The Decline of Remarriage: Evidence from German Village Populations in the Eighteenth and Nineteenth Centuries*

THE DECLINE OF REMARRIAGE: EVIDENCE FROM GERMAN VILLAGE POPULATIONS IN THE EIGHTEENTH AND NINETEENTH CENTURIES*

John Knodel and Katherine A. Lynch**

Abstract: Family reconstitution data for fourteen German village populations permit the examination of remarriage during the eighteenth and nineteenth centuries. The results provide compelling evidence for a secular decline in the tendency to remarry. Pronounced age and sex differentials in the likelihood of remarriage were evident: widows were far less likely to remarry than widowers, and the probability of remarriage declined rapidly with age, particularly for women. The probability of remarriage was also inversely associated with the number and age of children. There were, however, no clear differences in either the probability of remarriage or its tendency to decline over time among major occupational groups. The decline in remarriage probabilities was caused in part by declines in adult mortality, which gradually raised the ages of surviving spouses to levels at which remarriage has historically been rather unlikely. However, age-specific marriage probabilities also declined, affecting both men and women and all occupational groups, suggesting the presence of a social change of wide scope. Some comments on possible factors contributing to the decline of remarriage are presented. The need for a comprehensive explanation of remarriage trends and differentials remains an important challenge for family historians.

I. Introduction has been underscored in a number of important and influential studies (Hajnal, The importance of a distinctive marriage 1965, 1982; Laslett, 1977; Wrigley, pattern in preindustrial northwest Europe 1981). Most investigations of nuptiality in western Europe during the period cover- ing the seventeenth through the nineteenth centuries have focused on the rela- *An earlier version of this paper was presented late into first for at the Eighth Annual Meeting of the Social Science tively entry marriage History Association, October 27-30, 1983, both sexes and the prevailing, substantial Washington, D.C. levels of celibacy defined as proportions **John Knodel is Professor of Sociology at the never married. However, far less atten- University of Michigan and Research Associate at tion has been paid to patterns of marital the Population Studies Center there (Ann Arbor, dissolution and that were Michigan 48104-2590). His principal interests are remarriage historical demography and fertility transition in characteristic of prevailing socioeconomic developing countries. Katherine A. Lynch is Assis- and demographic systems despite their tant Professor in the History Department at potential significance for the individuals Carnegie-Mellon University (Pittsburgh, 15213). and societies involved (Baulant, 1976; Her research interests include the historical demography, and family history of modern Europe, Smith, 1979; Bideau, 1980). on in and the study of relationships among industrializa- Recent scholarship remarriage tion, demographic change, and social policy. Europe based on data on the composition 35

of marriages by types has revealed a sub- individual probabilities of remarriage also stantial decline in the proportion of mar- declined. If widows and widowers consti- riages involving people remarrying, and a tuted a smaller proportion of European concomitant growth in the proportion of populations in the nineteenth century primary marriages (those between never- compared to those of the sixteenth or married men and women). In their seventeenth centuries, then the propor- analysis of English trends, for example, tional importance of remarriage could Schofield and Wrigley (1981:212) argue decline while leaving individual probabili- that while 25-30 percent of those marry- ties of remarriage unaffected or even in- ing in the sixteenth century were remarry- creasing. ing, this proportion had declined to only The present paper examines the inci- 10 percent by the nineteenth century. dence of remarriage as well as probabili- Bellettini (1981:259-260) cites late nine- ties of remarriage based on the experience teenth-century European figures, showing of fourteen German village populations that there were substantial national dif- during the eighteenth and nineteenth cen- ferences in the relative importance of turies, studied with the use of family remarriage. However, he argues that the reconstitution data. The results provide number of remarriages has undergone compelling evidence for a secular decline &dquo;progressive diminution&dquo; since the seven- in the tendency to remarry. The data also teenth century (1981:259). Data from show that probabilities of remarriage local family reconstitution studies in were inversely associated with age and France, including those by Cabourdin number of children and that women were (1978, 1981) and Bideau (1980:29) con- far less likely to remarry than men. There firm the existence of a trend towards the were, however, no clear differences in decreasing, proportional importance of either the probability of remarriage, or its remarriages from the sixteenth to the tendency to decline over time among ma- nineteenth centuries. jor occupational groups. Explaining the Although a wide body of current causes of this decline and the presence or research points to a decline in the propor- absence of major differences in probabili- tional importance of remarriages in Euro- ties of remarriage is difficult, given the pean history, there has been considerably wide variety of socioeconomic and cultur- less work on actual probabilities and rates al causes that may have been at work. of remarriage (cf. Livi-Bacci, 1981; Dyr- II. Sources and vik, 1981). In their history of the Euro- Setting pean family, Mitterauer and Sieder (1982: The basic data for the present study are in 150) appear to assume that the probability the form of reconstituted family histories of remarriage underwent a &dquo;continuous&dquo; as available in village genealogies (Orts- decline over the last two centuries, al- sippenbücher), a source unique to Ger- though they provide no supporting evi- many. A total of fourteen such genealo- dence. Thus, most evidence to date on gies has been selected for study. The trends in remarriage over the last few cen- nature and quality of these genealogies, as turies has dealt with the proportionate well as the historical background of the share of marriages that have been remar- villages selected, have been discussed at riages. length elsewhere (Knodel, 1975; Knodel The finding that remarriages became a and Shorter, 1976; Knodel, forthcoming). progressively smaller proportion of total The fourteen villages are located in five marriages does not necessarily imply that different areas of Germany: four in 36

