Gender-Specific Life Expectancy in Europe 1850–2010 3515112588, 9783515112581

Citation preview

Gender-Specific Life Expectancy in Europe 1850–2010 e

Edited by Martin Dinges and Andreas Weigl MedGG-Beiheft 58

Franz Steiner Verlag Stuttgart

Gender-Specific Life Expectancy in Europe 1850–2010

Medizin, Gesellschaft und Geschichte Jahrbuch des Instituts für Geschichte der Medizin der Robert Bosch Stiftung herausgegeben von Robert Jütte Beiheft 58

Gender-Specific Life Expectancy in Europe 1850–2010 Edited by Martin Dinges / Andreas Weigl

Franz Steiner Verlag Stuttgart 2016

Gedruckt mit freundlicher Unterstützung der Robert Bosch Stiftung GmbH

Coverabbildung: Gruppenbild rauchende Soldaten in Esnes, Frankreich, 1916. © Privatarchiv Jan Selmer – www.zeitensprung.de

Bibliografische Information der Deutschen Nationalbibliothek: Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über abrufbar. Dieses Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist unzulässig und strafbar. © Franz Steiner Verlag, Stuttgart 2016 Druck: Laupp & Göbel GmbH, Nehren Gedruckt auf säurefreiem, alterungsbeständigem Papier. Printed in Germany ISBN 978-3-515-11258-1 (Print) ISBN 978-3-515-11275-8 (E-Book)

Contents Introduction Andreas Weigl The drifting apart of gender-specific life expectancies in Europe 1850–2010 7 Germany and Austria Marc Luy The impact of biological factors on sex differences in life expectancy: insights from a natural experiment.....................................................................17 Andreas Weigl The gender gap in life expectancy in Austria and the change in the working environment (c. 1900–1950) ..................................................... 47 Johannes Klotz The reduction of the gender gap in life expectancy in Austria since the 1980s: an educational phenomenon? ......................................................... 65 Western Europe, Belgium and the Netherlands Alice Reid, Eilidh Garrett, Chris Dibben, Lee Williamson Gender specific mortality trends over the epidemiological transition: a view from the British mainland 1850–2000 .................................................. 73 Patrick Deboosere The evolution of the gender gap in life expectancy in Belgium and the smoking epidemic (1841 to 2013) ........................................................ 89 Frans van Poppel, Fanny Janssen The mortality gender gap in the Netherlands 1850–2000 ............................ 111 Sweden Sam Willner Gender-specific life expectancies in Sweden 1810–1980 ................................131 Örjan Hemström Changes in the gender gap in life expectancy in Sweden: A cohort analysis with the most recent trends ............................................... 149

6

Contents

Switzerland Raymond Kohli The evolution of the gender gap in Switzerland since 1876.......................... 167 Christoph Junker How do causes of death influence the evolution of the gender gap in life expectancies in Switzerland? ..................................................................177 Conclusions Martin Dinges, Andreas Weigl Gender gap similarities and differences in Europe ........................................ 187 Authors ................................................................................................................217

The drifting apart of gender-specific life expectancies in Europe 1850–2010 Some introductory remarks Andreas Weigl At least as far as the last two or three decades are concerned there can be no doubt that the gender gap in life expectancy has become a topical research issue in a number of separate scientific disciplines, both in the natural and the social sciences. There is an obviously increasing demand for gender studies on health behaviour and its consequences on health politicies on a national, European and international level. The conclusive common ground is nevertheless limited, because these studies are based on two quite contradictory approaches: the biological on the one hand and the behavioural and environmental on the other hand. Due to the fact that methods and “scientific cultures” of these approaches differ widely it is not surprising that general conclusions are rare and often more or less simple enumerations of factors from both sides. But there are some exceptions. Some authors have attempted to calculate crude weightings of biological and “cultural” influences1, while others argued more cautiously by stressing the need for more “intersectional” approaches in studies of both health related behaviour and mortality.2 In 2005 a convincing model of gender health inequality was elaborated by Birgit Babitsch, who teaches New Public Health at the University of Osnabrück. The three level model (macro/meso/micro) integrated both behavioural and biological factors, though the focus is predominantly on the former.3 Unfortunately for the question of the gender gap in life expectancy, the results of Babitsch’s study based on German data is only to some extent helpful, because understanding morbidity differentials by either sex or gender does not necessarily lead to a full understanding of differentials in mortality. In general one receives the impression that there is still a wide methodological gap to overcome. It can be shown easily that even scientists who gained degrees in the humanities and in the natural sciences tend to stress or neglect either the biological or the behavioural factors.4 Therefore there are still a lot of open questions to be answered and multidisciplinarity remains “wishful thinking”. This does not only apply to research in the natural versus social sciences, but also to the problem of social and historical disciplines. The social sciences in particular have given little attention to the historical roots of gender-specific lifestyles and the long-term transformation of gender roles, while – since the “cultural turn” in the study of history 1 2 3 4

Johansson, 1991; Luy 2002a. Waldron, 2000, p. 177. Babitsch, 2005, pp. 138–141. Austad, 2006.

8

Andreas Weigl

– many historians have largely ignored quantitative research. Fortunately, there is one discipline that is rooted in both fields. Demography has – to a certain extent – managed to consider the historical view as well as include the findings of other social sciences and the natural sciences in its analyses. The late nineteenth and twentieth centuries are of particular interest for historical research into the gender gap in life expectancy in the industrialized world. Because this period is very close to the present it allows the use of historical findings as a means of enhancing recent research into women’s and men’s mortality. In addition, the encouraging availability of data – compared with the early modern period – and, not least, the remarkable development of the gender gap during that period offers the chance to gain a deeper understanding from international comparisons. The comparison of Northern, Western and Central Europe in this volume was chosen for pragmatic reasons because these countries offer a sufficiently wide range of gender gap variations and the actual state of research is sufficient for a relatively differentiated comparison.5 Nevertheless, it would certainly make sense in further research to include eastern and southern Europe as well as non-European industrialized countries in the picture. As mentioned before, specific research in the natural and cultural sciences proves that the gender gap can best be explained against the background of a combination of biological, genetic and behavioural factors. The significance of these factors is, however, disputed. Research in communities with very similar living conditions for men and women provides evidence that the gender gap has only a minor biological component. Following almost identical results from investigation of the general population and that of monasteries, it was rated for adults at one year or one and a half years at the most.6 This figure is higher, however, when we include the influence of biological factors on behavioural roles.7 Genetic-biological factors that influence the gender gap are best proved with babies, since in babies, and to a limited extent also in young children, acquired gender-specific behaviours hardly influence the excess mortality of boys. Although infant mortality decreased substantially in late nineteenth and twentieth century Europe, an advantage persisted for girls in their first year of life. Based on an analysis of infant mortality by major causes of death Ingrid Waldron has shown that there are multiple sex differences in biology and not all of them are a disadvantage for male infants. “The available evidence suggests that males have inherently greater vulnerability for mortality due to perinatal conditions and for total mortality in the neonatal period, but the assumption that males have a pervasive inherent disadvantage is incorrect for some types of congenital anomalies, and is of uncertain validity for

5 6 7

Unfortunately it was not possible to find colleagues to present the French case; but cf. Vallin / Meslé, 1988, chap. 12, pp. 467–505, for a differentiated analysis of the causes of death and their contribution to the gender gap until 1978. Luy, 2002a, p. 424. Ritzmann, 2001, p. 70.

The drifting apart of gender-specific life expectancies in Europe 1850–2010

9

infectious diseases and total mortality in early childhood.” 8 This indicates that, although the weight of infant mortality for the overall gender gap in mortality is almost fading away, further medical and biological studies in this field should not be ignored. Although the attempts to estimate the components of the gender gap are quite valuable to adjust downwards popular estimates in the medical sciences9 I would suggest that we should consider the combination of biological and behavioural factors not only as a simple equation of “a + b = c”. If newborn girls come with a basic gender-specific advantage – and there is much to support this assumption – this advantage can be tapped to a greater or lesser extent during their lives. Furthermore, recent medical research also shows that a person’s basic genetic disposition can be improved or made worse during their lifetime.10 Splitting the causes of the gender gap into behavioural-“social” and “biological” factors is therefore an auxiliary structure that supports a statistically more complex situation. While biological explanations of the gender gap due to the recent rise of genetics seem to me to gain greater importance again in medical research – which would mean that older medical explanatory models are staging a comeback – the social sciences, social medicine included, underscore the part played by gender-specific behaviour patterns and lifestyles. Their findings are corroborated by those of historians, but the interpretations of these findings differ widely. While some see the gender gap in life expectancy as proof of the provocative thesis, formulated by Martin van Creveld, that women are a “privileged sex”11 or that men are “health idiots”12, others, such as Maria Danielson and Gudrun Lindberg, speak of a “new gender paradox”: This new gender paradox implies that women live longer than men despite the fact that they continue to be discriminated against in society.13 The old paradox – dismissed by Danielson and Lindberg – states the alleged greater health impairment of women compared to men.14 From this old paradox it was no big step to DFLE (disability-free life expectancy) and HALE (health-adjusted life expectancy), concepts that basically made the gender gap in industrialized countries in the disability-free part of the life span disappear because, on average, women suffer from chronic disease for longer periods of time. This is a fact documented by a huge number of studies. But the definition of such severe health conditions remains vague and assessment of such values is biased as a result of national statistical traditions. In addition one gets the impression that the interpretation of the results of studies based on these concepts is strongly 8 9 10 11 12 13 14

Waldron, 1998, p. 79. See for instance Klotz, 1998, p. 101, who prorates 60 % of the gender gap to genetic-biological factors. Tammen / Friso / Choi, 2013. Van Creveld, 2003. Dinges, 2005, p. 512. Danielson / Lindberg, 2001, p. 61. Babitsch, 2005, p. 65.

10

Andreas Weigl

influenced by the current political gender agenda in the European Union, which is hardly focused on male disadvantages in public health.15 Introducing such concepts as the DFLE and the HALE may be justified with a view to gender-specific health policies, but it does not really help to explain the gender gap in life expectancy. On the contrary: it raises a new question: Why is it that women survive chronic illness for longer than men? If this question is not explained by gender-specific biological differences alone, it continues to be a field of research in social medicine. But what insights can be expected from a comparative historical view? From the late nineteenth century the widening of the gender gap in life expectancy became a European phenomenon that continued into the 1980s, in eastern European transition countries even beyond the fall of the Iron Curtain in 1989. This widening varied, however, in the various countries and groups of countries, a fact that points to the influence of “exogenous” factors such as the two world wars, but not least also to different economic and social developments with regard to the working world and gender roles. Richard Wilkinson and Kate Pickett have shown in a recent comparative study that within the group of countries with high GDPs per head a more even income distribution and an advanced health care system contribute to a general high level of life expectancy.16 This strengthens the argument that economic well-being is a key factor for a narrowing of the gap, but it reveals that the relationship is more complex. Even if the relatively small gender gap in life expectancy of some Scandinavian countries and the Netherlands needs to be seen in the context of economic vitality and modern societal structures, economic inequality, which was and is low in some North European welfare states, can clearly not adequately explain the development of the gender gap in industrialized countries during the twentieth century, because the last three decades saw a narrowing of the gender gap in life expectancy and a widening of economic inequality, not to mention the fact that beside undisputable discrimination of women in many developing countries the gender gap in favour of the female population was a global phenomenon in the early twenty-first century. I refer in particular to the headline of an editorial in a 2006 issue of the British Medical Journal: Life expectancy: Women are now on top everywhere!17 While many pertinent studies, such as the British Black Report of 198218, show a considerable gradient for life expectancy depending on income and education (as a proxy), the range is much smaller with women than it is with men. Studies from other countries, including Austria, show that the life expectancy of a female labourer was and still is as high or even higher as that of an

15 16 17 18

For a critical discussion see Dinges / Weigl, 2011, p. 198, fn. 20. Wilkinson / Pickett 2009. Barford et al., 2006. Black, 1992.

The drifting apart of gender-specific life expectancies in Europe 1850–2010

11

academic.19 According to Pierre Bourdieu’s capital theory20 this is a strong hint that the uneven distribution of “social capital” must be included in the analysis too. There is much to support the view that women have the better “social capital” due to their traditional gender roles and possibly also due to their biological make-up. As mentioned before, some demographic studies have successfully refuted explanations of the gender gap that are based on purely biological premises. For instance in the long-term comparison of monastic populations (monks and nuns) with the general population, Marc Luy was able to present compelling evidence that – at least for adults – it was mostly the male gender roles and the strains of working life (stress at work, consumption of addictive substances) that enhanced the widening of the gender gap during the twentieth century.21 Luy’s study results therefore refer to a specific field of inequality research, the research into people’s lifestyles. Unfortunately, gender-specific lifestyles have so far not been the preferred subject of sociological research.22 The few existing studies certainly present a wide range of differing gender “health behaviours”, in particular on the male side.23 On the basis of a cluster analysis a Swiss study showed a statistically significant connection, for men, between “somatic culture” and social milieu but not social class.24 In addition, this study stressed the fact that men in general still often need to justify body-sensitive health behaviours.25 This need for justification is clearly fed by the explanation pattern that became popular in the 1950s, which claims that the excess male mortality was the price to be paid for modern (Fordist) work situations and a male-dominated success-oriented working world – the rise of the male manager.26 Making the changes in the working world during the twentieth century alone responsible for the growing gap in life expectancy up to the 1980s would be a questionable step, however. The service societies that emerged in Europe in the last decades of the twentieth century and the parallel exodus of (heavy) industries to developing countries obviously reduced the physical hazards of the working world and this has affected men in particular. Accident statistics and the drop in what is termed “occupational diseases” also support this observation although the male/female-ratio in this respect is quite stable.27 But this says nothing about mental stress which seems to be an influential factor if one considers the prominent mortality due to cardiovascular disease and can19 20 21 22 23 24 25 26 27

Klotz, 2008, p. 150. Bourdieu, 1992, pp. 49–79. Luy, 2003, pp. 647–676. Abel, 1999, pp. 43–61. Robertson, 2007, p. 156; Nettleton / Watson, 1998; Ervø / Johansson, 2003; Schneider 2002. Nideröst, 2007, p. 97, 106. Robertson, 2007, p. 63. Forth, 2008, p. 205; Ehrenreich, 1983, p. 70. For instance in Austria in the time period 1970–2012 the ratio is about 3 to 1. See Statistik Austria, 2014, pp. 84–89.

12

Andreas Weigl

cer. Given the established high shares of high-ranking male managers and full time jobs in general this could explain the still existing higher mortality of men, at least in cardiovascular disease.28 The uneven distribution of time spent on childrearing and housework between female and male employees is still present and diminishing rather slowly.29 The question that remains unanswered is whether there are gender-specific differences in respect of stress-related mortality. For the last three decades the narrowing of the gender gap at least coincides with an interpretation of stress as an equalizer of male and female mortality, seeing that the incidence of “burnout-related diseases” has risen among the highly-qualified female workforce as well. The same is true for the shrinking of the gap in gender-specific consumption of nicotine and alcohol that has been demonstrated by studies for many industrialized countries.30 This brings us to another question that has, in my opinion, still not been sufficiently answered and that relates to the long-term effect of the two world wars. Although some crude estimates are available,31 the effect of chronic illness, war injuries and other morbidities caused by the war is not easy to measure and certainly goes beyond these “direct” consequences. As a recent study based on an evaluation of (West-)German, Austrian and Swiss “popular autobiographies” has shown, even the long term mental consequences of wars are a very complex issue. In her thesis on gender specific discourse and health-related lifestyles Susanne Hoffmann stressed the point that there is evidence of a positive judgement of severe war-injuries, because wounding could bring soldiers out of danger zones, at least as far as surviving soldiers are concerned.32 On the contrary, it is possible that the two wars played a part in retaining and reinforcing images of a “military masculinity” – after World War II transformed in male economic competition in the “battlefield” of the labour market33, at least among the men of age groups that were actively involved in the war or educated during war times, and therefore in further widening the gender gap. New insights will emerge in this volume from the comparison of countries and populations that were, to various degrees, involved in the wars. To summarize: for a long term perspective on the gender gap in life expectancy in the “long” twentieth century the following questions are particularly relevant: 1. Why did the gender gap in life expectancy widen in a time period when sectoral change and automation in manufacturing reduced the disadvantages of the male working population, at least as far as physical strains (in factories, in the construction business) were concerned.

28 29 30 31 32 33

Siegrist, 2010, p. 79. See i. e. McGinnity / Russell, 2008. Johansson 1991, p. 157. Haudidier, 1996; Höhn, 1996 Hoffmann, 2010, pp. 288–299. Hanisch, 2005, pp. 118–121.

The drifting apart of gender-specific life expectancies in Europe 1850–2010

13

2. What are the long-term effects of the two World Wars, since the effect of war and the consumption of addictive substances overlap? A comparison between nations involved in the war and neutral states would allow for a more accurate evaluation of this factor. 3. Considering the partly diametrically opposed developments of female labour force participation, a comparison of the gender gap in West and East Germany from 1949 to 1989 or between countries with a long tradition of high female labour force participation (Sweden) and countries with the opposite tendency (FRG, Switzerland, Austria) would be particularly interesting regarding the influence factor “world of work”. Comparing the state of research in the various countries into sub-disciplines relevant to this question – such as (historical) demography, economic and social history, the history of medicine, social medicine and gender research – might also yield synergies and new insights, especially as far as the reduction of the gender gap in life expectancy in highly developed countries in the last three decades is concerned. What links the demographic case studies collected in this volume is a long term perspective with a focus on the twentieth century. The course of the gender gap over 150 years in the advanced Atlantic economies of North-West Europe has been elaborated by Alice Reid and Chris Dibben in their study on gender specific mortality trends over the epidemiological transition on the British mainland including the “national patterns” of England/Wales and Scotland, and by Frans van Poppel and Fanny Janssen on the Netherlands. Given the differing involvement of these countries this also opens insights into the deteriorating effects of World War I. Based on a smaller set of data these effects were discussed by Andreas Weigl in his article on Austria in the first half of the twentieth century as well, though his main focus is on the changing working environment. The Austrian experience of a reduction of the gender gap in life expectancy since the 1980s is discussed by Johannes Klotz. Klotz stresses the importance of high education for the particular convergence of male and female life expectancy, reinforcing academics as a “vanguard group”. A different methodological approach is presented by Marc Luy. Luy’s comparison of Catholic order members and the overall German (East and West) population in the second half of the twentieth century is embedded in a broad overview of the biological/non biological factors-debate that traces, as he demonstrates, back to the eighteenth century. Furthermore, like several other contributions of this volume, Luy’s study also includes an analysis of non-biological factors like smoking. A comparison of the Benelux experience is another interesting issue. It is no big surprise that Patric Deboosere’s case study on the gender gap in Belgium in the nineteenth and twentieth century shows important similarities as well as differences compared to the Netherlands. The same could be said of the neutral countries Switzerland and Sweden. An overview of the much more fluctuating Swedish trends in the gender gap is given by Sam Willner, while Örjan Hemström (like Johannes Klotz in the case of Austria) discusses the reasons for the narrowing of the gap in the

14

Andreas Weigl

last decades. Although Switzerland, like Sweden, was not involved in both World Wars, the Swiss case is much more “Central European” than “Scandinavian” as the papers of Raymond Kohli in general and Christoph Junker by analysis of cause specific mortality show The questions I have outlined and those raised in the case studies and the editors’ résumé in this volume do, of course, not constitute all the questions that are still open with regard to the gender gap in life expectancy and not all of them should be seen in a long term perspective. But even if one focuses on the most recent changes in the gender gap, social history cannot be ignored. As an analysis based on meta-data of 72 recent studies on the subject reveals, specific male subpopulations are in all probability the core factor of the gender gap in life expectancy in present industrialized countries.34 If this conclusion is correct, public health programs focusing on these specific subpopulations certainly need a historical perspective for a deeper understanding of these risk groups including socialization, representations of masculinity and many more factors. Almost all these factors are in a wider sense embedded in social, economic, cultural history. To conclude: history matters in the long run. Bibliography Abel, T.: Gesundheitsrelevante Lebensstile: Zur Verbindung von handlungs- und strukturtheoretischen. In: Aspekten in der modernen Ungleichheitsforschung. In: Mäder, Christoph; Burton-Jeangros Claudine; Haour-Knipe, Mary (eds.): Gesundheit, Medizin und Gesellschaft. Beiträge zur Soziologie der Gesellschaft, Zurich 1999, pp. 43–61. Austad, Steven N.: Why Women Live Longer Than Men: Sex Differences in Longevity. In: Gender Medicine Vol. 3 No. 2, 2006, pp. 79–92. Babitsch, Birgit: Soziale Ungleichheit, Geschlecht und Gesundheit, Bern 2005. Barford, Anna et al.: Life expectancy: women now on top everywhere. In: British Medical Journal 332 (2006), p. 808. Black, Douglas et al.: The Black Report. In: Inequalities in Health, Harmondsworth 1992, pp. 29–213. Bourdieu, Pierre: Die verborgenen Mechanismen der Macht, Hamburg 1997. Danielson, Maria; Lindberg, Gudrun: Differences between men’s and women’s health: The old and the new gender paradoxon. In: Piroska, Östlin et al. (eds.): Gender Inequalities in Health. A Swedish Perspective, Cambridge (Mass.) 2001, pp. 23–66. Dinges, Martin: Die Gene sind nicht schuld. 200 Jahre Männergesundheit. In: Pflegezeitschrift 8/2005, pp. 508–512. Dinges, Martin; Weigl, Andreas: Männergesundheit als Forschungsthema der Sozial- und Kulturwissenschaften. In: Gesundheit und Geschlecht (Österreichische Zeitschrift für Geschichtswissenschaften 22/2 (2011), pp. 191–199. Dinkel, Rainer (ed.), Sterblichkeitsentwicklung unter besonderer Berücksichtigung des Kohortenansatzes, München 1996. Ervø, Søren; Johansson, Thomas (eds.): Bending Bodies, Moulding Masculinities, Hants/Burlington 2003. Ehrenreich, Barbara: The Hearts of Men: American Dreams and the Flight from Responsibility, New York 1983. 34 Luy / Gast, 2013.

The drifting apart of gender-specific life expectancies in Europe 1850–2010

15

Forth, Christopher E.: Masculinity in the Modern West. Gender, Civilization and the Body, Houndmills; New York 2008. Hanisch, Ernst: Männlichkeiten. Eine andere Geschichte des 20. Jahrhunderts, Vienna-Cologne-Weimar 2005. Haudidier, Benoit: Vergleich der Sterblichkeitsentwicklung in der Bundesrepublik Deutschland und Frankreich 1950–1898. In: Dinkel, 1996, pp. 139–152. Höhn, Charlotte: Kohortensterblichkeit unter besonderer Berücksichtigung der Weltkriege. In: Dinkel, 1996, pp. 45–65. Hoffmann, Susanne: Gesunder Alltag im 20. Jahrhundert? Geschlechterspezifische Diskurse und gesundheitsrelevante Verhaltensstile in deutschsprachigen Ländern (Medizin, Gesellschaft und Geschichte 36), Stuttgart 2010. Klotz, Johannes: Convergence or divergence of educational disparities in mortality and morbidity? The evolution of live expectancy by educational attainment in Austria in 1981– 2006. In: Vienna Yearbook of Population Research 8 (2010), pp. 139–174. Klotz, Thedor: Der frühe Tod des starken Geschlechts. Unterschiede im Gesundheits- und Krankheitszustand von Männern und Frauen, Göttingen 1998. Luy, Marc: Die geschlechtsspezifischen Sterblichkeitsunterschiede – Zeit für eine Zwischenbilanz. In: Zeitschrift für Gerontologische Geriatrie 35 (2002a), pp. 412–429. Luy, Marc: Warum Frauen länger leben. Antworten durch einen Vergleich von Kloster- und Allgemeinbevölkerung, Wiesbaden 2002b. Luy, Marc: Causes of Male Excess Mortality: Insights from Cloistered Populations. In: Population and Development Review 29 (2003), pp. 647–676. Luy Marc; Gast, Katrin: Do Women Live Longer or Do Men Die Earlier? Reflections on the Causes of Sex Differences in Life Expectancy. In: Gerontology November 2013. McGinnity, Frances; Russell, Helen: Gender Inequalities in Time Use. The Distribution of Caring, Housework and Employment Among Women and Men in Ireland, Dublin 2008. Nettleton, Sarah; Watson, Jonathan (eds.): The Body in Everyday Life, London; New York 1998. Nideröst, Sibylle: Männer, Körper und Gesundheit. Somatische Kultur und soziale Milieus bei Männern, Bern 2007. Ritzmann, Iris: Die Frage nach dem “kleinen Unterschied” vor dem Tod und seinen Hintergründen – von der göttlichen Ordnung zur chromosomalen Determination. In: Bulletin der Schweizerischen Gesellschaft für Anthropologie 7(2) (2001), pp. 51–71. Robertson, Steve: Understanding Men and Health. Masculinities, Identity and Well-being, Maidenhead 2007. Schneider, Sven: Lebensstil und Mortalität, Wiesbaden 2002. Siegrist, Johannes: Arbeit, Arbeitslosigkeit und Gesundheit. In: Bardehle, Doris; Stiehler, Matthias (eds.): Erster Deutscher Männergesundheitsbericht. Ein Pilotbericht, Munich 2010, pp. 72–86. Statistik Austria: Demographische Indikatoren 1961–2012. Erweiterte Zeitreihe für Österreich, Vienna 2014. Tammen, Stephanie A.; Friso, Simonetta; Choi, Sang-Woon: Epigenetics: the link between nature and nuture. In: Molecular Aspects of Medicine 34/4 (2013), 753–764. Van Creveld, Martin: Das bevorzugte Geschlecht, Munich 2003. Vallin, Jacques; Meslé, France: Les causes de décès en France de 1925 à 1978, Paris 1988. Waldron, Ingrid: Sex differences in infant and early childhood mortality: major causes of death and possible biological causes. In: United Nations: Too young to die: genes or gender, New York 1998, pp. 64–83. Waldron, Ingrid: Trends in gender differences in mortality: Relationships to changing gender differences in behaviour and other causal factors. In: Annandale, Ellen; Hunt, Kate (eds.): Gender Inequalities in health, Buckingham 2000, pp. 150–181. Wilkinson, Richard; Pickett, Kate: The Spirit Level. Why More Equal Societies Almost Always Do Better, London 2009.

