Exchange Rate, Credit Constraints and China’s International Trade 9811575215, 9789811575211

Citation preview

Exchange Rate, Credit Constraints and China’s International Trade m i ao j i e y u

Exchange Rate, Credit Constraints and China’s International Trade

Miaojie Yu

Exchange Rate, Credit Constraints and China’s International Trade

Miaojie Yu Peking University Beijing, Beijing, China

ISBN 978-981-15-7521-1    ISBN 978-981-15-7522-8 (eBook) https://doi.org/10.1007/978-981-15-7522-8 Jointly published with Peking University Press The print edition is not for sale in Mainland of China. Customers from Mainland of China please order the print book from: Peking University Press. © Springer Nature Singapore Pte Ltd. 2021 This work is subject to copyright. All rights are reserved by the Publishers, 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 publishers, 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 publishers 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 publishers remain neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Palgrave Macmillan 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

Contents

1 Does Appreciation of the RMB Decrease Imports to the US from China?  1 2 Revaluation of the Chinese Yuan and Triad Trade: A Gravity Assessment 31 3 Exports, Productivity, and Credit Constraints 55 4 Exports and Credit Constraints under Incomplete Information 97 5 Exchange Rate Movements and Exporter Profitability137 6 Export Tightening, Competition, and Firm Innovation163 7 Promotion Effect of CNY Appreciation on Export Quality195 8 Outward Directs Investment, Firm Productivity, and Credit Constraints229

v

vi 

Contents

9 The Effect of RMB Internationalization on Belt and Road Initiative245 10 The Potential Impacts of China–US BIT on China’s Manufacturing Industries261 References287

List of Figures

Fig. 1.1 Fig. 2.1 Fig. 2.2 Fig. 3.1 Fig. 3.2 Fig. 3.3 Fig. 5.1 Fig. 5.2 Fig. 5.3

Fig. 6.1

The American imports from China and the RMB appreciation trajectory (2002–2008). (Source: CEIC Database) 5 Bilateral trade and the exchange rate: China and Japan. (Source: CEIC Database) 33 China’s bilateral exports and exchange rates. (Source: CEIC Database)36 Log exports vs. log inward FDI at Chinese City Level during 2000–2007. (Source: Data are from Chinese City Statistical Yearbook, various years) 58 Log exports vs. log interest expenditure at two-digit industrial level during 2000–2007. (Source: Authors calculation based on the data set) 58 Firm’s productivity, credit constraints, and exports 71 Import tariffs of Chinese manufacturing firms. (Source: International Monetary Fund E-library) 140 Trends of profitability indicators for exporters and non-exporters143 Chinese firm’s revenue and interest payment by two-digit industry. (Notes: The average industrial revenue and interest are calculated over years 2000–2008 by two-digit level Chinese manufacturing sectors. Table 5.2 provides the detailed description for numbers of each sector) 158 Nominal exchange rate of the RMB, 2000–2008. (Note: This figure reports the RMB–dollar exchange rate index and the effective RMB exchange rate. The base period is January 2000, with the exchange rate index set to 100. An increase in the index implies an appreciation of the RMB) 169 vii

viii 

List of Figures

Fig. 6.2 Fig. 6.3 Fig. 7.1 Fig. 7.2 Fig. 8.1 Fig. 10.1 Fig. 10.2

(a) Log R&D expenditure and (b) New product revenue share, by exporters and non-exporters 173 βt, 2002–2007 174 Average export quality and real exchange rate 207 Quality discrete degree of industries with different quality differentiation214 Productivity distributions of ODI firms and non-ODI firms in China231 FDI in manufacturing sector, actually used, China, 2000–2013, $ million. Notes: No manufacturing data in 2004 and 2014. (Source: National Bureau of Statistics of China) 262 Size of production, manufacturing, 2007. (Source: Firm data set by NSB. See Table 10.5 for names of the industries) 271

List of Tables

Table 1.1 Table 1.2 Table 1.3 Table 1.4 Table 1.5 Table 2.1 Table 2.2 Table 2.3 Table 2.4 Table 2.5 Table 2.6 Table 3.1 Table 3.2 Table 3.3 Table 3.4 Table 3.5 Table 3.6 Table 3.7 Table 4.1 Table 4.2

Concordance of industries 12 Effects of RMB revaluation on the imports to the US from China14 More robustness checks for the imports to the US from China (2002–2008) 19 Alternative estimates of the American imports from China and other Asian countries (2002–2008) 24 Main notation for the models 27 Summary statistics 43 Time series properties of the data 44 Estimates of China’s exports to the US (2002–2007) 46 Simultaneous estimates between imports from China to the US and the exchange rate (2002–2007) 48 Estimates of China’s exports to Japan (2002–2006) 49 3SLS simultaneous estimates (2005–2007) 51 Summary statistics (2000–2007) 76 Benchmark estimates, log of exports as the dependent variable 79 IV fixed-effects estimates of interest expenditures on Firms5 exports84 More robustness checks 87 Estimates of ratio specifications 89 Main notation for the models 93 Total factor productivity of Chinese plants 93 Basic statistics for key variables (2000–2008) 119 Benchmark estimates for Chinese and foreign firms, 2000–2008122

ix

x 

List of Tables

Table 4.3

The Heckman two-step estimates of bivariate selection model, 2000–2008 125 Table 4.4 2SLS estimates by sea and non-sea shipments for Chinese firms, 2000–2006 129 Table 4.5 2SLS estimates with measures of sectoral productivity dispersion, 2000–2006 132 Table 5.1 Summary statistics of variables 142 Table 5.2 Effects of RMB appreciation on ROE 144 Table 5.3 Effects of RMB appreciation on Asset Turnover 145 Table 5.4 Effects of RMB appreciation on Profit Margin 145 Table 5.5 Total factor productivity of Chinese plants (2000–2008) 157 Table 6.1 Export tightening 170 Table 6.2 Growth of employment, profit and total sales (%) for exporters and non-exporters 171 Table 6.3 Summary statistics of major variables 176 Table 6.4 Benchmark regression result of Eq. (6.4) 177 Table 6.5 Using one year before and after the shock 179 Table 6.6 Excluding the influence of other confounding policies 180 Table 6.7 Placebo test 181 Table 6.8 Unbalanced sample regressions 182 Table 6.9 Industry heterogeneity 183 Table 6.10 Firm heterogeneity: continuing exporters and export quitters 184 Table 6.11 Processing versus nonprocessing exporters 186 Table 6.12 Growth difference of employment, profit, and total sales (%), by export intensity 189 Table 6.13 Regression results of equation (5), by export intensity 190 Table 6.14 Firm characteristics of continuing exporters and export quitters190 Table 6.15 Export growth and export intensity for continuing exporters in 2005 191 Table 7.1 Overall distribution of China’s export quality from 2000 to 2006206 Table 7.2 Descriptive statistics 210 Table 7.3 Benchmark regression result 211 Table 7.4 Classification of vertically differentiated industries 213 Table 7.5 Subsample analysis of quality differentiation 215 Table 7.6 Subsample results based on Rauch classification 216 Table 7.7 Exchange rate of CNY and the number of export enterprises 217 Table 7.8 Core products and non-core products 218 Table 7.9 Convergence effect of CNY appreciation 220 Table 7.10 Exchange rate system reform and the expired ATC 221

  List of Tables 

xi

Table 7.11 Considering processing trade enterprises 223 Table 7.12 Considering the proportion of import inputs 225 Table 8.1 ODI summary by industry 233 Table 8.2 Summary of statistics 235 Table 8.3 Multinomial logit estimates of basic model 237 Table 8.4 Multinomial logit estimates with interaction effect 238 Table 8.5 Multinomial logit estimates of first-time export/ODI sample 239 Table 8.6 Estimates of export and ODI value 241 Table 9.1 Summary statistics 251 Table 9.2 Estimate of the effects of dummies of whether signed a swap agreement on trade 252 Table 9.3 Estimate of the effect of the scale of swap agreement on trade 254 Table 9.4 Results of IV estimations, IV for whether exists swap agreements256 Table 9.5 Results of IV estimations, IV for relative scale of the swap agreements257 Table 9.6 Score of relationship between China and its strategic partners 258 Table 10.1 US Direct investment in China, actually used, in all sectors, 2005–2013, $ million 263 Table 10.2 Effects of FDI on performance of firms, productivity, China 268 Table 10.3 Estimates of the effect of BIT on firms’ productivity 269 Table 10.4 Effects of FDI on performance of firms, profitability and export propensity, China 270 Table 10.5 shares of foreign capital and restrictions, manufacturing, 2007 272 Table 10.6 Regression results: effects of policy variables on FDI 274 Table 10.7 Estimates of the effects of BIT on FDI under different scenarios274 Table 10.8 Current list of restricted and prohibited industries, manufacturing, 2011 278

CHAPTER 1

Does Appreciation of the RMB Decrease Imports to the US from China?

In 2005, China abated its fixed exchange rate against the US dollar and began to appreciate the Renminbi (RMB). In this chapter, I explore the effect of the appreciation of the RMB on imports to the United States (US) from China by augmenting the gravity model with the exchange rate. Using an industrial panel data set during the period 2002–2008 and controlling for the endogeneity of the bilateral exchange rate, this extensive empirical analysis suggests that the appreciation of the RMB against the US dollar significantly reduced imports to the US from China. This finding is robust to a variety of econometric methods and to coverage in different periods.

1   Introduction Exchange rate movement and its pass-through to changes in domestic prices have been topics of wide concern among economists. However, relatively few studies have empirically investigated the relationship between exchange rate movements and trade flow. This chapter fills this gap by investigating the effect of the appreciation of the Chinese Renminbi (RMB) on imports to the US from China.

This chapter is originally published in Contemporary Economic Policy (2012), 30(4), pp. 533–547 © The Author(s) 2021 M. Yu, Exchange Rate, Credit Constraints and China’s International Trade, https://doi.org/10.1007/978-981-15-7522-8_1

1

2 

M. YU

Today, China has replaced Mexico as the second-largest trading partner with the US. In July 2005, China abated its fixed exchange rate to the US dollar but pegged its currency to a basket of currencies. Since then, the RMB has appreciated by about 20% against the US dollar, from 8.3 to 6.8 RMB per dollar. Simultaneously, China’s bilateral trade surplus from the US decreased from US$232 billion in 2006 to US$114 billion in 2008. This raises the question: has the RMB appreciation decreased the imports to the US from China? The economic intuition behind this question seems straightforward: the appreciation of the RMB resulted in more expensive Chinese exports; consequently, exports diminished while imports increased. However, answering the question is not, by any means, trivial. It is widely recognized that bilateral trade volumes are affected by the trading countries’ GDP, declining trade costs, and trade liberalization (Feenstra 1998). The appreciation of the RMB would have a pass-through effect on American import prices, which in turn would affect the amount of imports to the US from China. By this means, the exchange rate has an effect on the domestic import price similar to that of tariffs, which has been recognized as the symmetry hypothesis between tariffs and the exchange rate (e.g., see Feenstra 1989). Therefore, the effect of exchange rate movements on bilateral trade remains an empirical issue. The gravity model is perhaps the one model that can successfully explain the growing trade volumes. In its simplest version, the gravity model suggests that the bilateral trade volume is directly proportional to the trading countries’ GDP (Tinbergen 1962). I therefore adopt a theoretical gravity model with general equilibrium to access the effect of appreciation of the RMB on Sino-US bilateral trade. The innovation of this chapter is that I explicitly introduce the exchange rate into the theoretical gravity framework; hence, I am able to estimate the effect of the yuan’s revaluation on imports to the US from China.1 Extensive analysis suggests that the revaluation of the Chinese yuan significantly reduced imports to the US from China. Chinese exchange rate movements are helpful in reducing the bilateral Sino-US trade imbalance and accordingly in avoiding a possible trade war between the two countries.

1  In this chapter I do not consider strategic trade policies used by either the home or foreign country to introduce the “terms of trade” changes. The only reason for terms of trade changes is the stylized fact that the US is the largest economy in the world today.

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

3

This chapter joins a growing literature on exchange rates and trade. As introduced by Goldberg and Knetter (1997), there are three related strands in the mainstream literature about exchange rates and goods prices. They cover the pass-through of exchange rates, the law of one price, and pricing-to-market. Feenstra (1989) finds that the symmetry hypothesis between tariffs and exchange rates is easily supported using Japanese and US data. This seminal work also suggests that there is a symmetric response of import prices to changes in import tariffs and bilateral exchange rates. Regarding the previous research on the Sino-US trade and exchange rate, Thorbecke and Zhang (2006) estimate that the Sino-US real exchange rate in the long run is around a unit. By including China’s 33 main trading partners, Thorbecke and Smith (2010) rationalize that the appreciation of the RMB helps to rebalance China’s trade. In particular, a 10% RMB appreciation leads to a decrease of 12% in ordinary exports and 4% in processing exports. The asymmetric effects of RMB appreciation on processing trade and ordinary trade are also explored by Mann and Plueck (2007). Bergin and Feenstra (2008) explore how a change in the share of US imports from a country like China with a fixed exchange rate could affect the pass-through of the exchange rate to import prices in the US. By way of comparison, the main aim of this chapter is to determine how movements of the exchange rate affect imports to the US from China when the terms of trade improvement for importers and the incomplete pass-through of the exchange rate are allowed. Last but not least, Yu (2009) suggests that the RMB appreciation against the dollar significantly reduced China’s exports to the US but had no significant effects on China’s exports to Japan by using three-stage least-square (3SLS) estimations. To explore fully the effect of the RMB exchange rate on imports to the US from China, my estimations are based on a theoretical gravity framework; however, I do not attempt to predict the exchange rate’s influence theoretically, but rather to use a tightly specified theory to inform the empirical analysis. It turns out that the structural parameters based on the theoretical framework help us to understand the impact of the exchange rate on trade. The remainder of this chapter is organized as follows. Section 2 briefly introduces China’s exchange rate reform in the past decade. Section 3 presents a theoretical gravity equation that includes the exchange rate.

4 

M. YU

Section 4 introduces the estimation methodology. Section 5 discusses the estimation results and presents robustness checks. Section 6 concludes the chapter.

2   China’s Exchange Rate Reform China claimed to move toward a market economy in 1992. Shortly afterwards, the exchange rate in China was fixed at the level of 8.3 RMB per dollar in January 1994. During the East Asian Financial Crisis (1997–1998), many countries depreciated their own currencies to mitigate the negative shocks caused by the crisis. For example, the Thai baht was depreciated by around 40%. In contrast, China insisted on maintaining the value of the RMB at the pre-crisis level. However, in July 2005 the RMB against the dollar was revaluated at 2%. In addition, the RMB was no longer solely pegged to the US dollar. The peg was changed to a basket of currencies, including the US dollar and the Japanese yen, among others. Since then, the Chinese currency has appreciated to 6.83 RMB per dollar in December 2008, a 20% revaluation. Why did the Chinese government revalue the RMB in 2005? One important reason was the surging bilateral trade imbalance with the US. From 2002 to 2006, the bilateral Sino-US annual trade growth rate was more than 20%. In 2007, China had already replaced Mexico as America’s second-largest trading partner when the bilateral trade total (including Hong Kong’s re-exports) reached US$318 billion. Simultaneously, China also maintained a huge trade surplus with the US. In 2004, the bilateral trade surplus was US$161 billion. Equally importantly, the Multi-Fiber Agreements, which set an upper bound for textile exports from China to the US, were automatically terminated in January 2005 according to the requirements set by the Agreement on Textiles and Clothing (ATC) in the Uruguay Round of the GATT. As a result, China’s textile exports to the US increased dramatically. In response to demands by special interest groups, such as labor unions, the US Congress threatened to impose trade sanctions on China if it did not “voluntarily” restrain its exports to the US.  In order to avoid a further bilateral trade war, the Chinese government agreed to revaluate its RMB against the dollar by 2% on July 21, 2005. In addition, the exchange rate was allowed to fluctuate within a restricted band.

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

5

Fig. 1.1  The American imports from China and the RMB appreciation trajectory (2002–2008). (Source: CEIC Database)

In this chapter, I focus on how the recent structural change in 2005 has affected the Sino-US bilateral trade. At first glance, as shown in Fig. 1.1, the imports to the US from China kept an increasing trend over the years 2002–2008. Simultaneously, the Sino-US exchange rate, measured as RMB per dollar, has kept declining since July 2005. Motivated by these observations, in Section 3, I develop a theoretical framework aimed at exploring the relationship between exchange rate movements and bilateral trade.

6 

M. YU

3   Theoretical Gravity Framework Following Yu (2010), suppose that each country produces unique product varieties. Let h represent the good, k represent the industry, and i represent the importer. The export of good h in industry k from country i to the importer (i.e., the US) is identical to the consumption of good h in industry k in the US. Exporter i = 1, …, I has K industries. Industry k ∈ K produces Nik commodities. The US faces an aggregate CES utility function: I

U

K Nik

   C

h i ,us , k





dhdkdi,    0  ,

(1.1)

i 1k 1h 1

where Cih,us , k is American consumption of good h in industry k produced by country i. The elasticity of substitution σ is denoted as σ = 1/(1 − ρ). I follow Anderson and van Wincoop (2003) and assume that, given each exporter i, pih,us , k  pih,us , k for all h and h′ in {1, …, Nik}, that is, all the goods in industry k imported by the US from country i have the same price pi, us, k.2 In addition, American consumption is identical over the entire line of products within industry k sold by country i, that is, Cih,us , k  Cih,us , k  Ci ,us , k , ∀h ∈ {1, …Nik}. Utility function (1.1) can then be expressed as I

U

K

  N C ik

i 1k 1

i ,us , k





dkdi.

(1.2)

The representative consumer in the US maximizes her utility (1.2) subject to the budget constraint: I

Y us 

K

 N

i 1k 1

ik

pi ,us , k Ci ,us , k dkdi,

(1.3)

where Yus is the US GDP. By solving this maximization problem, I obtain the demand function for each product: 2  Note that prices of varieties are allowed to differ across industries. This assumption is roughly consistent with the reality: the price of a Chrysler-type automobile is close to that of a Ford, but it is very different from the price of a pencil.

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

1

Ci ,us , k

p   1  Y us   i ,us , k    Pk   Pk

 , 

7

(1.4)

where the aggregate American price index, Pk, is defined as  I K  Pk     N ik  pi ,us , k   1 dkdi   i 1k 1 



 1 

.

(1.5)

Hence, the total value of American imports from China (i = ch) is Xusch, k 

N kch

p

h ch ,us , k

Cchh ,us , k dh  N kch pch,us , k Cch,us , k ,

(1.6)

h 1

where the first equality follows the definition of export value, and the second one is due to the equal price assumption across varieties of goods. Combining Eqs. (1.4), (1.5), and (1.6), I obtain the export value of industry k from China to the US: 

X

ch us , k

p  1   N Y  ch,us , k  .  Pk  ch us k k

(1.7)

However, bilateral trade is also affected by the number of varieties in the exporting country, N kch , which is unfortunately unobservable. For estimation, I consider the monopolistic competition model presented originally by Krugman (1979), which helps us to eliminate the number of exporting varieties in my gravity Eq. (1.7). Turning to the supply side, the representative firm in a country maximizes profits. Specifically, as in Krugman (1979), Baier and Bergstrand (2001), and Feenstra (2002), the production of goods ykch incurs a fixed cost  kch and a constant marginal cost  kch , given that labor lkch is the representative firm’s unique input in industry k:

 



 

lkch   kch   kch ykch .

 

 

(1.8)

8 

M. YU

The monopolistically competitive equilibrium implies two conditions for the representative firm. First, the marginal revenue should equal marginal cost for the representative firm. Since the elasticity of demand equals the elasticity of substitution, σ, when China’s number of goods N kch is large, I obtain the first equilibrium condition:

 pkch   kch w ch ,



(1.9)

where the wage in China is denoted as wch. Second, the representative firm obtains zero profits due to free entry. Given that the firm’s profit function in China is  kch  pkch ykch  w ch  kch   kch ykch , I obtain the equilibrium production level, ykch , for such a representative firm in industry k in China:





ykch 

 kch , 1    kch



where ykch is a constant number given that ρ, κ kch , and ϕ kch are all constant parameters. By denoting the bilateral exchange rate ($/RMB) as e, the 1 GDP in China measured in dollars is Y ch = ch eN kch pkch ykch , where skch is the sk output share of industry k in China. Substituting this into Eq. (1.7), I have: 

Xusch, k

  1 s chY chY us  p  k ch chk  ch,us , k  . epk yk  Pk 

(1.10)

Therefore, bilateral trade depends on the bilateral exchange rate as well as the trading countries’ GDP, China’s industrial output share, the fixed production of China’s representative firm, and various price indices. Note that in Eq. (1.10), I use disaggregated industrial output to measure American income but GDP to measure Chinese income. The reason is that I do not have data on disaggregated Chinese industrial data. For convenience, I include the main notation of the model in Appendix Table 1.5

4   Empirical Methodology To estimate the gravity Eq. (1.10), I specify the estimating equation by taking logs on both sides:

9

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 





ln Xusch, k  ln Y chYkus  ln e  ln pkch  ln skch  1    ln pch,us , k  1    ln Pk  ln ykch .





(1.11)

Like tariffs, the bilateral exchange rate serves as a kind of “iceberg” trade cost across borders (Samuelson 1952). The RMB appreciation would have a partial pass-through effect on the domestic import prices in the US. Put another way, like imposing a tariff on the imports of a large country, the movement of the exchange rate lowers the exporter’s (China)  prices. We shall consider pch,us , k  e pkch where δ   F or Prob. > χ2 Number of observations

i Regressand : ln XUS / YiYUS

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

15

16 

M. YU

5.3  Endogeneity Issues The bilateral exchange rate is not exogenously given but is indeed affected by the volume of imports to the US from China. In reality there may be a variety of channels through which bilateral trade would reversely affect the bilateral exchange rate. One possible channel is that China’s higher trade surplus from the US could increase the US political pressure on China to appreciate the RMB.  In early 2005, the termination of the Multi-Fiber Agreement led to a surge in textile exports from China into the US. As a result, the Sino-US trade imbalance increased dramatically, which in turn caused special interests groups in the US to demand that the domestic textile producers be protected. To avoid possible trade sanctions from the US, the Chinese government agreed to appreciate the RMB against the dollar by 2% in July 2005.13 Moreover, the RMB was no longer pegged to the US dollar alone but to a basket of currencies. Therefore, the volume of imports to the US from China reversely affected the bilateral exchange rate. To control for the endogeneity of the bilateral exchange rate, IV estimation is a powerful econometric method.14 To obtain accurate estimates, I chose China’s monetary stock (M1) as the instrument variable to perform the two-step general method of moments (GMM) estimation. The main reason for adopting the GMM was that it requires fewer assumptions about the error terms and has the ability to generate heteroskedasticity-­ robust standard errors as compared with the general least-squares method (Hall 2004). I report the estimation results of the second-stage GMM in columns (3) and (4) of Table 1.2. The economic rationale for choosing M1 as an instrument for the exchange rate follows that of Bergin and Feenstra (2008): with a tight monetary policy caused by a decreasing money supply, Chinese interest rates increase. As a result, the surging demand for the RMB pushes its

13  Though the Chinese officials would be reluctant to admit that the US diplomacy has a key role to play in the development of the RMB, I thank a referee for correctly pointing this out. 14  The IV approach is a good way to control the endogeneity issues raised by various possible sources: reverse causality (i.e., simultaneity), omitted variables, and measurement errors. Wooldridge (2002, chapter 5) carefully scrutinizes this topic. Therefore, the IV estimates here control for the endogeneity caused by the reverse causality of the bilateral exchange rate as well as the one caused by the omitted variables in (1.14).

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

17

exchange rate up.15 With a stronger RMB, the Chinese exports to the US are expected to decrease. To validate this instrument variable, I performed several statistical tests. First, the F-statistic test in the first stage shows that the instrument is highly statistically significant. The F-statistics are also definitely high enough to pass the F-test. Second, in columns (3) and (4) of Table 1.2, I go further to check whether or not such an exclusive instrument was “relevant,” that is, whether it is correlated with the endogenous regressor (i.e., the exchange rate). In my econometric model, the error term is assumed to be heteroskedastic:  ijt ~ N 0, ij2 . Therefore, the usual Anderson (1984) canonical correlation likelihood ratio test is invalid since it only works under the assumption. Instead, I use the Kleibergen and Paap (2006) Wald statistic to check whether the excluded instrument correlates with the endogenous regressors. The null hypothesis that the model is under-identified is rejected at the 1% significance level. Third, I test whether or not the instrument (i.e., Chinese M1) is weakly correlated with the exchange rate. If so, then the estimates will perform poorly in the IV estimate. The Kleibergen and Paap (2006) F-statistics provide strong evidence to reject the null hypothesis that the first stage is weakly identified at a highly significant level.16 Finally, both the Anderson and Rubin (1949) statistic (which is an LM test) and the Stock and Wright S statistic (which is a GMM distance test) reject the null hypothesis that the coefficient of the endogenous regressor is equal to zero. In short, these statistical tests provide sufficient evidence that the instrument performs well and therefore the specification is well justified. Column (3) of Table 1.2 reports the two-way fixed-effects estimation results using the Chinese M1 as an instrument. After controlling for the two-way fixed effects, the estimated magnitude of the log of the exchange rate was reduced to 1.53, which is also identical to its counterpart in column (1) without controlling for the endogeneity. In column (4), by including lags of exchange rate as additional regressions, the coefficient of





15  One caveat here is that China currently still, to some extent, has capital control. A possible related concern is that the historical link between the money supply and the exchange rate may be weak. However, the simple correlation between the two variables in my data set is quite sizable (corr. = 0.47), hence the concern mentioned above should not be so severe. I thank a referee for suggesting this check. 16  Note that the Cragg and Donald (1993) F-statistic is no longer valid since it only works under the i.i.d. assumption.

18 

M. YU

current exchange rate keeps stable as in previous estimations. In addition, the coefficients of the lags of exchange rate are, again, insignificant at the conventional statistical level. 5.4  Additional Robustness Checks To repeat, China’s exchange rate against the US dollar changed after July 2005. Therefore, it is reasonable to suspect that the pass-through of the exchange rate and accordingly its impact on the bilateral trade volume are underestimated when data from before the structural change are included in the model. I therefore re-estimate the effects by including only the samples after the 2005 change. Columns (5)–(8) of Table 1.2 report the Sino-U.S. estimations for the samples during 2005–2008. Briefly, the point elasticity of bilateral trade with respect to the exchange rate in all the specifications has the same statistically significant signs and close magnitudes to their counterparts shown in columns (1)–(4) of Table 1.2. In particular, in column (5), after controlling for the two-way fixed effects, the appreciation of the RMB was found to have a similar magnitude as its counterpart in column (1) once again. Similarly, the estimate in column (6) with lags of exchange rate ascertains that lags of exchange rate have no significant effects on bilateral trade. After controlling for the endogeneity and the two-way fixed-effects, in column (8) I find that the effect of RMB appreciation on the American import from China is slighter larger than its counterpart in column (4) by using the whole sample during 2002–2008. Moreover, columns (1) and (2) of Table  1.3 include both countries’ GDP per capita in the estimations to check if they have significant effects on bilateral trade as these variables are standard in recent gravity models (e.g., see Rose 2004; Subramanian and Wei 2007). In column (1), China’s GDP per capita has a significant and positive sign, whereas the US counterparts are insignificant at the conventional statistical level. Nevertheless, the appreciation of the RMB still has a significantly negative effect on the imports to the US from China. After controlling for the endogeneity issue in column (2), the coefficients of nominal exchange rate, as well as GDP per capita of both China and the US still have anticipated signs, though statistically insignificant.