Baden, one in Wurttemberg, three in detailed discussion of the criteria used is Bavaria, four in the former principality of provided elsewhere (Knodel, forthcom- Waldeck, and two in East Friesland. The ing). In addition, a somewhat reduced four villages in Baden and the three in coding scheme was applied to couples in Bavaria were Catholic while the other these villages. Of particular relevance for villages were Protestant. All were rather the present study is the fact that for those small in population size. They represent a villages for which the shorter coding considerable range of demographic condi- scheme was used, the event of remarriage tions as indicated by intervillage differ- was coded for couples but the exact date ences in child mortality, marital fertility, of remarriage was not. Thus, for analyses and age at first marriage and also show requiring knowledge of the date of remar- some diversity with respect to occupation- riage, we rely on the subset of six fully al distributions and inheritance systems. coded villages rather than the total sample Thus, while they cannot be considered of all fourteen villages. When the results either a random or representative sample are based on the complete sample of four- of rural Germany in any rigorous sense, teen villages, the same set of selection they at least span a reasonable range of criteria is applied to couples in all of the regional, demographic, and socioeconom- villages including the six fully coded ones. ic diversity and should provide an in- For the total sample, this eliminates ap- teresting variety of settings in which to ex- proximately half of the couples in the plore marital dissolution and remarriage. genealogies, although this proportion The data set on which the present study varies substantially from village to village is based has been developed in connection (Knodel, forthcoming). with a broader study of’demographic be- One additional problem with the source havior in the past. For six of the fourteen material deserves mention since it is of villages, virtually all data for all families relevance to our analysis. Information on were coded and processed into machine- marriage order in the genealogies is un- readable form. However, because coding doubtedly imperfect. The marital status the genealogies in full was extremely time- of a bride and groom at the time of mar- consuming and expensive, and because riage is generally not stated explicitly, and the emphasis of the broader study for unless there was an indication to the con- which these data were processed is on trary, marriages were assumed to be first reproductive behavior, for the remaining marriages for the purpose of the present eight villages, only couples for whom at analysis. For those individuals whose least some prespecified minimum amount marriages all occurred within the villages, of information was available were coded. there was no problem in correctly classify- Specifically excluded are those couples ing marriage order (barring gaps in regis- for whom sufficient information is un- tration and incorrect linkages between available to permit their use for fertility marriages). However, in cases where a analyses. Thus, with some minor qualifi- widow or widower from outside the cations, couples for whom the date of village remarried in the village, it is possi- marriage, the age of the wife, the date of ble that their marriage order was incor- the end of the union, or the birth dates of rectly classified as first marriage. Even in any known child could not be determined these cases, the individual’s marital status are excluded. Also excluded are the few or some reference to a previous marriage couples for whom divorce, separation, or was sometimes indicated. Thus, the extent marriage annulment were indicated. A to which higher-order marriages are in- 37 correctly classified as first marriages is ased downward to some extent because probably not extensive.’ This problem is they are based only on couples for whom not unique to village genealogies, but is the marriage date and both death dates present in all family reconstitution studies are known. Since the main reason death whenever they are based on marriage dates would not be known is out-migra- registers which do not state the marital tion from the village and since the longer status of brides and grooms as a matter of a couple lived the longer they would be at routine. risk of out-migration, the sooner after marriage a spouse died, the greater is the III. Marital Dissolution chance the couple would be included in Remarriage patterns are of necessity the analysis.2 The risk that such a bias closely linked to prevailing patterns of distorts the trend over time is less serious, marital dissolution. In German villages provided major changes in the extent to during the eighteenth and nineteenth cen- which this bias operated did not occur.’J turies, divorce, separation, and annulment were rare events and almost all mar- until the death of one of riages persisted ’The age at the end of marriage for couples who the spouses. Since marital dissolution was move in from elsewhere should be biased in the op- virtually synonymous with widow- or posite direction. However, such couples are general- widowerhood, the exclusion of the few ly excluded from the analysis since their marriage couples whose marriages ended otherwise dates are unknown. Moreover, unless their age at burial or at some other event occurring in the village creates no serious for the study problem is given, their age at the time of marital dissolution is of marital dissolution and remarriage for also unknown. the time period covered by the present study. Given the close link between the ’If out-migration of married varied two phenomena, we begin our analysis of persons over time, the trend could be affected. Another with an examination remarriage patterns potential bias, however, may also affect the trend. of marital dissolution. Calculations of the age at the end of union for fami- Of all marriages 1700-1899, 48 percent ly reconstitution data are potentially biased down- ended through the death of a wife and 52 ward for the initial decades following the establish- ment of parish records relative to later periods, to percent through the death of a husband. the extent ages at death are determined by linking The mean of husbands and age surviving the death of a spouse to the spouse’s birth date. The wives at the time a marital union ended is reason for this is that older persons marrying and presented in Table 1 for different mar- dying during this period will have been born prior to riage cohorts of the combined sample of the start of the birth register and thus their age will not be known, while younger persons will be more fourteen villages. Results are shown for likely to have been born after the start of the all combined and for first and marriages register. While part of the increase in the age at last marriages (of surviving spouse) sepa- marital dissolution between the cohorts married dur- rately. These estimates are probably bi- ing the first and second halves of the eighteenth century might be attributable to this bias, the rising trend is probably genuine, since age at death was ’To the extent that all marriages of an individu- often included in death entries in the parish registers, al are included in the genealogy, remarriages of dif- thus eliminating the potential truncation bias re- ferent orders can also be distinguished. For the total ferred to above. In addition, for a number of the sample of all villages, approximately 11 percent of villages, the birth registration began considerably marriages involving a remarriage included a spouse, before the eighteenth century, thus eliminating any generally the husband, who had already previously problem for calculations of age at end of union. The remarried. In the analysis, we do not treat this group problem would be more severe if the tabulations separately from those who remarry for the first time. were made according to year of marital dissolution 38

TABLE 1. MEAN AGE OF SURVIVING SPOUSE AT END OF MARRIAGE.a

NOTE: Results are restricted to couples for which both death dates are known. aBy sex and year of marriage, combined sample of all villages. All the tables that follow, except for Table 5, are a combined sample of all villages. bThe number of cases for the individual cohorts range from 208 surviving husbands of last marriages 1700-1749 to 1246 surviving wives of all marriages 1750-1799.

For marriages of all orders, the average nate the possible effect of changing prob- age of a surviving wife and a surviving abilities of remarriage on the age at the husband at the time of the end of the end of the union, we can examine the age union was quite similar and increased at the end of first marriages which is, of generally with successive marriage course, free of any influence of remar- cohorts. For surviving husbands, the im- riage. Changes in the age at marital dis- provement represents a six year increase solution for surviving spouses in first while for wives, a seven-and-a-half year marriages differ only modestly from the increase is evident. results for all marriages, indicating that Trends in the average age at the end of changing remarriage probabilities had a union for all marriages are potentially only a moderate impact on the latter. The influenced by the changes in the age at somewhat greater increase in the mean marriage, the extent of remarriage, and age at the end of first marriages compared the level of adult mortality. The average to all marriages reflects the decline in the age at the end of a union tends to be probability of remarriage examined in higher when age at marriage is later, Section IV. The reason for this is as remarriage more frequent, and adult mor- follows: the average age at the end of tality lower. Since age at marriage was remarriages is virtually always older than relatively stable (see Table 4), it is unlikely the average age at the end of first mar- to have exerted much influence.4 To elimi- riages. Since a decline in the probability of remarriage generally reduces the proportions of marriages that are remar- rather than by year of marriage. Since we are con- riages, the compositional shift tends to sidering only marriages which occurred no earlier lower the age at the end of all marriages, than 1700, we are dealing almost entirely with per- other sons born at the earliest in the second half of the things being equal. Hence, changes seventeenth century, and indeed predominantly in the probability of remarriage dampen those born after the third quarter of the seventeenth the observed increase in the age at the end century. of union for all marriages. 4For trends in the age at first marriage, see Undoubtedly, the predominant factor Knodel, 1983. responsible for the increase in the age at 39 marital dissolution was an improvement ably older than surviving wives at the end in adult mortality. Unfortunately, direct of last marriages. This is most likely a estimates of the level and trend in adult reflection of the greater likelihood of mortality (in contrast to infant and child husbands in last marriages to have been mortality) from family reconstitution previously married and thus for their last data are methodologically difficult and marriage to be a remarriage. For example, have not been attempted for the present among all last marriages during the eigh- analysis (see Wrigley, 1968). However, teenth and nineteenth centuries, 21 per- the fact that child mortality at ages above cent involved remarriages for husbands infancy declined during much of the compared to only 10 percent for wives. In period under study is consistent with the general, the results (whether for all, first, suggestion of improving adult survivor- or last marriages) indicate that over the ship in the sample villages (Knodel, 1983). course of the eighteenth and nineteenth Scattered evidence from various areas of centuries, both husbands and wives re- Germany also provides support for the mained in marital unions until reaching hypothesis of improvements in mortality progressively older ages. This, in turn, past infancy (Imhof, 1981; Lee, 1980). had important implications for trends in Increases in the age at dissolution of the remarriage. surviving spouse’s last marriage are also In the absence of divorce, annulment, of interest as they indicate the age at or separation, the average duration of a which permanent widow- or widower- marital union is determined by the age at hood began. However, these figures are entry into marriage and the age at marital particularly sensitive to the declining dissolution. Since the average age at chances of remarriage (especially since dissolution was increasing while the age at they include first marriages not followed marriage remained relatively stable, the by a remarriage). This explains why the average duration of a marital union age at the end of last marriage increased generally increased during the eighteenth somewhat less than for all marriages and and nineteenth centuries. This is shown in considerably less than for first marriages. Table 2, which indicates the average dura- In contrast to the average age at the end tion of marital unions by age at marriage of a first marriage, which was almost for both spouses. As would be expected, identical for surviving spouses of both the average duration of a marital union sexes, surviving husbands were consider- was inversely related to the age of entry

TABLE 2. AVERAGE DURATION OF MARITAL UNIONS AT TIME OF DISSOLUTION.

