'The number and quality of children': education and marital fertility in early-twentieth century Iowa.

Smith, Daniel Scott

Using remarkably rich data on the extent and type of schooling collected by the state census of Iowa in 1915, this essay explores the role of education in structuring the transition to low fertility in the United States. Its point of departure is a pioneering study that employed these data published for Iowa counties within a framework concerned with the quality of children. By also using information on individuals, the article attempts to isolate the distinctive role of schooling in influencing the fertility behavior of married women in Iowa.(1)

In American demographic history, very little is known directly about the impact of schooling on the fertility of women. Similarly, in American educational history, scarcely any research has focused on the impact of variation in education on any later behavior of individuals as adults. In addition to a narrow focus on educational institutions, the absence of appropriate data accounts for this gap in the history of education.(2) In 1940 and thereafter, the U.S. Census recorded the number of years of schooling for a sample of the population. Before that date, scholars must rely on the extremely time-consuming method of record linkage to tie status or behavior in adult life to the schooling received as a child.(3)

Education figures centrally in the general literature on the transition from high to low fertility for political, empirical, and theoretical reasons. First, education is, unlike other determinants of fertility, under the control of government policy makers. Since the 1950s, much of the research on the process of fertility decline has been framed by the need to curtail the rapid rates of population growth in Third World societies. In principle, the state can independently intervene to alter the extent and change the content of education. Thus, even though the magnitude of the impact of schooling on fertility might be less than other factors, it is important because it can be manipulated.

Second, education usually makes a substantial difference on fertility in studies across societies, over time, and within populations. While the relationship is not universal, it is robust. Fertility is typically found to be negatively related to the extent of schooling.(4) Education can affect more than fertility within marriage, making it necessary to control for a multiplicity of other demographic variables, especially nuptiality. Variation in schooling frequently is also intertwined with other attributes that capture socio-economic development. Thus, the relationship must be assessed within a framework that includes the several other factors that structure fertility behavior.

Third, the potential importance of education to fertility is complex and multifaceted. Given the significance of the topic, scholars from several social-science disciplines and theoretical traditions have offered diverse conceptual frameworks. The impact can occur later on adults of childbearing age or as the response to a change in the schooling of children themselves.(5) The role of schooling has been interpreted narrowly, or at the other extreme, has been advanced as the key to the revolution in fertility.(6) Current wisdom on population policy, as exemplified by the 1994 United Nations' conference in Cairo, Egypt, emphasizes the importance of education in demographic change, especially for the behavior of women.(7)

The Place of Iowa in Fertility and Educational Transitions:

In 1940, when the Census first recorded and tabulated data on the school attainment of adults, American fertility, which began its decline early in the nineteenth century, had plummeted to a level at which the population was barely replacing itself. At the end of the Great Depression, there was the familiar inverse relationship between education and both nuptiality and the number of children ever born. In order to gain insight into the role of education in the decline of American fertility, better information is needed for earlier phases of the transition. One state, Iowa, can provide the essential material on this relationship for the period before 1910. Prior to that time, Iowa was accelerating its pace of participation in the transition toward lower fertility. Child-woman ratios had been declining in Iowa virtually since the beginning of white settlement. As a state with a relatively late frontier period, its absolute decline in fertility was particularly rapid in the final decades of the nineteenth century and the first decade of the twentieth. The average of number of children ever born to ever-married white Iowa women dropped from 6.00 for those aged 70-74 in 1910 to 4.23 for those aged 40-44; only Missouri, Oregon, and Washington exceeded this magnitude of reduction over these three decades. The averages for all of the states in the country were 5.45 and 4.52 for these two age-groups.(8) For the five years before 1910, white Iowa women were very near the national average both in their level of marital fertility and in an index that measures the extent of birth control within marriage.(9)

During this era of Progressive reform nationally, Iowa was also pushing forward its vanguard position in the American educational transition. Not only was Iowa extending attendance (both years and days within years), educational leaders in the state were also attempting to substitute what they regarded as better schools - meaning centralized or consolidated ones, with trained and certified teachers - for poorer ones - one-room institutions, staffed by minimally-educated farmers' daughters, and controlled similarly by the common-school-educated adults of these small neighborhoods.

In the published volume for its 1915 census, the Iowa state government reported information for four educational variables for each of its 99 counties. Months of school attendance were tabulated for four lengths by sex for three age groups: 5-9, 10-17, and 18-20. For two age groups 5-9, and 10-21, information was published as to the class of school: common, private (mainly religious), high school, and college. Most innovatively, anticipating the education question on the national census by a quarter-century, the Iowa enumeration cards also requested information separately on years of schooling. The Iowa census card also divided the experience into separate queries on years spent in four different types of institutions: common schools, grammar schools, high or preparatory schools, and colleges.(10)

