Read an Excerpt
Family and Population in Nineteenth-Century America
By Tamara K. Hareven, Maris A. Vinovskis
PRINCETON UNIVERSITY PRESSCopyright © 1978 The Center for Advanced Study in the Behavioral Sciences, Stanford, California
All rights reserved.
Farms and Farm Families in Old and New Areas: The Northern States in 1860
RICHARD A. EASTERLIN, GEORGE ALTER, AND GRETCHEN A. CONDRAN
This essay exploits a remarkably rich data set to compare demographic and economic conditions in old and new farming areas in the northern United States in 1860 and assesses the implications of these conditions for the causes of farm family fertility differences between these areas. The analysis centers on a sample comprising 11,492 farm households, and this essay presents the results of the initial analysis of this sample.
Part I addresses such questions as the following. Are there relatively more husband-wife households on the frontier? What is the sex and age distribution of the population? Are young frontier women more likely to be married and to have higher marital fertility? How large is the number of persons per household — children, adults, and aged? Are nonnuclear family members in the household more or less frequent? How do schooling of children and adult literacy compare in new and old regions? What are the geographic origins of the residents of new versus old areas? What are the occupations of family members? How large are farms and what are they worth? How do farms compare in their use of labor and capital?
Part II considers some tentative implications of this analysis for fertility. Is higher fertility in new areas a result of "compositional influences" — special features of the population by, say, marital status, age, or ethnic origin — which raise measured fertility in new compared with old areas? Do statistical biases exaggerate fertility in new versus old areas? Or are there substantive factors, such as a lower burden of aged dependents or greater farm labor requirements, that encourage higher fertility behavior within farm families in new areas?
An essential first step in research is to construct a representative picture of the subject to be studied — guided, of course, by some tentative hypotheses, and with a formal model as the eventual goal. The analytical problem toward which this work is ultimately directed is the source of the dramatic decline in the fertility of American farm families in the nineteenth century. Some important clues to this may be discovered by investigating the causes of higher fertility in new versus long-settled agricultural areas. Our expectation is that the change in fertility behavior is bound up chiefly with changes in the economic circumstances of families as an area undergoes settlement. But before trying to develop a formal model, we want better to inform ourselves on the typical circumstances of these families, so that hypotheses can be framed as knowledgeably as possible. Part I thus addresses factual questions regarding household size and structure, fertility, migration, and farming characteristics. This section should also be useful to scholars interested in topics other than fertility behavior. In recent years, for example, there has been considerable work by social historians on family and household structure. Among economic historians, outputs, inputs, and farm productivity are a continuing concern. Although our analysis in Part I will not fully satisfy the needs of researchers in these areas, it should prove helpful. Following this, we examine in Part II the implications of our findings for the explanation of higher farm fertility in new versus old areas.
I Descriptive Analysis
DATA AND PROCEDURES
The sample. Our approach is what historians have come to call "collective biography." The data comprise a sample of rural households taken from the 1860 manuscript censuses by Fred Bateman and James D. Foust. The sample covers all households living on farms or in rural villages in each of 102 townships scattered across 16 northern states. Households living in towns or cities in these townships are excluded. In total there are 20,664 households in the sample, almost all of which are white. Table 1-1 shows the distribution of sample households by place of residence and occupation of head. Over half of the households' heads — those on whom the present essay focuses — live and work on farms. Those identifying themselves as farmers or tenants (line 3 — there are actually very few who report themselves as tenants) certainly work on farms, but the impression gained from perusal of some of the census schedules suggests that most laborers in the sample probably work on farms too. Probably this is because the nonfarm segment of the sample was confined to rural villages.
The data for each household included in the sample comprise almost all information on the "free population" census schedule. In addition, for each household living on a farm, data were obtained from the agricultural census. The availability for farm households of these matched demographic and economic data is a unique attraction of this sample.
Maryland and Missouri were slave states in 1860, and the returns from the slave schedules for the sample townships in these states were not yet available for this analysis. This omission does not seem serious for our purpose. The sample includes only one township in Maryland. Although there are seven sample townships in Missouri, the statewide ratio of slave to free population is only a little greater than one to ten. Since the present analysis relates to characteristics of white households, the chief problem arising from the omission is that of obtaining an adequate estimate of labor input on farms. The present findings are probably not much affected by this omission.
