The UPDB database on which this study relies includes nearly complete genealogical information for several hundred thousand individuals in families descended from the Utah pioneers of the mid-nineteenth century. The family and other relationships identified in the dataset link mothers to their own children and facilitate the linking of multiple generations.12
Figure 1 shows age-specific marital fertility rates for selected birth cohorts of women in the dataset. Total marital fertility (ages twenty to forty-nine) fell from 11.0 in the pre-1850 birth cohort to 7.2 in the 1890 to 1899 birth cohort. Marital fertility decline was modest through the 1870 to 1879 birth cohort but accelerated thereafter. Although marital fertility rates fell for all age groups, the faster decline for that of older age groups suggests the importance of both spacing and stopping behaviors. The marital fertility decline documented by the UPDB lagged behind that of the overall white population of the United States by several decades. Period estimates show that the total marital fertility of white women in the United States was 6.9 in 1880--a level that the women in the UPDB did not reach until c. 1905 to 1910. Ewbank's state-level estimates of marital fertility using the Public Use Sample of the 1910 census indicate that Utah was one of only two states in the nation with m
values less than 0.2 in the period from 1905 to 1910. Evidently, Utah couples were relatively late to adopt parity-dependent fertility control.13
We begin with an examination of simple bivariate correlations of children ever born and intermediate reproductive indicators between index women, their mothers, and mothers-in-law. Results are shown by birth cohort to determine whether the strength of intergenerational relationships changed during the course of the fertility transition. We then proceed to an event-history analysis of birth spacing by parity. Time-dependent and time-invariant variables are introduced to test various premises.
shows bivariate correlations of reproductive indicators between women in the UPDB and their mothers and mothers-in-law. The results are limited to cases in which index women, their mothers, and mothers-in-law survived to age forty-five, had nonpolygamous husbands, and remained married to first husbands beyond age forty-five. Most reproductive indicators were positively associated between generations. The correlation coefficient between index women and their mother's completed fertility (0.085) was on the low side of those typically reported. The correlation improved slightly when comparing index women and their mother's fertility relative to other women in a nine-year moving birth cohort centered on their respective years of birth (the definition of relative fertility
being “women's number of children ever born minus her birth cohort's mean number”).14
Intergenerational Correlation Coefficients of Reproduction Indicators
Other indicators of reproduction--age at first marriage, age at first childbirth, age at last childbirth, duration of fertile period, and relative duration of the reproduction period--were positively correlated across generations. The intergenerational correlation in age of first marriage was particularly strong, likely playing a large role in the correlation in children ever born. We should be cautious, however, about assuming that the relatively strong correlation in age at first marriage and its likely contribution to children ever born resulted from an intergenerational transmission in the use of marriage as a strategy to control family size. If fathers' and husbands' occupations were correlated (which seems likely), the correlation in age at first marriage between mothers and their daughters may simply reflect the effect of occupation and SES on the feasibility and desirability of marriage. Although weaker, the correlation in age at last birth is better evidence that values and attitudes about family limitation were transmitted across generations. Index women with a young age at last birth were more likely to have mothers with a young age at last birth. The association may have been due to intergenerational correlations in the practice of effective parity-dependent fertility control (that is, “stopping” behavior) or to correlations in the practice of effective inter-birth spacing, which can also lower the age at last birth. Despite the emphasis in earlier studies on the conscious strategy of spacing of births to limit family size by women in the UPDB, however, the bivariate correlations show no intergenerational relationship in the intensity of fertility—defined as the number of children born divided by the duration of the childbearing period—or the relative intensity of fertility.15
Figure 2 illustrates two findings consistent with other studies. First, although index women with fewer siblings tended to have smaller families of their own, and index women with more siblings tended to have larger families, the relationship was not linear. Index women whose mothers had ten or more children averaged between seven and eight children of their own. Yet, index women whose mothers had fewer than six children averaged more than six children. Second, the relationship between index women and their mothers' children ever born was more responsive for first-born daughters than for second and higher-order-born daughters. shows that correlation coefficients for first-born index women were typically greater than correlation coefficients for second- and higher-order-birth women. The differential in the strength of the relationship between children ever born by birth order, however, is largely eliminated by comparing relative fertility across generations, suggesting that much of the birth-order effects were due to cohort effects. Necessarily, first-born daughters were born closer to their mother's cohort than their younger siblings, who were born later in the fertility transition.