Discerning students of American historical statistics may guess that there must be something a bit odd about a place so compulsive in recording and reporting the educational status of its population. And there is. The state had attained a very high level of education, particularly at the basic level of schooling. But progressive Iowans, whose thinking is itself suggested by the novel questions and comprehensive published volume of its ambitious state census, were not content to leave well enough alone. In a period in which high-school attendance had not yet become usual in America, Iowa youth remained in school well into their teenage years. In Iowa in 1914, some 85.2% of those aged 10-17 were enrolled, and a fifth (20.3%) of those between 18 and 20 attended school.(11) In 1920, Iowa ranked only behind California in the percentage of its school-age children in high schools, an achievement that depended on having a very large number of high schools with very small enrollments.(12)

Educational reformers in Iowa in this period were not, however, centrally concerned with expanding schooling. In their opinion, a bigger problem was that there were too many schools. The 1915 census reports the astounding statistic of one ungraded (rural) school for every ninety of rural population. Thus the problem was not the absence of schooling, but that many rural Iowa children were not receiving what the reformers regarded as the right kind of education.(13)

In educationally-progressive Iowa, not all Iowans were "progressive." It was the last state outside of the former Confederacy to adopt, in 1902, a watered-down law mandating compulsory enrollment. While opponents of compulsory schooling included Catholics and other religious minorities, most important was the opposition in rural areas, whose one-room, ungraded, common schools were responsible for the high average level of basic education in the state.(14) The effect of the law, if any, was to force the enrollment of the tiny fraction (only several hundred in the view of its historian) not already in attendance.(15) What reformers sought, and the area in which they achieved considerable success in the period between 1913 and 1922, was the consolidation of schools in rural areas.(16)

In his foreword to a methodologically pioneering investigation of fertility variations among Iowa counties in 1915 by Hornell Hart,(17) Bird T. Baldwin, director of the Iowa Child Welfare Research Station of the University of Iowa, proclaimed that his office had as its goal to assist every child in developing to his or her "maximum ability consistent with its native endowment and special aptitudes."(18) These goals advanced beyond the more usual objectives of modern advocates of schooling of punctuality, thrift, self-denial, and so on.(19)

Anticipating the approach to fertility of the "new" home economists of a half-century later, Baldwin separated the issues of the number of children and their quality, and rhetorically queried the connection between education and fertility:

The problem of child welfare in Iowa is intimately bound up with the number and quality of children born into our Iowa homes. These two factors condition in a large measure the practical methods of child rearing.

This second study by Dr. Hart throws definite light on the general questions: What types of individuals in our State are becoming parents? Are more children born, proportionally, in the city than in the country? Are the larger families found among the native born, among the home owners, among those of average or superior school training?(20)

Baldwin and Hart asked relevant questions, and in a pointed, value-laden way. Just what effect did education have on the fertility of Iowans? Did "superior school training" of wives and husbands interpreted in terms of quality make a difference along with variations in its quantity? Of course, in contrast to their optimism about the capacity of social science to show definitive results, the last several decades of research have suggested that nothing works in education to alter later lives. At the least, scholars and policy makers now have severely chastened expectations about the magnitudes and import of quantitative results forthcoming from investigations of the effects of variations in schooling.(21)

Ecological Analyses:

Since schooling was collectively provided in Iowa, variation in educational attainment is partly a phenomenon existing at the aggregate rather than the individual level. It is also not implausible to believe that poorly-educated people in more highly-educated communities would share to some extent the values and aspirations arising from education. Education, like a variable such as job opportunities, can have causes and consequences at both aggregate and individual levels.

Using data for Iowa counties, this article first re-considers the ecological study of Hart that examined the effect of educational and other variables on marital fertility. In his overall assessment, Hart subsumed the effect of education on overall fertility (children under age 5 per 1,000 women aged 21-44) under a variable that captured the past differentials in fertility among counties.(22) Hart, who must be regarded as something of a hero of the statistical epoch of primitive cumulation, calculated up to seventh-order partial correlations between the child-woman index and ten independent variables, including education (three separate measures), religion (Catholic and non-Catholic per 1,000 population), nativity, urbanization, home ownership, and proportions married for the 99 counties of Iowa.

Until the very end of the monograph, Hart proceeds cogently, showing the extent to which the independent variables remain correlated when various controls are introduced. But he concludes with the seemingly parsimonious isolation of the three main predictors of variation in the child-woman ratio among Iowa counties: the proportion married, urbanization, and his age-distribution index. This statistically powerful demographic index - the ratio of the number of women aged 19-20 to the number over age 45 - is a measure of the steepness of the slope of the age distribution. Demographers now know that this slope is overwhelmingly determined by the history of past fertility.(23) Hart's explanation for the predictive capacity of this age-structure indicator was cast in terms of it being "an index of misery," a proxy for poor people:

Poor people in one generation tend to be the parents of the poor in the next. They also tend to be tenant farmers and to have low levels of education. Foreign-born immigrants are poor, and migrate into poor districts. . . (many of the correlations) express the same fact - namely that the poor and ill-educated - the unsuccessful in a word - are the highly fecund class.(24)

Hart's interpretation can hardly be said to be supported by the empirical analysis that he offered. No indicator measured the incidence of poverty or the extent of its heritable tendency. Since his argument was about the behavior of individuals as individuals rather than individuals embedded in particular places, one cannot save his interpretation by invoking the plausibility of a particularistic explanation resting on the distinctive cultural-historical legacies that differentiate locales.(25) Statistically, all Hart actually showed was that past fertility (the slope of the age-distribution) was the "cause" of more recent fertility measured by another part of the age-distribution. Thus, it is not at all surprising that his index was a better predictor than the more indirect variables such as tenancy, foreign-birth, and educational involvement.

