable: CES-D score (1)
(2)
(3)
(4)
(5)
Compared to relatives
Compared to schoolmates
Compared to colleagues
Compared to neighbors or the villagers
Compared to people in the city or county
−1.445
1.346
The relative economic status (base group: “About the same”) Better
−3.494** (1.162)
(2.418)
(2.136)
(1.040)
(1.996)
Worse
3.827***
4.749***
4.635***
3.786***
2.491***
(0.527)
(0.977)
(1.001)
(0.516)
(0.563)
0.031
0.021
−0.006
0.016
0.009
(0.027)
(0.036)
(0.036)
(0.027)
(0.029)
Other control Yes variables
Yes
Yes
Yes
Yes
N
4345
2730
2294
4261
3941
R2
0.236
0.228
0.227
0.237
0.233
Log(Family annual income per capita)
−1.421
−1.385
The five columns presented the results of the five regression models for the relative economic status compared to relatives, schoolmates, colleagues, neighbors or the villagers, and people in the city or town, respectively. Other control variables include age (and age square), gender, registered permanent residence (hukou), married, education level, working status, chronic disease, physical disability, family size, and province dummies. The spouses’ perceived relative income would impact the respondents’ mental health status through spouses’ symptoms. Thus, we try to control the possible pathway of the spouses’ mental health by adding the variables of the depressive symptoms and life satisfaction of the spouses in the IV regressions. The results are similar to the findings in Table 5.3. Clustered standard errors are in parentheses. *p < 0.05. **p < 0.01. ***p < 0.001
symptoms is similar among men and women, but it is larger among the groups of lower socioeconomic status such as the low-income group. Our findings reveal the significant influence of self-perceived relative economic status in affecting individuals’ mental health in China, which is consistent with most international studies (Wilkinson 1992; Ferrer-i-Carbonell 2005; Daly et al. 2013). Individuals’ mental well-being is more likely to be affected by comparisons with others who have similar backgrounds (such as education, work status, and living condition) and people with whom they keep in close contact, and the mental cost of “keeping up with the Jones” turn out to be substantial. In comparison, the absolute income tends to play a smaller role in determining people’s mental well-being. Our findings are in contradiction to Luo (2009) but consistent with Guan (2010). Both of these studies focus on the Chinese population but use different methods to measure relative income. Luo (2009) used subjective poverty line as the reference group of comparison, while Guan (2010) used self-rated relative income, which is more similar to our measure. Our findings give evidence that the “Easterlin paradox” exists
132
5 Relative Economic Status and the Mental Health Status Among Chinese …
among Chinese population that relative economic status for individuals in a society is more important for their mental well-being than absolute income (Easterlin 1974, 1995). Moreover, according to Strain Theory, when economically poor individuals realize that other people of the same or similar background are leading a much better life, they will experience deprivation strain (Zhang et al. 2014). Deprivation stain was found significantly associated with mental health problems and suicide (Zhang et al. 2011, 2014), and it could help explain our findings that favorable comparison had limited effect on improving mental health while worse relative economic status led to significant worse mental health. As the social comparison theory predicts, people’s mental well-being may remain unchanged when their absolute income increases and their relative economic status remain unchanged; what’s more, the mental well-being may even be reduced when people’s relative economic status deteriorates, and this may be caused by an enlarging income disparity and the inequalities of labor market opportunities (Eaterlin 1974, 1995). In the past 30 years, China’s income gap is continuously expanding despite the rapid growth of real economy. According to the calculation of China’s national bureau of statistics, the Gini coefficient of China was 0.2 in the 1980s, and rose to 0.469 in 2015, far higher than the international warning line. Our results help to understand the much larger role of relative economic status than absolute wealth in determining people’s mental well-being and explain why the prevalence of mental disorders keeps increasing despite the rapid economy growth. Although individuals are not likely to have precise information about the actual wealth distribution of the population or their reference groups, the perceived wealth inequality and the feeling of underperforming one’s peers in socioeconomic comparison may affect one’s mental well-being (Oshio and Urakawa 2014). Thus, policies that aim to reduce wealth inequality among people with similar backgrounds may also have the spillover benefit of improving population mental health status. For example, discrimination is an important reason for inequality (Lang and Lehmann 2012). In China, discriminations on hukou and gender are common (Sicular et al. 2007). People in the same work units with similar education background and ability can be paid differently due to hukou or gender differences. Therefore, anti-discrimination programs could help people with similar characteristics and abilities to get equal payments, which in turn can improve individual’s mental well-being as a whole. In addition, our study reveals that vulnerable groups, including the lower-income group and the unemployed group are more likely to be affected by income comparisons, which is different from the findings of McBride (2001) that the effect of relative income is smaller at lower-income levels in the United States. Thus, we should pay more attention to the mental health of low-income population and other vulnerable populations in China. Policy programs targeting at reducing income gaps within lower-income regions and lower-income populations may help improve the mental well-being among these disadvantaged groups.
Appendix
133
Our study has three main limitations. Firstly, we use self-reported CES-D to measure mental health rather than the clinical measure of depression based on standardized diagnosis, and the CES-D questionnaire covers only one week before the survey, which may cause potential measurement error among adults with severe mood variation or those with recent traumatic experiences (Lindell and Whitney 2001). Secondly, the potential endogeneity problem is an obstacle for the study of this topic, and most of the literature does not adequately address this problem. This study tries to use the IV method to identify the causal relationship between relative economic status and mental health. However, the IV model requires spouse information, limiting the sample to individuals who are married. Moreover, spouse’s perceived relative economic status might affect the respondent’s mental well-being although the statistical tests on the validity of IV are passed and the potential mechanisms underlying the effect of spouse’s relative economic status on mental health are controlled. Further studies would extend the current one by using other IV(s) for the whole population including the married and unmarried samples. China’s economy had a miraculous growth record in the past three decades, but at the same time wealth disparity was also rising at a fast pace, making people care much more about their relative position in the socioeconomic hierarchy than their absolute wealth. Our conclusion implies that this relative economic status comparison has a real and substantial impact on people’s mental well-being, especially for the low-income and unemployed population, and the negative impact of perceived lower relative income is comparatively larger and more significant than the positive impact of scoring higher in the income comparison. Thus, policies that aim to reduce wealth inequality among people with similar backgrounds (e.g. reducing income gaps within lower-income regions and lower-income populations) may be meaningful, and they can have the spillover benefits of improving the mental health status of the population as a whole.
Appendix See Tables 5.6, 5.7, 5.8 and 5.9.
134
5 Relative Economic Status and the Mental Health Status Among Chinese …
Table 5.6 Descriptive statistics of original sample and study sample Variable
Original sample N
Study sample (N = 6.414)
Mean
S.D
Mean
S.D
Age (Year)
10369
61.203
10.661
61.064
9.694
Male (Yes = 1)
10493
0.460
0.498
0.463
0.499
Agricultural hukou (Yes = 1)
9844
0.777
0.416
0.778
0.416
Married (Yes = 1)
9930
0.775
0.418
0.792
0.406
Primary school and below
10376
0.683
0.465
0.677
0.468
Middle school
10376
0.196
0.397
0.203
0.402
Senior high school and above
10376
0.121
0.327
0.121
0.326
Educational level
Working status Unemployed
9478
0.259
0.438
0.244
0.430
Government and gov. affiliated institutions
9478
0.031
0.174
0.029
0.168
Firm
9478
0.044
0.205
0.036
0.188
Self-employed
9478
0.498
0.500
0.534
0.499
Retired
9478
0.113
0.317
0.112
0.315
9478
0.054
0.226
0.045
0.207
Chronic disease (Yes = 1)
Others
10493
0.582
0.493
0.678
0.467
Disabled (Yes = 1)
10493
0.177
0.382
0.184
0.388
Family size
9819
3.388
1.819
3.246
1.745
Family annual income per capita (yuan)
8740
5.978
9.961
6.173
9.981
Table 5.7 Statistical tests on the validity of instrumental variables Test
(1)
(2)
(3)
(4)
(5)
The relative economic status compared to relatives
The relative economic status compared to schoolmates
The relative economic status compared to colleagues
The relative economic status compared to neighbors or the villagers
The relative economic status compared to people in the city or county
Tests of endogeneity (null hypothesis H0: the variable is exogenous) Wu-Hausman F value
14.807
7.630
6.844
9.309
1.209
(reject H0)
(reject H0)
(reject H0)
(reject H0)
(cannot reject H0)
Under-identification test (null hypothesis H0: IV is independent of endogenous variables) Kleibergen-Paap rk LM statistic
89.442
27.041
28.112
106.998
38.916
(reject H0)
(reject H0)
(reject H0)
(reject H0)
(reject H0) (continued)
Appendix
135
Table 5.7 (continued) Test
(1)
(2)
(3)
(4)
(5)
The relative economic status compared to relatives
The relative economic status compared to schoolmates
The relative economic status compared to colleagues
The relative economic status compared to neighbors or the villagers
The relative economic status compared to people in the city or county
Weak identification test (null hypothesis H0: weak correlation between IV and endogenous variables) Kleibergen-Paap rk Wald F statistic
