Stephan Klasen Harvard University and World Bank
Address: The World Bank T-9106 1818 H Street NW Washington, DC 20433 phone: (202) 473-5407 fax: (202) 522-3124 e-mail: SKLASEN@WorldBank.Org
I want to thank Jeffrey Williamson for continued support, comments, and suggestions at various stages of this project. In addition, I want to thank Robert Allen, Claudia Goldin, Jane Humphries, Guido Imbens, Jonathan Morduch, Sheila Ryan Johansson, Juliet Schor, Amartya Sen, Peter Timmer, participants of the 1993 Cliometrics Society Conference, members of the Economic History Workshop, the Economic History Tea, the Program on Ethics and the Professions at Harvard and participants at seminars at Bates College, University of Pennsylvania, University of Utah, Williams Colege, George Washington University, Universitat Pompeu Fabra, and Witwatersrand University for many helpful comments and suggestions. Funding from the Center for European Studies at Harvard and the World Institute for Development Economics Research in Helsinki is gratefully acknowledged.
This female survival disadvantage is evaluated by introducing and testing a bargaining model to intrahousehold resource allocation. The empirical section applies and confirms the bargaining approach and finds that EFM is related to women's position in the remarriage market, the perceived relative value of her work as well as differences in each parent's intensity of altruism vis-ˆ-vis their children. Agricultural change which brought about a relative devaluation of women's work appears to be one of the factors accounting for the rise in EFM in late 18th and early 19th century Germany.
If the cow kicks off, mighty cross. If the wife kicks off, no big loss. Folk wisdom from Hesse, Germany Got a dead wife? No big deal, Got a dead horse? How you squeal. Folk wisdom from Franconia, Germany [1]
These current trends have increased the interest of demographers and economic historians in the issue of female/male mortality ratios in Europe to determine whether similar developments took place during early stages of European development (Johansson 1984, Humphries 1991, Tabutin 1978). This line of research has shown that there were episodes of excess female mortality in the late 18th and throughout the 19th century in rural areas of almost all European countries. Individual studies show that the intensity and timing of the problem varied considerably: Ireland experienced the worst and most enduring problems (Kennedy 1973), Sweden experienced the latest and shortest problem (Johansson 1984), with England and France being intermediate cases (Tabutin 1978, Humphries 1991). [2] The Model Life Tables developed by Coale, Demeny, and Vaughan (1983) which are derived from the mortality experience of four regions in Europe in the late 19th and early 20th century also show excess female mortality in high mortality environments, with the regions "South" and "West" showing the worst problems and "North" showing the least (see Table 1). While much of the findings on EFM relate to infants and children, several researchers as well as some of the Model Life Tables have also discovered considerable EFM among adults (Kennedy, 1973, Humphries 1991, see Table 1). [3]
Given the rural concentration and the timing of the phenomenon, most researchers have linked episodes of excess female mortality to aspects of agricultural change that appeared to have hurt the relative economic position of women and girls which in turn reduced their share of resources within the household. Explanations include the effect of enclosures, farm amalgation, declining protoindustrialization, changing agricultural technologies and cropping patterns on women's work opportunities (Humphries 1987, 1991, Allen 1992, Snell 1981); the effect of the increased monetization of the rural economy on the perceived value of women's labor (Johansson 1984); and increasing work burden for women involved in agricultural production (Kennedy 1973, Imhof 1979, Shorter 1991). While this literature has added much insight to the understanding of excess female mortality and the economic roles of women during the 18th and 19th centuries, there have been few attempts to link the demographic accounts of excess female mortality (Kennedy 1973, Imhof 1979) with the economic history literature on the changing economic position of women (Humphries 1987, Snell 1981, Allen 1992). [4]
This paper examines excess female mortality among adults in six regions of 18th and 19th century rural Germany using a uniquely rich sample of recently assembled village genealogies. [5] These genealogies contain demographic, economic, and social data of about 160,000 people who lived in these regions between 1600 and 1930, thus providing a unique opportunity to link the demographic analysis of excess female mortality with concurrent economic changes in Germany in the 18th and 19th century.
The analysis will show that there was large and rising EFM among adults in Germany in the late 18th and early 19th century, which declined only in the later parts of the century. Moreover, excess female mortality, which was particularly severe among land-owning married peasants, was indeed due to inequalities in the intrahousehold distribution of survival-related resources which can be best understood in terms of a bargaining model of marriage. Finally, changes in agriculture that led to a relative devaluation of women's labor appear to be responsible for the emergence of high levels of EFM.
The paper proceeds as follows: after defining EFM, I will develop a model of excess female mortality based on a bargaining approach and compare it to altruistic models of intrahousehold resource allocation. The subsequent section will introduce the data and present first evidence on excess female mortality. Section 5 will concentrate on married adults and study whether unequal allocation of resources was indeed responsible for EFM. Sections 6 and 7 relate EFM to the predictions of the bargaining model and link the findings to concurrent changes in agriculture, arguing that these changes can indeed account for EFM in late 18th and early 19th century Germany.
In industrialized countries, women now outlive men by about 5 to 8 years and experience lower mortality rates than men in every age group (UNPD 1992). Although a considerable portion of this excess may be accounted for by higher rates of smoking, drinking, dangerous driving, work-related hazards, homicides and suicides among males, there is evidence that women have at least a slight biological survival advantage in all age groups (Waldron 1983, 1985). This apparently biological survival advantage is not just a 20th century phenomenon. The earliest demographers, working in the 17th and 18th centuries, had already commented on the higher male mortality rates and the greater female resistance in most age groups (Graunt 1661, Sü§milch 1765).
There is a question, however, as to whether one should expect women of child-bearing ages in the 18th and 19th century to have lower mortality than men given that high fertility and the considerable hazards associated with childbirth surely served to increase female mortality rates (Shorter 1991, Loudon 1992). Indeed, Table 1 shows that some of the Model Life Tables show higher rates of female mortality during child-bearing years in high mortality populations. But the difference is slight and by no means universal suggesting that the hazards of motherhood did not always outweigh the "natural" survival advantage enjoyed by women of this age group (Preston 1976, Coale and Demeny 1983, Reichsamt für Statistik 1894, Sü§milch 1765). Nor is it clear that the higher female mortality rates during child-bearing ages observed in some of the Model Life Tables are due to the "natural" hazards of motherhood rather than reduced access to survival-related resources within the household (Shorter 1991).
For the purposes of this study, I will use the female-male mortality ratios found in the Model Life Tables as a standard to determine the existence of excess female mortality. A particular advantage of this standard is that it is itself based on the historical experience of European countries so that many region- and time specific factors should be captured. At the same time, using the Model Life Tables presents two biases working in opposite directions. On the one hand, they are likely to understate EFM given they are based on populations that also discriminated against women and girls in the intrahousehold allocation of resources (Humphries 1991, Johansson 1984, see essay 1). In order to minimize this bias, I will rely on the Model Life Tables "North" based on Scandinavian populations which showed the smallest amount of gender bias in mortality (Johansson 1984, see Table 1). On the other hand, this standard may underestimate the importance of maternal mortality in the 18th and 19th century given that it is based on time periods when fertility and maternal mortality rates had already fallen considerably (Loudon 1992). In order to address this bias, later parts of this paper will specifically exclude the portion of EFM that is directly attributable to maternal mortality. [7]
Table 2 shows the "expected" female/male mortality rate ratios based on the Model Life Tables "North" for different mortality levels. Actual mortality rate ratios for a given level of life expectancy that are significantly higher than the expected levels in Table 2 will thus constitute EFM as defined here. Thus the standard for identifying and evaluating EFM here is:
EFM = F( FMA / MMA, FMN / MMN )
FMA : Actual female mortality rate
MMA : Actual male mortality rate
FMN : Expected female mortality rate ("North" model)
FMN : Expected male mortality rate ("North" model)
Although this particular way of defining EFM may be most suitable for an analysis of EFM in 18th and 19th century Germany, it is important to point out that the results presented below do not depend on the particular choice of Model Life Tables (see C oale, Demeny, and Vaughan 1983). Using regional Model Life Tables developed by the United Nations or simply defining EFM to exist whenever the female/male mortality ratio was greater than unity would lead to similar results (United Nations 1982).8
3. Altruism, Bargaining, and Intrahousehold Resource Allocation
Since the household is the most important locus in the allocation of survival-related resources, any theory of excess female mortality must focus on intrahousehold resource allocation.9
Models of intrahousehold resource allocation generally fall into two categories: altruistic models where the existence of an "effective altruist" ensures the allocation of resources within the family (Becker 1981),10 and bargaining models where the re source allocation is dependent on each partner's next best alternative to the current marriage (McElroy and Horney 1981, McElroy 1992, Manser and Brown 1980, Sen 1990, Nash 1950).11 Although these models differ in important ways, the altruist model can b e seen as a particular solution to a more general bargaining problem (Manser and Brown 1980, Thomas 1990).
Figure 1 illustrates the set-up for this marriage bargaining problem. Point A represents the next best alternative to the current marriage, i.e. remaining single or being married to someone else. Bargaining models refer to point A as the "breakdown position" or "threat point" and Becker would see it as the respective positions of the two partners in a competitive marriage market.