In my analysis, four indicators related to economic and cultural patterns account for just over half of the variation among counties in the marital fertility index ([ILLUSTRATION FOR FIGURE 1 OMITTED] and Table 1). Counties with more Catholics and Lutherans among religious adherents were particularly likely to have higher marital fertility while counties with a larger fraction living in incorporated areas had lower fertility. However, the relationship between a concentration of adherents to Lutheran religion and marital fertility was quite weak. These denominational variables capture, in a culturally-sensitive way, the legacy of the later pattern of fertility decline in Europe compared to in America.(26) Irrelevant to the outcome were other indicators of religious differentiation among Protestant groups (Methodists, or the denominations with New England origins).

Finally, at this late date in American agricultural settlement history, an indicator - the percentage increase in improved acres between 1890 and 1910 - of the effect of the "frontier" in making it easier for parents to provide inheritances to children was positively related to marital fertility.(27) Another agricultural indicator - the average value of land per farm - also had a positive relationship with marital fertility. One historian has conjectured that there should be a negative relationship between farm values and fertility, on the grounds that more expensive farms are less "available" to prospective farmers.(28) In conjunction with the "frontier" variable, it here can be interpreted more plausibly as a positive effect of greater parental assets on fertility. Wealthier farm counties could thus afford to have more children than poorer counties.

Controlling for these other variables, two of the educational indicators are negatively and significantly related to marital fertility: the fraction of adults with more than eight years of schooling and the enrollment rate of 16-17 year-olds in 1920. The former is more easily interpreted than the latter, as increased enrollment may be an effect of lower fertility rather than a cause. Since education was completed before marriage and childbearing began, it has, one may presume, an uni-directional impact on fertility. Baldwin correctly inferred from Hart's analysis of overall fertility that those counties with a "superior" education [TABULAR DATA FOR TABLE 1 OMITTED] quantitatively had lower fertility within marriage than those with only average educational training. Furthermore, Hart also found that the measure of past education for adults was more strongly related to fertility than was the incidence of current high-school attendance.(29)

Individual-Level Results:

In addition to these important aggregate patterns, schooling is an experience of individuals. The major empirical contribution of this paper depends on an individual-level data set that contains virtually all relevant variables for isolating the distinctive impact of the quantity and quality of education on fertility behavior.

DATA AND SAMPLE: Some 8,503 ever-married women were sampled from the 1910 federal manuscript census, which contained questions on the number of children ever born (CEB) and the order and duration of the current marriage on its extensive schedule. Sampling units included four of the eight cities with populations over 25,000 (Des Moines, Dubuque, Cedar Rapids, and Council Bluffs) and 10 rural areas (counties or the rural part of urban counties) that were selected on a stratified regional basis. In the rural sample were the counties of Boone, Butler, Cerro Gordo, Henry, Lee, Louisa, Marshall, and Story as well as the non-urban areas of Polk and Dubuque Counties.(30) In general, the characteristics of the urban and rural samples differ only marginally from those of the Iowa cases in the national public use sample.

The cards for some 4,730 or 56% of women sampled in 1910 were found in the 1915 microfilm. Although there was no statistically significant difference in the indicators of fertility between linked and unlinked cases, the differential in linkage rates was considerably more pronounced in its biasing effect than the modest departures of the 1910 stratified sample from the national public use sample of that year.

VARIATION IN SCHOOLING: Comparison across cohorts reveals the magnitude of the ongoing advance in the extension of schooling in Iowa. More than a third of the women over age 65 in 1910 (i.e., born before 1845) had gone to school for seven or fewer years, and only one in twelve had any high school experience. Of those under 25 at the enumeration in 1910 (born after 1885), fewer than one in five (18.3%) had less than eight years of schooling and more than one-fourth (26.3%) had some high school education. A modest discontinuity in the series appears for females born in the decade after the Civil War. Trends and levels of education for the husbands of women who were still married in 1915 were similar to those of their spouses. The modal length of common and grammar school was eight years, but nearly a quarter of the women reported nine or more years at the common or grammar level.(31)

Beyond its increase over time, schooling varied with the other delimiters of status in early twentieth-century Iowa. Given the novelty of these data for this time period, Table 2 reports, in considerable detail, the relationship of patterns of schooling to other characteristics. In virtually every group, going to school eight years was typical; such was the median for nearly every segment of the ever-married female population.