50.863
14.362
15.003
60.670
22.678
(reject H0)
(reject H0)
(reject H0)
(reject H0)
(reject H0)
Note The results of endogeneity tests show that most variables of relative economic status are endogenous except of the relative economic status compared to people in the city or county, so we need to employ IV methods to control the endogeneity problem. The IV in this study (the relative economic status of the respondent’s spouse) passed the under-identification test and weak identification test Table 5.8 Regression results (OLS) with same number of observations for five different relative status measures Variable
Dependent variable: CES-D score (1)
(2)
(3)
(4)
(5)
Compared to relatives
Compared to schoolmates
Compared to colleagues
Compared to neighbors or the villagers
Compared to people in the city or county
Panel A. using same sample size for all the five relative status measures The relative economic status (base group: “About the same”) −0.715*
−0.684*
−1.231***
−1.165***
−0.729
(0.292)
(0.331)
(0.292)
(0.276)
(0.395)
2.477***
2.020***
2.190***
2.181***
1.457***
(0.218)
(0.210)
(0.209)
(0.234)
(0.197)
Other control variables
Yes
Yes
Yes
Yes
Yes
N
3260
3260
3260
3260
3260
R2
0.180
0.167
0.173
0.171
0.150
−0.728*
Better Worse
Panel B. using the entire sample The relative economic status (base group: “About the same”) Better Worse
−0.829***
−0.715*
−1.197***
−1.148***
(0.234)
(0.283)
(0.273)
(0.218)
(0.346)
2.499***
2.123***
2.152***
2.385***
1.685***
(0.158)
(0.175)
(0.189)
(0.158)
(0.150) (continued)
136
5 Relative Economic Status and the Mental Health Status Among Chinese …
Table 5.8 (continued) Variable
Dependent variable: CES-D score (1)
(2)
(3)
(4)
(5)
Compared to relatives
Compared to schoolmates
Compared to colleagues
Compared to neighbors or the villagers
Compared to people in the city or county
1.181
0.818***
0.849***
−0.212
0.702**
(0.618)
(0.177)
(0.176)
(0.418)
(0.256)
Other control variables
Yes
Yes
Yes
Yes
Yes
N
6414
6414
6414
6414
6414
R2
0.168
0.147
0.144
0.164
0.140
Missing value
Note Other settings in regressions are same as those in Table 5.2
Table 5.9 OLS regressions results based on the married sample Variable
Dependent variable: CES-D score (1)
(2)
(3)
(4)
(5)
Compared to relatives
Compared to schoolmates
Compared to colleagues
Compared to neighbors or the villagers
Compared to people in the city or county
The relative economic status (base group: “About the same”) Better
−0.982***
−0.807*
−1.247***
−1.145***
−0.613
(0.255)
(0.369)
(0.347)
(0.262)
(0.395)
Worse
2.487***
2.171***
2.123***
2.227***
1.677***
(0.180)
(0.224)
(0.251)
(0.183)
(0.183)
N
4.345
2.730
2.294
4.261
3.941
R2
0.173
0.166
0.166
0.162
0.148
Note The five columns present results of the five regression models for the relative economic status compared to relatives, schoolmates, colleagues, neighbors or the villagers, and people in the city or town, respectively. Other control variables include age (and age square), registered permanent residence (hukou), married, education level, working status, chronic disease, physical disability, family size, family annual income per capita, and province dummies. Clustered standard errors are in parentheses. *p < 0.05 **p < 0.01 ***p < 0.001
References Akay A, Bargain O, Zimmermann KF (2012) Relative concerns of rural-to-urban migrants in China. J Econ Behav Organ 81(2):421–441 Andresen EM, Malmgren JA, Carter WB, Patrick DL (1994) Screening for depression in well older adults: evaluation of a short form of the CES-D. Am J Prev Med 10(2):77–84 Andrews G, Titov N (2007) Depression is very disabling. Lancet 370:808–809 Blanchflower DG, Oswald AJ (2004) Well-being over time in Britain and the USA. J Public Econ 88(7–8):1359–1386
References
137
Brockmann H, Delhey J, Welzel C, Yuan H (2009) The China puzzle: falling happiness in a rising economy. J Happiness Stud 10(4):387–405 Chesney E, Goodwin GM, Fazel S (2014) Risks of all-cause and suicide mortality in mental disorders: a meta-review. World Psychiatry 13(2):153–160 Clark AE, Senik C (2010) Who compares to whom? The anatomy of income comparisons in Europe. Econ J 120(544):573–594 Compbell A, Converse PE, Rodgers WL (1976) The quality of American life: perceptions, evaluations, and satisfactions. Russell Sage Foundation, NY Daly MC, Wilson DJ, Johnson NJ (2013) Relative status and well-being: evidence from U.S. suicide deaths. Rev Econom Stat 95(5):1480–1500 Dumludag D (2013) Life satisfaction and income comparison effects in Turkey. Soc Indic Res 114(3):1199–1210 Easterlin RA (1974) Does economic growth improve the human lot? Some empirical evidence. In: Growth PD, Reder M (ed) Published in nations and households in economic. Academic Press, New York Easterlin RA (1995) Will raising the incomes of all increase the happiness of all? J Econ Behav Organ 27(1):35–47 Everson SA, Maty SC, Lynch J, Kaplan GA (2002) Epidemiologic evidence for the relation between socioeconomic status and depression, obesity, and diabetes. J Psychosom Res 53(4):891–895 Ferrer-i-Carbonell A (2005) Income and well-being: an empirical analysis of the comparison income effect. J Public Econ 89:997–1019 Filho AD, Kawachi I, Wang Y, Viana MC, Andrade LH (2013) Does income inequality get under the skin? A multilevel analysis of depression, anxiety and mental disorders in São Paulo, Brazil. J Epidemiol Community Health 67(11):966–972 Firebaugh G, Schroeder MB (2009) Does your neighbor’s income affect your happiness? Am J Sociol 115(3):805–831 Frank RH (1990) Luxury fever: why money fails to satisfy in an era of excess. Free Press, NY Guan H (2010) The impact of income on happiness: absolute and relative measures. Nankai Econom Stud 5:56–70 (In Chinese) Hsieh C, Qin XZ (2018) Depression hurts, depression costs: the medical spending attributable to depression and depressive symptoms in China. Health Econ 27(3):525–544 Kahneman D, Krueger AB (2006) Developments in the measurement of subjective well-being. J Econom Perspect 20(1):3–24 Knight J, Song L, Gunatilaka R (2009) Subjective well-being and its determinants in rural China. China Econ Rev 20(4):635–649 Koster A, Bosma H, Kempen GI, Penninx BW, Beekman AT, Deeg DJ, Van Eijk JT (2006) Socioeconomic differences in incident depression in older adults: the role of psychosocial factors, physical health status, and behavioral factors. J Psychosom Res 61(5):619–627 Lang K, Lehmann JK (2012) Racial discrimination in the labor market: theory and empirics. J Econom Literat 50(4):959–1006 Lei XY, Sun XT, Strauss J, Zhang P, Zhao YH (2014) Depressive symptoms and SES among the mid-aged and elderly in China: evidence from the China health and retirement longitudinal study national baseline. Soc Sci Med 120:224–232 Li N, Chen G, Du W, Song X, Zhang L, Zheng, X (2011) Population-level prevalence estimate and characteristics of psychiatric disability among Chinese adults. J Psychiatr Res 45(11): 1530–1534 Lindell MK, Whitney DJ (2001) Accounting for common method variance in cross-sectional research designs. J Appl Psychol 86(1):114–121 Lorant V, Deliege D, Eaton WW, Robert A, Philippot P, Ansseau M (2003) Socioeconomic inequalities in depression: a meta-analysis. Am J Epidemiol 157(2):98–112 Luo CL (2009) Absolute income, relative income and subjective well-being: empirical test based on the sample data of urban and rural households in China. J Fin Econom 11:79–91 (in Chinese)
138
5 Relative Economic Status and the Mental Health Status Among Chinese …
McBride M (2001) Relative income effects on subjective well-being in the cross section. J Public Econ 45(3):251–278 Meltzer H, Bebbington P, Brugha T, Jenkins R, Mcmanus S, Stansfeld S (2010) Job insecurity, socio-economic circumstances and depression. Psychol Med 40(8):1401–1407 Muramatsu N (2003) County-level income inequality and depression among older Americans. Health Serv Res 38(6):1863–1883 Oshio T, Kobayashi M (2011) Area-level income inequality and individual happiness: evidence from Japan. J Happiness Stud 12(4):633–649 Oshio T, Urakawa K (2014) The association between perceived income inequality and subjective well-being: evidence from a social survey in Japan. Soc Indic Res 116(3):755–770 Oshio T, Nozaki K, Kobayashi M (2011) Relative income and happiness in Asia: evidence from nationwide surveys in China, Japan, and Korea. Soc Indic Res 104(3):351–367 Phillips MR (2004) Mental health in China: challenge and choice in the 21st century. Chinese J Nervous Mental Diseases 30:1–10 (in Chinese) Phillips MR, Zhang J, Shi Q et al (2009) Prevalence, treatment, and associated disability of mental disorders in four provinces in China during 2001–05: an epidemiological survey. Lancet 373(9680):2041–2053 Pickett KE, Wilkinson RG (2015) Income inequality and health: a causal review. Soc Sci Med 128:316–326 Qin X, Wang S, Hsieh C (2016a) The prevalence of depression and depressive symptoms among adults in China: estimation based on a national household survey. China Econ Rev 51:271–282 Qin XZ, Wang SY, Hsieh C (2016b) The prevalence of depression and depressive symptoms among adults in China: estimation based on a national household survey. China Econom Rev, In Press https://doi.org/10.1016/j.chieco.2016.04.001 Radloff LS (1977) The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol 1(3):385–401 Ren J, Ma AX, Liu GG (2014) Structure of simplified version of CES-D among middle aged and elderly population. Modern Bus Trade Industry 10:77–79 (in Chinese) Schor JB (1991) The overworked American: the unexpected decline of leisure. Basic Books, NY, ISBN: 0-465-05433-1 Senik C (2009) Direct evidence on income comparisons and their welfare effects. J Econ Behav Organ 72(1):408–424 Sicular T, Yue XM, Gustafsson B, Li S (2007) The urban-rural income gap and inequality in China. Rev Income Wealth 53(1):93–126 Wilkinson RG (1992) National mortality rates: the impact of inequality? Am J Public Health 82(8):1082–1084 Witoelar F, Strauss J, Sikoki B (2012) Socio-economic success and health in later life: evidence from the Indonesia family life survey. In: Majmundar M, Smith JP (eds) Aging in Asia: findings from new and emerging data initiatives, committee on policy research and data needs to meet the challenge of aging in Asia, national research council. National Academies Press, Washington D.C. World Health Organization (2012) Global burden of mental disorders and the need for a comprehensive, coordinated response from the health and social sectors at the country level. Document A65/10 Yin ZC, Gan L (2014) Smoking, drinking, and earnings in China. J Econ Res 10: 90–100. (in Chinese) Yin ZC, Song QY, Wu Y, Peng CY (2015) Financial literacy, entrepreneurial motivation and entrepreneurial decision. Manage World 1:87–98 (in Chinese) Zhang J, Xiao SY, Zhou L (2010) Mental disorders and suicide among young rural Chinese: a case-control psychological autopsy study. Am J Psych 167(7):773–781 Zhang J, Wieczorek WF, Conwell Y, Tu XM (2011) Psychological strains and youth suicide in rural China. Soc Sci Med 72(12):2003–2010