The population census collected no information on marital status and household relationship. A problem of some importance for the present analysis is the failure in transcribing the basic data to use surnames on the population census schedules to estimate family relationship. Because of this, we developed rules (based chiefly on the age of an individual and his position in the enumeration of a household) to establish type of household (whether husband-wife, other male head, or female head) and nuclear versus nonnuclear household members. These rules and some tests of them are presented in Appendix A. Appendix B deals with sample size. It gives the number of cases in each cell for those text tables where such information is not otherwise given.
Method. The townships, 102 in all, were classified into five groups according to degree of settlement. The "settlement classes," consisting of quintiles from 0 to 100 percent, were based on a calculation for each county containing a sample township of the percentage of ever-improved agricultural land not improved in 1860. Thus townships in counties where, say, in 1860, 95 percent of the agricultural land ever-improved was not yet under cultivation fell in the "newest" quintile (80-100 percent), while those with a 5 percent ratio in 1860 fell in the "oldest" quintile (0-19.9 percent). Table 1-2 shows the distribution by state of the number of counties and households in each quintile. In general, townships in the newest settlement classes are located in the newest states. However, there are a few townships in newer states in which settlement is fairly far along in 1860. Because of this, the newness of the state in which a township is located is not a reliable measure of a township's degree of agricultural settlement. This is why we have used a stage of settlement measure based on the ratio of land actually cultivated to that eventually cultivated. The measure was computed for the county in which a township was located rather than for the township itself, because data at the township level were not available.
Table 1-3 gives an approximate idea of the date of settlement of the counties in which the sample townships were located. The approximation is based on the date a country is first listed in the census, and is misleading in the case of a county that is formed by subdivision of a larger county long after settlement. Despite this problem, for which no correction was attempted here, there is a fairly clear pattern — counties in newer settlement classes tend to be established at later dates. The biggest exceptions are the counties in settlement classes III and IV, which show fairly similar dates of settlement.
All households in a township were assigned to the same settlement quintile as the county in which they were located. The data for all of the households in the townships falling within a quintile were then combined to obtain measures of the characteristics for that settlement class, such as household size, farm size, and so forth. The present analysis is based almost wholly on the arithmetic means so calculated and centers on households rather than individuals as the basic unit of study. In time, this analysis will be supplemented with measures of dispersion and will be extended with pertinent data for individuals. In what follows, we take up, in succession, household size and structure, fertility, migration, and farming characteristics.
Type and age distribution of households. In all regions the great proportion of households — four-fifths or more — were husband-wife households in which both spouses were present (Table 1-4). Most of our subsequent analysis will center on these households. About two-thirds of the remaining households were headed by a male and one-third by a female. The various settlement classes show quite similar distributions, except for the newest area where the proportion of male-headed households is somewhat higher and that of husband-wife households correspondingly lower.
While the type of farm household is similar among regions, there is a systematic difference in age characteristics. As shown by the age distribution of husband-wife households in Table 1-4, the newer the region, the younger the average household. Over half (56 percent) of the wives in the newest region are under 40, compared with about four-tenths in the oldest settlement. The figures for mean age of head show a similar pattern (Table 1-5). The term "conspicuously middle-aged," which Malin uses to characterize Kansas frontier households in 1860, does not seem appropriate (though our difference may rest largely on the meaning of "middle-aged").
Table 1-5 shows that in all regions the heads in non-husband-wife households are older than the average. This is probably because most of these households are former husband-wife households in which one of the spouses has died. The preponderance of male over female heads in non-husband-wife households should not be taken to indicate higher female mortality. It is probably because women heads left with a farm are more likely to remarry, pass the responsibilities of headship on to a son or son-in-law, or sell out and move in with relatives. The decline across regions in age of head in these non-husband-wife households probably reflects chiefly the differing age distributions of the population. In the newest region the somewhat larger proportion of male-headed households and noticeably younger age of head suggest that in this region there may be a disproportionate number of young male heads who have established farms either before marrying or while their wives are "back East."