Hart was correct in associating the least-educated with the foreign-born and the less affluent, but not with tenant farmers, whose relative youth was associated with more schooling. Wives of both farm owners and tenants resembled in educational attainment married women in the working classes. Wives of men in the middle classes had somewhat more education, but there was little differentiation by education for wives of men whose income was in the lower three-fourths of the distribution. Nor did education vary among those whose homes or farms were valued below the top two-thirds of the sample. Wives in families in the highest-income quarter and wealthiest third were more educated than those below them.

In terms of cultural background, women of southern birth or origin had less schooling on average. Those who were from New England or were a member of a religious denomination whose origins were in that region(32) were among the best educated. Implied in the figures for the foreign-stock (classified by the birthplace of the sampled woman's mother), second-generation women, especially from Britain and Ireland, narrowed the educational gap with native-born women of native parentage. Catholic and Lutheran women were virtually identical in education, having attended two years less on average than Yankee Protestants and one year less than women in other Protestant denominations.

FERTILITY MEASURES: To focus on marital fertility, the cases selected for analysis were further limited to those women who reported on duration of their first marriage as of 1910 and whose husbands were also found in 1915. The former restrictions were needed to construct fertility indices that take into account the varying exposure to the risk of childbearing by age and marital duration.(33) In the numerator of the two fertility indicators are the number of children ever born (CEB) and the number of "own-children" under age five in the household in 1910. The denominators of the rates are the number of children that would be expected if women followed the standard age-pattern of natural fertility over their period of exposure.(34) For convenient interpretation, this ratio is then multiplied by 11.05, the number of children expected by a woman who married at age 15 and bore children at the rates specified in the natural fertility model until age 50.

Calculating fertility rates in this way has several advantages. Most important, it translates observed behavior into a demographically meaningful standard. In terms of fertility, what matters is the potential inherent in a woman's age and exposure to the risk of childbearing that is heightened as a consequence of being married. Measured in natural-fertility-equivalent-years, two children born between ages 38 and 42 represent a higher fertility rate than two children born between ages 23 and 27. Second, it allows more of the cases to enter the analysis, as there is no need to restrict the sample to those who have been married for at least five years.(35) Third, it fails to control for the relationship between age and fertility but flunks in a way that is demographically and historically interpretable. Since family limitation was being practiced in Iowa, the expectation based on the schedule of natural fertility does not eliminate the effect of age. Indeed, a nearly linear relationship between age and children-ever-born is curvilinear in the transformed index. Controlled fertility departs more from the standard schedule in the later ages of childbearing than in the earlier. Thus the index declines for women between 15-19 and 40-44. However, since fertility reduction had been underway for decades, the index increases for women of completed fertility.

[TABULAR DATA FOR TABLE 2 OMITTED]

The trend in the age-pattern of recent fertility (based on the number of living children under age five) is exactly that of a population practicing what demographers identify as parity-specific control. The index, which has not been augmented to correct for mortality in infancy and early childhood, is nearly two-thirds (64%) of that expected for the younger ages (15-24) of childbearing. By the later ages of childbearing (35-44), the ratio is just over a third of that expected from the natural fertility schedule.

Schooling and Fertility:

BIVARIATE RESULTS: Table 3 reports the variation in the indices of childbearing by years of schooling and a quality of education measure that makes use of the important distinction between training that took place in common (rural, ungraded) schools and grammar (graded) schools. Education was inversely related to fertility. Women with seven or fewer years of schooling had a CEB-index of 8.45 41% higher than those with 12 or more years of schooling (6.00 implied [TABULAR DATA FOR TABLE 3 OMITTED] children). Women with only a common school education had 44% higher fertility (TMFR = 7.87, based on CEB) than those with high-school experience or more (5.48 implied children).(36) By the comparative standards of recent experience in the Third World, the educational differential in fertility among Iowa women was large. A gap in the total marital fertility rate of three children implies, given an average marriage age in the low 20s, a difference in completed family size among Iowa women of about two children.(37)

The small group of women with the most schooling, 16 or more years, apparently had higher fertility than the better educated generally. This departure from the linear pattern, however, can mostly be attributed to outliers.(38) In the index of recent fertility that is based on the number of own-children under age five, the women with the least education (four or fewer years) actually had lower fertility than those with some high school or college education. This anomaly may possibly explained by higher mortality among those with the least schooling or by a greater tendency of the young children of the least educated to live apart from their mothers.

Table 4 displays variations in the two indices of fertility by the duration and type of education and several of the other potentially important determinants of fertility (see the unadjusted columns). The pattern of variation is nearly identical for the CEB measure and the index computed from the number of own children aged 0-4 living in the household. For example, wives of tenants had higher fertility than wives of farm owners on both indicators; similarly, wives with eight or more years of grammar school experience had lower fertility than wives who had only attended common school. The absolute magnitude of the unadjusted differences is also very similar in both instances, although there is more divergence in percentage terms because of the smaller overall average for the second index.

It was not merely differences in the extent of schooling that structured the rate of childbearing within marriage. Indeed, virtually every dimension of social structure was related to variations in marital fertility. Among occupational categories, farmers and especially farm tenants had higher fertility, while women with husbands in the middle classes exhibited the lowest fertility. The CEB-index for wives of farm tenants (8.58) was 49% higher than that for spouses of middle-class husbands (5.75). In the center of the distribution were wives of men in the working classes.