References
139
Zhang J, Lu J, Zhao S, Lamis DA, Li N, Kong Y, Jia C, Zhou L, Ma Z (2014) Developing the psychological strain scales (PSS): reliability, validity, and preliminary hypothesis tests. Soc Ind Res 115(1):337–361 Zhang XZ, Cai GW (2011) Income, values and residents’ subjective well-being: empirical evidence from adult survey data in Guangdong. Manage World 9:63–73 (in Chinese) Zhao Y, Strauss J, Yang G, Giles J, Hu P, Hu Y, Lei X, Park A, Smith JP, Wang Y (2013) China Health and Retirement Longitudinal Study-2011–2012. National Baseline Users Guide. China Center for Economic Research, Peking University Zhao YH, Hu Y, Smith JP, Strauss J, Yang G (2014) Cohort profile: the China health and retirement longitudinal study. Int J Epidemiol 43(1):61–68 Zhou Q, Qin X, Liu GG (2020) Relative economic status and mental health among Chinese adults: evidence from the China health and retirement longitudinal study. Rev Dev Econ 24(4):1312– 1332 Zimmerman FJ, Katon W (2005) Socioeconomic status, depression disparities, and financial strain: what lies behind the income depression relationship? Health Econ 14(12):1197–1215
Chapter 6
Understanding and Addressing the Treatment Gap in Mental Healthcare: Economic Perspectives and Evidence from China
6.1 Introduction The prevalence of mental disorders such as depression has been increasing worldwide in the past decades (Vos et al. 2012). Around the world, more than 650 million people live with diagnosable mental disorders (WHO 2020). As a result, many studies have paid attention to the social consequences of the rising prevalence of mental illnesses, attempting to estimate the magnitude and the channels of such impacts. A general conclusion drawn from the existing studies is that mental disorders impose considerable costs (both direct and indirect costs) on the individuals, their families, and the society. In developed countries like the United States, estimation has shown that the economic burden of depression was as high as $210.5 billion in 2010 (Greenberg et al. 2015a, b). In developing countries such as China, it has been found that the costs of medical expenditures due to depressive mental illnesses account for 14.7% of China’s total medical spending in 2012 (Hsieh and Qin 2018). The deterioration of mental health has a significant negative impact on the patient’s labor force participation, which in turn reflects the opportunity (indirect) costs of depression in labor markets. Moreover, the socioeconomic costs of mental disorders may go well beyond the traditional boundary of disease burden estimation, presenting implications on the reduction of social trust, self-esteem, and life satisfaction (Hsieh et al. 2019). Compared to the rich evidence on the social consequences of mental disorders, there are relatively few studies that investigate the issue of low treatment rate in mental illness. In comparison with other non-communicable diseases (NCDs) with high disease burden and prevalence rate (such as hypertension and diabetes), the rate of treatment for mental disorders are quite low, indicating a large treatment gap, measured by the difference between the need for and the actual provision of treatment among patients with mental illness. It was estimated that at least 10% of the global population is affected by one or more mental disorders; however, according to the The content of this chapter is published in Qin and Hsieh (2020). Copyright Qin and Hsieh (2020). © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 X. Qin and C.-R. Hsieh, Economic Analysis of Mental Health in China, Applied Economics and Policy Studies, https://doi.org/10.1007/978-981-99-4209-1_6
141
142
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
estimation by the World Health Organization, more than three-quarters of people with severe mental disorders in low- and middle-income countries (LMICs) receive no medical treatment. Although the situation in high-income countries is better, there is still a high proportion, ranging from 35 to 50%, of people with mental illness who go untreated. The questions that this paper tries to answer are (1) why so many people with mental illness go untreated, and (2) how to bridge the treatment gap for them. We first provide economic perspectives to explain the potential causes of treatment gap in mental healthcare generally. Specifically, we propose a testable hypothesis that patients with mental illness face four major hurdles in accessing appropriate care, including stigma, high out-of-pocket payment, low availability of mental health resources, and the slow diffusion of new medical knowledge and technology. We then use China as a case study to show the evidence in support of this four-hurdle hypothesis, and we propose four policy options to bridge the treatment gap in mental healthcare. Although our empirical evidence and policy discussion lie in the context of China’s health system, our findings also have important implications for other low- and middle-income countries (LMICs) with similar development experience and challenges in the healthcare sector. Our work contributes to the growing body of research on the mental health policy designs in LMICs. Many studies have demonstrated a substantial treatment gap in mental health services (Kohn et al. 2004; Knapp et al. 2006; Patel et al. 2016), however, little research has investigated the potential determinants of the treatment gap in a systematic way. Several international agencies have identified major roadblocks to receiving treatment in mental health, including stigma, inadequate funding, and poor design of health system (WHO 2016a, b). However, there is limited research that unpacks the key factors that shape such roadblocks, especially in LMICs. Our research aims to provide a synthesis for the various academic endeavors and policy discussions on how to help people with mental illness get out of the shadow and receive appropriate diagnosis and treatment. It provides a relatively general framework by integrating institutional analysis with economic analysis of healthcareseeking behavior to achieve a better understanding on the potential causes of the treatment gap. Based on this analytical framework, we then propose several policy options on bridging the treatment gap in the mental health sector. The rest of the paper is organized as follows. Section 6.2 provides a conceptual framework to develop our testable four-hurdle hypotheses. Section 6.3 presents empirical evidence to support the existence of the four hurdles in accessing mental healthcare, using the current available data in China. Section 6.4 discusses four policy options for bridging the treatment gap in mental healthcare. The last section concludes the paper and discusses the implications of the findings.
6.2 Conceptual Framework
143
6.2 Conceptual Framework As first conceptualized by Michael Grossman, the demand for personal healthcare services is derived from the person’s demand for health, which in turn is determined by the optimal health capital stock at which the marginal cost of holding health capital equals the marginal efficiency (benefit) of health capital (Grossman 1972). Following this framework, the economic theory of healthcare-seeking behavior predicts that individuals will seek healthcare for their mental illness as long as the expected benefit of doing so exceeds its expected cost. Based on this framework, the low treatment rate of mental illness can be explained by two potential reasons: the cost is too high and/or the benefit is too low. A closer look of mental healthcare delivery indicates that the costs of seeking treatment include both monetary and nonmonetary costs. The nonmonetary cost is to a large extent due to the stigma associated with mental illness. The full monetary costs that the patients pay for receiving treatment can be further divided into two parts: (1) the financial prices of mental healthcare as reflected by the patient out-of-pocket payment and (2) the time prices of seeking mental healthcare as reflected by the opportunity cost of a patient’s time allocated to traveling, waiting, and receiving treatment (Sloan and Hsieh 2017). Compared to physical illness, the most distinctive feature associated with mental illness is stigma (Frank and McGuire 2000). Stigma indicates the co-occurrences of the following five components: labeling, stereotyping, separation, status loss, and discrimination (Link and Phelan 2001). In the first component, people distinguish and label the differences of persons with mental illness. In the second, negative stereotyping, such as unpredictable, unstable, dangerous, violent, and socially worthless, often surrounds the images of people with mental illness (McSween 2002). As a result, the stigma of mental illness creates a position of social distance or rejection (Link 1987). Specifically, negative consequences of stigma include a decrease in the opportunity of seeking employment and housing, an increase in family stress, and the lower quality of life. The fear of status loss and discrimination in turn becomes the internal cost and a major barrier for people with mental illness to overcome when they seek medical treatment. This implies that the non-monetary cost imposed by social stigma is the first hurdle in the access to mental healthcare. The second hurdle in seeking mental healthcare is the money price in the form of patient out-of-pocket payments for mental healthcare utilization. As shown in the WHO report, many countries (especially LMICs) suffer from the under-funding problem in the sense that the public sector allocates an extremely low share of health budgets into the mental health sector (WHO 2016a, b). According to a recent survey conducted by WHO, governments spend on average 3% of their health budgets on mental health (a figure much lower than the non-mental health sector such as hypertension and diabetes), with a distribution from 0.5% in the low-income countries to 5.1% in the high-income countries (WHO 2014a, b). A natural consequence of the under-funding problem is that patients with mental illness typically need to pay a higher out-of-pocket amount to finance their treatment compared to their counterparts with physical illness.