NOTES: Results are limited to marriages for which the date of the end of the union is known with certainty. Results based on fewer than 20 cases are omitted. 40 into the union. At the same time, for all last marriage is shown in Table 3. The lat- age-at-marriage categories, the average ter category includes both first marriages duration of marital unions increased for of persons who have married only once both husbands and wives between the (since their first marriage was also their eighteenth century and the second half of last marriage) and the last marriages of the nineteenth century. persons who remarried. In the cases of Only rarely did both spouses die at the both first and last marriages, the period same time. Thus, the dissolution of a mar- of survival is simply the interval between riage was typically followed by a period the death dates of the two spouses. Thus, during which the surviving spouse lived as following the end of the first marriage, a widow or widower, or if remarriage oc- the survival interval includes any time curred, as the partner of a new spouse. spent by the surviving spouse in a remar- During this period, the surviving spouse’s riage. The number of years spent between living conditions undoubtedly changed the end of the last marriage and the death substantially, particularly if remarriage of the surviving spouse refers exclusively did not occur, as he or she would no to a period when the surviving spouse was longer function as part of a conjugal unit. no longer part of a conjugal unit but In cases where remarriage occurred, the rather a permanent widower or widow un- period of widowhood was curtailed and til his or her own death. the surviving spouse once again func- As the results indicate, the surviving tioned as part of a single conjugal unit un- spouse, whether husband or wife, lived til the death of one of the spouses again between sixteen to nineteen years on the disrupted the marriage. While family average following the death of his or her reconstitution data are not suited for first spouse. There appears to be little dif- determining with whom, if anybody, a ference either over time or between the widow or widower lived following the dis- sexes in this respect. There is also no clear solution of a marriage, they can shed light trend over time in the mean duration of on the duration of widowhood and the survival following last marriage. The lack process of remarriage. of a difference in the lengths of husbands’ The mean number of years lived by the and wives’ survival interval following a surviving spouse following both first and first marriage is consistent with the

TABLE 3. MEAN NUMBER OF YEARS LIVED BY THE SURVIVING SPOUSE.

NOTES: Results include marriages prior to 1700 that ended 1700 onwards. Results are restricted to couples for whom both death dates are known.

. alncluding periods during which a remarriage may have occurred. 41

TABLE 4. PERCENT DISTRIBUTION OF MARRIAGES AND MEAN AGE AT MARRIAGE BY PRIOR MARITAL STATUS.

NOTES: Results in parentheses are based on fewer than 50 cases.

similar average ages of surviving spouses IV. Remarriage of both sexes at the time the first marriage ended, and indicates a concomitant simi- a. The distribution marriage types larity in average age at death for men and of women who survived their first marriage. The evolution of the distribution of mar- The average number of years survived by riages by marriage order combinations is a husband following the end of his last shown for the combined sample of all marriage, however, was less than for a villages in Table 4, along with the mean wife following her last marriage. This age at marriage according to the prior reflects differences in the average age of marital status of the spouses. There was a widows and widowers at the end of last steady increase in the proportion of marriages, which in turn was due to men’s primary marriages (bachelors with higher propensity to remarry. The fact spinsters) which rose over the eighteenth that first marriage survivors of either sex and nineteenth centuries from approxi- had similar ages at death indicates that mately two-thirds to four-fifths of all mortality differences between the sexes marriages while the proportion of remar- were relatively unimportant compared to riages, particularly those involving a differences in remarriage chances in ac- widow, decreased. Since divorces are ex- counting for the differences in the dura- cluded from these tabulations, and indeed tion of widowhood following last mar- were virtually nonexistent, they have no riage. effect on the figures. 42

Bideau (1980) has argued that remar- for the first half of the eighteenth century riages between widowers and spinsters and the second half of the nineteenth cen- gradually became the dominant form of tury. A similar, modest compositional ef- remarriage, with the proportion of fect due to the increasing share of primary widow/bachelor remarriages undergoing marriages affected the average age of the greatest decline in the period from the women for all marriages, which was late seventeenth to the mid-nineteenth slightly lower at the end of the period century, while the incidence of under observation than at the beginning. widower/widow remarriages remained b. Probabilities rather stable. Our data tend to confirm of remarriage Bideau’s findings. Considering only Following marital dissolution, the surviv- remarriages, we find that while ing spouse is subject to the competing widower I spinster marriages rose from 56 risks of remarriage and death. For this to 71 percent of all remarriages over the reason, life table techniques are well four periods, bachelor/widow marriages suited for a detailed examination of the declined from 31 to 22 percent of the probability of remarriage. The life table total. Widower/widow remarriages approach specifically allows for the fact declined somewhat, too, falling from 14 that the surviving spouse is at risk of to 8 percent of the total. If we shift our remarriage only up until the time that perspective slightly to examine the remarriage actually occurs or until the previous marital status of people whom time at which the surviving spouse dies, widowers and widows chose to marry, we whichever comes first. The surviving find that over the four time periods, a spouses who do not remarry prior to higher percentage of both widowers and death are eliminated from the denomi- widows chose single people as their part- nator of the probabilities of remarrying ners in a remarriage. The percentage of following their death, while those who do remarrying widows who chose a bachelor remarry are eliminated from the denomi- as their new husband increased from 69 to nator following their remarriage. For ex- 75 percent, while the percentage of remar- ample, imagine two populations with rying widowers who chose a spinster in- equal probabilities of remarriage at each creased from 80 to 91. duration following marital dissolution for The average age at marriage for both surviving spouses but different probabili- men and women in each of the prior ties of surviving to a given duration. marital status categories remained rela- Under such circumstances, a lower pro- tively constant over the period under portion of the survivors at the time of observation. There was only a very slight dissolution would remarry in the higher increase in the average age of men enter- mortality population because greater pro- ing primary marriages and an even less portions would die prior to having found pronounced increase for women entering a new marriage partner. Life table primary marriages. For men, a slight in- analysis, however, would reveal that the crease in the age at marriage is evident probability of remarriage at any given over the period covered for other prior duration since dissolution for those who marital status combinations as well, but continue to survive was equal in the two due to the increasing share of primary populations. marriages, which occur at younger ages To apply life table techniques, informa- than other types of marriages, the average tion is needed on the date of dissolution age of men for all marriages was identical of the union and the date of death of the 43

FIGURE 1. CUMULATIVE PROPORTIONS REMARRYING AS ESTIMATED BY LIFE TABLE TECHNIQUES, FOR SURVIVING SPOUSES OF MARRIAGES ENDING 1700-1899.