Given these results for farm wives, it is not surprising that women in rural places, defined as under 2,500 in population in 1915, had 26% higher fertility (7.75 children) than women in towns (6.21) or in cities (6.14). There was no discernible difference in fertility between the towns, which ranged in population from 2,686 to 17,152, and the four larger cities in the sample.

As was the case for occupation, a wide gap in marital fertility was obtained for Iowa women of different religions. Among denominations, Catholic (8.85) and Lutheran (8.61) women had substantially higher fertility than average, while women affiliated with denominations whose institutional origin was most distinctly that of New England (Unitarians, Universalists, Congregationalists, and Presbyterians) had the lowest fertility (5.75). In the nineteenth century, Yankees, descendants of the Puritans in New England and elsewhere in the [TABULAR DATA FOR TABLE 4 OMITTED] United States, were the vanguard group in the adoption of family limitation. The level of fertility for Catholic women was 54% higher than for those in the New England denominations.

Women whose mothers were born in Europe had the highest fertility among nativity groups. Those whose mothers were born in the northeastern United States (the New England and middle-Atlantic states) had the lowest fertility (6.25), although their recent childbearing experience was not too much below that of women whose mothers were born in Britain. The largest differential was 36%.(39)

To summarize by adding these patterns: In early twentieth-century Iowa, poorly-educated Catholic and Lutheran women whose mothers were born in Europe and who were married to farm tenants had the highest fertility. Conversely, well-educated, urban women whose mothers were born in the Northeast, who were members of Yankee religious denominations, and who had husbands in middle-class occupations had the lowest fertility.

MULTIPLE CLASSIFICATION ANALYSES: Because fertility was influenced by a multiplicity of characteristics, inferences from these simple bivariate comparisons must, in principle, be limited. However, the adjusted results reported in Table 4 show what would be expected when several independent variables are only modestly interrelated. While the extreme categories converge toward the mean, substantial differences persist after introducing controls.

For the CEB index, for example, a difference of 2.8 children in the implied total marital fertility rate between middle-class and farm tenants in the raw data narrows to 2.3 children when other factors are taken into account. A 3.1 child difference between Catholics and those in Yankee denominations closes to 2.3, and a 1.6 child gap between towns or cities and the countryside shrinks to 0.6.

For education, what these persistent differences in Table 4 mean is that schooling does not mediate entirely between the other sources of social status or economic position in Iowa and fertility. The higher fertility of farm women, for example, cannot be entirely attributed to their shorter duration of education in common schools. Unlike the case made by Hart for his tautological indicator of age-structure, there is no "magic" variable under which all of the others can be subsumed. Instead, we see a diversity of factors related to differential fertility. Since the statistical method in Table 4 of estimation assumes independence among factors, the reader can add across categories. For example, at one extreme, an old (over age 65), Catholic rural woman who had seven or fewer years of common schooling and who was married to a farm tenant is estimated to have an implied total marital fertility rate of 13.56 (7.22 + 2.89 + 1.47 + 0.20 + 0.69 + 1.09). At the other extreme, the rate for a 35-44 year old "New England Protestant" city woman with 9-12 years of schooling in grammar or high school and married to a middle-class husband was 2.93 (7.22 - 0.69 - 0.87 - 0.65 - 0.89 - 1.19).

Hart and Baldwin were partly correct. Both the extent and quality of schooling played definite, independent roles in the overall structure of marital fertility differentials among individual women in early twentieth-century Iowa. Treating duration of schooling as an interval variable yields the estimate that an additional year, controlling for type (common only versus some grammar) as well as the other factors in Table 4, reduces the implied total marital fertility rate by 0.08 children. In addition, having any experience in grammar-school education or beyond (high school or college) further reduces the rate by 0.75 children. Controlling for other factors, including religion, nativity and occupational group, a woman with exactly eight years of common schooling would have a slightly more than a one-child lead (1.07) in her TMFR over a woman with 12 years of education and some experience beyond the common schools.

Did women have lower fertility because of what they "learned" in school? To be sure, Iowans were not instructed in methods of family limitation in any type of school in this period. Rather, a goal of education in the modern sector of schools was to produce "modern" individuals. Even the common schools aimed at more than simply the 3 R's - seeking as well to produce rational, responsible individuals. Obviously the statistical results cannot tell us whether it was the kind of school and the extent of schooling that made a difference in fertility behavior or whether, despite the inclusion of other major determinants of fertility in the analysis, it was some other unmeasured factor or factors associated with variation in education. However, several additional results are of considerable relevance and suggestive interest. First, the educational indicators do not appear to be a proxy for social status that is not captured by the occupational and residential indicators. The education of the wife mattered more than that of her husband.(40) Summing the schooling of husband and wife and including the husband's advantage over the wife in schooling as an additional variable does not improve the fit of the model to the data. Combined education of spouses reduced fertility, while the husband's advantage over the wife increased it; including both variables in the analysis left each less significant statistically.(41)

Second, the role of education in structuring fertility appears stable over time as reflected in the nearly five decades of past and current childbearing experience measured by the six age groups of women. In technical language, there was no interaction between the educational variables and age group. Schooling had the same implications for fertility of women born before 1850 (those over age 65 in 1910) and those born after 1885 (under 25 in 1910). Nor did the educational indicator interact with any other of the variables included in Table 4. Education, for example, did not affect Catholic women differently from Protestant. The duration and type of schooling can thus be regarded as part of a stable process over this time span in the history of fertility decline in Iowa.