144
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
Thirdly, many countries do not have sufficient healthcare resources (including mental health personnel and facilities) to deliver the appropriate care to people with mental illness. The low availability of mental health resources is mainly reflected in two dimensions: (1) inadequate capacity building in the training of mental health professionals, which leads to the overall insufficient supply in mental healthcare; (2) the limited resources for mental healthcare, including both professionals and facilities, are usually concentrated in the densely populated urban areas within a country, indicating an uneven geographic distribution of mental healthcare resources. One of the significant consequences of the insufficiency and maldistribution of healthcare resources is the increase in the time cost for seeking mental healthcare, which in turn becomes the third hurdle for people suffering from mental health conditions, such as depression. Finally, the fourth hurdle in accessing mental healthcare is the low expected benefits of treatment arising from the technology gaps between the frontier of new knowledge in treatment procedures and the clinical practice available to patients. Although there has been a rapid development in medical knowledge and technology for mental healthcare in recent years, whether the frontier of this new technology and knowledge can be transmitted to become a prevailing local practice standard depends on the speed of knowledge diffusion and technology adoption. Many studies have shown the evidence that the incentives for innovation in general and the technology diffusion in particular are positively correlated with the market size (Acemoglu and Linn 2004; Berndt et al. 2014). As mentioned, many countries face common challenge of inadequate funding in their mental health sectors, indicating that mental healthcare has a relatively smaller market size compared to that of the general healthcare. As a result, mental health sector is in disadvantage in facilitating technology diffusion such as the launch of new prescription drugs and the provision of psychological treatments for mental illness (e.g. cognitive-behavioral therapy for mild depression). This in turn enlarges the gap between the frontier of treatment know-how and the local practice standards in those countries. The existence of such a knowledge gap and outdated clinical practice may reduce the potential benefits of mental health treatment, which in turn further decreases the incentive for the patients to seek medical assistance when in need. In summary, the above analyses indicate that people with mental illness face higher marginal costs of accessing mental healthcare than other patients, including the psychological cost imposed by social stigma, the out-of-pocket cost arising from the low public funding, the time cost due to the poor availability of mental healthcare resources. In addition, the perceived benefits of medical treatment may be lower due to the slow diffusion of new medical knowledge and technology. We hypothesize that these four hurdles largely explain why many people with mental illnesses tend to delay the treatment or go completely undiagnosed, a stylized fact in epidemiological studies in many countries (Bor 2015). In the next section, we use China as an example to show the evidence for this four-hurdle hypothesis.
6.3 Empirical Evidence from China
145
6.3 Empirical Evidence from China 6.3.1 Stigma WHO has pointed out the long-term negative effects of stigma, highlighting that stigma, as a major source of discrimination and exclusion, can damage people’s selfesteem, disrupt their family relationships, and consequently limit their ability and willingness to socialize, obtain housing and seek employment. Table 6.1 presents the comparative statistics among adult groups of different mental health status in China based on data from the China Family Panel Studies (2012), a nationally representative household survey. We classify the respondents’ mental health status based on their Center for Epidemiologic Studies Depression Scale (CES-D) scores. The table shows a significant correlation between the respondents’ mental health status and their ideology and social economic status. People with mental depression (CES-D at 28 or higher) or depressive symptoms (CES-D between 16 and 27) are shown to have significantly lower life satisfaction on their family and themselves, less selfconfidence and lower trust toward family members and other social groups. 58.4% of the mentally healthy people (CES-D at 15 or lower), compared to only 37.5% of people with depression, tend to believe that most people are trustworthy. In fact, stigma has been reported to hamper the prevention and treatment of mental health disorders and the promotion of mental well-being, which in turn results in poorer physical health, suicidality, and higher mortality rates (WHO 2020; Thornicroft 2007). Stigma can increase the feelings of worthlessness and despondency that increase the risk of depression and suicide, and potentially more damaging than the mental illness itself (Eagles et al. 2003). For example, based on interviews with close associates of people who committed suicide and of people who died from other injuries in China, it has been found that a high depression symptom score remains a significant predictor for suicidality after adjusting for sex, age, residential location, and other factors (Phillips et al. 2002). What’s more, the negative impacts of stigma are likely to extend from the daily life of patients to that of their family members, friends, and even mental health provider groups, implying a negative spill-over effect of social stigma (Corrigan et al. 2005). The long-term consequences of stigma among the mentally ill may also extend from health outcomes to labor market outcomes such as employment and income. Based on the National Co-morbidity Survey-Replicate (NCS-R) data, evaluations have shown that psychiatric disorders are associated with reductions of 9 and 14% of the labor force participation rate and the employment rate for male (Chatterji et al. 2011). It has also been found that depression leads to an annual work loss of about 1.4 days (accounting for 33% of total health-related workday loss) (Peng et al. 2015). The CFPS 2012 dataset provides additional evidence of stigma in China’s labor market: for example, the years of education and the levels of personal income are shown to be significantly lower for people with mental depression or depressive symptoms compared to the mentally healthy respondents. On average, individuals without mental illness acquire 7.6 years of formal schooling, which is almost twice as
146
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
Table 6.1 Differences in psychological and socioeconomic characteristics among three mental health groups in China, 2012 Mentally healthy
Depressive symptoms
Severe depression
3.645
3.197
2.754***
(0.976)
(1.053)
(1.218)
Life satisfaction and confidence Satisfaction of one’s family (from 1 to 5)
Satisfaction of one’s life (from 3.485 1 to 5) (0.99)
3.037
2.618***
(1.049)
(1.202)
Social status of oneself (from 1 to 5)
2.745
2.554
2.391***
(0.982)
(1.059)
(1.213)
Degree of confidence to one’s future (from 1 to 5)
3.874
3.367
2.736***
(1.006)
(1.153)
(1.361)
Tendency to trust other people Most people are trustworthy (1 0.584 = yes; 0 = no) (0.493)
0.482
0.375***
(0.5)
(0.484)
Do you trust your parents (from 0 to 10)
9.278
8.881
8.448***
(1.485)
(1.828)
(2.288)
Do you trust your neighbor (from 0 to 10)
6.567
6.015
5.65***
(2.138)
(2.241)
(2.638)
Do you trust the doctors (from 2.189 0 to 10) (6.686)
2.29
2.642***
(6.362)
(6.01)
Do you trust the cadres (from 0 to 10)
2.424
2.481
2.857***
(4.924)
(4.68)
(4.529)
Do you trust strangers (from 0 2.264 to 10) (2.14)
2.045
1.943***
(2.066)
(2.257)
Do you trust the American
2.526
2.426
2.678***
(2.564)
(2.439)
(2.465)
7.625
5.958
3.992*** (4.627)
Labor market outcomes Years of education by 2012
(4.764)
(4.953)
Personal annual income (in 1000 Yuan)
13.42
8.694
4.703***
(36.95)
(25.23)
(10.21)
Observations
16,503
6104
1114
Notes (1) Data Source: China Family Panel Studies (2012). (2) Mental health status (Mhs) is divided into three groups: mentally healthy group, group with depressive symptoms, and group suffering from severe depression, which are categorized using the CES-D score (mentally healthy = CES-D at 15 or lower; depressive symptoms = CES-D between 16 and 27; depression = CES-D at 28 or higher). (3) The statistics reported are sample means within each mental health status group, with standard deviation in parenthesis. ANOVA test with the null hypothesis that the mean values of different mentally health status groups are the same is provided. *** denotes statistical significance at 1% leve.
6.3 Empirical Evidence from China
147
much as the average education years of people with depression. The annual income of individuals without mental illness is 50% higher than those with depressive symptoms and triple that of people with depression (see Table 6.1).
6.3.2 Out-of-Pocket Costs of Mental Healthcare Although China has made significant progress in achieving the goal of universal healthcare coverage, the current system contains more than 3000 local health insurance plans that vary substantially in eligibility criteria, insurance benefits, and copayment schemes (Meng et al. 2015). More specifically, different health insurance plans, such as the New Rural Cooperative Medical Scheme (NCMS, a governmentsubsidized plan covering all rural families), the Urban Employee Basic Medical Insurance (UEBMI, a social insurance program financed by employers and employees covering urban workers in the formal sectors) and the Urban Resident Basic Medical Insurance (URBMI, an urban health insurance scheme covering informal sector workers and people without employment), differ in their enrollment criteria and co-insurance rates. Within each plan, coinsurance and copayment rates also differ by regions and types of treatment. Generally speaking, the copayment rates for outpatient visits are higher than those for inpatient admissions in China, especially for rural health insurance programs. Before 2012, the insurance coverage and reimbursement for mental healthcare are usually limited and dependent on the provincial government’s financial capacity. In 2012, the Chinese central government announced a decision to expand the coverage of the country’s health insurance system to include the treatment of critical illnesses including major mental diseases. Meanwhile, the Mental Health Law of China was launched in 2013, which formalizes the legal protection and treatment of people with mental disorders (Qin et al. 2018). After these milestone steps in strengthening mental healthcare, a significant portion of mental health outpatient and inpatient medical expenses were able to be covered by the national health insurance system. For example, Beijing covered six types of major mental illnesses (e.g. schizophrenia, bipolar disorder) in its insurance plan in 2014 and increased the reimbursement rates for the inpatient and outpatient healthcare for these conditions from 60 to 70% with no maximum payment limits. In addition, the essential drugs for the treatment of these major mental diseases are also made free to outpatients, which benefited more than 12,000 mental health patients by 2016. Shanghai included four mental diseases into its Critical Disease Insurance Plan in 2015, which provides a 50% reimbursement rate (increased to 55% in 2017) in supplement to the basic health insurance coverage. Several cities in China’s eastern coastal provinces, including Jinan, Zhanjiang, Foshan, and Dongguan also added mental diseases into their health insurance coverage in 2015. The city of Shenzhen covered six mental diseases in 2016 with maximum reimbursement rates of up to 90%. In the rural sector, NCMS started to launch pilot programs to cover mental diseases and other critical illnesses in 2013. Meanwhile, the average government financial support for NCMS increased from 320
148
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
RMB (about $53) per person in 2013 to 450 RMB (about $75) per person in 2017. However, there remains substantial variation in the reimbursement rates for mental illnesses across regions within NCMS. Despite the above-mentioned progress in extending health insurance coverage for mental health patients, China still suffers from a serious disparity in the coverage and reimbursement rates for mental diseases. High-income areas such as the eastern coastal regions and major urban cities usually have better coverage as well as higher reimbursement rates. For example, based on the data from 1989–2011 China Health and Nutrition Survey (CHNS), we estimate the effective reimbursement rates (total medical expenditure less the patient out-of-pocket payment) for mental healthcare and non-mental healthcare in China’s eleven provinces (see Fig. 6.1). As indicated, patients with mental health problems in the eastern provinces have a significantly higher effective reimbursement rate compared to their counterparts in the Northeast, Central, and West. In addition, compared with the reimbursement rates of physical conditions such as heart disease, tumor and respiratory diseases, the mental illness patients that are surveyed in the 1989–2011 CHNS received a lower average reimbursement rate (10.46 vs. 16.61%). Therefore, there is substantial variation not only in insurance coverage across regions, but also in the depth of benefits between the general healthcare and mental healthcare. Given such disparity of financial support from health insurance plans, people with mental health problems in lower-income areas would face higher out-of-pocket burden, which in turn deters the proper use of treatment (Lambregts and Vliet 2018).