Given that most models posit considerable gains to marriage such as the advantages derived from the specialization of labor, the existence of household-specific public goods as well as intangibles such as love and companionship (Manser and Brown 1980, Becker 1981), there is likely to be a potential arrangement that dominates being single, i.e. the "marriage-possibility frontier" includes point A in Figure 1. Area d (the shaded area) shows the allocations that are preferred by both partners to their n ext best alternative.12
If there is more than one point in area d, the question arises as to how these gains from marriage are distributed. While both altruist models as well as cooperative bargaining models suggest that the outcome will be pareto-optimal, i.e. the outcome of the marriage bargain will lead to an allocation on line BC in Figure 1, the particular solutions proposed are quite different. I will consider three solutions: Becker's model of altruism, the simple Nash-bargaining solution, and an augmented and infor mationally enriched cooperative bargaining solution derived from Sen (1990).
Becker's "effective altruist" solution to the bargaining problem is for the altruistic household head to dictatorially choose a point that maximizes his utility function (Manser and Brown 1980, Becker 1981).13 Given that his utility function includes the well-being of his family as one of the arguments, it is in his best interest to allocate resources to them as well.
Formally, the dictatorial altruist's problem is as follows:14
max Um ( Xim, Zm, Uk (Xik, Zk) ) k=f,ch
subject to the usual budget constraint:
Y = piXi = wm tm + ym + wf tf + yf
where: Zk=f(Xik)
The Xs refer to market produced goods, Z is a household-produced welfare outcome such as longevity or (as analyzed in the empirical section of this paper) the probability to survive a certain age interval. Clearly, this health and survival outcome de pends, among other things, on the allocation of X goods such as food and health care. W refers to wages (implicit or explicit) for market and household work, y is non-wage income.
Although wages, work hours, and non-wage income enter the budget constraint separately, an important implication of Becker's solution to the bargaining problem is that small changes in family income will have the identical effect on each partner's wel fare regardless of who contributed to the increase (McElroy 1992). Therefore changes in survival of husband and wife will also only depend on total income:15
(1) Zk = f (Y) k=m,f
In contrast, bargaining models would lead to a different solution. In the Nash cooperative bargaining solution, the partners maximize the product of their utility gains from marriage (representable by the largest possible area of rectangle AEFG in Fi gure 1):
N = [(Um (Xim, Zm) - U0m (Xim, Zm)] [(Uf (Xif, Zf) - U0f (Xif, Zf)]
U0 refers to the "breakdown position", i.e. Point A in Figure 1 (McElroy 1990, Thomas 1990, McElroy and Horney 1981, Sen 1990). Given that the ability to earn income as well as the potential position in a remarriage market clearly affects the breakdo wn position, the allocation of resources and Z within marriage will therefore depend not only on total income Y but also on who is contributing to this income. For example, an increase in women's wages or non-wage income will, ceteris paribus, lead to la rger improvements in her well-being than an equivalent increase in men's wages since the former improves her breakdown position. Hence:
(2) Zk = f (wk tk , yk , EEPk) k=m,f
where EEP refers to "extramarital environmental parameters" such as the legal, economic, and social circumstances influencing the marriage and remarriage market (McElroy 1990).
A particular problem associated with the Nash bargaining solution is that there is no enforcing mechanism in the framework of cooperative bargaining so that the question arises as to whether it is a plausible description of the real world.16 Moreover , Sen argues that the informational base upon which these solutions depend is too narrow and may therefore limit their predictive efficacy (Sen 1990). In order to enrich the informational base, Sen suggests a model that considers implicit notions of dese rt and legitimacy as important factors influencing the outcome of the bargaining problem. These notions of desert and legitimacy are, in turn, functions of perceived contributions to the household by the two partners. Perceptions may be very important h ere since the contributions of both partners are not easily quantifiable and will depend on shared notions of value for different types of labor (Sen 1990). For example, it can be easily imagined that labor that is remunerated in cash may be more valued than unpaid household labor. At the same time, actual contributions are also likely to influence the distributional outcome by changing the perceptions of each partner's legitimacy to claim resources.
In addition, Sen argues that differences in the partners' perceived interests can influence the distributional outcome (Sen 1990). If, for example, the wife is altruistic, particularly vis-ˆ-vis her children, while her husband is less so (or not at all), she may be willing to sacrifice for the good of the family by reducing her share of the gains from marriage while her husband may not (see Thomas 1990).
Formalizing and adapting such an informationally enriched bargaining model to an analysis of EFM will yield the following directional relations influencing survival of husband and wife:17
(3) Zk = f (EEPk, ak (wk tk , yk), bk U(Xch, Zch))
Here a is a perception parameter measuring how much the actual contributions are valued by the two partners. In the extreme case where both partners view non-market work to be completely worthless and only market work to be of value, ak can also be i nterpreted as the marketed share of each partner's work. The parameter b measures the respective altruism intensity of father and mother vis-ˆ-vis their children which will determine how many of their own resources they are willing to forgo further their children's welfare.18
Applying this model to EFM in Germany leads to a number of testable predictions regarding the correlates of EFM, some of which distinguish it from altruistic models or the Nash bargaining model.
First, as with most bargaining models but in contrast to altruist models, changes in the breakdown position caused by changes in the relative position of males and females in the marriage and remarriage market or the relative value of male and female labor will have an impact on EFM within marriage.19 Second, this informationally enriched bargaining model suggests that increasing monetization of the rural economy could influence EFM if it altered the perceptions of the relative value of male and f emale contributions (Johansson 1984, Sabean 1990).20 Third, the presence of children can affect EFM (net of maternal mortality) if fathers' and mothers' perceived interests vis-ˆ-vis their children differ (Thomas 1990).
All three implications turn out to play an important role in an explanation of excess female mortality in Germany. Before discussing them, I will first describe and present the data on EFM.
4. The Data
The data used in this analysis includes vital statistics of over 160,000 people who were born in six regions in Western Germany between 1600 and 1870. They are based on village genealogies from about 80 villages in Germany (Figure 2). The compilatio n of village genealogies was a project started under the Nazi rule in 1937 to serve a number of ideological purposes (Knodel 1975, Imhof 1990).21 The village genealogies include vital statistics from parish records, any official registration data, and us ually a description of relevant historical events in the locality. The plan was to compile 30,000 of those village genealogies, but by the end of the Nazi rule only 30 were finished and published (Knodel 1975:295). Local historians continued the unfinis hed projects in recent decades (for purposes entirely different than those pursued by the Nazi rulers) so that by the late 1980's about 100 village genealogies were published.
A first analysis of a portion of the data was undertaken by Knodel (1988), who studied fertility and child mortality in a sample of 14 villages, three of which are also included in the present analysis.22 Imhof collected and transcribed the genealogi es and made the entire set available in 1990, together with a detailed description of the set, the regions included, and a number of summary statistics (Imhof 1990).
The data not only include vital statistics for each individual, but more than half of the cases provide additional information about parents' occupation, own occupation, marital status, cause of death, social class, religion, legitimacy of the childre n, and birth order.23 Thus the data are not only of interest to demographers but can be used to explore a variety of economic and historical questions (Imhof 1990). Although Imhof used a small portion of the data for a demographic analysis of excess fem ale mortality (Imhof 1979), this paper is the first to use the entire data set for an analysis of the demographic, social, and economic correlates of EFM in Germany.24
Figure 3 gives a first impression of EFM using the pooled sample of all villages and comparing it to national mortality statistics which are available after 1870. Among the 20-45 and 45-65 age groups, there is considerable EFM throughout the time peri od covered. A comparison of the rural data used here and the national data suggests that EFM was much higher in rural areas.25 It appears to have peaked in the last decades of the 18th and the first decades of the 19th centuries, but is statistically si gnificant in every period.26
The data from the pooled sample mask considerable regional variation and sharp fluctuations (Table 3). Between the ages of 20 and 45 Herrenberg and Hartum experienced the lowest and Schwalm the highest incidence of EFM. Between the ages of 45 and 65 , Ostfriesland has the lowest, Schwalm the highest, and Ortenau the most persistent EFM. Although not present in every region, there is a perceptible rise of EFM in most regions in the late 18th and early 19th century, declining thereafter. EFM also app ears to differ among occupational groups. Figure 4 shows that in families where the male occupation was agriculture, EFM was particularly high in the decades between 1780 and 1800 after which it declined.27
In order to determine whether unequal allocation of resources is indeed responsible for women's higher mortality, more must be known about the causes of death. Unfortunately, the causes of death are known only for about 2% of the sample.28 These data , analyzed in appendix 2, indicate that adult women have between 26% and 124% higher mortality from causes related to nutritional intake, suggesting that women might indeed have suffered from higher rates of undernutrition relative to men.
This view is also supported by the seasonality of mortality for men and women. Figure 5, which shows the monthly female-male ratio of deaths as well as the seasonality of total deaths, demonstrates that females die at relatively higher rates during w inter while they die at lower rates in the summer. Since mortality during the winter months is dominated by pulmonary diseases which are highly nutrition-sensitive while the gastroenterological diseases that predominate in the summer are less dependent o n the nutritional status (Knodel 1988, Lunn 1991, Imhof 1981a, Prinzig 1900), these findings suggest that insufficient nutrition might be the cause of women's higher mortality during winter. Moreover, food supplies are more scarce in winter and spring th an they are in summer indicating that unequal allocation of food in times of acute food shortage might be the cause of women's relatively higher mortality in winter (BrŸgelmann 1982). The close (and statistically significant) link between the seasonality of the female/male mortality ratio and the overall seasonality of deaths supports this hypothesis (Figure 5).29
Finally, there is ample anecdotal evidence supporting these developments. Shorter (1991), Abel (1978), and Sabean (1990) report that women received lowest priority in food allocation, often even less attention than livestock. A doctor describing the medical situation in southern Germany in 1804 also comments on the vastly inferior nutrition women receive as well as the much reduced medical attention they get leading them to higher mortality rates from age 10 onwards (Metzler 1822).
