A two-tiered demographic system: ‘insiders’ and ‘outsiders’ in three Swabian communities, 1558–1914

Abstract

This paper presents first results from a project to reconstitute the demographic behavior of three villages in Württemberg (southern Germany) from the mid-sixteenth to the early twentieth century. Using high-quality registers of births, deaths, and marriages, and unusual ancillary sources, we improve on the family-reconstitution techniques pioneered by Louis Henry and applied to good effect by the Cambridge Group and other scholars. This paper focuses on simple, standard demographic measures, in order to provide a broad overview and support comparisons with other places. An extreme system of demographic regulation operated in these Württemberg communities until around 1870. This regulation created a two-tiered demographic system. A group of ‘insiders’ were able to marry, and experienced both high marital fertility and high infant and child mortality. A second group, consisting of ‘outsiders’, were prevented from marrying. Many, especially the males, left the community; those who stayed contributed to growing illegitimacy and associated levels of infant and child mortality that were even higher than for the offspring of ‘insiders’.

Keywords::

• western European marriage pattern
• politische Ehekonsens
• Wuerttemberg

Acknowledgements

We gratefully acknowledge financial support from the Economic and Social Research Council of Great Britain; the Leverhulme Trust; and the Economic Growth Center (Yale University). Markus Küpker and Janine Maegraith provided excellent research assistance. We are grateful to Adèle Rossouw and Marian Schmidt for additional help. For comments and suggestions we thank seminar participants at INED, Queen's University (Kingston), UCLA, the University of Copenhagen, and Williams College. This paper is part of a larger ongoing project; for detailed background information on the communities in question, see Ogilvie et al. (2009); for a more formal treatment of population and economic shocks, see Guinnane and Ogilvie (2008). We thank Cambridge University Press for permission to reproduce the map in Figure 1.

Notes

^1. Galor and Weil (2000) is the key reference in ‘unified growth theory,’ the most influential strand of recent theoretical work on long-run growth. Galor (2011) discusses more recent contributions. Guinnane (2011) discusses the demographic transition, stressing weaknesses in the way the growth-theoretical literature interprets the historical evidence.

^2. For suggestive evidence to this effect, see Ogilvie (2003), 196–200; Medick (1996), 359–360, 368–369.

^3. Examples of that literature include our own, earlier effort (Guinnane and Ogilvie (2008)), which uses two of the three sets of registers discussed here; Weir (1984); and Galloway (1988). The two volumes of the ‘Eurasia project’ (Bengtsson et al. (2004) and Tsuya et al. (2010)) include references to more recent studies of this type. The most famous project in this tradition is Wrigley & Schofield (1981).

^4. For Wildberg we have two distinct estimates for 1818 (1593 and 1594 persons) as well as for 1821 (1560 and 1786 persons). We use the average of the two totals, although the differences are too small to affect the results reported here. We discuss the sources underlying Figure 2, and the rest of the paper, in section 2 below.

^5. For more detail on the economic history of Wildberg, Ebhausen, and the immediate region, see Ogilvie (1997, 2003); Troeltsch (1897); Mantel (1974); Klaß (1987); Königliches Statistisch-topographisches Bureau (1862).

^6. For a more complete discussion of these demographic sources, see Ogilvie et al. (2009).

^7. See Ogilvie et al. (2009), pp. 85–86, p. 65. Catholics in Auingen remained under the pastoral care of the priest in nearby Magolsheim.

^8. That is, the value reported for 1750 is the sum of the counts for 1747–1753, divided by seven. Figure 3 includes all marriages. Below we discuss differences between first and higher-order marriages. Guinnane and Ogilvie (2008) report econometric analysis of marriages, deaths and births as reactions to economic ‘shocks,’ using the techniques employed by Weir (1984).

^9. Figure 5 includes both illegitimate births and stillbirths. The dividing line between stillbirths and live births is not always clear-cut, but the range of error is small relative to the levels in Figure 5. We discuss the issue in more detail in section 7 below.

^10. We discuss legitimacy and illegitimacy below.

^11. See Wrigley and Schofield (1981). Wrigley et al. (1997) undertake this with family reconstitutions.

^12. The year given on the horizontal axis in each graph is the beginning-year of the quarter-century in question: thus ‘1700’ indicates the quarter-century 1700–1724. Exceptionally, ‘1875’ indicates the period 1875–1914. To facilitate comparison to other studies, we present the raw numbers underlying Figures 678 in Appendix 2.