6.3.3 Mental Healthcare Resources Much of the access barriers for mental healthcare in China are due to the limited supply and unequal distribution of professional mental healthcare resources. For example, China only had 1.46 psychiatrists per 100,000 population in 2010, which was substantially below the global average mental health workforce (4.15 psychiatrists per 100,000 population) (Liu et al. 2011; Qian 2012). The lack of qualified mental health professionals may be partially due to the government control of medical education and accreditation, and it may also be attributable to the severe under-diagnosis of mental illnesses that results in the mismatch between supply and potential demand of mental healthcare. In China, over 100 million Chinese experience different kinds of mental disorders during a year, and these mental diseases account for over 20% of the total burden of diseases (Fan et al. 2013). Given the high prevalence rate of depression (4.08%) in China, the medical resource of mental healthcare is relatively scarce compared to the general healthcare (Qin et al. 2018). In addition to the overall undersupply of manpower, geographic maldistribution of available mental health resources in China and the concentration of qualified personnel in the urban-based psychiatric hospitals indicate that mental health services are quite limited in rural areas (Phillips et al. 2009).
6.3 Empirical Evidence from China
149
Fig. 6.1 Regional variation in the effective reimbursement rates for mental healthcare in eleven provinces of China (1989–2011). Notes (1) Data source: China Health and Nutrition Survey (CHNS, 1989–2011). (2) Effective reimbursement rate is calculated as the payment paid by the public payers (total medical spending less the patient out-of-pocket payment) expressed as a percentage of total medical spending for the most recent treatment for mental health conditions. (3) Caution: only 4 provinces in East China (Beijing, Shanghai, Jiangsu, Shandong), 2 provinces in Northeast China (Liaoning, Heilongjiang), 3 provinces in Middle China (Henan, Hubei, Hunan), and 3 provinces in West China (Guangxi, Guizhou, Chongqing) are covered in the sample. Sample may not be nationally representative
To illustrate the above points, Table 6.2 compares the mental healthcare resources and general healthcare resources between 2010 and 2015 in China. A cross-sectional comparison indicates that in 2015, the number of licensed doctors in the mental healthcare sector contributes to only 0.9% of the total supply of licensed doctors, and the number of hospital beds in the mental health sector accounts for only 1.1% of total hospital bed capacity in China. A time series comparison indicates that while the physician density of general healthcare has increased from 18.0 per 10,000 population in 2010 to 22.2 in 2015, the density of licensed mental healthcare physicians decreased from 0.234 per 10,000 population in 2010 to 0.199 in 2015. In contrast with the stable growth in the density of general healthcare doctors, the growth rate of licensed doctors in the mental health sector has fluctuated between 21.79% and 6.32% in recent years. The annual growth rate of hospital beds in mental health is also significantly lower than that in the general healthcare until 2014. One of the plausible reasons for the undersupply of mental healthcare manpower is that the profession offers less attractive wage payment and working environment compared to other specialties of medical care. In China, healthcare price is rigidly regulated at a very low level, and for a long time, doctors must rely on the kickbacks from drug prescriptions to compensate for their low earnings. In such a setting, certain specialists such as psychiatrists and pediatricians are at disadvantage, as they have very small room for drug over-prescription due to the nature of the specialties (Du
150
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
Table 6.2 Capacity and annual growth rate in healthcare resources in China: mental health sector versus general healthcare (2010–2015)
Number of licensed doctors (per 10,000 population)
Healthcare sector
2010
General healthcare
18.0
Mental healthcare
0.234
2011 18.3
2012
2013
2014
2015
19.4
20.4
21.2
22.2
0.183
0.174
0.185
0.190
0.199
Growth rate of General licensed doctors (%) healthcare
1.67
6.01
5.15
3.92
4.72
Mental healthcare
−21.79
−4.92
6.32
2.70
4.74
38.36
42.40
45.50
48.45
51.12
Number of hospital beds (per 10,000 population)
General healthcare Mental healthcare
0.45
0.48
0.49
0.54
0.58
Growth rate of hospital beds (%)
General healthcare
7.27
10.53
7.31
6.48
5.51
6.67
2.08
10.20
7.41
Mental healthcare
35.76
Notes (1) Data Source: Health Statistical Yearbook of China (2011–2016), National Bureau of Statistics of the People’s Republic of China. (2) The statistics reported are density of licensed doctors, density of hospital beds, and their annual growth rate from 2010 to 2015 for general healthcare sector and mental healthcare sector, respectively. General healthcare includes mental healthcare and other specialty such as internal medicine, paediatrics, and gynaecology.
and Zhu 2016; Zhu 2016). Thus, mental health professionals in China commonly earn less than their counterparts in other specialties. Table 6.3 presents the service revenue, service costs, and the implied gross profit rates of different specialty hospitals in China based on the public data in the national health statistical yearbook of 2016. Compared to the profit-generating specialties such as plastic surgery (83.3%), ophthalmology (52%), rehabilitation (42.4%), and psychiatric hospitals (16.8%) rank comparatively low in the profit rates in year 2015, despite their relatively high annual revenue of 29.6 million Yuan per hospital (Yang 2016). Given that most hospitals in China rely on their own profits for daily operation and physician employment, the above comparison indicates that the prospective income is lower for mental health doctors compared to doctors in other specialties, which suggests that the mental health profession may fail to attract sufficient personnel in the long term. In addition to the insufficiency of overall mental healthcare capacity, the geographic maldistribution of available mental health resources in China is also pronounced. Figure 6.2 maps the provincial density of hospital beds in psychiatric services in 2015. The figure indicates a dramatic disparity in mental healthcare resources across the country: the economically developed eastern provinces such as Shanghai and Zhejiang enjoy higher densities of psychiatric hospital beds, while the economically less developed inland regions in Central and Western China are in
6.3 Empirical Evidence from China
151
Table 6.3 Estimated profit rates of specialty hospitals in China, 2015 Number of Average medical service hospitals revenue (1000 yuan)
Average medical service costs (1000 yuan)
Profit rate (%)
Cosmetic hospital
228
19,649
8,228
138.8
Plastic surgery hospital
57
19,850
10,828
83.3
Ophthalmic/eye hospital
455
28,825
18,964
52.0
Rehabilitation hospital
453
12,354
8,675
42.4
Stomatological hospital
501
24,173
17,146
41.0
Others
1290
17,135
12,658
35.4
Hospital of dermatology
163
13,274
9,923
33.8
Obstetrics and gynecology hospital
703
27,640
20,878
32.4
Orthopaedic hospital
558
18,145
14,287
27.0
ENT hospital
89
24,061
19,246
25.0
Psychiatric hospital
920
29,606
25,354
16.8
Hematonosis hospital
10
99,115
85,802
15.5
Cardiovascular hospital
79
108,461
94,049
15.3
Occupational disease hospital
16
52,880
45,866
15.3
Tumor hospital
135
372,513
324,137
14.9
Tuberculosis hospital
34
121,546
109,549
11.0
Children’s hospital
114
236,575
218,898
8.1
Chest hospital
20
280,775
263,850
6.4
Leprosy hospital
31
4,765
4,670
2.0 (continued)
152
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
Table 6.3 (continued) Number of Average medical service hospitals revenue (1000 yuan)
Average medical service costs (1000 yuan)
Profit rate (%)
Hospital for infectious diseases
167
99,161
98,459
0.7
Specialty hospital
6023
38,811
31,977
21.4
Notes (1) Data Source: Health Statistical Yearbook of China (2016). (2) Statistics on medical service revenue and costs reflect the average revenue and costs per hospital for the particular medical specialty in year 2015 (Average medical service revenue = total medical service revenue/number of hospitals; Average medical service costs = total medical service costs/number of hospitals.); statistics for psychiatric hospitals and specialty hospitals are shown in bold. (3) The profit rates are based on the authors’ calculation. Profit rate = (average medical services revenue—average medical service costs)/average medical service costs
dire need of mental healthcare resources. The most underdeveloped provinces such as Qinghai, Gansu, Ningxia, and Guizhou have extremely low densities of hospital beds for professional mental health treatment. Given that the prevalence rates of depression and depressive symptoms are also higher in central and western provinces (Qin et al. 2018), the above findings indicate that the inland regions suffer from the most severe problem of unmet mental healthcare needs.