The impact of biological factors on sex differences in life expectancy: insights gained from a natural experiment Marc Luy 1. Introduction That women live longer than men has been known at least since the middle of the eighteenth century when Kersseboom mentioned his observation that the mortality experiences of males and females differ sufficiently to make it worthwhile using separate tables for calculating annuities.1 A few years later, the first sex-differentiating life tables by Struyck and Deparcieux added the corresponding empirical evidence.2 The finding of male excess mortality was confirmed with the introduction of official population statistics in all western societies and has been documented in Sweden from 1751 onwards.3 Until recently, a higher life expectancy at birth for men was known only for some countries in Africa and Asia, mainly due to an excessive female mortality among infants and in early childhood.4 A few years ago, Barford et al. announced in a British Medical Journal editorial entitled “Life expectancy: women now on top everywhere” that females outlive males now even in the poorest countries of the world.5 Men have higher mortality than women not only in terms of overall measures like life expectancy at birth but also – at least in industrialized societies – in all ages and leading causes of death. The mortality differences between women and men remained more or less constant until the end of the nineteenth century and started to increase during the twentieth century. This increase of the sex gap coincided with a rise among men in cardiovascular disease, cancer, and accidents, and a fall in maternal mortality and in causes of death related to pregnancy.6 However, since the beginning of the 1980s the gap between women and men in overall mortality has been slowly narrowing in the western world. In Eastern Europe, the trend reversal set in during the 1990s and only recently reached Japan, the sole laggard in the western world.7 The hypotheses advanced to explain male excess mortality can be divided into two basic categories, those concentrating on the biological factors (factors largely beyond human control which are also called “inherited risks”) and those concentrating on non-biological factors (behavioural, cultural and environmental factors, i. e. factors directly or indirectly influenced by human ac1 2 3 4 5 6 7

Kersseboom (1737), Kersseboom (1740). Deparcieux (1746), Struyck (1740). Tabutin (1978). See e. g. Aden et al. (1997), Langford (1984). Barford et al. (2006). Lopez (1983). Liu et al. (2013).

18

Marc Luy

tion which are also called “acquired risks”). The three following sections summarize the main biological (section 2) and non-biological factors (Section 3) discussed in the literature, as well as the various interactions between them (Section 4). Section 5 includes an overview of the existing attempts to quantify the impact of biological factors and section 6 provides a refined estimation of the biologically caused sex difference in life expectancy based on our data for Catholic order members from Germany and Austria. Section 7 demonstrates the empirical application of this estimate to data from Germany by isolating the impact of biological factors in order to estimate the impact of smoking on sex differences in life expectancy in comparison to other non-biological factors. The article ends with a discussion in which our estimates of the impact of biological factors are compared to those of other studies on sex differences in mortality among Catholic order members. 2. Biological factors Ever since the differences in mortality between the sexes have been known, one has assumed that women outlive men because humans are subject to a general “law of nature” according to which females enjoy a longevity advantage.8 There are several biological differences between the sexes which might be responsible for this natural female survival advantage.9 It is the established view that sex differences in mortality are built upon a “fundamental genetic basis”10 which is supposed to act along three basic axes: – The first relates to the implications of homogametic (XX) and heterogametic sex (XY) in the viability of the human organism.11 In each female cell, one X chromosome is randomly inactivated, protecting women against a double dose of X chromosome expressions as well as against disadvantageous genes on one X chromosome.12 – Another relevant genetic sex difference is seen in relation with the positive correlation between telomere length and length of life, as men exhibit shorter telomeres than women because of faster telomere attrition.13 This line of reasoning is directly related to the “costly growth hypothesis” stating that the costs of growing and the sexual size dimorphism lead to higher mortality in the sex with the larger body size.14 – Most recently, it has been proposed that the mitochondrial genome is optimized for function with the female genome through natural selection as 8 9 10 11 12 13 14

Trivers (1972). For an extensive overview see Seifarth et al. (2012). The summary of hypotheses and findings presented in this Section is based primarily on this reference. Hayflick (1982), p. 248. See e. g. Christensen et al. (2001), Puck / Willard (1998), Smith / Warner (1989). Seifarth et al. (2012). Barrett / Richardson (2011), Stindl (2004). Kalmbach et al. (2005), Owens (2002), Promislow (1992), Samaras et al. (2002).

The impact of biological factors on sex differences in life expectancy

19

humans inherit mitochondria from mothers only.15 As explicated by Seifarth et al.,16 this optimization of mitochondrial functionality in females might result in a survival advantage because mitochondrial dysfunction has been demonstrated to be related to ageing17 as well as to a number of mortality relevant diseases such as cancer18 and cardiovascular disease.19 Upon this genetic basis, the endogenous sex hormones testosterone and oestrogen comprise the second group of biological factors which are likely to affect the mortality of women and men. Again, there are several routes on which hormones are expected to contribute to sex differences in life expectancy: – One important sex-specific hormonal effect concerns the handling of lipids. Men tend to store more fat in the abdominal region, whereas women tend to store more fat in hips, thighs and buttocks. Moreover, women tend to have greater amounts of subcutaneous fat, whereas men are more likely to accumulate visceral adipose tissue which has been implicated in a number of diseases including metabolic syndrome,20 coronary artery disease21 and ischemic heart disease,22 among others.23 These sex differences in body fat distribution are complemented by sex differences in the lipoprotein metabolism. Most importantly, oestrogens have been shown to increase HDL (“good cholesterol”) and lower LDL (“bad cholesterol”) levels, whereas androgens lower HDL concentrations but raise those of LDL.24 Cholesterol is a well-known risk factor for atherosclerosis. Differences between women and men in sex hormone levels therefore lead to a sex differential in lipoprotein metabolism which is supposed to cause sex differences in cardiovascular disease and these, in turn, lead to sex differences in mortality.25 – Other studies indicate that hormonally caused sex differences in “immunocompetence”–i. e. an organism’s all-around ability to avoid the harmful effects of infections – may underlie male excess mortality, supposing that oestrogens are immunity enhancers, whereas androgens and progesterone are natural immunosuppressants.26 However, a study by Fairweather and Cihakova suggests that the female advantage in infection resistance may turn into a disadvantage when an immune response is initiated against host

15 16 17 18 19 20 21 22 23 24 25 26

Tower (2006). Seifarth et al. (2012). Trifunovic / Larsson (2008). Brandon et al. (2006). DiMauro / Andreu (2000). Albu et al. (1997). Nakamura et al. (1994). Matsuzawa et al. (1994). For more details see Seifarth et al. (2012), pp. 393–394. Hazzard (1989), Hazzard / Applebaum-Bowden (1990). See Hazzard (1986), p. 464. Caruso et al. (2013), Hamilton (1948), Hazzard (1989), Moore / Wilson (2002), Owens (2002).

20

Marc Luy

cells.27 Consequently, whereas males are more susceptible to infection, females may be more susceptible to autoimmune disease. Nonetheless, this contrasting susceptibility still favours women because infectious diseases, cancer and cardiovascular disease – all with significant male excess – produce much higher mortality than autoimmune diseases.28 – Another hypothesis based on sex hormones relates sex differences in mortality to sexually dimorphic mechanisms of combatting oxidative stress. The thesis is built on the observation that the combination of the antioxidant properties of oestrogen and associated antioxidant genes in females leads to more favourable handling of oxidative stress and its accumulation over the lifespan.29 – A further hormone based sex dimorphism has been suggested by physiological studies of the hypothalamic-pituitary-adrenal (HPA) axis stress response to psychosocial stress. For instance, Kirschbaum et al. found that the mere prospect of an upcoming psychological stress task produced a cortisol response in men, but not in women.30 This finding has been supported by Dahl et al., who found that male children exhibit a significantly higher cortisol response to corticotropin-releasing hormone ingestion.31 At the organ and cellular levels, several differences in the ability of female cells to deal with cellular perturbations have been documented.32 Although these findings were drawn from studies with rats they point to sex differences in the hormonal and cellular response to stress, which could contribute to the female advantage in life expectancy.33 Nielsen suggests that androgens affect foetal lung development via a mechanism dependent on the presence of androgen receptors within the HPA axis, causing male infants to be at greater risk of respiratory distress syndrome than female infants.34 Male excess mortality has been observed in very different animal species, including house flies,35 rats,36 chimpanzees,37 and many others.38 Among humans higher male mortality rates hold among children39 and even among infants and in the prenatal period, when higher rates cannot be caused by ac-

27 28 29 30 31 32 33 34 35 36 37 38

Fairweather / Cihakova (2009). For more details see Seifarth et al. (2012), pp. 395–396. Behl et al. (1997), Borrás et al. (2003), Proteggente et al. (2002). Kirschbaum et al. (1992). Dahl et al. (1992). Brown et al. (2005), Thorp et al. (2007). For details see Seifarth et al. (2012), pp. 396–397. Nielsen (1985). Rockstein / Lieberman (1959). Asdell et al. (1967). Hill et al. (2001). Ciocco (1940), Clutton-Brock / Isvaran (2007), Comfort (1979), Hamilton (1948), Judge / Carey (2000), Smith (1989). 39 Théré / Rohrbasser (2006).

The impact of biological factors on sex differences in life expectancy

21

quired risks.40 The existence of at least a biological basis for the female survival advantage is therefore undoubted. 3. Non-biological factors The biological sex differences in mortality are complemented by a number of non-biological impacts on the mortality of women and men. Corresponding research argues that society and culture influence men to lead lifestyles that are increasingly detrimental to health and life (in terms of smoking habits, alcohol consumption, reckless driving, diet, exercise, involvement in religious activities, etc.), that men are subjected to greater health risks at work, that environmental factors lead to survival disadvantages for men, and that men are generally more exposed and susceptible to different kinds of social and psychological stress than their female counterparts. Several causation lines can be found in the literature: – Many studies suggest that nicotine consumption is the health behaviour contributing most to increasing male excess mortality.41 In general, men have higher proportions of smokers than women, they start smoking earlier and therefore smoke longer, and they smoke more and stronger cigarettes than women. Smoking also appears to play a considerable role in the currently observable narrowing of the male-female differentials in mortality since the proportion of female smokers has increased greatly in recent decades,42 complemented by declining differences between women and men in alcohol consumption.43 – Estimates for occupational hazards show that risks caused at the workplace are related to a number of severe and fatal diseases.44 Men are exposed more to occupational hazards because more men are employed, and among those who are employed more men than women work in hazardous occupations. Reviewing several studies, Waldron concluded that approximately 95 % of fatal work accidents involve men and that these higher rates for work accidents account for roughly one-fifth of the sex differences in accident fatalities.45 In a multivariate analysis of Swedish labour force participants, Hemström arrived at the same estimate that sex differences in occupational fields account for approximately 20 % of men’s excess mortality from external causes.46 This contribution originated mainly from high job hazard levels in traditional male jobs including, for example, 40 Hazzard (1986), Hazzard / Applebaum-Bowden (1990), Kalben (2000), Lopez (1983), Waldron (1985), Wingard (1982). 41 E. g. McCartney et al. (2011), Payne (2001), Waldron (1986). 42 Luy / Wegner-Siegmundt (2015), Nathanson (1995), Pampel (2002). 43 Martelin et al. (2004), Simons-Morton et al. (2009). 44 E. g. Concha-Barrientos et al. (2004), Leigh (1988), Nurminen / Karjalainen (2001). 45 Waldron (1991). 46 Hemström (1999).

22

Marc Luy

heavy lifting, heavy shaking or vibration, contact with dirt, inadequate ventilation and exposure to gas, vapour or smoke and exposure to chemicals and carcinogens.47 – Social stress is seen as another basic causal factor for male excess mortality.48 In this context, Jenkins introduced the term “type A behaviour”, which is characterized by intensive striving for achievement, competitiveness, easily provoked impatience, time urgency, abruptness of gesture and speech, over-commitment to vocation or profession, and excess of drive and hostility.49 In western societies, type A behaviour is found more frequently among men since it is strongly linked to professional life and social status.50 – Because lifestyles generally differ with the level of social status, male-female differences in mortality could also be affected by the fact that men and women are not equally distributed within various social classes.51 Nathanson and Lopez hypothesized that the extent of male excess mortality is mainly determined by the harmful lifestyles of men of low socioeconomic status.52 Wingard et al.53 and Luy and Gast54 supported this hypothesis with different empirical approaches. – Finally, a survival advantage among women may also be inferred from the tendency in women to consult a doctor more often than men, both on noticing symptoms of illness and because of their health care needs related to childbearing.55 This could lead to an early detection of serious diseases with increased chances to treat them successfully.56 The contribution of this factor to the sex difference in mortality is discussed controversially, however.57 All these acquired risks develop in the context of the prevailing economic, social, cultural and political system that immediately affect the opportunity structures of individuals and groups and the external forces that affect their behaviour and life chances.58 Some research suggests that the population’s living environment itself is the central driver of the non-biologically caused difference in mortality between women and men. For instance, Preston argues that increasing sex differences in life expectancy during the twentieth century were an effect of the economic modernization of society improving the status of women more than that of men, and this leading to a greater reduction in 47 48 49 50 51 52 53 54 55 56 57 58

See also Waldron (1991). Jarvik (1963). Jenkins (1976). Luy / Di Giulio (2005), Luy / Di Giulio (2006), Nathanson (1984), Waldron (1978). E. g. Johansson (1991), McDonough et al. (1999), Vallin (1995). Nathanson / Lopez (1987). Wingard et al. (1983). Luy / Gast (2014). Galdas et al. (2005), Hazzard (1986), Verbrugge / Wingard (1987), Wallen et al. (1979). Lang et al. (1994). Johansson (1991), Verbrugge (1985). Angel (2011), Anson (2003).

The impact of biological factors on sex differences in life expectancy

23

mortality among women.59 Similarly, Ram60 and Medalia and Chang61 identified gender equality and the degree of modernization of society as the decisive causes for the extent of male excess mortality. In populations directly involved in the two World Wars the increasing sex gap in mortality might also be linked to the impact of the wars on risk selection of women and men (most male war victims were selected largely among healthy persons in good physical condition, whereas females who died during the two World Wars were subject to poor hygiene and medical treatment, which largely affected persons who were physically weaker and less healthy) or to the long-term consequences of poor nutrition and suffered traumas which affected women and men differently.62 4. Interaction between biological and non-biological factors Biological and non-biological factors are not operating in isolation of each other. In fact, there are several routes through which they are interacting: – Sex hormones are supposed to increase or mitigate the health effects of specific diseases which are strongly related to health lifestyles. For instance, smoking, unhealthy diets, excessive body weight in relation to height, lack of exercise and too much stress are thought to operate primarily by raising the mortality from coronary heart disease. As a result of the protective effect of female hormones against this kind of disease, biological differences between the sexes could play an important mediating role between non-biological factors and ultimate mortality.63 Manton mentions significant hormone-related differences between women and men when it comes to the effects of major risk factors on mortality from stroke (obesity, adult-onset diabetes, hypertension) compared to those for heart disease (hypercholesterolemia, hyperhomocystinemia) at later ages.64 Another example for interaction between sex hormones and behavioural factors is female breast cancer, where age-specific fertility behaviour – which is also socioeconomically and culturally regulated – determines the length of exposure of breast tissue to endogenous oestrogens.65 – Excessive male mortality due to “external causes” (accidents, injuries, homicides and suicides) has also been hypothesized to root in hormonal differences between women and men.66 One possible mechanism refers to the effects of prenatal and post-pubertal male hormones as drivers of sex differences in physical activity levels and physical aggressiveness, which 59 60 61 62 63 64 65 66

Preston (1976). Ram (1993). Medalia / Chang (2011). Dinkel (1984), Hart (1989), Haudidier (1996), Haudidier (2005), Höhn (1996), Horiuchi (1983), Luy / Zielonke (2009), Sibai et al. (2001). Retherford (1975). Manton (2000). See Manton (2000), p. 48. Waldron (1983b).

24

–

–

–

67 68 69 70

Marc Luy

may contribute to males being at higher risk for fatal accidents and other violent deaths.67 Recent research suggests that also sex differences in central cholesterol and serotonin levels contribute to higher average aggressiveness among males.68 Other scholars suggest that inherent sex differences in reproductive function (women’s role as bearers and carers of children versus males’ internal competition to mate with females) have influenced the cultural evolution of gender roles and consequently indirectly contributed to sex differences in behaviour, like risk-taking or heavy drinking.69 A specific mechanism behind this development is supposed to be traits shaped by sexual selection which interact with environmental conditions to cause sex differences in psychology, ageing rate and mortality.70 The process of sexual selection might be the basis for Johansson’s observation that the behaviour of women is, on average, more strongly determined by “positive freedoms” (defined loosely as the extent to which individuals are free to make fundamental decisions about their long-term welfare because they control the material resources necessary to support extended longevity, e. g. through education, work and health care), while male behaviours are rather expressed through “negative freedoms” (defined as the extent to which individuals are free to dispose of whatever resources they happen to control on a day-to-day basis, in order to satisfy their short run needs, and/or pursue various forms of pleasure).71 The hypothesized war impact is also partly due to interaction with biological factors. Horiuchi assumes that the impact of war on increasing male excess mortality operates primarily through the poor nutritional situation of the population in the immediate post-war years.72 As a result of sex-specific anatomic characteristics – greater female ability to store energy in the form of body fat73–the nutritional deficit affected mainly adolescent men at the end of the war by causing increased susceptibility to cardiovascular disease later in live. Last but not least, modern medical knowledge and technology opened up the possibility to influence the level of sex hormones through hormone treatments for both women and men. Future developments in the field of gene engineering might extent this possible influence of biological differences between the sexes to include the genetic level.

Lopez (1984). Wallner / Machatschke (2009). Carey / Lopreato (1995), Waldron (1983a), Waldron (1983b). Bonduriansky et al. (2008), Clutton-Brock / Isvaran (2007), Kruger / Nesse (2004), Kruger / Nesse (2006), Kruger / Nesse (2007), Promislow (1992), Promislow (2003), Wells (2000). 71 Johansson (1991). 72 Horiuchi (1983). 73 See also Speakman (2013).

The impact of biological factors on sex differences in life expectancy

25

Figure 1: Overview of the factors contributing to differences in life expectancy between women and men Differences  in  life  expectancy  between  women  and  men   Non-­‐biological  factors:   male  disadvantage  

Biological  factors:   female  advantage   GeneBc  sex  (XX-­‐XY)   (X  dose/inacBvaBon,  temolere   aFriBon,  mitochondrial  inheritance)  

Hormone   treatment  

Sex  hormones   OxidaBve   stress   Lipid   handling  

Lifestyle  

Immuno-­‐   competence   HPA   stress   response  

Living  environment   (economic/social/cultural/   poliBcal  system,  historical  events)  

Cellular   integrity  

Inter-­‐   acBon  

Sexual   selecBon  

Health   behaviours   OccupaBonal   risks   Social   stress  

Smoking   Use  of   medical  services   Social  class   distribuBon  

Gender  roles  

Disease  environment   (incl.  health  care  and  medical  technologies,  treatment  of  illnesses  and  diseases)  

Source: author’s own; illustration of biological factors adopted from Seifarth et al. (2013: 393), slightly modified

Figure 1 provides an overview of the complex network of factors contributing to sex differences in life expectancy. The graph summarizes the various biological and non-biological factors and the interactions between them. The effects of each factor develop within the prevailing disease environment which differs between populations and is subject to changes in time. Thus, the whole causation system and each of its elements have to be seen as dynamic with regard to the corresponding impacts on mortality differences between women and men.74 5. Quantitative estimations of the impact of biological factors on sex differences in life expectancy Until the middle of the twentieth century, sex differences in mortality were thought to be caused primarily by biological factors.75 However, with the extension of the sex gap during the second half of the century it became the established view that the natural female survival advantage is responsible for only a minor fraction of the increased differences in life expectancy between 74 75

See Johansson (1991). E. g. Madigan (1957), Scheinfeld (1958).

26

Marc Luy

women and men in developed countries.76 The strongest support for this inference comes from studies demonstrating that male excess mortality is significantly lower in subgroups of individuals among whom male and female lifestyles and social environments are more homogeneous than in the total population, such as non-smokers,77 Mormons,78 Seventh-day Adventists,79 Old Order Amish,80 Kibbutz members,81 members of Roman-Catholic orders,82 Israeli Jews83 and the traditionally living inhabitants of an isolated upland bog in Ireland84 as well as an old village in Sardinia.85 However, the quantification of the impact of biological factors is difficult because it is impossible to carry out pertinent experiments in human beings and research is therefore limited to what can be observed.86 Bourgeois-Pichat was the first who tried to assess the biologically caused difference in life expectancy between the sexes in the early 1950s.87 Based on an analysis of deaths due to causes that were considered unavoidable at that time, he found a natural difference of 1.9 years of life expectancy at birth. Around two decades later, Pressat presented an almost identical estimate by stating that biological factors are responsible for a female advantage in life expectancy of approximately two years.88 This assessment was based on observations of pre-industrial populations and of a 25–30 % higher mortality in males in the first year of life in the post-industrial era; both were identified by Pressat as being primarily biologically driven. Studying similar records of mortality conditions during the early stages of the epidemiologic transition and mortality levels reflected in model life tables, Trovato and Lalu surmised that biological factors cause women’s life expectancy to be in general 1–2 years, but under no conditions more than 2–3 years, higher than that of men.89 Waldron and Johnston did an in-depth analysis of causes-of-death data for the US population in 1967 to identify the main causes of the mortality differences between women and men.90 They attribute about 70 % of the overall sex differential to specific social and behavioural causes and estimate that genetic factors are responsible for 5 % of the difference. The remaining 25 % were not assigned to specific risk factors, but the authors note the contribution of the 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90

Among many others DesMeules et al. (2004), Enterline (1961), Hart (1989), Lemaire (2002), Lopez (1984), Stillion (1985), Waldron (1995), Wingard (1982). Miller / Gerstein (1983). Lyon / Nelson (1979). Berkel / de Waard (1983). Hamman et al. (1981). Leviatan / Cohen (1985). Luy (2003a). Staetsky / Hinde (2009). Casey / Casey (1970). Poulain et al. (2011). Austad (2006), Kalben (2002). Bourgeois-Pichat (1952). Pressat (1973). Trovato / Lalu (1996). Waldron (1976), Waldron / Johnston (1976).

The impact of biological factors on sex differences in life expectancy

27

Table 1: Estimates for the quantitative impact of biological factors on sex differences in life expectancy (LE) in the post-war period

Author(s) Bourgois-Pichat (1952) Pressat (1973) Waldron (1976), Waldron/ Johnston (1976) Trovato/Lalu (1996) Luy (2003) Christensen/Herskind (2007)

Research strategy Unavoidable deaths Pre-industrial populations, infant mortality

Estimate 1.9 years of LE 25–30 %, ~2 years of LE

Model life tables Catholic order members

5 % genetic + max. 25 % hormonal factors 1–2 years of LE (max. 2–3) 1–2 years of LE

Mono- vs. dizygotic twins

25 % (general variation)

Causes of death

Source: author’s own compilation

possibly protective role of female hormones. By analyzing the mortality of around 12,000 members of female and male Catholic orders in Germany between 1890 and 1995, Luy estimated that biological factors cause a maximum sex difference of 2 years in life expectancy at young adult ages.91 The estimates are summarized in Table 1. Each of them is definitely imprecise because environmental effects can never be completely isolated92 and each estimate is basically only valid for the specific conditions studied.93 Nonetheless, they all agree on the modest size of the naturally caused sex gap with only minor deviations in the estimated extent. Interestingly, all these estimates are similar to results obtained in twin studies, which attribute approximately 25 % of the general (not sex-specific) variation in both health94 and mortality95 to genetic factors. 6. A refined estimation of biologically caused sex difference in life expectancy We assessed the impact of biological factors on sex differences in life expectancy by relating the sex difference in life expectancy of Catholic order members (as estimate for the biologically caused differences) to the prevailing sex gap of the total population. Life tables for Catholic order members are based on the continuously extended nuns’ and monks’ mortality data base of the ”German-Austrian Cloister Study” collected from order archives.96 The data includes for each individual ever entering the participating monasteries the 91 92 93 94 95 96

Luy (2003a). Nobile (2007), Waldron (1983b). Rogers et al. (2000), Waldron (1995). Christensen et al. (1999). Christensen / Herskind (2007). Luy (2003a), Luy (2009).

28

Marc Luy

date of birth, date of entry and if applicable date of exit or death. At present, the data includes the life dates of 15,164 order members from 12 German communities (9,264 nuns and 5,900 monks) and 1,427 from 4 Austrian communities (305 nuns and 1,122 monks). The included communities belong predominantly to the so-called “semi-contemplative orders” – such as Franciscans, Dominicans and Benedictines – which can be characterized as institutionalized compromises between activity in the general public and strict contemplation.97 Life tables of the general populations of Germany and Austria are calculated based on data of the Human Mortality Database.98 The data for Germany and Austria are combined by averaging the life expectancies. Although it is known that cloistered life entails specific risk factors which influence women’s and men’s mortality differently – among them the high tuberculosis mortality in nuns in the first half of the twentieth century and the impact of smoking on the mortality of monks since the 1970s99 – it seems that nuns and monks are an almost ideal experimental setting for this research question.100 They comprise the group of women and men among whom behavioural and environmental conditions are probably as close to being equal as can be found in modern societies. Female and male order members have a “simple lifestyle” determined by vows (living in poverty, chastity and obedience), with similar daily regimes as regards time for sleep, work, study and recreation, and also with respect to diet, housing and medical care. The first vows are preceded by a novitiate of at least twelve months. During this trial period candidates are screened for psychological suitability as well as physical health. Hence, nuns and monks form a select group of individuals in good and stable mental and physical condition at the time of entry.101 Furthermore, cloistered life entails no sex-specific influences of financial burdens, reproductive roles, marital status or familial responsibilities. Nuns and monks live their whole life in a stable community and they are not burdened by the fear of becoming unemployed, the associated competitions, or the need for occupational advancement and social climbing. Cloistered life is not entirely free of pressures and stress, but the forms of interpersonal conflicts are very different from those of a worldly society. Aside from the typical internal occupations in male and female cloisters like household work, farming and similar occupations, most monks practise as priests while nuns mainly work as nurses or school and pre-school teachers. Despite these differences in the professions of nuns and monks, sex-specific mortality risks related to occupation are unlikely among the order members. The only exception is the outings of priests which

97 Klassen (2001). 98 Human Mortality Database. University of California, Berkeley (USA), and Max Planck Institute for Demographic Research (Germany). Available at www.mortality.org or www. humanmortality.de (data downloaded on 31 July 2009). 99 See Luy (2003a). 100 See also Hansson et al. (1997), Madigan / Vance (1957). 101 See also Gajewski / Poznanska (2008).