NOTES: Results are limited to the six fully coded villages and to couples for whom both death dates are known. Marriages ending in annulment or divorce in which migration of the surviving spouse is indicated are excluded. surviving spouse or the date of remar- The cumulative proportions of widows riage, whichever comes first. Family and widowers remarrying according to the reconstitution data typically contain such number of months since marital dissolu- information. For the present study, how- tion as estimated by the life table ap- ever, the date of remarriage was coded proach are shown in Figure 1 for the com- only for the six fully coded villages and bined sample of the six fully coded not for the remaining eight for which the villages. The results clearly show a strik- abbreviated scheme was applied. Thus, ing difference in the probability of remar- full life table techniques can be applied riage for men and women. Widowers were only to the six fully coded villages. For the far more likely to remarry than widows. remaining villages, however, information For the six villages taken together during on whether or not a remarriage occurred the period of the eighteenth and nine- was coded. Thus, a somewhat less de- teenth centuries, the probability of remar- tailed analysis of remarriage is possible riage during the ten years following the for all fourteen villages in the sample. end of a union was more than twice as Before turning to the complete set of high for men as for women. Moreover, fourteen villages, we examine the proba- widowers who remarried tended to do so bility of remarriage in detail based on an more quickly than widows who remar- application of life table techniques to the ried. six fully coded villages. For both sexes, the probability of 44 remarriage tapered off substantially by by the end of the first year following the end of five years following marital marital dissolution. dissolution. Thus, for men, the probabili- Table 5 indicates the probability of ty of remarriage within the first five years remarriage estimated by the life table ap- after the end of a previous marriage was proach for the six fully coded villages ac- approximately 42 percent, increasing to cording to the year of dissolution as well only 45 percent by the end of ten years. as for each village separately and accord- For women, the probability of remarriage ing to the age at dissolution. A clear time within the first five years following trend in remarriage probabilities is evi- marital dissolution was approximately 19 dent, with a sharp reduction in the pro- percent, increasing to only 20 percent by portions remarrying occurring from the the .end of ten years. Indeed, these life eighteenth century onward. Since all mar- table calculations show that among those riages were coded for the fully coded husbands who remarried by the end of ten villages including those occurring after years, more than half did so by nine 1900, it is possible to extend the analysis months following marital dissolution. For into the twentieth century, although much those women who remarried within the of our discussion focuses on the eigh- first ten years, more than half remarried teenth and nineteenth centuries.

TABLE 5. LIFE TABLE ESTIMATES OF THE PROPORTION OF WIDOWERS AND WIDOWS REMARRYING.a

acombined sample of six fully coded villages. NOTES: This table is limited to couples for whom both death dates are known. Marriages ending in annulment or divorce or in which migration of the surviving spouse is indicated are excluded. Results in parentheses are based on fewer than 100 persons entering into the initial life table calculation. 45

It is evident that the trend that was 16 percent by the first quarter of the twen- already strongly apparent before 1900 tieth century. For women, an even more continued at least until the first quarter of drastic reduction in the pace of remar- the twentieth century.5 For husbands riage occurred. The proportion of those whose marriage ended in the death of the women remarrying within ten years of wives during the eighteenth century, the marital dissolution who did so by the end probability of remarrying within two of the first year, declined from almost years following marital dissolution was half in the eighteenth century to only 3 almost 50 percent and close to 60 percent percent by the first quarter of the twen- by the end of ten years. In contrast, by the tieth century. This decline in the propor- second half of the nineteenth century, the tion who remarried at all, as well as the probability of remarrying within two sharp reduction in the pace of remarriage, years dropped to less than 30 percent and has been suggested by other historical within ten years to only a little more than studies (Flandrin, 1979:115-116; cf. Lee, one out of three. By the first quarter of 1981:96). the twentieth century these probabilities The probabilities of remarriage in the dropped substantially further. For six fully coded villages show considerable women, there was an even greater propor- variation. For men, remarriage was least tional decrease in the probability of common in the East Frisian village of remarriage. For those widows whose Middels and, moreover, occurred at a husbands died in the eighteenth century, relatively slow pace there. For women, approximately one-fifth married within remarriage was also relatively uncommon two years following marital dissolution in Middels, although a similar pattern was and almost a third by the end of ten years. evident for the village of Oeschelbronn as In contrast, by the end of the nineteenth well. The pace of remarriage for widows century only 5 percent remarried within was also very slow in Middels relative to two years and less than 10 percent by the the other villages. It is interesting to end of ten years. observe that the remarriage patterns did The changing nature of remarriage is not correspond strictly to religious dif- also reflected in the declining pace at ferences among the villages. While both which remarriage occurred. During the Oeschelbronn and Middels, the two eighteenth century, widows and widowers villages characterized by the least remar- remarried far more rapidly than in subse- riage, are Protestant, the Waldeck villages quent periods. For men, the life table of Braunsen and Massenhausen are also estimates indicate that the proportion Protestant and yet had considerably who remarried within six months, among higher rates of remarriage. The two Cath- all who remarried within ten years follow- olic villages of Kappel and Rust in Baden ing marital dissolution, declined from 42 were . also characterized by high rates of percent in the eighteenth century to only remarriage, but not noticeably higher than those for the two Waldeck villages. A number of historical demographic 5Given the increasing importance of divorce as a studies have established a strong negative source of marital dissolution during the twentieth association between the probability of the which is limited to century, present analysis, remarriage and the age of the surviving studying remarriage following marital dissolution as spouse at the of marital dissolution a result of death, is not extended beyond the first time quarter of the twentieth century even for the fully (e.g., Bideau, 1980:34-35; Cabourdin, coded villages for which such data are available. 1978:318-319). This same relationship is 46 apparent in German villages during the death. While in some cases this involves eighteenth and nineteenth centuries. As surviving spouses who moved out of the indicated by the life table estimates for the village and died elsewhere, in all likeli- six fully coded villages, the probability of hood if the death date were available to remarrying following marital dissolution the genealogist, information on whether declined sharply with age for both men or not that spouse had remarried was also and women. For men, the large majority available. who were under age fifty at the time of The life table approach applied to the marital dissolution remarried within ten fully coded villages had the advantage of years while only a small percentage did so estimating the probability of remarriage if they were age sixty or above at the time by duration since marital dissolution, of the death of their wife. For women, the after allowing for the competing probabil- large majority remarried only among ity of death. Because the date of remar- those who were under thirty at the time of riage has not been coded for the complete the death of their husband, although set of villages, it is not possible to almost half of those who were in their calculate probabilities of remarriage net thirties at the onset of widowhood remar- of the probability of death as is true with ried as well. Only a very small percentage the life table approach. Thus, results are of women over age forty at the time of influenced not only by the probability of marital dissolution remarried and almost remarriage but also by the probability of none over age fifty did. From the point of death. In order to minimize the impact of view of the reproductive potential lost the competing risk of death from the esti- through marital dissolution, it is of in- mates for the complete set of all villages, terest to note that a substantial propor- we have made separate estimates of the tion of women who became widows in probability of remarriage for those sur- their prime reproductive ages did remarry viving spouses who lived at least five years and have an opportunity to resume their after marital dissolution. As noted in con- reproductive careers. Thus, the low rates nection with the life table analysis of of remarriage for women above age forty, remarriage in the fully coded villages, the while perhaps having a substantial impact vast majority of remarriages took place on the social circumstances of the women within five years following marital involved, had little demographic signifi- dissolution. Thus, by restricting calcula- cance. tions to those who survived at least five While it is not possible to use the life years, we can achieve a close approxima- table approach to calculate the full set of tion to the probability of remarrying in- remarriage probabilities by duration dependent of the competing probability (since marital dissolution) for the com- of death. bined sample of fourteen villages, it is Estimates of the proportion remarrying’ nevertheless possible to estimate the pro- among all survivors and among those sur- portion who ever remarried. As with the viving at least five years for both men and life table estimates for the fully coded women are presented in Table 6 for the villages, it is necessary to restrict calcula- full set of fourteen villages combined. As tions to those couples for whom both expected, estimates of remarriage based death dates are known, in order to maxi- on spouses surviving at least five years are mize the likelihood that the surviving substantially higher than estimates of spouse was under observation for the full remarriage among all survivors. Given the period between marital dissolution and larger number of cases available for 47

TABLE 6. PROPORTION OF WIDOWERS AND WIDOWS REMARRYING BY YEAR OF MARITAL DISSOLUTION.

