1. Stephan Thernstrom, Poverty and Progress: Social Mobility in a Nineteenth Century City (New York, 1973), pp 154–157.Google Scholar

2. Thernstrom showed this for the sample as a whole. Had he controlled for ethnicity, the magnitude of the inverse association might have dropped since more of the Irish were homeowners, and the Irish children were probably less upwardly mobile than the native ones even when ownership is taken into account. But I doubt that the positive association would have disappeared altogether, judging from Tables 6, 13 and 14 of Poverty and Progress. Unfortunately, the table which shows the relation between property accumulation and mobility within each ethnic group, Table 14, combines, as I understand it, both father's and son's occupational mobility (it is entitled “Mobility of … laboring families”). Thernstrom used it to show a number of relationships between occupational and property mobility. However, in order to observe the relationship between property ownership and son's occupational advancement within ethnic group, occupational advancement should be defined only in terms of the son. Still, I strongly suspect that no matter how the Newburyport data were reshuffled they would support Thernstrom's conclusion of a negative relationship between ownership and son's advancement.Google Scholar

3. Thernstrom, Stephan, The Other Bostonians: Poverty and Progress in the American Metropolis, 1880–1970 (Cambridge, Mass., 1973), 97–104.Google Scholar

4. Hogan, David, “Education and the Making of the Chicago Working Class, 1880–1930”, History of Education Quarterly, 18 (Fall 1978): 244. To the best of my knowledge, Hogan does not refer to Thernstrom's work on the sacrifices necessitated by homeownership. This is a pity because their arguments on this one point seem to be very similar. If there is a difference between them, it might have been useful to make that difference explicit. And, of course, the results of Thernstrom's Boston analysis call into question the universality of the pattern which Thernstrom suggested for Newburyport and which Hogan is emphasizing.Google Scholar

5. At one point in his essay Hogan does try to show that the connection between ownership and child labor is closely connected by presenting additional evidence. He relates the rate of ownership in specific groups to the rate of child labor in these groups. But he does not do so systematically since, according to his argument, child labor could be replaced by income from boarders. He might still have investigated how well some combination of these sources of income predicted rates of ownership. But there is a more fundamental problem with this approach. Hogan's theory does not, I think, give us much reason to think the drive to attain the security of ownership would differ from one immigrant group to another. Hence differences in ownership rates between groups should be understood in terms of other factors, such as father's wage rate. Perhaps because of these complications Hogan restricts himself to pointing out that two groups with the highest rates of child labor (the Irish and the German) also had the highest rates of homeownership (Hogan, , “Education,” 243).Google Scholar Parenthetically, it is not clear to me why from Hogan's point of view the native working class should have had any less desire than the immigrants to obtain the security of ownership. They too were subject to the “viscissitudes of wage labor”.Google Scholar Finally, there is a problem in the way “child labor” is treated which is not germane to the theoretical issues but which may invalidate the evidence Hogan presented. As a measure of child labor, Hogan relies on the Immigration Commission's findings on the proportion of family income in each group that came from a child. But any child living at home who contributed to the family income would have been included in this count—regardless of his age. So the figures are not a good guide to “child labor”. Moreover, immigrant families were usually larger, I suspect, and more children would have been available to work, hence “child labor” would have been a larger proportion of family income.Google Scholar Finally the age of children left home, and living arrangements generally, would seriously affect the measure. One wonders, for example, how the Commission classified extended families, with two adult wage earners. These factors probably make “proportion of family income contributed by children” a poor guide to “child labor”. A hint that this is the case appears in the evidence Hogan presents. He points out that the Germans and the Irish had the highest proportion of income contributed by this source, and the highest proportion of home ownership. But another table shows that they also had very high proportions of their children enrolled in the higher grades in school (as judged by the ratio of 8th graders to 5th graders in 1908). See Hogan, , “Education”, 243 and Table 12, 248.Google Scholar I want to reemphasize that Hogan's ideas are not per se implausible. I return later in the text (and in note 18) to the difficult problems of finding appropriate evidence on these matters.Google Scholar

6. The research design is described briefly in Perlmann, Joel, “The Use of Student Records for the Study of American Educational History”, Historical Methods, 12, No. 2 (Spring 1979): 66–75 and in detail in “Education and the Social Structure of an American City: The Research Design and the Data Collection” E.R.I.C. Microfiche no. 170–220. There is no need to repeat any but the most essential details here. All Providence children in the sampled age range who fell on every nth page of the census schedules were included. For technical reasons (explained in the words just cited) the age range sampled in 1880 was 11–16, in 1900 13–16 and in the two later years 12–15. In 1880 and 1915 samples of girls as well as of boys were sampled. In each year about 1000 children were included in each single sex sample. Finally, in each year a supplemental sample including all the blacks in the age range was collected and in 1915 large supplemental samples of Irish, Italian and Russian Jewish children were collected in order to permit intensive analysis of the experience of these groups.Google Scholar