The impact of biological factors on sex differences in life expectancy

29

might increase the risk of being involved in traffic accidents.102 However, such accidents can have only minor influence on overall mortality of order members. As a consequence of these preconditions, a mortality analysis of nuns and monks can abstract from a whole range of factors that influence sex differences in life expectancy, including different behavioural patterns of women and men regarding pursuit of harmful lifestyles, different propensities of women and men to work in hazardous occupations, maternal mortality and causes of death linked to pregnancy, the unequal roles of husband and wives, the degree of modernization of society, the increased stress load connected with specific typical male occupations, nutritional differences between women and men, and differences in social status and their potential sex-specific impact. Thus, differences in life expectancy between nuns and monks can be supposed to be an approximate estimate of the impact of biological factors on sex differences in life expectancy. An analysis of external cause mortality among order members suggests that the interactions of biological and non-biological factors are inherent when using the differences between nuns and monks in this context.103 The best conditions for this natural experiment to isolate the impact of biological factors on sex differences in life expectancy prevailed in the period 1960–1970 for the following reasons: – The tuberculosis threat among nuns – mainly caused by their nursing activity – ceased to exist as bias for female order members’ mortality; – Smoking could not yet have caused high mortality in male religious communities where cigarette smoking only became common after World War II (note that even nowadays smoking is not tolerated in female communities); – During the first half of the twentieth century, female and male religious communities were affected by specific circumstances – high tuberculosis prevalence for nuns and two world wars for monks – which increased the mortality among their young members and led to similarly selected populations in the post-war periods. An additional advantage of a mortality study among Catholic order members is the existence of complete and trustworthy records of vital statistics of their members.104 Moreover, the data permits estimates of the naturally caused differences as a function of the overall sex gap in life expectancy. As outlined in Section 4, because biological factors interact with non-biological drivers of mortality, it seems plausible that their contribution is not stable but rather varies with the respective living and disease environments (see Figure 1). In order to take these interactions into account we estimated the impact of biological factors as a function of the overall difference in life expectancy between women and men at ages 20, 25, 30, …, 90 according to period life ta102 Luy (2009). 103 Ibid. 104 King / Bailar (1969), Madigan / Vance (1957).

30

Marc Luy

bles for 1960, 1965 and 1970 by relating the sex gap among order members to the corresponding sex gap of the total population. The use of estimates for different ages to simulate different conditions of mortality levels is a common procedure in indirect demographic estimation in cases of lacking observation. Such approximation is, for instance, the basis of the well-established “growth balance method” for the estimation of adult mortality in developing countries.105 Figure 2 shows the estimate based on life tables for order members from Germany and the total West German population. Each life table is calculated from the age-specific deaths and person years lived within the 30 calendar years around the midpoint years.106 The data suggests a logarithmic relationship between the impact of biological factors (estimated on the basis of the difference in life expectancy between nuns and monks) and the extent of the sex differences in life expectancy (indicated by the respective values for the total population). Note that this function is based on mortality experiences at adult ages only. Nonetheless, using it to quantify the impact of biological factors on sex differences in life expectancy at birth seems justified as it was recently reported by Pongou that sex differences in infant mortality are not caused by biological factors only either.107 His estimates indicate that the preconception environment, including age, health and smoking behaviour of parents as well as gestational weight of mothers, are almost as important as natural differences between the sexes. The thin line in Figure 2 indicates one quarter of the sex difference in life expectancy of the total population. This corresponds to the estimated 25 % for the biological impact identified in some other studies. It becomes apparent that our estimated logarithmic function and the constant 25 % line are very close for sex differences in life expectancy between 1.0 and 6.0 years, including the range prevailing in most western populations since the Second World War (see area marked with a rectangle in Fig. 2). However, the functions yield different estimates for the biological impact in populations with higher sex differences as they are found in Russia and other Eastern European countries.

105 See e. g. Moultrie et al. (2013). 106 For details see Luy (2003a). 107 Pongou (2013).

31

The impact of biological factors on sex differences in life expectancy Figure 2: Estimated contribution of biological factors to sex differences in life expectancy (nuns-monks difference) as function of the overall life expectancy gap between women and men (female-male difference in the general population), data for Germany

4.0 constant 25% difference

Nuns-monks difference

3.0 single observations for specific ages

2.0 1.0 0.0

range of current sex gap in western societies

-1.0

estimated logarithmic function

-2.0 0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

Female-male-difference in the general population Source: author’s own calculations with author’s own data (nuns and monks) and data from the Human Mortality Database (general population); NB: Estimates refer to life expectancy between ages 20 and 90 derived from period life tables based on data for 5-year age groups and covering 30 calendar years around the years 1960, 1965 and 1970.

Figure 3 includes identical estimates based on life tables for smaller observation periods (10 and 20 years around the mid-year) for the German order members (black lines) and the same estimates for total sample including order members from German and Austrian plus an estimate including only data for the specific period 1960–1970 (blue lines). The black line marked with “GER 5×30” corresponds to the estimate presented in Figure 2. It becomes apparent that each estimate suggests a similar logarithmic association between the sex differences in life expectancy among order members (as estimate for the biologically caused sex gap) and those prevailing in the total population. The differences between the estimates are only minor as can be seen in comparison to the constant 25 %-difference line as well as in the marked area of currently prevailing sex differences in life expectancy in western societies. In this area, the impact of biological factors on differences in life expectancy between women and men ranges between a minimum of around 0.5 years and a maximum of 1.5 years across all estimates.

32

Marc Luy

Figure 3: Estimated contribution of biological factors to sex differences in life expectancy (nuns-monks difference) as function of the overall life expectancy gap between women and men (female-male difference in the general population), data for Germany and Austria and different observation periods

4.0

Nuns-monks difference

3.0

GER 5x10

GER+AUT 5x10

GER 5x20

GER+AUT 5x20

GER 5x30

GER+AUT 5x30

constant 25% difference

2.0 1.0 0.0 range of current sex gap in western societies

-1.0

GER+AUT 1960-70 (5x10)

-2.0 0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

Female-male-difference in the general population Source: author’s own calculations with author’s own data (nuns and monks) and data from the Human Mortality Database (general population); NB: Functions derived from estimates for life expectancy between ages 20 and 90 for the years 1960, 1965 and 1970 (see Fig. 2); 5×10 = life tables for 5-year age groups and 10 calendar years around mid-year, 5×20 = life tables for 5-year age groups and 20 calendar years around mid-year, 5×30 = life tables for 5-year age groups and 30 calendar years around mid-year; GER = Germany, AUT = Austria

7. An application: causes of sex differences in life expectancy in Germany The derived function for the impact of biological factors on sex differences in life expectancy can be applied easily to any empiric data. This is demonstrated in this Section with mortality data for Germany between 1950 and 2012. Figure 4 shows the corresponding results separated for western Germany (Fig. 4a) and eastern Germany (Fig. 4b). In western Germany, sex differences in life expectancy at birth followed the typical western pattern as described briefly in the introduction.108 In 1950, women outlived men by 3.9 years. The following years were characterized by an almost constant increase of the sex gap up until the year 1980 when the difference between women and men reached its maximum of 6.8 years. In the subsequent years, the female surplus in life ex108 For details see Liu et al. (2013).

The impact of biological factors on sex differences in life expectancy

33

Figure 4: Estimated contribution of biological and non-biological factors to sex differences in life expectancy at birth in West and East Germany, 1950–2012 (b) East Germany

8.0

7.0

7.0 Sex difference in life expectancy

Sex difference in life expectancy

(a) West Germany

8.0

6.0 5.0 4.0 3.0

Non-biological factors

2.0 1.0 0.0 1950

1980 Year

1995

5.0 4.0 3.0

2010

Non-biological factors

2.0 1.0

Biological factors 1965

6.0

0.0 1950

Biological factors 1965

1980 Year

1995

2010

Source: author’s own calculations with data from Luy (2015)

pectancy decreased progressively to 4.6 years in 2012. Sex differences in eastern Germany followed the typical central-eastern European pattern, indicating the relevance of general economic and political conditions. Starting at similar sex difference levels of around 4 years, the gap increased for a longer period of time and to a higher level compared to western Germany. The maximum was reached in 1993 with the difference in life expectancy between women and men being 7.5 years. The subsequent years are characterized by an almost constant decrease of the sex gap to 5.8 years in 2012. Figure 4 includes the corresponding estimates for the impact of biological factors based on the logarithmic function derived in Figure 2. In both parts of Germany, these are estimated to be around one year across the whole observation period. Most of the sex gap in life expectancy and the trend changes are consequently due to non-biological factors. Because the estimation function derives the impact of biological factors from the overall sex difference in life expectancy, their contribution is slightly increasing with the opening and decreasing with the closing of the sex gap. This reflects the fact that in the logic of the derivation of the estimation function the biological factors include the interaction effects with non-biological factors. Nonetheless, the changes of the estimated impact of biological factors are minor compared to those of the overall life expectancy gap between women and men, indicating that also the trend changes were caused primarily by non-biological factors.

34

Marc Luy

Table 2: Estimated contribution of biological factors (BIOL), smoking (SMOKE) and other non-biological factors (ONBF) to sex differences in life expectancy at birth (SDLE) in West Germany, 1955/59–2005/09

Absolute contribution (in years) Relative contribution (in %) Period

SDLE

BIOL

SMOKE

ONBF

BIOL

SMOKE

ONBF

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

1955/59

5.03

1.08

1.99

1.96

21.4

39.6

39.0

1960/64

5.67

1.14

2.42

2.12

20.1

42.6

37.3

1965/69

6.07

1.17

2.71

2.19

19.3

44.6

36.1

1970/74

6.47

1.20

2.89

2.38

18.6

44.7

36.7

1975/79

6.70

1.22

3.06

2.42

18.2

45.6

36.2

1980/84

6.74

1.22

3.10

2.41

18.2

46.0

35.8

1985/89

6.61

1.22

3.08

2.31

18.4

46.6

35.0

1990/94

6.44

1.20

2.98

2.27

18.6

46.2

35.2

1995/99*

6.37

1.20

2.64

2.53

18.8

41.5

39.7

2000/04*

5.74

1.14

2.22

2.38

19.9

38.7

41.4

2005/09*

5.18

1.09

1.85

2.23

21.1

35.8

43.1

Source: Luy and Wegner-Siegmundt (2015) NB: * periods include data for all of Germany

The quantitative isolation of biological factors is useful when one is interested in a comparative analysis of specific non-biological risk factors and their impact on sex differences in life expectancy. Luy and Wegner-Siegmundt, for instance, analysed the impact of smoking on sex differences in life expectancy in comparison to other non-biological factors in Europe.109 This required an estimation of the impact of biological factors to reduce the sex gap in life expectancy to the portion caused by non-biological factors. Table 2 presents the corresponding estimates for western Germany from 1955–59 to 2005–09. Column (2) includes the prevailing sex differences in life expectancy at birth in years, columns (3)–(5) the estimated portions caused by biological factors, smoking and other non-biological factors in years of life expectancy, and columns (6)–(8) include the corresponding portions in percent of the sex gap. In relative terms, the impact of biological factors was around 20 percent of the overall sex difference in life expectancy. However, the trend of the relative portion caused by biological factors contrasts with the absolute trend. This is due to the fact that the trend was influenced more strongly by the non-biological factors, most importantly by the effects of tobacco consumption. Note, however, that the estimated impact of smoking declined strongly since the 1980s whereas the impact of other-non-biological factors has stalled. Consequently, since 2000–04 the other non-biological factors have contributed more 109 Luy / Wegner-Siegmundt (2015).

The impact of biological factors on sex differences in life expectancy

35

than smoking to sex differences in life expectancy in western Germany in absolute as well as in relative terms. 8. Discussion The main aim of this contribution was to provide a quantitative estimate of the impact of biological factors on sex differences in life expectancy in the postwar period. The latter specification is important because the impacts of all factors related to the mortality difference between women and men are strongly dependent on the prevailing disease environments. These changed considerably in the course of the epidemiologic transition. Consequently, even biological factors cannot be expected to operate independently of these conditions. This is also a result of the fact that biological and non-biological factors are interacting at several levels as described in the corresponding section of this article. Our review of the literature yielded only a handful of studies providing a quantitative estimate of the impact of biological factors. Interestingly, although they all pursued very different approaches, they resulted in almost identical estimates, which indicates that the impact of biological factors is unlikely to exceed 25 percent or 2 years of life expectancy difference between the sexes. In this paper we further elaborated our previous approach to estimate the contribution of biological factors to the sex gap in life expectancy based on the differences between female and male members of Catholic religious orders from Germany and Austria. The main advantage of this approach lies in the possibility to estimate the biologically caused sex gap for different levels of life expectancy and thereby to relax the assumption that the impact of biological factors is constant. In fact, our results indicate that the association between the impact of biological factors and the extent of the sex difference in life expectancy tends to follow a logarithmic relationship. This pattern reflects the expectation one should derive from the manifold interactions between biological and non-biological factors. The difference to the estimates derived from other approaches is nonetheless only minor in absolute as well as relative terms for most industrialized populations of the post-war era. The history of research on sex differences in mortality includes two other studies of differences in life expectancy between female and male order members. The first was Deparcieux’s search for data to arrive at general estimates for the life expectancy of women and men in the seventeenth and eighteenth century when comprehensive systems of population statistics were not yet established.110 His estimates based on lists for order members who died between 1685 and 1745 revealed a difference in life expectancy at age 20 between nuns and monks of 1.6 years. Although related to a very different time, this sex difference in life expectancy lies well within the range of our estimates for the 110 Deparcieux (1746).

36

Marc Luy

differences in life expectancy between female and male order members who lived in the second half of the twentieth century. The second study of sex differences in life expectancy among Catholic order members was conducted by Francis C. Madigan in the 1950s. In his doctoral thesis entitled “The differential mortality of the sexes, 1900–1954: Cultural and biological factors in the diverging life chances of American men and women”,111 he studied the same research question as we did in our original study, i. e. whether mainly biological or non-biological factors are responsible for increasing sex differences in life expectancy – a process that started in the U. S. already during the first half of the twentieth century.112 Being a Jesuit monk himself, Madigan followed the same experimental approach by comparing the mortality of female and male members of Catholic orders in the United States. The main results of this study have been published in two articles one year after the completion of his PhD thesis.113 In order to establish the ideal experimental setting, Madigan restricted his sample of order members to those who worked as teachers or in the administration of Catholic brotherhoods and sisterhoods engaged in educational work. Communities which were operating hospitals were excluded as were individuals who worked in household and manual duties, nurses, and those who had served in foreign missions. Moreover, he included only white order members and eliminated those from the study who entered the religious life on or after their 27th birthday. Based on this subsample of order members he found that both brothers and sisters showed a similarly lower mortality compared to their worldly counterparts in the United States throughout the observation period. As a consequence, the differences between sisters and brothers increased similarly to the sex gap prevailing in the total population. Madigan’s findings are therefore in sharp contrast to our findings as he concluded “(1) that biological factors are more important than sociocultural pressures and strains in relation to the differential sex death rates; and (2) that the greater sociocultural stresses associated with the male role in our society play only a small and unimportant part in producing the differentials between male and female death rates”.114 Madigan himself did not expect these findings because, according to him, the life of sisters appears to be more stressful than that of brothers in comparison to the women and men of the general population. Thus, according to Madigan, “even if sociocultural factors should be only of slight importance in relation to the observed sex mortality differentials of the general public, one would still not anticipate finding that young Sisters, at least, had experienced greater gains over females of the general public in mortality rates than Brothers had made over the corresponding males”.115 He therefore expected results 111 112 113 114 115

Madigan (1956). See Wiehl (1938). Madigan (1957), Madigan / Vance (1957). Madigan (1957), p. 209, italics in original. Madigan (1957), p. 209.

The impact of biological factors on sex differences in life expectancy

37

more in line with our findings according to which monks exhibit significant lower mortality than men of the total population, whereas nuns showed mortality similar to or higher than that of their worldly counterparts.116 Our results consequently correspond to the three conditions described by Madigan and Vance which would lead to a confirmation of the non-biological hypothesis: “First, the life expectations and death rates of Brothers should improve over those of United States white males as to reach a position approximately intermediate between the mortality records of American white males and females. Secondly, life expectations and death rates of Sisters should retrogress from the favoured position of United States white females so as to bring about considerable convergence with the mortality experience of Brothers. Thirdly, both parts of this relation should remain fairly constant over time for the period of observation […]. The relation ought to be reflected particularly in the middle and old age groups, because stress factors should effect greater and greater cumulative results with each older group from the beginning of middle age.”117

There are several issues that could explain the opposing results of Madigan’s research compared to our study. Firstly, Madigan studied sex differences in mortality in a different disease environment. During the first half of the twentieth century, infectious diseases were still the most important cause of death – in contrast to the post-war disease environment in which most deaths are caused by circulatory disease and neoplasms. Moreover, also the general living environments were very different. The most relevant non-biological factors causing specific stresses for men during the years covered by Madigan’s study have been assumed to be “an agricultural depression, an industrial depression of unprecedented magnitude, and the second world war”.118 In contrast, the post-war perspective assumes lifestyles and occupational risks to be the most important non-biological drivers of male excess mortality. The second issue concerns the specific selection of Madigan’s subsample of order members which has been criticized also by other scholars.119 The order members included in the Madigan study appear to be better protected against infectious diseases – in particular against tuberculosis – than those who were excluded from the sample. Moreover, according to the results from our study,120 the exclusion of nuns and monks who served in foreign mission creates a significantly sex-differentiated selection of order members regarding their mortality.121 Thirdly, according to Madigan, the brothers in the American communities were more likely than the sisters to smoke and have an occa116 117 118 119 120 121

Luy (2003a). Madigan / Vance (1957), pp. 197–198. Ibid., p. 195. King / Bailar (1969), Nathanson (1984), Waldron (1983). Luy (2003b). Missionating nuns exhibited even significantly lower mortality than non-missionating sisters, above all in higher and highest ages. This is due to the fact that female orders practised a strict health-based selection of missionaries by sending only the healthiest members abroad and keeping those with even minor current or former health problems at home. Male orders practised a less strict selection. As a consequence, missionating monks show an equal or – among the older cohorts – only slightly higher mortality compared to the brothers who never left Germany.

38

Marc Luy

sional drink. Among the order members of our sample, smoking started to become accepted only after the Second World War in male religious communities, whereas it has been prohibited in female communities until today.122 Fourthly, Madigan’s study suffers from low case numbers of deceased male order members despite the large overall sample size123 as well as from his methodological approach to apply the general population’s death rates for ages 85+ in an identical way to the male order members.124 Finally, one cannot exclude a bias in the Madigan study due to the fact that more than half of the data was compiled by the communities and not directly by the research team as is the case in our study. In sum, the opposing results and conclusions between the Madigan study and ours cannot be seen as qualification of the presented quantitative estimate of the impact of biological factors on sex differences in life expectancy in the post-war period. However, a decisive question for our estimates being accepted in general is whether Catholic order members are representative of women and men of the total population. According to our knowledge, this question has never been examined in detail. Timio et al. tested some medical conditions and health behaviours which might cause biases when comparing female order members with worldly women.125 They reported that the nuns and the general population women of their study were similar in most of these variables. Also Madigan – an insider himself because of being an order member – excluded the possibility of sex-specific selection effects among order members that might produce corresponding effects on mortality.126 Madigan and Vance explicated this assumption in more detail: “Are our experimental groups representative of the biological population from which they were drawn? It is our assumption that they are. First, in order to be admitted, the candidates have to conform to no rules regarding the physical structure […]. All that is required is that they be in normal good health and that insanity not be part of the family history. Of course they differ from the general population in that they must pass a health examination for admittance, but this makes no difference since it is equally true of both of the experimental groups. There is no evidence, moreover, that any special physical type is attracted particularly to the life of the teaching Brother or the teaching Sister. Actually, investigation has shown us the presence of all types. The randomness of the sample should be judged by the biological condition of the individual members at the point when they enter religious life, not after it. This is to say that clustering does not occur because of the group life. Rather this group life is the experimental factor itself, to which the individuals, equally selected at random, are exposed.”127

To conclude, according to all we know we can assume that Catholic order members comprise the best experimental setting for estimating the impact of biological differences on sex differences in mortality among humans. The esti122 123 124 125 126 127

Luy (2002), Luy (2003a). For details see Luy (2003a). Madigan (1956), pp. 111–112, footnote 3. Timio et al. (1999). Madigan (1957), p. 222. Madigan / Vance (1957), p. 196.

The impact of biological factors on sex differences in life expectancy

39

mates presented in this article can be used to isolate biological factors and to analyze the non-biologically caused portion of sex differences in life expectancy as demonstrated in Section 7 where we estimated the impact of biological factors on the sex gap in eastern and western Germany. Acknowledgement This research was supported by the European Research Council within the European Community’s Seventh Framework Programme (FP7/2007–2013), ERC Grant Agreement No. 262663 (HEMOX). Bibliography Aden, A. S.; Omar, M. M.; Högberg, U.; Persson, L. Å; Wall, S.: Excess female mortality in rural Somalia – is inequality in the household a risk factor? In: Social Science & Medicine 44, 5 (1997), pp. 709–715. Albu, Jeanine B.; Murphy, Laura; Frager, David H.; Johnson, Julia A.; Pi-Sunyer, F. Xavier: Visceral fat and race-dependent health risks in obese nondiabetic premenopausal women. In: Diabetes 46, 3 (1997), pp. 456–462. Angel, Ronald J.: Agency versus structure: genetics, group membership, and a new twist on an old debate. In: Social Science & Medicine 73, 5 (2011), pp. 632–635. Anson, Jon: Sex differences in mortality at the local level: an analysis of Belgian municipalities. In: European Journal of Population 19, 1 (2003), pp. 1–28. Asdell, S. A.; Doornenbal, H.; Joshi, S. R.; Sperling, G. A.: The effects of sex steroid hormones upon longevity in rats. In: Journal of Reproduction and Fertility 14, 1 (1967), pp. 113–120. Austad, Steven N.: Why women live longer than men: sex differences in longevity. In: Gender Medicine 3, 2 (2006), pp. 79–92. Barford, Anna; Dorling, Danny; Davey Smith, George; Shaw, Mary: Life expectancy: women now on top everywhere. In: British Medical Journal 332 (2006), pp. 808. Barrett, Emma L. B.; Richardson, David S.: Sex differences in telomeres and lifespan. In: Aging Cell 10, 6 (2011), pp. 913–921. Behl, Christian; Skutella, Thomas; Lezoualc’h, Frank; Post, Anke; Widmann, Martina; Newton, Christopher J.; Holsboer, Florian: Neuroprotection against oxidative stress by estrogens: Structure-activity relationship. In: Molecular Pharmacology 51, 4 (1997), pp. 535– 541. Berkel, J.; de Waard, F.: Mortality pattern and life expectancy of Seventh-day adventists in the Netherlands. In: International Journal of Epidemiology 12, 4 (1983), pp. 455–459. Bonduriansky, R.; Maklakov, A.; Zajitschek, F.; Brooks, R.: Sexual selection, sexual conflict and the evolution of ageing and life span. In: Functional Ecology 22, 3 (2008), pp. 443– 453. Borrás, Consuelo; Sastre, Juan; García-Sala, David; Lloret, Ana; Pallardó, Federico V.; Viña, José: Mitochondria from females exhibit higher antioxidant gene expression and lower oxidative damage than males. In: Free Radical Biology & Medicine 34, 5 (2003), pp. 546– 552. Bourgeois-Pichat, Jean: Essai sur la mortalité “biologique” de l’homme. In: Population 7, 3 (1952), pp. 381–394. Brandon, M.; Baldi, P.; Wallace, D. C.: Mitochondrial mutations in cancer. In: Oncogene 25, 34 (2006), pp. 4647–4662.

40

Marc Luy

Brown, David A.; Lynch, Joshua M.; Armstrong, Casey J.; Caruso, Nicholas M.; Ehlers, Lindsay B.; Johnson, Micah S.; Moore, Russell L.: Susceptibility of the heart to ischaemia reperfusion injury and exercise-induced cardioprotection are sex-dependent in the rat. In: The Journal of Physiology 564, 2 (2005), pp. 619–630. Carey, Arlen D.; Lopreato, Joseph: The biocultural evolution of the male-female mortality differential. In: The Mankind Quarterly 36, 1 (1995), pp. 3–28. Caruso, Calogero; Accardi, Giulia; Virruso, Claudia; Candore, Giuseppina: Sex, gender and immunosenescence: a key to understand the different lifespan between men and women? In: Immunity & Ageing 10 (2013), 20. Casey, Albert E.; Casey, Joanne G.: Long-lived male population with high cholesterol intake, Slieve Lougher, Ireland. In: Alabama Journal of Medical Sciences 7, 1 (1970), pp. 21–26. Christensen, Kaare; Herskind, Anne Maria: Genetic factors associated with individual life duration: heritability. In: Robine, Jean-Marie; Crimmins, Eileen M.; Horiuchi, Shiro; Zeng, Yi (Ed.): Human longevity, individual life duration, and the growth of the oldest-old population. Dordrecht; Springer 2007, pp. 237–250. Christensen, Kaare; Holm, Niels V.; McGue, Matt; Corder, Larry; Vaupel, James W.: A Danish population-based twin study on general health in the elderly. In: Journal of Aging and Health 11, 1 (1999), pp. 49–64. Christensen, Kaare; Ørstavik, Karen Helene; Vaupel, James W.: The X chromosome and the female survival advantage. An example of the intersection between genetics, epidemiology and demography. In: Annals of the New York Academy of Sciences 954 (2001), pp. 175–183. Ciocco, Antonio: Sex differences in morbidity and mortality. In: The Quarterly Review of Biology 15, 1 (1940), pp. 59–73. Clutton-Brock, T. H.; Isvaran, K.: Sex differences in ageing in natural populations of vertebrates. In: Proceedings of the Royal Society B: Biological Sciences 274, 1629 (2007), pp. 3097–3104. Comfort, Alexander: The biology of senescence. New York; Elsevier 1979. Concha-Barrientos, Marisol; Nelson, Deborah Imel; Driscoll, Timothy; Steenland, N. Kyle; Punnett, Laura; Fingerhut, Marilyn; Prüss-Üstün, Annette ; Leigh, James; Tak, Sang Woo; Corvalan, Carlos: Selected occupational risk factors. In: Ezzati, Majid; Lopez, Alan D.; Rodgers, Anthony; Murray, Christopher J. L. (Ed.): Comparative quantification of health risks. Global and regional burden of disease attributable to selected major risk factors. Geneva; World Health Organization 2004, pp. 1651–1801. Dahl, Ronald E.; Siegel, Selma F.; Williamson, Douglas E.; Lee, Peter A.; Perel, James; Birmaher, Boris; Ryan, Neal D.: Corticotropin releasing hormone stimulation test and nocturnal cortisol levels in normal children. In: Pediatric Research 32, 1 (1992), pp. 64–68. Deparcieux, Antoine: Essai sur les probabilités de la durée de la vie humaine. D’où l’on déduit la maniere de déterminer les rentes viagères, tant simples qu’en Tontines. Précédé d’une courte explication sur les rentes à terme, ou annuités. Paris; Frères Guerin 1746. DesMeules, Marie; Manuel, Douglas; Cho, Robert: Mortality: life and health expectancy of Canadian women. In: BMC Women’s Health 4, Supplement 1 (2004), p. S9. DiMauro, Salvatore; Andreu, Antoni L.: Mutations in mtDNA: Are we scraping the bottom of the barrel? In: Brain Pathology 10, 3 (2000), pp. 431–441. Dinkel, Reiner H.: Sterblichkeit in Perioden- und Kohortenbetrachtung – zugleich eine ansatzweise Berechnung der Kohortensterbetafel für Deutschland. In: Zeitschrift für Bevölkerungswissenschaft 10, 4 (1984), pp. 477–500. Enterline, Philip E.: Causes of death responsible for recent increases in sex mortality differentials in the United States. In: The Milbank Memorial Fund Quarterly 39, 2 (1961), pp. 312–328. Fairweather, DeLisa; Cihakova, Daniela: Alternatively activated macrophages in infection and autoimmunity. In: Journal of Autoimmunity 33, 3–4 (2009), pp. 222–230.

