European Journal of Population 13: 243-268,1997. 243 © 1997 Kluwer Academic Publishers. Printed in the Netherlands. Fertility and Culture in Eastern Europe: A Case Study of Riga, Latvia, 1867-1881 CHARLES WETHERELL1 and ANDREJS PLAKANS2 1Department of History, University of California, Riverside, Riverside, CA 92521, USA (E-mail: charles.wetherell@ucr.edu); 2 Department of History, Iowa State University, Ames, Iowa 5001, USA (E-mail: aplakans@iastate.edu) Wetherell, C. and Plakans, A. 1997. Fertility and Culture in Eastern Europe: A Case Study of Riga, Latvia, 1867-1881. European Journal of Population. Abstract. Recent research on the secular decline of fertility in historical Europe has focused on cultural explanations in the wake of the European Fertility Project's failure to confirm demographic transition theory. Using the city of Riga in present-day Latvia as a case study, the essay provides initial estimates of nuptiality and fertility for resident language and religious groups in 1867 and 1881, and reviews the prospects of future work. Despite obstacles, Eastern Europe offers researchers an exceptional opportunity to test major cultural and economic hypotheses about the fertility decline because sustained ethnic diversity coexisted with economic development. Charles Wetherell et Andrejs Plakans. Fécondité et culture en Europe de l'Est: une etude de cas ä Riga, Latvie, 1867-1881. Resume. Des recherches récentes sur le déclin séculaire de la fécondité en Europe se sont concentrées sur des explications culturelles devant ľéchec du Projet Européen sur la Fécondité ä confirmer la théorie de la transition démographique. A ľ aide de ľexemple de la ville de Riga, située dans ľactuelle Latvie, cet essai fournit des estimations initiales de la nuptialité et de la fécondité pour des groupes linguistiques et religieux presents en 1867 et 1881, et présente des perspectives de recherches futures. En dépit des difficultés, l'Europe de l'Est offre aux chercheurs une opportunité exceptionnelle de tester les principales hypotheses culturelles et économiques du déclin de la fécondité, du fait du maintien ďune diversité ethnique qui coexiste avec le développement économique. 1. Introduction In the past decade research on fertility in historical Europe has taken a new direction in response to the findings of the European Fertility Project (EFP). The EFP failed to substantiate demographic transition theory (Notestein, 1953), concluding that measurable economic forces associated with industrialization and urbanization were poor predictors of the timing of the secular decline in fertility in Europe from 1860 to 1960. Instead, cultural differences, most clearly observed through region, ethnicity and language, separated fertility regimes (Coale and Watkins, 1986). As new work focused more directly on cultural mechanisms, the once sharp distinction faded between what van de Kaa (1996) calls "economic narratives," which 244 CHARLES WETHERELL AND ANDREJS PLAKANS have couples adapting their fertility to changing economic conditions (Easterlin, 1978; Easterlin and Crimmins, 1985), and "social narratives," which often portray the transition as resulting from a diffusion of new ideas about controlling fertility (Knodel and van de Walle, 1979; Watkins, 1986, 1991). Although serious differences remain (Cleland and Wilson, 1987; Lesthaeghe and Surkyn, 1988; Alter, 1992; Poliak and Watkins, 1993; Greenhalgh, 1995b; Kertzer, 1995), studies advancing culturally grounded explanations of the decline of fertility in Europe now outnumber economic ones (van de Kaa, 1996, p. 396). Anthropologists, especially, helped to advance cultural explanations (Caldwell, 1976, 1978, 1982; Caldwell and Caldwell, 1987; Greenhalgh, 1990, 1995a). Hammel (1990), for example, argued that older conceptions of culture in existing research, such as that of culture as identifier (ethnicity, religion, class) or culture as context (serfdom, industrialization), do not render culture an effective explanatory variable. By contrast, newer notions that afford individuals agency to change behaviour which is specific to different ethnicities and contexts provide better ways of giving culture explanatory power. With respect to fertility control, therefore, culture might be viewed as the habits of reproduction, discussed and modified over time by individual couples in response to micro-level (the price of grain) and macro-level (urban or rural; agrarian or industrial) conditions, as well as new class-specific values surrounding family limitation, women, education, or childbearing (e.g., Woods, 1987; Garrett, 1990; Gillis, 1992; Szerter, 1996). Another feature of post-EFP research has been a critical examination of the measures the EFP employed and the development of new and more refined methods. The basic conclusion emerging from research evaluating EFP measures is that they have systematic flaws that can lead to erroneous conclusions. Guinnane et al. (1994), for example, demonstrated that the EFP's principal index of marital fertility, Ig, cannot measure well parity-dependent control nor can changes in Ig over time accurately detect the onset of control. Similar weaknesses in older measures spawned the development of new techniques and measures (Alter, 1988; David et al., 1988; David and Mroz, 1989a, 1989b; David and Sanderson, 1990; Weir, 1993, 1994; Okun, 1994a, 1994b, 1995) which utilize the higher quality individual level data that European states collected in the late nineteenth and twentieth centuries. Thus criticism of EFP methods has helped to create better measures of fertility behaviour in historical Europe. Among economic demographers, methodological criticism also led to a questioning of the EFP's major finding that economic factors