«CENTER DISCUSSION PAPER NO. 993 Social Class and the Fertility Transition: A Critical Comment on the Statistical Results Reported in Simon ...»
ECONOMIC GROWTH CENTER
P.O. Box 208629
New Haven, CT 06520-8269
CENTER DISCUSSION PAPER NO. 993
Social Class and the Fertility Transition:
A Critical Comment on the Statistical Results Reported in
Fertility, Class and Gender in Britain, 1986-1940 Geoffrey Barnes Yale University Timothy W. Guinnane Yale University November 2010 Notes: Center Discussion Papers are preliminary materials circulated to stimulate discussions and critical comments.
Barnes is a student in Yale College. Guinnane is the Philip Golden Bartlett Professor of Economic History in the Department of Economics, Yale University. Address correspondence to firstname.lastname@example.org. We are grateful to Simon Szreter for his comments on an earlier version of this note.
This paper can be downloaded without charge from the Social Science Research Network electronic library at: http://ssrn.com/abstract=1709594
An index to papers in the Economic Growth Center Discussion Paper Series is located at:
Social Class and the Fertility Transition:
A Critical Comment on the Statistical Results Reported in Simon Szreter’s Fertility, Class and Gender in Britain, 1860-1940 Geoffrey Barnes and Timothy W. Guinnane Abstract Simon Szreter’s book Fertility, Class, and Gender in Britain, 1860-1940 argues that social and economic class fails to explain the cross-sectional differences in marital fertility as reported in the 1911 census of England and Wales. Szreter’s conclusion made the book immediately influential, and it remains so. This finding matters a great deal for debates about the causes of the European fertility decline of the nineteenth and twentieth centuries. For decades scholars have argued whether the main forces at work were ideational or social and economic.
This note reports a simple re-analysis of Szreter’s own data, which suggests that social class does explain cross-sectional differences in English marital fertility in 1911.
Keywords: Fertility transition, 1911 Census of England and Wales JEL Codes: J13, N33 The 1996 publication of Simon Szreter’s Fertility, Class, and Gender brought the author acclaim seldom produced by any scholar's first book. The work was widely and almost uniformly positively reviewed. David Kertzer, for example, called it “something of a tour de force, combining a historian’s concern for local context, a demographer’s multivariate statistical techniques for the analysis of fertilityhistory data, and a social theorist’s engagement with epistemological issues.” 1 More than one reviewer remarked that the book reports research sufficient in scope and quality for at least two monographs. One consists of a thoughtful analysis of the way the British statistical authorities thought about "class," and the way those notions shaped the taking of the census. The second is the subject of our discussion here.
Szreter provides an analysis of the remarkable fertility survey undertaken as part of the 1911 Census of England and Wales. He argues that the “professional model” of class-based differential fertility held by the census officials and others at the time does not explain the actual patterns in the data. Many take Szreter to have demonstrated that socio-economic differences in fertility were not terribly significant in turn-of-the-century Britain, and that, by extension, socio-economic forces did not play a major role in the decline of fertility. 2 Only slightly less influential is Szreter’s positive conclusion that Britain underwent many simultaneous fertility transitions, each reflecting participation in a “communication community” rather than social class or economic status.
The statistical heart of the book relies on fertility measures drawn from the published fertility survey. This unusual exercise was published in multiple cross-classifications, allowing later scholars to examine differences in fertility according to the occupations of both husbands and wives. Social “class” reflects an amalgamation of groups of occupations. Szreter refers to the then-prevailing theory of fertility decline and its relationship to social class as the “professional model” and concludes, based on his statistical analysis that the model “fails.” We find that, on the contrary, that the professional model performs well by conventional standards. Since we rely entirely on the professional model as he
Journal of Interdisciplinary History, 28(1), Summer 1997, p.103. We located fourteen reviews of the book in scholarly journals. As of September 2010, the book had 209 citations on Google Scholar. The only generally negative review was by Charles Tilly, Population and Development Review 22(3), September 1996, pp. 557-66. 2 describes it, and the fertility measures he reports, the only differences at issue are statistical methods and interpretation of results. Our discussion should not be taken as a rejection of Fertility, Class and Gender in its entirety. Much of Szreter’s argument deals with the nature of fertility control, more particularly the issue of spacing versus stopping. This part of the book is not affected by the results we report here.
Moreover, we do not claim to have made our own, positive contribution to understanding the British fertility transition. Such a contribution would require an even more complete exploitation of the published tabulations than Szreter attempted, and would in any case be superseded by research based on the manuscript census schedules for 1911, which will soon be available to scholars.
1. Class, occupation, and fertility Part II of Fertility, Class and Gender is an intellectual and policy history of the making of the 1911 fertility survey. Szreter starts from the premise that most social scientists probably share, but few take as seriously: that the sources from which we gather our data were themselves devised and used with specific purposes in mind. That is, data collection reflects a theory of social behavior. If a census reports considerable information on religion, for example, it is likely that those responsible for its construction thought religion a powerful force. Those theories inform the categories used in the sources, as well as the selection of people for inclusion in the study.