6.3.4 Diffusion of New Medical Knowledge and Technology in Mental Healthcare Under the current practices in China’s healthcare sector, two institutional features may enlarge the technology gaps in the field of mental healthcare. First, due to the lack of government subsidy for low service fees charged by public hospitals, healthcare providers in China rely heavily on profits obtained from prescription drugs as their major sources of revenue, indicating that physicians may choose to prescribe drugs based not on efficacy, safety or cost, but solely on the extent of the profit margins that they or their institutions obtain (Yang 2016). Second, given the Essential Drug Policy and the regulated insurance reimbursement schedule, there may be a long delay in the launch of new mental healthcare drugs or treatment procedures in China; as a result, physicians may not be able to prescribe what proves to be the most effective treatment regimes, and this translates to another policy-induced barrier for the mental illness patients in China. In China, the market access process for pharmaceutical products (patented or differentiated generic drugs) is complex and involves the following steps: registration and approval of new drugs, pricing and bidding, reimbursement listing at the local and national level, and at last hospital listing (Burns and Liu 2017). More specifically,
6.3 Empirical Evidence from China
153
Fig. 6.2 Density of Hospital Beds in Psychiatric Departments per 1000 Population in China’s All Provinces, 2015. Data Source Health Statistical Yearbook of China (2016)
provincial bidding is held every two years or so; national reimbursement listing takes place every 4–5 years; another two years’ time is needed for the hospital listing process. As such, for a domestic or multinational pharmaceutical company to launch a new drug in China, it has to wait seven years on average for drug approval, launching, and listing in the target hospitals. Companies are not allowed to sell new drugs on the market until the above process is fully completed. This results in a wide gap in the launch of new and innovative drugs between China and high-income countries such as USA, Japan, and UK. This is illustrated in Fig. 6.3, which shows that the initial market share of new drugs in China (2.5%) is considerably lower than that in USA (56.3%), Japan (12.6%), UK (7.7%), Germany (6.5%), and Korea (3.1%) in year 2015. Given that the knowledge and technology frontier in the mental health treatment witnesses fast expansion in recent years, the above statistics suggest that the mental illness patients in China are less likely to benefit from the most innovative drugs and treatment options compared to their counterparts in the above-mentioned countries. As a result, this system may produce lower expected value for its patients, which in turn reduces the incentives for people with mental health conditions to utilize the system. There is ample evidence to illustrate the slow adoption of mental health drugs in China compared with the high-income countries, taking the United States as an example. First, as Table 6.4 illustrates, among 12 new molecular entities for central nervous system (CNS) diseases (the therapeutic category for mental illness) available
154
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
Percentage of initial market of NMEs
56.3%
12.6% 7.7% 6.5%
3.1% 2.5% 1.8% 1.5% 0.9% 0.9% 0.3%
Fig. 6.3 Initial market of new molecular entities (NMEs) as a percentage of All NME launches for various countries, 2007–2015. Notes (1) Data Source: Constructing a sustainable Chinese Pharmaceutical Innovation Ecosystem (2016), by China Pharmaceutical Enterprises Association, et al. (in Chinese) (2) Percentage of initial market of NMEs = NMEs launched in a certain country as initial market/total number of NMEs marketed globally. Only new molecular entities (NMEs) between 2007 and 2015 are included in the calculation. (3) Launching NMEs as initial market in a country partially indicates the drug R&D strength of the country, thus the percentage of initial market illustrated in the figure partially indicates the relative R&D strength for innovative pharmaceutical products in a country compared to other countries in the world
in the global market, only one was launched in China. By contrast, eight drugs are adopted in the United States. This indicates a difference of 0.583 (8/12 minus 1/12) in the adoption rate between USA and China. Furthermore, the difference in adoption rates between CNS drugs and drugs for other NCDs in China is 0.094, which is higher than that of USA (0.01). Therefore, not only China has a slow adoption of new drugs, but its adoption of drugs for mental illnesses is even slower than that for other noncommunicable diseases. Second, Table 6.5 takes 14 atypical antipsychotic (AAP) drugs (for the treatment of schizophrenia) as examples, and shows the years in which these mental health drugs were approved by the US Food and Drug Administration (USFDA) and China Food and Drug Administration (CFDA). As indicated, there is a significant time lag (an average of 6 years) between the approvals in the two countries, which suggests a long delay in the launch of new pharmaceutical products in China’s mental health sector. Another piece of evidence for the knowledge gap comes from the comparison of clinical guidelines for the first-line drug prescription on mental health conditions between China and the developed countries (see Table 6.6). Clinical guidelines for the treatment of mental depression started to be published in China from 2006, the first edition of which is only five pages long.1 The guideline was still officially 1
Chinese Medical Association (2006).
6.3 Empirical Evidence from China
155
Table 6.4 Estimated availability of new molecular entities (NME) for diseases of central nervous system (CNS) and other NCDs: China versus USA (2008–2012) Country NME/global NME
China USA
Difference (NME in USA—NME in China)
NMEs for CNS Diseases
1/12
0.583
NMEs for Other NCDs
11/62 42/62 0.500
8/12
Difference (Drugs for other NCDs—CNS 0.094 0.010 0.083 (Difference in difference) drugs) Notes (1) Data Source: Global Outlook for Medicines through 2018—IMS Institute for Healthcare Informatics (2014). (2) New molecular entities (NME) include small molecule and biological pharmaceutical products where at least one of the ingredients is novel. The availability of Global NMEs is measured by the number of NMEs with global launch in at least one country between 2008 and 2012. The availability of country NMEs is measured by the number of global NMEs available in a specific country by the end of 2013. (3) CNS drugs are drugs designed for treating illness in central nervous system, which are mainly related to mental health problems. NMEs for other non-communicable diseases (NCDs) include drugs for cardiovascular diseases, diabetes, and tumor
Table 6.5 Time lag between USA and China in the approval of new atypical antipsychotic (AAP) drugs for the treatment of schizophrenia, 1989–2019 AAP drugs
Year approved by USFDA
Year approved by CFDA
Time lag (in years)
Clozapine
1989
2002
13
Risperidone
1993
2000
7
Olanzapine
1996
1998
2
Quetiapine
1997
2000
3
Ziprasidone
2001
2007
6
Aripiprazole
2002
2006
4
Paliperidone
2006
2008
2
Iloperidone
2009
/
/
Asenapine
2009
/
/
Paliperidone palmitate
2009
2018
9
Lurasidone
2010
2019
9
Aripiprazole lauroxil
2015
/
/
Brexpiprazole
2015
/
/
Cariprazine
2015
/
/
Notes (1) Data Source: US Food and Drug Administration; China Food and Drug Administration
recommending the use of TCAs (a category of first-generation antidepressant with considerable adverse drug reaction), while at the same time second-generation antidepressants such as SSRIs and SNRIs had been widely recommended in USA and other developed countries for more than a decade due to their effective treatment and less
156
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
Table 6.6 Major antidepressants and whether they are recommended as first-line therapy for treating depression in different countries Whether recommended as US UK Canada China China Treatment first-line therapy for guideline guideline guideline guideline guideline practice in depressive disorders (2010) (2009) (2016) (2006) (2015) China MAOIs TCAs
Yes
Yes
TeCAs
Yes
Yes
Yes
SSRIs
Yes
Yes
SNRIs
Yes
Yes
Yes
NaSSAs
Yes
Yes
Yes
NDRIs
Yes
Yes
Yes
Yes
Notes (1) Data Source: Practice Guideline for the Treatment of Patients with Major Depressive Disorder (2010) by American Psychological Association; Depression in Adults with a chronic physical health problem: Treatment and Management (2009) by the National Institute for Health and Care Excellence (NICE); Canadian Network for Mood and Anxiety Treatments (CANMAT) 2016 Clinical Guidelines for the Management of Adults with Major Depressive Disorder: Introduction and Methods; Clinical Guidelines for treatment-Psychiatry (2006) by Chinese Medical Association (CMA); Guidelines for the Prevention and Treatment of Depression in China (2015) by Chinese Medical Association (CMA). (2) Drugs listed in the table are major categories of medicines used to treat depression. (3) The last column, treatment practice in China, reflects the main drugs in current usage for the majority of Chinese regions based on the reports in Guidelines for the Prevention and Treatment of Depression in China (2015)
side effects.2 This represents a lag in guideline development between China and developed countries as well as a technology gap in the pharmaceutical industry. The second edition of official guidelines for the treatment of depressive disorders was published in 2015, with much more detailed and up-to-date content, recommending SSRIs, SNRIs, and NaSSAs as first-line treatment options for mental depression.3 However, there still exists a large know-do gap between the official recommendations and the field practices in China, and first-generation therapies such as TCAs and TeCAs were still commonly prescribed by mental health doctors in various regions of China. Other than the regulation-induced barrier to the diffusion of medical knowledge and technology, the persistent under-funding problem of the mental health sector also enlarges the gap between the technology frontier and the local clinical practices in China. Figures 6.4 and 6.5 present the market shares (measured as the number of outpatient visits or inpatient discharges for a specific service type as a percentage of total number of outpatient visits or inpatient discharges) of various types of diseases among China’s medical institutions in 2015. As indicated, both outpatient and inpatient shares of psychiatry (mental health department) account for merely 1% among all types of healthcare services, suggesting that the mental healthcare sector accounts 2 3
American Psychiatric Association (2010). Chinese Medical Association (2015).
6.4 Policy Options for Bridging the Treatment Gap in Mental Healthcare
157
Psychiatry 1% Internal medicine 30%
Oncology 3% General practice 5%
Paediatrics 10%
Traditional Chinese Medicine (TCM) 11%
Gynaecology and obstetrics 12%
Surgery 17%
Fig. 6.4 Percentage of outpatient and emergency visits by types of healthcare services in China, 2015. Notes (1) Data Source: Health Statistical Yearbook of China (2016). (2) Services with percentage less than 1% are not annotated in this figure
for a very small size in the overall healthcare market in terms of patient volumes and service revenues. Given that the public and private funds tend to flow into major sectors with large market sizes (such as internal medicine and traditional Chinese medicine), the under-funding problem is expected to plague China’s mental health sector in the foreseeable future and in turn reduce the speed of technology adoption in the field. The vicious cycle of under-funding and under-treatment is thus exacerbated by the gap, leading to further reduction in the effectiveness of mental healthcare services in China.
6.4 Policy Options for Bridging the Treatment Gap in Mental Healthcare Given the evidence that high costs and low benefits are two main causes of undertreatment in mental healthcare, we offer two approaches to bridge the treatment gap: the “push incentives” and the “pull incentives”, which are designed to reduce the costs of treatment and to increase the benefits of treatment, respectively. For push incentives, we suggest three policy options to reduce the nonmonetary cost, out-of-pocket cost, and time cost in the mental healthcare-seeking process. For pull incentives, we suggest using the information and communication technology (ICT)
158
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
Ophthalmology Dermatology 2% 2% Stomatology 2% Emergency Medicine Others 3% 4%
Psychiatry 1%
Internal medicine 24%
Surgery 8% Traditional Chinese Medicine (TCM) 14%
Paediatrics 10% Gynaecology and obstetrics 10%
General practice 13%
Fig. 6.5 Percentage of hospital discharge by types of healthcare services in China, 2015. Notes (1) Data Source: Health Statistical Yearbook of China (2016). (2) Services with percentage less than 1% are not annotated in this figure
to speed up the technology diffusion and hence to increase the quality (benefit) of the treatment. We discuss all these policy options in the following subsections.