The impact of biological factors on sex differences in life expectancy

41

Gajewski, Antoni K.; Poznanska, Anna: Mortality of top athletes, actors and clergy in Poland: 1924–2000 follow-up study of the long term effect of physical activity. In: European Journal of Epidemiology 23, 5 (2008), pp. 335–340. Galdas, Paul M.; Cheater, Francine; Marshall, Paul: Men and health help-seeking behaviour: literature review. In: Journal of Advanced Nursing 49, 6 (2005), pp. 616–623. Hamilton, James B.: The role of testicular secretions as indicated by the effects of castration in man and by studies of pathological conditions and the short lifespan associated with maleness. In: Recent Progress in Hormone Research 3 (1948), pp. 257–322. Hamman, Richard F.; Barancik, Jerome I.; Lilienfeld, Abraham M.: Patterns of mortality in the Old Order Amish. I. Background and major causes of death. In: American Journal of Epidemiology 114, 6 (1981), pp. 845–861. Hansson, Lennart; Hedner, Thomas; Jern, Sverker: Psychological factors and blood pressure. Why are nuns better off? In: Blood Pressure 6, 2 (1997), p. 68. Hart, Nicky: Sex, gender and survival: inequalities of life chances between European men and women. In: Fox, John (Ed.): Health inequalities in European countries. Aldershot; Gower 1989, pp. 109–141. Haudidier, Benoît: Vergleich der Sterblichkeitsentwicklung in der Bundesrepublik Deutschland und in Frankreich. In: Dinkel, Reiner H.; Höhn, Charlotte; Scholz, Rembrandt D. (Ed.): Sterblichkeitsentwicklung unter besonderer Berücksichtigung des Kohortenansatzes. München; Boldt 1996, pp. 139–152. Haudidier, Benoît: Cause-specific mortality in France and in the former FRG, 1950–95. Similarities and differences. In: Population: English Edition 60, 1–2 (2005), pp. 119–145. Hayflick, Leonard: Biological aspects of aging. In: Preston, Samuel H. (Ed.): Biological and social aspects of mortality and the length of life. Proceedings of a seminar at Fiuggi, Italy, May 13–16, 1980. Liège; Ordina Editions 1982, pp. 223–258. Hazzard, William R.: Biological basis of the sex differential in longevity. In: Journal of the American Geriatrics Society 34, 6 (1986), pp. 455–471. Hazzard, William R.: Why do women live longer than men? Biological differences that influence longevity. In: Postgraduate Medicine 85, 5 (1989), pp. 271–283. Hazzard, William R.; Applebaum-Bowden, Deborah: Why women live longer than men: the biologic mechanism of the sex differential in longevity. In: Transactions of the American Clinical and Climatological Association 101 (1990), pp. 168–189. Hemström, Örjan: Does the work environment contribute to excess male mortality? In: Social Science & Medicine 49, 7 (1999), pp. 879–894. Hill, Kim; Boesch, Christophe; Goodall, Jane; Pusey, Anne; Williams, Jennifer; Wrangham, Richard: Mortality rates among wild chimpanzees. In: Journal of Human Evolution 40, 5 (2001), pp. 437–450. Höhn, Charlotte: Kohortensterblichkeit unter besonderer Berücksichtigung der Weltkriege. In: Dinkel, Reiner H.; Höhn, Charlotte; Scholz, Rembrandt D. (Ed.): Sterblichkeitsentwicklung unter besonderer Berücksichtigung des Kohortenansatzes. München; Boldt 1996, pp. 45–66. Horiuchi, Shiro: The long-term impact of war on mortality: old-age mortality of the First World War survivors in the Federal Republic of Germany. In: Population Bulletin of the United Nations 15 (1983), pp. 80–92. Jarvik, Lissy F.: Sex differences in longevity. In: Advances in Sex Research 1 (1963), pp. 151– 158. Jenkins, C. David: Recent evidence supporting psychologic and social risk factors for coronary disease. In: New England Journal of Medicine 294, 19 (1976), pp. 1033–1038. Johansson, S. Ryan: Welfare, mortality, and gender. Continuity and change in explanations for male/female mortality differences over three centuries. In: Continuity and Change 6, 2 (1991), pp. 135–177. Judge, Debra S.; Carey, James R.: Postreproductive life predicted by primate patterns. In: Journal of Gerontology: Biological Sciences 55A, 4 (2000), pp. B201-B209.

42

Marc Luy

Kalben, Barbara Blatt: Why men die younger: causes of mortality differences by sex. In: North American Actuarial Journal 4, 4 (2000), pp. 83–111. Kalben, Barbara Blatt: Why men die younger: causes of mortality differences by sex. Schaumburg, Illinois; Society of Actuaries 2002. Kalmbach, Ellen; Furness, Robert W.; Griffiths, Richard: Sex-biased environment sensitivity: natural and experimental evidence from a bird species with larger females. In: Behavioral Ecology 16, 2 (2005), pp. 442–449. Kersseboom, Willem: Vertoog van 6 May 1737, bewijzende dat de faculteit om de renten in t’Hollands negotiatie, bij wege van loterije van obligation en renten, te mogen nemen op lijven bij verkiezingen zonder eenige distinctie geen nadeel aan den lande heeft kunnen toebrengen, vervattende tegelijk een der exacte proeven van de zekere gronden op welke berust de waardije van Lijfrenten in proportie van Losrenten. 1737. Kersseboom, Willem: Observatien, waar in voornamentlyk getoont word wat is gelyktydigheid, dewelke vereischt word in alle calculatien, die tot voorwerp hebben de probable leevenskracht van persoonen van eenigen voorgestelden ouderdom. Hier is bygevoegt een proeve van de constante ondervinding, dat die van de vrouwelyke sexe doorgaans maar eenige weinige jaaren doormalkander langer leven, dan die van de mannelyke. The Hague; Jan van den Bergh 1740. King, Haitung; Bailar, John C.: The health of the clergy: a review of demographic literature. In: Demography 6, 1 (1969), pp. 27–43. Kirschbaum, C.; Wüst, S.; Faig, H. G.; Hellhammer, D. H.: Heritability of cortisol responses to human corticotropin-releasing hormone, ergometry, and psychological stress in humans. In: The Journal of Clinical Endocrinology & Metabolism 75, 6 (1992), pp. 1526–1530. Klassen, Sherri: Greying in the cloister: the Ursuline life course in eighteenth-century France. In: Journal of Women’s History 12, 4 (2001), pp. 87–112. Kruger, Daniel J.; Nesse, Randolph M.: Sexual selection and the male:female mortality ratio. In: Evolutionary Psychology 2 (2004), pp. 66–85. Kruger, Daniel J.; Nesse, Randolph M.: An evolutionary life-history framework for understanding sex differences in human mortality rates. In: Human Nature 17, 1 (2006), pp. 74–97. Kruger, Daniel J.; Nesse, Randolph M.: Economic transition, male competition, and sex differences in mortality rates. In: Evolutionary Psychology 5, 2 (2007), pp. 411–427. Lang, E.; Arnold, K.; Kupfer, P.: Frauen werden älter. Biologische, medizinische und soziologische Ursachen. In: Zeitschrift für Gerontologie 27, 1 (1994), pp. 10–15. Langford, C. M.: Sex differentials in mortality in Sri Lanka: changes since the 1920s. In: Journal of Biosocial Science 16 (1984), pp. 399–410. Leigh, J. P.: Odds ratios of work-related deaths in United States workers. In: British Journal of Industrial Medicine 45, 3 (1988), pp. 158–166. Lemaire, Jean: Why do females live longer than males? In: North American Actuarial Journal 6, 4 (2002), pp. 21–37. Leviatan, Uri; Cohen, Jiska: Gender differences in life expectancy among Kibbutz members. In: Social Science & Medicine 21, 5 (1985), pp. 545–551. Liu, Yan; Arai, Asuna; Kanda, Koji; Lee, Romeo B.; Glasser, Jay; Tamashiro, Hiko: Gender gaps in life expectancy: generalized trends and negative associations with development indices in OECD countries. In: European Journal of Public Health 23, 4 (2013), pp. 563– 568. Lopez, Alan D.: The sex mortality differential in developed countries. In: Lopez, Alan D.; Ruzicka, Lado T. (Ed.): Sex differences in mortality: trends, determinants and consequences. Canberra; Australian National University Press 1983, pp. 53–120. Lopez, Alan D.: Sex differentials in mortality. In: WHO Chronicle 38, 5 (1984), pp. 217–224. Luy, Marc: Warum Frauen länger leben. Erkenntnisse aus einem Vergleich von Kloster- und Allgemeinbevölkerung. Wiesbaden; Bundesinstitut für Bevölkerungsforschung 2002. Luy, Marc: Causes of male excess mortality: insights from cloistered populations. In: Population and Development Review 29, 4 (2003a), pp. 647–676.

The impact of biological factors on sex differences in life expectancy

43

Luy, Marc: Warum Frauen länger leben – wird ein Vergleich der Sterblichkeit von Klosterund Allgemeinbevölkerung durch Bildungsgrad und Missionstätigkeit der Ordnungsmitglieder beeinflusst? In: Zeitschrift für Bevölkerungswissenschaft 28, 1 (2003b), pp. 5–35. Luy, Marc: Unnatural deaths among nuns and monks: the biological force behind male external cause mortality. In: Journal of Biosocial Science 41, 6 (2009), pp. 831–844. Luy, Marc: Lebenserwartung in Deutschland. Available at www.lebenserwartung.info. (Data/ Information downloaded at 01.02.2015). Luy, Marc; Di Giulio, Paola: Der Einfluss von Verhaltensweisen und Lebensstilen auf die Mortalitätsdifferenzen der Geschlechter. In: Gärtner, Karla; Grünheid, Evelyn; Luy, Marc (Ed.): Lebensstile, Lebensphasen, Lebensqualität. Interdisziplinäre Analysen von Gesundheit und Sterblichkeit aus dem Lebenserwartungssurvey des BiB. Wiesbaden; VS Verlag für Sozialwissenschaften 2005, pp. 365–392. Luy, Marc; Di Giulio, Paola: The impact of health behaviors and life quality on gender differences in mortality. In: Geppert, Jochen; Kühl, Jutta (Ed.): Gender und Lebenserwartung. Bielefeld; Kleine 2006, pp. 113–147. Luy, Marc; Gast, Katrin: Do women live longer or do men die earlier? Reflections on the causes of sex differences in life expectancy. In: Gerontology 60, 2 (2014), pp. 143–153. Luy, Marc; Wegner-Siegmundt, Christian: The impact of smoking on gender differences in life expectancy: more heterogeneous than often stated. In: European Journal of Public Health 25, 4 (2015), pp. 706–710. Luy, Marc; Zielonke, Nadine: Die geschlechtsspezifischen Sterblichkeitsunterschiede in Westund Ostdeutschland unter besonderer Berücksichtigung der kriegsbedingten Langzeitfolgen auf die Kohortenmortalität. In: Cassens, Insa; Luy, Marc; Scholz, Rembrandt D. (Ed.): Die Bevölkerung in Ost- und Westdeutschland. Demografische, gesellschaftliche und wirtschaftliche Entwicklungen seit der Wende. Wiesbaden; VS-Verlag für Sozialwissenschaften 2009, pp. 169–198. Lyon, Joseph L.; Nelson, Steven: Mormon health. In: Dialogue: A Journal of Mormon Thought 12, 3 (1979), pp. 84–96. Madigan, Francis C.: The differential mortality of the sexes, 1900–1954: Cultural and biological factors in the diverging life chances of American men and women. Dissertation for the Degree of Doctor of Philosophy in the Department of Sociology and Anthropology of the University of North Carolina. Chapel Hill; University of North Carolina 1956. Madigan, Francis C.: Are sex mortality differences biologically caused? In: The Milbank Memorial Fund Quarterly 35, 2 (1957), pp. 202–223. Madigan, Francis C.; Vance, Rupert B.: Differential sex mortality: a research design. In: Social Forces 35, 3 (1957), pp. 193–199. Manton, Kenneth G.: Gender differences in the cross-sectional and cohort age dependence of cause-specific mortality: the United States, 1962 to 1995. In: Journal of Gender-specific Medicine 3, 4 (2000), pp. 47–54. Martelin, Tuija; Mäkelä, Pia; Valkonen, Tapani: Contribution of deaths related to alcohol or smoking to the gender difference in life expectancy. Finland in the early 1990s. In: European Journal of Public Health 14, 4 (2004), pp. 422–427. Matsuzawa, Yuji; Shimomura, Iichiro; Nakamura, Tadashi; Keno, Yoshiaki; Tokunaga, Katsuto: Pathophysiology and pathogenesis of visceral fat obesity. In: Diabetes Research and Clinical Practice 24, Supplement (1994), pp. S111-S116. McCartney, Gerry; Mahmood, Lamia; Leyland, Alastair H.; Batty, G. David; Hunt, Kate: Contribution of smoking-related and alcohol-related deaths to the gender gap in mortality: evidence from 30 European countries. In: Tobacco Control 20, 2 (2011), pp. 166–168. McDonough, Peggy; Williams, David R.; House, James S.; Duncan, Greg J.: Gender and the socioeconomic gradient in mortality. In: Journal of Health and Social Behavior 40 (1999), pp. 17–31. Medalia, Carla; Chang, Virginia W.: Gender equality, development, and cross-national sex gaps in life expectancy. In: International Journal of Comparative Sociology 52, 5 (2011), pp. 371–389.

44

Marc Luy

Miller, G. H.; Gerstein, Dean R.: The life expectancy of nonsmoking men and women. In: Public Health Reports 98, 4 (1983), pp. 343–349. Moore, Sarah L.; Wilson, Kenneth: Parasites as a viability cost of sexual selection in natural populations of mammals. In: Science 297, 5589 (2002), pp. 2015–2018. Moultrie, Tom; Dorrington, Rob; Hill, Allan; Hill, Kenneth; Timæus, Ian; Zaba, Basia: Tools for demographic estimation. Paris; IUSSP 2013. Nakamura, Tadashi; Tokunaga, Katsuto; Shimomura, Iichiro; Nishida, Makoto; Yoshida, Shingo; Kotani, Kazuaki; Islam, A. H. M. Waliul; Keno, Yoshiaki; Kobatake, Takashi; Nagai, Yoshiyuki; Fujioka, Shigenori; Tarui, Seiichiro; Matsuzawa, Yuji: Contribution of visceral fat accumulation to the development of coronary artery disease in non-obese men. In: Atherosclerosis 107, 2 (1994), pp. 239–246. Nathanson, Constance A.: Sex differences in mortality. In: Annual Review of Sociology 10 (1984), pp. 191–213. Nathanson, Constance A.: Mortality and the position of women in developed countries. In: Lopez, Alan D.; Valkonen, Tapani; Caselli, Graziella (Ed.): Adult mortality in developed countries. Oxford; Oxford University Press 1995, pp. 135–157. Nathanson, Constance A.; Lopez, Alan D.: The future of sex mortality differentials in industrialized countries: a structural hypothesis. In: Population Research and Policy Review 6, 2 (1987), pp. 123–136. Nielsen, Heber C.: Androgen receptors influence the production of pulmonary surfactant in the testicular feminization mouse fetus. In: Journal of Clinical Investigation 76, 1 (1985), pp. 177–181. Nobile, Annunziata: Male excess mortality between biology and culture. In: Pinnelli, Antonella; Racioppi, Filomena; Rettaroli, Rosella (Ed.): Genders in the life course. Demographic issues. Dordrecht; Springer 2007, pp. 249–281. Nurminen, Markku; Karjalainen, Antti: Epidemiologic estimate of the proportion of fatalities related to occupational factors in Finland. In: Scandinavian Journal of Work, Environment & Health 27, 3 (2001), pp. 161–213. Owens, Ian P. F.: Sex difference in mortality rate. In: Science 297 (2002), pp. 2008–2009. Pampel, Fred C.: Cigarette use and the narrowing sex differential in mortality. In: Population and Development Review 28, 1 (2002), pp. 77–104. Payne, Sarah: ‘Smoke like a man, die like a man’?: A review of the relationship between gender, sex and lung cancer. In: Social Science & Medicine 53 (2001), pp. 1067–1080. Pongou, Roland: Why is infant mortality higher in boys than in girls? A new hypothesis based on preconception environment and evidence from a large sample of twins. In: Demography 50, 2 (2013), pp. 421–444. Poulain, Michel; Pes, Gianni; Salaris, Luisa: A population where men live as long as women: Villagrande Strisaili, Sardinia. In: Journal of Aging Research (2011), ID 153756. Pressat, Roland: Surmortalité biologique et surmortalité sociale. In: Revue Française de Sociologie 14 (numéro spécial) (1973), pp. 103–110. Preston, Samuel H.: Mortality patterns in national populations. With special reference to recorded causes of death. New York; Academic Press 1976. Promislow, Daniel: Costs of sexual selection in natural populations of mammals. In: Proceedings of the Royal Society of London: Biological Sciences 247, 1320 (1992), pp. 203–210. Promislow, Daniel: Mate choice, sexual conflict, and evolution of senescence. In: Behavior Genetics 33, 2 (2003), pp. 191–201. Proteggente, Anna R.; England, Timothy G.; Rehman, Almas; Rice-Evans, Catherine A.; Halliwell, Barry: Gender differences in steady-state levels of oxidative damage to DNA in healthy individuals. In: Free Radical Research 36, 2 (2002), pp. 157–162. Puck, J. M.; Willard, H. F.: X inactivation in females with X-linked disease. In: New England Journal of Medicine 338, 5 (1998), pp. 325–328. Ram, Bali: Sex differences in mortality as a social indicator. In: Social Indicators Research 29, 1 (1993), pp. 83–108.

The impact of biological factors on sex differences in life expectancy

45

Retherford, Robert D.: The changing sex differential in mortality. Westport, London; Greenwood Press 1975. Rockstein, Morris; Lieberman, Harry M.: A life table for the common house fly, Musca domestica. In: Gerontologia 3 (1959), pp. 23–36. Rogers, Richard G.; Hummer, Robert A.; Nam, Charles B.: The sex differential in mortality. In: Rogers, Richard G.; Hummer, Robert A.; Nam, Charles B. (Ed.): Living and dying in the USA. Behavioral, health, and social differentials of adult mortality. San Diego et al.; Academic Press 2000, pp. 31–51. Samaras, Thomas T.; Storms, Lowell H.; Elrick, Harold: Longevity, mortality and body weight. In: Ageing Research Reviews 1, 4 (2002), pp. 673–691. Scheinfeld, Amram: The mortality of men and women. In: Scientific American 198, 2 (1958), pp. 22–27. Seifarth, Joshua E.; McGowan, Cheri L.; Milne, Kevin J.: Sex and life expectancy. In: Gender Medicine 9, 6 (2012), pp. 390–401. Sibai, Abla M.; Fletcher, Astrid, Armenian, Haroutune K.: Variations in the impact of longterm wartime stressors on mortality among the middle-aged and older population in Beirut, Lebanon, 1983–1993. In: American Journal of Epidemiology 154, 2 (2001), pp. 128– 137. Simons-Morton, Bruce G.; Farhat, Tilda; ter Bogt, Tom F. M.; Hublet, Anne; Kuntsche, Emmanuel; Gabhainn, Saoirse Nic; Godeau, Emmanuelle; Kokkevi, Anna; the Hbsc Risk Behaviour Focus Group: Gender specific trends in alcohol use: cross-cultural comparisons from 1998 to 2006 in 24 countries and regions. In: International Journal of Public Health 54, Supplement 2 (2009), pp. S199-S208. Smith, David W. E.: Is greater female longevity a general finding among animals? In: Biological Reviews 64 (1989), pp. 1–12. Smith, David W. E.; Warner, Huber R.: Does genotypic sex have a direct effect on longevity? In: Experimental Gerontology 24 (1989), pp. 277–288. Speakman, John R.: Sex- and age-related mortality profiles during famine: testing the ‘body fat’ hypothesis. In: Journal of Biosocial Science 45, 6 (2013), pp. 823–840. Staetsky, Laura; Hinde, Andrew: Unusually small sex differentials in mortality of Israeli Jews: what does the structure of causes of death tell us? In: Demographic Research 20, 11 (2009), pp. 209–252. Stillion, Judith M.: Death and the sexes: An examination of differential longevity, attitudes, behaviors, and coping skills. Washington, D. C.; Hemisphere Publishing Corporation 1985. Stindl, Reinhard: Tying it all together: telomeres, sexual size dimorphism and the gender gap in life expectancy. In: Medical Hypotheses 62, 1 (2004), pp. 151–154. Struyck, Nicolaas: Inleiding tot de Algemeene Geographie, benevens eenige sterrekundige en andere Verhandelingen. Amsterdam; Isaak Tirion 1740. Tabutin, Dominique: La surmortalité féminine en Europe avant 1940. In: Population 33, 1 (1978), pp. 121–148. Théré, Christine; Rohrbasser, Jean-Marc: Facing death in the early days of life: inequality between the sexes in enlightenment demographic thought. In: History of the Family 11, 4 (2006), pp. 199–210. Thorp, David B.; Haist, James V.; Leppard, Jennifer; Milne, Kevin J.; Karmazyn, Morris; Noble, Earl G.: Exercise training improves myocardial tolerance to ischemia in male but not in female rats. In: American Journal of Physiology – Regulatory, Integrative and Comparative Physiology 293, 1 (2007), pp. R363-R371. Timio, M.; Saronio, P.; Venanzi, S.; Gentili, S.; Verdura, C.; Timio, F.: Blood pressure in nuns in a secluded order: a 30-year follow-up. In: Mineral and Electrolyte Metabolism 25, 1–2 (1999), pp. 73–79. Tower, John: Sex-specific regulation of aging and apoptosis. In: Mechanisms of Ageing and Development 127, 9 (2006), pp. 705–718.

46

Marc Luy

Trifunovic, A.; Larsson, N. G.: Mitochondrial dysfunction as a cause of ageing. In: Journal of Internal Medicine 263, 2 (2008), pp. 167–178. Trivers, Robert L.: Parental investment and sexual selection. In: Campbell, Bernard (Ed.): Sexual selection and the descent of man, 1871–1971. Chicago; Aldine Publishing 1972, pp. 136–179. Trovato, Frank; Lalu, N. M.: Narrowing sex differentials in life expectancy in the industrialized world: early 1970’s to early 1990’s. In: Social Biology 43, 1–2 (1996), pp. 20–37. Vallin, Jacques: Can sex differentials in mortality be explained by socio-economic mortality differentials? In: Lopez, Alan D.; Caselli, Graziella; Valkonen, Tapani (Ed.): Adult mortality in developed countries: from description to explanation. Oxford; Clarendon Press 1995, pp. 179–200. Verbrugge, Lois M.: Gender and health: an update on hypotheses and evidence. In: Journal of Health and Social Behavior 26, 3 (1985), pp. 156–182. Verbrugge, Lois M.; Wingard, Deborah L.: Sex differentials in health and mortality. In: Women & Health 12, 2 (1987), pp. 103–145. Waldron, Ingrid: Why do women live longer than men? Part I. In: Journal of Human Stress 2, 1 (1976), pp. 2–13. Waldron, Ingrid: Type A behaviour and coronary heart disease in men and women. In: Social Science & Medicine 12B (1978), pp. 167–170. Waldron, Ingrid: The role of genetic and biological factors in sex differences in mortality. In: Lopez, Alan D.; Ruzicka, Lado T. (Ed.): Sex differentials in mortality: trends, determinants and consequences. Canberra; Australian National University Press 1983a, pp. 141–164. Waldron, Ingrid: Sex differences in human mortality: the role of genetic factors. In: Social Science & Medicine 17, 6 (1983b), pp. 321–333. Waldron, Ingrid: What do we know about causes of sex differences in mortality? A review of the literature. In: Population Bulletin of the United Nations 18 (1985), pp. 59–76. Waldron, Ingrid: The contribution of smoking to sex differences in mortality. In: Public Health Reports 101, 2 (1986), pp. 163–173. Waldron, Ingrid: Effects of labor force participation on sex differences in mortality and morbidity. In: Frankenhaeuser, Marianne; Lundberg, Ulf; Chesnay, Margaret (Ed.): Women, work, and health. New York, London; Plenum Press 1991, pp. 17–38. Waldron, Ingrid: Contributions of biological and behavioral factors to changing sex differences in ischaemic heart disease mortality. In: Lopez, Alan D.; Valkonen, Tapani; Caselli, Graziella (Ed.): Adult mortality in developed countries. Oxford; Oxford University Press 1995, pp. 161–178. Waldron, Ingrid; Johnston, Susan: Why do women live longer than men? Part II. In: Journal of Human Stress 2, 2 (1976), pp. 19–30. Wallen, J.; Waitzkin, H.; Stoeckle, J. D.: Physician stereotypes about female and male health and illness: a study of patient’s sex and the informative process during medical interviews. In: Women and Health 4, 2 (1979), pp. 135–146. Wallner, Bernard; Machatschke, Ivo H.: The evolution of violence in men: the function of central cholesterol and serotonin. In: Progress in Neuro-Psychopharmacology & Biological Psychiatry 33, 3 (2009), pp. 391–397. Wells, Jonathan C. K.: Natural selection and sex differences in morbidity and mortality in early life. In: Journal of Theoretical Biology 202 (2000), pp. 65–76. Wiehl, Dorothy G.: Sex differences in mortality in the United States. In: Milbank Memorial Fund Quaterly 16 (1938), pp. 145–155. Wingard, Deborah L.: The sex differential in mortality rates. Demographic and behavioral factors. In: American Journal of Epidemiology 115, 2 (1982), pp. 205–216. Wingard, Deborah L.; Suarez, Lucina; Barrett-Connor, Elizabeth: The sex differential in mortality from all causes and ischaemic heart disease. In: American Journal of Epidemiology 117, 2 (1983), pp. 165–172.