Size of household. On a priori grounds one might expect that if frontier fertility were higher, household size in new areas would be larger than in old, by giving rise to a larger number of young persons per household. However, this might be offset in some degree by a greater number of adults per household in older areas. Table 1-6 presents the relevant data.
To minimize the effect of age distribution in regional comparisons, let us compare the situation for husband-wife households in which the wife is in the same age bracket — 20 to 29, then 30 to 39, then 40 to 49. Here we find that children per household, those under 15, increase consistently as one moves to newer settlement classes, with the exception of the newest region, where the figure drops off somewhat (section B of Table 1-6). For adults, the picture is less uniform. It is clear that with regard to older adults — both those over 30 and over 55 — the older regions do, in general, have a somewhat larger number per household (sections D and E). It is worth noting, however, that for those households where the wife is in the reproductive age span, not more than one household in three has a person over 55, even in the oldest region.
When the 15 to 29 age group is added to those over 30, and one considers the number of persons aged 15 and over per household, there is still some tendency for the older regions to exceed the newer, but the pattern is no longer as consistent (section C). This is particularly so for households in which the wife is over 40 and the higher frontier fertility leaves its mark on the 15 and over age group via extra children over 15. Moreover, because the regional differences in persons 15 and over per household are smaller and less systematic, the differences in children per household dominate the regional differences in overall household size (section A). Thus, average household size typically increases as one moves from older to newer areas, except for the newest settlement class (where children per household also drops off).
Similarly the differences in young children dominate the dependency ratio — the number of persons under 15 or 65 and over per person 15 to 64(Table 1-7). This measure shows an increase from old to new regions, with the exception, again, of the newest region.
It is perhaps worth pointing out, finally, that the regional variations in average household size are fairly small. For husband-wife households in which the wife is in her 20s, the average size of household in all regions is not much different from four and a half persons; for wives in their 30s, and 40s, the range is from a little over six persons to somewhat under seven and a half.
Nuclear and nonnuclear household members. The nuclear family consists of a husband, wife, and their children. For husband-wife households with no children over 15, any figure for adults per household in excess of two necessarily implies the presence of nonnuclear adults. There may, in addition, be nonnuclear children in the household. To what extent do the figures on size of household reflect variations in the presence of relatives or unrelated persons in the farm household, and what roles do these persons perform? It is possible to form a rough idea from our data of the presence in each household of nonnuclear persons and their function. The present discussion focuses on those aged 15 and over.
In farm households where the wife is in her 30s, the average number of adults (those 15+) per household ranges from a little over three in old regions to somewhat under three in the newest. Table 1-8 indicates that in older regions the third person is considerably more likely to be nonnuclear than nuclear (lines 3 and 4). The probability of the third person being an older child increases, however, as one moves to the newer regions, and is about 50-50 in region IV. The third person, whether a child or nonnuclear household member, is more likely to be male than female (lines 7-8 and 11-12). The decline in adults per household as one moves to the newer regions is due to the diminished presence of nonnuclear persons, which more than offsets the increased presence of older children.
Our data at present do not permit precise identification of the characteristics of nonnuclear household members, but a reasonable approximation can be made, as shown in Table 1-9. For males, the figure on nonhead males reporting occupations comes out fairly closely to that for nonnuclear males 15+ (lines 1 and 2). The distribution by occupation (lines 3-5) suggests that most nonnculear males are in agriculture as laborers or farmers. Males in farm households reporting nonfarm occupations amount to only four or five per 100 households in every region. This suggests that nonfarm sources of income are not of importance for most of our farm households. It is true that there may be some contribution to the income of farm households from children not living on their parents' farms, but for households in this age group this is probably not of sizable importance.
Excerpted from Family and Population in Nineteenth-Century America by Tamara K. Hareven, Maris A. Vinovskis. Copyright © 1978 The Center for Advanced Study in the Behavioral Sciences, Stanford, California. Excerpted by permission of PRINCETON UNIVERSITY PRESS.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.