6.4.1 Out of the Shadow: Information Campaign for the Awareness of Mental Illnesses Given the high prevalence rates of mental disorders in China, it is important to educate the public through information campaigns to increase the awareness of mental illnesses. In addition, an anti-stigma campaign would be beneficial to reduce the nonmonetary cost of seeking mental healthcare. In the Chinese traditional culture, some forms of the stigma associated with mental disorders arise from the names of mental illnesses per se. Thus, an effective approach to mitigate the stigma is to rename the diseases to eliminate the negative bias inherent to the name tags and to give the medical condition a neutral image. This could be done in both psychiatric textbooks and popular culture, and hence change how doctors and the general public think about mental illnesses. Table 6.7 lists the traditional names of mental illnesses in the Chinese language (Mandarin) that contain a strong stigma as well as the suggested new names that may substantially reduce the stigma associated with the medical condition.
6.4 Policy Options for Bridging the Treatment Gap in Mental Healthcare
159
Table 6.7 Name tags of mental illnesses as a source of stigma in Chinese language English name for mental illness
Chinese name with stigmatic bias, followed by literal English translation
Neutral name that avoids stigma, followed by literal English translation
Dementia
痴呆症
失智症
Mentally retarded
Loss of mental capability
精神分裂症
思觉失调症
Mentally split
Early psychosis disorder
躁郁症
双向情感障碍
Choleric and depressed
Bipolar disorder
妄想症
偏执性精神障碍
Schizophrenia Bipolar disorder Paranoid disorder
Hallucination Alzheimer’s disease 老年痴呆症 Old-age mental retard
Paranoid disorder 阿尔茨海默氏症 Alzheimer’s disease
Notes (1) Chinese names with stigmatic bias are the name tags for mental illnesses commonly used in mainland China. (2) Neutral names for Dementia and Schizophrenia are name tags adopted in Taiwan, neutral names for other mental illnesses are the recommended name tags in Chinese
International experiences also suggest that mass media campaigns made by trusted sources (such as professional medical associations) can also contribute to reduce the social stigma and encourage patients with mental diseases to seek proper healthcare. For example, an advertisement campaign in Germany made by Phychenet features a patient suffering from mental illness, which demonstrates and explains the symptoms and prevalence of mental diseases with warm-hearted encouragement for people with such symptoms to seek help. This campaign has successfully raised the public awareness of mental diseases and let patients know that many other people are suffering from the same health conditions, which in turn helped to reduce the self-perceived stigma among these patients. Another example is the Canadian “Bell Let’s Talk” campaign that encourages discussions about mental health and raises funds (https:// letstalk.bell.ca/en/our-initiatives). Similar mass media campaigns have been experimented in various parts of China, with government-financed advertisement displayed on TV, on large advertisement boards in densely populated areas (such as subway stations) and within hospitals.
6.4.2 Increasing the Public Investment in Mental Healthcare Currently, the public funds allocated to the mental health sector only account for less than 1% of total health expenditure in China. Thus, China still has ample room for increasing the public investment in mental health resources, which can be achieved through two main channels: one is to use the general tax revenues to directly subsidize the mental healthcare institutions; the other is through an earmarked tax in the existing
160
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
health insurance programs by specifically enhancing the mental health benefit and financing levels. The advantage for the direct public subsidy to mental health facilities is to mitigate the price distortion and the related profit-seeking behaviors by physicians and hospitals. For example, the essential psychotropic medications for the treatment of mental illnesses are relatively inexpensive in LMICs, because many of them are alreadyoff patent and can thus be produced by local pharmaceutical firms. However, this does not mean that physicians in these countries have incentives to prescribe these cost-effective medicines under a profit-centered health system such as China’s, as the hospitals still rely on the profit of higher-priced prescriptions to resolve their funding gaps. The increase in public investment through direct government subsidy may thus help to reduce such behavioral distortion and hence increase the efficiency of mental healthcare. The advantage of the second financing channel (through an earmarked tax) is to reduce the out-of-pocket payment for the patients with mental illnesses, which in turn provides push incentives for reducing the under-treatment gap in mental healthcare. A recent study suggests that increasing public funds provides a strong return on investment, ranging from 2.3 to 5.7 USD per dollar invested (Chisholm et al. 2016). Although the argument is clear, the government needs to have a strong political willingness to take action. From a theoretical point of view, the need for public investment in the mental health sector arises from the “semi-public good” nature of mental healthcare, which contains an important component of public health with large social benefits (Qian 2012). This change of mentality and alignment of social awareness are necessary, and they can provide a justification for the government to increase the public investment in the mental health sector.
6.4.3 Integrated People-Centered Health System Many studies have shown that the current hospital-centered health system in China is not an efficient approach to bridge the treatment gap in mental healthcare. Rather, an effective intervention and treatment model is to deliver the mental healthcare at the primary and community level (Barber et al. 2014; Liang et al. 2018; Wang et al. 2018). There are at least three arguments to support the urgent need to restructure the current delivery system for mental health services. First, hospital-centered health system is more likely to be constrained by the maldistribution of healthcare resources across regions. By contrast, primary care facilities are relatively easy to access at the community level. As a result, a natural consequence of a shift from the hospital-centered to the primary-care-oriented system is a reduction in the time cost of diagnosis and treatment, which in turn provides strong push incentives to bridge the treatment gap in mental healthcare. Second, a people-centered system, which integrates primary, maternal, and the care for other NCDs together, is in a better position to address the co-morbidities of mental illnesses and the common co-existence of risk factors such as hypertension and obesity. Third, mental health
6.4 Policy Options for Bridging the Treatment Gap in Mental Healthcare
161
is also strongly correlated to economic poverty and poor lifestyle choices (such as malnutrition and physical inactivity), an integrated system is beneficial in the sense that it provides an effective treatment by integrating mental healthcare with antipoverty policies and other disease management programs. In sum, an integrated people-centered delivery model can be a viable choice to break the vicious cycle of economic poverty, under-treatment of mental illnesses, and the co-morbidity with other NCDs. In fact, strengthening primary healthcare has been put on China’s healthcare reform agenda since 2009 (Yip and Hsiao 2014a, b). However, little progress has been made for the transformation of healthcare system from a hospital-centered to a people-centered delivery mode (Du and Zhu 2016; Li et al. 2017; Zhu 2016). According to the recent reviews, there are two main factors that block the progress of this policy. First, China’s healthcare delivery system is still dominated by government-run hospitals (Zhu 2016). Second, due to the existence of multiple-tier medical education system, the quality of physicians is heterogeneous across hospitals. Good doctors with high-quality training are locked in larger urban hospitals, and hence primary care has long been perceived as low-quality and attracts few patients (Hsieh and Tang 2019). As a result, the free mobility of physicians across healthcare institutions, or the “multiple-point practice” model, may be a necessary step in setting up the integrated people-centered healthcare system.
6.4.4 E-health System As mentioned, one major barrier for developing the integrated people-centered primary care system in China is that the primary care is often perceived as low-quality care (Du and Zhu 2016). An effective policy option for breaking this perception is to develop an ICT-based platform, or the e-health system, to inform the public about mental healthcare options and to facilitate the remote and data-based healthcare practices. Properly managed, these ICT-based platforms can potentially lead to quality improvement and cost reduction in mental healthcare, with at least the following identifiable benefits. First, digital healthcare can be an effective approach to reducing the regional inequality in the accessibility of mental healthcare resources, especially in mitigating the quality and technological gaps between the urban and rural areas as well as between large hospitals and primary care institutions. Second, ICT offers alternative models of delivering mental healthcare by eliminating many access barriers in the current system, including the transportation barriers, the perceived stigma associated with visiting mental health clinics, clinician shortages, and the slow diffusion of medical technology from urban to rural areas. Third, ICT has a potential to bridge the treatment gap in mental healthcare by providing remote screening, diagnosis, monitoring, treatment, and even remote training for non-specialist healthcare workers.
162
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
6.5 Conclusions One of the common problems that plague the mental healthcare sectors in many developing countries is the substantial unmet healthcare needs or the large gap between the need for and the provision of mental healthcare treatment. This paper contributes to increasing our understanding on the potential causes of the treatment gap from the perspectives of economics. We hypothesize that mental health services face more access barriers than the general healthcare. Based on the institutional features in China’s health system, we find evidence to support our hypothesis on the four major hurdles in accessing mental healthcare, namely the nonmonetary costs associated with stigma, the monetary costs due to the limited insurance coverage and reimbursement, the time costs that result from the geographic maldistribution of healthcare resources, and the poor healthcare quality due to the slow diffusion of knowledge and technology.4 An important implication of our study is that removing access barriers to mental healthcare is a multi-dimensional task that requires coordination from mental health institutions, the healthcare planning and financing authorities, the patients, and the society in general. Previous policy efforts to remove access barriers have been focused on reducing the monetary cost of mental healthcare through expanding health insurance coverage and on reducing the time costs through a redistribution of healthcare resources. This approach that relies on a single policy tool proves to be insufficient to mitigate the treatment gap in mental healthcare. Our analysis indicates that more policy tools and further actions are needed. Specifically, we propose an information campaign for mental health awareness and we suggest properly renaming the mental health conditions in the Chinese language, both of which aim to reduce the social stigma in public perception and to reduce the nonmonetary costs of seeking mental healthcare. In addition, we also call for more policy efforts to accelerate the process of new drug launch and the adoption of new medical technology in the treatment of mental illnesses, which helps to improve the value and to close the treatment gap of mental healthcare.
References Acemoglu D, Linn J (2004) Market size in innovation: theory and evidence from the pharmaceutical industry. Quart J Econ 119(3):1049–1090 American Psychiatric Association (2010) Practice guideline for the treatment of patients with major depressive disorder. Washington, D.C
4
Admittedly, the evidence provided here is not sufficient for making causal claims on the four major hurdles of mental healthcare, and future studies based on micro-level datasets are invited to empirically assess the relative importance of these factors in causing the treatment gap in mental healthcare.