The gender gap in life expectancy in Austria and the change in the working environment (c. 1900–1950) Andreas Weigl Introduction: The data The first half of the twentieth century is of particular interest in the history of the gender gap in life expectancy in Austria, because it not only preludes the huge widening of the gap in the following decades but follows the same trend that accelerated later on. In many aspects the widening of the gap therefore roots in the period between 1900 and 1950. Unfortunately, studies on Austrian mortality decline in the first half of the twentieth century are limited due to a lack of individual census data, which means that they are mainly based on published aggregate data. Fortunately, calculations of life tables for the territory of the later Republic of Austria are available for the census years 1900, 1910, 1934 and for every year from 1947–2012.1 The life expectancies in 1900 and 1910 were calculated by Statistik Austria for the “Alpine lands” within the Habsburg monarchy, including Lower Styria, after 1918 part of Yugoslavia, later Slovenia, South Tyrol, after 1918 part of Italy, and excluding Burgenland, in the Habsburg monarchy a part of Western Hungary.2 Due to the fact that, in terms of social and economic indicators, the populations of these provinces and regions are quite similar to other parts of the Austrian Republic except for the capital Vienna, this bias does not matter much, at least as far as the gender gap in life expectancy is concerned. The general picture of the changes in the gender gap is therefore quite clear. Futhermore, in the case of Vienna I have calculated abridged life tables for 1919/1921 and 1921/1923 which offer additional information on the shortterm effects of World War I.3 Gaining deeper insight into the mortality transition in Austria is a more complicated scientific issue. In the first half of the twentieth century at state level only a crude differentation of causes of death was published by the Austrian Statistical Office from 1914 onwards.4 The data about causes of death includes no information on age distribution. Vienna, the capital, is fortunately an exception. The city’s Statistical Office collected and published a much broader demographic data set, although even in Vienna the published census data of 1920 and 1923 is very limited.5 In the case of Vienna a comparison of 1 2 3 4 5

Statistik Austria, 2014. Findl, 1979, pp. 425–452. Weigl, 2013, p. 69; Weigl, 2000, p. 165. Bundesamt für Statistik, 1923a, p. 93; Bundesamt für Statistik, 1920–1937; Österreichisches Statistisches Landesamt, 1938; Österreichisches Statistisches Zentralamt, 1950– 1952. Bundesamt für Statistik, 1921a; Bundesamt für Statistik, 1921b; Bundesamt für Statistik 1923b.

48

Andreas Weigl

death rates by cause of death and age is possible, but these data represent only 30 % of the entire Austrian population and they hardly represent rural Austria. Additional information could be derived from the statistics of death by profession available for Austria in the mid-1930s and Vienna for the time period c. 1900–1955, unfortunately in the pre-World War I period not cross tabulated with age groups. Before 1914 surveys by co-operative health insurance funds and contemporary social scientists offer some additional information. For contemporary Austrian scientists in the field of social medicine sex differentials in mortality were of almost no importance. In the interwar period the increase of female activity rates initiated a hype in papers and monographs dealing with the link between female health and women at work, which a majority of the authors opposed.6 But beside their conservative attitudes the trend in activity rates did not change, which led to the conclusion, that the gender gap in life expectancy will vanish in the long run, because of the deteriorating effects of female work in factories.7 To summarize, analysis of the gender gap in life expectancy in Austria in the first half of the twentieth century is partly based on incomplete data. Notwithstanding the puzzle this presents, these data show some general trends which might help to understand the widening of the gap until the 1950s and presumably the 1960s and 1970s too. A general survey If we look at the development of the gender gap in Austria (before 1918 in the so-called Austrian hereditary lands, which are geographically more or less congruent with the territory of the later Austrian Republic) in the first half of the twentieth century, its extent in the early twentieth century can be seen as typical of European industrialized countries. Around the year 1910 the gender gap was 3–3.5 years, which is comparable with Germany, Switzerland, France, England and Wales, and also with Denmark, Norway and Iceland. One essential difference to the countries listed was that the overall life expectancy in the Habsburg Monarchy in general, and the Alpine regions in particular, was clearly lower and this was due primarily to the unusually high infant mortality.8 The subsequent decades up until 1950 were marked in Austria by a further increase by around two years in the gap between male and female life expectancy. This development corresponded closely to that in France, while the gap widened less in England and Wales, for instance, or in Switzerland.9 In contrast to the development in Austria as a whole, the gender gap was, at 4.5 years in 1910, much more pronounced in Vienna than in the entire Alpine region. Between 1910 and 1950 it widened by only one year which meant that it moved closer to the Austrian average. 6 7 8 9

Weigl, 2011, p. 113. Niedermeyer, 1949, p. 27. Mayr, 1926, p. 429. United Nations, 1957, pp. 566–571; Statistik Austria, 2014.

49

The gender gap in life expectancy in Austria

The immediate effect of the two world wars on the gender gap can be reconstructed only for Vienna in the case of World War I, and for both Austria and Vienna in the case of World War II. This shows a disproportionate mortality rise for women during the First World War, a development that was reversed in the early 1920s. But the gender gap was nonetheless slightly smaller then than it had been previous to 1914. Fig. 1: Life expectancy in Vienna by sex 1909–12, 1919–21 and 1921–2310

remaining life expectancy at age … year

0

1

20

60

0

1

male

20

60

female

1909/12

43.8

52.0

39.7

12.5

48.4

55.9

44.4

14.5

1919/21

41.4

47.4

35.8

8.2

44.5

49.8

38.9

9.5

1921/23

49.8

56.3

41.4

12.2

53.9

59.5

44.9

13.8

1909/12

-4.6

-3.9

-4.7

-2.0

1919/21

-3.1

-2.4

-3.1

-1.3

1921/23

-4.1

-3.2

-3.5

-1.6

gender gap

A look at the Vienna life tables for the years 1930/1933 and 1949/195111 suggests that this temporary decrease in the gender gap is unlikely to have repeated itself in Vienna as a direct consequence of World War II. It certainly did not re-occur in Austria as a whole as indicated by the life tables for Austria that go back to 1947, a period directly after the war marked by hunger and hardship. Fig. 2: Life expectancy in Austria 1930–33, 1947–50, 196012

at birth

age 60

year

m

f

GG

m

f

GG

1930/33

54.5

58.5

-4.0

14.2

15.4

-1.2

1947

59.4

64.4

-5.0

15.1

17.2

-2.0

1948

60.7

66.0

-5.2

15.7

17.8

-2.0

1949

60.9

65.6

-4.8

15.1

17.1

-1.9

1950

62.2

67.3

-5.1

15.3

17.5

-2.2

1960

66.5

72.8

-6.4

15.5

19.0

-3.5

10 11 12

Weigl, 2013, 69; Weigl, 2000, p. 165. Statistisches Amt der Stadt Wien, 1953a, pp. 8–11. Statistik Austria, 2014.

50

Andreas Weigl

In the late nineteenth century the gender gap was very much a consequence of the uneven infant mortality of boys and girls, but at the eve of World War I the picture changed gradually. Beside the still higher infant mortality of boys the mortality of the higher working ages became more and more important now. In 1909–12 the male/female-mortality-ratio at an age of about 50 peaked at 140 (female mortality = 100).13 Later on in the time period c. 1910–1950, as Fig. 3 shows, it was definitly not infant mortality that caused the widening of the gap. On the contrary: infant mortality decreased considerably more for boys than for girls. Therefore the widening of the gender gap in the remaining life expectancy peaked at age 1, rising to 3.2 years. The plus at birth was only 2.3 years. In general the gap widened substantially for all age groups except the oldest. Fig. 3: Gender Gap in Austria per age, c. 1900–195014 4

3

difference 1900 : 1950 2

1

GG m - f

0

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90

-1

-2

1950

-3

1900 -4

-5

-6

Causes of death According to the wellknown epidemiologic transition model of Abel Omran15 the Austrian population reached the final stage 3 of the transition dominated by diseases of the circulatory system and cancer in the interwar period with some backlashes at the end of and after World War II. During the economic crisis of the 1930s the share of infectious deseases in 1933–34 was about 25 % 13 14 15

Findl, 1979, pp. 442–443. Own calculations based on Statistik Austria, 2014. Omran, 1971, pp. 509–538.

The gender gap in life expectancy in Austria

51

including lung inflamation, 16 % excluding lung inflamation, which could be infectious and non-infectious, supposedly including also hidden cases of lung tuberculosis.16 Unfortunately, mortality data by causes of death are quite patchy for the Republic of Austria with the exception of the 1950s. But even crude comparisons based on abridged ICDs without any age distribution show some interesting trends. At least one simple question could be resolved: the influence of accidents. The sex ratio of these causes of death dropped down from 4.4 to 2.1 between 1914 and 1952.17 This means that the decline of industrial injuries and other casulities reduced the gender gap. As the age-specific Viennese mortality for “injuries and poisoning” indicates this applies to all relevant age groups.18 Another noteworthy factor for the gender gap was death in childbirth. From c. 1910–1950 the age standardized mortality rate of the male population in Vienna decreased by 40.5 %, the female by 44.1 %. By excluding death in childbirth the decrease of the female death rate would be almost the same, 44.2 %.19 It is possible that the specific mortality rate for death in childbirth was slightly higher in other parts of Austria, but hardly to a degree that could change the overall picture. The available aggregate data in Vienna allows the computation of mortality rates for the period between c.1910 and 1950 based on causes of death and age groups. With some restrictions the contemporary ICDs can be translated to relatively “modern” ICDs of 1979.20

16 17 18 19 20

Bundesamt für Statistik, 1923a, p. 93; Bundesamt für Statistik, 1937, p. 22. Bundesamt für Statistik, 1923a, p. 93; Österreichisches Statistisches Zentralamt 1953, 39. Weigl, 2011, p. 122. Weigl, 2011, p. 121. For 1909–1911 the ICDs of Bertillon (1–186) were reassigned to the ICD-Groups 1979: I: 1–9, 11–19, 21–35; II: 39–46; III: 36; V: 56; VI: 60–76; VII: 47, 77–85; VIII: 10, 86–98; IX: 99–118; X: 119–133; XI: 134–141; XII: 142–145; XIII: 48–55, 146–149; XIV: 150; XV: 151–153; XVI: 154, 187–189; XVII: 20, 57–59, 155–186; for 1950–1952 the ICDs 1938 (1–87) were reassigned and recalculated to the ICD-Groups 1979 : I: I – 12; II: II + 32; III: III – 25, 26; IV: IV – 32; VI: VI – 35, 36, 37; VII: VII + 25, 35, 36, 37, 52 * 0,619; VIII: VIII + 12–52 * 0,619; XIII: XIII + 26; see Hohenegger / Weigl, 1988, p. 19.

52

Andreas Weigl

Fig. 4: Age-standardized death rates by selected causes of death, Vienna 1909–11 and 1950–5221

ICD-Group

male

1909/11 Age distribution Infectious diseases Cancer Diseases of the circulatory s. Diseases of the respiratory s. Accidents and adverse effects crude death rate

1951

1910

female

1950/52 1951

1909/11 1951

1910

male female Change 1909/11–1950/2 1950/52 % 1951

1951

407.5 108.0

108.4

310.1

81.0

65.2

-73.4

-79.0

217.0

65.1

340.6

273.1

56.3

286.3

57.0

4.8

387.5 136.2

716.5

487.5

125.5

625.9

84.9

28.4

270.0 148.9

73.6

282.2

122.7

60.1

-72.7

-78.7

87.0

21.9

108.4

42.5

10.1

68.5

24.6

61.2

20.9

18.1

15.8

21.8

15.4

12.9

-24.4

-40.8

The general picture of epidemiologic transition in Vienna in the first half of the twentieth century is quite clear: a massive decline in the mortality of infectious diseases and diseases of the respiratory system for both sexes on the one hand, a massive increase in cancer mortality and mortality of diseases of the circulatory system for the male population on the other hand, while the increase was much smaller on the female side. This was exactly the reason for the widening of the gender gap, although limited to some extent by the fact that the gap between the mortality from accidents, drug abuse and suicide of females and males narrowed. There is much to suggest that the data for Vienna concerning primary causes of death are generally representative of the whole of Austria, even if one considers the regional particularities that definitely existed according to the Atlas of Mortality in Austria by Causes of Death, which has only been available since the 1970s. Fig. 5 shows that the age pattern in the gender gap change that occurred in Vienna between 1910 and 1950 slightly deviated from that of Austria taken as a whole. Apart from the peak in remaining life expectancy at age 1 of approximately 1.4 there is another peak in life expectancy at age 55–60 of 1.5. This means that the deaths that occur towards the end of the working life are particularly significant for the increase of the gender gap.

21

(Statistisches) Jahrbuch Wien 1911–1913; 1952–1953.

53

The gender gap in life expectancy in Austria Fig. 5: Gender Gap in Vienna by age, c. 1910–195022 2

1

0

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80 +

GG m - f

-1 GG 1910 -2

GG 1951 difference GG

-3

-4

-5

-6

As concerns the causes of death, the decreasing age-standardized death rates for infectious and respiratory diseases show no gender-specific differences. The increase in cardiovascular disease and cancer (malignant neoplasms), on the other hand, is much more pronounced in the male population, particularly among the higher age groups. The age-specific death rates due to degenerative disease show that from the age of about 60, cancer deaths were the most conspicuous in the widening of the gender gap. The increase may also be due to the fact that cancer diagnostics were not very advanced at the turn of the century and improved in the interwar period. While, before World War I, male excess mortality from cancer among the higher age groups was still relatively moderate, the gap now widened drastically, reaching a ratio of almost 2 to 1. Although this is not greatly relevant for overall mortality: female excess cancer mortality prevailed before World War I among the age groups of up to around 55, but in the mid-twentieth century this no longer applied to any age group. The pattern was very similar with cardiovascular disease; but here the female excess mortality was restricted to under 40s, while the divergence in the higher age groups was less extreme.23

22 Statistisches Amt der Stadt Wien, 1953a, pp. 8–11. 23 Weigl, 2011, p. 122.

54

Andreas Weigl

The influence of gender specific occupational patterns The strong impact of degenerative disease on the widening of the gender gap points to the influence of (in a wider sense) long-term damage due to occupational processes and gender-specific leisure activities. Occupation-specific death rates therefore offer important additional information on the possible reasons for the widening of the gender gap in the period under observation. This indicator is particular instructive in the first half of the twentieth century when life expectancy was only just above retirement age, which means that diverse living conditions and possible differences in health behaviours during retirement were of relatively little importance. Fig. 6: Occupation-specific death rates of workers, day-labourers, apprentices in Vienna by sex 1909–1124

Industry

Exploitation of earth and stone Metal and machine production Wood processing Textile industry Foodstuffs and beverages, accomodation Clothing industry Other industry Construction Manufacturing and construction total Trade

Occupation-specific death rates m f 18.3 28.2 14.8 111.6 18.7 86.4 17.0 10.1 26.1 19.1 72.1 26.6

26.8 15.1 89.5 79.7

32.9 10.2

n M

f

351

146

3856

1319

1975

789

450

279

889

300

2777

2656

1253

405

2979

953

39.1

22979

13009

10.6

1030

392

In Figure 6 I have calculated occupation-specific death rates according to sex. However, these figures have the problem of correctly matching numerators and denominators of occupational and mortality statistics. But with the help of some additional information, like the age distribution by economic sector and sex, one can gain a relatively clear picture of mortality rates among the workforce. According to that, mortality was clearly higher in the production sector than in trade. Even if one looks at the whole production sector, excess female mortality prevailed in numerous industries, with the exception of the textile and clothing industry which is, however, particularly relevant for the female workforce. Nevertheless, female workers in “heavy” industries influenced 24 Weigl, 2007, p. 229.

The gender gap in life expectancy in Austria

55

overall mortality substantially. Though female excess mortality in some industrial branches might be due to an age factor, but – as the proportion of over 50-year old working women in manufacturing in 1910 was 11 % and that of working men 13 % – it cannot be due to the mortality at a higher working age overall in the production sector.25 Fig. 7: Mortality of members of co-operative health and insurance funds in Vienna by age 1892–190026

Age group

death rate (number of death per 1000 members) male

female

GG

under 16

3.1

3.2

-0.1

16–20

6.0

9.2

-3.2

21–25

7.2

10.1

-2.9

26–30

7.3

9.4

-2.1

31–35

8.9

8.6

0.3

36–40

11.1

10.2

0.9

41–45

14.4

12.2

2.2

46–50

20.3

16.9

3.4

51–55

24.7

18.1

6.6

56–60

32.5

23.3

9.2

61–65

42.3

34.1

8.2

66–70

61.6

50.0

11.6

According to a contemporary study into membership of the major Viennese health insurance schemes for the years 1892 to 1900 – around the year 1900 about 54 % of the entire male and 26 % of the female workforce27 – the result described above can be itemized in terms of age patterns. In contrast to the death rates mentioned earlier, the deaths and the working population in the health insurance statistics refer to the same overall population, the only restriction being that the chronically ill dropped out of the insurance after a fixed period of time, which meant that their deaths were not included in the health insurers’ statistics.28 This fact would, however, only diminish the validity of the gender-specific mortality rates if these insurance cancellations were disproportionately gender-specific and this is unlikely.

25 K. k. statistische Zentralkommission, 1916, pp. 62–65. 26 Rosenfeld, 1905, p. 860. 27 Author’s own calculations based on Rosenfeld, 1905, pp. 822–824; k. k. statistische Zentralkommission, 1904, p. 68. The insurance funds are Bezirkskrankenkasse, Allgemeine Arbeiterkrankenkasse, Genossenschaftskrankenkassen, Gremialkrankenkassen. 28 Rosenfeld, Gesundheitsverhältnisse, p. 863.

56

Andreas Weigl

The statistics of the major Vienna health insurers show excess mortality of young female workers up to the age of 30, a phenomenon that was not restricted to Vienna. In the Austrian half of the Habsburg empire female excess mortality applied up to even the age of 40. Among the older working population, where more deaths occurred, there was excess male mortality. Fig. 8: Mortality of members of co-operative health and insurance funds in Austria (Cisleithania) by age 1891–189529

age group

mortality rate (per 1000) male

female

15–20

5.2

8.1

21–30

6.6

9.2

31–40

9.0

10.0

41–50

14.2

12.4

51–60

23.1

17.1

Overall, before World War I, the evidence points to excess female mortality, certainly in industry and trade, among the younger age groups, in Vienna as well as in the Alpine regions. This disadvantage of working women in manufacturies was more than equalized in overall mortality by sex due to – among other things – low death rates in domestic service – dominated more or less exclusivly by maids – and among the non-active population. In 1910 43 % of the Viennese female adult population (age 16 and above) were housewives and 12 % maids.30 Comparison of age-specific death rates in the overall population with those in health insurance members revealed that the mortality of non-members – in other words, primarily housewives – was clearly lower than that in female health insurance members, especially at a lower age. The same age pattern of the mortality rates existed in the whole “Austrian” part of Austria-Hungary. Especially the mortality rates of young female workers in the textile and tobacco industry were very high. Overall it exceeded the male mortality by about 30 %. According to the expertise of contemporary physicians, female execess-mortality in these industries can be explained by both the very low economic status of the workers and (nicotine-) poisoning.31

29 Prinzing, 1906, p. 492. 30 K. k. statistische Zentralkommission, 1916, p. 67; own calculations. 31 Prinzing, 1906, pp. 492–493.

57

The gender gap in life expectancy in Austria Fig. 9: Comparison of the overall mortality 1891/1900 with the mortality of members of co-operative health and insurance funds by age 1894/190232

age group

mortality rate of overall mortality difference members members – overall mortality m

f

m

f

m

f

15–20

6.0

9.2

5.8

5.8

0.2

3.4

21–25

7.2

10.1

7.0

6.7

0.2

3.4

26–30

7.3

9.4

8.1

7.6

-0.8

1.8

31–35

8.9

8.6

10.3

8.4

-1.4

0.2

36–40

11.1

10.2

14.0

10.1

-2.9

0.1

41–45

14.4

12.2

17.4

11.1

-3.0

1.1

46–50

20.3

16.9

22.0

13.8

-1.7

3.1

51–55

24.7

18.1

27.5

17.3

-2.8

0.8

56–60

32.5

23.3

34.4

23.9

-1.9

-0.6

61–65

42.3

34.1

45.3

32.7

-3.0

1.4

66–70

61.6

50.0

63.0

50.3

-1.4

-0.3

1) monthly average in 1900: 313 401 men, 96 499 women.

In the interwar period the gender gap in the labour force changed profoundly. The 1934 census established occupation-specific death rates in some occupational groups for all of Austria. These figures can be seen as accurate since the census team, in one transaction, allocated occupational codes for the statistics of deaths and of individuals in occupations. With one exception, the occupation-specific death rates now show definite male excess mortality for all age groups. Only the comparison of male physicians and female nurses below the age of 40 showed male mortality to be lower. The picture is similar for Vienna, although in the aggregate data there is no differentiation according to age available.33 In Vienna it was the group of female civil servants that formed the exception. This cannot be explained with the age structure of the workforce but with their level of education. In the 1930s women were still largely underrepresented among academics and high-school graduates in public service in Vienna.

32 Rosenfeld, 1905, pp. 860–863. 33 Weigl, 2011, p. 127.

58

Andreas Weigl

Fig. 10: Occupation specific mortality rates by sex and age in Austria 1933/3434

mortality rate (per 1000) Occupation

status

20–29

30–39

40–49

50–59

60–69

Taylor

self employed, m white collar f

4.7

4.8

8.9

17.4

34.7

2.1

2.7

4.6

10.1

15.6

m

6.2

8.8

13

28.0

58.3

f

Innkeeper

self employed

2.5

3.5

5.3

13.0

29.9

Waiter

white collar

m

3.9

8.2

14.1

22.5

47.7

Waitress

white collar f self employed. w. collar m

3.6

2.2

5.0

5.0

16.7

3.0

4.7

7.9

22.8

40.3

white collar

f

4.5

4.9

6.5

12.3

37.0

m

3.7

2.7

6.1

15.6

63.6

f

3.6

2.5

5.2

8.7

33.8

Physician Nurse teacher. Education, art, entertainment

white collar

Based on the 1951 census and the death statistics it is possible, at least for Vienna, to work out occupation-specific death rates according to age, which were counted according to a standard occupational classification for the postwar period too.35 Again, the rates show definite male excess mortality for all age and occupational groups. And again, the exception lies with the public service – unfortunately without any detailed statistical classification – where the death rates are clearly higher for women. Also – compared with the time before 1914 – comparison between overall mortality and labour force mortality reveals a change in death rates. In contrast to the pre- World War I period, the mortality of the overall female population of working age were higher in all age groups than those of the female workforce in the 1950s. The same is true for working men, but in their case the advantage also applied pre-1914. This means that the “housewife bonus” vanished in the first half of the twentieth century. This supports the view that the severely chronically ill left the workforce – after 1939 not only employees but also workers joined old age, survivors’ and invalidity insurances36 – while this had not been the case before World War I, at least not for female workers under different epidemiological conditions – particulary the importance of tuberculosis.

34 Bundesamt für Statistik, 1937, p. 23. 35 See Fig. 11. 36 Peissl, 1994, p. 136.

59

The gender gap in life expectancy in Austria Fig. 11: Occupation-specific death rates by age, 1951–5337

Occupation

occupation-specific mortality rates 1)2) all under 50

agriculture and forestry Industry catering, hotel ind. commercial jobs, clerks Transport financial services cleaner etc. health care, welfare work teaching, education, art, entertainment law and business consultant public services home services unskilled labour all

Weigl, 2011, p. 131.

65 +

13.5

3.0

13.9

23.5

80.5

f

5.2

1.4

2.6

8.4

35.5

m

8.3

2.4

12.8

24.3

59.4

f

2.5

1.3

4.0

10.0

28.7

m

17.7

5.0

20.1

45.5

76.1

f

5.7

1.8

5.9

13.7

50.5

m

10.1

2.9

12.8

24.5

57.9

f

3.8

1.3

5.3

12.8

41.4

m

6.8

2.9

12.1

28.8

55.1

f

3.9

1.9

4.8

3)

84.6

m

7.2

1.5

10.7

23.1

51.0

f

4.1

2.2

10.8

1)

1)

m

7.6

2.8

8.1

26.1

32.1

f

1.7

0.6

2.2

6.3

7.8

m

11.4

2.9

15.3

27.0

74.0

f

3.0

1.9

4.5

11.3

19.8

m

9.0

2.0

10.3

23.4

68.2

f

3.7

1.8

5.4

11.3

28.5

m

11.4

2.1

14.7

21.6

38.3

f

1)

1)

1)

1)

1)

m

8.7

3.0

16.6

30.7

75.2

f

20.6

5.2

21.9

42.2 160.5

m

19.9

1)

1)

1)

42.1

7.6

1.9

7.3

15.1

35.4

m

12.8

5.3

18.7

29.9

37.3

f

8.5

5.8

15.5

20.1

34.2

m

9.1

2.7

13.5

25.9

59.2

f

4.4

1.7

5.8

12.6

38.1

f

1) n < 10. Number of deaths total male: 13 308, female: 4 058.