 −1.81 (−1.10) –

−1.69** (−10.50) – – – – −0.56 (−0.41) 0.86* (1.76) 1.00 (0.67) 1.36 (0.84) −0.10** (−4.05) Yes Yes Yes 15.74a 15.72a 15.87a 1.14

(2)

OLS

(1)

OLS

– – – −0.45** (−3.10 Log US price index 0.83** (5.03) Log GDP per capita of US 1.06 (0.82) Log GDP per capita of China 1.24** (2.83) Time trend −0.03** (−2.88) Year-specific fixed effects Yes Quarter-specific fixed effects Yes Industry-specific fixed effects Yes First-stage F-statistics – Kleibergen-Paap rk LM statistic – Kleibergen-Paap rk Wald statistic – Anderson-Rubin χ2 statistic –

Log real exchange rate (1-Lag) Log real exchange rate (2-Lag) Log real exchange rate (3-Lag) Log China’s price index

Log real exchange rate

Log exchange rate ($/RMB)

i Regressand : ln XUS / YiYUS

 – −0.52** (−2.61) −0.31 (−1.21) 0.35 (1.36) −0.15 (−0.60) – – – – −0.006** (−4.58) Yes Yes Yes – – – –

−0.56** (−4.23) – – – – – – – −0.003** (−3.14) Yes Yes Yes – – – –

(4)

IV

–

(3)

IV

−0.261 (−0.13) Yes Yes Yes 0.00 0.01 0.01 29.44

–

–

–

524.9 (0.06) – – – –

–

(5)

OLS

−0.066 (−0.29) Yes Yes Yes 1.97 0.85 0.85 106.2a

–

–

–

34.7 (1.21) −18.85 (−0.51) 35.11 (1.00) −15.64 (−0.73) –

–

(6)

OLS

Table 1.3  More robustness checks for the imports to the US from China (2002–2008)

1.31 (0.91) 0.27 (0.57) −0.015 (−1.28) Yes Yes Yes – – – –

–

−0.57** (−4.16) – – – –

–

(7)

IV

(continued)

1.31 (0.91) 0.27 (0.57) −0.19** (−12.30) Yes Yes Yes 172.2a 157.1a 173.4a 29.64a

−0.57** (−4.18) – – – –

–

(8)

IV

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

19

 – 0.00 1730

(1)

OLS

1.14 0.00 1730

(2)

OLS

– 0.00 1730

(3)

IV

– 0.00 1544

(4)

IV

28.93 0.00 1544

(5)

OLS

98.94 0.00 1544

(6)

OLS

a

(8) 29.12a 0.00 1730

– 0.00 1730

IV

(7)

IV

a

The p-value of the statistic is less than 0.01

Notes: Numbers in parenthesis are t-values. *Significant at 1%; **significant at 5%. The first-stage F-statistic value reports the F-statistic result in the first-stage regression with log of current exchange rate as the regressand

Stock-Wright LM S statistic Prob. > F or Prob. > χ2 Number of observations

i Regressand : ln XUS / YiYUS



Table 1.3  (continued)

20  M. YU

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

21

5.5  Alternative Measures on Exchange Rate Since the inflation rate in China and the US certainly did not track exactly over 2002–2008, it is worthwhile exploring how the real Sino-US exchange rate affects the American imports from China. Following previous works such as Zhang (2001), I proxy the real exchange rate as the product of the nominal exchange rate (e) and a fraction of the American producer price index (PPIUS) in the denominator and China’s producer price index (PPICH) in the numerator: e × (PPICH/PPIUS). The fixed-­effects estimate in column (3) of Table 1.3 suggests that real exchange rate appreciation leads to low American imports from China. These results are still robust even when the three-quarter lags of real exchange rate realizations are included, as shown in column (4). After controlling for the endogeneity of real exchange rate, the fixed-­ effects estimate in column (5) shows that the effect of real exchange rate on the American import from China is no longer significant. Adding the lag variables of real exchange rate in column (6) does not change the results substantially. This is possibly due to the lack of consideration of GDP per capita. Therefore, in columns (7) and (8), by including trading countries’ GDP per capita, I find that the coefficients of real exchange rate turn to be significant at the conventional statistical level. 5.6  Further Estimates on Sectoral Heterogeneity17 In all the estimations above, the exchange rate variable varies over years but does not change across industries. The homogeneity assumption on the exchange rate coefficient may be acceptable if the aggregate trade flow is of interest. However, the exchange rate pass-through, as a function of market (pricing) power, would vary considerably across industries. Hence, it is important for us to study the heterogenous effect of the exchange rate on the industry-level bilateral trade. The common correlated effects (CCE) approach is a good way to identify such heterogenous effects of the exchange rate across industries. As introduced by Pesaran (2006), the basic idea is to filter the industry-­ specific regressors by means of cross-section averages. In this way, as the number of industries becomes increasingly larger, the differential effects of 17  I am most grateful to two anonymous referees for their insightful suggestions on Sects. 5.6 and 5.7.

22 

M. YU

unobserved common factors converge to zero asymptotically. In particular, the CCE estimator is obtained by two steps following Eberhardt and Teal (2009). First, I perform 62 OLS estimations by each industry i and obtain its coefficients bˆ i . Second, the CCE estimators are those averaged across sectors: bˆ CCE  i bˆ i / 62 . Columns (1)–(4) of Table  1.4 report the estimation results by using this common factor approach to spatial heterogeneity, in which I adopt the import ratio divided by the product of trading countries’ GDPs as the regressand. The point elasticity of bilateral imports with respect to the exchange rate is −0.96 in column (1), which is smaller than its counterpart obtained by OLS in column (1) of Table 1.2. However, once the trading partners’ per capita GDPs are considered in column (2), the coefficient of the exchange rate, −3.25, turns to be much larger than its counterpart in column (1) of Table 1.3: −1.69. Column (4) replaces nominal exchange rate with real exchange rate and still obtains an exact identical magnitude of the CCE estimator of the exchange rate as in column (2). Nevertheless, in any case, all the CCE estimates suggest that the appreciation of the RMB against the dollar significantly reduced the imports to the US from China. 5.7  Additional Estimates with Other Competing Trading Partners As highlighted by Anderson and van Wincoop (2003), to estimate the gravity model precisely, it is essential for researchers to control for the “multilateral resistance.” The basic idea is that the bilateral trade flow is not simply affected by the two trading countries’ economic factors but is also affected by factors from all other trading countries. That is, trade volumes are determined by relative export barriers but not by absolute trade barriers. Although the theoretical model above suggests that the American imports from China explicitly depend on the US and Chinese incomes, the Sino-US exchange rate, and the prices of traded goods in China and the US, it also implies that the American imports from China are also affected by imports from other countries.18 In fact, it is possible that the American imports from China are affected by its imports from some Asian 18  To see this point, note that the American aggregate industrial price index in the derived gravity equation (Logarithm Gravity) depends on many exporters’ numbers of varieties, as shown in (Aggregate Price Index).

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

23

countries that have patterns of exports similar to China.19 Indeed, the exchange rates in such countries also adjust after the dollar depreciation against the RMB. Therefore, it is worthwhile seeing how the variations of such an American exporters’ exchange rate as well as that of the RMB vis-­ à-­vis the US dollar affect the US imports. To address this concern, I include data of Indonesia, Japan, Korea, Thailand, and Vietnam as well as China in the sample.20 Columns (5)–(8) of Table 1.4 report the estimation results using such Asian sample in which number of observations increases to 6305. Note that here the regressands, once again, are the ratio of American import from its Asian trading partners over the product of GDPs. Column (5) reports the estimation results by including country-specific, industry-specific, and time-specific fixed effects. It turns out that the coefficient of the exchange rate has an anticipated negative sign but statistically insignificant. I suspect that this is because of the lack of the complete control for the “multilateral resistance” effect in my gravity model. I therefore follow Baldwin and Taglioni (2006) to perform the estimates with the time-varying country-specific fixed effects as well as the regular industry-specific fixed effects in column (6). Since the panel in my sample includes 6 American trading countries with 28 time spans, I generate 168 (i.e., 6 × 28) dummies for unidirectional trade data (e.g., exports from China to the US) in addition to the regular industry-specific fixed effects. The estimation results there clearly suggest that the appreciation of the exporter’s exchange rate against the US dollar decreases the ratio of American imports from such Asian trading partners (i.e., the US imports over the product of the two trading countries’ GDPs). In particular, a 10% appreciation of the exporter’s exchange rate is associated with an 11% decrease in the ratio of US imports from such countries. Finally, it is also worthwhile to check the effect of real exchange rate movement on the import ratio of the US from such trading countries. By including industry-­ specific and time-specific fixed effects, the estimate in column (7) clearly suggests that real exchange appreciation significantly reduces the US imports from such Asian trading partners. Finally, the last column of Table  1.4 reports the estimate with time-varying country fixed-effects.  I thank a referee for insightfully suggesting this point.  Here data of Hong Kong are not included since Hong Kong kept a fixed exchange rate against the US dollar over time and hence it is impossible to explore the effects of the movement of the exchange rate on bilateral trade. 19 20

Year-specific fixed effects

Time trend

−0.01** (−2.48) Yes

–

Log US price index

Log GDP per capita of China

−3.25** (−5.60) –

−0.96** (−3.08) –

Log GDP per capita of US

(2)

(1)

−1.94* (−3.56) 2.53** (3.41) −0.37 (−0.34) 2.05** (4.56) −0.03** (−4.36) Yes

CCE

CCE

US-China sample

0.41 −1.22 1.23* (1.89) –



Log exporter’s price index

Log real exchange rate

Log exchange rate

i Regressand : ln XUS / YiYUS



−0.01** (−3.36) Yes

–

–

2.72** (6.84) 2.73** (2.57) −0.06** (−8.12) Yes

–

−3.25** (−6.13) –

−0.67** (−2.41) – –

–

(4)

CCE

–

(3)

CCE

−0.005 (−0.66) Yes

–

0.28 (0.27) 0.67 (0.53) –

−1.05 (−1.26) –

(5)

OLS

Yes

–

0.75 (0.39) 0.61 (0.43) –

−1.10** (−2.86) –

(6)

OLS

Asian sample

−0.000 (−0.16) Yes

–

–

–

−2.15** (−21.56) –

–

(7)

OLS

Table 1.4  Alternative estimates of the American imports from China and other Asian countries (2002–2008)

No

–

–

–

–

0.16 (0.13) –

–

(8)

OLS

24  M. YU



 Yes No No No 0.00 1736

(2)

(1) Yes No No No 0.00 1736

CCE

CCE

US-China sample

Yes No No No 0.00 1736

(3)

CCE

Yes No No No 0.00 1736

(4)

CCE

Yes No No No 0.00 6305

(5)

OLS

i

Yes No No No 0.00 6305

(6)

OLS

Asian sample

Yes No No No 0.00 6305

(7)

OLS

No No No No 0.00 6305

(8)

OLS

Notes: Columns (1)–(4) adopt the common correlated effects (CCE) approach to perform the estimations, in which XUS denotes imports to the US from i China. In columns (5)–(8), XUS denotes imports to the US from exporter i and the exporters include China, Indonesia, Japan, Korea, Thailand, and Vietnam. The exchange rates (ei) in columns (1)–(4) are defined as dollar per RMB whereas those in columns (5)–(8) are defined as per exporter i’s currency. There are 168 (i.e., 6 × 28) time-varying country dummies and 68 industrial dummies in the FE estimations. Numbers in parenthesis are t-values. *Significant at 1%; `significant at 5%

Quarter-specific fixed effects Industry-specific fixed effects Country-specific fixed effects Time-varying country fixed effects Prob. > F Number of observations

i Regressand : ln XUS / YiYUS

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

25

26 

M. YU

The coefficient of log real exchange rate seems to have an unanticipated positive sign. However, one does not need to worry much on that, given that it is statistically insignificant.

6   Concluding Remarks In this chapter, I investigate the effect of the RMB appreciation on imports to the US from China using industrial panel data from 2002 to 2008. Differently from other pure reduced-form estimations, my estimations are guided by an augmented theoretical gravity model. Structural parameters based on a theoretical framework will help us to understand the magnitude of RMB revaluation on Sino-US bilateral trade. The estimation results clearly suggest that the RMB appreciation against the dollar significantly reduced imports to the US from China. These findings are robust to different econometric methods and different data periods. This finding has policy implications. Firstly, if appreciation of the RMB does significantly reduce the Sino-US bilateral trade imbalance, then it would have the beneficial effect of relieving the trade tensions between the two giants. Secondly, RMB appreciation would make it more difficult for Chinese exporters to export to the US ceteris paribus, which in turn would require Chinese exporting firms to make every effort to boost their productivity to survive in the global competition. Several extensions and possible generalizations merit special consideration. One of them is to replace the industrial price index with actual unit-­value f.o.b. prices, if the data are available. In this manner, the exchange rate pass-through can be more precisely identified. Another possible extension is to include import tariffs in the model and to examine the symmetry hypothesis between the exchange rate and the tariffs. Due to the data constraint, I am not able to explore these issues here. However, these are some possible research topics to pursue in the future.

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

27

Appendix Table 1.5   Main notation for the models Symbol

Definition

Panel A: Theoretical framework Cih,us ,k Amount of goods h of industry k produced in country i and consumed in the US Nik Number of goods of industry k produced in country i σ Elasticity of substitution, σ > 1 e Sino-US bilateral exchange rate ($/RMB) Ych Level of GDP in China Ykus Output level of industry k in the US pch,us,k Price of industry k on an American c.i.f. basis pch,k Price of industry k on a f.o.b. basis Yusch, k Value of exports of industry k from China to the US Pk American aggregate price index of industry k wch Wages in China lkch Labor input for the representative firm of industry k in China ykch Output of China’s representative firm of industry k, which is a fixed number in equilibrium: ykch = ykch ch κk Fixed cost for the representative firm of industry k in China skch Industry k’s output share in China φkch Constant marginal cost for the representative firm of industry k in China Panel B: Empirical specification αk Unspecified industrial bilateral border effect εkt Error term in specification (estimate) ηk Industry-specific random variable φyt Year-specific random variable φqt Quarter-specific random variable νkt Industrial idiosyncratic random variable

References Anderson, James and Eric van Wincoop (2003), Gravity with Gravitas: A Solution to the Border Puzzle, American Economic Review 93(1), pp. 170–192. Anderson, T. W., and H. Rubin (1949), “Estimation of the Parameters of a Single Equation in a Complete System of Stochastic Equations,” Annals of Mathematical Statistics 20, pp. 46–63. Anderson, T.W. (1984), Introduction to Multivariate Statistical Analysis, 2nd ed. New York: John Wiley & Sons.

28 

M. YU

Baier, Scott L. and Jeffery H. Bergstrand (2001), The Growth of World Trade: Tariffs, Transport Costs, and Income Similarity, Journal of International Economics 53, pp. 1–27. Baldwin, Richard and Daria Taglioni (2006), “Gravity for Dummies and Dummies for Gravity Equations,” NBER Working Papers, No. 12516. Bergin Paul R. and Robert C. Feenstra (2008), “Pass-through of Exchange Rates and Competition between Floaters and Fixers,” Journal of Money, Credit and Banking, 41(s1), pp. 35–70. Cragg, J.G. and S.G. Donald (1993), “Testing Identfiability and Specification in Instrumental Variables Models,” Econometric Theory 9, pp. 222–240. Eberhardt, Markus and Francis Teal (2009), “A Common Factor Approach to Spatial Heterogeneity in Agricultural Productivity Analysis,” University of Oxford, CSAE WPS/2009-05. Feenstra, Robert and Gordon Hanson (2004), Intermediaries in Entrepôt Trade: Hong Kong Re-Exports of Chinese Goods, Journal of Economics & Management Strategy, 13(1), pp. 3–35. Feenstra, Robert C. (1989), Symmetric Pass-through of Tariffs and exchange Rates under Imperfect Competition: An Empirical Test, Journal of International Economics 27, pp. 25–45. Feenstra, Robert C. (1998), “Integration and Disintegration in the Global Economy,” Journal of Economic Perspectives 12, pp. 31–50. Feenstra, Robert C. (2002), Border Effects and the Gravity Equation: Consistent Methods for Estimation, Scottish Journal of Political Economics 49, pp. 491–506. Goldberg, P.  K. & Knetter, M.  M., 1997, “Goods Prices and Exchange Rates: What Have We Learned?”, Journal of Economic Literature, Vol. 35(3), pp. 1243–1272. Hall, R. Alastair (2004), Generalized Method of Moments, Oxford University Press. Kleibergen, Frank and Richard Paap (2006), “Generalized Reduced Rank Tests Using the Singular Value Decomposition,” Journal of Econometrics, 133(1), pp. 97–126. Krugman, P. (1979), Increasing Returns, Monopolistic Competition, and International Trade, Journal of International Economics 9, pp. 469–479. Mann, C. and Plueck, K. (2007), “The U.S.  Trade Deficit: A Disaggregated Perspective,” in R. Clarida (eds.), G7 Current Account Imbalances: Sustainability and Adjustment. University of Chicago Press. Pesaran, M. Hashem (2006), “Estimation and Inference in Large Heterogeneous Panels with a Multifactor Error Structure,” Econometrica 74(4), pp.967–1012. Rose, Andrew K.(2004), Do We Really Know That the WTO Increases Trade? American Economic Review 94(1), pp. 98–114. Samuelson, Paul (1952), The Transfer Problem and Transport Costs: The Terms of Trade When Impediments are Absent, Economic Journal 62, pp. 278–304.

1  DOES APPRECIATION OF THE RMB DECREASE IMPORTS… 

29

Subramanian, Arvind and Wei, Shang-Jin (2007), The WTO Promotes Trade, Strongly but Unevenly, Journal of International Economics 72(1), pp.151–175. Thorbecke, W. and Smith G. (2010), “How would an Appreciation of the Renminbi and other East Asian Currencies Affect China’s Exports?” Review of International Economics, 18(1), pp.95–108 Thorbecke, W. and Zhang H. (2006), “How Would an Appreciation of the Renminbi Affect the U.S.  Trade Deficit with China?” BE Press Topics in Macroeconomics 6(3), pp. 1–15. Tinbergen, Jan (1962), Shaping the World Economy, New  York: Twentieth Century Fund. Wooldridge, Jeffery M. (2002). Econometric analysis of cross section and panel data. Cambridge, Massachusetts: MIT Press. Yu, M. 2010. Trade, democracy, and the gravity equation. Journal of Development Economics, 91(2), 289–300. Yu, Miaojie (2009), “Revaluation of the Chinese Yuan and Triad Trade: A Gravity Assessment,” Journal of Asian Economics 20, pp. 655–668. Zhang, Zhichao (2001), “Real Exchange Rate Misalignment in China: An Empirical Investigation,” Journal of Comparative Economics, 29, pp. 80–94.

CHAPTER 2

Revaluation of the Chinese Yuan and Triad Trade: A Gravity Assessment

The literature has paid little attention to the endogenous nexus between exchange rates and bilateral trade. In this chapter, I use a gravity model to investigate the two-way causality between exchange rates and bilateral trade with data from China, Japan, and the US during the 2002–2007 period. After controlling for the simultaneous bias between exchange rates and bilateral trade, the extensive empirical evidence shows that the revaluation of the Chinese yuan against the dollar significantly reduced China’s exports to the US but had no significant effect on China’s exports to Japan. These findings are robust to different measures, econometric methods, and period coverage.

1   Introduction After China acceded to the World Trade Organization (WTO) in 2001, China’s exports increased dramatically. The annual export growth rate was 274% during 2002–2007. China’s exports to Japan and the US, the two largest trading economies in the world, also grew very quickly. Specifically, China’s exports to the US increased from US$53.2 billion to $232.7 billion, a 35.9% annual growth rate during this period. By way of

This chapter was originally published in Journal of Asian Economics, 2009, 20, pp. 655–668. © The Author(s) 2021 M. Yu, Exchange Rate, Credit Constraints and China’s International Trade, https://doi.org/10.1007/978-981-15-7522-8_2

31

32 

M. YU

comparison, China’s exports to Japan increased from US$48.4 billion to $102.0  billion, a 10.7% annual growth rate during this period. Simultaneously, the exchange rate of the Chinese yuan (RMB) against the US dollar changed by around 20% during this period due to revaluation. As shown in Fig. 2.1, after the RMB’s revaluation against the US dollar in 2005, the proportion of China’s exports to the US compared with China’s overall export volume followed a downward trend whereas that of China’s exports to Japan continued to decrease. It is therefore interesting to ask whether the revaluation of the RMB reduced bilateral trade among China, Japan, and the US. This chapter seeks to understand the endogenous nexus between the movements of the bilateral exchange rates and bilateral trade among the triad: China, Japan, and the US. The intuition seems straightforward: the increase in RMB valuation against the US dollar made Chinese exports to the US more expensive, which in turn decreased China’s exports to the US.  However, there is a more fundamental mechanism underlying this conventional wisdom: the bilateral exchange rate is not exogenous itself. Surging Chinese exports could result in strong pressure to protect markets raised by the import-competing special interest groups in the importing country. Accordingly, the government in the importing country would push the exporting country to revalue its exchange rate. Put another way, exports have a reverse causality on bilateral exchange rates. Ignoring this fact may make estimation results imprecise. Previous studies have paid little attention to this two-way causality. Most only mention one of the two causal connections. Some works focus on the impact of the bilateral exchange rate on bilateral trade, especially through the pass-through effect of the exchange rate (Goldberg and Knetter 1997). When the nominal bilateral exchange rate is changed, it has a pass-through effect on the price of the imports, which in turn would affect bilateral trade. Previous studies like Feenstra (1989) find empirical evidence that the effect of the bilateral exchange rate on bilateral trade is like that of a tariff. Bergin and Feenstra (2008) also explored how a change in the share of US imports from a country with a fixed exchange rate like China could affect the pass-through of the exchange rate to import prices in the US. On the other hand, a variety of papers consider the determinants of the bilateral exchange rate (Meese and Rose 1991). Bilateral trade, among others, is one of the most important determinants. As mentioned above, in considering the endogenous nexus between these two variables, studies on a one-way causality would lead to simultaneous bias.

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

33

Fig. 2.1  Bilateral trade and the exchange rate: China and Japan. (Source: CEIC Database)

34 

M. YU

I therefore estimate the bidirectional causality between the bilateral exchange rate and bilateral trade using data from China, Japan, and the US during the period 2002–2007. My first estimation equation in the system is the bilateral export equation. Since the gravity model is successful in explaining the growing trade volume since World War II (Feenstra 2003), I therefore use an augmented gravity equation to estimate the effect of the bilateral exchange rate on exports. My second estimation equation is the bilateral exchange rate equation, which takes various determinants into account. I use a three-stage least-squares (3SLS) estimation to take full account of the joint correlations of the error terms between the two equations. Overall, I find that the revaluation of the RMB against the US dollar reduces China’s exports to the US whereas there is no significant impact on China’s exports to Japan. Simultaneously, the effect of exports on the bilateral exchange rate is insignificant for both the Sino-US and the Sino-­ Japanese cases. Various robustness checks confirm these findings by using different measures of exchange rates, econometric methods, and data period coverage. The rest of this chapter is organized as follows. Section 2 introduces the evolution of China’s bilateral exchange rate and its trade with Japan and the US. Section 3 discusses the determinants of both bilateral trade and the exchange rates. The main estimation results and robustness checks are presented in Sect. 4. Section 5 concludes the chapter.

2   China’s Exchange Rate and Triad Trade According to China’s Statistical Yearbook, published by the National Bureau of Statistics (2008), the bilateral trade between China and Japan increased dramatically since 2002. After China acceded to the WTO in 2001, the bilateral trade volume (i.e., exports plus imports) between China and Japan reached US$101.9 billion, with a 16.1% annual growth. Japan was China’s largest trading partner in 2002: the bilateral trade volume accounted for 16.5% of China’s overall trade volume, which was higher than the 12.9% with the US. Since then, the average growth rate of Sino-Japanese bilateral trade has been about 25%. In 2006, bilateral trade between China and Japan reached US$207.4 billion, which accounted for 11.7% of China’s overall trade volume. This volume was smaller than the Sino-US trade volume, worth US$262.7  billion, making Japan China’s second largest trading partner in the world since 2004.

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

35

In the overall whole trade volume, Japan has maintained a modest trade surplus with China in the new century. The bilateral trade imbalance was US$5  billion in 2002. China then became Japan’s largest source for imports for a share of 18.3% of Japan’s total import volume, which is higher than the 17.1% in the case of the US. The imbalance gap has widened over time. In 2006, China had a trade deficit with Japan worth US$24  billion, which accounted for around 12% of the overall bilateral trade volume. According to reports by China’s General Administration of Customs and the Department of Commerce in the US, Sino-US bilateral trade also increased rapidly after China acceded to the WTO. Simultaneously, China maintained a huge bilateral trade surplus with the US. In 2004, the bilateral trade was worth US$161 billion. More importantly, the Multi-Fiber Agreements, which set an upper bound for textile exports from China to the US in the Uruguay round of the GATT, were automatically terminated in January 2005. Accordingly, China’s textile exports to the US increased dramatically soon after that. In 2005, the trade imbalance gap widened to around US$200  billion. Due in part to appreciation of the RMB, China’s bilateral trade surplus with the US was reduced from US$232 billion in 2006 to US$213 billion in 2007. In 2008, the Sino-US trade volume did not increase by very much because of the stronger RMB and the shrinking demand in the US caused by the financial crisis. However, China still maintained a US$170  billion trade surplus with the US, accounting for 57.8% of China’s total trade surplus. Due to the surge in the Sino-US bilateral trade, special interest groups, such as labor unions, in the US appealed to the US government by arguing that China had manipulated its currency at an unreasonable level. They argued that China had a serious real exchange rate misalignment such that China could maintain a huge bilateral trade surplus. In response to the demand by special interest groups, the US congress threatened to impose trade sanctions on China if China did not “voluntarily” restrain its exports to the US, or revalue the RMB.  To avoid a possible trade war, China adjusted its fixed exchange rate against the US dollar, which had been adopted for one decade. In July 2005, the RMB against the dollar was revalued by 2% (see Fig. 2.2). It was pegged no longer solely to the US dollar but to a basket of currencies including, among others, the US dollar and the Japanese yen. Since then, the RMB was continuously revalued. In the next three years, the RMB against the dollar was revalued by around 20% from 8.3 to 6.8 RMB per dollar.

36 

M. YU

Fig. 2.2  China’s bilateral exports and exchange rates. (Source: CEIC Database)

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

37

3   Review of Related Theory This section specifies the bilateral trade equation, the exchange rate equation, and the simultaneous bilateral export and exchange rate ­ equations. 3.1  The Determinants of Bilateral Trade In a seminal work, Feenstra (2003) highlighted three reasons that explain the growing bilateral trade since World War II: growing GDP, declining transportation costs, and trade liberalization. The gravity model is expected to be the only successful model to explain the growing trade volume. It is easy to understand that the GDP growth of two trading partners plays a significant role in determining their bilateral trade. The gravity model suggests that larger countries trade more since they produce more commodities. Also, two countries trade more if the sizes of their economies are similar (Helpman 1987). In addition, rich countries (i.e., countries with higher per capita GDP) are expected to trade more. Later, Anderson and van Wincoop (2003) provided a solid theoretical micro-­ foundation for the typical gravity model by carefully introducing the “multilateral trade resistance” term, which specifies the implicit price indices in the gravity equation. Traditional wisdom suggests that international trade agreements foster international trade. After a 15-year long march, China successfully acceded to the WTO as its 143rd member in 2001. The impact of WTO accession on the Chinese economy has been substantial. Some researchers like Woo (2001) argued that WTO accession was a key component to reconstruct the Chinese economy. At the very least, the accession to WTO helped China enjoy a larger foreign market, which in turn fostered exports. Besides multilateral trade agreements, trade liberalization, such as tariff reduction and nontariff barriers, is important to bilateral trade growth. Shortly after it began its economic reforms, China set up a whole system of tariffs in the 1980s. After the Uruguay Round of the WTO, China experienced huge tariff reductions due, in large part, to its eagerness for WTO accession. China cut its tariffs from 35% in 1994 to around 17% in 1997. In 2001, China further cut its average tariff rate from 16.4% to 15.3%. Equally importantly, the bilateral exchange rate plays another key role in bilateral trade. Previous studies like Feenstra (1989) argued that there

38 

M. YU

is a symmetric response of import prices to changes in import tariffs and the bilateral exchange rates. This hypothesis is supported by Japanese and the US industrial data. The economic rationale for the effect of the exchange rate on bilateral trade seems straightforward. A change in the nominal bilateral exchange rate has a pass-through effect on import prices. Accordingly, both of these changes affect the bilateral trade. Put another way, an appreciation in the real exchange rate, which is defined as an increase in the relative price of tradable to nontradable goods, would lead to a decrease in bilateral trade. Of particular interest in this chapter is the effect of movements in China’s exchange rate on its bilateral trade with two other giants: Japan and the US. To estimate its effect, I control GDP of China and its trading partner in the estimations inspired by the gravity model. I drop data before 2002 to avoid the structural change in the Chinese economy caused by China’s WTO accession in 2001. Also, to keep the model neat, the usual variables of transportation costs such as bilateral geographic distance are captured in an error term in the empirical model.1 3.2   The Determinants of the Exchange Rate As summarized by Meese and Rose (1991), there are five models to explain the determinants of nominal exchange rates: a flexible-price monetary model, two sticky-price models, and two Lucas-type (1982) models. In all of these models, the bilateral spot exchange rate (ej) is determined, at least, by both the nominal domestic (i.e., China) money supply relative to foreign (MCH/Mj) and domestic industrial production relative to the foreign money supply (YCH/Yj). These common variables gain special theoretical support in Lucas’s (1982) model of a two-good, two-country, pure exchange economy. Shortly after that, Hodrick (1988) extended Lucas’s (1982) and Svensson’s (1985) models to include the change in the relative money growth rate in the model to capture the timing of money market transactions. The other three models have different extensions to the benchmark setup introduced by Lucas (1982). In particular, the flexible-price monetary model includes a nominal interest differential since it assumes that the 1  The inclusion of various bilateral geographic variables did not change my estimation results since such variables will be dropped automatically in the two-way fixed-effects estimations.