These pieces of evidence point to unequal allocation of resources as the most likely candidate for EFM among adults. It is now crucial to ascertain the determinants of this survival disadvantage for women.
5. Marriage, Maternal Mortality, and Excess Female Mortality
To isolate the factors influencing adult EFM, I estimate a logistic regression model that can be used to predict the conditional probability for an individual to die in two adult age intervals (ages 20 to 45 and ages 45 to 65) given that they were a live at the beginning of the interval. Independent variables include (husband's) occupations and (husband's) occupational class.30 Class and occupation should offer important clues about factors influencing (perceived) contributions, while the marital s tatus category will determine whether distributional struggles within marriage are indeed at the heart of EFM. In addition, I include dummies for religion, birth cohorts, and region. Occupation, class, marital status, and religion were then interacted wi th sex to determine sex differentials.
Table 4 shows the results for the 20 to 45 age group.31 To facilitate interpretation of the results, I calculate predicted male and female death probabilities at the omitted categories. While the occupational interactions all turn out to be insignif icant, the occupational class variable is significant with lower class women being relatively better off than their middle-or upper-class equivalents, a finding similar to contemporary developing countries (Das Gupta 1987, Basu 1992, and Agarwal 1985).
By far the biggest determinant of EFM, however, turns out to be marital status. While married men have less than a third of the mortality of single men, the effect is much smaller for women. Consequently, there is considerably higher male mortali ty for single people and a 90% higher female mortality for married people. This is quite an extreme result and in contrast to contemporary findings about the relative longevity of single and married people. In Germany in the 1970's, married people, both men and women, outlived their single contemporaries by about 7 years with only a slightly higher benefit from marriage for men (Imhof 1981b).32 In the historical data presented here, however, men profit tremendously while women suffer from very high EFM within marriage. Interpreting these results in terms of the bargaining framework discussed above, it appears that the survival gains from marriage--while present for both as expected--disproportionately accrue to men putting them closer to point C than point B in Figure 1.
This result deserves further scrutiny. One possible factor influencing the unusually large mortality for single people is that the regression includes all those single people who died before reaching marriageable ages. Given that the average age at marriage in the sample is between 24 and 28 years (Knodel 1988) many people who died between 20 and 28 did not have the opportunity to get married, thereby overestimating the survival advantage of married people.
In order to omit this selection bias, I restrict the analysis to people between the ages of 30 and 45 to see how much this affects the results. Sure enough, Table 5 shows that the difference between married and single people now becomes considerably smaller. Men reduce their mortality by about 40% by marrying, while the reduction for women is only 10%.33 But the "marriage hazard" for women remains huge, leading to a 63.5% higher mortality for married women over their married husbands (at the omitte d categories). Thus while women continue to gain from marriage, they get a much smaller share of the gains and thus are the huge relative losers of marriage.
In order to further investigate the female marriage hazard, I checked whether maternal mortality might play an important role in the relative disadvantage of married women. It is important to note at the outset, however, that maternal mortality by it self cannot be the sole explanation of EFM within marriage since married women still do better than single women. Moreover, the bigger factor in the relative disadvantage of married women is derived from the huge absolute advantage men derive from marria ge which is presumably unrelated to the hazards of childbearing.
There are two possible ways of testing whether maternal mortality affects EFM within marriage. The first is to restrict the analysis to childless couples, which make up about 10% of the families in the data set. The results, presented in Table 6, sh ow that the absence of children does not change the female survival disadvantage within marriage. Both males and females gain from marriage, but men seem to capture the majority of the gains, leading to 75% higher female mortality within marriage.
Another way to determine the effect of child-bearing and maternal mortality is to directly identify those women that died as a result of childbirth. The United Nations Population Division defines maternal mortality as the death of a mother during preg nancy and within 42 days of childbirth from diseases directly related to the pregnancy and childbirth (UNPD 1992).34
Table 7 shows that maternal mortality is indeed substantial. The average rate is 1126 per 100,000 live births, a rate similar to the 10 least developed countries in the world today and higher than in other European countries at the time (UNICEF 1993, Loudon 1992). The rate varies considerably by birth order showing high maternal mortality for first-borns, lower for second, third, and fourth children, and then increasing slowly again.35 Table 7 also shows time trends in maternal mortality. When sor ted by mother's birth cohorts, the rate drops from 1278 for mothers born between 1720 and 1760 to 922 for mothers born between 1760 and 1800, after which increases to 1197 again (see also Knodel 1988).
While much of maternal mortality is certainly a result of the poor state of medical knowledge, many childbirths, and general poverty (Shorter 1991), this very high rate of maternal mortality and the fact that it increases towards the end of the period could at least partially be the result of poor and worsening nutrition and health conditions for mothers leading to these high maternal mortality rates. Moreover, there are numerous contemporary reports citing the unwillingness of husbands to call and p ay for doctors in the case of birth complications suggesting that part of maternal mortality was in fact due to economic choice rather than biological circumstance.36
In order to check to what extent maternal mortality accounts for EFM within marriage, I exclude all women who died of maternal mortality from the analysis. Table 8 shows that, in spite of high rates of maternal mortality, it accounts for only a small portion of EFM. EFM within marriage drops from 90% in Table 6 to 70% in Table 8.37 This shows quite clearly that factors other than the hazards of childbirth account for most of this extraordinary survival disadvantage of married women.
Table 9 shows the factors influencing adult mortality between the ages of 45 and 65. Women continue to be the relative losers of marriage, although the marital status variable is less significant and also smaller in magnitude. Moreover, similar to the findings for the younger age group, lower class women do better, although not significantly so.38
The two most important observations emerging from this discussion are the reduced incidence of EFM among lower class people and the considerable relative mortality hazard suffered by married women of both adult age groups. While both husband and wife are better off than their unmarried contemporaries, men appear to reap most of the gains from the marriage arrangement. While maternal mortality matters, EFM is only partially accounted for by high rates of maternal mortality.
The task ahead is to explain these observations in terms of the parameters of the model described above, namely breakdown position, perceived interests, and perceived contributions.
7. Breakdown Positions and Perceived Interests
Following the theoretical discussion in section 3, differences in the breakdown position could account for the disparity in the intra-marriage distribution of resources. One indication of the better breakdown position of men is their superior positio n in the remarriage market. In fact, of men who become widowed in this age interval, close to 100% remarried while only 48% of women widowed before age 45 were able to find another mate (see also Sabean 1990).39 Among middle-class people consisting main ly of landowning peasants, this disparity is even larger with middle-class widowers remarrying 3.7 times as often as middle-class widows. For lower-class people the male-female disparity is much lower, with widowers remarrying twice as often as widows. Remarriage occurred usually very fast, within the first 6 months for most men and tended to involve considerably younger women (see also Lee 1977:284, Knodel 1988, Imhof 1981c) suggesting that, in contrast to widows, widowers were in a considerably better economic position to attract a new spouse (Moser 1980, Sabean 1990).
The superior position of men in the remarriage market is a clear indicator of men's improved breakdown position within marriage. Contemporary references to the seeming indifference of husbands to the death of a wife and the frequent complaints by wiv es over severe mistreatment by their husbands (Sabean 1990) underscores that wives were believed to be easily replaceable and, in fact, quickly replaced (Shorter 1991, Kennedy 1973, Segalen 1981, see also the folk wisdoms at the beginning of this paper).4 0
In order to test whether the remarriage market is linked to EFM within marriage, I investigate whether remarriage rates can explain the regional and temporal variation in EFM. Figure 6 shows that the remarriage ratio (defined as percentage of widower s remarrying divided by the percentage of widows remarrying) in the various regions closely tracks EFM between 20 and 45 in those regions.41
To formally test this link, I regress a panel of EFM between 20 and 45 on a remarriage ratios in the regions at four time periods. The results, shown in Table 10, indicate that the remarriage ratio significantly influences EFM between 20 and 45.42
Given that lower class people have a smaller remarriage ratio, the smaller EFM among lower-class people observed in the last section could be a result of the smaller difference in breakdown positions of married people.43 Indeed the correlation coeffi cient between the remarriage ratio and the percentage of lower-class people in the region is 0.41 suggesting considerable correlation.44 The question arises whether the differences in the remarriage market can account for the lower EFM among lower-class people.
If the lower EFM among lower-class people were entirely due to the remarriage market, including the percentage of people who are lower-class in the panel regression should increase the standard errors of the remarriage and class variable considerable and add little to the R-squared. Table 10 shows, however, that adding the percentage of people who are lower-class only slightly changes the standard errors and increases R-squared considerably suggesting that in addition to the smaller gap in remarriage prospects other factors contribute to the relatively better situation of women in the lower classes.45
Turning to the effect of perceived interests on EFM within marriage, the attitudes of mothers and fathers towards their children could influence EFM within marriage. Table 11 can sort out some of those issues. Restricting the analysis to married cou ples with children, I test whether and how family size and sex composition affects the mortality of both parents as a proxy of how much parents are willing to sacrifice for their children.46
The results are striking. The presence of children raises the mortality of both parents significantly, which is to be expected given that family resources have to be shared among more people. But the mortality of fathers increases by much less as a result of more children than the mortality of mothers. This is not due to maternal mortality since I specifically excluded women who died as a result of childbirth. It thus appears that mothers are more willing than fathers to share their resources with their children, an impression confirmed by contemporary accounts and also found in today's developing world (van DŸlmen 1990, Thomas 1990).47
The sex composition of the family also appears to influence the mortality of mothers and fathers. In particular, a family with more girls leads, ceteris paribus, to lower mortality for fathers than a family with mostly boys indicating that fathers mi ght be more willing to reduce their share of household resources in favor of boys (see also Thomas 1990).