37

50–59 60–64

m

60

Andreas Weigl

Fig. 12: Comparison of the overall mortality of the mortality of employed persons in Vienna 1951–5338

1951/53 age

employed

overall

difference 1)

m

f

m

f

m

f

under 18

0.9

0.4

2.8

2.3

-2.0

-1.9

18–29

1.4

1.0

1.5

1.1

-0.1

-0.1

30–49

3.5

2.2

4.0

2.9

-0.5

-0.7

50–59

13.5

5.8

16.1

8.0

-2.7

-2.3

60–64

25.9

12.6

30.6

15.7

-4.7

-3.1

65 +

59.2

38.1

81.3

60.8

-22.1

-22.7

1) Employed – overall mortality rate.

The shift to white collar jobs If one now relates the demographic results to the economic change in Austria during the first half of the twentieth century, the process of tertiarization seems to lend itself, at first, as an explanatory factor. But this was not the case in the interwar or postwar periods. During these periods there was a rise in people working in the production sector in Austria while the service sector fluctuated between 27 and 32 %.39 In Vienna, the proportion of production workers stagnated in the pre- World War I era and the interwar period at a level of 50 % and decreased slightly after World War II. Even in the capital city there was no tertiarization to start with.40 Sectoral shifts can therefore not be responsible for the gender gap in the mortality of the working population. Things were different with regard to people’s occupational status. Between 1910 and 1951 the proportion of employees and civil servants in Austria rose from 7.0 % to 19.8 %. This was not due to a shift towards sectors or jobs with a high proportion of employees, but to the expansion of the number of employees following the Austrian Employment Act of 1921 and the Employees Pension Act of 1926.41 As can be demonstrated in the case of Vienna, working women benefitted disproportionately from these advancements in occupational status. While the ratio of employees among the male workforce rose from 14.9 % to 36.7 %, that of the female workforce came close when, around 1950, the proportion rose abruptly from 7.1 % to 34.9 %.42 This was relevant for the gender gap because the death rates for both sexes were most favourable among employees. Furthermore excess male mortality in Vienna was 38 39 40 41 42

Weigl, 2011, p. 134. Österreichisches Statistisches Zentralamt, 1953, p. 48; Mesch / Weigl, 2012, p. 7. Statistisches Amt der Stadt Wien, 1953b, p. 72. Mesch / Weigl, 2011, pp. 4–5. Own calculations based on k. k.statistische Zentralkommission, 1916, p. 54 and Magistrat der Stadt Wien, 1954, p. 11.

The gender gap in life expectancy in Austria

61

more pronounced in employees than in labourers. In the mid-1930s the male/ female-ratio in the mortality rates for the self-employed was 10.9 to 4.1, for white collar employees and civil servants 4.8 to 1.8, for labourers (and apprentices) 4.9 to 2.3.43 At the beginning of the 1950s a similar mortality pattern by sex occurred.44 These findings seem to support the view that the influences occupation had on mortality differed for male and female employees. Although – or maybe even because – female employees, up to the 1950s, rarely reached senior positions and their material living conditions remained precarious due to discrimination in income,45 they benefited disproportionately from their status and the corresponding improvements with regard to physical stress and legal privileges. The raising of qualification levels in wage-earning women, on the other hand, did not have much influence on survival rates in the first half of the twentieth century. In 1951 only 5.9 % of women over 14 in Vienna had secondary school qualifications and only 0.9 % had an academic degree.46 Despite these matching results for sex differences in occupational death rates, their explanatory power should not be exaggerated. Since, in the 1950s, life expectancy at age 20 in Austria was close to 69 for men and 72 for women47, it is necessary to slightly qualify the effects of the working life on the gender gap around the mid-twentieth century, seeing that around the year 1950 37 % of men and 10 % of women died during their working life.48 Additional sources are therefore required to strenghten or weaken the argument. Unfortunately, data on gender specific living styles in Austria before the 1960s are patchy. One exception is a survey of a contemporary researcher, the merchant, economist and politician Alfred von Lindheim. His study provides interesting insights into the health state of very old workers in closed geriatric care in Vienna before World War I.49 Of the interviewees in 1908 the overwhelming majority lived in homes for the elderly in Vienna. Their living and health conditions were statistically registered with the support of the respective director’s offices and the City Council on the basis of a standardized questionnaire. Participants only included persons capable of answering the questions on their own. The results which I refer to are founded on the analysis of a partial sample of 601 persons who spent the evening of their lives in various homes for the elderly in Vienna. This sample corresponds to the age group 80 years and above representing 8.7 % of the male and 7.1 % of the female population of 1910. All in all, the production sector is overrepresented. The living habits of the interviewed partly showed great gender-specific differences. It did not come as a surprise that among men a diet dominated by meat 43 Author’s own calculations based on Statistisches Jahrbuch Wien 1937, p. 37; Bundesamt für Statistik, 1935, pp. 18–30. 44 Author’s own calculation based on Statistisches Jahrbuch der Stadt Wien, 1952, p. 55; 1953, p. 13, 315; 1954, p. 366–367. 45 Appelt, 1985, pp. 41–120; Peissl, 1994, pp. 152–155. 46 Author’s own calculations based on Statistisches Amt der Stadt Wien, 1953b, p. 124. 47 Österreichisches Statistisches Zentralamt, 1953, p. 24. 48 Author’s own calculations based on Statistisches Jahrbuch der Stadt Wien, 1952, p. 55; 1953, p. 315; 1954, p. 366 f. 49 Lindheim, 1909.

62

Andreas Weigl

was more common than among women. Not least on account of the bad water quality before the opening of the Vienna Spring Water Ducts both genders frequently consumed alcoholic beverages. An important difference existed though with regard to smoking. Whereas smoking was commonplace among almost nine out of ten men, this held true only for 10 % of the women.50 With a different methological approach a recently published modern study by Susanne Hoffmann entitled “Did work make men sick” also offers additional information. The study was based on 155 unpublished popular autobiographies from the German speaking countries, of which one third were born in the Austrian hereditary lands or the Republic of Austria.51 The results fit in very well with the demographic analysis. They reveal that in the interwar period it was not what could be called “occupational diseases” but a work-centered lifestyle that was the core risk factor for the male population born at the turn of the century and the following two decades. As Hoffmann showed for the male cohorts born in the 1920s and 1930s that this workaholic lifestyle lost its importance because they were increasingly reflecting the conflict between work and health.52 Conclusion At the eve of World War I in Austria, particulary in Vienna, all age-specific death rates of the male population were above the female level. This, however, was not the case with the small group of female workers in manufacturing and construction. Especially young female workers faced higher death risks than the male workers of the same age. Lung tuberculosis played a major role in these differences in mortality.53 In contrast to that the mortality of maids and housewives was below the average and contributed to the overall gender gap in favor of the female population. Female excess mortality at working age largely disappeared in the time between the two World Wars. In the early 1950s, it played a marginal part only in the small group of older employees actively engaged in public service. Occupation-specific death rates for particular occupations in the 1930s show that this was mostly due to the fact that less qualified women also benefited disproportionately from the change in the working world as a result of the expansion of the employee status. The long-term effect of this occupational advancement becomes apparent when compared with the situation in the early 1980s. Around the year 1980, female civil servants and female non-manual workers had the lowest death risk in Austria, while skilled as well as unskilled female labourers were most at risk at a young age.54 50 51 52 53 54

Weigl, 2007, pp. 234–237. Hoffmann, 2010, p. 49. Hoffmann, 2011, pp. 140–167. Johannsson, 1991, p. 159. Doblhammer / Rau / Kytir, 2005, p. 475.

The gender gap in life expectancy in Austria

63

These findings disprove to some extent what could be called the “double burden thesis” in social history. This thesis argues that the rise of the “respectable working class family” causes a double and triple burden for working class women. In addition to physical demands at the workplace they had to cope with increased standards of homework and child care.55 There can be no doubt that this double burden existed, but nothing seems to point to a deterioration in the interwar period and during the 1950s. A reduction of distress due to the fall of the birth rate and improving working and living conditions should be taken in consideration too. Therefore the overall picture is that of a widening of the gender gap in life exptectancy in Austria in the first half of the twentieth century. Bibliography Appelt, Erna: Von Ladenmädchen, Schreibfräulein und Gouvernanten. Die weiblichen Angestellten Wiens zwischen 1900 und 1934, Vienna 1985. Bundesamt für Statistik (ed.): Statistisches Handbuch für die Republik Österreich 1–17, Vienna 1920–1937. Bundesamt für Statistik (ed.): Ergebnisse der außerordentlichen Volkszählung vom 31. Jänner 1920. Alter und Familienstand, Wohnparteien. Vienna 1921a. Bundesamt für Statistik (ed.): Ergebnisse der außerordentlichen Volkszählung vom 31. Jänner 1920. Endgültige Ergebnisse samt Nachtragszählungen. Vienna 1921b. Bundesamt für Statistik (ed.): Die Bewegung der Bevölkerung in den Jahren 1914 bis 1921. Vienna 1923a. Bundesamt für Statistik (ed.): Vorläufige Ergebnisse der Volkszählung vom 7. März 1923. Vienna 1923b. Bundesamt für Statistik (Bearb.), Die Ergebnisse der Volkszählung vom 22. März 1934, Wien, Vienna 1935. Doblhammer, Gabriele / Rau, Roland / Kytir, Josef: Trends in educational and occupational differentials in all-cause mortality in Austria between 1981/82 and 1991/92. In: Wiener klinische Wochenschrift 117 (2005), pp. 468–479. Ehmer, Josef: Frauenarbeit und Arbeiterfamilie in Wien. Vom Vormärz bis 1934. In: Geschichte und Gesellschaft 7 (1981), pp. 438–473. Findl, Peter: Mortalität und Lebenserwartung in den österreichischen Alpenländern im Zeitalter der Hochindustrialisierung (1868 bis 1912). In: Österreichisches Statistisches Zentralamt (ed.): Geschichte und Ergebnisse der zentralen amtlichen Statistik in Österreich 1829–1979 (Beiträge zur österreichischen Statistik 550, 550a), Vienna 1979, pp. 425–452, Tabellenanhang, pp. 33–50. Gruber, Helmut: Red Vienna. Experiment in Working-Class Culture 1919–1934, New York-Oxford 1991. Hoffmann, Susanne: Gesunder Alltag im 20. Jahrhundert? Geschlechterspezifische Diskurse und gesundheitsrelevante Verhaltensstile in deutschsprachigen Ländern, Stuttgart 2010. Hoffmann, Susanne: Machte Arbeit Männer krank? Erwerbsarbeit, Männlichkeit und Gesundheit im 20. Jahrhundert. In: Dinges, Martin / Weigl, Andreas (ed.): Gesundheit und Geschlecht (Österreichische Zeitschrift für Geschichtswissenschaften 22/2), Innsbruck-Vienna-Bolzano 2011, pp. 140–167. Hohenegger, Martin, Weigl, Andreas: Aspekte der Mortalität und Morbidität der Wiener Bevölkerung. In: Statistische Mitteilungen der Stadt Wien 1988/2, pp. 8–19. 55 Ehmer, 1981, pp. 464–469; Gruber, 1991, pp. 151–152.

64

Andreas Weigl

Ryan Johannsson, S. Ryan: Welfare, mortality and gender. Continuity and change in explanations for male/female mortality differences over three centuries. In: Continuity and Change 6 (1991), pp. 135–177. k. k. statistische Zentralkommission (ed.): Berufsstatistik nach den Ergebnissen der Volkszählung vom 31. Dezember 1900 in den im Reichsrate vertretenen Königreichen und Ländern: Niederösterreich (Österreichische Statistik 66/2), Vienna 1904. k. k. statistische Zentralkommission (ed.): Berufsstatistik nach den Ergebnissen der Volkszählung vom 31. Dezember 1910 (Österreichische Statistik NS 3/2), Vienna 1916. Lindheim, Alfred von: Saluti sencetutis. Die Bedeutung der menschlichen Lebensdauer im modernen Staate. Eine sozial-statistische Untersuchung, Leipzig-Vienna 1909. Magistrat der Stadt Wien, MA 66–102/54: Die wichtigsten Ergebnisse der Volkszählung vom 1. Juni 1951 in den nach dem Gebietsänderungsgesetz bei Wien verbleibenden und in den an Niederösterreich abgetretenen Gebieten. Mayr, Georg von: Statistik und Gesellschaftslehre, sec. ed., Tübingen 1926. Mesch, Michael, Weigl, Andreas: Angestellte, Beamte und der Wandel der Beschäftigtenstruktur in Österreich in der ersten Hälfte des 20. Jahrhunderts (Materialien zu Wirtschaft und Gesellschaft 115), Vienna 2012. Mesch, Michael, Weigl, Andreas: Angestellte und Tertiärisierung in Österreich 1910–51. In: Wirtschaft und Gesellschaft 37 (2011), pp. 95–138. Österreichisches Statistisches Landesamt (ed.): Statistisches Jahrbuch für Österreich 1938, Vienna 1938. Österreichisches Statistisches Zentralamt (ed.): Statistisches Handbuch für die Republik Österreich 1–3, Vienna 1950–1952. Österreichisches Statistisches Zentralamt (ed.): Ergebnisse der Volkszählung vom 1. Juni 1951 (Volszählungsergebnisse 1951 Heft 14), Vienna 1953. Omran, Abel R.: The Epidemiologic Transition. A Theory of the Epidemiology of Population Change. In: Milbank Memorial Fund Quarterly 49/1 (1971), pp. 509–538. Peissl, Walter: Das „bessere“ Proletariat. Angestellte im 20. Jahrhundert (Studien zur Gesellschafts- und Kulturgeschichte 4), Vienna 1994. Prinzing, Friedrich: Handbuch der medizinischen Statistik. Jena 1906. Siegfried Rosenfeld, Die Gesundheitsverhältnisse der Wiener Arbeiterschaft. In: Statistische Monatsschrift NF 10 (1905), pp. 725–753, 821–863, 881–914. Statistik Austria: Sterbetafeln 1868/71 bis 2010/12 nach dem Geschlecht; Jährliche Sterbetafeln 1947 bis 2013 für Österreich http://www.statistik.at/web_de/statistiken/bevoelkerung/sterbetafeln/index.html (15.7.2014). Statistisches Amt der Stadt Wien (ed.): Wiener Sterbetafeln (Mitteilungen aus Statistik und Verwaltung der Stadt Wien Jg. 1953a Sonderheft 1). Statistisches Amt der Stadt Wien: Die Häuser-, Wohnungs- und Volkszählung in Wien vom 1. Juni 1951 (Mitteilungen aus Statistik und Verwaltung der Stadt Wien Jg. 1953 Sonderheft 3), Vienna 1953b. Statistisches Jahrbuch der Stadt Wien 27–29 (1909–1911), Vienna 1911–1913; NS 11–12 (1950–1951); Jahrbuch der Stadt Wien 1952, Vienna 1952–1953. United Nations, Demographic Yearbook. Vol. 9, New York 1957. Weigl, Andreas: Demographischer Wandel und Modernisierung in Wien (Kommentare zum historischen Atlas von Wien 1), Vienna 2000. Weigl, Andreas: Dank Keuschheit ein langes Leben? Hochbetagte Arbeiterinnen und Arbeiter im Wien der Jahrhundertwende. In: Dinges, Martin (ed.): Männlichkeit und Gesundheit im historischen Wandel ca. 1800 – ca. 2000 (Medizin, Gesellschaft und Geschichte Beiheft 27), Stuttgart 2007, pp. 227–241. Weigl, Andreas: Arbeit, Lebenserwartung, Geschlecht: Wien 1900–1950. In: Dinges, Martin / Weigl, Andreas (ed.): Gesundheit und Geschlecht (Österreichische Zeitschrift für Geschichtswissenschaften 22/2), Innsbruck-Vienna-Bolzano 2011, pp. 112–139. Weigl, Andreas: Eine Stadt stirbt nicht so schnell. Demographische Fieberkurven am Rande des Abgrunds. In: Pfoser, Alfred / Weigl, Andreas (ed.): Im Epizentrum des Zusammenbruchs. Wien im Ersten Weltkrieg, Vienna 2013, pp. 62–71.

The reduction of the gender gap in life expectancy in Austria since the 1980s: an educational phenomenon? Johannes Klotz One of the most stable and widely known findings of applied demography is that, on average, women live longer than men. In Austria in 2013, life expectancy at birth was 83.6 years for females but only 78.5 years for males, which is a difference of just over 5 years. The past 150 years or so have witnessed an unprecedented increase in life expectancy in Austria. Both male and female life expectancy has more than doubled, from less than 40 years to around 80 years. Despite this fundamental change in human mortality risks, a female life expectancy advantage has been observed since the beginnings of Austrian life tables in the nineteenth century (Figure 1). A look at the post-war years reveals an arch-shaped evolution of the gender gap in life expectancy at birth in Austria. The difference between female and male life expectancy increased considerably in the 1950s and 1960s and plateaued at around 7 years in the 1970s and early 1980s. Since then the figure has diminished by 2 years. Figure 1a: Life expectancy at birth by sex, Austria 1868/1871–2013 Males  

Females  

80   70   60   50   40   30   20  

Calendar  year  

2013  

2008  

2003  

1998  

1993  

1988  

1983  

1978  

1973  

1968  

1963  

1958  

1953  

1948  

1943  

1938  

1933  

1928  

1923  

1918  

1913  

1908  

1903  

1898  

1893  

1888  

1883  

1878  

0  

1873  

10   1868  

Life  expectancy  in  years  

90  

66

Johannes Klotz

Figure 1b: Gender gap in life expectancy at birth, Austria 1868/71–2013 8  

Difference  in  years  

7   6   5   4   3   2  

0  

1868   1872   1876   1880   1884   1888   1892   1896   1900   1904   1908   1912   1916   1920   1924   1928   1932   1936   1940   1944   1948   1952   1956   1960   1964   1968   1972   1976   1980   1984   1988   1992   1996   2000   2004   2008   2012  

1  

Calendar  year  

Source: Author’s own diagrams based on data from Statistics Austria.

A comparison with other western1 European countries reveals that there is nothing special about the arching evolution of the Austrian gender gap in life expectancy after the Second World War. Indeed, such a pattern can be observed quite generally. Figure 2 shows the evolution of the gender gap by quinquennial periods for six western European countries from 1950–55 to 2005–10. The first panel refers to more northerly countries and the second panel to Mediterranean countries. All countries shown have high general levels of life expectancy. Figure 2: Gender gap in life expectancy, western European countries 1950/1955–2005/2010 Sweden

United Kingdom

Netherlands

7,00

Difference in years

6,00 5,00 4,00 3,00 2,00

1,00 0,00

1

“Western” here means the traditional capitalist countries. In former communist countries in Central and Eastern Europe, the gender gap in life expectancy evolved in specific ways after 1945.

The reduction of the gender gap in life expectancy in Austria since the 1980s France

Italy

67

Spain

9,00

Difference in years

8,00 7,00

6,00 5,00 4,00 3,00 2,00 1,00 0,00

Source: Author’s own diagrams based on data from United Nations (2013).

Despite relatively different starting levels in 1950/1955, the gender gap first increased, then plateaued or peaked, and finally decreased in all six countries. Yet, what differs is the period in which the maximum gender gap occurred. The United Kingdom was the first European country to attain its maximum gender gap, as early as 1970–75. In contrast, France and Spain attained their maximum only in 1990–95. In general, it seems that the gap has started to narrow earlier in northern than in southern countries. Although a certain part of the gender gap in life expectancy is attributable to biological disparities between men and women,2 it is clear that huge variations in the gender gap between western European countries and, in particular, within countries over time must be essentially caused by non-biological social factors. In that respect one may interpret the arching of the gender gap as a summary of certain social processes related to the changing roles of men and women in a population. One may think, for example, of gender disparities in smoking behavior or labor force participation. The timing of the maximum gender gap may then be understood to index the state of development of different countries regarding these social processes.3 The past years have seen an increasing interest in intra-country mortality differentials by socio-economic criteria. The setup of the Eurostat Task Force “Life expectancy by socio-economic status” in 2009 needs to be mentioned here, charged with regular production and dissemination by Eurostat of mortality indicators by educational attainment. Although only a few countries are yet able to provide regular (annual) mortality statistics by socio-economic criteria, many are now in a position to calculate at least periodic values for census follow-up periods. A typical finding is that there are large disparities in life 2 3

Luy (2003). For a comparable application see Kuntsche/Gmel (2005).

68

Johannes Klotz

expectancy by socio-economic status (however measured), that these disparities are larger among men than women and that they have in general not declined over time. A breakdown of mortality statistics by socio-economic criteria, in particular by educational level, is also interesting regarding the gender gap in life expectancy. People with higher education are often seen as a “vanguard group”4 of the population, so their evolution of demographic outcomes in the past may serve as a reference for potential developments in the whole of society in the future. For Austria, education-specific life tables are available for one-year periods following the censuses of 1981, 1991, 2001, 2006 (test census), and 2011. All death records during the 12-month follow-up periods were matched with the individual census records, allowing for each census record to assess whether this person survived or died during the follow-up period. Information on the person’s educational level is obtained from the census. Details on the data sources and linkage are described elsewhere.5 In detail the method differed between the censuses; it may be assumed that the accuracy of the mortality statistics has improved over time. Since the highest educational level is not completed at birth but later in life, education-specific life tables have to start at an appropriate minimum age. We chose to start at 35 years, which is the age when around 90 percent of all Austrians with university education have attained their degree. The minimum age implies that the part of the overall gender gap which is caused by mortality differentials below the age of 35 is not covered by our analysis. Also, for data quality reasons uniform death probabilities were assumed for all educational groups at ages 95 and older. We distinguish five levels of educational attainment, which is the highest level of education successfully completed. Basic education means compulsory education and includes people who did not successfully complete compulsory education. Apprenticeship is the second-lowest category. Lower secondary education means intermediate technical and vocational school, whereas higher secondary education covers academic secondary school, higher technical and vocational college, including post-secondary courses. Finally, tertiary education means university and Fachhochschule, including post-secondary colleges. Education-specific gender gaps in remaining life expectancy at age 35 are given in Table 1 and Figure 3. We see that the gender gap is larger among the lower educational classes. Moreover, this educational gradient has increased over time. Whereas among Austrians with basic education the gender gap declined merely from 6.6 years in 1981 to 6.2 years in 2011, it dropped from 4.7 to 2.0 years among Austrians with tertiary education. The three intermediate educational groups showed intermediate reductions in the gender gap. As a result, the variation in the gender gap by educational level has increased over time. 4 5

Sholnikov et al. (2009). Klotz/Asamer (2014) and the references therein (in German).

The reduction of the gender gap in life expectancy in Austria since the 1980s

69

Table 1: Education-specific gender gap in life expectancy at age 35, Austria 1981–2011

Census

Gender gap in life expectancy at age 35 in years Total

Educational attainment Basic

Apprenticeship

Lower Secondary

Higher Secondary

Tertiary

1981

5.9

6.6

6.8

5.2

4.7

4.7

1991

5.7

6.8

6.5

5.3

5.1

3.9

2001

5.2

6.5

5.8

4.7

4.3

3.0

2006

5.0

6.2

5.5

4.2

4.0

2.6

2011

4.7

6.2

5.5

3.7

3.9

2.0

Source: Statistics Austria. In 2006 a test census was conducted. Figure 3: Education-specific gender gap in life expectancy at age 35, Austria 1981–2011 8,0  

Gender  gap  in  years  

7,0   6,0  

Basic  

5,0  

Appren7ceship  

4,0  

Lower  Secondary  

3,0  

Higher  Secondary  

2,0  

Ter7ary  

1,0   0,0   1981  

1986  

1991  

1996  

2001  

2006  

2011  

Source: Author’s own diagram based on data from Statistics Austria.

Comparing the initial values in 1981 with the final values in 2011 provides a first summary of the development over time. Yet the three observations in between should not be neglected. To bring out more information from the data, linear regressions of the gender gap on time were fitted for each educational group. The observational periods were coded as 0, 1, 2, 2.5 and 3, which means that a regression coefficient indicates the average change in the gender gap per decade. Calculations were made in IBM SPSS Statistics, Version 20, using the GENLIN command. Robust (“sandwich”) standard errors were calculated.6 Results are also given in months of gender gap (Table 2). 6

Ziegler (2011), pp. 51–77. In this context randomness in the data means random deviations from linearity. The observed gender gaps are therefore implicitly assumed to be

70

Johannes Klotz

Table 2: Average reduction of the gender gap per decade, Austria 1981–2011

Average reduction of the gender gap per decade Total

Educational attainment Basic

in years in months 95 % CI lower limit 95 % CI upper limit

Apprenticeship

Lower Higher Tertiary Secondary Secondary

0.418

0.181

0.482

0.524

0.359

0.874

5.0

2.2

5.8

6.3

4.3

10.5

4.0

0.9

4.7

3.8

1.9

9.8

6.0

3.4

6.9

8.8

6.7

11.1

Source: Statistics Austria. CI indicates confidence interval.

The educational gradient of the decline in the gender gap is apparent also from the regression coefficients. If one assumes a linear decline in 1981–2011, then the average reduction per decade was 5 months overall, but ranged between 2 months for basic education and 10 months for tertiary education. The 95 % confidence interval limits indicate that this disparity is too large to be the outcome of pure chance. Again, the intermediate educational classes show intermediate values. A slight deviation from the general finding is the coefficient for those with higher secondary education, which is smaller than for apprenticeship and lower secondary education. Yet standard errors reveal that this could also be a chance outcome. Given the educational gradient of the gender gap and its decline in Austria, it is natural to ask for the rationale. The author does not believe that complex social phenomena as discussed in this article can be fully covered by one particular explanation. Instead, several changes in society over the last decades may play a role. Firstly, it may be that Austrians with tertiary education are indeed a vanguard group in terms of the reduction of the gender gap. This means that within this educational group the maximum gender gap was attained earlier than in the Austrian general population – perhaps in the 1970s as in the United Kingdom general population given in Figure 2. In contrast, the Austrians with basic education are laggards regarding the reduction of the gender gap; their maximum was not attained until the 1990s, which resembles the pattern of the Mediterranean countries in the second panel of Figure 2. Clearly, the past decades have witnessed some convergence of male and female lifestyles and social behavior, for instance regarding labor market participation, occupational distribution, smoking behavior or motor vehicle driving. There is reatrue values, without measurement error or inherent variation as discussed by Brillinger (1986). If one interpreted the measured gender gaps themselves as only estimates of true values, the regression standard errors would have to be adapted accordingly.