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

39

purchasing power parity (PPP) still holds when the home country faces an exogenous real exchange rate shock. In contrast, the two sticky-price-type models emphasize that the adjustment of goods prices is lower than that of asset prices. Therefore, one of the sticky-price-type models assumes that the real interest differential, measured by the difference between interest rates and inflation, is included in the estimation. Another sticky-price-type model instead argues that the relative cumulated trade balance (TBj) is an appropriate explanatory variable. That is,

e j  f  TB j ,MCH / M j ,YCH / Y j   error

(2.1)

where the bilateral nominal exchange rate, ej, is measured as China’s price of a unit of domestic exchange. Since my main interest in this chapter is to explore the endogenous nexus between bilateral trade and bilateral exchange rates, I therefore adopt specification (2.1) to capture the effect of China’s bilateral trade on its exchange rate. Previous research on real exchange rates takes special interest in the extent of its misalignment. It is usually believed that there exists an equilibrium exchange rate in which both internal and external balances are achieved. The gap between the estimated equilibrium and the actual exchange rate is the so-called real exchange rate misalignment (Williamson 1994). There are two major approaches to identifying the misalignment (Zhang 2001). One of them is based on the idea that the equilibrium concept is derived from the macroeconomic balance. Based on this, the misalignment is measured either by PPP or the black market exchange rate. Another approach is the so-called Behavioral Equilibrium Exchange Rate (BEER): the equilibrium exchange rate is determined by a variety of explanatory variables of economic fundamentals. Since my objective in this chapter is to estimate the effect of the exchange rate on trade, I do not attempt to measure the misalignment of China’s real exchange rate. However, I use the real exchange rate as another indicator of the exchange rate to estimate its effect on bilateral trade. 3.3  Empirical Methodology Accordingly, I introduce the following simultaneous equation model (SEM) for the estimations:

40 

M. YU

CH ln X CH   3 ln Y jkt   k  t  kt jkt   0  1e jt   2 ln Ykt





(2.2)

CH e jt   0   1 ln X CH   3 ln Y jkt   4 ln M tCH   5 ln M jt   kt (2.3) jkt   2 ln Ykt

where X CH jkt denotes China’s k exports to country j in year t. Correspondingly, the new variables in the bilateral exchange rate Eq. (2.3) M tCH and Mjt are China and its trading partner’s j monetary bases, respectively. I also include, though not listed in the equations above, the j-period time lag of the exchange rate, ejt-1, in both equations as robustness checks. Following Feenstra (1989), the expected exchange rate in each quarter is a log-linear function of the current and past three quarterly-average spot rates. As theoretically recognized by Anderson and van Wincoop (2003), standard gravity estimations on bilateral trade could suffer from the bias caused by “multilateral trade resistance,” which measures the implicit price indices in the gravity model. When the data set is a panel, the regular OLS estimates are biased if the trade resistance is ignored.2 To control for multilateral resistance among trading partners, inspired by Rose and van Wincoop (2001), I use fixed effects to control for other unobservable features within each industry of the trading partners over time. In particular, ηk captures the unobserved industry-specific time-invariant fixed effects, whereas φt is the time-varying fixed effects. Since the samples are quarterly data, both year-specific and quarter-specific fixed effects are included to completely capture the time-specific fixed effects. Turning to the exchange rate equation, in addition to bilateral trade, the bilateral exchange rate is affected by the trading partners’ GDP and monetary base, as inspired by the sticky-price-type models. The error terms (μkt, εkt) are a bivariate residual vector. In this SEM, the coefficients β1 and γ1 consider the simultaneous feedback from bilateral trade and the exchange rate, which are my main interests. Since the error terms in the SEM are generally correlated with the dependant variables, the conventional methods such as Ordinary Least Squares (OLS) and Generalized Least Squares (GLS) are inconsistent. In this case, as introduced in Wooldridge (2002), the 3SLS is appropriate to take the error-term correlations between the two equations into account.

2  Such an estimation bias is called the “gold metal error” of estimating the gravity trade model (Baldwin and Taglioni 2006).

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

41

4   Data, Econometrics, and Results In this section, I first describe the data sets used in the chapter, followed by a discussion of the Sino-US estimations and the Sino-Japanese estimations. I then address the possible endogeneity problem. Finally, I close the section with various robustness checks. 4.1  Data My data coverage is from the first quarter of 2002 to the last quarter of 2007. The economic rationale of focusing on this window is that China’s economy was significantly affected by its WTO accession in 2001, whereas China’s exchange rate was stable at that time. Since my objective in this chapter is to estimate the effect of the exchange rate on bilateral trade, I therefore drop observations before 2001 to avoid the possible structural change caused by the WTO accession shock. I use log directional industrial imports of the US (or Japan) from China to measure bilateral trade among China, Japan, and the US. This is because directional imports are consistent with the prediction of the gravity model, which only emphasizes one-way trade flow (Baldwin and Taglioni 2006). It is recognized that there is a mismatch problem between using data on China’s exports and American imports due to China’s re-export (via Hong Kong) problem (Feenstra and Hanson 2004): exports from China via Hong Kong are counted as American imports from China but they are not counted as China’s exports to the US.  To be consistent with previous works using the standard gravity model, I use American import data to measure Sino-US trade. In addition, I use quarterly-average rates to measure the bilateral nominal spot exchange rate. In this way, I can avoid the daily random error caused by adopting spot rates instead (Feenstra 1989). Unless specified, all data are from the CEIC database, which is publicly available.3 The directional import data is at the SITC two-digit level. Trading partners’ GDPs are measured in constant US dollars. Data on American GDP is disaggregated by sectors (NAICS) and are available from the Bureau of Economic Analysis (BEA). China’s producer price index can be accessed from China’s Statistical Yearbook (2008), published by the National Bureau of Statistics of China, which specifies the base year of the PPI as 1995. The American PPI is obtained from the Bureau of  Data source: http://www.ceicdata.com.

3

42 

M. YU

Labor Statistics, which specifies the base year of the PPI as 2000. Similarly, I get the Japanese data from the Ministry of Economy, Trade, and Industry (METI). Finally, as usual, I use M1 to measure the monetary base. Sixty-two industries are covered in the Sino-US trade observations for six years (i.e., 2002–2007), whereas only 41 industries are covered in the Sino-Japanese trade observations for five years (i.e., 2002–2006). I have 1,482 quarterly observations for the Sino-US estimations, whereas I have only 770 quarterly observations for the Sino-Japanese estimations.4 In Table 2.1, the upper half presents the descriptive statistics for each variable in the Sino-US bilateral trade estimations whereas the lower half reports those in the Sino-Japanese bilateral trade estimations. 4.2  The Sino-US Estimates Table 2.2 presents the estimation results in which Eqs. (2.2) and (2.3) are estimated separately. These serve as a benchmark for comparison with the estimated results from the simultaneous equation method (SEM). Column (1) is the simple pooled OLS estimate. The coefficient of the log exchange rate is −1.226, which is significant at the conventional statistical level. The economic rationale is that a percent increase in the dollar to RMB exchange tends to have a 1.226% decrease in China’s exports to the US. As introduced in Eq. (2.2), I also control for both trading partners’ GDP.  It turns out that the coefficients on trading partners’ GDP have an anticipated positive sign, which is consistent with the theoretical prediction that larger countries trade more. Since my data set is panel data, I also perform the fixed-effects estimation in column (3) and find that the coefficient of the log of the exchange rate is still negative and significant. We might expect that the previous realization of exchange rates could play a role in affecting current exports, inspired by the typical J-curve argument: in response to a domestic devaluation, a country’s trade balance typically worsens first before becoming better. I therefore include the past three quarterly-average spot rates in column (2) for the pooled OLS estimates and column (4) for the fixed-effect estimates. It turns out that most of those past exchange rates do not have significant effects on exports. In any case, China’s RMB appreciation (i.e., an increase in the log exchange rate) is associated with lower exports from China to the US. Also, 4  Six observations are missing from the Sino-US bilateral trade, and 50 are missing from the Sino-Japanese bilateral trade.

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

43

Table 2.1  Summary statistics Variables

Observation

Mean

Std. Dev.

Minimum Maximum

Panel A: basic statistics of Sino-US bilateral trade (2002–2007) Log GDP of US (million) 1488 3.967 0.531 2.679 Log GDP of China 1488 5.723 0.120 5.560 PPI for China 1488 100.250 32.752 27.328 PPI for the US 1312 143.349 33.626 82.133 Log exchange rate ($/RMB) 1488 −0.888 0.102 −0.918 Log real exchange rate ($/ 1312 −1.043 0.039 −1.325 RMB) 1-Lag of log exchange rate 1426 −0.909 0.012 −0.918 ($/RMB) 2-Lag of log exchange rate 1364 −0.911 0.012 −0.918 ($/RMB) 3-Lag of log exchange rate 1302 −0.913 0.008 −0.918 ($/RMB) Log China’s Monetary Base 1488 3.968 0.110 3.775 (M1) Log American Monetary 1488 3.121 0.022 3.075 Base (M1) Year 1488 2.004 1.708 2002 Industrial Code for Sino-US 1488 31.5 17.901 1 Trade Panel B: basic statistics of Sino-Japanese bilateral trade (2002–2006) Log GDP of Japan (million) 820 3.636 0.373 2.903 Log GDP of China 820 6.599 0.088 6.478 PPI for China 820 111.440 48.360 81.412 PPI for Japan 780 103.038 14.060 97 Log exchange rate (¥/RMB) 820 1.147 0.028 1.101 Log real exchange rate (¥/ 780 1.151 0.136 0.856 RMB) Log exchange rate (¥/$) 820 2.059 0.026 2.012 Log exchange rate ($/RMB) 820 −0.913 0.006 −0.924 Log Chinese Monetary Base 820 3.935 0.089 3.775 (M1) Log Japanese Monetary Base 820 6.037 0.031 5.949 (M1) Year 820 2004 1.415 2002 Industrial Code for 820 12 11.839 1 Sino-­Japanese Trade

5.147 5.919 313.420 286.933 −0.398 −0.632 −0.889 −0.885 −0.889 4.171 3.140 2007 62

4.447 6.724 303.105 184.3 1.204 1.661 2.118 −0.901 4.086 6.078 2006 41

2.257 (1.000) −2.472 (0.341)

1.921 (0.998)

−1.531 (0.517)

ADF test with trend

−2.074 (0.255)

3.427 (1.000)

Phillips–Perron test

0.414 0.063

10.810 1.174

Trace statistics

15.41 3.76

5% Critical value

ADF test with trend

0.490 0.326

Trace statistics 18.167 6.719

HQIC −12.871 −13.805 −14.816* −14.548

AIC −12.871 −13.825 −14.855* −14.607 First difference Eigenvalue

−3.975** (0.009) −2.971* (0.140)

−3.181* (0.021) −2.915* (0.043)

2.206 (0.998) 1.234 (1.000) −4.880** −4.751** (0.000) (0.000)

First difference ADF test

15.41 3.76

5% Critical value

SBIC −12.871 −13.629 −14.462* −14.019

−19.513** (0.000)

−2.991* (0.035)

−4.858** (0.000)

2.241 (0.998)

Phillips–Perron test

Notes: Numbers in parenthesis are p-value in panel A. Double asterisks (**) denote the significance at 1% level, and single asterisk (*) denotes significance at 5% level

0 1 2

Eigenvalue

Sino-Japanese data Log exchange rate −1.716 (0.423) −1.272(0.895) −2.052(0.264) (¥/RMB) Log Imports to −0.129 (0.946) −2.437 (0.360) −4.377 (0.000) Japan from China Panel B: determination of order of lags using Sino-Japanese data Number of lags Likelihood Likelihood ratio FPE 0 61.161 3.3e-06 1 73.271 24.221 1.3e-06 2 86.023 25.505* 4.6e-07* 3 87.916 3.7855 6.4e-07 Panel C: Johansen tests for cointegration using Sino-Japanese data Maximum rank Level

Sino-US data Log exchange rate ($/RMB) US from China

ADF

Panel A: unit root tests Variables Level

Table 2.2  Time series properties of the data 44  M. YU

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

45

the coefficient of the bilateral exchange rate remains stable across these estimates.5 As predicted by the theoretical model (1), the bilateral exchange rate is not exogenously given but is indeed affected by bilateral trade. As discussed in the previous section, I include other determinants of the bilateral exchange rate such as trading partners’ GDP and monetary base (M1) in columns (5) and (6). The coefficients of exports on the bilateral exchange rate are significantly negative. This is a striking unexpected result: with greater exports from China, the US government is more eager to push the Chinese government to revalue the RMB. Therefore, if China’s surging exports play a role in the Sino-US bilateral exchange rate, it should lead to a revaluation of the RMB. The unexpected estimated results in columns (5) and (6) are possibly because of the lack of a control for the simultaneous bias in the estimations. Table 2.3 reports the 3SLS estimates of the simultaneous equation system. In specification (2.1), I adopt a simple form of the gravity equation to estimate the effect of the bilateral exchange rate on bilateral trade. Aside from the bilateral exchange rate, Sino-US bilateral trade is affected by the two trading countries’ GDPs. Simultaneously, the bilateral exchange rate is affected by the trading countries’ GDPs and monetary bases. In the bilateral trade equation, the coefficient of the bilateral exchange rate is significantly negative. The elasticity of the exchange rate on bilateral trade is −0.563, which is much lower in absolute value than the result in column (3) of Table 2.3. This suggests that the effect of the current exchange rate on bilateral trade is overestimated if we do not control for the simultaneous bias caused by the reverse causality of bilateral trade on the exchange rate. Turning to the exchange rate equation, the coefficient of China’s exports to the US is not significant. Also, its magnitude is close to nil. The idea behind this estimated result is that, once the simultaneous bias between the exchange rate and bilateral trade is taken into account, China’s surging exports to the US did not have a significant effect on the

5  It is worth pointing out that the sign and magnitude of the bilateral exchange rate do not change substantially even when the trading partners’ per-capita GDPs are included in the estimations, though the magnitude of trading partners’ GDPs would be affected instead. To save space, I do not report the estimation results with per-capita GDP in the text, though they are available upon request.

46 

M. YU

Table 2.3  Estimates of China’s exports to the US (2002–2007) Separate estimation Log imports from China to the US Log exchange rate ($/ RMB) Log exchange rate (1-lag) Log exchange rate (2-lag) Log exchange rate (3-lag) Log GDP of US Log GDP of China Log M1 of China Log M1 of US Year-specific fixed effects Industry-­ specific fixed effects Prob. > F or Prob. > χ2 Number of observations R2

Bilateral export equation

Exchange rate equation

(1)

(4)

(2)

(3)

−1.226** −1.095** −1.020** −1.194** (−3.31) (−2.19) (−16.05) (−14.39)

(5)

(6)

−0.005 ** (?2.70)

−0.003**(−2.15)

−0.749 (−0.03)

6.061* (1.67)

19.844** (12.26)

1.744 (0.05)

−7.692 (−1.24)

−19.661** (−6.38)

3.434 (−0.13)

8.454* (1.87)

23.821** (7.19)

No

No

Yes

Yes

−0.023** (−6.64) −0.682** (−2.68) 1.981** (6.03) −4.385** (−12.19) No

No

No

Yes

Yes

No

Yes

0.00

0.00

0.00

0.00

0.00

0.00

1482

1296

1482

1296

1482

1296

0.03

0.02

0.48

0.46

0.45

0.59

0.189** (2.72) 1.116** (3.46)

0.172** (2.21) 1.079 (1.44)

0.065** (2.41) 1.035** (2.61)

0.060** (2.26) 0.923** (2.41)

−0.035** (−8.59) −0.329 (−1.08) −2.070** (−4.58) 6.421** (11.58) Yes

Notes: Numbers in parenthesis are t-values. Asterisk * (**) indicates significance at the 1 (5) percent level

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

47

bilateral exchange rate, though the revaluation of the RMB did decrease China’s exports to the US. I now look for more evidence by adopting different specifications. I first check for the endogenous nexus between the exchange rate and bilateral trade by including previous exchange rate realizations in both equations. In this way, the past exchange rates are allowed to affect the current bilateral trade and provide additional information to form the current exchange rate due to rational expectations and other reasons. As shown in column (2) of Table 2.3, the past three-period Sino-US bilateral exchange rates have no significant effects on China’s exports to the US, though they significantly affect the formation of its current bilateral exchange rate. 4.3  The Sino-Japanese Estimates Turning to the Sino-Japanese case, I first estimate the effect of the Sino-­ Japanese bilateral exchange rate on China’s exports to Japan by using bilateral trade data from 41 industries during 2002–2006. As shown in column (1) of Table 2.4, the coefficient of the exchange rate on trade is negative but insignificant. I suspect that this is due to the lack of consideration of the specific characteristics of the panel data set. I therefore perform the two-way fixed-effects estimation in column (2). Strikingly enough, the coefficient of the Sino-Japanese bilateral exchange rate on bilateral exports is significant and positive. Given that the RMB maintained a fixed exchange rate with the US dollar before 2005, the variation in the Sino-Japanese exchange rate must come from the movement of Japanese yen against the US dollar. I therefore decompose the bilateral exchange rate of the Japanese yen against the RMB into two components: the bilateral exchange rate of the Japanese yen against the US dollar and that of the US dollar against the the RMB.  Column (3) reports the estimated coefficients of these two variables. It turns out that the Japan–US bilateral exchange rate did not have a significant effect on China’s exports to Japan. The bilateral Sino-US exchange rate instead had an expected positive effect on the Sino-Japanese bilateral trade. However, careful readers will observe that the variable of China’s GDP drops in columns (2) and (3) possibly due to multicollinearity in the estimations. This serves as side evidence that the simultaneous bias between trade and the exchange rate must be large. I therefore perform the 3SLS to control for this simultaneous bias. As shown in column (4) of Table 2.4, I include the Sino-Japanese bilateral

48 

M. YU

Table 2.4  Simultaneous estimates between imports from China to the US and the exchange rate (2002–2007) Joint estimations

(1) Export

Log imports from China to the US Log exchange rate ($/ −0.563** RMB) (−4.27) Log exchange rate (1-lag) Log exchange rate (2-lag) Log exchange rate (3-lag) Log GDP of US 0.094** (2.79) Log GDP of China 1.135** (2.77) Log M1 of China

Exchange rate

Export

−0.001 (−0.81) −1.366** (−4.30) 7.990 (1.63)

Yes Yes

−0.024* (−5.54) −0.713** (−9.57) 2.014** (21.90) −4.391** (−23.97) No No

0.00 1482 0.98

0.00 1482 0.45

Log M1 of US Year-specific fixed effects Industry-specific fixed effects Prob. > F or Prob. > χ2 Number of observations R2

(2) Exchange rate −0.001 (−0.59)

Yes Yes

19.915** (12.48) −19.809* (−9.40) 23.963** (13.80) −0.035** (−8.32) −0.319** (−3.75) −2.086** (−9.15) 6.414** (11.27) No No

0.00 1296 0.98

0.00 1296 0.59

−7.982 (−1.24) 9.655** (2.19) 0.47 (1.17) 0.926** (2.35)

Notes: Numbers in parenthesis are t-values. Asterisk * (**) indicates significance at the 1 (5) percent level

exchange rate and trading countries’ GDP in the export equation. In the exchange rate equation, I instead add the trading partners’ monetary base in the estimation. The coefficient of the Sino-Japanese bilateral exchange rate on the bilateral trade is positive but insignificant. Simultaneously, the coefficient of Sino-Japanese trade on the bilateral exchange rate is negative and also insignificant. This suggests that, even when the endogenous nexus between trade and export is taken into account, the Sino-Japanese bilateral exchange rate still has no significant effect on China’s exports to Japan.

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

49

4.4  Additional Robustness Checks Finally, given that China’s exchange rate against the US dollar was changed only after July 2005, the impact of the bilateral exchange rate on the bilateral trade volume is underestimated if data before that are included. I therefore reestimate the effects by including only the data since 2005. Table  2.5 first reports the Sino-US 3SLS estimations for data for 2005–2007. I find few differences by comparing these results to the estimation results in column (1) of Table 2.3 in terms of magnitude and sign. Table 2.5  Estimates of China’s exports to Japan (2002–2006) Separate/joint estimations

OLS

FE

(1)

(2)

Log Imports to Japan from China Log exchange rate −0.483 (¥/RMB) (−0.61) Log exchange rate (¥/$) Log exchange rate ($/RMB) Log GDP of Japan −0.092 (−1.23) Log GDP of China 1.273 (3.21) Log M1 of China

3SLS (3)

(4-1)

−0.001 (−0.58) 1.075** (2.29)

0.528 (0.25)

1.605** (5.49) –

0.582 (1.11) 3.223** (2.92) 1.447** (4.81) –

No

Yes

Yes

Yes

−0.003 (−1.30) 0.303** (7.60) −0.126** (−2.63) −0.780** (−9.15) No

No

Yes

Yes

Yes

No

0.0000

0.0000

0.0000

0.0000

770

770

770

770

770

0.02

0.39

0.40

0.97

0.22

1.650** (2.24) −0.627 (−0.87)

Log M1 of Japan Year-specific fixed effects Industry-specific fixed effects Prob. > F or Prob. > χ2 Number of observations R2

(4-2)

Notes: Numbers in parenthesis are t-values. Asterisk * (**) indicates significance at the 1 (5) percent level. In estimate (4), the dependent variable in the first equation is China’s export to Japan whereas that in the second one is the Sino-Japanese exchange rate

50 

M. YU

In the export equation, the coefficient of the Sino-US bilateral exchange rate is still significantly negative. However, its magnitude, in absolute value, is higher than its counterpart in column (1) of Table 2.3. This finding confirms that the effect of the bilateral exchange rate on bilateral trade is underestimated when we look at a longer period. Simultaneously, the effect of bilateral trade on the exchange rate is still insignificant. Turning to the Sino-Japanese case, we observe a very similar finding compared with column (4) of Table 2.4. The Sino-Japanese bilateral exchange rate did not have a significant effect on China’s exports to Japan, and vice versa. As mentioned above, the movement of the nominal bilateral exchange rate has a pass-through effect on the price of the imports, which in turn would affect bilateral trade. Therefore, it is worthwhile to examine the joint effects of both nominal exchange rate and price change on bilateral trade. Put another way, one can check for the effect of the real bilateral exchange rate on bilateral trade directly. Table 2.6 reports the 3SLS simultaneous estimates between trade and real exchange rate among the triad during the 2002–2007 period. Following the literature, I proxy the real bilateral exchange rate (rej) as the product of the nominal bilateral exchange rate and a fraction consisting of China’s producer price index (PPICH) in the denominator and its importer’s producer price index (PPIj) in the numerator. That is, rej = ­ej(PPIj/PPICH). In addition, I include the past real exchange rate realizations in estimations to allow them to affect the current real rate formation.6 Estimation results in the SEM for the Sino-US case show that the current real exchange rate did not have a significant influence on their bilateral trade. However, the real spot rate of half a year earlier (i.e., a two-period lag) has a significantly negative coefficient, which suggests that the past real exchange rate appreciation reduced the Sino-US bilateral trade. Simultaneously, China’s exports to the US had no significant effect on real exchange rate formation. A similar finding can be obtained from the 3SLS estimates for the Sino-Japanese case, as shown in Table 2.6. In a nutshell, all my results are robust to show that the revaluation of the RMB against the dollar significantly reduced China’s exports to the US, but there were no significant effects on China’s exports to Japan. These findings are robust to different measures, econometric methods, and data period coverage. 6  Using the wholesale price index (WPI) or the consumer price index (CPI) does not substantially change the estimation results in Table 2.6.

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

51

Table 2.6  3SLS simultaneous estimates (2005–2007) Joint estimations

China and US Export

China and Japan Exchange rate

Log Imports from China Log exchange rate ($/ −0.675** RMB) (−6.57) Log exchange rate (¥/ RMB) Log GDP of China 0.949** (2.76) Log GDP of Japan

−0.002 (−0.73)

Log GDP of US

−0.645** (−4.34) 2.145** (12.88)

0.098** (3.19)

Log M1 of China

−0.056 (−0.73)

Export

−0.000 (0.34)

0.670 (0.23) 1.067 (0.47) −1.756 (−0.50)

−0.287** (−10.01) −0.000 (−0.39)

0.589** (26.11) 1.533** (18.11)

Log M1 of Japan

Yes

−15.020** (−5.62) No

Yes

No

Yes

No

Yes

No

0.00 556

0.00 556

0.00 317

0.00 317

0.98

0.79

0.98

0.90

Log M1 of US Year-specific fixed effects Industry-specific fixed effects Prob. > F or Prob. > χ2 Number of observations R2

Exchange rate

Notes: Numbers in parenthesis are t-value. Asterisk * (**) indicates significance at 1 (5) percent level

5   Concluding Remarks In this chapter I investigate the effect of the revaluation of the RMB on bilateral trade among China, Japan, and the US by using industrial panel data from the 2002–2007 period. Different from previous one-way estimations, I use simultaneous equation methods to take into account the endogenous nexus between bilateral exchange rates and bilateral trade. Thanks to this method, I am able to explain the results both statistically and economically. The estimation results clearly suggest that the

52 

M. YU

revaluation of the RMB against the dollar significantly reduced Sino-US trade but it had no significant effects on Sino-Japanese bilateral trade. The policy implication for this finding is that the revaluation of the RMB was helpful in reducing the bilateral Sino-US trade imbalance and accordingly in avoiding a possible trade war between the two countries. Several extensions and possible generalizations merit special consideration. One of them is to introduce a theoretical gravity model to serve as the empirical estimate. This way, the exchange rate pass-through channel can be more precisely presented. Another possible extension is to include other protection instruments like export tax rebates into the two equations so that the model can be closer to reality. Due to data restrictions, I am not able to explore these issues here. However, these are the topics that I will pursue in future work.