Szreter documents a growing fear about fertility decline in late-nineteenth century Britain, especially the fear that the “lower orders” were providing an ever-larger share of each birth cohort. These developments justified the expense and trouble of the special fertility survey undertaken in connection with the census. The manner in which the census authorities decided to publish the data reflects, as Szreter stresses, their understanding of the causal mechanisms underlying fertility patterns. The census tabulations of fertility by social class reflected a shared, almost dominant understanding of differential fertility at the time: fertility decline had gone furthest among the upper classes. In discussing Szreter’s argument and results, it is important to stress that the professional model implies regular gradations by
that simply claims social Class I has fertility different from Class II.
The 1911 census organized the British population into 465 occupations. The social-class groupings were amalgamations of subsets of these 465 occupations. The underlying idea was really one of three classes: Class I, thought to be an “upper and middle class;” Class III, comprising skilled workmen;
and Class V, the unskilled. The remaining two classes were inserted as Class II and Class IV. 3 Classes II and IV consisted of couples who did not fit together as naturally as is the case for the other three Classes, and it is not clear that they serve as anything more than intermediates. Szreter discusses several different groups of couples defined by their age at marriage and marital duration as of 1911, but in his analysis he stresses one particular group, those married at ages 20-24 in the period 1881-1885. This is the largest single age-at-marriage category. By 1911, these couples were nearing the end of their reproductive lives but were not so old that one would have to worry much about selective mortality affecting the chances that higher-fertility wives would still survive. 4
2. Szreter’s analysis We focus on the discussion presented in Chapter Six of Fertility Class and Gender (entitled “A test of the coherence of the professional model of class-differential fertility decline”). The 1911 census published complete parity distributions for each marriage-age and duration category, but the occupational cross-tabulations only present the total number of couples and total births (and child deaths) in each age-duration category. 5 Thus Szreter works
The system was devised by the Registrar-General, who also ran the census, and was first described in the Registrar-General’s Annual Report for 1911 (published in 1913) Szreter (1996, pp. 255-256). The reference to the key source in Szreter’s bibliography is not quite right; the proper full citation is in the list of references. We restrict our analysis in this note to the sub-population Szreter stresses. But we should note that age at marriage itself could reflect the influence of social class, or more broadly social and economic forces, on the desire to limit family size. Elsewhere in his book Szreter stresses precisely this point. One could use the parity distributions to look for heterogeneity in the way couples achieved smaller families, for example. Others have used the similar fertility survey from the 1911 Census of Ireland to examine questions similar to this one. Paul David, Warren Sanderson, and their collaborators have developed an entirely new statistical methodology that exploits all the information in those parity distributions for the purpose of arriving at measures of 4 with the mean number of children ever born (CEB). In evaluating the social-class model we need to bear in mind that the census reports cross-classification of fertility by occupation for a sub-set of occupations. The census itself counts 465 different occupations, but the fertility crossclassifications restrict themselves to 206 occupations chosen, Szreter (1996, p. 291) reports, “for a combination of their quantitative importance and their specificity of occupational description.” They include 87.5 percent of couples enumerated in the census. Szreter argues that Classes I-V are graded, and thus focuses on the 176 occupations in those classes. Including Classes VI-VIII (which we discuss below) increases the total number of occupations to 195.
Szreter uses these fertility measures drawn from the 176 different occupational groups to demonstrate that the social-class model of fertility, in his word, fails. Table 1 reports descriptive statistics for mean CEB in the five classes. The table confirms that fertility was graded by social class. Yet one could suggest, as Szreter does, that there is too much variation in CEB within each class to view them as
A strong and relatively unequivocal corroboration for a theory of class-differentials would be provided by a pattern in which fertility levels of the component occupations within each class proved to be reasonably tightly bunched around their class mean, demonstrating that the classes approximated to both homogenous and discrete entities. Of course, there would be some outliers in each class but a high degree of continuity in the run of occupational fertility values, without any clustering, certainly would not conform to this kind of highly positive confirmation of the class-differential model.
Secondly, it might be found that there was no bunching of occupational fertility values around their class mean and correspondingly no clear separation between the graded social classes, as each one merged into the next. But it might still be the case that the fertility levels of the individual occupations followed a rank ordering more or less in accord with their graded class affiliations: class I occupations appearing on a rank-ordered list before those of class II and so on.
This would be a qualified success for the professional model: class-graded rather than classdifferential fertility decline.
Thirdly, a still greater degree of overlap between the occupations of the adjacent graded classes might present itself. More than a merging at the edges, there might be such substantial overlapping of the fertility values recorded by many of the individual occupations drawn from different classes that some of the social classes could hardly be distinguished from each other in