Our expectation that correlations through the family of the index woman's husband would rival the size and statistical significance of correlations between the index woman and her mother received weak support. In general, the size of the correlation coefficients between index women and their mothers-in-law were noticeably weaker than correlations between index women and their mothers. The correlation coefficients between the children ever born and the relative number of children ever born of index women and of their mothers-in-law, for example, were approximately two-thirds of those observed between index women and their mothers.
Correlations in other reproductive indicators between index women and their mothers-in-law were even weaker, ranging from about one-third to two-thirds of the size of the coefficients between index women and their mothers. Although more modest than the intergenerational correlation between index women and their mothers' age at first marriage, the correlation between the marriage age of index women and that of their mothers-in-law was statistically significant, indicating the possibility of assortive marriage patterns by occupation, ethnicity, or location. The correlation likely played some role in the correlation in children ever born. Nevertheless, the statistically significant results for most reproductive indicators, including age at last birth, suggests that reproductive behaviors were transmitted from the husband's family of origin.
examines correlations between index women and their mothers and mothers-in-law by birth cohort. For the most part, the results confirmed our expectation about the growing importance of intergenerational transmission throughout the course of the fertility transition. Although mostly positive, correlation coefficients between index women and their mothers in birth cohorts before 1870 were not significantly different from zero for most reproductive indicators. This finding is consistent with the literature on pre-transition populations that found little or no intergenerational correlation in fertility. Thereafter, with the exception of the intensity of fertility measures—which remained statistically insignificant throughout the study period— correlation coefficients were positive and statistically significant.
Inter generational Correlation Coefficients of Reproductive Indicators by Birth Cohort
As was true for the overall analysis, age at first marriage had the strongest correlation between index women and their mothers. The size of the coefficient was significantly different from zero for index women in the 1850 to 1859 birth cohort and was larger in subsequent birth cohorts, reaching 0.198 in the 1880 to 1889 birth cohort. Our expectation that marital fertility decline would allow more room for conscious family-limiting behaviors to be passed across generations received support from the age at last childbirth correlations. Although the coefficient was about one-third the size of the correlation for age at first marriage, it was statistically significant for index women in birth cohorts after 1850, suggesting the increasing importance of the transmission of stopping behavior between index women and their mothers.
Correlation coefficients for index women and their mothers-in-law also increased from statistically insignificant levels in early cohorts to positive and statistically significant levels in later cohorts. Once again, the correlation coefficients between index women and their mothers-in-law were noticeably weaker than those between index women and their mothers. As a result, some of the variables were not statistically significant until relatively later-born cohorts of index women. The correlation between the age at last birth of index women and that of their mothers-in-law, for example, was roughly about one-half the size of the correlation in age at last birth between index women and their mothers. Although positive and significant in the 1850 to 1859 and 1890 to 1899 birth cohorts, the correlation was not statistically significant for index women in the 1870 to 1879 and 1880 to 1889 birth cohorts. Nevertheless, the positive correlation between index women and their mothers-in-law in two of the four cohorts suggests the transmission of stopping behavior across families was not limited to transmission between mothers and their daughters. In some small part, stopping behavior, which likely required significant male cooperation, was transmitted from the husband's family of origin.
The bivariate regression results shown in and are in mixed agreement with the results from Reher, Ortega, and Sanz-Gimeno's study of the fertility transition in the Spanish town of Aranjuez. Their results indicated stronger intergenerational correlations between the duration of reproduction and the age at last birth and no correlation in age at first childbirth, which served as a proxy in their study for age at first marriage. Both studies, however, document stronger correlations between index women and their mothers than between index women and their mothers-in-law and increasing strength in correlation coefficients throughout the course of their respective fertility transitions.16
Although valuable, bivariate correlations are limited. Selection restrictions to index women, mothers, and mothers-in-law whose marriages remained intact beyond their childbearing years result in limited use of the data available in the UPDB. Analyses that examine only bivariate correlations in children ever born also fail to consider that couples' reproductive decisions were part of a sequential process. Although some couples might have commenced childbearing with a desired number (and timing) of children in mind, might never have adjusted their target during their childbearing years, might never have suffered any unexpected losses of children in infancy and childhood, and might have effectively stopped their childbearing after meeting their target number, such couples most likely constituted a small minority. Most couples continually re-evaluated their reproductive goals and adjusted their behavior, whether to accelerate or postpone a birth or to cease childbearing altogether, in response to current conditions and future prospects. Important factors might have included their current number of surviving children, the mother's health, the family's economic condition, and the ability of older children and kin to assist in child rearing and household duties.