^13. The idea of marriage as central to demographic behavior goes back to Malthus. Hajnal (1965, 1982) argued that a distinctive ‘western European marriage pattern’ was central to population patterns in that region. Wrigley and Schofield (1981) is just one study that views nuptiality as the primary control in a pre-modern population.

^14. Guinnane (1997) focuses on post-Famine Ireland, but includes comparative discussion for much of Europe.

^15. One way to think about the politische Ehekonsens is to recall Malthus' objections to ‘early and improvident marriages’ and the English Poor Laws' supposed role in fostering such marriages. Malthus objected to some marriages and many of his contemporaries doubtless felt the same way. But they had no legal right to forbid them, in contrast to the right possessed and exercised by the Württemberg state and devolved by it to the local communities. On the legal framework and concrete implementation of the politische Ehekonsens in Württemberg, see Matz, 1980, esp. 44–5, 120–121, 181, 191; Borscheid, 1982, esp. 243–260; Kaschuba & Lipp, 1982, esp. 312–362; Schraut, 1989, esp. 91–121; Ehmer, 1991, esp. 53–55; Ogilvie, 1997, 61–63; Ogilvie, 2003, 51–54.

^16. Tables 345 include a bride or groom if she or he has a valid birthdate; that is, the number of events recorded from a given marriage need not be equal. For the seventeenth century, especially in Ebhausen, the register did not record prior marital status at marriage for some brides and grooms. Based on the conventions governing how individuals were identified in written documents in this society, however, brides of unrecorded marital status can be assumed to be unmarried in almost all cases, while grooms of unrecorded marital status can be assumed to be widowers. The ages at marriage reported in our tables are not sensitive to variations in treatment of this problem. If we simply exclude all brides and grooms of unrecorded marital status, the sample becomes smaller without appreciable changes in the mean or median age at marriage reported in the tables.

^17. Flinn, 1981 (Appendix 7) and Dennison & Ogilvie, 2013 (Table 1) report only a handful of studies with a female mean age at first marriage outside the range observed in our three communities. These two surveys do not report male age at marriage. Medick, 1996, Table 4.3, reports a similar pattern for Laichingen, with female ages at first marriage peaking at a mean of 27.4 in 1850–74. Knodel, 1988, 122–123 (Table 6.1), finds the same pattern for Öschelbronn, with female marriage age peaking at 27.4 in 1850–74. Kaschuba & Lipp, 1982, 331 (Table II.5), find a very similar pattern but even female higher marriage ages in Kiebingen, with a peak of 30.8 in 1850–54. A similarly high marriage age, typically peaking in the period 1850–70, is reported by Schomerus, 1976, 175 (Table 1) for women marrying factory workers in Esslingen and by Borscheid, 1982, 260 (Table 1), for brides of craftsmen in Nürtingen.

^18. The figure cited in the text includes stillbirths and covers the entire period of Knodel's study, roughly the eighteenth, nineteenth, and early twentieth centuries. Öschelbronn experienced a sharp fall in infant and child mortality in the late nineteenth century, similar to the decline we document below for our communities.

^19. Medick's estimates apparently include illegitimate children, and he uses a different approach to dealing with children whose date of death is not known, so his figures are not directly comparable to ours.

^20. Wrigley et al. (1997) estimate adult mortality from their family reconstitutions. The challenge for this type of analysis is to estimate the numbers and ages of adults at risk of death, which in a mobile population requires potentially heroic assumptions. A second approach is to rely on model life tables to infer adult mortality from the mortality levels estimated for children. We prefer this approach, like Knodel (1988), and return to it at the end of this section.

^21. In Ebhausen we lack death dates for 2225 of 9779 (or 23%) of the children linked to a birth. Of those, 686 were born after 1850. In Wildberg we have 11,377 births, of which 34% are not linked to a death. Of these 3942 births missing a death date, 1202 were born after 1850.

^22. One might be tempted to infer that a child remained in the community on the basis of information concerning the birth or death of siblings. We deliberately do not use such information, as it would definitely be selecting on individuals whose parents had more births and perhaps higher rates of infant mortality.

^23. A similar late-nineteenth-century decline in infant mortality is observed for the Württemberg village of Kiebingen by Kaschuba & Lipp, 1982 (Table II.93).