NOTES: Results include marriages prior to 1700 that ended in 1700 or later. Results are restricted to couples for whom both death dates are known. analysis from the entire set of fourteen both the subset of fully coded villages and villages compared to the subset of six fully the complete sample of all villages do coded villages, it is possible to trace trends agree, however, in indicating a very sub- in probability of remarriage in greater stantial reduction in the probability of detail. remarriage during the period under obser- Whether we consider all survivors or vation. In general, the results based on the only those surviving at least five years, the sample of all villages suggest somewhat results indicate that the probability of higher probabilities of remarriage for any remarriage for both men and women given period of time than do the results declined steadily during the eighteenth based on the subset of six fully coded and nineteenth centuries. Since the full villages. This is apparent from the fact sample of all villages is based only on that the proportion remarrying among marriages occurring through 1899, esti- those surviving at least five years in the mates of the probability of remarriage combined sample of all villages was gener- during the twentieth century would be ally higher for equivalent time periods seriously biased due to sample censorship than the probability of remarriage within and thus are not indicated.’ Results of ten years following marital dissolution based on the life table results for the six .. ~ fully coded villages.’7 6Since only marriages prior to 1900 are included in the data set for the full sample, marriages ending during successive years in the twentieth century problem is not present at the beginning of the eigh- would be to couples that on average would be pro- teenth century, since the results are based on mar- gressively older simply because couples who were riages regardless of whether they were contracted younger at the time of dissolution during the twenti- prior to or subsequent to the turn of the eighteenth eth century would be more likely to have married century. during the twentieth century. Given that the probability of remarriage is strongly related to the age at 7These two estimates are of course not precisely which marital dissolution occurs, such a selection of equivalent. Those based on the life table approach as cases would seriously bias downward the probability applied to the fully coded villages estimate the prob- of remarriage for marriages dissolving during the ability of remarriage only within the first ten years twentieth century if based only on marriages con- following marital dissolution and thus do not incor- tracted prior to the twentieth century. The reverse porate the fact that a small proportion of persons 48

TABLE 7. PROPORTIONS OF WIDOWERS AND WIDOWS REMARRYING BY AGE AT DISSOLUTION AND YEAR OF DISSOLUTION.

NOTES: Results in parentheses are based on 20-49 cases; results based on fewer than 20 cases are omitted. Results include marriages prior to 1700 that ended 1700 onwards. Results are restricted to couples for whom both death dates are known.

The probabilities of remarrying accord- to the. of the at will remarry after ten years. In contrast, the esti- ing age surviving spouse mates based on the set of all villages incorporate the time of marriage dissolution as well as remarriages at any duration including those occur- according to the year of marital dissolu- ring after ten years. On the other hand, the life table tion are shown in Table 7 for the com- estimates eliminate from the probability of remar- bined sample of all villages. Again, results riage the competing risk of dying during any time are both for all survivors and prior to the ten-year period subsequent to marital presented dissolution, while those based on the complete set of for those who survived at least five years villages eliminate only the competing risk of dying following the time of marital dissolution. during the first five years following marital dissolu- As was true for the subset of the six ful- tion. Since little remarriage occurs after five years ly coded villages, the results for the com- and even less after ten years, and since these biases bined of all show a are in opposite directions, it is unlikely that they sample villages very have a serious effect on the comparability of the two strong relationship between age at end of sets of results. union and probability of remarriage. 49

Moreover, this is true throughout the en- For both men and women, the propor- tire period under observation. At any tionate decline in the probability of given age at the time of marital dissolu- remarriage was directly related to the age tion, the probability of remarriage for a of the spouse at the end of the union. A man was greater than for a woman. Sex much greater proportionate decline in the differences were particularly pronounced probability of remarriage occurred for at the older ages. For men or women who both men and women at older ages than were widowed under age thirty, the vast for those who were younger at the time of majority remarried, although during the marital dissolution. For men under thirty, nineteenth century there was a noticeable the probability of remarrying remained decline in the probability of remarriage virtually unchanged during the two cen- even among widowed women under thir- turies. Moreover, there was only a modest ty. For both sexes, remarriage was not decline for men who were in their thirties common after age sixty, even after con- at the time of the end of the previous trolling for the competing probability of union. However, our results indicate a death by limiting consideration to those very substantial reduction in the probabil- who survived at least five years. For sur- ity of remarriage for men who were older viving spouses who were in their forties at when their marriage ended. For women, the time of marital dissolution, remar- there was a similar relationship between riage was quite common for men but rela- the proportionate decline in the probabili- tively rare for women, except during the ty of remarrying and the age at the time of first half of the eighteenth century. In ad- marital dissolution although, unlike men, dition, while a significant proportion of women in their thirties experienced a sub- men who were in their fifties at the time of stantial reduction in remarriage probabili- marital dissolution remarried, only a very ties. Indeed, the proportion of women in small minority of women of the same age this group remarrying during the second did so. half of the nineteenth century was less As indicated above, the average age of than half of the proportion remarrying the surviving spouse at the time of marital during the first half of the eighteenth cen- dissolution increased during the period tury. For women above age fifty, the under observation. Given the fact that probability of remarrying was reduced to there is a strong negative association be- very close to zero by the end of the nine- tween age at dissolution and probability teenth century. of remarriage, the increasing age at end of c. and union would contribute to a decline in the Remarriage, occupation, social status overall probability of remarriage even if no changes in age-specific probabilities of The proportion remarrying for the com- remarriage occurred. As is evident in bined sample of all villages among men Table 7, however, such a compositional and women surviving at least five years effect can explain only a part of the following the death of a spouse is shown decline in the overall probability of in Table 8 according to the occupational remarriage that occurred during the eigh- group, occupational subcategory, and teenth and nineteenth centuries in Ger- village leadership status of the husband.g man villages. With the sole exception of men under the of thirty, probability 8A detailed description of the occupational and remarrying for each age group of surviv- village leadership status categories and how they ing spouses declined during the period. were coded is provided elsewhere (Knodel, Forth- 50 coming, Appendix C). It should be stressed that Given the strong association of remar- there are a number of difficulties in constructing riage probabilities with age at the time of meaningful occupational groupings for past popula- marital dissolution and the decline in tions from limited information derived from parish remarriage chances over the two century registers: some occupational titles refer to occupa- period under observation, results are tions which no longer exist or are local expressions use of which are difficult to interpret; the social signifi- shown adjusted by multiple classifi- cance of any particular designation may vary over cation analysis for both of these influ- time and space; even within the same village and ences. For the four broad general occupa- time period the same designation can encompass a tional groupings, the probability of variety of situations. Moreover, changes in occupa- for both men and women was tions over the lifecourse create additional difficulties remarriage similar whether or not since the occupational designations derived from a remarkably adjust- marriage or burial entry may not represent a lifetime ment is made for age at and year of dis- occupation. Thus, occupational designations as solution. Even for the occupational sub- given in the village genealogies are necessarily im- categories, the probabilities of remarriage precise and incorporate a variety of ambiguities. were generally fairly similar, with only a The more than 1000 different occupational designations included in the genealogies comprising the few exceptions. sample were coded into 12 categories including a Among all occupational subcategories, residual category for all those that did not fit conve- the unadjusted probability of husbands into the other 11 Table niently (see 8). Categories remarrying was at least 45 percent and in either correspond to an occupational grouping one exceeded 60 per- which, on the basis of theoretical considerations or only subcategory previous empirical research, is expected to be associ- cent. After results are adjusted for age at ated with demographic behavior or because they cor- and year of dissolution, considerable respond to frequently mentioned occupations or oc- similarity remains in the probability of clusters in the cupational genealogies. Moreover, remarriage, although two groups, sailors they were chosen in such a way that they could be and professionals, fell below the 40 per- combined into three broader groups thought to cent level. The results for sailors be reflect, at least to some modest extent, common may social and economic circumstances. Nevertheless, in part a result of the small number of there is clearly variation within the categories them- cases involved. selves and, even moreso within the broader groups Regardless of their former husbands’ with respect to social and economic standing. occupation, the women’s probabilities of Farmers form both a category and a broad group- were far lower than men’s. In ing. Their presumed distinguishing feature is that remarriage they earn their livelihood primarily through agricul- most cases, both unadjusted and adjusted tural pursuits on land which is essentially under their results indicate a probability of remar- control. This category undoubtedly incorporates a variety of different situations even among persons with identical occupational designations. The prob- tisans and Skilled, includes artisans, fishermen, lem is reduced in villages where farmers and cot- businessmen, and professionals. All others, in- tagers can be distinguished and assigned to different cluding unknowns and those with two occupations categories. The second broad grouping is labeled not in the same broad grouping, were placed in a Proletarians and is intended to embrace workers fourth residual category. who depend for survival on the sale of their labor In each village genealogy, information about a power and who do not have extensive control over small number of husbands indicated that they held property or the means of production (see Tilly, 1981, some position of leadership in the village administra- Chapter 7). This grouping includes cottagers on the tion or in the local church. Presumably these posi- presumption that many cottagers needed to supple- tions were largely honorary and involved the most ment their income either by selling their services as prominent and respected members of the communi- agricultural laborers or in some other way. The ty. This information was used to determine a simple other categories subsumed under Proletarians are indication of village leadership status. Approximate- day laborers, other unskilled, sailors, weavers, and ly 5 percent of all husbands were categorized as other home industry. The third broad grouping, Ar- village leaders. 51