In recent years, the study of fertility has moved from reliance on cross-sectional aggregate data to longitudinal microdata and from analyses of the number of births to the timing and likelihood of the next birth. The shift in focus was made possible with advances in event-history methods, which allow the modeling of repeated events. Although the methods require longitudinal data, which are relatively rare for historical populations, the benefits are considerable. First, the ability to include censored observations allows much more of the available data to be analyzed. Women whose childbearing was interrupted by death or divorce, for example, can be included in the analysis. Second, the ability to include time-dependent variables allows researchers to examine the relevance of a wider variety of biological, demographic, and socioeconomic variables on couples' reproductive behavior in a multivariate framework that treats fertility as a dynamic, sequential process.
We rely on event-history methods to examine the intergenerational transmission of reproductive behavior in the UPDB. Because we know the precise timing of demographic events—including the exact date of births, marriages, and deaths across multiple generations— we construct a Cox continuous-time proportional-hazard model. The model assumes that the effects of covariates are proportional to the baseline hazard--the hazard of a woman giving birth since her marriage or last birth. We model each birth interval separately. The approach thus estimates the influence of independent covariates on the relative risk of an index woman giving birth in the interval.
Covariates from the index woman's mother and mother-in-law include their relative fertility, relative age at marriage, length of first birth interval, and a time-dependent covariate for vital status (0=dead, 1=living). For index women, fixed covariates include birth cohort, age at marriage, age differential from spouse, and whether a member of the LDS. Time-dependent variables include the index woman's five-year age group and whether her previous child died before the age of two in the interval. The latter variable is an attempt to control for the effect of differential breast-feeding. The premature death of an infant terminates breast-feeding, which results in a shortened duration of postpartum amenorrhea. Women whose previous child died before the age of two are expected to have shorter birth intervals and are more likely to progress to the next birth, all else being equal.
The model universe includes all index women married to nonpolygamous males linked to mothers and mothers-in-law married to nonpolygamous males. Because we rely on the relative fertility of mothers and mothers-in-law, we exclude index women with mothers or mothers-in-law whose first marriage ended before age forty-five. Most mothers and mothers-in-law still living when their daughters or daughters-in-law began childbearing were either no longer bearing children of their own or would soon be finished. Assuming that they lived in proximity to their daughters or daughters-in-law, most of them would have been able to help with the raising of their grandchildren or with other household duties.17
Hazard ratios for attributes of index women, their mothers, and mothers-in-law are shown in . Values greater than one indicate a greater hazard of birth (implying a shorter interval to next birth and a greater likelihood of having another child). With the exception of the interval between marriage and first birth, the hazard ratio for the relative fertility of index women's mothers is greater than one and statistically significant in all birth intervals, indicating that index women whose mothers achieved higher fertility than other women in their cohort were more likely to have another child and progress faster to the next birth, all else being equal. Index women whose mothers had one child more than the average for their cohort had, on average, a 1.9 percent increased hazard of having a child in birth intervals between one and nine. The hazard ratio remained at the same approximate level in higher birth intervals, despite the attrition of women with low fertility from the analysis.18
Proportional Model of Fertility Timing by Birth Interval and Family History (Hazard Ratios)
Index women whose mothers were still living had a significantly higher hazard rate of giving birth in all intervals, indicating that living mothers exerted a significant pronatal affect on their daughter's fertility. Coefficients ranged between 1.154 and 1.314; index women with living mothers experienced a 15 to 31 percent higher hazard rate of birth (the average for intervals between one and nine was 1.203, or a 20.3 percent increased hazard). The relative fertility and vital status of mothers-in-law of index women exerted similar, albeit slightly less, influence on the hazard ratio of giving birth. With the exception of birth intervals above nine, the relative fertility of mothers-in-law was positive and statistically significant in all birth intervals. Index women whose mothers-in-law had one child more than the average for women in their cohorts had, on average, a 1.4 percent increased hazard of having a child in birth intervals between one and nine (78 percent of the increased hazard associated with the relative fertility of index women's mothers). Index women with living mothers-in-law experienced an approximately 11 percent higher hazard of giving birth (55 percent of the increased hazard associated with the vital status of index women's mothers).