7. On the problem of tracing individuals with common names see Thernstrom, , The Other Bostonians, Appendix A and Perlmann, , “The Research Design and the Data Collection”, Section III.Google Scholar It was necessary to identify individuals in the Tax Book by their name only. Those with common names therefore had to be excluded. They include a substantial proportion of the family heads—roughly a third in 1880, slightly fewer in other periods. Since commonness of name was closely associated with ethnicity, more individuals in some ethnic groups than in others were excluded. However, there is no reason to think that the relationship between homeownership and schooling—within ethnic group and father's occupational stratum—was different for those included than for those excluded. We can therefore be reasonably confident that the technical difficulties of tracing did not substantially bias the results.Google Scholar The Tax Books listed the value of real and personal property. I assumed ownership of real property was equivalent to homeownership. No doubt in some cases the assumption is erroneous, but probably not in many. At the very least those who owned land can reasonably be thought to be on their way to homeownership.Google Scholar Nevertheless, a comparison of the figures for homeownership based on the trace to the Tax Book, and the figures based on the U.S. census returns of 1890–1920 serves as a warning about the quality of the data (Table III). The fluctuation in the former is not very clearly reflected in the latter. It is possible, of course, that patterns of ownership may have changed more among families of adolescents than among all families in the city. But I think it is more reasonable to attribute the difference to other sources: sampling variability, the accuracy of the assessment lists, the difficulties of tracing individuals from one source to another, and the assumption that property ownership was equivalent to homeownership. For our purposes the major issue is whether these imperfections systematically bias the findings on the relationship between schooling and homeownership. I do not think they do, for two reasons. First, the material almost surely does present us with two groups in each sample, one in which a great majority were homeowners, the other in which a great majority were not. A comparison of the rates of schooling in these groups should provide a reasonable indication of what the rates would be if the minority of misclassified cases could have been correctly assigned (there is no reason to assume something special about the relationship between homeownership and schooling among the misclassified). The second reason for confidence is that the 1880, 1915 and 1925 evidence, based on the Tax Books, points unmistakably in the same direction as the 1900 evidence, which is based on the census records. The consistency with regard to the basic issue—the relation between homeownership and schooling—is reassuring.Google Scholar The tables showing homeownership rates across ethnic groups should ideally have been weighted in such a way that each ethnic group constituted as large a proportion of the group with common names as it did of the total group. But the weighting was only carried out in the first table. It made so little difference to the results that it was not undertaken for subsequent tables (a considerable amount of computation is involved). In the tables which exploit log-linear analysis the weighting is especially unimportant. There the relationship between homeownership and schooling is examined within each ethnic group (and that relation did not vary in important ways across groups—see note 13—so that small differences in the weighting of the results from each ethnic group would have an insignificant effect on the multivariate analysis).Google Scholar

8. Thernstrom, , The Other Bostonians, Appendix B.Google Scholar

9. The 1925 figures may be affected by an anomaly in the data for that year. The 1925 Census did not include occupational data. That was obtained from the city directory. But for about 8% of the heads the trace to the directory was unsuccessful. If these individuals were disproportionately blue collar non-owners, which is a reasonable assumption, 1925 blue collar ownership rates are somewhat inflated in the table.Google Scholar

10. I have contrasted owners with non-owners—rather than attempting to control for the value of the property owned. The best data available, that derived from the 1900 census, does not indicate the value of the home. Also, in other years, so few working-class families owned substantial amounts of property (assessed value over $5,000) that it hardly seemed worth distinguishing their experiences from the others.Google Scholar

11. For 1880, school attendance, as opposed to high school enrollment, rates are also available: Yankees 76%, Irish 41%. On the difference between these two measures of schooling, see text, below (next paragraph).Google Scholar

12. Several clear expositions of log linear analysis are available. See especially Davis, J.A., “Hierarchical Models for Significance Tests in Multivariate Contingency Tables: An Exegesis of Goodman's Recent Papers” in Costner, Herbert L. (ed.), Sociological Methodology 1973–1974 (San Francisco, 1974), pp 189–231. Also helpful are Page, William F., “Interpretation of Goodman's Log Linear Effects: An Odds Ratio Approach,” Sociological Methods and Research, 5 (May, 1977): 419–435. Page argues for the use of odds ratios derived from expected values of log linear models as a measure of association in multivariate tables. This is the measure on which I have relied in the fourth column of Tables V and VII. Also helpful is Duncan, Beverly and Duncan, Otis Dudley, Sex Typing and Social Roles: A Research Report (New York, 1978), pp 335–366. The standard work in this field, for the mathematically trained reader, is Bishop, Yvonne M., Fienberg, Stephan E., and Holland, Paul W., Discrete Multivariate Analysis: Theory and Practice (Cambridge, Mass., 1975).Google Scholar

13. In none of the 6 comparisons were ethnic variations in the relation between schooling and property ownership statistically significant. (Nor could I find any intriguing hints in those variations.) The important point is that the relationship was nearly always positive within each occupational stratum and ethnic group (eg, among unskilled Irish) whenever cell size was not trivial. Consider, for example, the Providence Irish of 1880. As already noted, they, like the Irish of midcentury Newburyport were especially likely to own homes and their children were less likely to enroll in school than children in other groups. But among the Irish the relationship between school attendance and homeownership was, once again, positive: Google Scholar