were not central to the decline of fertility in Europe. Critics argue that if the EFP's estimates of the start of the fertility transition are wrong because their measures are flawed, then their conclusion that economic forces were not responsible for the decline may also be wrong (Guinnane et al, 1994). Galloway et al. (1994; cf. Lee et al., 1994), in an analysis of Prussia between 1875 and 1901, identified religion and ethnicity as key indicators of the level of fertility, but poor predictors of its decline. Instead, structural variables such as education, female labour-force participation, and the FERTILITY AND CULTURE IN EASTERN EUROPE 245 growth of reliable financial institutions proved to be better predictors of the decline. In a similar analysis of Slavonia in Croatia in the 18th and 19th centuries, Hammel (1995) came to the same conclusion. He found that only when ethnic or religious designations corresponded closely to different positions in economic and political structures were ethnicity or religion good predictors of fertility behaviour. In sum, although cultural hypotheses are gaining ground in empirical work and theoretical debates, economic demographers continue to advance structural rather than cultural change as the essential cause of the decline of fertility in Europe. While developments signal a promising future for research on historical fertility in Western Europe, the same cannot be said for Eastern Europe for two reasons. First, historical demographic research on Eastern Europe has long lagged behind that on Western Europe. Before the collapse of the Soviet Union in 1989-1991, work on Eastern European populations not produced by Soviet bloc scholars, who were wedded to Marxist paradigms, focused primarily on family history (Lastlett and Wall, 1972; Czap, 1982,1983; Wall, 1983; Hoch, 1986; Plakans and Wetherell, 1988). Western scholars outside the EFP were simply not conducting fertility research on nineteenth- and early twentieth-century Eastern European populations. Important exceptions exist (Andorka, 1971, 1979; Kruminš, 1993; Galloway et al., 1994; Lee et al, 1994; Hammel, 1993, 1995; Hammel and Wächter, 1996a, 1996b), but most are recent. Entire areas of Eastern Europe - the Baltic States, Poland, the Czech Republic, much of Slovenia, Bosnia, Croatia, Macedonia, and Yugoslavia, as well as Belarus, Ukraine, and Moldova in the western parts of the former Soviet Union - lack demographic histories. In short, there is no body of historical demographic research on Eastern Europe that is remotely comparable to that on Western Europe. The second reason stems from the first. Because non-EFP researchers did not pursue the historical demography of Eastern Europe, existing sources of data that might be used to study of fertility over time in more than a single region before the twentieth century were not systematically exploited. Scattered community studies exist, but their findings are plagued by questions of representativeness (Czap, 1978, 1982, 1983; Hoch, 1982, 1986; Andorka and Balazs-Kovacs, 1986; Kuklo, 1990; Plakans and Wetherell, 1988, 1992, 1995; Wetherell, Plakans and Wellman, 1994). Galloway's Prussian (Galloway et al., 1994; Lee et al., 1994) and Hammeľs Slavonia data (Hammel, 1993, 1995; Hammel and Wächter, 1996a, 1996b) are important exceptions to the generalization that small, particularized data sources underlie Eastern European historical demographic research. Yet if Watkins' (1991) vision of demographic diversity across Western and Eastern Europe in the nineteenth and early twentieth centuries is correct, then the utility of community based studies for understanding broad fertility regimes in Eastern Europe remains questionable. A related liability in Eastern European demographic research involves the necessity of using older measures on existing data. Because newly developed measures of fertility invariably require individual level data, and because Eastern Europe 246 CHARLES WETHERELL AND ANDREJS PLAKANS suffers from a lack of systematically developed individual level data, historical demographers are currently forced to use older measures that do not meet current standards of acceptable explanatory power. In sum, historical demographers of Eastern Europe face major obstacles. Not only do they need to catch up substantively, but they also need to develop new sources of data that can answer critical demographic questions and deflect methodological criticisms. Paradoxically, however, historic Eastern Europe may well prove to be an ideal laboratory for exploring competing explanations of the fertility decline because its inhabitants remained ethnically and linguistically diverse while they modernized economically and demographically. The three Russian Baltic provinces of Livland (Livonia), Kurland (Courland), and Estland (Estonia), for example (Figure 1), variously part of Swedish, Polish and Russian empires from the sixteenth to the twentieth centuries, contained nine identifiably distinct ethnic nations - Latvians, Estonians, Lithuanians, Germans, Russians, Jews, Belorussians, Ukrainians, and Livs - that maintained their own traditions and languages well into the twentieth century. Critical political and social transformations also occurred in the nineteenth-century Baltic that fostered immigration, urbanization and industrialization. The abolition of serfdom between 1816 and 1819, the removal of restrictions on geographical movement in the 1830s, and the introduction of private ownership of land by peasants in the 1850s prompted heavy migration to Baltic cities. Between 1806 and 1853, the population of Riga, the metropolitan centre of the nineteenth-century Russian Baltic provinces (Henrikkson, 1986), grew at a modest annual rate