One may plausibly regard an educational history limited to experience in the common school as a rough index of rural socialization and experience in grammar school or beyond as a measure of some contact in childhood with town or urban life. The boundary between town and country in early twentieth-century Iowa was permeable in both directions. Some 40% of women living in places above 2,500 in population had an education that did not go beyond the common school, and a full 26% of rural women had some experience in grammar schools (weighted sample, bottom panel of Table 5).

For cohorts of non-farm white women born in the first two, and possibly three, decades of the twentieth century, the educational differential in fertility was quite modest for non-farm women who were not farm-born; this pattern seemingly did not continue for cohorts of childbearing age during the Baby Boom of the 1940s and 1950s. The overall differential was due to the migration of less-educated women from the farm.(42) Although the measures employed in this article differ [TABULAR DATA FOR TABLE 5 OMITTED] from studies treating what is called the "two-generation urbanite hypothesis," the results here suggest that the weak effect of education on the fertility of non-farm women with non-farm origins characterized only a brief interlude in the history of American fertility. It appears to be merely a distinctive peculiarity of the very end of the long secular transition in fertility.

As Table 5 shows, both place of socialization and place of current residence affected fertility. Urban women educated in the common schools had lower fertility than city women with grammar-school experience and rural women with common-school training. Rural women educated in or beyond grammar school had lower fertility than rural women with only common schooling.

Schooling of higher quality or longer duration was not necessary for lower fertility. Common-school-educated (rurally socialized) women who moved to the city adapted substantially to the lower fertility pattern there, while urban-socialized women who moved to the countryside carried most of their lower fertility pattern with them. In an era in which fertility was declining rapidly, combinations - rural residence and poor education - are needed to sustain substantially higher fertility. In the countryside and in towns or cities, more years of schooling of either type reduced marital fertility. These differences are particularly large for the analysis that only uses age as an additional control variable.

Something happened to girls who went to school that helped them, as married women, in setting or attaining lower fertility goals. Despite the skepticism among historians about the impact of education and the programs of educational reformers, and a nostalgia by scholars of disparate persuasions for the fabled rural common school, the fertility behavior of Iowa women was altered by their contact with longer and/or a more modern form of schooling. Whether by leading them to desire fewer children or making them more careful users of birth control, those women who enjoyed more and better schooling were able more effectively to limit their rate of childbearing. In Iowa, at least, education worked.

Department of History Chicago, IL 60607

ENDNOTES

Acknowledgements: The research in this article was supported by Grant R01 HD23824 from the National Institute of Child Health and Human Development and was assisted by Ellen Caldwell and N. Sue Weiler. Earlier versions were presented at the 1992 Denver, Colorado, meetings of the Population Association of America, and at the Newberry Library Seminar in Rural History in 1993. The author thanks those who provided comments on both occasions but takes responsibility for the errors and flaws that remain.

1. Hornell Norris Hart, "Differential Fecundity in Iowa: A Study in Partial Correlation," University of Iowa Studies in Child Welfare, II, number 2, 1923 (New York, 1969).

2. Of course, it is vastly more complicated to isolate and assess the broader outcomes of a system of education on the views or behavior of the adults in a society, as the brief comments in Lawrence A. Cremin, American Education: The Metropolitan Experience, 1876-1980 (New York, 1988), 660-64, suggest.

3. Joel Perlmann, Ethnic Differences: Schooling and Social Structure among the Irish, Italians, Jews and Blacks in an American City, 1880-1935 (New York, 1988), x, cites the fourteen years of research involved in such a record-linkage project. Census questions and tabulations related to education through 1960 are conveniently compiled in John K. Folger and Charles B. Nam, Education of the American Population (Washington, 1967), 253-76.

4. Susan H. Cochrane, Fertility and Education: What Do We Really Know? (Baltimore, 1979). For a summary of results from the World Fertility Survey, see Susheela Singh and John Casterline, "The Socio-Economic Determinants of Fertility," in Reproductive Change in Developing Countries, eds. John Cleland and John Hobcraft (Oxford, 1985), 201-10. For a more cautious assessment about the strength of the education-fertility relationship, see Ronald Freedman, "Fertility Determinants," in The World Fertility Survey: An Assessment, eds. John Cleland and Chris Scott (Oxford, 1987), 783-87.