References Anderson, James and Eric van Wincoop (2003), Gravity with Gravitas: A Solution to the Border Puzzle, American Economic Review 93(1), pp. 170–192. Baldwin, Richard and Daria Taglioni (2006), “Gravity for Dummies and Dummies for Gravity Equations,” NBER Working Papers, No. 12516. Bergin Paul R. and Robert C. Feenstra (2008), “Pass-through of Exchange Rates and Competition between Floaters and Fixers,” Journal of Money, Credit and Banking, 41(s1), pp. 35–70. Feenstra, Robert and Gordon Hanson (2004), Intermediaries in Entrepôt Trade: Hong Kong Re-Exports of Chinese Goods, Journal of Economics & Management Strategy, 13(1), pp. 3–35. Feenstra, Robert C. (1989), Symmetric Pass-through of Tariffs and exchange Rates under Imperfect Competition: An Empirical Test, Journal of International Economics 27, pp. 25–45. Feenstra, Robert C. (2003), Advanced International Trade: Theory and Evidence. Princeton University Press. Goldberg, P.  K. & Knetter, M.  M., 1997, “Goods Prices and Exchange Rates: What Have We Learned?”, Journal of Economic Literature, Vol. 35(3), pp. 1243–1272. Helpman, Elhanan (1987). Imperfect competition and international trade: Evidence from fourteen industrial countries. Journal of the Japanese and International Economics, 1, 62–81. Hodrick, R. (1988). Risk, uncertainty and exchange rates. Journal of Monetary Economics, 23, 433–459. Lucas, Robert E., Jr. (1982). Interest rate and currency prices in a two-country world. Journal of Monetary Economics, 10, 335–359.

2  REVALUATION OF THE CHINESE YUAN AND TRIAD TRADE: A GRAVITY… 

53

Meese, Richard A., & Rose, Andrew K. (1991). An empirical assessment of non-­ linearities in models of exchange rate determination. Review of Economic Studies, 58, 603–619. National Bureau of Statistics (2008). China’s Statistical Yearbook 2008. Beijing: National Bureau of Statistics. Rose, Andrew K., & van Wincoop, Eric (2001). National money as a barrier to international trade: The real case for currency union. American Economic Review, 91(2), 385–390. Svensson, Lars (1985). Money and asset prices in a cash-in-advance economy. Journal of Political Economy, 93, 919–944. Williamson, John (1994). Estimating equilibrium exchange rates. Washington: Institute for International Economics. Woo, Wing Thye (2001). Recent claims of China’s economic exceptionalism: Reflections inspired by WTO accession. China Economic Review, 12, 107–36. Wooldridge, Jeffery M. (2002). Econometric analysis of cross section and panel data. Cambridge, Massachusetts: MIT Press. Zhang, Zhichao (2001), “Real Exchange Rate Misalignment in China: An Empirical Investigation,” Journal of Comparative Economics, 29, pp. 80–94.

CHAPTER 3

Exports, Productivity, and Credit Constraints

Recent Melitz-type (2003) intra-industry heterogenous trade models argue that a firm’s productivity has significant effects on the firm’s exports. This chapter examines how a firm’s credit constraints as well as its productivity affect its export decisions. We imbed the firm’s credit constraints into a Melitz-type general-equilibrium model by endogenizing the probability of the success of firm-specific projects. We show that, all else equal, it is easier for firms to enter the export market if (1) the probability of the success of their project is higher and consequently they have easier access to external finance from financial intermediaries; or (2) they have alternative sources, other than from financial intermediaries, to obtain funds. We test these theoretical hypotheses using firm-level data from Chinese manufacturing industries and find strong evidence supporting the predictions of the model.

1   Introduction A widely accepted belief about a firm’s heterogeneous export behavior is that a firm with higher productivity would generate greater revenue and thus be able to shoulder the fixed costs of market entry (Melitz 2003). This interpretation abstracts from financial frictions that may arise from This chapter is coauthored with Zhiyuan Li at Fudan University, China, in 2010. © The Author(s) 2021 M. Yu, Exchange Rate, Credit Constraints and China’s International Trade, https://doi.org/10.1007/978-981-15-7522-8_3

55

56 

M. YU

the firm’s access to liquidity or to external finance. In the presence of financial frictions, however, firms face different borrowing constraints, which could in turn affect their capabilities to finance the upfront entry costs. In addition, inward foreign direct investment (FDI) may provide an alternative source of low-cost funds and alleviate credit constraints faced by firms. These possibilities raise the following questions: How do firms’ credit constraints affect their exports? Do credit constraints vary across firms’ types? This chapter addresses the issue by providing a general-equilibrium Melitz-type (2003) model to analyze the impact of a firm’s heterogeneous external finance capability on its exports. There are two main innovations in our theoretical model. First, firms’ export projects have different probabilities of success. More importantly, the heterogeneous probabilities of success induce different capabilities of firms to obtain external finance from financial intermediaries. Second, different types of firms have different abilities to obtain low-cost funds from sources other than financial intermediaries. In particular, subsidiaries of multinational corporations are able to receive financial support from their parent firms to bypass domestic credit constraints. Inspired by the theoretical predictions, we estimate the effect of a firm’s credit constraints on its exports, using a very rich panel data set of Chinese manufacturing firms over the period of 2000–2007. We obtain robust empirical evidence that firms with fewer credit constraints export more, while controlling for the endogeneity of credit constraints. Overall, 1% increase in a firm’s interest expenditures, which is used as a proxy for the firm’s capability to borrow, increases its exports by 1.068%, ceteris paribus. In addition, foreign-invested-enterprises (FIEs) in China export more and are less sensitive to the availability of external finance indicating that the presence of foreign capital does alleviate the credit constraints faced by firms. Finally, various sensitivity checks strongly support our empirical findings. This work adds to a small but growing literature on trade and finance, including work done by, among others, Qiu (1999), Chaney (2005), Manova (2008), Muûls (2008), and Buch et al. (2008). Inspired by Melitz (2003), Chaney (2005) first introduced liquidity constraints into the heterogeneous firms model and predicted that firms with higher liquidity endowment would face fewer financial constraints and it would consequently be easier for such firms to enter the export market. As a pioneering work in the field, Chaney (2005) theoretically focused only on the impact of a firm’s internal funds on its exports. Then, Manova (2008) took a step forward to consider financial contracts and asset tangibility in

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

57

a framework of firm heterogeneity. She found that, using sector-level panel data of bilateral exports, those industries in more financially developed countries are more likely to export bilaterally and to ship greater volumes. In addition, Muûls (2008) combined both liquidity endowments and external financial contracts into a general equilibrium model and showed that the credit rating, the Coface score from a credit insuring company, has significant effects on a firm’s exports. However, Muûls’s theoretical model assumes equal cost of external finance following Manova (2008). That is, the repayment required for an identical principal is the same across all firms. Last but not least, Buch et al. (2008) investigated the impact of credit constraints on exports and outward FDI, including firm heterogeneity in borrowing costs in a partial equilibrium framework. In retrospect, all these works have enriched our understanding of the impact of a firm’s credit constraints on its export behavior. Most of the previous studies assume that, other than productivity, firms are different only in terms of the principal amount they need to borrow. Once this amount is given, there are no differences in repayments. This stems from the assumption made by these studies that the default rates across firms are the same. Our work introduces two more sources of heterogeneous credit constraints. Project-specific risks lead to different borrowing capabilities and capital from foreign parent firms causes firms to be less dependent on external finance. These additional sources of credit constraints lead to a better fit to reality. Equally importantly, though some of the previous studies worked on firm heterogeneity theoretically, they estimated their models only at the country and sector levels, in part due to data restrictions. In this chapter, we are able to test our theoretical predictions using disaggregated firm-level data from China. These findings are of special interest because China’s exports experienced unprecedented growth over the past decade whereas Chinese firms faced severe credit constraints. Over the period of 2000–2007, the annual growth rate of China’s exports was 25%. However, according to the World Business Environment Survey during 1999–2000 and the Investment Climate Assessment surveys in 1999 and 2002, China was among the group of countries that had the worst financing obstacles (Claessens and Tzioumis 2006). Our results suggest that easier access to external finance through financial intermediaries and/or FDI would make the engine of China’s exports more powerful. Figure 3.1 illustrates this point. During 2000–2007, FDI is positively associated with exports across Chinese cities as shown in Fig. 3.1, whereas interest expenditures are positively associated with exports across industries, as shown in Fig. 3.2.

Fig. 3.1  Log exports vs. log inward FDI at Chinese City Level during 2000–2007. (Source: Data are from Chinese City Statistical Yearbook, various years)

Fig. 3.2  Log exports vs. log interest expenditure at two-digit industrial level during 2000–2007. (Source: Authors calculation based on the data set)

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

59

The remainder of the chapter is organized as follows. Section 2 describes mechanisms by which a firm’s credit constraints affect its exports. Section 3 constructs a model introducing a firm’s heterogeneous access to external finance and the impact of external finance impact on the firm’s exports. Section 4 presents the econometric specification and data. Section 5 discusses the estimation results and provides robustness checks. Section 6 concludes the chapter.

2   Exploring the Impact of Credit Constraints on Exports The role of a firm’s productivity in its exports has been discussed in the literature. The seminal work by Melitz (2003) highlights that the most efficient firms export and earn extra profits abroad. Less efficient firms can only serve the domestic market since the entry to the foreign market would generate losses due to fixed entry costs. In contrast, the least efficient firms die and exit from the market. Inspired by pioneering works by Chaney (2005) and Manova (2008), in this chapter we take a step forward to identify that a firm’s credit constraints are one of the most important channels through which a firm’s productivity affects its exports. Firms have heterogeneous capabilities in obtaining external finance. Firms with higher productivity would be more capable of obtaining loans. The intuition is straightforward. First, the more efficient firms are expected to have higher probability of success in export projects. Thus for the same size loan and the same amount of repayment, financial intermediaries are more likely to extend loans to firms with higher productivity because it is more likely that the intermediaries will be repaid. With projects with low probability of success, investors expect higher risks and thus are reluctant to extend loans to these projects. Incapable of obtaining external finance, less efficient firms would face more severe credit constraints and might be prevented from entering the export market, or they have to export less if they also face credit constraints in financing variable costs as well. In addition to borrowing from financial intermediaries, firms are able to finance internally with their domestic profits. Aside from this, one interesting observation is that the firm’s type also plays a role to its access to alternative sources of finance. One good example is that foreign-invested-­ enterprises (FIEs) in China, compared with non-FIEs, may have easier

60 

M. YU

access to funds from their parent firms.1 As a result, FIEs are expected to export more given all other constants. Indeed, the share of exports by FIEs in China increased from 48% in 2000 to 57% in 2007. The average annual growth rate (i.e., 29.6%) is higher than that of China’s total exports (i.e., 25%) during 2000–2007. In summary, all else equal, it is easier for firms to enter the export market if they have higher productivity, easier access to external finance from financial intermediaries, or easier access to alternative sources of funds. In this chapter, we first theoretically show that firms that have higher probabilities of success probabilities, especially FIEs, are less financially constrained and thus less likely to be prevented from exporting. The intuition is that, when firms need to borrow from financial intermediaries, the low probability of success leads to high repayment requirements by the intermediaries to cover investment risks. This in turn increases the export threshold. Therefore, firms that otherwise are productive enough to export could be prevented from exporting due to higher requirements for repayment. FIEs that have access to alternative funds from their parent firms need to borrow less and are less sensitive to external finance. In this sense, FIEs are less credit constrained. We evaluate the importance of access to external finance to firms’ export behaviors by testing two hypotheses: (1) firms for which it is easier to borrow from financial intermediaries export more; and (2) FIEs export more and are less sensitive to the availability of external finance from financial intermediaries. By choosing firms’ interest expenditures as a proxy for firms’ external financing capabilities, we are able to explore the nexus among interest expenditures, access to foreign capital, and exports. Equally importantly, we also pin down the effect of foreign capital access on exports through the channel of credit constraints. All the estimation results have the expected sign, are statistically significant and are robust to different specifications. Finally, we address the endogeneity issue of the effect of the firm’s exports on its interest expenditures. The main idea is that, with more exports, firms need more upfront fixed costs. We choose the scale-weighted money supply in the previous period as an instrument of interest expenditures. Various statistical tests confirm that the instrument is well justified. 1  Héricourt and Poncet (2009) show that inward foreign direct investment in China plays an important role in alleviating the domestic firms’ credit constraints. Similar evidence is provided by Harrison and McMillan (2003) using data from the Ivory Coast.

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

61

After controlling for endogeneity problems, interest expenditures are shown to have an even greater impact on firms’ exports. Motivated by these observations, in the next section we develop a general equilibrium model aimed at capturing the impact of a firm’s credit constraints on its exports.

3   The Model 3.1  Domestic Demand and Production The economy consists of two countries: home and foreign (henceforth, foreign counterparts of the variables are denoted with an asterisk *). Labor is the only input factor for production. The population is of size L at home. There are two sectors. One sector produces a single homogeneous good that is freely traded. Production in this sector is characterized by constant returns to scale with q0 = wl0, where l0 is the labor used to produce quantity q0 of the good. Thus, labor productivity in this sector determines the wage level w at home. We assume that both countries produce in this sector and that wages are thus fixed by the productivity in this sector. The second sector produces a continuum of differentiated goods as in Melitz (2003). Each firm supplies one of these goods and is a monopolist of its variety. As in Melitz (2003), consumers are endowed with one unit of labor and the preference over the differentiated good displays constant elasticity of substitution. The utility function of the representative consumer is σ



µ

σ −1  σ −1 U = q10− µ  ∫ q (ω ) σ dω  ,  ω∈Ω 

where ω denotes each variety and Ω is the set of varieties available to the consumer. The utility has a Cobb-Douglas form with share 1 − μ on the numeraire sector. The constant elasticity of substitution across each variety in the non-numeraire sector is denoted by σ (>1). The aggregate price index at home is thus 1



  1 −σ 1 −σ P =  ∫ p (ω ) dω  ,  ω∈Ω 

62 

M. YU

where p(ω) is the price of each variety. Accordingly, the demand for each variety is  p ( ω ) −σ q (ω ) = µ wL  1−σ  P 



   

and the revenue of each firm is 1 −σ

 p (ω )  r (ω ) = µ wL    P 



,

where wL is the total expenditure on the differentiated good at home. Firms face a fixed cost of entering the domestic market (Cd), which can be fully financed by using their internal liquidity. The cost function of serving the domestic market is cd ( q d ) = q d



w + wCd , x

where x is the productivity of the firm. As is common for monopolistic competition, each firm sets the domestic price, pd, with a constant markσ over the unit cost ­up σ −1 pd ( x ) =



σ w . σ −1 x

Therefore, firms generate profits at home:



πd ( x) =

rd ( x )

σ

1 −σ

− wCd =

µ  σ w  wL   σ  σ − 1 xP 

− wCd ,



where rd(x) is the firms’ domestic revenues. In order to survive in the domestic market, firms must have high enough productivity so that they can make positive profits. The cutoff productivity level, xd , is determined by the zero profits condition,

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

π d ( xd ) = 0



63



or 1

σ w  σ Cd σ −1 xd =   σ −1 P  µL 



(3.1)

3.2  The Decision to Export When a firm wants to export, the upfront fixed cost, wCE, can be financed internally using a fraction, d, of its domestic profits πd. Firms could also use a certain amount of zero-interest, zero-collateral funding from the parent firm if the firm is a FIE. Finally, they could raise funds externally from financial intermediaries. For simplicity, we assume that FIEs’ parent firms provide them a fraction of the fixed costs of the exports without interest, δiw∗CE, where w∗CE is the fixed cost of exports.2 The parameter δi, ∀i = F, NF, is the fraction and F, NF are the FIEs and non-FIEs, respectively. We assume that δF > δNF = 0, given the fact that non-FIEs do not have alternative source of funds. If the project is successful, firms pay back this specific loan. Otherwise, the loss will be covered by the parent firms. When firms borrow funds from financial intermediaries, they face different costs. First, the export project is subject to firm-specific risk and the probability of success, λ(x) ∈ [0, 1], is public knowledge. λ(x) is assumed to be an increasing function of a firm’s productivity x.3 Firms must offer investors a repayment, GE(x), such that investors can break even. The minimum level of the repayment firms can offer depends on the project’s probability of success. The external funds lent by investors require pledgeable collateral in case that the project fails. The collateral that a firm can provide is a fraction of the domestic fixed cost, twCd, where wCd is the domestic fixed cost and t is the fraction of the fixed cost that is pledgeable. The firm will never  For example, setting up a distribution network in foreign country requires foreign labor.  The assumption that the probability of success is a function of productivity is only for the sake of theoretical simplicity. In reality, the probability of success of a project might be related to other variables. 2 3

64 

M. YU

default on the repayment if the project is successful. If the project fails, investors could get back only the collateral. Financial contracting proceeds as follows. In the beginning of each period, every firm makes a take-it-or-leave-it offer to a potential investor. This contract specifies the amount the firm needs to borrow, the repayment GE in case the project is successful, and the collateral in case of failure. Revenues are then realized and the investor receives payments at the end of the period. The firms maximize their expected export profits, subject to four constraints:



  q ( x )τ w E (π E ( x ) ) = λ ( x )  pE ( x ) qE ( x ) − E − (1 − kE ) w∗CE − GE ( x )  x   − (1 − λ ( x ) ) twCd s.t.λ ( x ) GE ( x ) + − (1 − λ ( x ) ) twCd = kE w CE

(3.2)

∗



pE ( x ) qE ( x ) −

q E ( x )τ w x

− (1 − kE ) w∗CE ≥ GE ( x )

dπ d ( x ) + δ i w∗CE w ∗C E qE ( x ) =



= 1 − kE

µ w L pE ( x ) ∗ ∗

P ∗1−σ

σ

,



where τ is the iceberg transportation cost, 1 − kE is the firm-specific fraction of the fixed cost that could be financed internally or through loans from parent firms of FIEs. The first constraint is the investor break-even equation, stating that investors are perfectly competitive so they receive zero profits. The second constraint indicates that, if the project is successful, the firm must have enough net revenue to pay the repayment GE(x) to the investors. Notice that if expected exports profits, E(πE(x)), are greater than or equal to zero, this constraint is not binding. The third constraint specifies the portion of funds that need to be externally financed. Firms could use a fraction of domestic profits and loans from parent firms, if any, to cover part of fixed costs of the exports. Aside from these, the leftover, kEw∗CE, must be financed externally. The last constraint is the derived demand function.

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

65

This maximization problem thus provides the sources of credit constraints. First, in order to export and make a profit, firms must have the capability to generate large enough revenue so that they can offer a large enough repayment to investors. With a repayment that would allow investors to gain more than the break-even amount, investors would choose to extend financing to the firms. If the firm’s productivity is too low, it cannot generate large enough profits and hence cannot offer repayments that will allow the investors to break even, and thus they will not get any financing. Second, for the same level of repayment, investors are more likely to extend financing to firms that have larger probabilities of success because it is more likely that they will be repaid. In sum, the easier for the firm to get external financing, the more likely that the firm will enter the export market. The investor’s break-even equation determines the repayment that investors demand: GE ( x ) = twCd +

1 kE w∗CE − twCd λ ( x)

(

)

(3.3)

We substitute Eq. (3.3) into firm’s expected profits   q ( x )τ w E (π E ( x ) ) = λ ( x )  pE ( x ) qE ( x ) − E − w∗CE  − (1 − λ ( x ) ) kE w∗CE x   This equation indicates that as long as firms need to borrow from investors, that is, kE > 0, they must have higher expected profits to survive in the export market. The extra cost due to credit constraints depends on the amount they borrowed (kEw∗CE) and the probability of success of the project (λ(x)). The expected profits maximization problem is thus equivalent to maximizing   q E ( x )τ w − w∗CE   pE ( x ) qE ( x ) − x    1  − 1 1 − δ i ) w∗CE − dπ d ( x ) −  λ ( x )  (  

(



)

(3.4)

66 

M. YU

By defining the operating export profits as



π E ( x ) ≡ pE ( x ) qE ( x ) −

qE ( x )τ w x

− w ∗C E



(3.5)

the maximization problem is in turn equivalent to maximizing the operating exports profits since productivity, domestic profits, and the fraction of fixed costs covered by parent firms are predetermined from the perspective of the firms when they make export decisions.4 Consequently, the results of Melitz (2003) such as the optimal price pE(x), the quantity qE(x), the operating exports profits πE(x), and revenue rE(x), hold for the active exporters:5 pE ( x ) =



 σ τw  qE ( x ) =    σ −1 x 



σ τw σ −1 x

πE ( x) =

rE ( x )

σ

−σ

µ w∗ L∗ P ∗1−σ σ

− w ∗C E =

µ ∗ ∗  σ τw  wL  − w ∗C E ∗  σ  σ − 1 xP  1 −σ



 σ τw  rE ( x ) = µ w∗ L∗  ∗   σ − 1 xP 



Firms might or might not be restricted by credit constraints. If firms have high enough productivity so that they can generate high enough domestic profits, or if loans from parent firms are high enough, so that there is no need for externally raise funds, firms are not subject to credit constraints. Without credit constraints, the threshold productivity level of the exporter, xE , is determined by π E ( xE ) = 0 , or

4  An implicit assumption is that a firm’s productivity leads to exports. That is, good firms export, which is a common assumption for all Melitz-type (2003) models. 5  Notice that exporting firms sell in the foreign market and thus the foreign price level (P ∗), rather than the domestic price level, is present in the solutions.

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

σ −1

 σ 1  xE σ −1 =  ∗   σ −1 P 



σ 1 CE µ L∗ (τ w )1−σ

67

(3.6)

Firms might face credit constraints. Firms with productivity level x such that

(1 − δ i ) w∗CE − dπ d ( x ) > 0



need to borrow money externally in order to enter the exports market. Within these firms, only a subset could successfully enter the export market by making a positive expected export profits. Note that if d → 0, all firms that want to export must raise funds externally. The capability of firms to secure external loans determines whether or not the firm can export. In order to make positive expected profits, firms must have a productivity level x such that  1  πE ( x) −  − 1  1 − δ i ) w∗CE − dπ d ( x )  > 0  λ ( x )  (  



to survive the export market. Consequently, the cutoff productivity of exporting firms that are subject to credit constraints is determined by  1  π E ( xCE ) −  − 1  (1 − δ i ) w∗CE − dπ d ( xCE )  = 0  λ (x )  CE  



(3.7)

Compared to the cutoffs without credit constraints in Eq. (3.6), credit constraints virtually “increase” the fixed costs of exports for marginal firms, and hence the difficulty of entering the market increases. Solving this equation, the cutoff productivity for firms subject to credit constraints is6

6

 See the Appendix for details.

68 

M. YU

1

∗ σ −1  σ  σ σ −1  (1 − δ i + λ ( xCE ) δ i ) w CE  1 = + − 1  dwCd      λ (x )   σ −1  µ   λ ( xCE ) CE    1

xCE

1

1− σ 1−σ  ∗ ∗  τ w 1− σ  1  w  − 1  dwL     w L  ∗  +    P   P    λ ( xCE )  



(3.8)

Firms with productivity below xCE cannot export due to credit constraints, although some of them are sufficiently productive to do so profitably without credit constraints. 3.3  The Equilibrium To examine the equilibrium, we assume that the price level only depends on domestic firms’ prices and that foreign firms do not face any credit constraints by following Chaney (2005) and Muûls (2008). This assumption indicates that 1

  1 −σ 1 −σ P ≈  ∫ pd ( x ) LdF ( x )  ,  x≥ x   d 



where L is the home population and F(x) is the cumulated density function of the firm’s productivity at home. Given that the cutoff productivity level has an implicit form, we define the function h(∙) by σ h ( ⋅) : x σ −1 =  µ



∫x

x≥ x

σ −1

 dF ( x )  C ⇔ x = h ( C ) 

with h′ > 0. Assuming that the distribution of productivity is the same in the foreign country as at home, F(x) = F∗(x), the function h(∙) will be the same for the foreign country. It turns out that the cutoffs in Eqs. (3.1), (3.6), and (3.8) are solved as7

7

 See the Appendix.

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

xd = h ( Cd ) ,



69

(3.9)



1

 C σ −1 τ w xE =  ∗E  h C ∗d , ∗ C w  d w ∗C E (1 − δ i + λ ( xCE ) δ i ) w + (1 − λ ( xCE ) ) dCd , σ  w ∗  ∗ 1 −σ ∗ 1 −σ C d + (1 − λ ( xCE ) ) dCd h ( Cd )   C dh  w

( )

xCE =

τ

1 −σ

( )

(3.10)

(3.11)

where Eq. (3.11) indicates that xCE is an implicit decreasing function of δi, which is denoted as xCE (δ i ) . From solutions (3.10) and (3.11), it is easy to show that xCE = xE provided that λ = 1. Instead, if λ = 0, then xCE (δ i ) is the solution of equation (1 − δi)w∗CE = dπd(x), defined as x NEF , where NEF stands for “no external finance.” This is because x NEF is the cutoff productivity level with which firms’ domestic profits and parent firms’ loans are sufficient to finance their fixed costs to export. That is, firms with productivity above this level do not need any external finance. It is worth noting that when d → 0 as in Manova (2008), x NEF → ∞ . In this case, all firms need external loans in order to export. The capability of obtaining loans determines the capability of firms to export. When λ ∈ (0, 1), xCE (δ i ) is in between of xE and x NEF . From this fact, we can identify the sufficient condition such that there are some potential exporting firms prevented from exporting as stated in the following proposition. Proposition 1  If x is continuously distributed from [0, ∞], and if

1



 w∗C ∗d C ∗d σ −1  h ( Cd ) 1 − δ + ( )    i dwCd CE   h C ∗d  

( )

 τw > ∗  w 

then there exists a nonempty set of firms that are prevented from exporting even though they are profitable enough to do so if without credit constraints.

70 

M. YU

Proof. This proposition can be easily proved by substituting equation (3.11) and (1-) into inequality xE < xCE (δ i ) and making λ = 0. Proposition 1 suggests that, when the fraction of domestic profits used to finance the fixed costs to export is close to zero (d → 0), then there are always firms that are prevented from exporting. This is very realistic because firms might not be able to use a large portion of domestic profits to finance export projects. One example of this is that in the presence of principal–agent problems, stockholders may demand dividends at the end of each period instead of entrusting management with the utilization of retained earnings. This indicates that the capability of obtaining external finance is essential in determining whether or not firms could export. Figure 3.3 illustrates firms’ exporting behaviors. The vertical axis is the expected export profits and the horizontal axis is the productivity. The EM curve is the expected export profits curve if there is no financial friction, as in Melitz (2003). Firms with a productivity level greater than xE can profitably export if there is no credit constraint. However, when there are financial frictions, some of these firms can no longer export because they cannot obtain external loans to finance the fixed costs to export. Curves EF and ENF are the expected exports profits curves for FIEs and non-FIEs, respectively. Thus, potential FIE exporters and potential non-FIE exporters that are prevented from exporting are firms with productivity levels in the range of  xE ,xCE (δ F )  and  xE ,xCE (δ NF )  , respectively. These firms cannot offer large enough repayments to investors and consequently cannot obtain external loans. Thus, they are prevented from exporting. Instead, FIEs and non-FIEs with productivity levels greater than xCE (δ F ) and xCE (δ NF ) , respectively, are able to obtain external loans and enter the export market. Firms with productivity levels greater than x NEF have sufficient domestic profits to finance the fixed costs to export internally and need not borrow from investors. However, as noted above, if the fraction of domestic profits that can be used to finance an export project is sufficiently small, then all exporting firms are subject to credit constraints. Firms’ capabilities of obtaining external loans then determine their capabilities of exporting. Therefore, our model predicts that firms face different levels of credit constraints. The credit constraints are related to firms’ previous domestic profits. They affect the amount of internal liquidity that can be used to finance the entry costs to export as suggested by Chaney (2005). Aside from this, there still exist other important channels that affect firms’ credit constraints.

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

EM

71

EF ENF

0

External finance and export

No external finance and no export

xE

xCE (

F

)

xCE (

NF

)

X

xNEF

Fig. 3.3  Firm’s productivity, credit constraints, and exports

First, firms with higher productivity are more capable of obtaining external loans. Higher productivity firms would generate larger export revenues and thus they would be able to offer a larger repayment for the same amount of principal. Thus, they are less credit constrained. More importantly, the higher productivity introduces the higher probability of success of the projects. As a result, investors are more likely to break even when they extend loans to more productive firms, given the level of the principal and the repayment. Investors thus would be more inclined to extend loans to more productive firms. Less productive firms are then prevented from entering the export market. Again, the capability of obtaining external loans is the key to exporting in the presence of financial frictions.8 Second, our model clearly suggests that the firm’s type plays a role in its access to alternative funds, which in turn affects the firm’s export decision. 8  Of course, in reality, a firm’s capability of raising funds externally may be affected by other factors such as automatic customer insolvency, bad debts, overdue accounts, commercial risks, and even political risks. However, the key lesson is that firms that are capable of obtaining external loans will be able to export.