of 1.41 percent, from 33,665 to 65,413. In 1867, the city's population had expanded to 102,590, reflecting an annual rate of growth of 3.2 percent. And by 1881, Riga's population had increased to 169,320, at an extremely high rate of growth of 3.5 percent per year (Krastinš, 1978, p. 110). The EFP also identified the guberniias of Estland, Livland, and Kurland among the handful of Russian provinces where the decline in fertility began earlier than in the Empire as a whole: Kurland in 1860-1870, and Estland and Livland both in 1880-1890 (Coale, Anderson and Härm, 1979, p. 157; Coale and Treadway, 1986, pp. 39-40; Coale and Watkins, 1986, Map 1). Thus despite methodological limitations, the Baltic States are a good place to begin to explore the cultural and structural components of fertility in Eastern Europe at the micro-level. We have chosen to use Riga as a case study of both the problems and prospects of such an enterprise. Our preliminary exploration of fertility and culture in the Riga does three things. First, it serves as an introduction to some of the problematic evidence that historical demographers of Eastern Europe will need to confront and exploit. It reveals the weaknesses of existing data but also the analytic possibilities that better data might afford. Second, it provides basic measures of nuptiality and fertility among resident national and religious groups in Riga in 1867 and 1881 that can serve as initial points of reference for future work. As one might expect, fertility differed among national and religious groups, but not always in predictable FERTILITY AND CULTURE IN EASTERN EUROPE 247 BALTIC SEA ST. PETERSBURG Figure 1. The Russian Baltic Provinces about 1850. Source: Plakans, 1995, p. 62. Used with permission of the Hoover Institution Press. Note: Place names are from the mid-nineteenth century. Modern equivalents: Estland + Northern Livland = Estonia; Southern Livland + Courland + Western Vitebsk = Latvia; Kovno = Lithuania; Ösel = Saaremaa; Reval = Tallinn; Pernau = Pärnu; Dorpat = Tartu; Wolmar = Valmiera; Wenden = Cěsis; Dünaburg = Daugavpils; Mitau = Jelgava; Goldingen = Kuldiga; Libau = Liepaj a. ways. Nuptiality conformed to a Western European marriage pattern rather than an Eastern European one (Hajnal, 1965, 1982), confirming work which argues that the dynamics governing nuptiality were different in urban areas (Lynch, 1991) but also raising questions about the analytical utility of Hajnal's line between eastern and western European populations. Third and finally, it explores the structural context of fertility and nuptiality in the Baltic context. In the process, we reflect, from the viewpoint of historians interested in a particular area, on the problems that the exercise as a whole presents in terms of sources and methodologies. Our larger purpose is to raise issues and explore evidence that may ultimately lead to a fuller understanding of the fertility transition in Eastern Europe, rather than 248 CHARLES WETHERELL AND ANDREJS PLAKANS to offer any definitive interpretation. The state of historical demographic research on Eastern Europe and the available data do not currently allow more. However, further study of fertility in the nineteenth-century Baltic area may help to refine our understanding of the cultural and structural determinants of fertility and the fertility transition in Eastern Europe. 2. Data and measures 2.1. Data The data we employ are from two censuses taken in 1867 and 1881 (von Jung Stilling, 1869, 1883-85). Local and provincial administration in the three Russian Baltic provinces of Livland, Kurland, and Estland was entirely in the hands of the three Baltic noble corporations {Ritterschaften), each of which had organized a bureau to gather statistical information on the provincial population and economy. These committees became very active in the middle decades of the century, so that the Baltic provinces tended to be better documented statistically than other regions of the Russian Empire. The Livland Statistical Committee conducted the 1867 census only for nine Livland cities (including Riga), and not the countryside. In the morning hours of March 3, 1867, enumerators visited all inhabited places in Riga and, following elaborate instructions, took information by household {Haushaltung) on each of the members: name, sex, age, relation to head, marriage status, religion {Confession), utilized language, social estate {Stand), occupation, place of current residence, and place of registered residence {Hingehorigkeit), thus combining individual and household level information. In the 1869 published summary, von Jung Stilling, the director of the Livland Committee, explained that language was being used as an indicator {Kennzeichen) of nationality, with the result that all the published cross-tabulations were grouped separately in terms of the German, Russian, Latvian, Estonian, Jewish, and "other" language/nationality designations (von Jung Stilling, 1869, p. vi). Most published tables were three-way (e.g., Russian population by age, sex, and, marital status), with identical table-clusters for each of the four main districts of the city, and for each borough of each district. This highly structured information, of course, limited the number of cross-tabulations that could reasonably be published and so prohibits any analysis that might confirm or deny residential clustering among resident language groups within the city. By contrast, the 1881 census was organized by all three Statistical Committees, and carried out in the cities and rural districts on December 29. The published results of the census in each province were tabulated somewhat differently by each Committee and published separately over a five-year period. This