References
163
Barber SL, Borowitz M, Bekedam H, Ma J (2014) The hospital of the future in China: China’s reform of public hospitals and trends from industrialized countries. Health Policy Plan 29(3):367–378. https://doi.org/10.1093/heapol/czt023 Berndt ER, Cockburn IM (2014) The hidden cost of low prices: limited access to new drugs in India. Health Aff (millwood) 33(9):1567–1575. https://doi.org/10.1377/hlthaff.2013.1307 Bor JS (2015) Among the elderly, many mental illnesses go undiagnosed. Health Aff 34(5):727–731 Burns LR, Liu GG (2017) China’s healthcare system and reform. Cambridge University Press, Cambridge Chatterji P, Alegria M, Takeuchi D (2011) Psychiatric disorders and labor market outcomes: evidence from the national comorbidity survey-replication. J Health Econ 30(5):858–868. https:/ /doi.org/10.1016/j.jhealeco.2011.06.006 Chinese Medical Association (2006) Clinical guidelines for treatment-psychiatry. People’s Medical Publishing House (In Chinese) Chinese Medical Association (2015) Guidelines for the prevention and treatment of depression in China. Chinese Medical Multimedia Press (In Chinese) Corrigan PW, Kerr A, Knudsen L (2005) The stigma of mental illness: explanatory models and methods for change. Appl Prev Psychol 11(3):179–190. https://doi.org/10.1016/j.appsy.2005. 07.001 Du C, Zhu HP (2016) The logic behind the evolvement of healthcare delivery system in Urban China. Soc Sci China 08:45–68 (in Chinese) Eagles JM, Carson DP, Begg A, Naji SA (2003) Suicide prevention: a study of patients’ views. Br J Psychiatry 182:261–265. https://doi.org/10.1192/bjp.182.3.261 Fan P, Pei J, Hou Z et al (2013) The discussion of current mental health status and mental health management strategy in China. J Pract Med Tech 20(8):911–912 (in Chinese) Frank RG, McGuire TG (2000) Economics and mental health. Handb Health Econ 1:893–954 Greenberg PE, Fournier AA, Sisitsky T, Pike CT, Kessler RC (2015a) The economic burden of adults with major depressive disorder in the United States (2005 and 2010). J Clin Psychiatry 76(2):155–162 Greenberg PE, Fournier A-A, Sisitsky T et al (2015b) The economic burden of adults with major depressive disorder in the United States (2005 and 2010). J Clin Psychiatry 76(2):155–162 Grossman M (1972) On the concept of health capital and the demand for health. J Polit Econ 80(2):223–255. https://doi.org/10.1086/259880 Hsieh C, Qin XZ (2018) Depression hurts, depression costs: the medical spending attributable to depression and depressive symptoms in China. Health Econ 27(3):525–544 Hsieh CR, Tang C (2019) The multi-tiered medical education system and its influence on the health care market-China’s Flexner report. Hum Resour Health 17(1):50. https://doi.org/10.1186/s12 960-019-0382-4 Hsieh CR, Liu S, Qin X (2019) The hidden costs of mental depression: implication on social trust and life satisfaction”. Manchester School 87(2):259–296 Knapp M, Funk M, Curran C, Prince M, Grigg M, McDaid D (2006) Economic barriers to better mental health practice and policy. Health Policy Plan 21(3):157–170. https://doi.org/10.1093/ heapol/czl003 Kohn R, Saxena S, Levav I, Saraceno B (2004) The treatment gap in mental health care. Bull World Health Organ 82(11):858–866 Lambregts TR, van Vliet RCJA (2018) The impact of copayments on mental healthcare utilization: a natural experiment. Eur J Health Econ 19(6):775–784. https://doi.org/10.1007/s10198-0170921-7 Li X, Lu J, Hu S et al (2017) The primary health-care system in China. Lancet 390(10112):2584– 2594. https://doi.org/10.1016/S0140-6736(17)33109-4 Liang D, Mays VM, Hwang WC (2018) Integrated mental health services in China: challenges and planning for the future. Health Policy Plan 33(1):107–122. https://doi.org/10.1093/heapol/ czx137
164
6 Understanding and Addressing the Treatment Gap in Mental Healthcare …
Link BG (1987) Understanding labeling effects in the area of mental disorders: an assessment of the effects of expectations of Rejection. Am Sociol Rev 52(1):96. https://doi.org/10.2307/209 5395 Link BG, Phelan JC (2001) Conceptualizing stigma. Annu Rev Sociol 27(1):363–385. https://doi. org/10.1146/annurev.soc.27.1.363 Liu J, Ma H, He YL et al (2011) Mental health system in China: history, recent service reform and future challenges. World Psychiatry 10(3):210–216 McSween JL (2002) The role of group interest, identity, and stigma in determining mental health policy preferences. J Health Polit Policy Law 27(5):773–800. https://doi.org/10.1215/0361687827-5-773 Meng QY, Fang H, Liu X, Yuan B, Xu J (2015) Consolidating the social health insurance schemes in China: towards an equitable and efficient health system. Lancet 386(10002):1484–1492 National Bureau of Statistics of the People’s Republic of China. 2016. Health Statistical Yearbook of China (2011–2016) China, Beijing. Patel V, Xiao S, Chen H et al (2016a) The magnitude of and health system responses to the mental health treatment gap in adults in India and China. Lancet 388(10063):3074–3084. https://doi. org/10.1016/S0140-6736(16)00160-4 Patel V, Chisholm D, Parikh R et al (2016b) Addressing the burden of mental, neurological, and substance use disorders: key messages from Disease Control Priorities, 3rd edn. Lancet 387(10028):1672–1685 Peng L, Meyerhoefer CD, Zuvekas SH (2015) The short-term effect of depressive symptoms on labor market outcomes. Health Econ 25(10):1223–1238. https://doi.org/10.1002/hec.3224 Phillips MR, Yang G, Zhang Y, Wang L, Ji H, Zhou M (2002) Risk factors for suicide in China: a national case-control psychological autopsy study. Lancet 360(9347):1728–1736. https://doi. org/10.1016/S0140-6736(02)11681-3 Phillips MR, Zhang J, Shi Q et al (2009) Prevalence, treatment, and associated disability of mental disorders in four provinces in China during 2001–05: an epidemiological survey. Lancet 373:2041–2053 Qian J (2012) Mental health care in China: providing services for under-treated patients. J Ment Health Policy Econ 15(4):179–186 Qin X, Hsieh CR (2020) Understanding and addressing the treatment gap in mental healthcare: economic perspectives and evidence from China. Inquiry 57:1–15 Qin X, Wang S, Hsieh CR (2018a) The prevalence of depression and depressive symptoms among adults in China: estimation based on a national household survey. China Econ Rev 51:271–282 Sloan F, Hsieh CR (2017) Health economics, 2nd edn. The MIT Press, Cambridge The Institute of Social Science Survey 2012. China Family Panel Study (2012) Bejing: Peking University Thornicroft G (2007) Most people with mental illness are not treated. Lancet 370(9590):807–808. https://doi.org/10.1016/s0140-6736(07)61392-0 Vos T, Flaxman AD, Naghavi M et al (2012) Years lived with disability (YLDs) for 1160 sequelae of 289 diseases and injuries 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 380(9859):2163–2196. https://doi.org/10.1016/S0140-6736(12)61729-2 WHO (2016a) Mental disorders: fact sheet. [Accessed on 2017 January 2017]; Available from: http://www.who.int/mediacentre/factsheets/fs396/en/ World Health Organization (2016b) Out of the shadows: making mental health a global development priority. https://www.who.int/mediacentre/events/2016/mental-health-meeting/en/. Published April 12, 2016. Accessed 19 August 2020 Wang X, Sun X, Birch S et al (2018) People-centred integrated care in urban China. Bull World Health Organ 96(12):843–852. https://doi.org/10.2471/BLT.18.214908 World Health Organization (2014a) Stigma and discrimination. http://www.euro.who.int/en/whatwe-do/health-topics/noncommunicable-diseases/mental-health/activities/stigma-and-discrimin ation. Published October 10 2014a. Accessed 19 Aug 2020
References
165
World Health Organization (2014b) Mental Health Atlas 2014b. https://www.who.int/mental_hea lth/evidence/atlas/mental_health_atlas_2014b/en/. Published January 21, 2019. Accessed 19 Aug 2020 World Health Organization Mental Health Action Plan 2013–2020. https://www.who.int/publicati ons/i/item/9789241506021, published September 17, 2015. Accessed 19 Aug 2020 Yang W (2016) How does the pharmaceutical industry influence prescription? A qualitative study of provider payment incentives and drug remunerations in hospitals in Shanghai. Health Econ Policy Law 11(4):379–395. https://doi.org/10.1017/S1744133116000086 Yip W, Hsiao W (2014a) Harnessing the privatization of China’s fragmented health-care delivery. The Lancet 384:805–818 Yip W, Hsiao W (2014b) Harnessing the privatisation of China’s fragmented health-care delivery. Lancet 384(9945):805–818. https://doi.org/10.1016/S0140-6736(14)61120-X Zhu HP (2016) Leading a reform on medical care system by encouraging innovation on medical service patterns. Financial Minds 01:35–48 (in Chinese)
Correction to: Economic Analysis of Mental Health in China Xuezheng Qin and Chee-Ruey Hsieh
Correction to: X. Qin and C.-R. Hsieh, Economic Analysis of Mental Health in China, Applied Economics and Policy Studies, https://doi.org/10.1007/978-981-99-4209-1 Due to an unfortunate production mistake the correct version of this book was not published initially. Country name has been removed in this updated version. We apologize for the inconvenience caused.
The updated original version for this book can be found at https://doi.org/10.1007/978-981-99-4209-1 © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 X. Qin and C.-R. Hsieh, Economic Analysis of Mental Health in China, Applied Economics and Policy Studies, https://doi.org/10.1007/978-981-99-4209-1_7
C1