^24. Illegitimate children were approximately twice as likely as legitimate ones to be marked as stillborn in all three communities. This difference probably reflects both conditions of pregnancy (since unmarried women were even more likely to have to keep working late in pregnancy than married ones) and the conditions under which such children were born (since unmarried pregnant women were more likely to be poor and ‘outsiders’ to the mainstream community).

^25. Table 11 excludes stillbirths. In the next section we report very high fertility rates early in marriage, another indicator of low breastfeeding rates.

^26. The table reports only the proportion of children born in a given month who died in the first month of life. The basic patterns shown in Table 12 are similar to the patterns shown by the proportion that died within the first week or the first year of life. In discussing seasonality we focus on mortality up to the age of one month in order to reduce ambiguity concerning which ‘season’ the child experiences.

^27. The historical demography literature uses the term ‘endogenously’ differently. See Bourgeois-Pichat (1951).

^28. A simple summary is that in our communities infant mortality is higher relative to child mortality than the ‘East’ model life tables imply. Thus any estimate of adult mortality will depend on whether it is based on infant or on child mortality estimates.

^29. Our estimates of mortality to age five imply somewhat less severe mortality, ranging between levels 4 and 10.

^30. Following Knodel and others, if the first birth interval is shorter than nine months, we ‘back date’ the marriage so that the first interval has the same length as the mean for first births that did not involve bridal pregnancy.

^31. The estimates of age-specific fertility in Figure 10 and Table 14 exclude women who contribute less than one year of observation to a cell. This avoids the situation where, for example, a woman has a birth just before turning 25, thus producing a rate that is correct but misleadingly high.

^32. Since ‘m’ measures the reduction of fertility at older ages, it is best viewed as an indicator of ‘stopping.’ Earlier studies estimated ‘M’ and ‘m’ by taking logs and fitting a straight line. The estimates we report are based on the preferred approach of estimating the model in ‘levels’ by nonlinear least squares. As a rule of thumb most demographers consider a Coale–Trussell ‘m’ greater than 0.2 as evidence of parity-dependent fertility control.

^33. We also re-estimated ‘M’ and ‘m’ with weights for the number of woman-years of exposure in each cell. This did not yield appreciably different results.

^34. Borscheid (1982, p. 243ff and note 88) suggests a more precise periodization of the waxing and waning enforcement of the politische Ehekonsens. Building on his dating, we attempted a rough statistical test of the regulations' impact. For each community we estimated a separate AR(2) model of the number of marriages regressed on a fourth-degree polynomial in time plus dummies for every year in the range 1860–80. For Ebhausen and Wildberg the results show strong (positive) departures from trend for the period from the early 1860s to the late 1870s, as suggested by the graph. For Auingen the results are less clear-cut, in part because of the mini-boom in the 1840s noted above. We also adapted this approach to the periods Borscheid suggests; the results suggest that the periodization he describes does not account for the data from any of our communities. We do not believe these econometric analyses (or visual inspection of graphs) are the proper way to test the effects of the politische Ehekonsens. At the top of our research agenda is the question of how the restrictions' repeal affected marriage ages and marriage chances as revealed by the micro-level reconstitution data.

^35. Schlumbohm (1997), 649–650, 660–661; referring to Foucault (1975).

^36. In 1852, 67.3% of the German population lived in settlements under 2,000 inhabitants. See Twarog, 1997, pp. 288–289.

^37. The most famous version of this debate focused on Bavaria, but the same arguments are relevant for our communities. See Lee (1977b, 1978) and Shorter (1971, 1973, 1978).

^38. One might expect the end of the politische Ehekonsens to result in a reduction in average ages at marriage. Tables 3–5 actually suggest such a reduction, except for women in Auingen. But on closer examination the reduction in age at marriage during the quarter-century 1875–99 took place long after the repeal of the politische Ehekonsens, and cannot properly be attributed to it. This is not the case in our reconstitutions; if anything, there is at first an increase in the number of older brides and grooms. This observation is consistent with the idea that the repeal's major effect was on a cohort of people in their twenties and early thirties who would not have been allowed to marry under the old regime, and who had been forced to wait to marry past the age when they would have preferred to do so.

^39. At least this belief is persistent in the historical demography literature; for example, Kintner (1988, p. 237) claims ‘they [Catholics] tended to register stillbirths as infant deaths, thereby inflating the recorded infant mortality rate.’