time, enumerators filled out a form for each individual, asking for name, sex, marital status, nationality, religion, used language, literacy, main occupation and ancillary occupations, place of current residence, place of registered residence, and disability. Information from the individual forms was then grouped into household forms. As in 1867, FERTILITY AND CULTURE IN EASTERN EUROPE 249 the 1881 published results for Riga (von Jung Stilling, 1883-85) used the city's districts and boroughs for grouping purposes. Unfortunately "nationality" was used infrequently. Therefore, while we can estimate nuptiality and fertility for language groups in 1867, we are limited to religious groups in 1881. From the explanatory material and classifications in both censuses one also detects an unease with the concept of "nationality," in part because the Baltic German elite argued in the political polemics of the day that the Baltic area was a Ständestaat (society of social orders), rather than a. Nationalstaat (society of national groups). However, information on "social estate," which might have provided a way to analyze any breakdown in the traditional "social order," was not collected. The motivation for the 1867 census was not reported in the published volumes, although it was the logical next step to the work of the Committee, which had carried out a series of trial censuses (Probezählungen) in various localities beginning in 1860. No doubt the sheer growth of Riga's population played a role as well, as the native born were losing ground to immigrants by 1867. The 1881 enumeration may have been carried out in preparation for the large-scale Imperial inspection of the three provinces, the so-called "Manasein Revision," in 1882-1883. That Imperial revision turned out to be the first step of a policy of "russification" aimed toward the Baltic provinces from 1885 to 1914 (Haltzel, 1981; Plakans, 1981; for contemporary discussions of problems surrounding the 1881 census, see Stieda (18 81) and Wittschewsky (18 81)). Aside from the Baltic censuses, to our knowledge there were no other major censuses taken in the Empire prior to 1881; if there were others, they were provincial or local affairs. The first coordinated Imperial census was taken in 1897 and it is with this census that EFP researchers began to document fertility in the Russian empire and later Soviet Union (Coale et al, 1979; Coale and Watkins, 1986). In the case of the Baltic, the EFP extrapolated backwards in time from 1897 using reverse projection, and did not report using either census, despite the 1881 coverage of both rural and urban populations. Although ours is the first attempt to exploit these censuses, major limitations exist. For one, the information in the published versions of the 1867 and 1881 censuses is only partially comparable. For another, all data are aggregated. We have seen some of the 1881 individual household schedules for agricultural estates in the Latvian National Archives in Riga, but we have been unable to determine if all the individual schedules have survived. They are certainly not collected into an identifiable record group or special archive. The aggregate nature of the published data limits the demographic measures we can use and prohibits any multi-variate analyses, but they do allow some preliminary estimates. 2.2. Measures The measures we use are mainly those that Coale developed for the EFP: the index of general fertility, If, the index of marital fertility, Ig, the index of illegitimate 250 CHARLES WETHERELL AND ANDREJS PLAKANS fertility, Iu, and the proportion married, Im (Coale, 1967, p. 209; Coale, 1969, pp. 4-6; Coale and Treadway, 1986, pp. 153-162). Each is interpreted as relative to the Hutterites, an Anabaptist sect that settled in the upper, central U.S. and lower central Canada in the late nineteenth-century, and which has a reliable record of socially unrestrained fertility (Eaton and Mayer, 1954, pp. 1-2). This has the value of making the measures essentially parametric and practically limited at 1.0. An Ig of 0.65 thus indicates that the population under study has a level of marital fertility that is 65 percent of Hutterite marital fertility. Although Ig has no inherent capability to indicate the onset of fertility control (Guinnane et al, 1994, p. 13), different values of Ig nonetheless indicate different levels of fertility; they also have been widely used and so can be compared to a wide variety of other historical populations.1 Measures of nuptiality include the Singulate Mean Age at Marriage (SMAM) and the median age at marriage (Hajnal, 1953; Shryock and Siegel et al, 1975,1, pp. 292-293). The SMAM is "analogous to an expectation of life as a single person" (Haines, 1996, p. 19) but implicitly assumes a closed, non-migrating population (Hajnal, 1953, pp. 130, 132). By contrast, the median age at marriage does not assume that the population under study is closed and thus is arguably a better measure of age at marriage since Riga experienced heavy immigration in the last half of the nineteenth century. 