The bottom of Table 11 combines the effects of size and composition on parents' mortality. Assuming a typical father and mother presiding over the mean number of children with the mean sex composition, it asks how the addition of a child of undetermi ned sex, a girl or a boy will change mortality for each parent. The addition of a child increases the mother's mortality by more than twice the rate of the father's increase. But the sex of the child leads to considerable differences in the father's and mother's mortality. For the father, the addition of a boy will increase mortality by 10% more than when a girl is added, indicating that the father does not appear to reduce his share of resources as much if a girl is born. For the mother, the story i s the reverse. The addition of a boy increases mortality by 10.2%, while a girl increases her mortality by 14.2%.
Both of these observations shed light on how the perceived interests of the marriage partners might influence the solution to the bargaining problem. In particular, the results suggest that mothers take a stronger interest in the well-being of her ch ildren and a higher interest in girls, while fathers take a weaker interest in the well-being of their children and focus it more heavily on boys.48 Both of these results concord with the augmented cooperative bargaining model while they do not appear to be consistent with a standard Nash bargaining solution which does not allow for diverging interests and motivations.49
Apart from differences in the breakdown position as well as the perceived interests of the two partners, differences in perceived contributions may account for the difference in the mortality outcomes of married middle-class adults and, more important ly, may account for the time trends in EFM alluded to in section 4. In order to evaluate the influence of this factor, more must be known about agricultural change in Germany in the 18th and 19th century.
7. Agricultural Change, Perceived Contributions, and Excess Female Mortality
While there is little reliable information about agricultural change in the 18th and early 19th century, four developments took place that all are likely to have changed intrahousehold resource distribution by changing male and female contributions to the peasant household. These developments seem to have contributed to the particularly high levels of EFM among peasants in the late 18th and early 19th century.
In agriculture, the sexual division of labor at the time dictated that women were mainly in charge of livestock production and small-scale subsistence agriculture while men were involved in grain production (Imhof 1979, Moser 1984, van DŸlmen 1990, Sabean 1990). This division was not complete as women continued to help with some tasks in grain production. Similar to developments described by Snell (1981) in England, changes in the use of tools in late 18th century Germany served to intensify the s exual division of labor. In particular, the move away from the sickle to heavier harvest instruments reduced the ability of women and children of their ability to contribute to grain production (Oberschelp 1982: 59, see also Boserup 1970). Such a redu ction in women's work opportunities is likely to have reduced the contributions of women to the household, thereby lowering their share in the distributional bargain. There is very little precise data on the timing and regional variation of this technolo gical change so that this theory cannot be tested formally at this point.
This intensified sexual division of labor did not affect all sectors equally. Contemporary accounts suggest that among lower-class people (landless laborers and people involved in spinning and weaving), the sexual division of labor was much less pron ounced (Moser 1984, Oberschelp 1982). In fact, many social commentators complained that men's willingness to do female labor and vice versa was usurping the natural order of things (van DŸlmen 1990, Moser 1984). This similarity in tasks equalized actual contributions as well as the perceptions of both partners regarding their legitimacy to claim resources which may be the other reason lower class women appeared to do relatively better in terms of EFM than middle-class women.
A second development with potential repercussions on the intrahousehold resource distribution was the increasing monetization of the rural economy throughout the 18th century.50 The driving force behind this development was the commutation of in-kind services into monetary payments to the feudal lords (BrŸgelmann 1982). Henning (1978) claims that by the end of the 18th century, virtually all in-kind services had been eliminated (see also Sabean 1990). Given that peasants were required to deliver up to 40% of their production to the feudal lord (Moser 1984), they were now required to increasingly market their produce. Data on marketed surplus clearly indicate that it increased substantially in the late 18th century (Hagen 1986). Given that the vas t majority of the marketed surplus consisted of grain sales, the male-controlled portion of agricultural production was increasingly monetized while the female subsistence and livestock production continued to remain outside of the market. Moreover, even when the produce was mainly based on women's labor (such as livestock products as well as some root crops), men did most of the marketing and were therefore able to get control over the proceeds, a source of bitter complaints and marital disputes for ma ny women (Sabean 1990).51 As suggested by the model, such a change would increase EFM if the monetized male contribution to household production was considered more valuable than the non-monetized female contribution and if men had more control over the proceeds from the marketed surplus (Johansson 1984, Sen 1990).52
In addition to these two factors, the relative value of women's work in livestock production and men's work in grain production diverged in the period under analysis. In particular, there was a large shift in relative prices, allegedly already starti ng in 1650, favoring grain and hurting animal products (Abel 1978). Between 1670 and 1780, grain prices rose by 170% while the price of beef rose by only 55% (Abel 1978). This shift in relative prices intensified during the late 18th century when grain prices accelerated. In Figure 7, I have used price series from Gšttingen (Lower Saxony) to construct relative prices. The graph represent the 5-year moving average of the price of a "malter" of rye per "pound" of beef in Gšttingen (Gerhard 1990).53
Two potential explanations could account for this. One could be that large improvements in livestock production, relative to grain production, increased livestock supply, thereby increasing relative (grain/livestock) prices. Since the demand for live stock products is typically elastic, that would lead to increasing revenues from livestock production. If this were the case, women should have been in a better position.
A number of factors speak against this hypothesis and suggest that, instead, livestock production was declining relative to grain production, thereby further eroding women's earnings. First, most agricultural historians comment on increasing grain yi elds and improved technology in grain production, while livestock production did not improve much until the middle of the 19th century (Abel 1979, Sabean 1990). An estimate of the total product in agriculture in Prussia around 1800 shows that only 24.1% was derived from livestock production while 53.3% came from grain production (Abel 1979). This low proportion of livestock production is believed to be considerably below the level of the mid-18th century.
One important factor in the low and apparently falling livestock production in the late 18th and early 19th century is the beginnings of enclosures in Germany which significantly reduced the common areas which were exclusively used for livestock produ ction. That reduced the number of livestock that could be sustained in the villages, thereby reducing work opportunities for women (Henning 1979, Abel 1979, Moser 1984, Sabean 1990). Thus a development that is believed to have contributed to EFM in Brit ain is likely to have been at work in Germany as well (Humphries 1991).
Data about quantities of livestock also support a story of stagnant or declining livestock production. Although there is very little reliable data on livestock production from the 18th century, Gerhard (1990) was able to find data about livestock sa les in Gšttingen for six decades between 1737 and 1797. Although they do indicate slightly increasing sales, when adjusted for increases in the rural population during that time, livestock sales per peasant actually declined for cows, calves, and sheep throughout the 18th century in Gšttingen.54
In addition, livestock production dropped dramatically at the beginning of the 19th century due to the losses of the Napoleonic wars. Estimates of the loss of livestock between 1800 and 1820 vary from 15% (Henning 1979) to more than 30% (Finckenstein 1951).
Finally, estimates of meat consumption in Germany also indicate that livestock supply was not rising during that time. Agricultural historians estimate that meat consumption was declining ever since the 15th century when it was believed to be ab out 100kg per person (Abel 1979, van DŸlmen 1990). In the late 18th century, meat consumption continued its rapid decline dropping from 28kg in 1770 to 18kg in 1800, reaching its lowest point at the end of the Napoleonic Wars with 13.6kg per person (Abel 1978).
Thus it appears that supply is more likely to have contracted rather than expanded during the period of analysis and can therefore hardly explain the shift in relative prices. Instead, it appears to be the case that rapidly increasing population was the main source of a demand-induced shift in relative prices (Moser 1984, Abel 1978). After a reduction of the population by about a third as a result of the 30-Year War (1618-48), Germany's population regained the prewar level of 16 million by about 172 0 and thereafter population growth accelerated leading to increases of about 80% between 1740 and 1800 (Abel 1978, 1979, Marschalck 1984). Labor demand did not keep up with rising supply and real wages began to fall. The same source from Gšttingen (Gerh ard 1984, 1990) shows that real wages did indeed fall drastically in the late 18th and early 19th century and continued to be low until the middle of the 19th century (see also Henning 1979, Abel 1979, Sabean 1990). A consequence of these falling real wa ges and the ensuing rise in poverty is for the population to substitute cheaper grain for expensive meat which is in accord with the falling meat consumption mentioned earlier. To test this link, I correlate change in real wages to changes in relative pr ices and find the correlation coefficient to be 0.58.
Shifts in demand towards grain and away from livestock boosted revenues from grain production, thereby reducing the relative value of women's labor. In response to rising prices, increasing grain acreage further augmented the men's perceived contribu tions to the household, again shifting the intrahousehold resource allocation in their favor (Henning 1979, Abel 1979).
Starting in the 1820's and accelerating after the 1850's, the situation began to reverse itself. Real wages began to rise, leading to increasing demand for livestock production (Abel 1976). At the same time, livestock production increased rapidly, f ar outpacing gains in grain production (Finckenstein 1951, see Figure 9). These gains came largely as a result of barn feeding, improved animal feed, and improved varieties of livestock (Henning 1979, Sabean 1990). Meat consumption is estimated to have risen to 21.6kg in 1840 and 29.4kg in 1873 (Abel 1978). These factors all served to increase the contributions of women to the household and are therefore likely to have contributed to the decline in EFM, particularly among peasants.
As a partial test of this link between relative prices and EFM, I regress decadal averages of EFM among agriculturists on decadal averages of relative prices. The results, shown in Table 12, give some confirmation to the explanation advanced above. In particular, there is a positive, sizeable, and marginally significant link between relative prices and adult mortality for people of families that report agriculture as their main profession.55 Such a link is a further confirmation of the bargaining a pproach developed here and does not appear to support Becker's "altruist" model in which changes in the relative value of male and female labor should not affect the intrahousehold distribution of resources.