5. Susan H. Cochrane, "Effects of Education and Urbanization on Fertility," in Determinants of Fertility in Developing Countries, eds. Rodolfo A. Bulatao and Ronald D. Lee (New York, 1983), II, 587-626. Jee Peng Tan and Michael Haines, "Schooling and Demand for Children: Historical Perspectives," World Bank Staff Working Papers No. 697 (1984).

6. John C. Caldwell, "Mass Education as a Determinant of the Timing of Fertility Decline," Population and Development Review 6 (June 1980): 225-255.

7. "Program of Action of the 1994 International Conference on Population and Development," chapter XI, reprinted in Population and Development Review, 21 (June 1995): 444-447.

8. Avery M. Guest and Stewart E. Tolnay, "Children's Roles and Fertility: Late Nineteenth-Century United States," Social Science History 7 (Fall 1983): Table 1, p. 366.

9. Women in populations in which fertility within marriage is not consciously controlled vary substantially in the rate of childbearing. Compared to the pattern of natural fertility, the pace of childbearing in populations practicing modern birth control is reduced more markedly at older than younger ages within the childbearing span; women bear their last child at earlier ages than 40 which is characteristic of natural fertility. Thus, demographers have found it useful to divide the curve of marital fertility by age into two components: its level, indexed by "M", and its shape, indexed by "m". In Iowa, M was .73, compared to .72 for the United States, while m was .47 and .54 respectively. Douglas Ewbank, "The Marital Fertility of American Whites before 1920," Historical Methods 24 (Fall 1991): Table 2, p. 144.

10. State of Iowa, Executive Council, Census of Iowa for the Year 1915 (Des Moines, 1916).

11. Ibid., xix, xlii.

12. David R. Reynolds, "What Price Community: Some Observations on the Political Economy and Geography of the Rural School Consolidation Movement in Iowa, 1895-1925" (Unpublished Paper Presented at the Newberry Library Seminar in Rural History, 1991).

13. Ibid; Richard Jensen and Mark Friedberger, "Education and Social Structure: An Historical Study of Iowa, 1870-1930," (Newberry Library, Chicago, Report to the National Institute of Education, 1976.)

14. As was the case generally in the Midwest. Wayne E. Fuller, The Old Country School: The Story of Rural Education in the Middle West (Chicago, 1982), 233-39.

15. Carroll Engelhardt, "Compulsory Education in Iowa, 1872-1919," Annals of Iowa 49 (Summer-Fall 1987): 58-76.

16. Reynolds, "What Price Community."

17. Hart, "Differential Fecundity in Iowa."

18. Bird T. Baldwin, "Foreword," University of Iowa Studies in Child Welfare, II, number 2, 1923 (New York, 1969), 5. On the reform impulse behind the work of Baldwin and colleagues, see Hamilton Cravens, "Child-Saving in the Age of Professionalism, 1915-1930," in American Childhood: A Research Guide and Historical Handbook, eds. Joseph M. Hawes and N. Ray Hiner (Westport, CT., 1987), 415-88.

19. Carroll Engelhardt, "Schools and Character: Educational Reform and Industrial Virtue in Iowa, 1890-1930," Annals of Iowa 47 (Winter 1985): 618-36. Keach Johnson, "The State of Elementary and Secondary Education in Iowa in 1900," Annals of Iowa 49 (Summer-Fall 1987): 26-57. Idem., "Roots of Modernization: Educational Reform in Iowa at the Turn of the Century," Annals of Iowa 50 (Spring 1991): 892-918.

20. Baldwin, "Foreword," 5.

21. Already, by the period from World War I to the early 1920s, educational consolidators were disappointed by the limited (though usually present) advantage on knowledge tests that students in graded grammar schools had over pupils in rural one-room schools. See Fuller, Old Country School, 240-45. However, using current procedures, these limited results are probably exaggerated as family background of urban and rural students was not controlled. For a summary of evidence for the "almost nothing works" perspective in education, see Eric A. Hanushek, "The Economics of Schooling: Production and Efficiency in Public Schools," Journal of Economic Literature 24 (September 1986): 1141-77.

22. While Hart used the overall child-woman ratio as the dependent variable, an index of marriage was included as an important independent variable. Consequently, the other independent variables tapped variations in marital fertility. A separate analysis, not reported in this article, showed that educational indicators did not contribute to the statistical explanation of variation in the index of marriage among Iowa counties.

23. Ansley J. Coale, "How a Population Ages or Grows Younger," in Population: The Vital Revolution, ed. Ronald Freedman (New York, 1964), 47-58.

24. Hart, "Differential Fecundity in Iowa," 34-35.

25. John A. Agnew, Place and Politics: The Geographic Mediation of State and Society (Boston, 1987).

26. These religious variables were more effective predictors than the proportion foreign-born in 1915 or the proportion German-stock and foreign non-German stock in 1930.

27. Yasukichi Yasuba, Birth Rates of the White Population in the United States, 1800-1860: An Economic Study (Baltimore, 1961); Richard Easterlin, Population Change and Farm Settlement in the Northern United States." Journal of Economic History 34 (March 1976): 45-65.