TABLE 8. PROPORTION OF WIDOWERS AND WIDOWS REMARRYING, BY OCCUPATIONAL GROUP, OCCUPATIONAL SUBCATEGORY, AND STATUS IN VILLAGE.

NOTES: Results include marriages prior to 1700 that ended after 1700 and are restricted to couples for whom both death dates are known. Results in parentheses are based on fewer than 50 cases. Adjustment for age at marital dissolution was made through Multiple Classification Analysis and was done separately for men and women. Unadjusted results for widowers include cases in which the age at dissolution is unknown while the adjusted results are based only on widowers for whom age at dissolution is known. In the case of widows, since age at dissolution is known for all, adjusted and unadjusted results are based on the same number of cases. alncludes some who were both farmers and fishermen, businessmen or professionals. blncludes some who were both proletarians and fishermen, businessmen or professionals. riage for women well below 30 percent. group is partially eliminated by adjust- The most notable exception is the unusu- ment. Thus, there appears to be little ally high probability of remarriage of systematic variation in patterns of remar- wives of. fishermen, which, although riage according to husband’s occupation. reduced, remained higher than for other It is worth noting that proletarian hus- groups even after adjustment. A some- bands did not suffer the same disadvan- what lower-than-average chance of tage as their wives in terms of chances of remarriage for women in the proletarian remarriage. 52

A comparison of the unadjusted proba- the village, if such women hesitated to bility of remarriage according to the hus- marry men who were from less advan- band’s village leadership status reveals taged positions than their former hus- roughly similar probabilities of remar- bands. In contrast, village leaders them- riage for those who were leaders and those selves may have been in a particularly who were not, but considerably lower favorable position to remarry as their pre- probabilities of remarriage for the wives sumably more privileged status in the of village leaders. The lower remarriage village might have made them particularly chances of wives of village leaders prior to attractive candidates even when they were adjustment undoubtedly reflects, in part, somewhat older. their more advanced age at the time of Time trends in the probability of remar- dissolution compared to wives of non- riage also show little consistent difference leaders (56.7 versus 49.9 respectively). In- by occupation. As shown in Table 9, the terestingly, there was far less difference in decline in the proportions remarrying oc- the age at the end of marriage between the curred in each major occupational group- village leaders and nonleaders themselves ing. Results are shown both unadjusted if they survived their wives (52.9 versus and adjusted for age at dissolution. While 51.4 respectively) and adjustment has adjusting for the changing average age at considerably less effect. Even after ad- dissolution over the eighteenth and nine- justment, widows of village leaders teenth centuries diminishes the extent of remarried somewhat less than widows of the decline in remarriage, it by no means nonleaders, perhaps reflecting a shortage eliminates it. F,or women, the adjusted of appropriate candidates for remarriage results indicate a remarkably similar for women in the more privileged strata of magnitude of decline between t4e first

TABLE 9. PROPORTION OF WIDOWERS AND WIDOWS REMARRYING, BY OCCUPATIONAL GROUP AND YEAR OF DISSOLUTION.

NOTES: Results include marriages prior to 1700 that ended from 1700 onwards and are restricted to couples for whom both death dates are known. Results in parentheses are based on 20-49 cases. Husbands, and wives of husbands with two occupations in different groups were coded in a separate mixed category not shown but are included in the results that refer to all husbands or wives. Adjustment for age at marital dissolution was made through Multiple Classification Analysis and was done separately for men and women. Unadjusted results for widowers include cases in which the age at dissolution is unknown, while the adjusted results are based only on widowers for whom age at dissolution is known. In the case of widows, adjusted and unadjusted results are based on the same number of cases since age at dissolution is known for all. 53 half of the eighteenth and second half of more than five years prior to marital the nineteenth centuries for all three ma- dissolution might have died before the jor occupational groups. end of the parents’ marriage but given the relatively low mortality of children past age five, this number is bound to be quite d. and the number Remarriage small. of children Table 10 shows the percent remarrying Among the circumstances that might sub- according to the number of children who stantially affect the changes of remarrying survived to age five, controlling for age at is the number of children that the widow widow- or widowerhood. Though mar- or widower has at the time of widowhood. riage probabilities were universally higher A thorough analysis of this would take in- for men than for women of the same age to account not only the number of surviv- group and with the same number of chil- ing children, but also their ages, since we dren, a similar inverse association charac- might expect quite different effects on terized both sexes in each age group. In chances of remarriage if children are still every age group, those with no children young and dependent compared to situa- surviving to age five were the most likely tions where the children are already old to remarry and indeed, except for enough to be in the labor force and thus widowers under forty and widows under provide support for the widowed parent. thirty, the reduction in the chances of Because of the way in which the data have remarrying between those with no chil- been coded for the present study, we dren and those with just one or two chil- relate the probability of remarriage to the dren is quite pronounced. It is interesting number of children surviving to age five to note further that a strong negative who had been born to the couple prior to association between age at dissolution and marital dissolution. It should be noted remarriage persists within each category that some of these children who were born of number of children.

TABLE 10. PROPORTION OF WIDOWERS AND WIDOWS REMARRYING, BY AGE AT MARITAL DISSOLUTION AND NUMBER OF CHILDREN SURVIVING TO AGE FIVE, MARRIAGES ENDING 1700-1899.

NOTES: Results include marriages prior to 1700 that ended after 1700 and are restricted to couples for whom both death dates are known. Results based on fewer than 20 cases are omitted. aTotal for widowers includes cases in which the age at dissolution is unknown; in the case of widows, age at dissolution is known for all. 54

While we are not controlling directly different time periods. Given the impor- for age of children at the time of marital tance of age at marital dissolution for the dissolution, presumably most if not all of chances of remarrying and the fact that the children of widowers or widows under the age at marital dissolution increased age forty would still be in dependent ages. over the time period being covered, we In contrast, most of the children of present results that are statistically ad- parents who became widowed after age justed for age at marital dissolution. fifty would likely be old enough to make When the eighteenth and nineteenth cen- significant contributions to the support of turies are broken. into four half-century the surviving parent, either in terms of periods, in the last three a substantial labor to the household if they were still negative association between number of living with their surviving parent or children and chances of remarriage is evi- through remittances if they were old dent. In contrast, for the first half of the enough to leave home as a servant, in eighteenth century, no consistent associa- some other employed status or to set up a tion appears. This may imply that at that household of their own. We do not have time, and earlier on; having children was any direct evidence that assistance was in not a discouragement to remarriage. fact provided, but the fact remains that However, these results for the first half- such assistance would only be possible century period need to be regarded with from children who had passed childhood. some caution since they are based on While the negative association between fever cases than the subsequent half- probability of remarriage and number of centuries. Moreover, for each of the sub- children holds at all ages, for both widows sequent half-centuries, the negative asso- and widowers, it is noticeably weaker for ciation is fairly similar and there’ is no widowers under age forty and widows trend toward a strengthening relationship, under age thirty. as would be expected if the nature of the In Table 11, we examine the association relationship were evolving in a linear of remarriage and number of children for fashion over time.