Another indication about changes in the relative value of women's and men's labor is to study relative wages. The very little information on relative wages also supports a decline in the value of women's labor in the late 18th century. In Speyer, a small town between the Saar area and the Ortenau region, relative female/male wages for unskilled agricultural day labor declined by about 15% between 1711 and 1789 (Elsas 1935).
Clearly, this kind of evidence does not definitively settle the issues surrounding the causes of EFM. But in the absence of more reliable data from 18th and 19th century German economic history, this test appears to confirm that changes in agricultur e account for the high levels of EFM in late 18th and early 19th century Germany.
8. Conclusion
This study has shown that excess female mortality was a pervasive phenomenon in rural Germany in the 18th and 19th century. It was concentrated among middle-class married women and among peasants who suffered from a combination of the hazards of chil dbirth as well as much reduced shares of the gains from marriage. While EFM continued throughout, it was largest in the late 18th and early 19th century. Childbirth and maternal mortality did play an important role in raising mortality of married women, but cause of death data, the seasonality of deaths, the analyses explicitly excluding maternal mortality, as well as a wealth of anecdotal evidence indicate that married women suffered from unequal allocation of survival-related resources.
The bargaining approach developed here appears to explain the existence and change in EFM quite well. In particular, women's precarious position in the remarriage market as well as differences in the perceived interests served to weaken their bargain ing position considerably. In addition, shifts in agriculture, including technological change, increasing monetization, and shifts in relative prices led to a decline in the perceived contributions of women and appear to be able to account for the time t rend of excess female mortality.
Given the findings of this paper, further research should be directed into analyzing the question whether shifts in production, sex-specific work opportunities, and relative prices in other countries led to similar episodes of EFM.56 Similarly, mor e research should investigate the potential links between agricultural change and EFM on the one hand, and fertility decisions on the other to understand the relations between the temporal patterns of EFM and the fertility transition in Europe. Finally, analyses of sex bias and EFM in developing countries might also take note of this research and investigate what role technological change, shifts in the sexual division of labor, and shifts in the relative value of male and female production play in gener ating gender inequality in the intrahousehold distribution of resources.
Bibliography
Abel, Wilhelm. 1979. Geschichte der Deutschen Landwirtschaft (Stuttgart: Verlag Eugen Ulmer).
-----. 1978. Agrarkrisen und Agrarkonjunktur (Hamburg: Verlag Paul Parley).
Ahmad, Asiv and Jonathan Morduch. 1993. Identifying Sex Bias in the Intra-Household Resource Distribution in Bangladesh. Harvard University. Mimeographed.
Agarwal, Bina. 1985. "Women and Technological Change in Agriculture: The Asian and African Experience," in Iftikhar Ahmed (Ed.) Technology and Rural Women (London: Allen and Unwin).
Allen, Robert 1992. Enclosure and the Yeoman (New York: Oxford University Press).
Basu, Alaka M. 1992. Culture, the Status of Women, and Demographic Behavior (Oxford: Oxford University Press).
Becker, Gary S. 1981. A Treatise on the Family (Cambridge: Harvard University Press).
Behrman, Jere. 1988. "Intrahousehold Allocation of Nutrients in Rural India: Are Boys Favored? Do Parents Exhibit Inequality Aversion?" Oxford Economic Papers 40: 32-54.
Blasius, Dirk. 1987. Ehescheidung in Deutschland (Gšttingen: Vandenhoeck&Ruprecht).
Boserup, Esther. 1970. Women's Role in Economic Development (New York: Academic Press).
BrŸgelmann, Jan. 1982. Der Blick des Arztes auf die Krankheit im Alltag 1779-1850. Inauguraldissertation (Berlin).
Chen, Lincoln et al. 1981. "Sex Bias in the Family Allocation of Food and Health Care in Rural Bangladesh." Population and Development Review 7: 55-70.
Coale, Ansley, Paul Demeny, and Barbara Vaughan. 1983. Regional Model Life Tables and Stable Populations (Princeton: Princeton University Press).
Das Gupta, Monica. 1987. "Selective Discrimination against Female Children in Rural Punjab, India." Population and Development Review 13: 77-100.
Deaton, Angus, 1989. "Looking for Boy-Girl Discrimination in Household Expenditure Data." World Bank Economic Review 3:1-15.
D'Souza, Stan and Lincoln C. Chen. 1980. "Sex Differences in Mortality in Rural Bangladesh." Population and Development Review 6: 257-270.
van DŸlmen, Richard. 1990. Kultur und Alltag in der FrŸhen Neuzeit (Munich: C.H. Beck)
Elsas, Moritz. 1936. Umri§ einer Geschichte der Lšhne und Preise in Deutschland (Leiden: A.W. Sijthoff's Vitgeversmaatschappij).
Finckenstein, H. Graf. 1951. Die Entwicklung der Landwirtschaft in Preussen und Deutschland (Berlin).
Folbre, Nancy. 1992. Rotten Kids, Bad Daddies, and Public Policy. Paper presented at the IFPRI-World Bank Conference, 12-14 February 1992 (Washington, D.C.).
Gerhard, F. 1984. Lšhne im Vor-und FrŸhindustriellen Deutschland (Gšttingen: Otto Schwartz).
-----. 1990. Preise im Vor-und FrŸhindustriellen Deutschland (Gšttingen: Otto Schwartz).
Graunt, John. 1661. Natural and Political Observations Made upon the Bills of Mortality (London).
Hagen, William H. 1986. "The Junkers' Faithless Servants: Peasant Insubordination and the Breakdown of Serfdom in Brandenburg-Prussia, 1763-1811" in Richard Evans and W. R. Lee (Eds.) The German Peasantry (London: Croom Helm).
Henning, F. H. 1979. Landwirtschaft und Lþndliche Gesellschaft in Deutschland (Paderborn:Schšningh).
Humphries, Jane. 1991. "Bread and a Pennyworth of Treacle: Excess Female Mortality in England in the 1840's." Cambridge Journal of Economics 15: 451-473.
-----. 1990. "Enclosures, Common Rights, and Women." Journal of Economic History 50: 17-42.
Imhof, Arthur E. 1990. Lebenserwartungen in Deutschland vom 17. bis 19. Jahrhundert (Weinheim: VCH Verlagsanstalt).
-----. 1981a. "Unterschiedliche Sþuglingssterblichkeit in Deutschland, 18. bis 20. Jahrhundert- Warum?" Zeitschrift fŸr Bevšlkerungswissenschaft 7:343-382.
-----. 1981b. Die Gewonnenen Jahre (MŸnchen: Beck).
-----. 1981c. "Remarriage in Rural Populations and in Urban Middle and Upper Strata in Germany from the Sixteenth to the Twentieth Century," in J. Dupaquier (Ed.) Marriage and Remarriage in Populations of the Past (New York: Academic Press).
-----. 1979. "Die †bersterblichkeit Verheirateter Frauen im Fruchtbaren Alter." Zeitschrift fŸr Bevšlkerungswissenschaft 5: 487-510.
Johansson, Sheila R. 1991. "Welfare, Mortality, and Gender. Continuity and Change in the Explanation of Male/Female Mortality Differentials over 3 Centuries." Continuity and Change 6 :135-177.
-----. 1984. "Deferred Infanticide: Excess Female Mortality in Childhood" in Glenn Hausfater and Sarah B. Hrdy (Eds.) Infanticide (New York: Aladin).
Kennedy, Robert 1973. The Irish: Emigration, Marriage, and Fertility (Berkeley: University of California Press).
Klasen, Stephan. 1994a. "'Missing Women' Reconsidered." World Development (forthcoming).
-----. 1994b. "Family Composition, Gender Preference, and Parental 'Investment': Excess Female Mortality Among Infants and Children During Early German Development." Harvard University, Mimeographed.
-----. 1993. Gender Inequality and Development Strategies: Lessons from the Past and Policy Issues for the Future. World Employment Programme Research Working Paper No.41 (Geneva: ILO).
Knodel, John 1975. "OrtssippenbŸcher als Quelle fŸr die Historische Demographie." Geschichte und Gesellschaft 1:288-324.
-----. 1974. The Decline of Fertility in Germany 1871-1939 (Princeton: Princeton University Press).
-----. 1988. Demographic Behavior in the Past (Cambridge: Cambridge University Press).
Kooreman, Peter and Aria Kapteyn. 1990. "On the Empirical Implementation of Some Game Theoretic Models of Household Labor Supply." Journal of Human Resources 25: 584-598.
Kynch, Jocelyn and Amartya K. Sen. 1983. "Indian Women: Well-Being and Survival." Cambridge Journal of Economics 7:363-380.
Lee, W.R. 1977. Population Growth, Economic Development, and Social Change in Bavaria 1750-1850 (New York: Arno Press).
-----. 1984. "The Impact of Agrarian Change on Women's Work and Child Care in Early-Nineteenth-Century Prussia in Fout (Ed.) German Women in the 19th Century (New York: Holmes and Meyer).
Loudon, Irvine. 1992. Death in Childbirth (Oxford: Clarendon Press).
Lunn, Peter G. 1991. "Nutrition, Immunity, and Infection," in Roger Schofield et al. (Eds.) The Decline of Mortality in Europe (Oxford: Clarendon Press).
Manser, Marilyn and Murray Brown. 1980. "Marriage and Household Decision-Making: A Bargaining Analysis." International Economic Review 21: 31-44.