72 

M. YU

Since xCE (δ i ) is a decreasing function of δi, firms that have atlternative channels (e.g., FIEs get loans from their parent firms) to obtain low-cost loans would depend less on external financing from financial intermediaries and thus have lower requirements of productivity levels to enter the export market. Given that δF  >  δNF as assumed, FIEs might be able to export more compared to non-FIEs even when they both have identical loans from financial intermediates, while otherwise they would be prevented from exporting. In sum, we make the following proposition. Proposition 2  All else equal, it is easier for firms to enter the export market if (1) they have higher probability of success of the project and consequently easier access to external financing from financial intermediaries; or (2) they have alternative sources, other than from financial intermediaries, to obtain funds.

4   Econometrics, Data, and Measures 4.1  Empirical Specification Our theoretical model introduced above clearly predicts that it is easier for firms with higher probability of success to get loans from financial intermediaries and it is thus easier for them to enter the export market. Put another way, the model predicts that firms with larger interest expenditures, an index of their capability of obtaining loans, would export more. In addition, firms with higher productivity also export more. We therefore consider a specification as follows: ln EX it = β 0 + β1 ln IEit + β 2 ln TFPit −1 + β 3 FIEi + β 4 FIEi ln IEit + β 5 ln Dprofit −1 + θ Xit + ςi + ϑt + it ,



(3.12)

where i is the firm and t denotes year, lnEX, lnIE, and lnTFP are the firm’s log exports, log interest expenditures, and log total factor productivity, respectively. FIEi is a dummy variable, equal to a unit if firm i is a foreign-­ invested-­enterprise and zero otherwise. Interest expenditures, TFP and FIE status all are predicted to contribute to exports. Thus, coefficients β1, β2, and β3, all are expected to be positive. The coefficient of the interaction

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

73

term of FIE and lnIE captures the differential impact of interest expenditures on exports for different types of firms. The model suggests that FIEs are less dependent on external borrowing. Thus, β4 is expected to be negative. Firms might also be able to finance the fixed costs to export using domestic profits. In order to capture this, we also include in the firms’ log domestic profits (ln Dprof ). Since domestic profits as well as productivity are both predetermined before the firms make export decisions, we include values for the previous period for ln  TFP and ln  Dprof in the empirical specification (3.12).9 In contrast, current period interest expenditures affect firms’ exports, Thus, we use its current realization in the empirical model. In addition, X denotes other control variables. In particular, traditional wisdom says that SOEs in China play a significant role in its economy, although this is not formally specified in the theoretical model. To shed light on this point, we also include a dummy for SOEs and its interaction term with interest expenditures, SOEi ×  ln IEit, in the model. Finally, all other unspecified factors are absorbed in the error term, which can be decomposed into the three following components: (1) firm-­ specific fixed effects ςi to control for time-invariant factors; (2) year-­specific fixed effects ϑt to control for firm-invariant factors; and (3) an idiosyncratic effect ϵit with normal distribution ϵit~N(0, σi2) to control for other unspecified factors. The impact of credit constraints on exports is the focus of this chapter; thus, the coefficient of β1 is our main interest. 4.2  Data The sample used in this chapter comes from a rich firm-level panel data set that covers more than 160,000 manufacturing firms per year for the years 2000–2007. The number of firms doubled from 162,885  in 2000 to 336,768  in 2007. The data are collected and maintained by China’s National Bureau of Statistics in an annual survey of manufacturing enterprises. It covers two types of manufacturing firms: (1) all SOEs; (2) non-­ SOEs whose annual sales are more than five million yuan.10 The data set 9  This setup also enjoys an extra advantage that we do not need to worry about the reverse causality of exports and productivity in the estimations. 10  Indeed, aggregated data on the industrial sector in the annual China’s Statistical Yearbook by the Natural Bureau of Statistics (NBS) are compiled from this data set.

74 

M. YU

includes more than 100 financial variables listed in the main accounting sheets of all these firms.11 Although this data set contains rich information, a few samples in the data set are noisy and misleading due, in large part, to the misreporting by some firms.12 For example, information on some family-based firms, which usually did not set up formal accounting systems, is based on a unit of one yuan, whereas the official requirement is a unit of 1000 yuan. We clean the sample and rule out outliers by using the following criteria: first, observations whose key financial variables (such as total assets, net value of fixed assets, sales, gross value of industrial output) cannot be missing; otherwise, they will be dropped. Second, the number of employees hired for a firm must not be less than 10 people.13 Following Cai and Liu (2009), and guided by the General Accepted Accounting Principles (GAAP), we delete observations if any of the following rules are violated: (1) the total assets must be higher than the liquid assets; (2) the total assets must be larger than the total fixed assets; (3) the total assets must be larger than the net value of the fixed assets; (4) a firm’s identification number cannot be missing and must be unique; and (5) the established time must be valid. In particular, observations in which the opening year is after 2007 or the opening month is later than December or earlier than January are dropped as well. Since FIEs as well as SOEs play a significant role in our model, it is an advantage to take a careful look at these firms. We first construct a dummy for foreign-invested firms (FIEs) to distinguish domestic from nondomestic firms. Here, we consider a broad classification of FIEs. In particular, the FIEs include both foreign firms and Hong Kong/Macao/Taiwan (H/M/T)-owned firms.14 In a robustness check, we consider a narrower classification of FIEs by excluding the H/M/T-owned firms. Turning to 11  Following Levinsohn and Petrin (2003), plants were treated as firms. In this chapter, we do not capture scope economics due to their multi-plant nature. This remains a topic for future research. 12  Holz (2004) offers careful scrutiny on possible measurement problems in Chinese data, especially on the aggregated level. 13  Levinsohn and Petrin (2003) suggest covering all Chilean plants with at least 10 workers. Here, we follow their criterion. 14  Specifically, FIEs include the following firms: foreign-invested joint-stock corporations (code: 310), foreign-invested joint venture enterprises (320), fully foreign-invested enterprise (330), foreign-invested limited corporations (340), H/M/T/ joint-stock corporations (code: 210), H/M/T joint venture enterprises (220), fully H/M/T-invested enterprises (230), and H/M/T-invested limited corporations (340).

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

75

the dummy for SOEs,15 to avoid a possible “adverse selection” problem faced by small SOEs in borrowing external loans, we use “above-scale” SOEs as a default sample by dropping SOE observations whose operation scales are smaller than the “above-scale” threshold. In particular, the observations are dropped if any of following indicators is lower than 5 million yuan: (1) the value of the firm’s sales; or (2) the value of the total assets; or (3) the value of the fixed assets. In robustness checks, we include all SOEs. After this very rigorous filter, we obtain a sample of 1,294,596 observations from the original sample of 1,898,958, which accounts for 69.2% of the original data set.16 All nominal terms are originally measured in current Chinese yuan. We therefore use the producer price index (PPI) by sector, which is measured at the two-digit Chinese industrial classification and obtained from the National Bureau Statistics, as the GDP deflator by choosing year 2000 as the base year. Panel A of Table 3.1 provides some basic statistical information about the Chinese firm data. After the filter, the number of “above-scale” FIEs is 273,160, which accounts for 21.1% of the full samples. In particular, the number of FIEs excluding H/M/T-owned firms is 132,826, 10.2% of all “above-scale” firms. The remaining 79% are domestic firms, of which 9.1% are SOEs and 5.8% are “above-scale” SOEs. In the sample, some firms export while some do not. Only 360,106 firms have positive exports, which accounts for 27.7% of all “above-scale” firms. However, 880,026 firms, a 67.3% of all firms, have positive interest expenditures. We plot the logarithm of exports against the logarithm of interest expenditures at the SIC two-digit industrial level as shown in Fig. 3.2. We can clearly observe a positive correlation between exports and interest expenditures across sectors. It seems difficult to find obvious outliers in the scatter plots. This suggests that the positive correlation between the two is unlikely due to outliers. Finally, a firm’s domestic profits are measured as the difference between the firm’s overall profits and its export value. A firm’s productivity is measured by using both total factor productivity (TFP) and labor productivity. In particular, the measure of TFP, by definition, requires that firms’  By definition, SOEs include firms such as domestic SOEs (code: 110); state-owned joint venture enterprises (141); state-owned and collective joint venture enterprises (143); and state-owned limited corporations. 16  Including small- and medium-sized SOEs whose scales are lower than the threshold would increase our sample to 1,401,569, which accounts for around 73.8% of the original size of the data set. 15

76 

M. YU

Table 3.1  Summary statistics (2000–2007) Panel A: Basic statistics for key variables Variables

Obs.

Mean

Std. dev.

Min

Max

Year Firm’s exports Firm’s interest expenditure Log exports (ln EXit) Log interest expenditure (ln IEit) Log domestic profits in previous year Log TFP (Olley–Pakes) in previous year Log labor productivity in previous year Log ratio of exports to sales Log ratio of interest expenditure to sales Log ratio of domestic profits to sales SOEs dummy (all scale) SOEs dummy (above scale: SOEi) FIEs dummy (inclusive H/M/T: FIEi) FIEs dummy (exclusive H/M/T) FIEi ×  ln IEit SOEi ×  ln IEit WTO dummy Money supply (M1) Weight of firm’s scale (lijt − 1/∑i ∈ jlijt − 1) Log weighted money supply

1,294,596 1,294,596 1,294,596 360,106 872,208

2004.348 17,303.12 1198.837 9.516 5.398

2.244 334,443.5 14,325.02 1.691 1.896

2000 0 0 0 0

2007 1.81e+08 5,363,291 19.014 15.495

525,533 6.533

1.905

0

17.557

840,826 −0.031

0.322

−9.746

2.444

846,495 5.250

0.999

−1.421

13.417

360,106 −965 872,208 −4.888

1.412 1.512

−15.194 −15.704

3.890 2.486

812,966 −3.564

1.446

−13.147

3.338

1401,569 0.091 1,294,596 0.058

0.288 0.234

0 0

1 1

1,294,596 0.211

0.408

0

1

1,294,596 0.102

0.303

0

1

872,208 872,208 1,294,596 1,294,596 846,495

2.355 1.732 0.350 3.17e+09 0.001

0 0 0 5.31e+09 3.07e-07

15.495 14.600 1 1.53e+10 0.040

1.271

7.398

19.773

1.038 0.444 0.857 1.51e+10 0.000

846,495 11.484

Panel B: Simple correlations for some key variables Log exports Log interest expenditure Log lag TFP Weighed M1 Log exports Log interest expenditure Log lag TFP Weighed M1

1.000 0.328 0.197 0.390

1.000 0.083 0.487

1.000 0.094

1.000

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

77

output be measured by real but not nominal terms. Following Amiti and Konings (2007), we therefore use China’s PPI at the SIC two-digit level as the deflator to convert data from nominal terms to real terms. To be consistent with other data, the base year of the deflator is also chosen as 2000. 4.3  Measures of TFP As one of the most important control variables in our regressions, it is essential for us to measure TFP precisely. However, many existing works on measuring TFP show that it is imprecise and biased (Amiti and Konings 2007). TFP is usually measured as the Solow residual, defined as the difference between the observed output and its fitted value calculated via OLS. However, this method suffers from two biases: a simultaneity bias and a selection bias. The first bias comes from the fact that a profit-­ maximizing firm would readjust its input decision as a response to productivity shocks that are observed by firms but not by econometricians. Second, all firms covered in the samples are those that have relatively high productivity and that survived during the period of investigation. Those firms that had low productivity, shut down, and left the market were not observed nor included in the sample. Put another way, the sample covered in the regressions is not randomly selected. Hence, all related estimates would suffer from a selection bias. To overcome these two empirical challenges, we use the augmented Olley–Pakes (1996) approach to estimate and calculate the firms’ TFP. Given China’s WTO accession in 2001, which was a positive demand shock for China’s exports, we also include a WTO dummy in the Olley– Pakes estimations to capture the effect of the WTO accession. We report the detailed estimation procedure in Appendix 2. Table 3.7 also presents the estimated survival probability of the firms in the next year by industry at the SIC two-digit level. Its mean of 0.993 suggests that the problem of firms’ exiting is not so severe during this period. The rest of Table 3.7 reports the difference in the estimated coefficients for labor, material, and capital by using a regular OLS approach and the OP methodology. We cover 36 manufacturing industries coded from 6 to 46, according to China’s new industrial classifications (GB/

78 

M. YU

T4754), which were adopted in 2002.17 Compared to the original firm-­ level data set, four industries are not covered after this filter process.18

5   Empirical Results 5.1  Main Estimation Results Table 3.2 presents the estimation results for Eq. (3.8). To consider the effect of a firm’s interest expenditures on its exports, we first run a simple OLS regression of firms’ exports on firms’ log interest expenditures, log TFP in the previous year, and FIE dummy as a benchmark. The estimated  in Eq. (3.8) is 0.290, which is significant at the convencoefficient β 1 tional statistical level. This suggests that a 1% increase in a firm’s interest expenditures is associated with a 0.290% increase in its exports. The benchmark finding is consistent with the simple cross-sectional plot in Fig. 3.2. The positive effect of TFP is also consistent with the previous findings like those of Bernard and Jensen (1999) that firms with higher productivity export more. In column (1), we also observe that FIEs are shown to export more than non-FIEs. However, the finding that FIEs have more exports could be due to their quick learning, better technology adoption, or higher-­ quality inputs (Amiti and Konings 2007), or even simply because they are established to export per se. However, FIEs might export more because of their easier access to low-cost loans (e.g., financial support from their parent firms). If this is correct, FIEs should be less sensitive to external finance. Similarly, it is also interesting to ask whether SOEs export more compared to non-SOEs and check whether or not they are sensitive to external finance.We then consider a specification including four more variables: the SOEs dummy, the interaction terms between FIEs and the log interest expenditure (FIEi ×  ln IEit) and between SOEs and log interest expenditures (SOEi ×  ln IEit), and the firm’s log domestic profit in the previous period (ln Dprofit − 1). The OLS estimation results show that the interaction term between FIEs and interest expenditures is negative, which 17  Firm data before 2002 were clustered into industrial data by adopting the old industrial classification. We concord such data so that they are consistent with data after 2002. 18  The five industries dropped include extraction of petroleum and natural gas (7), mining of other ores (11), processing of food from agricultural products (12), and recycling and disposal of waste (43).

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

79

Table 3.2  Benchmark estimates, log of exports as the dependent variable Regressand

OLS

Log firm’s exports (ln EXit)

(1)

Log interest expenditure (ln IEit)

0.290** (137.19) 0.960** (50.69) 0.775** (100.29)

FE (2)

0.170** (24.33) Log TFP ( ln TFPitOP 0.470** −1 ) (10.81) FIEs dummy (FIEi) 0.596** (7.81) FIEi ×  ln IEit −0.032** (−2.63) Log domestic profit (ln Dprofit − 1) 0.145** (22.88) SOEs dummy (SOEi) −1.650** (−8.59) SOEi ×  ln IEit 0.174** (7.23) Firm fixed effects No No Year fixed effects No No Observations 178,136 33,492 Root MSE 1.62 1.85 (Pseudo) R-squared 0.16 0.11

PPML+FE

(3)

(4)

(5)

0.105** (29.64) 0.215** (15.99) –

0.112** (5.87) 0.204** (3.05) –

0.351** (14,406.35) 0.635** (4422.97) –

−0.007 (−0.46) 0.008 (0.70) –

0.087** (12,237.53) 0.386** (16,775.06) –

−0.018 (−1.08) Yes Yes Yes Yes 178,136 33,492 0.000 0.000 0.09 0.03

−0.023** (−2326.87) No Yes 33,493 0.000 0.50

Notes: Robust t-values corrected for clustering at the firm level in parentheses. *(**) indicates significance at the 10(5) percent level. Column (5) is the Poisson pseudo-maximum likelihood (PPML) fixed-effects estimation in which the regressand is EXit > 0

suggests that exports of FIEs, compared to those of non-FIEs, are less sensitive to external finance. The variable of FIEs itself is positively correlated with exports as expected. In addition, the significantly positive sign of a firm’s domestic profits is consistent with the prediction suggested by the theoretical model: firms with high predetermined domestic profits do not suffer from credit constraints as much and hence they export more. In addition, the significant negative sign of SOEs suggests that SOEs, compared to non-SOEs, do not export more. There exist several possible reasons for this finding. First, China’s SOEs may suffer from an inefficient incentive mechanism (Lin 2003). Second, the SOEs enjoyed less preferential treatment in access to external finance after China’s WTO accession in 2001 (Bajona and Chu 2004). Accordingly, SOEs have lower productivity and export less.

80 

M. YU

Aside from the variables introduced above, other variables may affect a firm’s exports as well, though they are not specified in the theoretical model for simplicity. Such variables include time-varying but firm-­ insensitive factors like the Chinese yuan (RMB) appreciation.19 Similarly, some other unspecified time-invariant but firm-varying factors like a firm’s location may also have impact on the firm’s exports. To control for such factors, we include both year-specific and firm-specific fixed effects in the estimations. Columns (3) and (4) of Table  3.2 report the fixed-effect estimation results. Since the dummies of FIEs and SOEs are time invariant, they are automatically dropped from the estimations. In all specifications, the firm’s TFP is shown to be significantly positively associated with its exports.  are all positive and significant at the conventional The key coefficients β 1 statistical levels. In terms of economic magnitude, the coefficients of the firms’ interest expenditures in columns (3) are much smaller than the coefficients in column (1). However, after including more control vari in column (4) is close to its counterpart in column (2). Aside ables, β 1 from this, most other control variables in columns (1)–(4) have the expected signs, though they are insignificant. We suspect that the insignificance is due to the lack of a control for the endogeneity of interest expenditures, which will be handled shortly. 5.2  The Zero Trade Problem As mentioned earlier, more than 73% of firms had zero exports in our sample. Recent works like those by Santos Silva and Tenreyro (2006), Helpman et al. (2007), and Yu (2009) have argued that OLS estimates can lead to serious biases due to zero trade volume. The log-linearization of firms’ exports may cause some bias since the entire portion of the data with zero trade is dropped. In addition, the zero trade problem in our study could reflect the rounding down of firms’ exports, which creates an endogenous censoring problem. To address the empirical challenges raised by the zero trade problem, Santos Silva and Tenreyro (2006) proposed a truncated Poisson pseudo-­ maximum likelihood (PPML) estimation to address the zero trade problem. We therefore perform the PPML fixed-effects estimation by adopting firms’ exports EXi directly as the regressand. As shown in column (5) of Table  3.2, the coefficients of interest expenditures and TFP are still  Chinese yuan was revalued against the US dollar by around 20% during 2005–2007.

19

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

81

significantly positive in the PPML fixed-effects estimates. Compared to the OLS fixed-effects estimates in column (4), the coefficient of the predetermined domestic profits turns out to be highly significant with the expected sign. The interaction term between interest expenditures and SOEs turns to be significantly negative. However, the interaction term between FIEs and interest expenditures has a strikingly positive sign. We again suspect that this is due to the lack of consideration of the endogeneity of interest expenditures. We now turn to this issue. 5.3  Endogeneity Issues A firm’s interest expenditures are not exogenously given, but affected by its exports. With more exports, firms need more upfront fixed costs (e.g., firms need to have a greater distribution network abroad when its exports increase), which in turn require firms to obtain more debt to do business. One needs to control for the endogeneity of interest expenditures in order to obtain accurate estimated effects of a firm’s interest expenditures on exports. Otherwise, the related estimates would be suspect. Instrumental variable (IV) estimation is a powerful econometric method that can address this problem.20 An economic indicator, a firm’s weighted monetary supply, is constructed to serve as the instrument for interest expenditures. This indicator is defined as (lijt  −  1/∑i  ∈  jlijt  −  1)M1t  −  1, where lijt  −  1 is the number of employees of firm i in industry j, ∑i ∈ jlijt − 1 is the total number of employees in industry j, and M1t − 1 is China’s base monetary supply (M1) in the previous year t − 1. The last row in Table 3.1 lists the basic statistical information for this indicator.21 The intuition is straightforward. Monetary expansion increases the supply of investments, making external finance less costly. As a result, firms would finance more externally and interest expenditures would increase. Moreover, marginal firms, i.e., previous credit-constrained firms, are now able to borrow from financial intermediaries and they might be able to enter the export market. Since the money supply is exogenous to exports but it affects interest expenditures, it makes good economic sense to use it as an IV.  We note that the previous-period money supply is used to 20  The IV approach is a good way to control for endogeneity issues. Wooldridge (2002, Chapter 5) provided a careful scrutiny of this topic. 21  For robustness, we construct an alternative indicator using a firm scale, relative to the total economy, as a weight of the monetary supply. The results are similar.

82 

M. YU

construct the IV based on the fact that the monetary policy has a time lag to take effect.22 However, the effects of monetary expansion on external finance are not homogeneous across firms. Larger firms might be more competitive in the competition for loans and thus they may enjoy larger benefits from easier access to external finance. We therefore consider a firm’s relative size when constructing the IV. One might worry that a firm’s sales, a usual index for measuring a firm’s scale, is highly correlated with the dependent variable (i.e., the firm’s exports) and hence not exogenous. We address this concern by using the number of employees as a proxy for a firm’s size. More importantly, we use its industrial weight, lijt − 1/∑i ∈ jlijt − 1, rather than the national economy’s weight, to control for firm size.23 In this way, it is less likely that the level of exports is highly correlated with the weight. Indeed, their simple correlation is very small in our sample (corr = 0.04). Also, as observed in Panel B of Table 3.1, the log of the weighted money supply has a small simple correlation with the log of exports (corr = 0.39), which is lower than the simple correlation between the log of the weighted money supply and the log of the interest expenditure (corr = 0.49). We perform several tests to verify the quality of our instrument. First, we checked whether the excluded instrument (i.e., the firm’s weighted monetary supply) is “relevant.” That is, whether this instrument is correlated with the endogenous regressor (i.e., interest expenditures). In our econometric model, the error term is assumed to be heteroskedastic: ϵit~N(0, σi2). Therefore, the usual Anderson (1984) canonical correlation likelihood-ratio test is not valid since it only works under the i.i.d. assumption. Instead, we use the Kleibergen–Paap (2006) Wald statistic to check whether or not the excluded instrument is correlated with the endogenous regressors (i.e., interest expenditures). The null hypothesis that the model is under-identified is rejected at the 1% significance level. 22  One may worry that monetary expansion might not affect the external financing of small-sized firms in China, because they might not have any access to formal financial intermediaries whatsoever. However, this is not a severe problem for two reasons. First, monetary expansion stimulates investments from informal financial intermediaries as well. Small-sized firms would have easier external financing from these intermediaries following a monetary expansion. Second, all observation used in the default sample are “above-scale” firms, which usually have access to formal financial intermediaries. We will include small- and mediumsized firms later for robustness check. 23  The reason for using industrial weight and not national weight (lit − 1/∑ilit − 1) is that, in China, the competition to borrow from banks is usually stronger in the same industry. However, using a national weight to construct the IV also delivers very similar findings, which are not reported here to save space but available upon request.

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

83

Second, we also tested whether or not the weighted monetary supply is weakly correlated to interest expenditures. If so, then the estimates will perform poorly in the IV estimate. The Kleibergen–Paap (2006) F-statistics provide strong evidence to reject the null hypothesis that the first stage is weakly identified at a highly significant level.24 Third, the Anderson and Rubin (1949) χ2 statistics reject the null hypothesis that the coefficient of the endogenous regressor is equal to zero. In short, these statistical tests give sufficient evidence that the instrument performs well, and therefore, the specification is well justified. The IV fixed-effects estimates in columns (1) and (2) of Table 3.3 show that, after controlling for endogeneity, interest expenditures still have a positive effect on a firm’s exports. In particular, the corresponding coefficients of the IV (i.e., the logarithm of the lagged weighted money supply) in the first stage are also statistically significant and positive as expected. Their first-stage excluded F-statistics are definitely high enough to pass the  are higher than their counterparts in Table 3.2 F-tests. The coefficients β 1 without controlling for the endogeneity. The elasticity of a firm’s interest expenditures on its exports is around a unit, which clearly suggests that high interest expenditures lead to more exports. Equally important, in column (2), the interaction term between FIEs and interest expenditures is β4 = −0.256 , which is significant at the conventional level, indicating that FIEs face less severe credit constraints. On average, a 1% increase of non-FIEs’ interest expenditure leads to a 1.068% increase in its exports, ceteris paribus. By way of comparison, a 1% increase in an FIE’s interest expenditures leads to a 0.812% increase in its exports since 1.068–0.256 = 0.812. Similarly, the interaction term between SOEs and interest expenditures is significantly negative, which is consistent with the PPML fixed-effect finding in Table 3.2. The economic rationale, arguably, is that, compared to non-SOEs, SOEs are less sensitive to external finance due in part to their relatively easier access to subsidies or low-cost loans from the government. After controlling for the endogeneity and fixed effects, the coefficient of domestic profits in column (2) has a very small magnitude with an unexpected but insignificant sign. To some extent, this serves as side evidence that domestic profits is not an important channel through which firms finance their fixed costs for exports. Instead, the main channel to alleviate the credit constraints is borrowing externally.

24  Note that the Cragg and Donald (1993) F-statistic is no longer valid since it only works under the i.i.d. assumption.

84 

M. YU

Table 3.3  IV fixed-effects estimates of interest expenditures on Firms5 exports Regressand:

“Above-scale” estimates

Log firm’s exports (lnEXit)

Broad FIE

Broad FIE

Narrow FIEs

(1)

(2)

(3)

(4)

1.068** (4.67) 0.156** (2.01) – −0.256** (−4.18) −0.006

1.123** (6.37) 0.165** (2.08) – −0.157** (−5.26) −0.008

0.753** (3.76) 0.135 (1.47) – −178** (−3.16) 0.012

(−0.45) – −0.107** (−3.93) 0.445**

(−0.58) – −0.098** (−4.36) 0.430**

(0.81) – −0.098** (−3.32) 0.475**

(223.64) (64.36) 50,014.16a 4142.19a

(62.15) 7636.93a

(60.16) 3619.71a

23,282.69a 2914.84a 50,015.29a 4143.18a

3014.38a 3862.19a

2720.54a 3620.77a

25,804.69a Yes Yes 178,130 0.24

1067.70a Yes Yes 33,488 0.12

657.38a Yes Yes 33,493 0.14

602.62a Yes Yes 27,291 0.15

0.11

0.05

0.07

0.06

Log interest expenditure (ln IEit) 0.928** (28.72) Log TFP ( ln TFPitOP 0.193** −1 ) (13.78) FIEs dummy (FIEi) – FIEi ×  ln IEit Log domestic profit (ln Dprofit − 1) SOEs dummy (SOEi) SOEi ×  ln IEit Log weighted ml (IV in the first-stage) Excluded F statistic in the first-stage Kleibergen-Paap rk LM statistic Kleibergen-Paap Wald rk F statistic Anderson-Rubin χ2 statistic Firm fixed effects Year fixed effects Observations Partical R-squared in the 1st stage R-squared in the 2nd stage

0.717**

“All scale” estimate

Notes: Robust t-values corrected for clustering at the firm level in parentheses. *(**) indicates significance at the 10(5) percent level P-value is less than 0.01

a

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

85

5.4  Additional Robustness Checks  stimates with Narrower Definition of FIEs E As discussed above, being an FIE is important to a firm’s exports in the sense that being such can introduce alternative financing channels such as funding from parent firms. To shed light on this point, we use a broad classification of FIEs in the estimations above. That is, firms with funding from H/M/T are considered as foreign-invested firms. Given that Hong Kong is one of the most important sources of China’s foreign direct investment, our estimations so far thus provide an upper bound for the role of FIEs in exports. Yet, it is still a plus to understand the effect of FIEs on exports if we exclude the impact of funding from H/M/T. We therefore redefine a narrow classification of FIEs by excluding H/M/T-owned firms and present the corresponding estimation results in column (3) of Table 3.3. Compared with the findings in column (2) using the broad definition of FIEs, the effect of interest expenditures is slightly    =  1.123. In contrast, the effect of the interaction term, larger: β 1 FIEi ×  ln IEit, is much lower. These findings together suggest that a 1% increase in FIEs’ interest expenditures leads to a 0.966% increase in its exports since 1.123–0.157  =  0.966. This makes good economic sense. With fewer alternative channels through which to raise financing, firms, on average, have to rely much more on external borrowing to finance their fixed costs for exports.  stimates Using “All-Scale” Data E As mentioned above, the data used for the estimations above are all “above-scale” because they include firms with annual sales higher than 5 million yuan. One advantage of using all “above-scale” data is to avoid a possible “adverse selection” problem. In China, it is difficult for small- and medium-sized firms to obtain external funds since financial intermediaries like banks are usually unwilling to lend to them due to the high risk of default. However, large firms usually do not have this problem. The trade-off of using “above-scale” data is that the role of SOEs is overestimated since some small- and medium-sized SOEs are dropped. Equally importantly, the problem of the exiting of firms cannot be well controlled: it does not necessarily mean that a large firm that was included in the early years of the study period but is not included in the later years and exited from the market. It could be just that its scale is smaller in the later period than the above-scale threshold and hence not counted in the truncated data set.