2.3. Adjustments The calculation of Coale's indices requires data on age, marital status, and births, as well as some estimate of infant and adult female mortality. Both the 1867 and 1881 censuses list the population by sex, marital status, and single year of age; for persons under one, they list ages by month. However, neither census contains any data on deaths, which constitutes a serious problem because Ig is sensitive to even modest fluctuations in the number of births. Rashin's (1956) classic statistical summary of the Russian empire provides data on deaths that can be used to estimate mortality. Coale et al. (1979, pp. 207-208) reported using Rashin's estimates of crude birth and death rates in 1897 to select a model life table and a corresponding estimate infant mortality, \qo (the probability of dying between birth and exact age one), but reported using neither the enumeration of infants under one in the 1881 census nor Rashin's estimates of deaths prior to 1897. Rashin's estimates for Livland between 1866 and 1885 (Rashin, 1956, Table 123, p. 67) indicate an "East" model life table for females. The particular model life table levels (7 and 10) indicate Iqos of 0.259 for females and 0.306 formales in 1867, and 0.198 for females and 0.235 for males in 1881. The change in levels represents noticeably improving mortality, but life expectancies at birth (eo) were still below 45 (Coale and Demeny, 1983, 273-274). While our estimates of infant mortality may be an improvement over the EFP's, they are not without problems. For one thing, Woods (1993) recently showed that FERTILITY AND CULTURE IN EASTERN EUROPE 251 Table I. Urban multipliers for infant mortality in Riga, 1767-1881, and selected parts of Western Europe, 1799-1902 Locale and period Rural Urban multiplier Urban qo qo Bavaria, 1867-1885l'2 0.425 1.03 0.438 Finland, 1901-19021'2 0.151 1.25 0.189 Prussia, 1901''2 0.226 1.18 0.267 Germany, 1709-18991'2 0.222 - Sweden, 18702'3 0.141 1.53 0.216 Sweden/Stockholm4 1861-1870 0.136 1.97 0.268 1871-1880 0.132 2.08 0.275 1881-1890 0.113 1.17 0.194 Latvia/Riga5 1867 0.283 0.352 1881 0.217 1.25 0.270 Sources: (1) van de Walle, 1986, p. 209, Table 4.3, p. 217. (2) Reher, 1995, Table 1, p. 523; (3) Lynch and Greenhouse, 1994, p. 121; (4) Woods, 1993, Appendix Table 2, p. 219. (5) Latvian qos represent the averages of the estimates for males and females at East, Level 7 (1867) and Level 10 (1881); Coale andDemeny (1983, pp. 273-274) the expected relationship between infant and adult mortality inherent in model life tables has not always applied historically. In nineteenth-century Stockholm, for example, adult and infant mortality varied independently, suggesting a pattern that occurs "when both are at very high levels, but especially infant mortality" (Woods, 1993, p. 202). Woods also showed how improvements in infant mortality might outpace those in adult mortality. In short, Woods' findings suggest extreme caution in estimating infant mortality using life tables as we have done, especially during a period of rapid change in mortality generally (Schofield and Reher, 1991). Yet without additional data on deaths that would allow us to construct independent estimates of mortality, we have no choice but to use estimates derived from model life tables and to urge interpretive caution. An additional problem exists. Van de Walle (1986) demonstrated that infant mortality was higher in cities than in rural areas. Since our the estimates are for the entire province of Livland, they reflect a largely rural experience. We thus need to adjust infant mortality in Riga upward using an urban multiplier. Picking a multiplier, however, is risky business, largely for the reasons Woods outlined. Table I reveals the few instances in historic Europe where estimates of qo can be matched to an urban multiplier. Collectively the multipliers reveal no clear pattern that would lead us to choose one over another for Riga. The values of \qo for both sexes in Livland, 0.283 in 1867 and 0.217 in 1881, are closest to those for Prussia in 1901 (0.226) and Germany 252 CHARLES WETHERELL AND ANDREJS PLAKANS from 1709-1899 (0.222), and suggest a multiplier between 1.20-1.25. The figures derived from Sweden and Stockholm between 1860 and 1890 would favour a much higher multiplier on the order of 1.75-2.0. The rapid growth of Riga and the absence of a modern public health regime in the city might also favour choosing a high urban multiplier; although the presence of a military garrison would also suggest some public health measures were in place (Kunitz, 1986). The Stockholm example is one of extremely high infant mortality. By contrast, the other urban multiplier for Sweden in 1870 (1.53) comprehends other urban areas. The population of Riga was also predominantly German, Latvian, and Russian, which would favour the northeast European multipliers. Applying the high Sweden/Stockholm multipliers, moreover, pushes/^ above 1.0, which is simply not credible. Accordingly, we have chosen to use the simple average of the urban multipliers for Bavaria, Prussia, Finland and Sweden (van de Walle, 1996, p. 217), and have adjusted our estimates of the essentially rural \qo in Livland by 1.25 to arrive at qos for both sexes in Riga of 0.352 in 1867, and 0.270 in 1881. 3. Fertility and nuptiality Table II presents basic estimates of fertility for Riga in 1867 and 1881, and for selected Russian and Baltic provinces in 1870 and 1897. Two matters warrant comment. First, while our estimates are higher than the EFP's because we used different estimates of mortality, they are consistent with a decline in fertility after 1880. From ahigh of 0.692 in 1867,/,, fell to 0.623 in 1881, and to 0.519 in urban Livland in 1897. Moreover, we believe that our estimates are better than the EFP's because they are made on the basis of actual enumerations rather than projected populations, and they embody corrections for mortality for the periods surrounding the enumerations. Our measures of fertility may change when better estimates of mortality become available, but we believe they are the best we can currently make. Second, nuptiality in Riga, as indicated by the SMAMS, median ages at marriage, and proportions single aged 45-54 displayed in Table III, reflects Hajnaľs (1965) western European marriage pattern of relatively late ages of marriage for both men (late-20s) and women (mid-20s), and substantial