Marschalck, Peter. 1984. Bevšlkerungsgeschichte Deutschlands im 19. und 20. Jahrhundert (Frankfurt: Suhrkamp).
McCloskey, Donald C. 1985. The Rhetoric of Economics (Madison: University of Wisconsin Press).
McCrate, Elaine. 1987. "Trade, Merger, and Employment: Economic Theory on Marriage." Review of Radical Political Economy 19:73-89.
McElroy, Marjorie B. 1992. The Policy Implications of Family Bargaining and Marriage Markets. Paper presented at the IFPRI-World Bank Conference 12-14 February 1992 (Washington, D.C.).
-----. 1990. "The Empirical Content of Nash-Bargained Household Behavior." Journal of Human Resources 25: 559-583.
----- and Mary J. Horney. 1981. "Nash-Bargained Household Decisions: Toward a Generalization of the Theory of Demand." International Economic Review 22: 333-349.
Meitzen, August 1866. Der Boden und die Landwirtschaftlichen Verhþltnisse des Preussischen Staates (Berlin: Verlag von Paul).
Metzler, Franz. 1822. Versuch einer medizinischen Topographie der Stadt Sigmaringen. (Freiburg: Verlag Herder).
Miller, Barbara. 1981. The Endangered Sex: Neglect of Female Children in Rural North India (Ithaca: Cornell University Press).
Mokyr, Joel. 1993. Technological Progress and the Decline of European Mortality. Paper presented at the AEA Meeting, Anaheim, California.
Moser, Josef. 1984. Lþndliche Klassengesellschaft 1770-1848 (Gšttingen: Vandenhoeck&Ruprecht).
-----. 1980. "Familie und Soziale Plazierung in der Lþndlichen Gesellschaft am Beispiel des Kirchspiels Quernheim im 19. Jahrhundert," in JŸrgen Kocka (Ed.) Familie und Soziale Plazierung (Opladen: Westdeutscher Verlag).
Nash J. 1950. "The Bargaining Problem." Econometrica 18:155-62.
Nicholas, Stephen and Deborah Oxley. 1993. "The Living Standards of Women during the Industrial Revolution, 1795-1820." Economic History Review 46: 723-749.
Oberschelp, Reinhard. 1982. Niedersachsen 1760-1820 (Hildesheim: Verlag August Lax).
Oren, L. 1974. "The Welfare of Women in Labouring Families: England 1860-1950 in M. Hartman M. and L.Banner (Eds.) Clio's Consciousness Raised: New Perspectives on the History of Women (New York: Harper&Row).
Preston, Samuel H. 1976. Mortality Patterns in National Populations (New York: Academic Press).
----- and Michael Haines. 1991. Fatal Years (Princeton: Princeton University Press).
Prinzig, Friedrich. 1900. "Die Kindersterblichkeit in Stadt und Land." JahrbŸcher fŸr Nationalškonomie und Statistik 75:593-645.
Reichsamt fŸr Statistik. 1894. Statistik des Deutschen Reiches. Neue Folge, Band 44 (Berlin).
Sabean, David. 1990. Property, Production, and Family in Neckarhausen, 1700-1870 (Cambridge: Cambridge University Press).
Samuelson, Paul A. 1956. "Social Indifference Curves." Quarterly Journal of Economics 70: 1-22.
Schultz, T. Paul. 1981. The Economics of Population (Chicago: University of Chicago Press).
Segalen, Martine. 1984. "'Avoir sa part': Sibling Relations in Partible Inheritance Brittany," in Hans Medick and David Sabean (Eds.) Interest and Emotion (Cambridge: Cambridge University Press).
-----. 1981. "MentalitŽ Populaire et Remarriage en Europe Occidentale," in J. Dupaquier (Ed.) Marriage and Remarriage in Populations of the Past (New York: Academic Press).
Schofield, R, D. Rehrer, and A.Bideau (Eds.). 1991. The Decline of Mortality in Europe (New York: Clarendon Press).
Sen, Amartya K. 1990. "Gender and Cooperative Conflicts" in Irene Tinker (Ed.) Persistent Inequalities (Oxford: Clarendon).
-----. 1989. "Women's Survival as a Development Problem." Bulletin of the American Academy of Arts and Sciences 43: 14-29.
Shorter, Edward. 1991. A History of Women's Bodies (New York: Basic Books).
Sivard, Ruth Leger. 1991. World Military and Social Expenditures (Washington, DC: World Priorities Inc., 1991).
Smith, Ken R. and Cathleen D. Zick. 1994. "Linked Lives, Dependent Demise? Survival Analysis of Husbands and Wives." Demography 31:1-13.
Snell, Keith 1981. "Agricultural Seasonal Unemployment, the Standard of Living and Women's Work in in the South and East 1690-1860." Economic History Review 34, No. 3.
Statistisches Bundesamt. 1972. Bevšlkerung und Wirtschaft 1872-1972 (Stuttgart: Kohlhammer).
SŸ§milch, Johann Peter. 1765. Die Gšttliche Ordnung in den Verþnderungen des Menschlichen Geschlechts (Berlin: J.C. Spener).
Tabutin, Dominique. 1978. "La SurmortalitŽ feminine en Europe avant 1940." Population 33: 121-148.
Thomas, Duncan. 1990. "Intra-Household Resource Allocation." Journal of Human Resources 25: 634-664.
Tilly, Louise and Joan Scott. 1978. Women, Work, and Family (New York: Holt, Rinehart, and Winston).
UNICEF. 1993. The State of the World's Children (New York: Oxford University Press).
United Nations Population Division (UNPD). 1977-1993. Demographic Yearbook (New York: United Nations).
United Nations. 1982. Unabridged Model Life Tables (New York: United Nations).
Visaria, Pravin. 1967. "Sex Ratio at Birth in Territories with Relatively Complete Registration." Eugenics Quarterly 14: 132-142.
Waldron, Ingrid. 1985. "What do we Know about Causes of Sex Differences in Mortality? A Review of the Literature." Population Bulletin of the United Nations 18: 59-76
-----. 1983. "The Role of Genetic and Biological Factors In Sex Differences in Mortality," in Alan D. Lopez and Lado T. Ruzicka (Eds.) Sex Differentials in Mortality (Canberra: Australian National University).
Ware, Helen. 1981. Women, Demography, and Development (Canberra: Australian National University).
Williamson, Jeffrey G. 1990. Coping with City Growth During the First Industrial Revolution (Cambridge: Harvard University Press).
Appendix 1: Data Description, Reliability, and Selection
The villages are clustered in six regions of rural Germany. These regions cannot be considered a random sample of all of Germany for a number of reasons. First, they are all in regions of former West Germany thus omitting the Eastern provinces (see Figure 2). Second, the set is made up of entirely rural communities which, according to Knodel (1975) can be considered to be roughly representative of rural Western Germany (Knodel 1975:298), thus omitting developments in the former Eastern provinces (f ormer GDR, Western Poland, former East Prussia).
Knodel also compared 8 of the village genealogies (two of which are included in the current data set) with other available data and finds that fertility indices, infant mortality, overall mortality, nuptuality, and illegitimacy in those 8 villages are consistent with other available data (Knodel 1975).
Imhof also scrutinized the data set for a number of inconsistencies which are most likely due to underregistration. To deal with this issue, he suggested a number of time periods to be eliminated in the data set. All of the recommendations mentioned by Imhof were implemented before an analysis of the data was undertaken.
The data set does not include all entries from the village genealogies but only those people whose parents' date of death is known and occurred within the region of analysis. Data from marriages that occurred after 1850 were not included (Imhof 1990: 64).57 Imhof shows that this selection leads to an underrepresentation of illegitimate births, while it does not appear to influence other relevant demographic or socioeconomic variables (Imhof 1990: 64).
About 25% of all people left the village they were born in. In many cases, individuals migrated between the villages of the regions so that their death was recorded. In cases when they did not, their date of death is often not known. In order to co nstruct a life table, assumptions must be made as to when these people should be taken out of the so-called "risk population" (those that could potentially have died), i.e. when they presumably migrated. Following Imhof, the following assumptions are mad e. If the date of marriage or the birth of a child is known, migration is assumed to have taken place a year after the wedding or the birth of the last child. The bias this may introduce is that infertile couples may have staid in the region much longer after marriage leading to a slight upward bias in the mortality rates for adults.
For those for whom the marriage date is not known, they were assumed to have migrated between ages of 15 and 45, where the actual emigration data was randomly assigned with equal probability of migration in each age bracket (Imhof 1990:74).
A final problem is how to deal with stillbirths, which are excluded in a life table analysis. The village genealogies only inconsistently report whether a child that was recorded to have born and died on the same day was actually born alive or dead a nd such reporting partially depends on cultural and religious practises of the region. Imhof assumes that about 25% of the recorded children that died on the day they were born were not stillbirths and should thus be included. This leads to a stillbirth ratio of 24.3/1000 which appears to be slightly lower but generally in line with other available estimates of the time (Imhof 1990:72). In any case, errors made in dealing with this issue will not affect estimates of adult mortality.
In the remainder of this appendix, I will present a brief overview of the mortality experience. Table A-1 shows mortality rates for males and females in the entire sample sorted by periods.58 It shows little or no improvements in mortality rates thr oughout the 18th and 19th centuries. Infant mortality rates appear to increase throughout the period, while adult mortality appears to decline starting in the mid-19th century. These findings coincide with Knodel's analysis of 14 villages in Germany as well as findings from other European countries (Schofield et al. 1991).