Although the results support the evolutionary biology literature stressing the roles of grandmothers in providing mothers with assistance in child rearing and the pronatal consequences of that care, the lesser influence of mothers-in-law relative to mothers is slightly at odds with a recent literature review of the extent to which kin affect fertility in modern pre-transitional populations. Although many studies report that the presence of maternal grandmothers has a pro-natal effect, Sear and Matthews' systematic literature review found a more consistent pronatal effect from paternal grandmothers, who might on the whole be less concerned about the effect of childbearing on the health of the mother.19
Covariates associated with index women were also significant. Age made the largest difference. That index women with membership in LDS had higher hazard ratios than non-Church women means a positive relationship between membership in the church and marital fertility. With a few exceptions, index women whose previous child died in the interval had higher hazard rates of giving birth. The result suggests either a shortened period of postpartum amenorrhea related to a shortened period of breast-feeding or an attempt on the part of couples to replace the lost child.
Bivariate correlations in children ever born and other indicators of reproduction between index women and their mothers in the UPDB signal the intergenerational transmission of reproductive behaviors in birth cohorts beginning in the late nineteenth century. Although modest, correlations in children ever born between index women and their mothers were statistically significant for index women in birth cohorts after 1860. Among other reproductive indicators, age at marriage had the strongest association across generations and was significantly correlated in birth cohorts after 1850. The statistically significant bivariate correlations in age at last childbirth in birth cohorts after 1860 suggest that the practice of parity-dependent marital-fertility control, or possibly inter-birth spacing behavior, was transmitted between mothers and daughters. Correlations between first-born index women and their mothers were stronger than correlations between second- and higher-birth-order index women and their mothers, suggestive of first-born daughters' greater socialization and greater identification with their mothers.20
This study also found modest bivariate correlations between children ever born and other indicators of reproduction between index women and their mothers-in-law, including correlations in age at last childbirth. Although the correlations were smaller than those between index women and their mothers, the results confirm the existence of significant intergenerational transmission of reproductive behavior from husbands' families; husbands (and mothers-in-law) probably played a significant role in determining reproductive behavior. The result is not unexpected, given the need for men's cooperation in most of the effective means of birth control during the fertility transition in the United States.
Results of the event-history analysis further confirm that reproductive behavior was transmitted across generations. Although hazard ratios associated with the relative fertility of mothers and mothers-in-law appear modest—averaging 1.019 for mother's fertility relative to other women in her birth cohort and 1.014 for that of mothers-in-law—the hazard ratio shown was for a one-child change in relative fertility. The standard deviation for the relative fertility of both mothers and mothers-in-law was 2.9 children throughout the period. Thus, index women whose mother's relative fertility was one standard deviation above the mean had a hazard ratio approximately 11 percent higher than that of index women whose mother's relative fertility was one standard deviation below the mean. For index women with mothers-in-law whose relative fertility was one standard deviation above the mean, the hazard rate was approximately 8 percent higher than for women with mothers in-law whose relative fertility was one standard deviation below the mean. Moreover, index women's increased hazard of birth was repeated in each birth interval.
The event-history analysis also confirms that the vital status of index women's mothers and mothers-in-law was strongly associated with the hazard of giving birth in each birth interval. Index women with living mothers and mothers-in-law were much more likely to progress to the next birth, all else being equal. Although we lack detailed residence information for index women and their mothers, the result supports the hypothesis that close kin, particularly mothers and mothers-in-law, represented critical help in the rearing of children. As “cooperative breeders,” Utah women born in the nineteenth century were not only influenced by the past example of the fertility of their mothers and mothers-in-law; they also adjusted the quantum and tempo of their childbearing in response to the presence of these relations.