TABLE 11. PROPORTIONS OF WIDOWERS AND WIDOWS REMARRYING AMONG THOSE SURVIVING AT LEAST FIVE YEARS AFTER MARITAL DISSOLUTION, BY NUMBER OF CHILDREN SURVIVING TO AGE FIVE, AD- JUSTED FOR AGE AT TIME OF DISSOLUTION.

Given these various relationships be- adjusting for age at marital dissolution tween year of marital dissolution, age at and number of children. The inverse asso- marital dissolution, and number of chil- ciation between age at marital dissolution dren at the time of marital dissolution, it and remarriage persists after adjustment is useful to examine the probability of for year of dissolution and number of remarriage within a limited multivariate children; and the inverse association be- framework incorporating all three of tween remarriage and number of children these variables. Table 12 presents the is still evident after year of dissolution results of a multiple classification analysis and age at marital dissolution are con- in which the probability of remarrying is trolled. related to each of these three variables ad- V. Discussion justing for the effects of the other two. The analysis is done separately for A number of clear and interesting results widowers and widows. While, for each have emerged from the analysis. The variable, the association with the proba- secular decline in probabilities of remar- bility of remarriage is reduced by statisti- riage over the course of two centuries in cally adjusting for the other two vari- the German villages was accompanied by ables, substantial associations still persist an increase in the average age of both in all three cases. Thus, the probability of husbands and wives at the time death ter- remarriage declines over time even after minated their marriages. Since the age of

TABLE 12. UNADJUSTED AND ADJUSTED PROPORTIONS OF WIDOWERS AND WIDOWS REMARRYING, BY YEAR OF DISSOLUTION, AGE AT DISSOLUTION, AND NUMBER OF CHILDREN SURVIVING TO AGE FIVE.

entry into marriage was relatively stable, children. Often, they hypothesize, the this meant that the average duration of a most realistic alternative under such cir- marital union also increased. The fre- cumstances was prompt remarriage. quency and pace of remarriage following They further argue that the pressure to the death of a spouse declined substantial- remarry for both widows and widowers ly. Moreover, pronounced age and sex subsided as a &dquo;consequence of the differentials in the likelihood of remar- decrease in the social and economic im- riage were evident: widows were far less portance of filling the central positions in likely to remarry than widowers, and the the family, which has allowed for a probability of remarriage declined rapidly greater number of ’incomplete’ families&dquo; with age, particularly for women. (Mitterauer and Sieder, 1983:150). The decline in remarriage probabilities Widow- and widower-headed families be- was doubtless caused in part by improve- came more viable as the larger European ments in adult mortality, which gradually economy changed from one based on raised the ages of surviving spouses to domestic production to a more capitalistic levels at which remarriage has historically economy based on larger units of produc- been rather unlikely. However, declines in tion. adult mortality cannot account for the According to Mitterauer and Sieder’s substantial declines that were also evident view, the pressure to remarry would have in age-specific probabilities of remar- weighed most heavily on widows and riage. widowers with children, especially small . The fact that declines in age-specific children. Interestingly, however, data marriage probabilities affected both men from the German villages indicate a clear and women as well as all occupational inverse association between the number of groups in our sample of German villages children and the chances of remarriage suggests the presence of a social change of for both widows and widowers. Thus, wide scope. Previous scholars have sug- those widowed persons who would appear gested several explanations for the decline in most need of a new spouse were pre- in remarriage which range from the eco- cisely those least likely to remarry (see nomic to the psychological. Mitterauer also Schofield and Wrigley, 1981 :218- and Sieder (I983:I50), for example, argue 219). Data on remarriage by number of that declines in remarriage are linked to children suggest that understanding the aggregate trends in the European econo- possible economic or social pressures to my, particularly to the decline of the remarry provides only a partial explana- family as the primary unit of production, tion of the decline in remarriage. and the advent of capitalist economies. Though family reconstitution data do Their analysis is based on the idea that the not permit an assessment of the popula- preindustrial, rural household was one in tion at risk to marry, changes in the struc- which the entire family formed an integral ture of local marriage markets over two economic unit of production, and as such centuries may well have played a part in required that the two central adult roles, reducing the likelihood of remarriage, one male and one female, be filled given prevailing marriage customs. A (Mitterauer and Sieder, 1982; Sieder and number of studies have shown that the Mitterauer, 1983). When the unit was new spouses of both widows and broken, it was difficult for the surviving widowers were typically bachelors and spouse to, continue operations effectively, spinsters. In fact, as we have demon- especially if he or she was left with small strated, there was an increase over time in 57 the proportion of remarriages that allied some scholars, may gradually have re- widows and widowers to never-married duced the desire of young, single people persons. This trend may, of course, have to enter into marriage with a widow or a been the simple result of the fact that widower. It is noteworthy that analysis of there was a greater proportion of single the age gap between spouses in the sample than previously married people in the of villages shows a considerable reduction marriage &dquo;market.&dquo; However, the fact in the proportion of marriages in which that widowed persons with few children spouses were of sharply disparate ages were also most likely to remarry even (Knodel, forthcoming). after age is controlled suggests that these Mitterauer and Sieder (1983:62), as well two phenomena-the marriageability of as Jean-Louis Flandrin ( 1979:11 S) have both single persons and widows and argued that the historic growth of widowers with few children-may be stronger affective bonds between hus- related. As Bideau (1980:39) has argued, bands and wives over the course of mar- both single people and widows or riages that were themselves of longer widowers with few children may have duration contributed to the decline of re- been sought out as partners for remar- marriage and to the growth of intervals riage because of the difficulties of in- between widowhood and remarriage for tegrating two sets of children into one both men. and women. However, the household. Furthermore, there could also growth in marriage duration over the en- be difficulties in resolving complex ques- tire period under study was rather modest tions of inheritance among different sets for our sample village populations. of children. In populations where high rates of If our assessment of the importance of death or divorce lead to the frequent never-married people (or those with no or dissolution of marriage before the end of few children) as the most frequent remar- the reproductive span, remarried persons riage partners of widows or widowers is have the potential to contribute measur- correct, then a decline in the availability ably to the overall level of fertility and of single people or those with few children therefore to have a noticeable, if not would, all else being equal, reduce the large, demographic impact (Coale, 1981). probabilities of remarriage for widows Given the fact that remarriage in our and widowers. In fact, there is some evi- sample villages during most of the period dence from the village genealogies that under observation was relatively common emigration from the countryside may for women under age forty, the extent to have reduced the number of single per- which marital dissolution through the ear- sons of both sexes from rural marriage ly death of a husband reduced the repro- markets, particularly in the course of the ductive capacity of the population was nineteenth century (Knodel, forthcom- considerably mediated, at least until the ing). late nineteenth century. At the same time, The effects of the lower availability of the low probabilities of remarriage for single persons may also have been exacer- women forty or over had little or no im- bated by a growing reluctance on their pact on reproduction because such part to marry widows or widowers, par- women were either already terminally ticularly those much older than them- sterile or past their most fecund years. selves. A possible growth of sentiment in The reasons underlying the widely marriage and a rise in the ideal of a com- documented, sharp differences between panionate marriage, as hypothesized by widows and widowers in probabilities of 58 remarriage clearly merit further explora- suggested, future analyses of this impor- tion. If it is true that a substantial sexual tant subject should take into considera- division of labor and the typical economic tion not only pressures to remarry, but functioning of rural couples during the changes in the structure of local marriage eighteenth and nineteenth centuries re- markets and the values and attitudes quired both a man and a woman to func- towards marriage of single and previously tion efficiently, one might expect that married people. women would be under the same pressures to remarry as men (Baulant, 1976; Kennedy, 1973). It may have been easier BIBLIOGRAPHY for women to take on the male tasks for themselves than for men to take on female Baulant, Micheline tasks (Segalen, 1981:69). Segalen has also 1976 "The Scattered Family: Another Aspect of suggested that the cultural association Seventeenth Century Demography." In R. Foster and D. Ranum, eds., From between and may marriage childbearing Family to Society. Baltimore: Johns have reduced the likelihood of remarriage Hopkins Press. for women past the childbearing age Bellettini, Athos (1981:68). Nevertheless, this argument 1981 "Les remariages dans la ville and dans la would not explain the significant decline campagne de Bologne au dix-neuvième siècle." In J. Dupâquier, E. Hélin, in for women remarriage probabilities P. Laslett, M. Livi-Bacci, and S. Sogner, between ages 30 and 39, unless nascent eds., Marriage and Remarriage in Popula- fertility control was already lowering tions of the Past. New York: Academic women’s and men’s idea of the proper age Press. Alain for women to end childbearing (see Bideau, 1980 "A Demographic and Social Analysis of Knodel, 1983). Widowhood and Remarriage: The Exam- The phenomenon whose explanation ple of the Castellany of Thoissey-en- presents the greatest challenge, however, Dombes, 1670-1840." Journal of Family is the secular decline in remarriage itself. History 5:28-43. G. Although there is reason to believe that Cabourdin, 1978 "Le Remariage." Annales de the decline was than a rural larger purely Démographie Historique, 1978:305-336. the of phenomenon, continuous decline 1981 "Le remariage en France sous l’Ancien remarriage probabilities there seems most Régime (seizième-dix-huitième siècles)." In puzzling, given the longer persistence of J. Dupâquier, E. Hélin, P. Laslett, M. and S. Mar- domestic systems of production. In par- Livi-Bacci, Sogner, eds., riage and Remarriage in Populations of the we that and ticular, might expect peasants Past. New York: Academic Press. artisans were groups within which the Coale, Ansley J. family functioned more and longer as a 1981 "Introduction to Part III." In J. Dupâ- productive unit in comparison to pro- quier, E. Hélin, P. Laslett, M. Livi-Bacci, letarians. Yet there is little difference in and S. Sogner, eds., Marriage and Remar- riage in Populations of the Past. New levels of of these two the remarriage York: Academic Press. groups. Given the fact that the changing Dyrvik, Ståle nature of the economy must have affected 1981 "Gagne-pain ou sentiments? Trait du occupational groups differently, their remariage en Norvège au dix-neuvième similar patterns of decline in the probabil- siècle." In J. Dupâquier, E. Hélin, P. Laslett, M. Livi-Bacci, and S. Sogner, ities of lead us to believe that remarriage eds., Marriage and Remarriage in Popula- more than just economic changes lay tions of the Past. New York: Academic behind these trends. As our discussion has Press. 59