86 

M. YU

As observed in Table 3.1, once we consider the small- and medium-­sized SOEs, the number of SOEs increases to 127,598, which accounts for 9.1% of the sample. When including small- and medium-sized SOEs, we first recalculated firms’ average Olley–Pakes (1996)-type TFP, and then performed all related estimations. The results are reported in column (4) in Table 3.3. All coefficients have identical signs as the previous findings. The  , is shown to be effect of interest expenditures for non-FIEs on exports, β 1 smaller than the estimates using the “above-scale” data. However, the effect  + β , are insensitive to the of interest expenditures for FIEs on exports, β 1 4 estimation results using the “above-scale” data. In any case, our previous findings of the positive impact of interest expenditures on exports is robust.  stimates Without Productivity E Regardless of different empirical specifications, all the estimates from Tables 3.2 and 3.3 show that higher productivity leads to more exports. Our theoretical model also clearly shows that a firm’s pre-realized productivity is the main driving force for exporting. With higher productivity, the probability of success of a firm’s project is higher, which in turn helps the firm to secure outside loans and to export more. In short, credit constraints are one of the channels by which firm’s productivity affects its exports. It is interesting to consider the importance of the channel of credit constraints in relation to other channels. After controlling for fixed effects and reverse causality of interest expenditures, column (2) of Table  3.3 shows that the economic magnitude of interest expenditures is much   = 1.068 whereas β  = 0.156 ). higher than TFP (e.g., in column (5), β 1 2 This raises a question: if ignoring the impacts of other channels of productivity on exports, do credit constraints pick up the residual effects of productivity on exports? If so, then our estimations would be suspect since they are sensitive to different specifications. By dropping the variable of productivity, Columns (1)–(3) of Table 3.4 provide more evidence on this point. We report the estimation results using different econometric methods: OLS, fixed effects, and fixed-effect IV.  All of the results obtained from these methods are not substantially quantitatively different from their counterparts in Tables 3.2 and 3.3. In particular, the IV fixed effect estimates in column (3) of Table 3.4 show that the magnitude of interest   = 1.063, is almost idenexpenditures by excluding the variable of TFP, β 1 tical to its counterpart in column (2) of Table 3.3 with the presence of   = 1.068. This thus provides strong evidence that interest expenTFP: β 1 ditures do not pick up the residual effect of productivity on a firm’s exports.

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

87

Table 3.4  More robustness checks Regressand

Without productivity

Log firm’s exports (ln EXit)

OLS

FE

(1)

(2)

0.166** (23.76)

0.113** 1.063** (5.91) (4.65)

0.108** 1.206** (5.64) (7.11)

–

–

0.179** 0.057 (4.75) (1.30) – –

−0.009 (−0.53) 0.017 (1.41) –

−0.257** (−4.17) 0.001 (0.08) –

−0.007 (−0.44) 0.001 (0.07) –

−0.295** (−6.25) −0.006 (−0.46) –

−0.018 (−1.10) Yes Yes 33,604 0.000 0.15

−0.107** (−3.92) Yes Yes 33,600 0.000 0.05

−0.016 (−0.96) Yes Yes 33,604 0.000 0.15

−0.120** (−5.01) Yes Yes 33,600 0.000 0.05

Log interest expenditure (ln IEit)

Labor

Productivity

FE+IV

FE

FE+IV

(3)

(4)

(5)

Log TFP ( ln TFPitOP −1 ) Log labor productivity (ln LPit − 1) FIEs dummy (FIEi)

0.605** (7.94) FIEi ×  ln IEit −0.030** (−0.2.55) Log domestic profit (ln Dprofit − 1) 0.167** (27.90) SOEs dummy (SOEi) −1.703** (−8.88) SOEi ×  ln IEit 0.181** (7.53) Firm fixed effects No Year fixed effects No Observations 33,604 Prob.>F 0.000 R-squared 0.11

Notes: Robust t-values corrected for clustering at the firm level in parentheses. *(**) indicates significance at the 10(5) percent level

 stimation with Labor Productivity E In the previous section, we use total factor productivity (TFP) to measure productivity since it is close to reality. However, since our theoretical model essentially is a one-input (i.e., labor) Krugman (1979) model, it is useful to use labor productivity to measure productivity as well. Columns (4) and (5) of Table 3.4 report the estimation results using labor productivity to measure productivity. As shown in column (4), with the control of both firm-specific and year-specific fixed effects, the coefficient of labor productivity, 0.174, is smaller than its counterpart in column (4) of Table 3.2: 0.204. The result makes good economic sense given that labor is only one of the inputs. After controlling for the endogeneity of a firm’s interest expenditures, the

88 

M. YU

IV fixed-effects estimates in column (5) also present the anticipated positive sign. However, it is insignificant, which in turn suggests that the TFP may be a more ideal measure of a firm’s productivity.  dditional Estimates with Ratio Specifications A Thus far, all estimates suggest that high interest expenditures lead to more exports. Yet, the key variables in the estimates above are measured by levels. It is useful to consider estimations with ratio specifications to check if our previous findings hold. We therefore take the log ratio of a firm’s exports to its sales as the regressand. By the same token, we construct the log ratio of interest expenditures to sales and the log ratio of domestic profits to sales as the regressors. We then perform the fixed-effects and IV fixed-effects estimations to reconsider the impact of credit constraints on exports. Column (1) of Table 3.5 presents the fixed-effects estimation results for our three key variables: interest expenditures, TFP, and the interaction term between interest expenditures and FIEs. We can clearly observe that, even when measured in ratios, all coefficients still have the anticipated signs and are statistically significant. Adding more control variables in column (2) does not change our findings qualitatively. After controlling for the endogeneity issue, estimates in columns (3) and (4) are still in line with our previous findings. The only slightly surprising finding is that, in column (4), the coefficients on the ratio of interest expenditures to sales and TFP are insignificant. However, they still have the anticipated positive signs. In sum, all our robustness checks suggest that our findings are robust regardless of specification.

6   Concluding Remarks In this chapter, we first constructed a theoretical model to consider how a firm’s credit constraints affect its exports. Firms with higher productivity have higher probabilities for success of its projects. They thus are more capable of obtaining external finance from financial intermediates. Consequently, they are less credit constrained and are able to export more abroad. Moreover, FIEs are less sensitive to the availability of external finance since they have access to funds from their foreign parent firms. We then test the theoretical predictions using a very rich Chinese firm-­ level data set. After controlling for productivity and the endogeneity of a firm’s capacity for external borrowing, we find strong empirical evidence

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

89

Table 3.5  Estimates of ratio specifications Regressand

FE

Log ratio of exports to sales (ln (EX/Sales)it)

(1)

(2)

Log ratio of interest expenditure to sales (ln (IE/Sales)it)

0.019**

0.049** 0.535**

0.562

(6.66) 0.035** (3.73) – −0.013** (−4.09)

(2.77) 0.003 (0.05) – −0.024 (−1.19) −0.051

(1.16) 0.062 (0.70) – −0.272 (−1.15) −0.036*

Log TFP ( ln TFPitOP −1 ) FIEs dummy (FIEi) FIEi ×  ln (IE/Sales)it Log ratio of domestic profits to sales  Dprofit −1  ( ln  )  Salesit  SOEs dummy (SOEi) SOEi ×  ln (IE/Sales)it Firm fixed effects Year fixed effects Observations Prob.>F R-squared

FE + IV

(−4.21) – 0.019 (0.73) Yes Yes Yes Yes 178,136 33,492 0.000 0.000 0.01 0.01

(3)

(4.64) 0.091** (5.54) – −0.327** (−4.66)

(4)

(−1.87) – −0.078 (−0.81) Yes Yes Yes Yes 178,130 33,488 0.000 0.000 0.01 0.01

Notes: Robust t-values corrected for clustering at the firm level in parentheses. *(**) means significant at the 10(5) percent level

to support our theoretical argument. In particular, firms that are more capable of obtaining external finance are shown to have higher levels of exports. Moreover, the exports of FIEs are less sensitive to the firm’s access to external finance from financial intermediaries. All of these findings are robust to different measures and econometric settings. Our work enriches our understanding of impact of productivity on exports. Productivity could affect exports through credit constraints. Different productivity levels induce different project-specific success probabilities. Observing this, financial intermediaries prefer to extend financing to projects with less risk, given the same levels of principal and repayment. Thus, firms with higher productivity face less severe credit constraints and are able to export and to export more. In short, higher productivity leads to easier access to external finance, and, consequently, to more exports. Our work contributes to the literature in three important ways. First, our theoretical model suggests that project-specific risk leads to different

90 

M. YU

levels of credit constraints faced by firms. Second, empirically, we offer firm-level evidence that credit constraints are an important determining factor of a firm’s exports. Firms with easier access to external finance or alternative sources of funds face fewer credit constraints, and, consequently, they export more. Third, we offer evidence that foreign capital inflow affects firms’ exports through credit constraints. Several extensions and possible generalizations merit special consideration. One of them is to consider outward foreign direct investment (FDI) into the model in the sense that firms with higher productivity would perform outward FDI in addition to exports. Another possible extension is to consider how policy shocks like exchange rate changes affect a firm’s exports and FDI decisions in the presence of credit constraints. These are the topics that we will pursue in the future.

Appendix 1: Solving the Cutoffs Domestic cutoff σ −1

xdσ −1 = wσ −1



σ  σ 1 µ L  σ − 1 P 

σ −1

1 Cd = wσ −1   w

σ xσ −1dFx ( x ) Cd ≡ hσ −1 (Cd ) µ x ≥∫xd



• Exports cutoff without credit constraints σ −1

xEσ −1 =

σ  σ τw  µ L∗  σ − 1 P ∗  σ −1

=



CE σ  σ −1 (τ w )   Cd∗  σ −1

σ −1

CE =

CE σ  σ −1 (τ w )   σ Cd∗  −1  1−σ

 σ w∗  σ   ∫ µ x≥x ∗  σ − 1 x 

dFx ( x ) Cd∗ =

d

• Exports cutoff with credit constraints

σ P ∗1−σ Cd∗ µ L∗ σ −1

CE  τ w  Cd∗  w∗ 

( )

hσ −1 Cd∗



91

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

rE ( xCE )

σ



rE ( xCE )

σ



 1  − w ∗C E =  − 1 1 − δ i ) w∗CE − dπ d ( xCE )  λ ( x )  ( CE  

=

(

(1 − δ

i

+ λ ( xCE ) δ i ) w∗CE

λ ( xCE )

)

 1  − − 1 dπ x  λ ( x )  d ( CE ) CE  

Substituting in revenue and profit, we get 1

∗ σ −1  σ  σ σ −1  (1 − δ i + λ ( xCE ) δ i ) w CE  1  1 = + − dwC     d   λ (x )   σ − 1  µ   λ ( xCE ) CE    1

xCE

1 −σ  ∗ ∗  τ w 1−σ  1  w − 1  dwL    w L  ∗  +    P  P  λ ( xCE )  



1

 1 −σ   

In equilibrium, by substituting in P and P∗, we can solve the cutoffs: 1

xCE

 (1 − δ i + λ ( xCE ) δ i ) w∗CE  1 σ −1  = + − 1  dwCd   λ (x )    λ ( xCE ) CE    

1 −σ 1 −σ  ∗ ∗  σ 1−σ µ  τ w 1−σ  1   σ  µw 1 − dwL + w L           ∗  λ (x )    σ −1  σ  P   σ −1  σ  P  CE   

1

 1 −σ   

1



 (1 − δ i + λ ( xCE ) δ i ) w∗CE  1 σ −1  = + − 1  dwCd   λ (x )    λ ( xCE ) CE     1

1 −σ 1−σ  1 −σ  ∗ ∗   1 1−σ  1  1 − 1  dw 2 −σ Cd   hσ −1 (Cd )   w Cd (τ w )  ∗  hσ −1 Cd∗ +   λ(x )    w  w CE    

( )

92 

M. YU



 σ −1 w∗   (1 − δ i + λ ( xCE ) δ i ) w CE + (1 − λ ( xCE ) ) dCd   =  ∗ σ  λ ( x )τ 1−σ  w  C ∗ h1−σ C ∗ + 1 − λ ( x ) dC h1−σ ( C )  ( )   d d CE d d  CE   w  

1

( )

Appendix 2: TFP Calculation by the Olley–Pakes (1996) Approach One of the most important features of the Olley–Pakes approach is to model investment as a function of unobserved productivity as well as capital input. By assuming that three factor inputs (i.e., capital, labor, and material) are used to produce goods, the Olley–Pakes estimation method includes three steps. First, a semi-parametric empirical method is used to estimate the coefficients of both labor and material inputs. In particular, the unobserved productivity shock is modeled as an inverse function of investment, which is characterized by a fourth-order polynomial. Given China’s WTO accession in 2001, which was a positive demand shock for China’s exports, we also include the WTO dummy in this polynomial to capture the effect of the WTO accession. In particular, we consider a polynomial g(∙) as follows: 4



4

g ( kit ,I it ,EFit ,WTOt ) = (1 + WTOt + EFit ) ∑∑θ hq kith I itq h=0 q =0



where the polynomial depends on capital k, investment I, the dummy of the exporting firm EF, and a WTO dummy that equals one since 2002. The coefficients θhq, ∀ h, ∀ q are of interest to estimate. For a more detailed introduction to this approach, see Yu (2008). Second, after the coefficients of labor and material inputs are obtained, the coefficient of capital can be estimated by using a nonlinear square estimation. In this way, Olley and Pakes (1996) show that the simultaneity problem is well controlled. Third, to control the selection bias problem, we first use a probit function to estimate the probability of a firm’s survival in the next period. Once the fitted value of firms’ survival ratio is obtained, we put it into the inverse investment function once again to estimate all the three input coefficients. Finally, the residual between the data and the fitted value obtained from the three estimated input coefficients is the Olley–Pakes TFP.

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

93

Table 3.6  Main notation for the models Symbol

Definition

q0 ω Ω σ μ p P L, L∗ r(ω) Cd, CE

Panel A: Theoretical framework Quantity of homogeneous good Variety of differentiated good Overall set of varieties available to the consumer Elasticity of substitution between differentiated goods, σ > 1 Expenditure share in homogeneous good Price of each variety Price index of countries Home and foreign population Revenue of each firm producing variety ω Fixed entry cost of domestic market and exports market

w, w∗ x π λ G δi t τ d lijt M1 ςi ϑt ϵit

Home and foreign wage Firm’s productivity Profit Success possibility of project Repayment demanded by financial intermediaries The portion of fixed entry cost financed by alternative external fund The fraction of the domestic fixed cost pledgeable as collateral 丁 Iceberg transportation cost The fraction of the domestic profit used as internal finance Panel B: Empirical specification Number of employees of firm i in industry j in year t China base money supply Firm-specific fixed effect Year-specific fixed effect Idiosyncratic error term

Table 3.7  Total factor productivity of Chinese plants Industry (code)

Mining & washing of coal (6) Mining & processing of ferrous metal ores (8) Mining & processing of nonferrous metal (9) Mining & processing of nonmetal ores (10) Processing of food (13) Manufacture of foods (14)

Est. Prob.

Labor OLS

OP

Materials

Capital

OLS

OLS

OP

OP

0.992 0.063 0.043 0.834 0.813 0.059 0.081 0.998 0.096 0.092 0.872 0.898 0.040 0.038 0.999 0.058 0.072 0.889 0.876 0.042 0.101 0.995 0.083 0.066 0.819 0.791 0.044 0.099 0.994 0.068 0.043 0.833 0.890 0.048 0.058 0.995 0.057 0.058 0.850 0.840 0.049 0.023 (continued)

94 

M. YU

Table 3.7  (continued) Industry (code)

Manufacture of beverages (15) Manufacture of tobacco (16) Manufacture of textile (17) Manufacture of apparel, footware, & caps (18) Manufacture of leather, fur, & feather (19) Processing of timber, manufacture of wood, bamboo, rattan, palm & straw products (20) Manufacture of furniture (21) Manufacture of paper & paper products (22) Printing, reproduction of recording media (23) Manufacture of articles for culture, education, & sport activities (24) Processing of petroleum, coking, &fuel (25) Manufacture of raw chemical materials (26) Manufacture of medicines (27) Manufacture of chemical fibers (28) Manufacture of rubber (29) Manufacture of plastics (30) Manufacture of nonmetallic mineral goods (31) Smelting & pressing of ferrous metals (32) Smelting & pressing of nonferrous metals (33) Manufacture of metal products (34) Manufacture of general purpose machinery (35) Manufacture of special purpose machinery (36) Manufacture of transport equipment (37) Electrical machinery & e quipment (39)

Est. Prob.

Labor OLS

0.994 0.999 0.994 0.993

0.089 0.053 0.066 0.100

Materials

Capital

OP

OLS

OP

OLS

OP

0.068 0.048 0.056 0.096

0.820 0.848 0.863 0.792

0.855 0.854 0.879 0.796

0.052 0.161 0.033 0.053

0.044 0.182 0.036 0.019

0.989 0.082 0.082 0.846 0.842 0.043 0.078 0.989 0.074 0.051 0.841 0.881 0.038 0.045

0.989 0.107 0.154 0.802 0.732 0.046 0.077 0.990 0.066 0.061 0.851 0.849 0.044 0.048 0.994 0.088 0.063 0.796 0.847 0.068 0.052 0.991 0.086 0.068 0.822 0.827 0.049 0.045 0.992 0.035 0.041 0.864 0.906 0.062 0.061 0.991 0.053 0.031 0.830 0.857 0.063 0.074 0.990 0.995 0.996 0.994 0.993

0.101 0.047 0.078 0.079 0.049

0.064 0.029 0.089 0.074 0.038

0.785 0.901 0.801 0.821 0.858

0.803 0.923 0.729 0.816 0.870

0.060 0.028 0.067 0.056 0.040

0.002 0.032 0.142 0.051 0.870

0.991 0.054 0.043 0.891 0.921 0.036 0.036 0.995 0.052 0.038 0.887 0.889 0.031 0.052 0.994 0.078 0.102 0.793 0.710 0.067 0.063 0.995 0.066 0.049 0.827 0.835 0.057 0.058 0.993 0.067 0.029 0.809 0.868 0.060 0.070 0.992 0.086 0.077 0.809 0.804 0.065 0.058 0.996 0.085 0.068 0.812 0.833 0.063 0.119 (continued)

3  EXPORTS, PRODUCTIVITY, AND CREDIT CONSTRAINTS 

95

Table 3.7  (continued) Industry (code)

Manufacture of communication equipment, computers & other electronic equipment (40) Manufacture of measuring instruments & machinery for cultural activity & office work (41) Manufacture of artwork (42) Electric power & heat power (44) Production & supply of gas (45) Production & supply of water (46) All industries

Est. Prob.

Labor OLS

OP

Materials

Capital

OLS

OLS

OP

OP

0.994 0.103 0.094 0.776 0.785 0.082 0.148

0.992 0.089 0.049 0.724 0.815 0.096 0.050

0.992 0.996 0.999 0.981 0.993

0.084 0.156 0.072 0.046 0.068

0.073 0.140 0.035 0.019 0.061

0.821 0.611 0.653 0.671 0.825

0.849 0.590 0.558 0.636 0.828

0.046 0.219 0.184 0.172 0.062

0.045 0.217 0.275 0.163 0.075

Notes: I do not report standard errors for each coefficient to save space, which are available upon request

References Amiti, M. and J.  Konings, “Trade Liberalization, Intermediate Inputs, and Productivity: Evidence from Indonesia”, American Economic Review, 2007, 97(5), 1611–1638. Anderson, T. W., and H. Rubin (1949), “Estimation of the Parameters of a Single Equation in a Complete System of Stochastic Equations,” Annals of Mathematical Statistics 20, pp. 46–63. Anderson, T.W. (1984), Introduction to Multivariate Statistical Analysis, 2nd ed. New York: John Wiley & Sons. Bajona, Claustre and Tianshu Chu (2004), China’s WTO Accession and Its Effect on State-owned Enterprises, East-West Center Working Paper, No. 70. Bernard, Andrew and Bradford Jensen (1999), “Exceptional Exporter Performance: Cause, Effect, or Both?” Journal of International Economics, 47(1), 1–25. Buch, Claudia, Iris Kesternich, Alexander Lipponer, and Monika Schnitzer (2008), “Real versus Financial Barriers to Multinational Activity,” mimeo, University of Tuebingen. Cai, Hongbin and Qiao Liu (2009), “Competition and Corporate Tax Avoidance: Evidence from Chinese Industrial Firms,” Economic Journal 119, pp.764–795. Chaney, Thomas (2005), “Liquidity Constrained Exporters”, mimeo, University of Chicago. Claessens, Stijn and Tzioumis, Konstantinos (2006), “Measuring Firms’ Access to Finance,” mimeo, World Bank and Brooking conference paper.

96 

M. YU

Cragg, J.G. and S.G. Donald (1993), “Testing Identfiability and Specification in Instrumental Variables Models,” Econometric Theory 9, pp. 222–240. Harrison, Ann E. and Margaret S.  McMillan (2003), “Does Direct Foreign Investment Affect Domestic Credit constraints?” Journal of International Economics 61, 73–100. Helpman, Elhanan, Marc Melitz, and Yona Rubinstein (2007), “Estimating Trade Flows: Trading Partners and Trading Volumes,” Quarterly Journal of Economics 123, pp. 441–87 Héricourt, Jérôme and Sandra Poncet (2009), “FDI and Credit Constraints: Firm Level Evidence in China”, Economic Systems 33 (1), pp. 1–21. Holz, Carsten (2004), “China’s Statistical System in Transition: Challenges, Data Problems, and Institutional Innovations,” Review of Income and Wealth 50(3), pp. 381–409. Kleibergen, Frank and Richard Paap (2006), “Generalized Reduced Rank Tests Using the Singular Value Decomposition,” Journal of Econometrics, 133(1), pp. 97–126. Krugman, P. (1979), Increasing Returns, Monopolistic Competition, and International Trade, Journal of International Economics 9, pp. 469–479. Levinsohn, James and Amil Petrin (2003), “Estimating Production Functions Using Inputs to Control for Unobservable,” Review of Economic Studies 70(2), pp. 317–341. Lin, Justin Yifu (2003), “Development Strategy, Viability, and Economic Convergence,” Economic Development and Cultural Change 51(2), pp. 277–308. Manova, Kalina (2008), “Credit Constraints, Heterogeneous Firms and International Trade,” NBER Working Paper, No. 14531. Melitz, M. J., 2003, “The Impact of Trade on Intra-industry Reallocations and Aggregate Industry Productivity”, Econometrica, Vol. 71(6), pp. 1695–1725. Muûls, Mirabelle (2008), “Exporters and Credit Constraints: A firm-level Approach,” National Bank of Belgium working paper number 139. Olley S. and A.  Pakes, 1996, The Dynamics of Productivity in the Telecommunications Equipment Industry, Econometrica, 64(6), pp. 1263–97. Qiu, Larry D. (1999), “Credit Rationing and Patterns of New Product Trade,” Journal of Economic Integration, 14(1), pp. 75–95. Silva, J.M.C. Santos and Tenreyro, Silvana (2006), “The Log of Gravity,” Review of Economics and Statistics 88 (4). pp. 641–658. Wooldridge, Jeffery M. (2002). Econometric analysis of cross section and panel data. Cambridge, Massachusetts: MIT Press. Yu, Miaojie (2008), “Trade Liberalization, Firm Exits, and Productivity: Evidence from Chinese Plants,” Available at SSRN: http://ssrn.com/abstract=1146786. Yu, Miaojie (2009), “Revaluation of the Chinese Yuan and Triad Trade: A Gravity Assessment,” Journal of Asian Economics 20, pp. 655–668.

CHAPTER 4

Exports and Credit Constraints under Incomplete Information

This chapter examines why credit constraints for domestic and exporting firms arise in a setting where banks do not observe firms’ productivities. To maintain incentive-compatibility, banks lend below the amount that firms would need for optimal production. The longer time needed for export shipments induces a tighter credit constraint on exporters than on purely domestic firms, even in the exporters’ home market. In our application to Chinese firms, we find that the credit constraint is more stringent as a firm’s export share grows, as the time to ship for exports is lengthened, and as there is greater dispersion of firms’ productivities reflecting more incomplete information.

1   Introduction The financial crisis of 2008 has led researchers to ask whether credit constraints faced by exporters played a significant role in the fall in world trade. There are a wide range of answers: Amiti and Weinstein (2011) argue that trade finance was important in the earlier Japanese financial crisis of the 1990s and for the US recently, and Chor and Manova (2012) find that financially vulnerable sectors in source countries did indeed

This chapter is coauthored with Robert Feenstra and Zhiyuan Li and originally published in Review of Economics and Statistics, 2014, 96(4), pp. 729–744. © The Author(s) 2021 M. Yu, Exchange Rate, Credit Constraints and China’s International Trade, https://doi.org/10.1007/978-981-15-7522-8_4

97

98 

M. YU

experience a sharper drop in monthly export to the US.  In contrast, Levchenko et al. (2010) find no evidence that trade credit played a role in restricting imports or exports for the US, while for Belgium, Behrens et al. (2010) argue that to the extent that financial variables impacted exports, they also impacted domestic sales to the same extent. Of course, the potential causal link between financial development and international trade at country level was recognized long before the recent crisis. For example, Kletzer and Bardhan (1987; see also Beck 2002; Matsuyama 2005) argued that credit-market imperfections would adversely affect exporters needing more finance and hence influence trade patterns. That theme was modeled by Chaney (2005) in a Melitz (2003) framework, and implemented by Manova (2012), who argues that credit constraints have systematically different effects depending on the financial vulnerability of the exporter’s sector and financial development of their country.1 In view of the divergent findings on the role of credit constraints during the crisis, we believe that it is useful to go back to the theory and ask why credit for exports should be allocated any differently than credit for domestic sales. Amiti and Weinstein (2011) argue forcefully for two reasons: there is a longer time-lag between production and the receipt of sales revenue; and exporters also face inherently more risk, since it is more difficult to enforce payment across country boundaries. They define trade finance (as distinct from trade credit) to be the financial contracts that arise to offset these risks for exporters. We will pick up on the first of these reasons, the longer time to ship for exports, which is also discussed in relation to the financial crisis by Berman et  al. (2012a).2 The goal of this chapter is to build time to ship into a model of heterogeneous firms obtaining working-capital loans from a bank, to see whether exports are indeed treated differently from domestic sales in theory. We test the predictions of the model using firm-level data for China.