28. Maris A. Vinovskis, "Socio-Economic Determinants of Interstate Fertility Differentials in the United States in 1850 and 1860," Journal of Interdisciplinary History 6 (Winter 1976): 375-96.

29. The 1920 data for attendance of 16-17 year-olds were used because Hart's denominator included the broad age group between 10 and 17. Since the slope of the age distribution captures the past history of fertility, and since school attendance declines within this age group, his indicator seemed too ambiguous to interpret. Using a more appropriate measure of the level of education - the proportion of adults with college, high school or eight years of grammar or common schooling - results, however, in only a slightly higher regression coefficient and barely increases the overall variance explained by the equation.

30. The first sample was drawn from the 1910 enumeration rather than from the 1915 census because the individual cards from the latter were filed and later microfilmed in alphabetical order by surname for each county and each city over 25,000.

31. In the decades preceding 1910, however, schooling was becoming somewhat more effective in the sense that fewer years were being spent to attain a given level of education. More than a fourth of women aged 55 and older who did not go to high school had attended nine or more years compared to 18 per cent of similarly educated women under age 25.

32. Congregationalists, Presbyterians, Unitarians, and Universalists are the denominations that are grouped into this category.

33. Bryan Boulier and Mark R. Rosenzweig, "Age, Biological Factors, and Socioeconomic Determinants of Fertility: A New Measure of Cumulative Fertility for Use in the Socioeconomic Analysis of Family Size," Demography 15 (November 1978): 487-97.

34. For example, the denominator of the fertility index for a 32-year old woman who had been married 11 years is 4.98[(4.0 x 0.460) + (5.0 x 0.431) + (2.5 x .395)], taking into account respectively her years of experience in the 20-24, 25-29 and 30-34 age groups weighted by the rate of natural fertility. If she was the mother of three children, her CEB fertility index was 0.60 (3/4.98) times 11.05 or 6.63. For the "natural" marital fertility rates by age, see Ansley J. Coale and T. James Trussell, "Technical Note: Finding the two Parameters that Specify a Model Schedule of Marital Fertility," Population Index 44 (April 1978): 205.

35. However, it is possible, as will be noted below, for mothers married only one or two years to have artificially high values on the indicators. Extreme outliers were dampened by making a rate four times that of natural fertility (44.2 children) the maximum. Remaining extreme values, which can noticeably effect the average in a small group, can be suppressed by limiting the sample to women married five or more years or by weighting the sample to reflect the fertility-relevant duration of marriage (the number of years married converted to children expected on the basis of natural fertility) of each woman.

36. The implied total marital fertility based on the number of own-children under age five is substantially lower than the CEB index for four reasons. First, the former does not take into account children who have died or, presumably of lesser importance, who were not living with their mothers in 1910. Second, the CEB index incorporates the higher levels of fertility of the more distant past, in the experience of women over the age of 50. The own-child measure only uses the experience of women of childbearing age in the past five years and was not calculated for women over age 50 in 1910. Finally, for women under age 50 who had been married for more than five years, the implied number of children ever born is higher than that implied by fertility in the past five years because conscious family limitation concentrates births in the earlier years of marriage.

37. This contrast neglects the probable fact that better-educated women in Iowa married later than their less well-educated sisters. The larger Third World differentials in completed marital fertility between those with no schooling and those who finished secondary schooling are in the 2-3 child range. They are found in Latin America and the Caribbean, Jordan and Turkey, and Thailand, Malaysia and South Korea. In the remainder of Asia and in Africa, this difference tends to be under one child. Singh and Casterline, "Socio-Economic Determinants," 202.

38. Weighting by the fertility-relevant duration of marriage reduces the index for those with 16 or more years of schooling to 5.6, compared to 6.7 in overall sample and 5.4 for those with 12-15 years of education. Limiting the sample to those married five or more years produces estimates of 6.4, 7.1, and 5.5 children respectively.

39. Age should be regarded as only a control variable.

40. A result also found in recent studies of fertility in the Third World. Singh and Casterline, "Socio-Economic Determinants," 203-06.

41. An additional year of parental education lowered fertility by .041 children, while a year of husband's advantage increased it by 0.77. The former indicator was significant at .063 and the latter at .027. An alternate formulation that used the years of schooling for the wife as a variable as well as the difference between the schooling of the spouses also indicates that the advantage a husband had over his wife in years of schooling slightly increased her fertility. However, the difference in schooling was not statistically significant (F=0.89, sig.=0.345).

42. David Goldberg, "The Fertility of Two-Generation Urbanites," Population Studies 12 (March 1959): 214-22. Otis Dudley Duncan, "Farm Background and Differential Fertility," Demography 2 (1965): 240-49. P. Neal Ritchey and C. Shannon Stokes, "Residence Background, Socioeconomic Status, and Fertility," Demography 8 (August 1971): 369-77. Nancy J. McGirr and Charles Hirschman, "The Two Generation Urbanite Hypothesis Revisited," Demography 16 (February 1979): 27-35.