proportions of both sexes remaining single (13-16 percent). Minor differences existed among Riga's various language and religious groups, but the commonly understood Eastern European marriage pattern of early and universal marriage did not prevail in post-1850 Riga. This finding underlines the problems in applying Hajnaľs dichotomous model to the wide variety of demographic behaviour in the European east as well as to the complexities of urban nuptiality. Lynch (1991) recently argued that the evidence of Hajnaľs western European marriage in cities could be explained by economic constraints and values that operated in different ways for different groups. Instead of lowering the age of marriage among urban immigrants because young couples could secure independent incomes in industrial occupations as Hajnal reasoned, constraints imposed FERTILITY AND CULTURE IN EASTERN EUROPE 253 Table II. Estimates of fertility in Riga, 1867 and 1881, and selected Russian cities andLivland, 1870 and 1897 Locale Year If h 4 írn Riga 1867 0.361 0.692 0.064 0.473 1881 0.338 0.623 0.073 0.482 Moscow 1870 0.531 0.781 0.150 0.604 1897 0.439 0.645 0.125 0.604 Kiev 1870 0.559 0.787 0.063 0.685 1897 0.543 0.752 0.060 0.685 St. Petersburg 1870 0.393 0.659 0.119 0.507 1897 0.330 0.553 0.101 0.507 Livland 1870 0.335 0.670 0.043 0.466 Rural 1897 0.305 0.605 - 0.474 Urban 1897 0.256 0.519 - 0.452 Sources: Coale, Anderson and, Härm 1979, Table 2.2, p. 20; Coale and Treadway, 1986, Appendix A, pp. 137-138; von Jung Stilling (ed) 1869, Table 64, n.p. (Between 292 and 299); von Jung Stilling and Anders, (eds) 1885-1887, Vol. 1, Table 11, pp. 23-26. Table III. Estimates of nuptiality in Riga, Latvia, 1867 and 1881 1867 1881 Males Females Males Females Proportion, 44-54, single 0.162 0.160 0.137 0.131 SMAM 30.4 25.6 29.8 25.0 Median age at marriage 28.7 23.2 28.1 24.2 Difference (SMAM - median age) 1.7 2.4 1.7 0.8 Total, 15^19 31,570 27,858 45,196 45,291 Total population 52,047 50,543 85,821 83,499 102,590 169,320 Sources: von Jung Stilling (ed) 1869, Table 64, n.p. (Between 292 and 299); von Jung Stilling and Anders (eds) 1885-1887, Vol. 1, Table 11, pp. 23-26. by the high cost of housing or the requirements of education and training may actually have raised the age of marriage for many. The net result, in Lynch's view, is that a heterogeneity of experience prevailed and that simple accounts of the age of marriage in urban Europe are dangerous. Geographically, Riga sits above the Hajnal line that runs from St. Petersburg to Trieste. Although inhabited by peoples from both Eastern and Western Europe, Riga and the Baltic states have long been viewed as part of Eastern Europe. Certainly as part of the Russian empire, Riga would be considered part of Eastern Europe, but political divisions do not always 254 CHARLES WETHERELL AND ANDREJS PLAKANS translate into identifiable behavioural patterns. In parts of rural Latvia, the Western European marriage pattern definitely prevailed (Plakans and Wetherell, 1988). We have dealt with the issue of what geographic areas might be construed as Eastern Europe elsewhere (Plakans and Wetherell, 1997), and have argued that distinguishing Eastern from Western European demographic patterns in border areas such as the Baltic is an extremely difficult, if not a currently intractable, problem. 4. Nationality, language, and religion In order to comprehend the cultural divisions that might have existed in Riga and elsewhere in Eastern Europe, it is helpful to understand both the profound social changes that occurred in the Baltic in the nineteenth century and the attending rise of national sentiments. Taken together, they arguably worked to sustain differences in language and religion that may have helped to sustain national and ethnic sensibilities and behaviours at a time when urbanization and industrialization was breaking down such barriers elsewhere. In short, cultural identifiers such as language and ethnicity may have real analytic value in Eastern Europe as Watkins (1991) contends. Imperial internal passport reforms of the 1850s loosened restrictions on movement within the Russian Empire and produced migration to urban areas in the 1860s and 1870s. The existence of the two urban censuses that we are using is itself a symptom of the disquiet which the Baltic German provincial elites experienced with the advent of population redistributions in the Baltic area during the second half of the nineteenth century. Rural people were on the move, urban industry was expanding rapidly, the nature of the labour force was changing, and all of these aspects of "modernization" called for thorough population inventories so that effective control could be maintained. Moreover, these same decades marked significant upturns in the creation of new enterprise in the Empire generally and in Riga particularly, increasing in the long run the proportion of Riga's residents who worked in factories and craft enterprises by about 141 percent (1861-1897) (Krastinš, 1978, Table 4, p. 22). Finally, these decades also witnessed the sharpening of national consciousness in the Baltic area, as Latvian and Estonian nationalists successfully popularized the idea that language groups were "nationalities" and thereby forced German-speakers to articulate a group ideology of their own. Strong Slavophile sentiments in Imperial government circles also made the "nationalities" of the Baltic area increasingly aware of being governed by another nationality - the Russians - rather than simply Imperial administrators who spoke a different language (Plakans, 1995). In the third quarter of the nineteenth century, the Riga population grew principally through in-migration. In 1867, 64.2 percent of the population had been born there, while