The reduction of the gender gap in life expectancy in Austria since the 1980s

71

son to believe that altogether this convergence has started earlier among those with higher education. The case of smoking in Switzerland is analyzed in detail by Kuntsche and Gmel (2005). A second explanation is selection effects. The educational expansion of the Austrian population since the 1970s7 affected both men and women, but women to a greater extent. Females with tertiary education in 1981 were still a very small group, much smaller than males with tertiary education. Accordingly, in 1981 many more women than men had only basic education. Since then the distribution of educational attainment has become relatively even between the sexes. If one interprets educational attainment primarily as an indicator of the relative socio-economic position in a population, then in 1981 tertiary education had a quite different meaning among males than among females. In contrast, today’s implied socio-economic positions are better comparable, so education-specific gender gaps in mortality are now influenced to a smaller extent by gender gaps in educational attainment. Figure 4: Age-specific male excess mortality for educational groups, Austria 1981 and 2011 Basic 1981

Basic 2011

Tertiary 1981

Tertiary 2011

Ten-year probability of dying: males/females

3,00 2,50 2,00 1,50 1,00 0,50 0,00 35-44

45-54

55-64

65-74

75-84

85-94

Source: Author’s own diagram based on data from Statistics Austria.

An age-specific mortality analysis may give some insight. Figure 4 provides age-specific excess mortality (ratio of 10qx of males divided by females) for the two marginal educational groups in 1981 and in 2011. We see that among those educated at tertiary level, male excess mortality declined rather similarly in all age groups (the erratic values for the youngest ages are likely to be caused by low death counts). Yet among those with basic education, male excess mortality declined only for the two youngest age groups, 35–44 and 45– 54, whereas it increased for those aged 55 and older. Altogether the data somehow support the “vanguard group hypothesis”: Social changes may start in different subpopulations at different times, which may cause unusually large 7

Statistics Austria (2014), pp. 34–35.

72

Johannes Klotz

variation between such subpopulations in a cross-sectional perspective. Thus the large variation of the gender gap by educational level in 2011. Given this postulate, one may expect, in future, a further narrowing of the gender gap among those with basic or intermediate education, whereas the contraction should level off among those with tertiary education, where the current gender gap of 2.0 years is already close to what can be expected for non-social biological reasons.8 Bibliography Brillinger, David: A Biometrics invited paper with discussion: The natural variability of vital rates and associated statistics. In. Biometrics 42(4) (1986), pp. 693–734 (with discussion). Klotz, Johannes; Asamer, Eva-Maria: Bildungsspezifische Sterbetafeln 2006/2007 sowie 2011/2012. In: Statistische Nachrichten 69(3) (2014), pp. 209–214. Kuntsche, Sandra; Gmel, Gerhard: The smoking epidemic in Switzerland: An empirical examination of the theory of diffusion of innovations. In: Sozial- und Präventivmedizin 50(6) (2005), pp. 344–354. Luy, Marc: Causes of male excess mortality: insights from cloistered populations. In: Population and Development Review 29(4) (2003), pp. 647–676. Shkolnikov, Vladimir M.; Andreev, Evgueni M.; Jdanov, Dmitri A.; Jasilionis, Domantas; Valkonen, Tapani: To what extent do rising mortality inequalities by education and marital status attenuate the general mortality decline? The case of Finland in 1971–2030. Max Planck Institute for Demographic Research Working Paper 2009–018, Rostock 2009. Statistics Austria: Census 2011 Austria. Results of the register-based census. Vienna, 2014. United Nations, Department of Economic and Social Affairs, Population Division. World Population Prospects: The 2012 Revision, DVD Edition, 2013. Ziegler, Andreas: Generalized estimating equations. New York-Heidelberg-Dordrecht-London 2011.

8

Luy (2003).

Gender specific mortality trends over the epidemiological transition: a view from the British mainland 1850–2000 Alice Reid, Eilidh Garrett, Chris Dibben, Lee Williamson* Great Britain is undoubtedly a part of Europe: it is linguistically, culturally, politically and of course demographically similar, particularly to its nearest neighbours in the North-Sea basin. However, the maritime boundary between Britain and the European mainland is sometimes thought to have allowed the former to remain more distinct, and one aim of this paper is to explore the extent to which the British experience of the evolution of gender specific mortality differentials is different to that in the rest of North-West Europe. A second aim is to examine some of the differences within the constituent countries of the British Isles, specifically between England and Wales on the one hand and Scotland on the other. The British Isles lie off the north-west coast of continental Europe, and consist of two major islands, Great Britain and Ireland, and numerous smaller islands. The largest island, Great Britain, contains England, Wales, Scotland and several smaller offshore islands, while Ireland comprises Northern Ireland and the Republic of Ireland. In 1850 all these islands were part of the same state, but in 1922 Ireland established independence and the sovereign state of United Kingdom was formed from Great Britain and Northern Ireland. Even before Irish Independence, however, Scotland and Ireland were administered separately in several important respects, including their census and civil registration machinery with independent Registrars General, and this has carried on until the present. This has both advantages and disadvantages in terms of demographic comparisons within the British Isles. It is clearly an advantage that separate mortality data series exist from the various dates at which civil registration was introduced: 1837 in England and Wales, 1855 in Scotland, and 1864 for Ireland, enabling easy comparison of national datasets. Comparisons with Ireland are problematic as the Irish series are unified until partition in 1921, so this paper concentrates on trends and differentials within Great Britain, specifically between England and Wales (treated together) and Scotland. The paper uses the numbers of deaths by age group and single year produced by the respective offices of the Registrars General and published in

*

This work is supported by the Wellcome Trust, through the Scottish Health Informatics Programme (SHIP) Grant (Ref WT086113), Research Programme 4: demographic, Socio-Economic and Environmental Data Linkage.

74

Alice Reid, Eilidh Garrett, Chris Dibben, Lee Williamson

their Annual Reports or Statistical Reviews.1 Population data for denominators are based on decennial censuses conducted by each country.2 The most pertinent disadvantage of the independent registration machinery in the different countries is that although their data collection was similar, the nosologies they used to classify deaths were, at least in the early years of registration, significantly different. In England William Farr developed an aetiologically based nosology: a result of his belief in the miasmatic theory of disease causation. However the Scottish public health officials, represented by the Scottish Registrar General, James Stark, were more attached to contagion as a cause, and moreover felt that it was unrealistic to expect entirely accurate recording of causes of death; they therefore developed a nosology which classified diseases according to their ‘seat’ or position in the body. Thus in Scotland pneumonia would be classified under ‘diseases of the chest’, whether it was diagnosed by a doctor as pneumonia or termed ‘affliction of the lungs’ by a relative. This system would also avoid the problem that doctor-diagnosed deaths might be misleadingly placed under their immediate cause (eg ‘peritonitis’) rather than their underlying cause (eg ‘childbirth’).3 This resulted in nosologies which could not be compared across the border between England and Wales and Scotland, particularly for the early years of registration before the adoption of the International Classification of Diseases (ICD) system in the very early twentieth century. Furthermore, both countries underwent significant revision of their nosologies at various points in time, with causes being moved between causal groups in the light of new medical knowledge, severely hampering the ability to compare causes of death over time. In order to produce comparable series of causes of death for both countries, the individual causes for each country have been regrouped so that continuity over time in the constitution of each group is retained. However it is worth remembering that the underlying rationales behind the different cause of death approaches might have led to differences, for example in the way that multiple causes of death were assigned to a single category. It is also only possible to work with the categories offered by the Registrar Generals’ publications: in some cases, particularly near the start of registration when far fewer individual causes were reported, these were themselves vague and contained amalgamations of causes (for example typhus and typhoid were not distinguished until 1865 in Scotland and 1869 in England and Wales). We have at-

1

2 3

For England and Wales 1855 to 1900, and for Scotland 1855 until 1949 this is available through projects awarded to Dr Romola Davenport and available through the UK Data Archive (SN5705) and . The data for England and Wales 1900–1949 have been deposited in the UK Data Archive by the Office for National Statistics (SN2902) and the data for both countries since 1950 are have been obtained from through the Human Mortality Database (www.mortality.org). Obtained from the Human Mortality Database (www.mortality.org). Crowther (2006), p.6.

Gender specific mortality trends over the epidemiological transition

75

tempted to strike a balance between keeping individual causes in the same causal groups and retaining meaningful groups over time.4 Comparisons over time are particularly hampered by the changing proportion of ill-defined causes. In the nineteenth century many causes were vague or poorly defined, due to a combination of the lack of routine autopsy, rudimentary knowledge of the symptoms of many conditions, the certifying doctor not having visited the patient during their final illness, and the death not having been medically certified at all. It must be remembered that these ill-defined deaths should most probably belong in a more precise category, and that declines in such imprecise causes will be balanced by increases in better defined categories. This will be discussed in more detail below. The top series of lines in Figure 1 shows the evolution of increasing expectation of life at birth for England and Wales (red) and Scotland (blue) in the North-West European context. Only three additional countries, France, Sweden and the Netherlands, are shown to avoid congestion on the graph. The general trend shows remarkable similarity in the timing and pace of improving mortality. Life expectancy began to increase in all these countries in about 1870, with Sweden leading the way and maintaining a high position throughout. The pace of change picked up in the Netherlands after the turn of the century, England and Wales narrowed the gap after the First World War and France from about 1960. The pace of change in all countries slowed simultaneously in the 1950s, by which time Scotland had established a firm position at the bottom of this small group. In contrast to the uniformity of the trend in life expectancy, the lower series of lines on Figure 1 (depicting the ratio between male and female life expectancy) shows quite differently gendered paths of improvement in the different countries. A value of one on the right hand axis indicates parity of life expectancy for men and women, the further the value is above one, the higher is female life expectancy compared to male. There was relatively wide dispersion of the gender ratio at the beginning of the period while the convergence at the start of the secular improvement in life expectancy might be coincidence, the divergence thereafter suggests that males and females benefitted to different degrees from mortality declines in the different countries. The early mortality decline in Sweden and particularly the Netherlands appears to have been greater for men than for women, but the opposite was true for Scotland and particularly for France. The gender disparity in life expectancy was largest at the outbreak of the Second World War, and thereafter all these countries exhibit a common pattern of greater improvement for women until the 1970s or 1980s, followed by greater improvement for men, although the timing of the change and the levels of the ratio varied. The countries maintained the rankings they had achieved in 1939, with Sweden and the Netherlands having the smallest gender gap and France the greatest. France has maintained a significant difference to the rest of these countries, but an earlier turning point in 4

See Reid et al.: A confession (2015), and Reid et al.: A century (forthcoming) for more detail on the regrouping.

76

Alice Reid, Eilidh Garrett, Chris Dibben, Lee Williamson

the gender ratio in the British Isles has allowed convergence with Sweden and the Netherlands. Figure 1: The evolution of life expectancy and gender differentials: England and Wales and Scotland in the North-West European context Source: Human Mortality Database. University of California, Berkeley (USA), and Max Planck 1,4 80

1,35 70

Life expectancy at birth (left hand axis)

1,3

Life expectancy at birth

1,25 50

France 1,2

Sweden

40

Netherlands England and Wales 1,15

Scotland

30

Ratio between male and female life expectancy

60

Ratio between male and female life expectancy (right hand axis) 1,1

20

1,05

10

0

1 1850

1860

1870

1880

1890

1900

1910

1920

1930

1940

1950

1960

1970

1980

1990

2000

2010

Institute for Demographic Research (Germany). Available at www.mortality.org (data downloaded on 25/11/2014). Note: Gender ratios have been smoothed using a 5 year moving average, and extreme points affected by the following military conflicts are all excluded to ease visualization of the general pattern: Franco-Prussian War (France), First World War (France, England and Wales), Second World War (all except Sweden).

The rest of the paper will concentrate on the age and cause-specific patterns within the British Isles, but the reasons for some of these inter-country differences will be commented on as these are explored. Although the differences between Scotland and its more southerly neighbour are smaller than those between these two countries and the other European countries shown in Figure 1, there are significant differences. England and Wales established consistently better life expectancy from the early 1900s, and the advantage was greater for women south of the border until the 1960s, since when the reversal of the trend in the gender ratio has been stronger for England and Wales.

Gender specific mortality trends over the epidemiological transition

77

Figure 2 shows mortality rates for selected age groups by gender, together with the sex ratio in the mortality rates, for England and Wales and Scotland from 1841–2011. These are again calculated so that a higher ratio indicates greater male disadvantage. Gender parity (ratio=1) is marked on the graphs using a horizontal line. Please note that the left hand axes measuring mortality have different scales in each panel of the figure, but that the right hand y-axes measuring the gender ratio are always the same. In these and the figures which follow, Scottish series are always blue and those for England and Wales are always red; gender differentiated mortality series are thick lines with a solid line for females and a dashed line for males, with the scale on the left hand axis; gender ratios are thin lines with the scale on the right hand axis. Examination of age-specific mortality rates by gender allows us to identify the age groups with the most extreme gender patterns. Note, however, that a large gender disparity in mortality rates which are very low might not contribute much to the overall gender differences and that gender ratios tend to be very unstable when mortality rates are very low, due to small numbers. This is particularly true for Scotland where there are fewer absolute numbers of deaths. The male disadvantage in infant mortality is a long-standing feature of the demographic regime,5 and is relatively constant, with male mortality about 30 per cent higher than female, despite major declines in the level of infant mortality. We know that males are more prone to prematurity, respiratory problems and to foetal distress and the slight decline in the sex ratio of mortality in the last two or three decades is probably due to advances in obstetric and neonatal care which are likely to have benefitted males more than females.6 At the start of the period infant mortality was significantly higher in England and Wales than in Scotland, but the situation was reversed in the early twentieth century. In early childhood (one to four years), a small male mortality disadvantage emerged as mortality fell, probably due to the elimination of gender neutral causes of childhood death, such as infections, leaving a larger role for more masculine causes such as accidents. The emerging male disadvantage in post-war Britain is much more prominent among late adolescents and adults. The differences between the two countries are interesting here: until the First World War, Scottish mortality in this age group was significantly higher than English for both genders, but these converged in the interwar period and a different split, this time between the genders, emerged in the second half of the twentieth century. Adolescence is a period of very low mortality rates, and higher male mortality is usually ascribed to male risk-taking behaviour and occupational risks. It is interesting, however, that this gender difference did not emerge until after the second world war when it made a very sudden appearance, despite continued advances in safety and in medical care for victims of accidents. Were males not risk-taking before then? Did females become more risk averse? Or was there a sudden explosion of more dangerous and more gendered opportunities, such 5 6

Woods and Shelton (1997), p. 134. Naeye et al (1971) p. 902; Waldron (1985); Drevenstedt (2008).

78

Alice Reid, Eilidh Garrett, Chris Dibben, Lee Williamson

as motorcycling? This phenomenon has puzzled researchers, who have retreated to explanations such as the counterbalancing decline female in dominated causes such as tuberculosis and maternal mortality.7 We shall see later that the fall in maternal mortality, generally attributed to the introduction of sulphonamides in the late 1930s and penicillin in the early 1940s, was sudden and dramatic, and these antibiotics were also effective against infections such as tuberculosis which might have been more common or more likely to have been treated among women.8 This male disadvantage has attenuated in recent years. Figure 2: Age- and gender- specific mortality rates (selected ages) and gender ratios in mortality rates, England and Wales and Scotland.

2 1.5

50

1

1941

1961

1981

2001

Mortality age 15-19

9 8

6

0 1841

1861

1881

1901

1921

1941

1961

1981

30

5

0.5

0

0.5 1841

1861

1881

1901

1921

1941

1961

1981

2001

1 0.5 1841

1

0

4

1.5

1

1.5

5

Mortality age 25-29

2

2

10

2001

10

25

15

1981

2.5

2.5

20

1961

15

3

35

1941

3

3.5

40

1921

20

1861

1881

1901

1921

1941

1961

1981

2001

Mortality age 80-84

250

4

45

1901

3.5

2001

Mortality age 60-64

50

1881

25

1.5

1

0.5 1861

30

2

2

1

3.5

5

3

1.5

35

2.5

4

2 20

4

3

7

2.5

30

1841

deaths per 1,000 population

1921

ratio male to female death rate

1901

3

0

deaths per 1,000 population

1881

3.5

10

0.5 1861

40

4 3.5

200

3 150

2.5 2

100

1.5 50

1

0

ratio male to female death rate

100

4

E&W Female E&W Male Scotland Female Scotland Male E&W ratio Scotland ratio

ratio male to female death rate

2.5

Early child mortality

ratio male to female death rate

50 deaths per 1,000 population

150

10

deaths per 1,000 population

3.5 3

0 1841

deaths per 1,000 population

60

ratio male to female death rate

infant deaths per 1,000 births

200

4

ratio male to female death rate

Infant mortality

250

0.5 1841

1861

1881

1901

1921

1941

1961

1981

2001

Source: Human Mortality Database. University of California, Berkeley (USA), and Max Planck Institute for Demographic Research (Germany). Available at www.mortality.org (data downloaded on 15/02/2014). 7 8

Martin (1951); Retherford (1975), p.11. However, Wisser and Vaupel (2014) point out that small and changing denominators can produce large changes in ratios which can be misleading. It is important to remember that mortality rates in these ages are very low. For the effect of antibiotics, see Løkke (2012); Loudon (1992). On the greater propensity of women to access health services, see Bertakis et al (2000).

Gender specific mortality trends over the epidemiological transition

79

Another notable feature of both the young adult age groups shown here is the devastating effect of the 1918 influenza pandemic which was unusual in its particular lethality among young adults.9 It appears to have been slightly more dangerous for men, and this might have been due to troop movements facilitating transmission. The Second World War shows up as slightly heightened male mortality and greatly heightened male disadvantage: the data refer to civilians only, so this is likely to be a selection effect whereby those in poor health did not fight. In adulthood mortality rises with age, so although the gender ratios at older ages are lower than in early adulthood, the contributions of the older age groups to the overall gender disparities are much larger. Mortality among 60–64 year old women decreased fairly constantly from the later nineteenth century, but although men shared in the early part of the decline, their mortality decline stalled after the First World War in England and Wales, and slightly later in Scotland where it even increased. The resulting rise in the male disadvantage was therefore particularly marked in Scotland. This pattern has been established to be the result of cohort smoking trends.10 Men took up smoking earlier and in a more widespread fashion than women, and the cohort born at the turn of the century and who took up smoking during the First World War suffered particularly when they reached middle and old age.11 They are again prominent among those aged 80–84 where the peak male disadvantage occurs 20 years later, confirming the cohort effect. Returning to the 60–64 age group, it is possible to see that the turning point in the gender ratio is due both to a decline in male mortality, perhaps due to a reduction in smoking rates, and to a stalling in female mortality as their own smoking epidemic began. The timing of the smoking epidemic is likely to explain some of the differences between Britain and the other countries in Figure 1: smoking was adopted earlier in the UK both among men and women, hence the earlier turning points in the late twentieth century.12 The rest of this paper examines trends in cause-specific mortality for men and women using standardized mortality rates with the 1901 Scottish population taken as the standard. Before examining gender ratios it is instructive to look at the general pattern and consider the implications of changes in the recording and grouping of causes of death. Figure 3 shows the time series of age-standardized cause-specific mortality rates for Scottish women; Scottish men and both genders in England and Wales show the same general pattern.

9 10 11 12

Johnson (2006) p.83. Retherford (1975); Waldron (1985); Pampel (2002). Forey et al (2012). Pampel (2002), p. 91. The cohort pattern is clearly visible in a lexis map, see for example Gjonça et al (2005).

80

Alice Reid, Eilidh Garrett, Chris Dibben, Lee Williamson

Figure 3: Standardized mortality rates for Scottish women, by cause 25 Violence Ill defined

deaths per thousand people

20

Perinatal Other Childbirth

15

Genitourinary Digestive Diarrhoea

10

Infectious Tuberculosis Respiratory

5

Neoplasms Circulatory Nervous system

0

Old age

Sources: see footnote 1.

Our re-classification of causes of death has ended up with fifteen broad categories which are largely consistent over time. Instead of taking the approach of only uniting causes we absolutely know to be identical, which places a large proportion of causes in the early period in ‘nosologically not meaningful’ categories,13 such as ‘ill-defined’, we have tried to minimize the ‘ill-defined’ category by uniting causes which today would be considered too vague to be allocated to a precise cause, with other similar causes.14 For example we placed ‘pleurisy’ with ‘diseases of the respiratory system’ and ‘syncope’ with ‘circulatory diseases’. Nevertheless, even after doing this and separating out ‘old age’ into a category of its own, we are still left with a sizeable ‘ill-defined’ category. The colours and arrangement of causal groups in the figure are designed to group similar causes together, where we suspect that there might be transfer between groups over time due to increasing precision in diagnosis. Diarrhoea and diseases of the digestive system are both coloured green as there is likely to be confusion between infective diarrhoea and non-infective gastro-enteritis. Infectious diseases and respiratory diseases are coloured blue along with tuberculosis as it is likely that some deaths from bronchitis and pneumonia were the sequelae of common infectious diseases such as measles. The purple coloured causes at the bottom of the graph are those most commonly found among the elderly, and displaying them like this very strongly suggests that the decline in deaths from ‘old age’ is largely counterbalanced by a rise in neoplasms and circulatory diseases among the elderly.15 Old people’s deaths 13 14 15

Wolleswinkel Van-den Bosch et al. (1996). See Reid et al.: A confession (2015), and Reid et al.: A century (forthcoming) for more detail on the regrouping. This is even clearer if deaths among the elderly are examined, see Reid et al.: A century (forthcoming).

Gender specific mortality trends over the epidemiological transition

81

previously commonly attributed to ‘old age’ were increasingly given a more precise cause, and we should therefore be cautious about interpreting trends in causes of death such as neoplasms and circulatory diseases.16 Figure 4: Gender ratios in standardized cause-specific mortality Scotland

4,0

Genitourinary Nervous system

ratio: male to female mortality rates

3,5

Digestive 3,0

Childbirth

Perinatal 2,5

Old age

Tuberculosis

2,0

Circulatory

Neoplasms

1,5

Infectious Diarrhoea

1,0

Respiratory Violence

0,5

Other 0,0

1850s1860s1870s1880s1890s1900s1910s1920s1930s1940s1950s1960s1970s1980s1990s2000s

Ill defined

England and Wales

4,0

Genitourinary

ratio: male to female mortality rates

3,5

Nervous system Digestive

3,0

Childbirth Perinatal

2,5

Old age Tuberculosis

2,0

Circulatory Neoplasms

1,5

Infectious Diarrhoea

1,0

Respiratory Violence

0,5

Other Ill defined

0,0 1850s1860s1870s1880s1890s1900s1910s1920s1930s1940s1950s1960s1970s1980s1990s2000s

Source: see footnote 1.

Figure 4 shows the gender ratios in age-standardized cause-specific mortality rates for Scotland and for England and Wales over time. Most causal groups show masculinization in the early twentieth century but deaths from ‘violence’ (including suicide and accidents) are a major exception. Contrary to our speculation earlier that young men in the nineteenth century may not have indulged in as much risky behaviour, this figure shows that risk taking was more widespread or, more probably, more likely to have been fatal in the nineteenth century. Mortality from ‘violence’ fell for both sexes, but particularly among men throughout the period. The increase in the gender ratio in the last 16

See Reid et al.: A confession (2015), and Reid et al.: A century (forthcoming) for further elaboration of this argument.

82

Alice Reid, Eilidh Garrett, Chris Dibben, Lee Williamson

few decades reflects a slight increase in the pace of decline among women rather than any change in the pace of decline among men. The other major cause which is unusual in becoming more feminine throughout the twentieth century is ‘old age’, however here we must bear in mind our hypothesis that deaths formerly attributed to ‘old age’ were increasingly more accurately placed in ‘circulatory diseases’ or ‘neoplasms’. This trend is likely to have started first among the younger old, where mortality rates are higher for men, with the result that ‘old age’ as a cause of death was increasingly concentrated in the oldest ages and among women. The top left panel of Figure 5 confirms that this form of mortality was declining rapidly among both sexes, but that the persistent female disadvantage intensified during the twentieth century. The top right hand panel of Figure 5 shows the trends and gender ratios in mortality from neoplasms. In common with other causes prominent among the elderly, cancers were more common among women until the second half of the twentieth century: the risks increased for both sexes, probably largely due to transfers from ‘old age’, but rates for women levelled off and started to fall some forty or fifty years before those for men, producing a sharply increasing gender ratio of mortality. Lung cancer related to smoking is a primary cause of the continued rise in male mortality from neoplasms and the peak in masculinity in the late twentieth century, but it should be remembered that the nineteenth and early twentieth century rise and early masculinization is likely to be an artefact of improving certification of death. It is very unlikely that all ‘old age’ deaths in earlier times were actually due to cancer – some might have been caused by circulatory diseases (heart disease, stroke and other cardiovascular conditions) and the lower left hand panel of Figure 5 shows mortality rates and the gender ratio from this group of causes. The sharp increases in circulatory disease mortality in the nineteenth century could plausibly be due to transference from ‘old age’, as could the smaller increases across the first half of the twentieth century. This could also explain higher rates for women until the post-war period. Since then, the delay in the male decline from this broad group of causes and consequent masculinization is also affected by the influence of smoking on cardiovascular mortality.17 If there is a transfer between ‘old age’ (and possibly also diseases of the nervous system) to neoplasms and circulatory diseases, it will be instructive to examine these causes as a combined group, and this is shown in the lower right hand panel of Figure 5. This shows a much less dramatic, and in many ways less puzzling picture than those for neoplasms and cancers separately. If general health was improving as a result of better nutrition and fewer infections, it seems unlikely that these causes would be increasing, except for survival into older ages, and age-standardization removes this element from these graphs. Taking this extremely broad group of causes together has two important consequences for our conception of the changing risks of mortality from 17

Waldron (1995); Gjonca et al (2005); Pampel (2002), p. 82.