Flandrin, Jean-Louis German Family: Essays on the Social 1976 Families in Former Times: Kinship, House- History of the Family in Nineteenth- and hold and Sexuality. Cambridge: Cam- Twentiety-Century Germany. London: bridge University Press. Croom Helm. Goody, Jack Livi-Bacci, Massimo 1983 The Development of the Family and Mar- 1981 "On the Frequency of Remarriage in Nine- riage in Europe. Cambridge: Cambridge teenth Century Italy: Methods and University Press. Results." In J. Dupâquier, E. Hélin, Hajnal, John P. Laslett, M. Livi-Bacci, and S. Sogner, 1953 "Age at Marriage and Proportions Marry- eds., Marriage and Remarriage in Popula- ing." Population Studies 7:111-136. tions of the Past. New York: Academic 1965 "European Marriage Patterns in Perspec- Press. tive." In D. V. Glass and D. E. C. Mitterauer, M. and R. Sieder Eversley, eds., Population in History. Lon- 1982 The European Family. Oxford: Basil don : Edward Arnold. Blackwell. 1982 "Two Kinds of Pre-Industrial Household Schofield, Roger and Edward A. Wrigley Formation Systems." Population and 1981 "Remarriage Intervals and the Effect of Development Review 8:449-494. Marriage Order on Fertility." In J. Dupâ- Imhof, Arthur quier, E. Hélin, P. Laslett, M. Livi-Bacci, 1981 Die gewonnenen Jahre. Munich: C. H. and S. Sogner, eds., Marriage and Remar- Beck. riage in Populations of the Past. New Kennedy, Robert E., Jr. York: Academic Press. 1973 The Irish Emigration, Marriage, and Fer- Segalen, Martine tility. Berkeley: University of California 1981 "Mentalité populaire et remariage en Press. Europe occidentale." In J. Dupâuier, Knodel, John E. Hélin, P. Laslett, M. Livi-Bacci, and 1975 "Ortssippenbücher als Daten für die S. Sogner, eds., Marriage and Remarriage historische Demographie." Gesellschaft in Populations of the Past. New York: und Geschichte 1:288-324. Academic Press. 1983 "Demographic Transitions in German Sieder, R. and M. Mitterauer Villages." Research Report No. 83-45, 1983 "The Reconstruction of the Family Life Population Studies Center, University of Course: Theoretical Problems and Empiri- Michigan. Also scheduled for publication cal Results." In R. Wall, J. Robin, and in A. J. Coale and Susan Watkins, eds., P. Laslett, eds., Family Forms in Historic The Decline of European Fertility. Prince- Europe, 309-345. Cambridge: Cambridge ton : Princeton University Press (forthcom- University Press. ing). Smith, Richard M. Forth- Demographic Behavior in the Past: A 1979 "Some Reflections on the Evidence for the com- Study of 14 German Village Populations in Origins of the European Marriage Pattern ing the Eighteenth and Nineteenth Century. In in England." In C. C. Harris, ed., The preparation, title tentative. Sociology of the Family: New Directions Knodel, John and Edward Shorter for Britain, 74-112. Sociological Review 1976 "The Reliability of Family Reconstitution Monograph 28. Keele: University of Keele. Data in German Village Genealogies (Orts- Tilly, Charles sippenbücher)." Annales de Démographie 1981 As Sociology Meets History. New York: Historique 1976:115-154. Academic Press. Laslett, Peter Wrigley, E. A. 1977 "Characteristics of the Western Family 1968 "Mortality in Pre-industrial England: The Considered Over Time." Journal of Fami- Example of Colyton, Devon, over Three ly History 2:89-116. Centuries." Daedalus 97:546-580. Lee, W. Robert 1981 "Marriage, Fertility and Population 1980 "The Mechanism of Mortality Change in Growth in Eighteenth-Century England." Germany, 1750-1850." Medizin- In R. B. Outhwaite, ed., Marriage and historisches Journal 15:244-268. Society, 137-185. London: Europa Publi- 1981 "Family and ’Modernisation’." In cations. Richard J. Evans and W. R. Lee, eds., The