in 1881 that share had declined to 39.9 percent. Where these immigrants came from is easier to determine than who they were. In 1867, the main source of immigrants (34.1 percent of all immigrants) had been Livland province. FERTILITY AND CULTURE IN EASTERN EUROPE 255 The next largest source had been the province of Kurland (27.0 percent), adjoining Livland to the southwest; the remainder had come from other Russian provinces (21.6 percent) or elsewhere in Europe (16.3 percent). By 1881, the trajectory of migration had changed. The dominant supplier now was Kurland (31.5 percent) which lay south of Riga, followed by Livland itself (27.6 percent) and other Russian provinces, such as Kovno (12.4), which lay directly south of Kurland (Figure 1). Although Riga's officials did not keep records on individual immigrants, the language classifications in 1867 and 1881 help to answer the question of who the immigrants were. In 1867 the dominant language in Riga was German, which was spoken by 42.9 percent of all residents. Russian was the second most commonly spoken language (25.1 percent), followed by Latvian (23.6 percent), and Yiddish (5.1 percent) a distant fourth. In 1881, these proportions had changed at the expense of German and Russian. German-speakers still dominated, but their share had fallen to 39.4 percent of the population. The second largest language group now were Latvian-speakers (29.5 percent), with Russian-speakers assuming third place (18.9 percent). Because Riga's total population had also increased, all language groups contained larger absolute numbers, but proportional gains are instructive. While the increase in the number of German-speakers from 1867 to 1881 stood at 51.8 percent, Latvian speakers increased 106.5 percent and Yiddish-speakers 170.7 percent. The increase in Latvian-speakers in part reflects the expectation among the rural landless that Riga would provide them with employment. The increase for Yiddish-speakers fits with the noted southerly-to-northerly shift in the trajectory of migration: the Pale of Settlement ended at the Kovno-Kurland border, and Riga lay only 75 kilometers to the north. The steady in-flow of both Latvian and Yiddish speakers suggests that forces for retaining ethnic ways were strong as new immigrants reinforced rural habits of mind and behaviour- a common feature among U.S. immigrants during the late nineteenth and twentieth centuries (e.g., Zunz, 1982; Bodnar, 1985; Morawska, 1985,1996; Massey, 1995; cf Economakis, 1997). At the same time, the overriding and analytically troublesome fact is that the populations of the Baltic provinces in the second half of the nineteenth century are impossible to separate cleanly into groups defined by these designations. Baltic administrative boundaries corresponded very badly with the boundaries of putative national, linguistic, or religious "communities." Estonians lived in Estland and adjoining Livland in almost equal numbers; Latvians lived in Livland and Kurland and perhaps 20 percent could be found in the western districts of adjoining Vitebsk province. Baltic Germans, Russians, and Jews were scattered throughout rural and urban areas of the three provinces, though there were fewer Jews in Estland than elsewhere in the Baltic. Most Baltic Germans, Estonians, and Latvians were Lutheran Protestants, yet all three groups contained important sub-populations of Roman Catholics and persons of the Orthodox faith. Moreover, minorities of each nationality group identified the language of some other nationality group as the language of the their own homes. Although only 1.1 percent of Germans declared 256 CHARLES WETHERELL AND ANDREJS PLAKANS their language to be something other than German, the propensity was higher among other nationalities: 9.0 percent of Russians, 12.0 percent of Latvians, and 28.8 percent of all Jews. The EFP discouraged explanations of fertility that relied only on nationality by showing that the factors depressing fertility in any given area, including the Baltic provinces, were reflected in similar levels for all nationality groups in one area in comparison to groups of like nationality in other areas. In other words, "whatever it was that caused Latvians to have low fertility when there were few non-Latvians around also led non-Latvians to have low fertility in the presence of Latvians" (Anderson, 1986, p. 309). The phrase "whatever it was" invites the use of cultural propensities that affected all groups. At the same time, and especially in the Imperial borderlands in the nineteenth century, the assumption that language-group membership can be defended, as the correlation (r = 0.76) between language and nationality in 1881 suggests. One thing is certain. National identity was in flux between 1867 and 1881, and enumeration techniques could not have been flexible enough to catch the processes of change in individual self-identification muddying the area's cultural-political history. The same persons, mostly male household heads, who provided evidence to enumerators about their own and their families' group membership made up the age groups we use to examine fertility, and they are also the same people who were most vulnerable to changes in national identity. While estimates of fertility do not require subjective interpretations, national identity is subject even to short-term change. To use language as an indicator of nationality, and to assume, in this period, continuity in the latter may be misleading. This observation would not hold true as much for the Jews (considered as a nationality) as for the German- and Latvian-speakers. In sum, nationality is fraught with uncertainties in this transitional period. To make matters worse, the boundaries that Watkins (1986, 1991) has argued separated fertility regimes in Western Europe were bound to prove more permeable in a cosmopolitan city such as Riga. Bi- and even tri-lingualism (commonly German, Russian, and one other language) had long been an element in the region's history, but assessing that specific influence on fertility behavior remains impossible given the current state of our knowledge. 