14. Thernstrom, , Poverty and Progress, 122.Google Scholar

15. These supplemental analyses were limited to school attendance rates. High school enrollment rates in 1900 were too low to allow for meaningful log-linear analysis when the sample was subdivided into categories of so many control variables.Google Scholar

16. Hogan, , “Education,” 237–40 on lodging; on women's work see for example Table 7, page 240, or Modell, John, “Patterns of Consumption, Acculturation and Family Income Strategies in Late Nineteenth Century America,” in Hareven, Tamara K. and Vinovskis, Maris A. (eds.), Family and Population in Nineteenth America (Princeton, 1978), pp 218–19.Google Scholar

17. For an example of an attempt to study both sources of income and patterns of consumption see Modell, , “Patterns of Consumption.” Google Scholar

18. I have suggested that a relationship between relatively low aggregate rates of homeownership and relatively high rates of schooling in immigrant groups (compared to natives) may be spurious. I offered as evidence of spuriousness the finding that the children of homeowners were actually more likely to attend school (or enroll in high school) than the children of nonowners—even when father's occupation is controlled.Google Scholar The purpose of controlling father's occupation is to equalize insofar as possible the level of income available to a family due to the labor of the head of household. But it is possible that controlling father's occupation only achieves this purpose in part, because workers in the same occupational stratum differed in wages systematically by social group. The group with the higher earnings would clearly have had less need of child labor to purchase a home. Suppose that (a) the wages among homeowners were higher than among nonowners in the same stratum and (b) the wages among the Yankees were higher than among the Irish in the same stratum. If (c) the wage differences were large enough they could invalidate our conclusions. We might find that within the same occupational stratum and at the same level of wage homeowners were less likely to keep their children in school than nonowners (that the former did indeed sacrifice their children's schooling to the purchase of a home). In that case controlling father's occupation would simply not be controlling enough: the observed positive relationship between homeownership and schooling with father's occupation controlled would itself be spurious. And the crude, aggregate rates—showing lower school rates and higher homeowning rates among the Irish than among the Yankees—would be pointing in the right direction: with father's occupation and wage controlled the greater Irish desire for homeowning was satisfied at the expense of their children's schooling.Google Scholar But assuming (a), (b) and (c) above—assuming such large and consistent differences in income within the same occupational stratum—is assuming a great deal. Moreover, it would still be possible that ethnic differences in father's income were directly responsible for most of the ethnic differences in child labor—quite apart from issues of homeownership. And it is also perfectly possible that ethnic factors unrelated to either father's income or to homeownership—cultural factors involving the perceived value of schooling—were responsible for most of the ethnic differences in schooling. I referred to these in the text (page 4).Google Scholar The issue of the importance of cultural factors as opposed to factors arising directly from the American class structure clearly concerns David Hogan. He would like to argue that both played a part but the class structural factors were more fundamental. His argument, however, seems to rest on cultural differences in the desire for homoownership (the immigrant working class having sought it more, hence the more rapid attrition of their children from school in order to work for it). Moreover, he assumes cultural differences in the mode of response: some groups exploited child labor, others income from lodging. If these cultural differences existed, why not simply assume also that cultural differences between ethnic groups in the perceived value of schooling, unrelated to desire for homeownership, is the key determinant of ethnic differences in schooling (or of those that persist after father's occupation is controlled)? Google Scholar The argument in this note implies that if it were possible to control for father's wage and not merely for father's occupation, it would be advantageous to do so. Income from father's wage (or father's occupation used as a proxy for it) should be distinguished from other factors noted on pages 22–24: whether the wife worked, whether boarders were taken in, how many children the family included, etc. These factors involve strategies by which families could supplement the head's income or reduce their financial needs—and ways in which they could thus struggle toward homeownership. Hence, the text argues, these latter (in contrast to father's income or occupation) should not be controlled in assessing the importance of the relation between homeownership and schooling. 19. Finally, what of Newburyport? It should be recalled, first of all, that the Newburyport pattern of lower rates of mobility among the children of property owners apparently did not exist in late nineteenth century Boston. Mobility patterns may simply have differed in different cities, or they may have changed over time, of course. More central to the concerns of this paper is the point that Thernstrom's direct evidence pertained only to occupational mobility rates—not to rates of schooling among the children of homeowners and of nonowners. Possibly, therefore, less schooling may not have been the reason for less occupational mobility among the children of the homeowners—nor among the children of the Irish (who were more eager to own). The Irish, for example, may have kept their children out of the schools because the institutions appeared to be dominated by Protestantism—that is, for reasons unrelated to homeownership.Google Scholar Thernstrom, , Poverty and Progress, p. 24 notes the significance of the religious factor. I am grateful to Michael Katz for emphasizing to me its importance in relation to an understanding of a possible tradeoff between schooling and homeownership in Newburyport.Google Scholar And then again, possibly the tradeoff between homeownership and schooling did indeed take the form Thernstrom suggested in the midnineteenth century—but changed later when a belief in the value of schooling became more prevalent. The Newburyport data, in any case, do not allow us to say so since they do not include direct evidence on schooling.Google Scholar