Two factors influence the mortality experience. One is the general disease environment and the other are factors such as individual nutrition and health provisions. While the paper focuses on differences in the second factor, quite clearly the disea se environment also plays a significant role in influencing mortality, particularly at a time when the state of medical knowledge was still limited (Mokyr 1992).59 Therefore one would expect that regional differences are much more important for the morta lity experience than occupational differences within a given region. Table A-2 confirms these expectations. The regional differences are much larger than the occupational difference within the Saar region (this is also true for other regions). Moreover , within agriculture, the mortality rate difference between a landowning peasant and a landless laborer within the same village are very small as well and indeed sometimes nonexistent (see below, see also Knodel 1988:71).
Table A-3 presents the full results for the logistic regression predicting death probabilities between 20 and 45. Also here, the regional variation clearly dominates the occupational variation. While only some of the occupational categories are sign ificant predictors of death probabilities, the regional categories are highly significant and large in magnitude. While Hartum, Ostfriesland, and Saar have higher mortality than the people in the omitted Ortenau category, Schwalm has much lower and Herre nberg similar levels of adult mortality. Religious differences also matter. Lutherans and Jews have much lower, and Reformed Protestants insignificantly higher mortality than the omitted Catholics.60
Finally, the birth cohort dummies indicate the time trend already suggested in Table A-2. While there is little change in overall mortality throughout the 18th and early 19th century, the cohorts born after 1840 have much lower death probabilities in their age group.
Appendix 2: Causes of Death Data
For about 2% of the cases, causes of death data are available. Even among those, some are of limited quality given that the local priest rather than a trained medical professional noted down the cause. Also, some reported "causes" such as "died as i nfant" or "dropped dead suddenly" do not give precise clues of what the actual medical reason for the death was (BrŸgelmann 1982).
Moreover, the cases for which causes of death are known are unlikely to be a random sample of the entire data set. In particular, women are underrepresented making up only a third of the sample with recorded deaths. Also, it is likely that unnatural deaths such as accidents, suicides, war victims, etc are overrepresented given that they make up 25% of the sample of all recorded deaths.
Nevertheless the classification of the causes allows some limited inferences as to the importance of intrahousehold allocation in generating EFM. I classified causes of death into four categories. The first group are causes where nutrition is likely to have played a role in the death of the person. These mostly include tuberculosis, pneumonia, and other diseases of the pulmonary system as well as people who are reported to have died of "weakness", "starvation", and "consumption" (see Johansson 199 1, Lancaster 1990, Lunn 1991, BrŸgelmann 1982). A second group just refers to causes of death related to childbirth and is thus specific to women. A third group encompasses all causes of death where nutrition is unlikely to have played a role such as s mall pox, heart disease, cancer, and all diseases where the description in the genealogy does not allow a precise determination of the diagnosis, while a fourth group combines all unnatural causes of death such as homicide, suicide, war, beaten, drowned, and frozen to death as well as all accidents. This last group makes up a surprising 25% of all recorded causes of death with surprising numbers of suicides (2.5%), accidents (8%), drowned (5%), and death in war (3%). While these unnatural deaths are ver y likely oversampled, the sex ratio of these four groups of death allow us to further refine the analysis of excess female mortality.
Table A-4 shows that men and boys are four times more likely to die from unnatural causes than adult women. This suggests that the sex-specific mortality rates reported above understate the female mortality disadvantage due to inadequate health and n utrition. Tabutin (1978) also found this to be true for England in the 19th century. Moreover, it appears that indeed women are suffering more from mortality associated with inadequate nutrition given that the adjusted female/male death ratio among adul ts ranges from 1.26 and 2.24. Interestingly enough, the implied nutrition disadvantage is highest during child-bearing ages where women have a 124% higher fatality from nutrition-sensitive diseases.61
2 Anthropometric data on the relative heights of English and Irish women and men confirm the relative deprivation of Irish and English rural women in the late 18th and early 19th century (Nicholas and Oxley 1993).
3 The Model Life Tables extra- and intrapolate from actual life tables taken from these four regions in order to generate a hypothetical mortality pattern at 24 different levels of life expectancy. For details, see Coale, Demeny, and Vaughan (1983).
4 An exception is Humphries (1991) who links data on EFM in Britain in the 1840's to the effect of enclosures and industrialization on women's economic roles.
5 Klasen (1994b) usus the same data to investigate excess female mortality among children.
6 "Excessive" here refers to the portion of female mortality that is a consequence of unequal intrahousehold allocation of food and medical care. Equal allocation does not imply that husband and wife receive exactly the same amount of food and health care, but that, given their different needs, are similarly adequately nourished and healthy. Thus EFM can develop if, for example, women are burdened with an increasing workload while receiving the same amount of food as before so that they are suffering from a larger nutritional shortfall (Shorter 1991, Imhof 1979, Brügelmann 1982). For a discussion on definitional and measurement issues surrounding the notion of "equal treatment", see Chen et al. (1981), Johansson (1991), and essay 1.
7 For details, refer to section 5 of this paper. Apart from maternal mortality, other factors, such as gendered behavioral patterns leading females to different exposure to disease, might also account for changes in sex-specific mortality patterns. For a discussion, see Johansson (1991).
8 Note also that, from a policy perspective, it may be impossible to reduce excess female mortality by redirecting resources to women without increasing male mortality as a result.
9 This is especially true in the past where the household was the key provider of health and nutrition expenditures. Today, the state may also play an important role in allocating survival-related resources and may therefore be in a position to generate and combat gender inequalities. For a discussion, see Klasen (1993).
10 For an altruist to be "effective" the equilibrium utility of the "beneficiary" must be larger than in initial distribution, i.e. the altruist augmented the beneficiary's utility via a transfer. For this and other conditions, see Becker (1981) and McCrate (1987). Given the conditions, the"effective altruist" in Becker's model is typically the male household head.
11 Another well-known model assumes that intrahousehold allocation "takes into account the 'deservingness' of consumption of different members" (Samuelson 1956)
12 The metric of measurements typically are utilities or non-utility based welfare indicators such as 'functionings' (Sen 1990). Given the link between resource distribution and survival in countries where food and health care is scarce and undernutri tion rampant, longevity, health, and survival may be the most important well-being indicators (Thomas 1990).
13 For different justifications for this particular solution, see Becker (1981) and Thomas (1990).
14 For a similar formulation, see Thomas (1990), McElroy and Horney (1981), and Becker (1981).
15 Note that this is only true for small changes in wages or non-wage income and in the case of interior solutions. Large changes might affect the optimal sorting in the marriage market or might reverse the positions of altruist and beneficiary (McElr oy 1992, Becker 1981). Also, since I have no reliable data, I neglect the influence of labor-leisure choices on intrahousehold resource allocation.
16 In fact, Nash himself did not see his solution as empirically accurate, but rather thought of it as normatively desirable (Sen 1990).
17 By abandoning the Nash solution, it will be difficult to arrive at a closed-form solution without imposing more structure on the model (Kooreman and Kapteyn 1990), so that only directional relations can be determined.
18 The amount of resources they are willing to forgo will also depend on the number of children.
19 In today's world, the possibility of easily accessible divorce provides the direct link between the position of a person in the remarriage market and his/her breakdown position within marriage. In the 18th and 19th century, divorce was less common (although legal and surprisingly frequent particularly after 1800, see Blasius 1987, van DŸlmen 1990, and Sabean 1990) so that the link between the remarriage market and the breakdown position is less direct.
20 It is important to note that allocation decisions are unlikely to respond directly and immediately to changing economic circumstances, but will be mediated by institutions and cultural practises (Humphries 1991, Sen 1990).
21 Knodel determined, however, that the ideological interests of those who authorized the studies did not alter the veracity of the demographic and socioeconomic entries (Knodel 1975).
22 Knodel's sample also includes villages from regions not included in the analysis, namely 4 villages in Waldeck in the state of Hesse and 3 villages from Bavaria. Throughout this paper, I will compare my findings to Knodel's, wherever there is an ov erlap in the analysis. While in many cases, the findings are similar, there are also important differences.
23 The prevalence of these socioeconomic variables varies across regions. There is complete data for virtually all cases in the Ortenau region, while there is occupational data for only about 10% of the population in the Schwalm region, with the remai ning regions lying inbetween. Consequently, nearly 40% of all cases with occupational data come from Ortenau. I carefully checked whether there is a systematic bias in excess female mortality in cases with occupational data and cases without. The bias turned out to be insignificant so that the sample with occupational data can be considered representative of the entire data set.
24 A number of questions arise as to the reliability of these data as well as necessary corrections and adjustments that were needed before they could be analyzed. Appendix 1 takes up these issues in detail and provides an overview of the mortality ex perience.
25 National mortality statistics disaggregated by states show that EFM was higher in predominantly rural states than in urban areas (Reichsamt fŸr Statistik, 1894).
26 Statistical significance was determined by assuming the null hypothesis that the actual female-male mortality ratio does not differ from the "expected" ratio derived from the "North" Tables. In addition, the results were checked using the delta-met hod to determine whether the EFM-ratio (the female/male mortality ratio divided by the "expected" ratio) is significantly different from 1. Note, however, that in this context the notion of statistical significance is of somewhat limited use. Given that the underlying data represents all people whose parents died in the villages in the time period under analysis, we are dealing with a population and not a sample so that statistical inference does not have an obvious interpretation. Nor could one say, a s mentioned in the appendix, that these villages are a random sample of Germany. One could potentially assume that the actual data is a sample from all possible populations that could have lived in these villages, but this assumption has been criticized severely (McCloskey 1985).
27 The village genealogies unfortunately do not record women's occupation (Imhof 1990). Occupational data is available only for a portion of the data set but I could not detect a systematic bias.