1  Other papers dealing with trade and finance include Qiu (1999), Greenaway et al. (2007), Harrison and McMillan (2003), Muûls (2008), Buch et al. (2008), Héricourt and Poncet (2009), Poncet et al. (2009), and Egger and Keuschnigg (2011). 2  In our working paper (Feenstra et al. 2011), we also included the risk faced by exporters in international markets. But because that risk was taken as exogenous (in contrast to Ahn (2011), for example), it had little impact on the theory and could not be tested with our Chinese firm-level data, so that extension is omitted here. Berman et al. (2012a) also take the risk of default as exogenous but model it as depending on the time to ship, so that it plays an important role in their model and estimation.

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

99

The key feature of our model is that the bank has incomplete knowledge of firms, in two respects. First, the bank cannot observe the productivity of firms. We believe this assumption is realistic in rapidly growing economies such as China with rapid entry, and perhaps more generally. The bank will confront firms with a schedule specifying the amount of the loan and the interest payments to maximize its own profits. From the revelation principle, without loss of generality we can restrict attention to schedules that induce firms to truthfully reveal their productivity. Second, the bank cannot verify whether the loan is used to cover the costs of production for domestic sales or for exports. This second assumption means that we are not really modeling the loans from the bank as trade finance: such loans would typically specify the names of the buying and selling party, at least, so the bank could presumably verify whether the loan was for exports or not.3 Rather, the loans being made by the bank are for working capital, to cover the costs of current production, regardless of where the output is sold. The assumption that banks cannot follow a loan once the money enters the firm is made in a different context by Bolton and Scharfstein (1990), for example. With these assumptions, in Sect. 2 we derive the incentive-compatible loan schedule by the bank that maximizes its own profits. Sales revenue of firms is less than would occur at their optimal production, that is, the incentive-compatible loans impose credit constraints on firms. The reason for these credit constraints is that a firm suffers only a second-order loss in profits from producing slightly less than the production with complete information and borrowing less from the bank, but obtains a first-order gain from reducing its interest payments in this way. So a firm that is not credit constrained will never reveal its true productivity and borrow enough to produce at the level with complete information; hence, incentive-­ compatibility requires that the firm is credit constrained. Furthermore, because banks cannot follow a loan once it enters the firm, the credit constraint applies to the exports and domestic sales of a firm engaged in both these activities—which we refer to as an exporting firm. Because exports take longer in shipment, such exporting firms face a tighter credit constraint on both markets than purely domestic firms. So our answer to the question: is credit for exports and domestic sales treated differently, is nuanced; when these activities occur in the same firm, the bank treats them equally; but when these activities occur in an 3

 Ahn (2011) provides an information-based model of trade finance.

100 

M. YU

exporting firm and a purely domestic firm respectively, they are indeed treated differently. The tighter credit constraint on exporting firms comes from the longer time-lag between production and receipt of sales revenue, and reduces exports on both the intensive and extensive margins. These theoretical results are tested using a rich panel data set of Chinese manufacturing firms over the period 2000–2008, in Sects. 3 and 4. This application is of special interest because China’s exports experienced unprecedented growth over the past decades, while it is believed that Chinese firms faced severe credit constraints: according to the Investment Climate Assessment surveys in 2002, China was among the group of countries that had the worst financing obstacles (Claessens and Tzioumis 2006). We estimate a structural equation under which sales revenue depends on interest payments, the export share and other variables. We obtain robust empirical evidence that exporting firms face more severe credit constraints than purely domestic firms. The credit constraint is more stringent as a firm’s export share grows, as the time to ship for exports is lengthened, and as there is greater dispersion of firms’ productivities reflecting information incompleteness. These results go beyond Manova (2012), who focuses on the financial vulnerability of sectoral exports, by showing how production characteristics of the firm (i.e., its export share and mode of transport) and industry (i.e., information incompleteness) influence the credit constraint. But as in Manova (2012), we find that higher collateral can offset the credit constraint and expand exports. Conclusions and directions for further research are discussed in Sect. 5; and an online Appendix includes additional theoretical and empirical results.4

2   Incentive-Compatible Loans 2.1  The Model We suppose there are two countries, home and foreign (henceforth foreign counterparts of the variables are denoted with an asterisk *). Labor is the only factor for production and the population is of size L at home. There are two sectors where the first produces a single homogeneous good that is freely traded and chosen as numeraire. Both countries 4

 The Appendix is available at http://www.econ.ucdavis.edu/faculty/fzfeens/papers.html.

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

101

produce in this sector with constant return to scale technology and thus home wage (w) is fixed by productivity in this sector. The second sector produces a continuum of differentiated goods under monopolistic competition, as in Melitz (2003). Consumers Consumers are endowed with one unit of labor and the preference over the differentiated good displays a constant elasticity of substitution. The utility function of the representative consumer is s



m

s -1 æ ös -1 U = q10- m ç ò q (w ) s dw ÷ è wÎW ø

where ω denotes each variety, Ω is the set of varieties available to the consumer, σ > 1 is the constant elasticity of substitution between each variety, and μ is the share of expenditure on the differentiated sector. Accordingly, the demand for each variety is:



Y æ p (w ) ö q (w ) = çç ÷ P è P ÷ø

-s

(4.1)

where Y ≡ μwL is the total expenditure on the differentiated good at home, 1

æ ö 1 -s 1 -s p(ω) is the price of each variety, and P º ç ò p (w ) dw ÷ is the aggre è wÎW ø gate price index in the differentiated sector. Firms and the Bank Firms in the differentiated sector need to borrow working capital to finance a fraction δ of their fixed and variable costs. Firms borrow from a single, monopolistic bank, and the bank charges interest payments to maximize its profits. The timing of events is as follows. The bank specifies a loan and interest payment schedule based on publicly known productivity distribution. Then the firms draw their productivities and borrow from the bank. When borrowing from the bank, a firm will claim a productivity level to maximize its profit taking the loan and interest payment schedule

102 

M. YU

as given. With the resulting loans, firms choose markets to serve and produce. Revenues are then realized and the bank collects payments. Notice that the loan and interest payment schedules are worked out initially by the bank, and then firms self-select into the export market and choose the quantity to produce accordingly. Thus, the bank cannot take into account firm’s export status and production as extra information when it chooses the loan and interest payment schedule. But under the incentive-compatible loan contract, the bank can perfectly predict whether a firm will be an exporter or not. The bank faces an opportunity cost of i—the interest rate—on its loans. We assume that the loans for domestic (export) projects are paid back after τd(τe) periods, and further assume that τe > τd, reflecting the longer time-­ lags involved in the shipping of exports. 2.2  Domestic Firms’ Decision Under incomplete information, the bank does not observe the productivity level, x, of a firm coming to it for a loan. In order to maximize profits, the bank will design a schedule of loans Md(x′) and interest payments Id(x′) contingent on firms’ announced productivity level x′. By the revelation principle, the bank can do no better than to design a loan–interest payment schedule that induces firms to reveal their true productivity, x′ = x. Adding this incentive compatibility condition as a constraint, the domestic firm’s profit maximization problem is:



æq w ö max x ¢,qd p d ( x,x ¢ ) = pd qd - (1 - d ) ç d + Cd ÷ - ( M d ( x ¢ ) + I d ( x ¢ ) ) (4.2) x è ø s.t. p d ( x,x ) ³ p d ( x,x ¢ )

p d ( x ,x ) ³ 0



æ qd w ö Md ( x¢) ³ d ç + Cd ÷ è x ø

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

103

and also subject to the domestic demand function in (domestic demand), where Cd is the fixed cost.5 The first constraint is the incentive compatibility constraint, the second ensures that profits are nonnegative, and the third specifies that the amount of the loan must cover the fraction δ of fixed and variable costs at the chosen production level qd. Using the fact that the third constraint above will be binding in equilibrium, we take the derivative of the profit respect to announced productivity, x′, to obtain the first-order condition:



éFd ( x,M d ( x ) ) - 1ù ë û

M d¢ ( x )

d

= I d¢ ( x )

(4.3)



where





é æ s - 1 öù w Fd ( x,M d ( x ) ) º ê pd ç ÷ú / ë è s øû x ö æ s - 1 ö æ Md ( x ) =ç - Cd ÷÷ çç ÷ è s øè d ø

-

1 s

æ xP ö ´ç ÷ è w ø

s -1 s

1

Ys

The value of Φd in the first line of (measure of constraint) is recognized as the ratio of marginal revenue to marginal costs. A firm without any need to borrow will produce where Φd = 1, while a firm that produces less due to insufficient loans will have Φd > 1. This means that Φd is a measure of the firm’s credit constraint; and the larger the Φd, the lower is the quantity produced due to this constraint. The second line of (measure of constraint) is obtained by using the binding quantity level in the third constraint and its corresponding price from demand in (domestic demand). It is apparent that having lower loans Md(x) will raise Φd, indicating that the credit constraint is tightened. We can now develop some intuition as to why the bank might need to impose credit constraints. Let us suppose that the bank lends more to

5  Notice here we assume away risks. Including risks and collateral in the problem would not affect our main results, as shown in a more comprehensive version of the model in Feenstra et al. (2011).

104 

M. YU

higher productivity firms, and also collects more in interest payments.6 Then in (IC solution), both M′d(x) and I′d(x) are positive. It follows that the expression in brackets on the left must be positive, so it follows that the firm must be credit constrained, that is, Φd > 1. The reason this condition is needed is that, if the bank specifies loan and interest schedules such that firms are not credit constrained and all profits are paid back to the bank, a firm that is supposed to produce at the monopoly optimum with marginal revenue equal to marginal cost would have only a second-order loss in profits from announcing a slightly smaller productivity x′, and producing slightly less. But the firm would have a first-order gain from the reduction in interest payments I′d(x) > 0. So a firm at the monopoly optimum would always understate its productivity, and it follows that a credit constraint is needed to ensure incentive compatibility. 2.3  Exporters’ Decision We assume that the monopolistic bank cannot enforce different contracts to separate loans for domestic market and export market. Rather, exporters are free to determine how to allocate the loan to both markets. An exporter thus chooses quantities to produce at domestic market and export market and claims a productivity x′ to maximize its profit: max x ¢,qd ,qe p e ( x,x ¢ ) = pd qd + pe qe - (1 - d )

qw æq w ö ´ ç d + Cd + e + Ce ÷ - ( M e ( x ¢ ) + I e ( x ¢ ) ) (4.4) x x è ø s.t. p e ( x,x ) ³ p e ( x,x ¢ )

p e ( x,x ) ³ p d ( x,x )



qe w æ qd w ö Me ( x¢) ³ d ç + Cd + + Ce ÷ x è x ø

* and subject to export demand, q = Y æ pe ö e ç ÷ P* è P* ø

-s

, where Y∗ is the foreign

6  We show in the Appendix that these monotonicity conditions hold in the optimal schedules for the bank.

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

105

total expenditure on the differentiated good.7 The total loan received by the exporter is denoted by Me and total interest payments are Ie, while Ce is the fixed cost of exporting. The first two constraints above are analogous to those for the domestic firm, but the third constraint is different and important. It states that the total amount of the loan given to the exporter must cover the workingcapital needs of both domestic and export production costs. From the exporting firm’s perspective, these funds are fully fungible so the bank is making a single loan and likewise receiving a single interest payment. Solving the problem for the choice of qd and qe, it is readily shown that the firm will maximize its profit by choosing quantities in the two markets such that: æ s -1 ö æ s -1 ö pd ç = pe ç ÷ ÷ è s ø è s ø



This condition states that the loan will be allocated within the firm so that marginal revenue in the domestic and export markets are equalized. It means that for any given loan, the bank will know exactly how production is allocated between the two markets. Thus, for notational convenience, we break up the total loan Me(x’) into the component intended to cover domestic costs M ed x’ , and the component intended to cover export costs M ee x’ . That is, we will define the loans allocated to each market as

( )

( )



æq w ö M ed ( x ¢ ) º d ç d + Cd ÷ , x è ø



æq w ö M ee ( x ¢ ) º d ç e + Ce ÷ . x è ø

(4.5)

7  We do not make explicit the transportation costs to the export market for expositional convenience, but that iceberg cost can readily be incorporated into the definition of the effective foreign expenditure on the differentiated good Y∗. That is, including iceberg trans-s , which equals that shown æ ˜* ö port costs τ > 1, then export demand is qe = çç Y / P * ÷÷ t pe / P * è ø ˜* * -s in the export demand by defining Y = Y t .

(

)

106 

M. YU

Using domestic and export demand, combined with the requirement from (phi relation) that the prices pd and pe are equalized, it immediately follows that the loans to the two markets are related by M ee ( x ) / d - Ce

M ed ( x ) / d - Cd



=

he , hd

(4.6)

where we define the shares of demand coming from the domestic and foreign markets as

hd =



YPs -1 Y * P *s -1 and h = e YPs -1 + Y * P *s -1 YPs -1 + Y * P *s -1

(4.7)

Using the optimal quantity sold in each market from (loans) and its associated price, we can rewrite the firms’ profits as a function of productivity, x, and the amount borrowed for domestic market, M ed x’ . Similar to the problem for domestic firms, by taking derivative of profits respect to x’, we obtain the first-order condition for incentive compatibility:

( )

(

)

éFed x,M ed ( x ) - 1ù ë û

M ed ¢ ( x )

d

(

)

+ éëFee x,M ee ( x ) - 1ùû

M ee¢ ( x )

d

= I e¢ ( x )

(4.8)



where s -1 -1 d ö s æ xP ö s 1 é æ s - 1 öù w æ s - 1 ö æ Me ( x ) d d Ys Fe x,M e ( x ) = ê pd ç - Cd ÷ ç ÷ ÷ú / = ç ÷ çç ÷ èwø ë è s øû x è s ø è d ø

(

(

Fee x,M ee ( x )

)

)

e ö é æ s - 1 öù w æ s - 1 ö æ Me ( x ) / =ç º ê pe ç - Ce ÷ çç ÷ ÷ ú ÷ ë è s øû x è s ø è d ø

-

(4.9) 1 s

æ xP * ö ç ÷ è w ø

s -1 s

Y

*

1 s

and from the equality of marginal revenues in (6) we have that

(

)

(

Fed x,M ed ( x ) = Fee x,M ee ( x )

)

(4.10)

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

107

The interpretation of these conditions is analogous to what we obtained for domestic firms. The values Fed and Fee are the ratio of marginal revenue to marginal costs in the two markets served by the exporter. Credit constraints would mean that Fee = Fed > 1 , so the firm would be selling less in both markets than would be optimal in the absence of any constraints. We now determine the magnitude of credit constraints that are optimal for the bank. 2.4  Bank’s Decision The monopolistic bank chooses the loans given to domestic firms subject to the incentive-compatibility condition (4.3), and chooses the loans given to exporters for the domestic market ( M ed ( x ) ) and for export market ( M ee ( x ) ), subject to the incentive-compatibility conditions (4.8) and the equality of marginal revenue (4.10). The bank’s problem is then to choose Md(x), M ed ( x ) , M ee ( x ) , Id(x), and Ie(x) to maximize its profits: Xe

max M , I

ò ( I ( x ) - it d

d

M d ( x ) ) f ( x ) dx

Xd

¥

+

ò ( I ( x ) - it e

Xe

d

)

M ed ( x ) - it e M ee ( x ) f ( x ) dx

(4.11)

s.t. (4.3) if x Î éë xd , xe ) , and (4.8) and (4.10) if x Î éë xe , ¥ ) where f(x) is the probability density function of firms’ productivity distribution. The variables xd and xe are the productivities of the cutoff domestic firm and the cutoff exporter, respectively. As in the Melitz (2003) model, firms will enter into domestic production and export based on the profitability of these activities. This means that the cutoff domestic firm with productivity xd is defined by the zero-­cutoff-­profit condition p d ( xd , xd ) = 0 and the cutoff exporter with productivity xe by the condition p d ( xe , xe ) = p e ( xe , xe ) . These cutoff productivities can differ from the Melitz (2003) model, of course, because here they are influenced by the credit conditions offered by the bank. The maximization problem Eq. (4.11) is solved in two steps. First, we determine the loan schedule that maximizes bank’s profit, which is an optimal control problem analyzed in the Appendix. But that still leaves

108 

M. YU

open the initial level of interest payments for the cutoff domestic and exporting firms: these initial interest payments will in fact determine the productivity levels xd and xe for these firms. So the second step in the optimization problem for the bank is to determine the optimal initial interest payments for these cutoff firms, or equivalently, solving for the optimal cutoff productivities and consequently obtain the implied initial interest payments. To simplify the solution, we consider a Pareto distribution for firms q æ1ö productivity, F ( x ) = 1 - ç ÷ , x ³ 1 , where θ is the shape parameter.8 It is èxø shown in the Appendix that the optimal loan schedules for the bank are such that -1

æ s -1 ö Fd ( x,M d ( x ) ) = Fd º (1 + idt d ) ç 1 sq ÷ø è



-1

æ s -1 ö Fed x,M ed ( x ) = Fee x,M ee ( x ) = Fe º éë1 + id (t dh d + t ehe ) ùû ç 1 (4.12) sq ÷ø è

(

)

(

)

Examining the features of these solutions, we see that credit constraints for domestic firms and exporters apply, meaning that Fd > 1 and Fe > 1 , even if i = 0 in Eq. (4.12). Thus, even when the bank has no opportunity cost of making loans, a credit constraint is still needed to ensure incentive compatibility. When i > 0 then the credit constraint is further increased, and it is intuitive that the bank will restrict credit more as its opportunity cost rises. The opportunity cost is measured relative to the time required for the domestic and foreign loans, or τd and τe, respectively. We have assumed that τe > τd, from which it follows that the credit constraint Fe for exporters in either their domestic or export markets exceeds Fd for domestic firms in Eq. (4.12), when i > 0. The extra constraint faced by exporters will be the key testable implication in our empirical application. While the solution for the credit constraints implies the slope for the interest payment schedules, from Eqs. (4.11) and (4.8), we still need to 8

 We assume θ > 1 as is needed for the mean of the Pareto distribution to be finite.

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

109

determine the initial interest payments. Considering first domestic firms, by taking the first derivative of Eq. (4.3) with respect to xd , we can obtain



I d ( xd ) = ( Fd - 1)

M d ( xd )

d



Consequently, from (4.3) and (4.12), the interest payment for domestic firms is



I d ( x ) = ( Fd - 1)

Md ( x )

d

(4.13)



It is shown in the Appendix that the lowest productivity domestic firm, xd , is above the cutoff productivity in Melitz (2003). Similarly, taking the first derivative of Eq. (4.3) with respect to xe , we obtain the solution for the initial interest payment for the cutoff exporter:



I e ( xe ) = ( Fe - 1)

M e ( xe )

d

+ iQ,



(4.14)

where the final parameter in the above equation is Qº

d (t e - t d ) (hd Ce - heCd ) æ s -1 ö ç 1 - oq ÷ è ø

Consequently, the interest payment schedule for exporters is



I e ( x ) = ( Fe - 1)

Me ( x )

d

+ iQ



It is also shown in the Appendix that the lowest productivity exporting firm, xe , is above the cutoff productivity for exporters in the Melitz model.

110 

M. YU

3   Estimating Equation and Data 3.1  Empirical Specification We can use our results above to derive an equation linking the revenue of the firm to its interest payments, and we shall estimate that equation using data on Chinese firms. The basic relationship between firms’ revenue and interest payments is linear in these variables, as we show below, but the coefficient on interest payments is a nonlinear function of the credit constraints faced by domestic firms and exporters. The credit constraint, in turn, depends on the firms’ share of exports as shown by ηe and ηd = 1 − ηe in Eq. (4.12). So we will end up with an estimating equation that is nonlinear in the export share, which we treat as an endogenous variable: both these features create complications in the estimation that we shall address. To derive the basic relationship between firms’ revenue and interest payments, start with domestic firms. The loans Md(x)/δ are needed to finance total costs, so Md(x)/δ − Cd are needed for variable costs. The ratio of marginal revenue to marginal costs is Fd , and the ratio of price to marginal revenue for CES demand is σ/(σ − 1). Therefore, the total sales revenue pdqd obtained from the working-capital loans of Md(x) are pd qd = éë M d ( x ) / d - Cd ùû Fds / (s - 1) . Substituting from Eq. (4.13), we obtain: pd qd =

ö æ I ( x) s Fd çç d - Cd ÷÷ s - 1 è Fd - 1 ø



A similar line of argument will show that the relationship between revenue and interest payments for an exporting firm is pd qd + pe qe =

ö æ I ( x ) - iQ s Fe çç e - Cd - Ce ÷÷ s - 1 è Fe - 1 ø



Summarizing the above relations, let us denote the interest payments and firm revenue as



ìï I ( x ) I ( x) º í d îï I e ( x )

if x Î [ x d ,x e ] , if x Î [ x e ,¥ ]



4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 



ïì p q r ( x) º í d d ïî pd qd + pe qe

if x Î [ x d ,x e ] if x Î [ x e ,¥ ]

111

.

Using these, we obtain a linear relation between revenue and interest for firm j in year t:

r ( x jt ) = b 0Cd + b1 I ( x jt ) + g1 jt I ( x jt ) + g2 jt Cd + g3 jt ,

(4.15)



where the coefficients are obtained from above as



b0 = b1 =



s Fd < 0, s -1

s æ Fd ö ç ÷>0 s - 1 è Fd - 1 ø

(4.16)



and g1 jt = g1 (hejt ) =

g2 jt = g2 (hejt ) = g3 jt = g3 (hejt ) = -



Fd ö s æ Fe ç ÷ £ 0, s - 1 è Fe - 1 Fd - 1 ø

(4.17)

s ( Fe - Fd ) £ 0, s -1

ù s éæ Fe ö êç ÷ Qit + FeCe ú ´ 1{x ji ³ xe } s - 1 êëè Fe - 1 ø úû



We define 1{x ³ xe } as an indicator variable that takes one for x ³ xe and zero otherwise, and the term Fe appearing above depends on the export share ηejt from Eq. (4.12). The coefficient β0 is negative because higher fixed costs reduce the amount of the loan available to cover variable costs, and therefore reduce revenue. The coefficient β1, which multiplies the bank payments, is positive, indicating that larger payments are associated with larger revenues. The remaining variables in Eq. (4.15) have coefficients gijt, i = 1, 2, 3, that are actually functions of the export share ηejt. Notice that from the definition of the credit constraints in Eq. (4.12), gijt(0) = 0, while for i = 1, 2,

112 

M. YU

these functions are strictly negative for positive export shares provided that τe > τd and i > 0, so that Fe > Fd . Thus, the extra terms involving gijt in Eq. (4.15) apply only to exporters and indicate additional credit constraints on those firms. To interpret these extra terms, consider first the function g1(ηejt), which is negative for exporters under the condition mentioned above but less than β1 in absolute value. So for exporters, bank payments of I(xjt) are associated with revenue of β1 + g1(ηejt), which is positive but less than β1. This reduced coefficient on payments therefore lowers the sales revenue for exporters, reflecting the extra credit constraint imposed on them. A similar logic applies to the fixed costs on domestic sales Cd that all firms face, which reduces revenue by the amount β0 + g2(ηejt) for exporters but only by β0 for domestic firms. So exporters are constrained in what they can earn due to the extra credit constraint that they face via both their bank payments and the fixed costs Cd. In addition, exporters face a reduction in revenue from any increase in the interest rate it, as shown by the final term g3(ηejt) appearing in Eq.(4.15), which also incorporates the extra fixed costs Ce faced by exporters. The presence of this term can be traced back to Θ in Eq. (4.14), which determines the interest payments for the cutoff exporter. As interest rates rise, or the time-lag for exports increases, the bank faces higher opportunity costs in making export loans and passes these on as higher interest payments, thereby reducing the extensive margin of exports. While Eq. (4.15) summarizes the basic equilibrium relationship between firms’ interest payments and revenue in our model, we must confront three challenges in its estimation. First, as it is written this equation has no error term: it holds exactly in the model. That limitation occurs because revenue r(xjt) appearing on the left depends on the productivity x that is known by each firm: we can think of this as ex ante productivity, and distinguish it from ex post productivity that would incorporate a host of random factors outside our model, including unanticipated problems in production, abnormal delays in shipping, government intervention, etc. So we denote by Rjt the actual revenue earned each firm, which differs from anticipated revenues by Rjt = r(xjt) + ϵjt with E(ϵjt| xjt) = 0, which will introduce an error term into Eq. (4.15). The presence of this error term, however, immediately leads to endogeneity issues in our explanatory variables. We expect that the observed interest payments Ijt in the data differ from the theoretical schedule I(xjt) so we write Ijt = I(xjt) + ujt with E(ujt| xjt) = 0. The error ujt is likely

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

113

correlated with the error ϵjt in revenue, because unanticipated problems of production and delivery can equally well impact interest payments to the bank. Accordingly, we treat interest payments as endogenous and so we need an instrument that is uncorrelated with the errors ϵjt and ujt. One such variable is the ex ante productivity that is anticipated by firms. We will use the technique of Olley and Pakes (1996) to make a distinction between total factor productivity (TFP) of the firm inclusive of the unanticipated, random productivity shocks (what we call TFP1), and TFP of the firms exclusive of these unanticipated shocks (what we call TFP2). The first of these is the standard firm-level measure of productivity, whereas the second makes use of the firm’s investment decision to infer the productivity that is anticipated by the firm, so it is correlated with xjt but not with the unanticipated shocks ϵjt and ujt. A second challenge arises from the coefficients gijt = gi(ηejt), i = 1, 2, 3, that are functions of the export shares and differ across firms due to these shares. These coefficients should therefore be treated as random across firms, and so the goal of our estimation will be to estimate a mean value of the coefficients. But the decision to export is endogenous in our model through the determination of xe in Eq.(4.14), so that only firms with productivity x jt > xe are exporters. The export share ηejt is therefore endogenous. Our estimating equation thus has random coefficients that are correlated with the endogenous export share, so it is a correlated random coefficients (CRC) model. To see the challenge that this creates in estimation, substitute Rjt = r(xjt) + ϵjt and Ijt = I(xjt) + ujt into Eq. (4.15) to obtain

R jt = b 0Cd + b1 I jt + g1 jt I jt + g2 jt Cd + g3 jt - ( b1 + g1 jt ) u jt + e jt



(4.18)

Even with E(ϵjt| xjt) = 0, we would not expect to have E(g1jtujt| xjt) = 0 because of the correlation between g1jt and ujt. It follows that xjt is no longer a valid instrument on its own. Heckman and Vytlacil (1998) recommend replacing the endogenous variable in a CRC model—or the export share in our case—with its predicted value. In the next section, we will estimate the export share with a Type-II Tobit model, or Heckman procedure, using the exogenous variables Zjt that include xjt. Let us therefore rewrite the functions gijt using their expected values as, gijt = E(gijt| Zjt) + vijt with E(vijt| Zjt) = 0, i = 1, 2, 3. We substitute these relations into Eq. (4.18) and simplify to obtain:

114 

M. YU

R jt = b 0Cd + b1 I jt + E ( g1 jt |Z jt ) I jt + E ( g2 jt |Z jt ) Cd + E ( g3 jt |Z jt ) + w jt ,



(4.19)

where the error term is wjt = v1jtI(xjt) + v2jtCd + v3jt−. All the terms appearing within this error have zero expected value conditional on Zjt, so that wjt is conditionally uncorrelated with these instruments and they can be used for estimation.9 The final challenge is to deal with the nonlinear form of the functions gi(ηejt), as seen from the credit constraints in Eq.(4.12). Estimating Eq. (4.15) as a nonlinear structural equation in the presence of endogenous explanatory variables, as well as a first-stage Heckman procedure, is computationally burdensome. Accordingly, we simplify the estimation by taking certain approximations to the functions gi(ηejt), as described in the remainder of this section. We will simplify the functions gi, i  =  1, 2, 3, in different ways. Substituting from Eq. (4.12), we express g1 as g1 (hejt ) = -

idhejt (t e - t d ) s ´ (4.20) s -1 ö s - 1ù æ s -1 é i + i + + dt 1 d t h t h ( ) d ejt e ejt ê sq úû çè d sq ÷ø ë

(

)

We take into account the nonlinearity of g1(ηejt) in the estimation by using a second-order Taylor series approximation around the point ææ ö ηejt = 0: ç ç id (t e - t d ) ÷ 1 s g1 (hejt )  çç ÷h o - 1 ÷ ejt s -1 ö çç s -1 æ + i dt ç idt d + oq ÷ ç çè d oq ø÷ è øè 2 ö æ ö ç id (t e - t d ) ÷ 2 ÷ 2 -ç ÷ hejt ÷ º b 2hejt + b 3hejt 1 s ÷ çç idt + ÷÷ d ÷ oq ø è ø 9  Note that the troublesome term v1jtujt appears twice in wjt after the substitutions are made, but with opposite sign, so it cancels out. That occurs because, unlike Heckman and Vytlacil (1998), we start with an exact theoretical relation in (4est) and then add the errors. The term analogous to v1jtujt did not vanish in Heckman and Vytlacil, so they had to make a conditional homoskedasticity assumption on it to ensure that it would not bias the estimation. That additional assumption is not needed here.