5. Language, religion, and fertility Nonetheless, as Table IV reveals, there were differences in the fertility of the major language groups in Riga in 1867, although no clear pattern. Marital fertility was highest among Yiddish speakers (Ig = 0.771) and lowest among Latvians (Ig = 0.656). Estimates of nuptiality further suggest a divide between Yiddish speakers (Jews) and other language/nationality groups. Following Lynch (1991), one might expect male German speakers, who held the majority of jobs in the municipal bureaucracy and professions, to have a high age of marriage as the requirements of training and education would work to delay marriage. The same might be true of FERTILITY AND CULTURE IN EASTERN EUROPE 257 Latvian-speakers who dominated industrial and factory occupations, but for whom the high cost of housing in Riga worked to delay marriage. By contrast, Yiddish speakers, who were concentrated in craft industries and whose female labour force participation was less than 7 percent, might tend to have lower ages of marriage. The extent of inter-faith marriage in Riga suggests the social relevance of religion for the study of fertility. Between 1881 and 1885, the overall incidence of mixed marriages {Mischehen) was only slightly more than one in ten (cf. McCaa, 1993). Of 6,952 unions contracted, only 12.5 percent (876) were between partners of different faiths (von Jung Stilling, ed. 1887, pp. 207, 218-221,224). The degree to which persons of different religions decided to marry, however, differed according to sex and opportunity. Protestants, who were primarily German- and Latvian-speakers, were not the most likely to enter such unions, as we might expect on religious grounds: of all Protestant men married during the period only 5.9 percent married women of a different faith, and of the Protestant women only 9.9 percent did so. By contrast, 42.4 percent of Orthodox men married partners of a different faith along with 25.4 percent of Orthodox women. Of the main confessional groups, the highest incidence of interfaith marriage was among Protestants and Jews. If we interpret these results as reflecting preferences, we get support for group identity. Nonetheless, Protestants comprised 62 percent of the population and enjoyed a reasonably balanced sex ratio (98.2). By contrast, Orthodox and Catholic sex ratios were extremely unbalanced, 170.6 and 129.6 respectively, indicating that opportunity in the marriage market governed perhaps more than preferences. Fertility varied across religion as we might expect. Among the major religious groups in Table 5, Catholics had the lowest marital fertility (Ig = 0.493), Jews the highest (Ig = 0.778), and Protestants about halfway in between (Ig = 0.638). Logically, Orthodox and Raskolniks ('Old Believers') would tend to be Russian, thus untangling national propensities from religion is even more problematic. No clear explanation for the difference exists in the age structures or the ages of marriage of the different groups, with the exception of the Jews. Jewish women married roughly two years earlier than Protestant, Catholic or Greek Orthodox women, but that difference could not axiomatically account for their higher marital fertility. Raskolnik women married at about the same age as Jewish women, yet their marital fertility was considerably lower. Differences in the age structure of Jews and other religious groups stem from the presence of greater proportions of children, which suggests similar or even higher Jewish marital fertility in the previous two decades.2 Although religion differentiates between levels of marital fertility in Riga, the question remains of whether those differences are large enough to highlight religion as a cultural feature that differentiates childbearing behaviour among groups. The only group whose indices are consistently different are the Jews. They have substantially higher fertility and proportions married, and lower illegitimacy, proportions 45-54 single, and female labour force participation. Yet we simply do not know whether the differences among the other religious groups are substantial enough to 00 Table IV. Estimates of fertility and nuptiality by major language group, Riga, 1867 German Russian Latvian Yiddish Fertility If 0.342 0.374 0.383 0.530 h 0.721 0.715 0.656 0.771 In 0.056 0.067 0.076 0.034 Irn 0.430 0.474 0.656 0.673 M F M F M F M F Nuptiality SMAM 30.6 25.3 29.8 25.2 31.3 26.7 25.1 21.5 Median age at: marriage 29.3 23.3 26.4 23.5 30.0 24.2 24.1 19.7 Difference (SMAM -median age) 1.3 2.0 3.4 1.7 1.3 2.5 1.0 1.8 Proportion, 45- -54, í angle 0.144 0.203 0.221 0.169 0.132 0.092 0.030 0.016 Sample sizes Ages 15^19 11,510 12,783 9,280 6,315 7,553 6,661 1,533 1,296 Births 847 868 450 453 467 521 133 143 Total population 20,682 23,308 14,420 11,352 12,174 12,025 2,769 2,465 o > r M in M H ffi M fö M r Sources: von Jung Stilling (ed) 1969, Vol. 1, Tables 69, 74,78, 84, 89, 94,100, n.p. (between 292 and 299). Notes: SMAM indicates £ Singulate Mean Age at Marriage (Hajnal, 1953). Births are adjusted for mortality; see text. "Z, $ Ö fö M ^H IT! "tí r > r. FERTILITY AND CULTURE IN EASTERN EUROPE 259 ■B ~o ■a pi O y. o O B Ph m co ^i in ^r t~~ (N m ^r t~~ O m 00 00 m oo on o o m vo o ^r t-~ t-; O o in ^O ^ 00 on ^r oo m oo vo o o o ^ ^H t-~ ^h" —i m íN in on m m ^ ^ on m t~~ ON in ON vo o O t-~ oo„ m °\ vo" o" ON