28 See appendix 2 for details on the classification of diseases, and the analysis of the cause of death data.
29 On similar issues in developing countries, see Behrman (1988), Sen (1990) and Chen et al. (1981).
30 Occupational class was determined by Imhof based on the actual occupational description appearing in the parish record. Middle-class refers to land-owning peasants, craftsmen, and small traders, lower class to landless laborers, textile workers, an d unskilled labor.
31 The Wald Statistic reported in column 3 is distributed as Chi-square. Table 4 reports only a selection of the results. For a full tabulation and discussion of the results, refer to appendix 1.
32 For similar evidence from the US, see Smith and Zick (1994).
33 Part of this marriage bonus could be a result of the fact that less healthy people have a lower chance of getting married which is held to be one of the reasons for the survival advantage of married people in contemporary industrialized societies. This would influence the substantial results of this paper only if less healthy women have a higher chance of getting married than less healthy men. Studies about the selection bias in today's USA suggest that this is not a likely possibility (Smith and Zick 1994).
34 I cannot determine whether pregnancy or childbirth caused the death of a women unless the birth of the child (even if stillborn) is recorded. On the other hand, I also will not be able to distinguish between causes of death that occur within 42 da ys of childbirth. The first problem underestimates maternal mortality, while the second overestimates it, since women may have died from diseases unrelated to childbirth. For this analysis, data from Ortenau were, once again, excluded.
35 Knodel (1988) finds slightly lower overall levels, but the same parity and time pattern in his 14 villages, where maternal mortality was highest between 1825 and 1874.
36 For a discussion, see Shorter (1991) and Metzler (1822). Loudon (1992), however, places small importance on poverty as a determinant of maternal mortality rates.
37 Unfortunately, I could not determine maternal mortality rates in Ortenau so that I excluded Ortenau from the analysis in Table 8. This renders the results in Table 8 not entirely comparable to Table 6. Therefore, I also redid the regression in Tab le 6 excluding the Ortenau data. In that regression, EFM within marriage is 79.7% so that, in reality, maternal mortality only reduces EFM by ten percentage points.
38 I also checked whether the number of children mattered for female mortality in this age group under the assumption that many children hurt the physical health of a women and might lead to higher mortality even in the post-childbearing years. But th is is actually not the case. Using the number of children as the independent variable, it turns out that women with more children have a slightly (and insignificantly) lower mortality. Two factors may be at work here that off-set delayed health effects of multiple child births. One is that the women who suffered most under many childbirths are already dead and only the strongest survived to age 45 which then represents a particularly healthy selection. The other is that large families might pay off i n terms of increased support from their children in this age group where women might have to rely on them already. As with the 20 to 45 age group, controlling for children does not eradicate the marriage hazard although it is much smaller now.
Occupational factors do not have a significant influence on mortality rates, but female agriculturists appear to have among the highest EFM. Moreover, women in textile production also suffer from highly elevated mortality rates compared to their hus bands in this age group.
39 Among people of all age groups, 34.1% of all widowers remarried while only 14.9% of all widows did.
40 See Shorter (1991) and Sabean (1990) for a discussion.
41 The easy availability of remarriage for men might not only affect their breakdown position, but also reduce their perceived interests in their wife's well-being as well as their view of their wife's perceived contributions. Note that in the Schwalm region, the close relation between EFM and remarriage ratios does not hold. It turns out that other tests for gender bias among children, that perform extremely well for the other regions, also do not show the expected relations for the Schwalm region w hich suggests either data problems or a poor fit of the economic explanation of EFM in that particular region (see Klasen 1994b).
42 Note that this result could not obtain if causation ran from EFM to the remarriage ratio. In fact, any likely causation between EFM and the remarriage ratio would suggest that the higher EFM, the lower the remarriage ratio since EFM reduces the rel ative number of eligible women and should therefore increase the remarriage chances of widows.
Note also that two mechanisms could account for this link between remarriage ratios, the breakdown position, and EFM. The one emphasized earlier was that all men can use their high remarriage chances as a "bargaining chip" within their current marria ge, thereby driving down the share of resources going to women. Another potential mechanism is that the high remarriage ratio implies some marriages where men get married for the second, third, or fourth time to much younger women who have not been marri ed before. The large age, resource, and "marriage count" gap in these marriages may lead to particularly large gap in the breakdown position in these marriages and could thus contribute to the link between remarriage ratio and EFM (Sabean 1990).
In order to test for the latter effect, I replace the martial status variable with two alternatives in the logistic regression predicting death between 20 and 45 (not reported here). One is the gap in "marriage count" and the other the age difference s of the spouses. It turns out that, indeed, women are considerably worse off if they are in their first marriage and married to someone who is in his second, third, or fourth marriage. When the age difference is included, women who are married to older men do worse than women who are married to men with the same age. These effects are considerable and always significant suggesting that the difference in breakdown positions in these marriages is correlated with high EFM. They do not, however, account for all of the "marriage hazard" described above. Even after controlling for these factors, married women whose spouses are the same age and have the same "marriage count" still have EFM on the order of 30-50%.
43 The reason for such a difference is likely to be that male propertyless peasants laborers and textile workers have fewer economic resources to attract a new mate than middle-class peasants do (Moser 1984).
44 When the Schwalm region is excluded, the correlation rises to 0.47. Sabean (1990) who closely examined a village neighboring the Ortenau region finds another relation between marriage and economic class which supports the findings in the present st udy. In particular, he finds that divorces were largely concentrated among the richer groups of the village he studied and was particularly concentrated among propertied peasants suggesting that marital relations were particularly strained among these gr oups (Sabean 1990).
45 When data from Schwalm are excluded, the higher collinearity of the remarriage ratio and EFM in the other regions makes the coefficient on class insignificant when class and the remarriage ratio are included. Also, the R-squared does not increase b y much suggesting that in the four remaining regions, the lower remarriage ratio among lower class people may explain their lower EFM.
46 This test amounts to a variant of a test for sex bias frequently used in the development economics literature where researchers look for changes in the consumption of adult goods in response to additions of male or female children. For a discussion , see Ahmad and Morduch (1993), Deaton (1989). Note that in Table 11 I include the number of children born per year alive in the age interval as a proxy for family size instead of the overall number of children born. Using the overall number of children born would introduce a serious selection bias given that mothers and/or fathers who died early in the interval had little opportunity to have children.
47 But quite clearly, the burden does not rest with mothers alone given that fathers also suffer higher mortality in larger families. Similar results have been found for developing countries. See Thomas (1990).
48 This conclusion does not imply that given that the outcome was shaped by the diverging interests of the marriage partners we should be unconcerned by this aspect of EFM. Higher mortality as a result of unequal treatment should be seen as a social p roblem even if women might, in some sense, be understood to have consented to it. For a discussion, see Sen (1990) and McCrate (1987).
It is possible that the an additional factor influences this link between the higher mortality of women in the presence of additional children. In particular, if high fertility rates are a consequence of a low opportunity cost of women's time and EFM is another consequence of such a low value of women's time, then this link between EFM and additional children could be a reflection of such a link between fertility and excess female mortality. I am currently in the process of investigating this potent ial link between EFM and the demographic transition.
There is also an alternative hypothesis which could account for the apparent gender preference effect leading mothers to die at higher rates if a girl is born. If both parents, and fathers in particular, have a strong son preference and the mother fa iled to deliver a male infant, she might be punished by receiving smaller shares of resources (or feels that she "deserves" fewer resources because she failed to produce a son, see Kynch and Sen 1983, Miller 1981). Klasen (1994b) deals with the issue of gender preference in more detail.
49 Also, this model questions Becker's assumption of altruistic (male) household heads and suggests that women appear to act more altruistically (see also Folbre 1992).
50 Note that at the time, serfdom still prevailed in most of rural Germany. While it was on the retreat, it was finally abolished only in the first two decades of the 19th century (Henning 1978, Abel 1978).
51 For a survey of similar issues in the context of developing countries see Ware (1981).
52 Also, just the mere fact that males were given larger amounts of cash, they were able to spend more on their own personal consumption. For a discussion in the context of developing countries, see Thomas (1990).
53 Gerhard adjusted the prices for changes in the definition of these weights as well as differences in the silver content of currency. "Malter" defines a volume similar to a barrel. "Pound" is a weight close to 500 grams.
54 This came at a time when the population of Gšttingen was rising rapidly suggesting that meat consumption was clearly on the decline.
55 The small magnitude of the coefficient is deceptive given that the mean relative price is 67.53. At the mean, a 10% increase in relative prices would lead to about a 13% increase in EFM in the 20 to 45 age group and to a 10% increase in EFM in the 20 to 65 age group.
56 For example, did the large rise in grain prices in England and most European countries at the turn of the 18th to 19th century contribute to the rise in EFM there as well? Did the Corn Laws in England sustain EFM by raising the value of male agricu ltural production?
57 These selections were made by Imhof before he made the set available for use.
58 The life tables produced by the computer are calculated by cohorts which are then turned into period statistics using the method of demographic translation (see Imhof 1990).
59 Unless there are diseases for which women and men have a very different susceptibility, it seems doubtful that the local disease environment should have a major influence on EFM.
60 Only the Ortenau region has a considerable Jewish population concentrated in one village, namely Altdorf. Moreover, four regions (Ostfriesland, Schwalm, Hartum, and Herrenberg) have an overwhelming majority of one denomination so that this variable partially picks up regional differences.
61 Given that males are overrepresented in the sample with recorded causes of death, I adjusted the actual ratio of deaths from the four groups of causes by dividing it by the total ratio of females to males in the age group.