Gender specific mortality trends over the epidemiological transition

83

broad causal groups over the epidemiological transition. Firstly it shows that the risk of death from such causes combined did not significantly increase during the epidemiological transition, although of course, they were increasing dramatically relative to other causes.18 Secondly, it shifts the start of improving mortality from these causes back to an earlier date, both sexes experienced a mortality decline in the early twentieth century, one which continued for women but was arrested among men as a result of their early adoption of smoking behaviour. Figure 5: Gender ratios in standardized mortality for specific causes common in older people Old age

Scotland, gender ratio England and Wales, gender ratio

2

1,4

1,2

3

1,2

3

1,0

2

0,8

2

0,6

1

0,4

1

0,2

0

0,0

1,0 0,8

2

0,6

1

0,4

1

0,2

0

0,0

Old age, neoplasms, nervous system & circulatory diseases 1,4

7

1,4

3

1,2

6

1,2

3

1,0

5

1,0

2

0,8

4

0,8

2

0,6

3

0,6

1

0,4

2

0,4

1

0,2

1

0,2

0

0,0

0

0,0

Standardized mortality rates

4

Ratio: male to female mortality

Standardized mortality rates

Circulatory diseases

Ratio: male to female mortality

England and Wales, female mortality

3

4

Ratio: male to female mortality

England and Wales, male mortality

1,4

Standardized mortality rates

Scotland, female mortality

3 Standardized mortality rates

Neoplasms Scotland, male mortality

Ratio: male to female mortality

4

Source: see footnote 1.

Returning to Figure 4, tuberculosis is another cause which stands out as having a dramatically increased gender ratio in the 1960s and 1970s. A similar, although less extreme, pattern occurs in respiratory diseases in general (including bronchitis and pneumonia among other causes). Figure 6 shows mortality rates and causes for both tuberculosis and respiratory diseases. The masculinization of these causes occurs slightly earlier than the peak in cancers and circulatory diseases, but it is highly likely that these respiratory causes, which were by then at very low levels, were connected to the more immediate effects of cigarette smoking and greater vulnerability of the male lungs.19 18 19

However, the fact that there is still a small increase in the nineteenth century suggests that there is room for more transfer, probably from the ‘ill-defined’ category. This is confirmed by analysis by Davenport (2013), who demonstrates that both tuberculosis and lung cancer peak in the same cohorts.

84

Alice Reid, Eilidh Garrett, Chris Dibben, Lee Williamson

Figure 6: Gender ratios in standardized mortality for tuberculosis and respiratory diseases Respiratory diseases 5

2,5

2,0

4

Ratio: male to female mortality

Standardized mortality rates

4

3

1,5

3 2

1,0

2

Scotland, male mortality

1

England and Wales, male mortality

1 0

Scotland, female mortality England and Wales, female mortality

0,5

Scotland, gender ratio England and Wales, gender ratio

0,0

Tuberculosis 2,5

4

2,0

3

1,5

2 2

1,0 Scotland, male mortality

1

Scotland, female mortality England and Wales, male mortality

1 0

Ratio: male to female mortality

Standardized mortality rates

3

England and Wales, female mortality

0,5

Scotland, gender ratio England and Wales, gender ratio

0,0

Source: see footnote 1.

The image of young women wasting away with respiratory tuberculosis, or phthisis, is familiar in Victorian literature, and phthisis has been identified as the main cause of relative female disadvantage in substantial parts of England and Wales.20 However, the higher rates among young women (and of not so young women in some areas) are more than balanced by higher rates for older men, except for a thirty year period in late nineteenth century Scotland. From the 1880s males in England and Wales had consistently higher standardized risks of death from tuberculosis than females. It is plausible that perceptions of both contemporaries and more recent analysts have been swayed by higher rates of tuberculosis mortality among young women than young men, at ages when mortality was otherwise very low, while the fact that older men were suffering disproportionately was less well recognized. The fact that women had higher tuberculosis mortality at a younger age means tuberculosis will

20 Anderson (1990) 19; McNay et al. (2005); Hinde (2010).

Gender specific mortality trends over the epidemiological transition

85

have had a larger effect on the life expectancy of females than of males, and this might also contribute to an emphasis on a female disadvantage. Any discussion of gender differences in cause-specific mortality should not overlook maternal mortality. It is, of course, impossible to calculate a gender ratio for this cause as it is applicable only to women. Instead, Figure 7 shows the maternal mortality rate calculated not per 1,000 women in the population, but per 10,000 births. Calculating this per 1,000 women might in some ways provide better comparison with the other mortality rates shown here, but it is a less accurate reflection of the risks of deaths to parturient women as declining birth rates from the 1880s create the illusion that the risks associated with giving birth were declining. Figure 7 shows that the real risks associated with giving birth did not substantially decline until the widespread use of antibiotics after the Second World War.21 Figure 7: Maternal mortality per 10,000 births, and the gender ratio in non-puerperal septicaemia. 70

1,4

60

1,2

50

1,0

40

0,8

30

Scotland

0,6

20

E&W

0,4

10

gender ratio in nonpuerperal septicaemia, Scotland

0,2

0

0,0

gender ratio in nonpuerperal septicaemia, Scotland

maternal deaths per 10,000 births

Maternal mortality

Source: see footnote 1.

The nineteenth century Registrar General for Scotland, William Stark, argued that the maternal mortality rate in England and Wales was artificially depressed by the tendency of Farr’s classification system to allocate deaths in childbed to other, more immediate causes such as peritonitis.22 Detailed investigation by the English and Welsh Registrar General found there to be a substantial problem with maternal deaths, particularly those due to puerperal fever, which were ‘hidden’ among other causes, such as septicaemia. But was Scotland any better? Up to and including the 1940s, Scotland had a cause of death category for non-puerperal septicaemia, a condition which would gener21 Loudon (1992); Løkke (2012). 22 Crowther (2006).

86

Alice Reid, Eilidh Garrett, Chris Dibben, Lee Williamson

ally have been the result of infected wounds and plausibly therefore more common among men who, as we have seen, were more at risk of accidents and violence. Contrary to expectation, however, for most of the nineteenth century (except for the first five years when medical certification was low) non-puerperal septicaemia was more common amongst women than men, which would be consistent with some maternal deaths being wrongly recorded as septicaemia. The under-recording of maternal mortality is further substantiated by a temporal coincidence between the rise in the gender ratio from non-puerperal septicaemia and a rise in the recorded maternal mortality rate. As the recording of maternal deaths improved, so maternal deaths were removed from the non-puerperal septicaemia category, which increased the sex ratio in that cause. The cause of death categories for England and Wales do not allow this to be explored south of the border, so we can only speculate on the relative trends in misreporting south of the border. This cautionary tale warns us that even seemingly easily diagnosed causes such as maternal mortality should be treated with caution when considering historical mortality series, and are likely to bear on interpretation of trends in gender differentials from specific causes. The discussion so far has compared the evolution of the gender differentials in mortality in Scotland and England and Wales with selected other North-West European countries, and examined trends in mortality and gender ratios by particular age groups and for particular causes. Our analysis indicates that caution must be used when interpreting historical trends and comparisons in cause-specific mortality, and that behavioural changes such as smoking can affect mortality from many different causes. In our primarily descriptive analysis we have so far made few attempts to explain the differences between different countries considered here. The main debate in the literature when it comes to explaining the trends in gender disparities in mortality relates to the relative importance of biological and behavioural factors. We do not have space to do justice to this debate here but it seems plausible that although there is some degree of ‘universal biological force’23 behind the sex difference in mortality, differences between places and across time are governed more by different sex-specific distributions of behavioural and environmental risk-factors, such as violence, dangerous driving, tobacco and alcohol use.24 Such factors are also affected by socio-economic status, marital status, and urban or rural living. It is widely accepted that the different trends in smoking behaviour between men and women can explain the late twentieth century peak in the gender ratio in all countries, and the timing of the smoking epidemic in different countries can broadly explain the relative timing of the peaks in different places.25 What is less easy to explain is the divergence in the gender ratio in 23 Wisser and Vaupel (2014), p.1; see also Luy (2003). 24 Rogers et al. (2010) 25 Pampel (2002) further argues that although later take-up of smoking among women is contributing to a relative increase in smoking-related mortality among women, among

Gender specific mortality trends over the epidemiological transition

87

the first half of the twentieth century and the small but persistent difference between Scotland and England and Wales. It is beyond the scope of this paper to do more than suggest possibilities, but the early twentieth-century divergence between Sweden and the Netherlands on the one hand and France on the other, with England and Wales and Scotland in the middle, would fit with lower maternal mortality in Sweden and the Netherlands in the late nineteenth and early twentieth centuries, generally attributed to better midwifery.26 The difference between Scotland and England and Wales is likely to lie in behavioural and environmental factors. Since the 1960s, Scotland had a larger male disadvantage in mortality, possibly due to a delay in the reduction of smoking among men or to other gendered risk behaviours. A higher Scottish mortality rate has been in evidence for much longer, however, ever since the infant mortality rate in England and Wales dropped below that in Scotland. Possible reasons include differential migration, socio-economic composition, urbanization and behaviour, including smoking, drinking and diet.27 However, we must also remember that the difference between Scotland and England and Wales is not a clear-cut one. There is a North-South divide in health but it does not run neatly along the national border; many parts of the North of England are more similar in terms of health and survival to Scotland than they are to the South-East.28 Bibliography Anderson, Michael: The social implications of demographic change. In Thompson, F. M. L. (ed.): The Cambridge Social History of Britain 1750–1950, Volume 2: People and their Environment. Cambridge 1990, pp.1–70. Bertakis, K. D.; Azari, R.; Helms, L. J.; Callahan, E. J.; & Robbins, J. A: Gender differences in the utilization of health care services. In: Journal of Family Practice 49 (2000), pp. 147–152. Crowther, Anne: ‘By Death Divided. Scottish and English approaches to death certification in the nineteenth century’, A paper presented at the Society for the Social History of Medicine Conference (2006), http://www.gla.ac.uk/media/media_82267_en.pdf (accessed 23/11/2014). Davenport, Romola J.: Year of Birth Effects in the Historical Decline of Tuberculosis Mortality: A Reconsideration. In: PLoS ONE 8 (2013), e81797. De Brouwere, Vincent; Tonglet, René; Van Lerberghe, Wim: Strategies for reducing maternal mortality in developing countries: what can we learn from the history of the industrializing West? In: Tropical Medicine and International Health 3 (1998), pp. 771–782. Drevenstedt, GL; Crimmins, EM; Vasunilashorn, S; Finch, CE: The rise and fall of excess male infant mortality. In: Proceedings of the National Academy of Sciences of the United States of America 105 (2008), pp. 5016–5021. Forey, Barbara; Hamling, Jan; Hamling, John; Thornton, Alison; Lee, Peter: International Smoking Statistics, Web Edition, A collection of worldwide historical data, United King-

non-smoking related causes the female advantage is continuing to grow at most ages. 26 Loudon (1992); de Brouwere et al (1998). 27 The Scottish Office (1993). 28 Shaw et al. (2008).

88

Alice Reid, Eilidh Garrett, Chris Dibben, Lee Williamson

dom. (2012) http://www.pnlee.co.uk/Downloads/ISS/ISS-UnitedKingdom_120111.pdf. (accessed 26/11/2014). Gjonça, Arjan; Tomassini, Cecilia; Toson, Barbara; Smallwood, Steve: Sex differences in mortality, a comparison of the United Kingdom and other developed countries. In: Health Statistics Quarterly 26 (2005), pp. 6–16. Hinde, Andrew: Sex differentials in mortality in nineteenth-century England and Wales. Working Paper A10/04, University of Southampton. eprints.soton.ac.uk/162637/1/s3ri. workingpaper-A10-04.pdf Johnson, Niall: Britain and the 1918–19 Influenza Pandemic: A Dark Epilogue, London 2006. Løkke, Anne: The antibiotic transformation of Danish obstetrics: the hidden links between the decline in perinatal mortality and maternal mortality in the mid-twentieth century. In: Annales de démographie historique 2012/1 (2012), pp. 205–224. Loudon, Irvine: Death in childbirth: an international study of maternal care and maternal mortality, 1800–1950, Oxford/New York 1992. Luy, Marc: Causes of Male Excess Mortality: Insights from Cloistered Populations. In: Population and Development Review 29 (2003), pp. 647–676. McNay, Kirsty; Humphries, Jane; Klasen, Stephan: Excess Female Mortality in Nineteenth-Century England and Wales: A Regional Analysis. In: Social Science History 29 (2005), pp. 649–681. Martin, W. J: (1951). A comparison of the trends of male and female mortality. In: Journal of the Royal Statistical Society, Series A (General) 114 (1951), pp. 287–306. Naeye, Richard L; Burt, Leslie S.; Wright, David L.; Blanc, William A.; Tatter, Dorothy: Neonatal mortality, the male disadvantage. In: Pediatrics 48 (1971), pp. 902–6. Pampel, Fred C: Cigarette Use and the Narrowing Sex Differential in Mortality. In: Population and Development Review 28 (2002), pp. 77–104. Reid, Alice; Garrett, Eilidh; Dibben, Chris; Williamon, Lee: ‘A confession of ignorance’: deaths from old age and deciphering cause of death statistics in Scotland 1855–1949. In: History of the Family 20 (2015), pp. 320–344. Reid, Alice; Garrett, Eilidh; Dibben, Chris; Williamon, Lee: A century of deaths, Scotland 1855–1955: a view from the civil registers. In Jupp, Peter (Ed.): Death in Modern Scotland, forthcoming. Retherford, Robert D.: The changing sex differential in mortality. Studies in Population and Urban Demography No.1, Westport/London 1975. Rogers, Richard; Everett, Bethany; Saint Onge, Jarron M.; Kreuger Patrick M.: Social, behavioral, and biological factors, and sex differences in mortality. In: Demography 47 (2010), pp. 555–578. Shaw, Mary; Thomas, Bethan; Davey Smith, George; Dorling, Danny: The Grim Reaper’s road map: An atlas of mortality in Britain, Bristol 2008. The Scottish Office: Scotland’s Health – A Challenge to Us All: The Scottish Diet. Edinburgh (1993). www.healthscotland.com/documents/1181.aspx (accessed 30/11/2014). Waldron, Ingrid: What do we know about causes of sex differentials in mortality? A review of the literature. In: Population Bulletin of the UN 18 (1985), pp. 59–76. Waldron, Ingrid: Contributions of biological and behavioural factors to changing sex differences in ischaemic heart disease mortality. In Lopez, A D; Caselli, G; Volkonen, T: Adult mortality in developed countries: from description to explanation, Oxford 1995, pp. 161–78. Wisser, Oliver; Vaupel, James: The sex differential in mortality: A historical comparison of the adult-age pattern of the ratio and the difference. MPIDR WORKING PAPER WP 2014–005, JUNE 2014 http://www.demogr.mpg.de/papers/working/wp-2014-005.pdf (accessed 29/11/2014). Wolleswinkel-Van den Bosch, J; Van Poppel, F; Mackenbach, J: Reclassifying causes of death to study the epidemiological transition in the Netherlands, 1875–1992. In: European Journal of Population 12 (1996), pp. 327–361. Woods, Robert; Shelton, Nicola: An Atlas of Victorian Mortality, Liverpool 1997.

The evolution of the gender gap in life expectancy in Belgium and the smoking epidemic (1841 to 2013) Patrick Deboosere In this chapter we want to describe the global evolution of the gender differential in mortality in Belgium over the past 170 years and the main factors that contributed to it. After an overview of the evolution in life expectancy we will proceed with a decomposition of the age-specific contributions and the mortality causes involved. The discussion section starts with a review of different approaches to understand the gender differential in mortality. In a second part, as we heavily rely on cause-specific mortality to sketch the past evolutions, the registration and diagnosing of mortality at the end of the nineteenth century is briefly treated. Finally we zoom in on the role of the smoking epidemic as a main contributing factor in the recent evolution of gender-specific mortality in Belgium. A history of interaction between biological differences and social relations Basically, four periods can be distinguished in the evolution of the gender gap in life expectancy in Belgium (figure 1). The first period starts in the middle of the nineteenth century with the emergence of a male deficit in life expectancy progressively increasing to 3.5 years. This period is followed by a long plateau, roughly covering the years 1890–1930, with a gender life expectancy difference hovering around 3.5 years. From the thirties onwards the male life expectancy deficit starts to grow again and reaches a peak of 6.83 years in 1984. After a short stabilisation period of about ten years, the male deficit in life expectancy finally starts to decline from the mid-nineties onwards. In 2013 gender difference in life expectancy dropped below 5 years, returning to the 1951 level. Although the physiological differences between sexes obviously contribute to different health and mortality risks, variations in human mortality can only be understood as the outcome of an interaction process between biological potential and social environment. Sex differentials in mortality are the result of social differentiation in living and behaving of men and women and are modelled by the relationship between the sexes, a relationship that is constantly evolving over time. This explains the difficulty in differentiating between the impact of biological or genetic factors and of social or cultural factors. To understand the emergence of a male life expectancy deficit we need to analyse the evolution in gender specific mortality. The data used in the first series of figures (1–4) are based on the Human Mortality Database (HMD).

90

Patrick Deboosere

Figure 1: Deficit in life expectancy of Belgian men compared to women 1841–2012 (Human Mortality Database) 2  

1  

Difference  in  life  expectancy  at  birth  

0  

-­‐1  

-­‐2  

-­‐3  

-­‐4  

Dif  le0  

-­‐5  

-­‐7  

1841   1845   1849   1853   1857   1861   1865   1869   1873   1877   1881   1885   1889   1893   1897   1901   1905   1909   1913   1917   1921   1925   1929   1933   1937   1941   1945   1949   1953   1957   1961   1965   1969   1973   1977   1981   1985   1989   1993   1997   2001   2005   2009  

-­‐6  

Mortality data in the HMD have been standardized and are easily comparable between countries and over time. Figure 2 plots female life expectancy along the horizontal axis and the male life expectancy deficit along the vertical axis. The overall pattern shows an increase in female life expectancy going hand in hand with a growing male life expectancy deficit. As this figure does not contain a time dimension, years have been added on important inflection points. Figure 2: Belgium evolution of female life expectancy and male life expectancy gap 1841–2012 (source: HMD) 2  

Difference  in  life  expectancy  males  feamels    

0  

30  

40  

50  

60  

70  

80  

90  

1934  

-­‐2  

1952  

-­‐4  

-­‐6  

1993  

-­‐8  

1946   -­‐10  

1985  

Female  life  expectancy  

Belgium  data:HMD  

The evolution of the gender gap in life expectancy in Belgium

91

During the first half of the century progress in life expectancy is very slow and, while women are gaining about 6 years in life expectancy, men add only 2 years creating a gap of almost 4 years. Evolution in life expectancy in this period is not only slow but also very uneven, reflecting the mortality crises of the second half of the nineteenth century. From 1885 to 1890 onwards there is a strong growth in life expectancy for men and women alike, with a gain of 16 years over about half a century until 1934. The simultaneous progress among men and women results in a stabilisation of the gender gap. The evolution remains somewhat erratic, however, with the First World War and the Spanish flu as important mortality episodes. The next half-century (1934–1985) again adds 16 years in female life expectancy, but now, the gender gap strongly increases to a maximum of almost 7 years in 1984. The development is extremely uneven, with stagnation in female life expectancy during the crisis of the thirties and a male life expectancy collapse in the first war year. After a short post-war sub-period (1946–1952), with the gender gap slightly decreasing notwithstanding strong progress in female life expectancy (5.5 years increase over 6 years), the evolution starts to slow down and to become more homogeneous. In the next 33 years, between 1952 and 1985, females will add 7 years, males only 5. The year 1985 is a turning point. Male progress in life expectancy equals female progress and, from the mid-nineties onwards, males are catching up with female life expectancy, reducing the gap to under 5 years in 2013. To fully understand what is going on we have to quit the summary measure of life expectancy and take a look at age-specific mortality. A Lexis plot with the age-specific male female mortality rate ratios on the x-axis from 1841 until 2012 and age on the vertical axis sheds light on the uneven development in gender specific mortality by age. The interruption in the figure represents the missing data during the First World War and is a good point of reference. Yellow dots point to the ages and years where male and female mortality rates are almost equal. Red dots represent female excess mortality and green dots male excess mortality. The light colors stand for an excess mortality of less than 50 per cent, the darker red and green for excess mortality between 50 % and 100 % and the darkest green for a male mortality more than double the female mortality rate. The result tells the history of the evolution of mortality in Belgium across age-specific gender differences. Before the First World War, clear excess mortality was more common among women than among men. Women did have high excess mortality between ages 25 and 50 during the first 25 years of observation in the mid-nineteenth century. Maternal mortality was the most important risk factor. Young girls, from 6 to 19 years old, were also hit by excess mortality, a pattern persisting up until 1930 while retracting to the 12–18 years age group. The overall lower life expectancy in men was due to the older age group, 45 and beyond, with excess male mortality that gradually spread for all ages between 40 and 85. But a major contribution to male excess mortality is almost invisible in this Lexis plot. It is the first age group. With excess mortal-

92

Patrick Deboosere

ity of 20 to 30 % compared to girls, this group contributed heavily to the total male excess mortality. Mortality rates were indeed high among young children and staggeringly high amongst the newborn. The higher male birth rate was almost immediately neutralized in the first weeks after birth. That one in six children died before the first anniversary coupled with a 20–30 % excess mortality among boys explained half of the life expectancy gap between men and women during most of the nineteenth century. In that way, excess mortality in men and women alike was strongly biologically defined. The extra parts in the double X-chromosome structure in women are a better protection against degenerative diseases and birth deficiencies. But through their specific contribution in human reproduction, women have much higher mortality risks during their fertile years. The nineteenth century started to change this biological burden for women by improving the conditions of childbirth. At the start of the observation period, maternal mortality represented around 1.25 % of live births according to Vandenbroeke, but the steady improvement of hygiene and the introduction of aseptic techniques around 1880 progressively contributed to a decrease in maternal mortality during the second half of the nineteenth century.1 However, the industrialization of this very same nineteenth century was probably the main factor contributing to excess mortality of adult males. In Belgium, life expectancy for men became even worse during the period of intense industrialization before 1885. As for the excess mortality in young girls, this has been pointed out by several Belgian demographers2 and has been documented for other countries as well3. Although excess mortality before 1914 was much more pronounced in women of some specific age ranges, it occurred at ages when mortality rates are relatively low. Quite the opposite was true for men. The rather moderate higher male mortality under 1 year and among older males occurred at ages where mortality rates were high. The net result was gender equilibrium in life expectancy in the mid-nineteenth century that evolved towards a male life expectancy deficit of almost 4 years in 1885. The very late decrease of child mortality in Belgium at the end of the nineteenth century will contribute to the long period of equilibrium in the gender gap in life expectancy. While conditions are worsening for men at older ages and still improving for women during their reproductive years, the gradual decrease in child mortality will strongly contribute to a decrease in male mortality with a high impact on life expectancy.

1 2 3

Vandenbroeke 1978; Masuy-Stroobant and Humblet 2004. Eggerickx and Tabutin 1994; Poulain and Tabutin 1977. Wrigley et al. 2005.

The evolution of the gender gap in life expectancy in Belgium

93

Figure 3: Lexis plot time by age of mortality rate ratios of men and women in Belgium from 1841–2012 (HMD) 100

0 1841

WWI

2012

With the start of the Second World War a sharp and sudden relative increase of male mortality appears compared to women. This is particularly true at young ages (20 to 33) where the jump goes from virtual no difference (yellow) or less than 50 % before the war to 3 or 4 times higher mortality rates. Older men between 35 and 60 were also affected by the war, but they started from a higher relative mortality level. During the fifties, the relative risk of mortality of men clearly starts to take off in two distinct age groups: older men above age 55 and young men in their twenties. The excess mortality above age 55 follows a pattern that suggests a strong cohort effect over and above the period effect. Analyzing cause-specific mortality helps to explain this evolution and the smoking epidemic is probably the single most important factor as we will illustrate later. The youth peak starts at age 16 sharp, coinciding with the age young people were allowed to drive light motorbikes, and ends in the mid-thirties. The excess male mortality of the youth peak is mainly driven by the new evolution in individual mobility that conquers Europe in the post-war period. The impact on the life expectancy gap is not trivial, as young males are much more prone than young females to die in traffic accidents. Interestingly, in the same period, since the Second World War, the pattern below age 18 becomes more and more blurred, reflecting the very low mortality at young ages and the increasingly random pattern of gender specific mortality. But above age 18, the

94

Patrick Deboosere

declining number of traffic deaths starting in the seventies did not change the pattern of male excess mortality. It is important to bear in mind that we are looking at a pattern of relative mortality with age-specific mortality declining in the latter quarter of the twentieth century for women as well as for men. Depending on the slower or faster age-specific declines, the impact on the gender differential in life expectancy will be different. To link the patterns in the Lexis plot with the gap in life expectancy, the decomposition method of Arriaga allows decomposing life expectancy differences for a given year in the age specific contributions4. In the left hand panel of figure 4 the decomposition is given for 1910, reflecting the mortality pattern of the early industrialisation in Belgium before the disturbance of the war years. The total difference in life expectancy between men and women is 3.29 years. Young girls and young women do still have some excess mortality, but this is largely compensated by adult male excess mortality above age 35. However, child mortality under age 1 is the single most important contribution to the gap in life expectancy accounting for no less than 44 % of the gender gap. Each of the other panels illustrates the evolution in contribution by age over time by plotting two subsequent decompositions. The gender gap in 1950 was just over 5 years with still 17 % due to child mortality under age 1. But while the importance of child mortality is retracting, male mortality now contributes almost for each single year to the life expectancy gap. In 1985 the life expectancy gap arrives at a maximum with 6.8 years of difference. Child mortality under age 1 still contributes with 2.5 %, but is now surpassed by mortality for all ages between 62 and 78. However, the evolution between 1950 and 1985 is not that bleak for males in all ages. Between ages 42 and 55, male mortality improved faster than among women, partially neutralizing the extreme increasing gap in older ages. After 1985 the gap starts to stabilise and finally declines; mortality in women over age 80 still improves faster than among men, but this is largely compensated by a strong improvement in male mortality at a much faster pace than among women under age 80. Child mortality under age 1 contributes to 1.75 %, a lower contribution than all ages between 62 till 86, and the gap is almost reduced to the 1950 level. The evolution as depicted in figure 4 clearly shows how a largely biological factor in the gender gap, as expressed in the under age 1 mortality, is gradually declining in importance from a contribution of 44 % to the gender differential in 1910 to less than 2 % 100 years later. This decrease is the result of the tremendous progress in infant mortality over the period 1910–2010. On top of this, improved prenatal monitoring is probably also contributing not only to a decline in congenital diseases, but also to a decline in the gender differential in infant mortality.

4

Arriaga 1984.

The evolution of the gender gap in life expectancy in Belgium

95

Figure 4: Belgium: decomposition by age of the differences in life expectancy between males and females, 1910, 1950, 1985 and 2010 – red hatching for an age specific increase in the gender gap between observations, green hatching for an age specific decrease. Belgium 1910: Age specific contribu7on to 3.29 years gender LE gap (44%