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

115

From this definition of the coefficients β2 and β3, it follows that we can obtain an exact value for the function g1 in Eq. (4.20) as g1 (hejt ) =

b 22 b3

æ ö 1 ç ÷. ç 1 - é b / ( b h )ù ÷ 2 3 ejt ûø è ë

(4.21)

To be consistent with our model, we should find that β2  0. That sign pattern will be enough to ensure that g1(ηejt)  0 from Eq. (4.21), so that exporters face an additional credit constraint. In addition, we can use formula (4.21) to check that ∣g1(ηejt) ∣   0. When we estimate Eq. (4.29) over the entire 2000–2008 sample (not reported), we lose significance of the key coefficient bˆ2 on the interaction of the interest payments and the fitted export share. Likewise, the coefficients bˆ7 and bˆ8 on the interactions of collateral with fitted export shares are also insignificant. One reason for this may be that the last year of our sample (2008) has preliminary data.29 Accordingly, for the remainder of the chapter we focus for the earlier years 2000–2006, during which time 28  The introduction of the success rate of projects ρ, and the default rate (1 − ρ) leads to a slightly different definition of the credit constraints Fe and Fd . But the definitions of the coefficients in Eqs. (4.16) and (4.17) still hold: see Feenstra et al. (2011) for details. 29  As explained in note 19, the data for 2008 are a trial version, so that TFP cannot be computed in the same manner as for earlier years.

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

129

Table 4.4  2SLS estimates by sea and non-sea shipments for Chinese firms, 2000–2006 Type of firm

Matched Chinese firms All matched firms

Regressand: Firm’s revenue (1) 77.78*** (52.20) Interest payment × Fitted −252.1*** export share (β2) (−6.78) Interest payment × Fitted 281.8*** square of export share (β3) (6.74) Fitted export share (β4) 38,538*** (5.95) Export indicator (β5) 859.9*** (2.77) Tangible asset ratio (β6) 16,511*** (9.69) Tangible asset −44,435*** ratio × Fitted export share (−6.39) (β7) Tangible asset 2723 ratio × Fitted square of (1.56) export share (β8) Intangible asset indicator −207.9 (−0.73) Mean of (positive) export 0.446 share hem Estimated value of g1 hem −75.06 Year fixed effects Yes Industry fixed effects Yes Number of observations 536,064

Interact with sea dummy

Interact with non-sea dummy

(2)

(3)

Interest payment (β1)

( )

−335.0*** (−2.52) 432.9*** (2.11) 55,279*** (7.34)

78.12*** (49.44)

−166.0*** (−4.51) 93.76* (2.08) 58,972*** (7.96)

899.1*** (2.63) 17,380*** (9.61) −61,889*** (−7.81) 1909 (0.99) −286.2*** (−0.97) 0.481 −99.45

0.441

Yes Yes 536,064

−58.70

Notes: T-values shown in parentheses are obtained using bootstrapped standard errors, corrected for clustering at the firm level. Significant at *10%, **5%, and ***1%. We use firm-level data for 2000–2006 and match with customs transaction-level trade data. The regression reported in columns (2) and (3) includes Interest Payment × Fitted Export Share × Sea Indicator (and Non-Sea Indicator); Interest Payment × Fitted Square of Export Share × Sea Indicator (and Non-Sea Indicator); and Export Share × Sea Indicator (and Non-Sea Indicator) interactions. The Sea Indicator is defined as 1 if the share of the firm’s exports directly by sea relative to its total exports is higher than 50% and 0 otherwise. The Non-Sea dummy is defined as (1—Sea). The instruments used extend those described in Table 4.2 by interacting with Sea and Non-Sea. Industry fixed effects at the one-digit level are included in all estimates while Hong Kong/Taiwan/Macao m m firms are excluded. The estimated values of g1 he are obtained by inserting the mean ( he ) of the fitted export share into Eq. (4.21). In columns (2) and (3), the interactions of fitted export share and one-­ digit industry indicators and the interactions of fitted export share and year indicators are included

( )

130 

M. YU

we can conveniently merge with Chinese firm-level trade data as needed in the rest of Table 4.4. Thus, column (1) of Table 4.4 reports the 2SLS estimates with collateral over the 2000–2006 sample, using the sample of matched firms in our earlier data set and the firm-level trade data. The sample is reduced to 536,064 observations due to the omitted years 2007–2008 and this matching of firms.30 We find that all of the results in column (1) are consistent with our theoretical predictions. Firms with more collateral, as measured by tangible assets ratio, have higher revenue, bˆ6 > 0 . When interacting the tangible asset ratio with export share, the tangible assets ratio raises revenue less for firms with greater export share, bˆ7 < 0 . The economic magnitudes for the key coefficients (β1 to β3) are also consistent with our theoretical predictions, though we now find that g1 hem = 75.1 is only slightly below bˆ1 = 77.8 .

( )

4.5  Exports by Mode of Transport As a second extension, we consider breaking up exports into their mode of transport, as done by Amiti and Weinstein (2011). Our theory suggests that exporters are more constrained than domestic firms due to the longer time needed for export shipments. In reality, firms would have many types of shipments: by air, sea, truck, and their combination. Usually sea shipments are the slowest and have the longest time-lag to receive payment. It is reasonable to expect that if a firm relies more on sea shipments, then it would face more stringent credit constraints. To examine whether the credit constraint is more stringent as the time to ship for exporters is lengthened, we generate an indicator, sea, which is defined as one if the share of firm’s exports directly by sea relative to its total exports is higher than 50% and zero otherwise. Analogously, we introduce another indicator, non-sea, which equals (1—Sea).31 We then run a single regression, reported in columns (2)–(3), in which we interact 30  In addition, in Table 4.4 we exclude the Hong Kong/Macao/Taiwan-invested firms, since shipping by sea for those firms may involve only very short distances. Those firms are included again in Table 4.5. 31  Our estimation results are essentially unchanged if we take other proportion of sea shipment such as 75%, 90%, or 95%, to form the sea indicator. We have found, however, that if we try to distinguish air shipments as a separate category, then those results are not robust.

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

131

interest payments times the fitted export share and share squared with the sea and non-sea indicators, respectively. It turns out all the key coefficients are statistically significant and of desirable signs as predicted by our model. Turning to the economic magnitudes for each key variable, the estimated coefficients bˆ2 and bˆ3 for sea estimates in column (2) are much higher than their counterparts for non-sea estimates in column (3). Accordingly, the estimated credit constraint for firms that heavily rely on sea shipment is g1 hem = -99.5 , which is 70% larger than that obtained for non-sea transport mode, g1 hem = -58.7 . These findings are strongly consistent with our hypothesis that exporters are more credit constrained due to the longer time needed for sea shipments.

( )

( )

4.6  Incomplete Information So far, we have seen evidence that the credit constraint is more stringent as a firm’s export share grows and as the time to ship for exports is lengthened. Still, it is possible that the extent of incomplete information could be worse in some sectors than in others. In our theory, a reduction in the Pareto parameter θ leads to an increase in the dispersion of firms’ productivity, and corresponds to tighter credit constraints in (domestic constraint). To test this prediction, we make use of TFP2, which governs productivity levels that are known by the firms, but not observed by the bank. We compute its variance across firms within an industry, and then rank all the sectors by this variance, obtaining different percentiles to split the sample for estimation.32 Table 4.5 reports the 2SLS estimates with different percentiles of the variance of productivity. The dispersion of measured variance lies in the range between 0.376 and 4.77. We then present estimation results using four different ranges (all, >10th, and >25th percentile) to examine the role of credit constraints on firm revenue in successively higher variance industries. We find that, again, all the structural coefficients have the anticipated signs and magnitudes. By taking the mean of fitted export share in each column, we see that the measured credit constraint g1 hem increases monotonically with the rise of sectoral variance of firm productivity, consistent with the idea that more incomplete information leads to tighter

( )

 See the last column of Appendix Table A1.

32

132 

M. YU

Table 4.5  2SLS estimates with measures of sectoral productivity dispersion, 2000–2006 Regressand: Firm’s revenue Percentile of sectoral variance of TFP2

(1) AII

(2) >10th

(3) >25th

Interest payment (β1)

82.60*** (31.41) −144.8* (−1.85) 151.4 (1.37) 22,309*** (2.76) 574.9*** (2.00) 13,540*** (8.09) −22,877*** (−2.94) −362.0 (−0.17) 1083*** (3.08) >0.367 0.487 −18.73 Yes Yes 604,154

85.55*** (32.90) −200.1*** (−2.79) 219.7*** (2.09) 17,517* (1.83) 732.7** (2.03) 14,166*** (7.71) −18,396** (−2.00) 132.8 (0.06) −879.8*** (−2.26) >0.567 0.399 −61.71 Yes Yes 542,893

87.89*** (33.29) −293.8*** (−4.06) 399.9*** (3.66) 7126 (0.68) 878.4** (2.22) 13,452*** (6.97) −11,968 (−1.20) 302.8 (0.13) −961.8*** (−2.33) >0.670 0.466 −80.62 Yes Yes 450,599

Interest payment × Fitted export share (β2) Interest payment × Fitted square of export share(β3) Fitted export share (β4) Export indicator (β5) Tangible asset ratio (β6) Tangible asset ratio × Fitted export share (β7) Tangible asset ratio × Fitted square of export share (β8) Intangible asset indicator Cutoffs of sectoral variance of TFP2 Mean of (positive) export share hem Estimated value of g1 hem Year fixed effects Industry fixed effects Observations

( )

Notes: T-values shown in parentheses are obtained using bootstrapped standard errors, corrected for clustering at the firm level. Significant at *10%, **5%, and ***1%. The sample is the same as in Table 4.4, but now including Hong Kong/Taiwan/Macao firms. To measure the extent of incomplete information in each sector, we take the variance of log TFP2 across firms within an industry, then rank the CIC two-digit industries by the variance of productivity, while choosing those percentages as cutoffs to run the regresm u sions. The estimated values of g1 he are obtained by inserting the mean ( he ) of the fitted export share into Eq. (4.21). Industry fixed effects at the two-digit CIC level are included

( )

credit constraints. Moreover, all the estimated credit constraints obtained in each regression are less in absolute value than the coefficients of interest payment themselves, showing that our estimates fit with our model predictions.

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

133

5   Conclusions In this chapter, we have asked why firms will face credit constraints on their domestic sales and exports. We rely on the idea that firms must obtain working capital prior to production and that their productivity is private information. From the revelation principle, the bank can do no better than to offer loan and interest schedule that lead the firms to truthfully reveal this information. We argue that such incentive-compatible schedules will lead to credit constraints on the firms. The reason for this is that a firm that is not credit constrained would suffer only a second-order loss in profits by producing slightly less and borrowing less, but would have a first-order reduction in interest payments. Thus, such a firm would never truthfully reveal its productivity. We rely on a key reason why export sales differ from domestic sales: a longer time-lag in exports between production and sales (Berman et al. 2012b). This time-lag leads the bank to impose a more stringent credit constraint on exporters, for both their exports and domestic sales, than on purely domestic firms. The credit constraint reduces both the intensive margin and the extensive margin of exports. In our estimation we find that the credit constraint becomes tighter as a firm’s export share grows, as the time to ship for exports is lengthened, and as there is greater dispersion of firms’ productivities reflecting more incomplete information. Our theoretical result that the exports and domestic sales of an exporting firm should face the same credit constraint corresponds most closely to the empirical finding of Behrens et al. (2010) for Belgium, who show that financial variables impact both types of sales equally within a firm. This contrasts to the empirical findings of Amiti and Weinstein (2011) for Japan, however, who show that the health of the main bank has a five-­ times greater impact on firm-level exports than domestic sales. One reason for this difference is that Amiti and Weinstein (2011) are arguably capturing the trade finance activities of these banks, targeted specifically at exports, whereas our model and empirical work deals with working-capital loans in general. One limitation of our model is that it is static, whereas other theoretical literature focuses on the dynamic characteristics of credit constraints. Clementi and Hopenhayn (2006) characterize incentive-compatible credit constraints in a dynamic model, and show how such constraints affect firm’s growth and survival. In this setting, a firm’s credit constraint is relaxed when it increases its cash flow. Gross and Verani (2012) show how

134 

M. YU

the firm revenue function used in Clementi and Hopenhayn (2006) can arise from a Melitz-style model, and drawing on Verani (2011), solve for the dynamics of domestic and exporting firms. None of these papers, however, introduce the distinctions between domestic firms and exporters—in the time-lag of shipments—that we use here. We anticipate that our results would apply in some form to these dynamic models, too, but that is beyond the scope of this chapter.

References Ahn, J., Khandelwal, A.  K. & Wei, S.  J., 2011, “The Role of Intermediaries in Facilitating Trade”, Journal of International Economics, Vol. 84(1), pp. 73–85. Amiti, M., and D.  E. Weinstein (2011) “Exports and Financial Shocks”, The Quarterly Journal of Economics 126(4), 1841–1877. Beck, Thorsten (2002), “Financial development and international trade: is there a link?” Journal of International Economics 57, pp. 107–131. Behrens, Kristian, Gregory Corcos and Giordano Mion (2010), “Trade Crisis? What Trade Crisis?” NBB Working Paper No. 195. Berman, N., Martin, P. & Mayer, T., 2012a, “How do Different Exporters React to Exchange Rate Changes?”, Quarterly Journal of Economics 127(1), pp. 437–492. Berman, N., De Sousa, J., Martin, P. and Mayer, T., 2012b, “Time to Ship during Financial Crises,” NBER working paper no. 18274. Bernard, Andrew, Bradford Jensen, Stephen Redding, and Peter Schott (2007), “Firms in International Trade,” Journal of Economic Perspectives, 21(3), 105–130. Bolton, Partrick and David S. Scharfstein (1990), “A Theory of Predation Based on Agency Problems in Financial Contracting,” American Economic Review 80 (1), pp. 93–106. Brandt, L., Van Biesebroeck, J. & Zhang, Y., 2012, “Creative Accounting or Creative Destruction? Firm-level Productivity Growth in Chinese Manufacturing”, Journal of Development Economics, Vol. 97(2), pp. 339–351. Buch, Claudia, Iris Kesternich, Alexander Lipponer, and Monika Schnitzer (2008), “Real versus Financial Barriers to Multinational Activity,” mimeo, University of Tuebingen. Cai, Hongbin and Qiao Liu (2009), “Competition and Corporate Tax Avoidance: Evidence from Chinese Industrial Firms,” Economic Journal 119, pp.764–795. Chaney, Thomas (2005), “Liquidity Constrained Exporters”, mimeo, University of Chicago.

4  EXPORTS AND CREDIT CONSTRAINTS UNDER INCOMPLETE INFORMATION 

135

Chor, Davin and Kalina Manova (2012) “Off the Cliff and Back? Credit Conditions and International Trade during the Global Financial Crisis,” Journal of International Economics 87(1), pp. 117–133. Claessens, Stijn and Tzioumis, Konstantinos (2006), “Measuring Firms’ Access to Finance,” mimeo, World Bank and Brooking conference paper. Clementi, Gian Luca and Hugo A. Hopenhayn (2006), “A Theory of Financing Constraints and Firm Dynamics,” Quarterly Journal of Economics 121(1), pp. 229–265. Dai, M., M. Maitra and M. Yu (2012) Unexceptional Exporter Performance in China? The Role of Processing Trade, Working Paper, Peking University. Egger, Peter and Christian Keuschnigg (2011), “Access to Credit and Comparative Advantage,” ETH Zuerich and University of St. Gallen. Feenstra, Robert and Gordon Hanson (2005), Ownership and Control in Outsourcing to China: Estimating the Property-Rights Theory of the Firm, Quarterly Journal of Economics 120(2), pp. 729–762. Feenstra, Robert, Zhiyuan Li and Miaojie Yu (2011), “Exports and Credit Constraints under Incomplete Information: Theory and Evidence from China,” NBER Working Paper 16940. Greenaway, David, Alessandra Guariglia, and Richard Kneller (2007), “Financial Factors and Exporting Decisions,” Journal of International Economics 73, pp. 377–395. Gross, Till and Stéphane Verani (2012), “Financing Constraints, Firm Dynamics, and International Trade,” Finance and Economics Discussion Series 2012-68. Board of Governors of the Federal Reserve System. Harrison, Ann E. and Margaret S.  McMillan (2003), “Does Direct Foreign Investment Affect Domestic Credit constraints?” Journal of International Economics 61, 73–100. Heckman, James and Edward Vytlacil (1998) “Instrumental Variables Methods for the Correlated Random Coefficient Model: Estimating the Average Rates of Return to Schooling when the Return is Correlated with Schooling,” The Journal of Human Resources, 33(4), Autumn, 974–987. Héricourt, Jérôme and Sandra Poncet (2009), “FDI and Credit Constraints: Firm Level Evidence in China”, Economic Systems 33 (1), pp. 1–21. Kletzer, Kenneth and Pranab Bardhan (1987), “Credit Markets and Patterns of International Trade”, Journal of Development Economics 27 (1–2), pp. 57–70. Levchenko, Andrei A., Logan T. Lewis and Linda L. Tesar (2010), “The Collapse of International Trade During the 2008–2009 Crisis: In Search of the Smoking Gun,” NBER Working Papers 16006. Levinsohn, James and Amil Petrin (2003), “Estimating Production Functions Using Inputs to Control for Unobservable,” Review of Economic Studies 70(2), pp. 317–341.

136 

M. YU

Lu, Dan (2011), “Exceptional Exporter Performance? Evidence from Chinese Manufacturing Firms,” mimeo, University of Chicago. Manova, Kalina (2012), “Credit Constraints, Heterogeneous Firms and International Trade,” Review of Economic Studies, forthcoming. Manova, Kalina, Shang-Jin Wei, and Zhiwei Zhang (2011), “Firm Exports and Multinational Activity under Credit Constraints,” NBER working paper no. 16905. Matsuyama, Kiminori (2005), “Credit Market Imperfections and Patterns of International Trade and Capital Flows,” Journal of the European Economic Association, 3(2–3), 714–723. Melitz, M. J., 2003, “The Impact of Trade on Intra-industry Reallocations and Aggregate Industry Productivity”, Econometrica, Vol. 71(6), pp. 1695–1725. Muûls, Mirabelle (2008), “Exporters and Credit Constraints: A firm-level Approach,” National Bank of Belgium working paper number 139. Olley S. and A.  Pakes, 1996, The Dynamics of Productivity in the Telecommunications Equipment Industry, Econometrica, 64(6), pp. 1263–97. Poncet, Sandra, Walter Steingress and Hylke Vandenbussche (2009), “Credit Allocation in China: Firm-Level Evidence”, MET 17(2), pp. 3–7. Qiu, Larry D. (1999), “Credit Rationing and Patterns of New Product Trade,” Journal of Economic Integration, 14(1), pp. 75–95. Verani, Stéphane (2011), “Aggregate Consequences of Firm-Level Financing Constraints,” Board of Governors of the Federal Reserve System. Yu, Miaojie (2011), “Processing Trade, Firm Productivity, and Tariff Reductions: Evidence from Chinese Products,” CCER Working Paper, No. E201006, Peking University.

CHAPTER 5

Exchange Rate Movements and Exporter Profitability

This chapter examines how RMB appreciation affects profitabilities of exporting firms. Based on highly disaggregated firm-level data from 2001 to 2007, we take RMB appreciation during 2005–2007 as a natural experiment and adopt a difference-in-difference method for exporters and non-­ exporters. Compared to those of non-exporters, returns on equity in exporting firms dropped by about 5% after RMB appreciation. The more dependence firms have on exporting, the more declines there are in their profitabilities. RMB appreciation raised the relative price of Chinese products, which in turn decreased the capacities of exporters to generate sales.

1   Introduction On July 21, 2005, China began to implement the floating exchange rate system, based on market supply and demand and referring to a basket of currencies, then RMB appreciated by 2.1%. By July 2008, yuan–dollar exchange rate had increased by 21%. This arouses broad concerns that RMB appreciation may have negative effects on exporting firms. According to one report released by China Customs, significant declines had taken place in exports of SMEs in Zhejiang Province. Many of these ­enterprises

This chapter is coauthored with Zhonghua Liang and originally published in China Economic Journal, 2014, 7(2), pp. 214–221. © The Author(s) 2021 M. Yu, Exchange Rate, Credit Constraints and China’s International Trade, https://doi.org/10.1007/978-981-15-7522-8_5

137

138 

M. YU

attributed the underperformance to RMB appreciation. Another report revealed that with 1% RMB appreciation the profit margins in cotton, wool, and clothing industries would drop by 3.19%, 2.27%, and 6.18%, respectively.1 Nevertheless, other people argue that RMB appreciation would reduce importing costs and provide more access to capital goods, intermediate goods, and high technologies. As has been reported, 5% RMB appreciation would result in 1.1 billion cost saving in paper-­making industries. Considering the weakening external demand and rising labor costs, it is meaningful to explore whether RMB appreciation has negative or positive effects on exporters’ profitabilities. This chapter intends to shed some light on this topic. Extensive literature has investigated the effects of exchange rate movements on exporting firms. However, controversial conclusions have been reached. Theoretical researches like Marshall (1923) and Lerner (1944) argue that currency appreciation and exports are negatively correlated. Large numbers of empirical supports are provided (Berman et al. 2012b; Das et al. 2007; Greenaway et al. 2007a; Yu 2012). Nevertheless, there is also extensive empirical literature implying insensitivity of exports to exchange rate (Campa 2004; Hooper et al. 1998; Dell’Ariccia 1999). The inconsistent findings are due to various data and methods that different researches employ. Almost all of the existing literature uses macrodata at country, province, or industry levels. Much more factors might affect macro exports. Therefore, prior work is likely to suffer from problems of endogeneities. Different from existing literature, this chapter uses China’s firmlevel data between 2001 and 2007 in manufacturing sectors to investigate the effects of RMB appreciation on profitability. We implement a difference-­in-­difference (DID) method. After 11 years of strictly pegging the RMB to the US dollar at an exchange rate of 8.28, on July 21, 2005, the People’s Bank of China (PBOC) announced a revaluation of the currency and a reform of the exchange rate regime. RMB entered a new period of gradual appreciation. As the exchange rate in China is highly regulated by PBOC, we can take RMB appreciation during 2005–2007 as an exogenous natural experiment. We categorize firms into exporters and non-­exporters. Considering that sales of exporters are more dependent on foreign trade, they are affected directly by exchange rate movements and can be taken as treatment group. We 1  See http://www.mofcom.gov.cn/ar ticle/i/dxfw/gzzd/201308/201308002 38560.shtml.

5  EXCHANGE RATE MOVEMENTS AND EXPORTER PROFITABILITY 

139

take non-exporters as control group. By using DID, we compare the growth of profitabilities for treatment group to the growth of profitabilities for control group following RMB appreciation. This method isolates the influence of exchange rate movements from other factors affecting firm-level profitabilities more broadly. We find out RMB appreciation decreased the profitabilities of exporting firms. Following the 2005–2007 RMB appreciation, return on equity (ROE) of exporters (treatment group) is found to drop by about 5% compared to that of non-exporters. According to DuPont Analysis, ROE can be decomposed into asset turnover, sales margin, and leverage ratio. We also implement DID method with these three factors as outcome variables and find that ROE declines of exporters were mainly caused by decreasing asset turnovers. The more the dependence of firms’ revenues on exporting, the more are the declines in their profitabilities after the exchange rate movement. RMB appreciation raised the relative price of Chinese products, which in turn decreased the capacities of exporters to generate sales. After some empirical tests, our findings remain robust. This chapter contributes to the literature in at least two important ways. First, it enriches the understanding of the economic growth of China, the second largest economy in the world. On the one hand, with weakening external demand and rising labor costs, many Chinese manufacturing firms have run into trouble. Many economists in China and abroad are concerned that RMB appreciation may reduce price competiveness of “Made in China.” It is meaningful to investigate the effects of exchange rate movements on firms’ behaviors. On the other hand, profitability is the most important factor that affects firms’ entry into and exit from exporting market. Very little literature has dealt with this topic from the perspective of profitability. Second, rarely has firm-level data been used to investigate the effects of exchange rate movements. As far as we know, our chapter is among the first in the literature that attempts to explore this topic with microdata. The prior work is likely to suffer from problems like endogeneities. To avoid these biases, this chapter uses highly disaggregated firm-level production data to perform estimations. Last but not least, we use DID method to test assumptions, which in a large part tackles the endogeneity problem. The DID method allows us to control many unobserved factors affecting the control group and treatment group systematically and symmetrically.

140 

M. YU

The remainder of the chapter proceeds as follows. Section 2 introduces the econometric method. Section 3 describes the data sets used in our empirical analysis. Section 4 presents the empirical results and robustness checks. Finally, Sect. 5 concludes.

2   Model Specification We will use DID methods to implement empirical tests. The exchange rate in China is highly regulated by PBOC. As we can see from Fig. 5.1, before 2005, the nominal exchange rate of yuan against US dollar remained constant. It rose gradually after the reform in 2005. The real exchange rate of RMB had been decreasing since 2001 and began to increase after 2005. RMB appreciation during 2005–2007 can be taken as a natural experiment. We divide the whole sample into two periods (2001–2004 and 2005–2007) and categorize the firms into exporters and non-exporters. We consider exporters as “treatment group” and non-exporters as “control group.” Using DuPont Analysis, we decompose ROE (Net_Income/Equity) into asset turnover (Sales/Total_Asset), sales margin (Net_Income/ Sales), and leverage ratio (Total_Asset/Equity). ROE is useful for comparing the profitabilities of firms. It measures a firm’s profitability by revealing how much profit a firm generates with the money shareholders

Fig. 5.1  Import tariffs of Chinese manufacturing firms. (Source: International Monetary Fund E-library)

5  EXCHANGE RATE MOVEMENTS AND EXPORTER PROFITABILITY 

141

have invested. Asset Turnover is an indicator of the efficiency with which a firm is deploying its assets. It measures the amount of sales or revenues generated per dollar of assets. Sales margin measures how much out of every dollar of sales a firm actually keeps in earnings. A higher profit margin indicates a more profitable firm that has better control over its costs compared to its competitors. Leverage ratio is used to calculate the financial leverage of a firm to get an idea of the firm’s methods of financing or to measure its ability to meet financial obligations. We employ DID methods to test how these indicators changed for exporters and non-exporters around the RMB appreciation. A multi-period DID empirical equation is specified as follows: y ft = β 0 + β1 Post t × Exporterf + α f + λt +  ft



(5.1)

We let f denote firms and t denote years. yft is the outcome variable, here referring to ROE, asset turnover, sales margin, or leverage ratio. Postt is a dummy variable indicating whether or not year t is after the cutoff year, that is, 2005. Exporterf is an indicator of firms that engage in exporting. They are expected to have a high level of exposure to exchange rate volatility. The decisions whether to export and how much to export might be correlated with firm-level profitability. We identify firm f as an exporter if it took up exporting trade for the period 2001–2004, during which the RMB exchange rate remained nearly constant (see Fig. 5.1). αf and λt are firm-specific and year-specific fixed effects. ϵft is an idiosyncratic error term. If RMB appreciation raised profitability of exporters compared to non-­ exporters, β1