Ross M. Stolzenberg

Tools for Intuition about Sample Selection Bias and Its Correction

The authors provide mathematical tools to assist intuition about selection bias in concrete empirical analyses. These new tools do not offer a general solution to the selection bias problem; no method now does that. Rather, the techniques they present offer a new decomposition of selection bias. This decomposition permits an analyst to develop intuition and make reasoned judgments about the sources, severity, and direction of sample selection bias in a particular analysis. When combined with simulation results, also presented in this paper, their decomposition of bias also permits a reasoned, empirically-informed judgment of when the well-known two-step estimator of J. Heckman is likely to increase or decrease the accuracy of regression coefficient estimates. The authors also use simulations to confirm mathematical derivations

Intended childbearing and labor force participation of young women: insights from nonrecursive models.

Analysis of 3589 women in their mid-20s indicates that the number of children planned is more often determined by labor force participation planned than the reverse. A statistical model is constructed which 1) allows labor force participation plans to cause fertility expectations of young women; 2) simultaneously allows fertility expectations to cause labor force participation plans; 3) allows certain background factors to account for the relationship between fertility expectations and labor force participation plans. One model allows a womans plans for labor force participation to be caused by income and attitude of husband toward labor force participation. The conclusion reached by the model is that the correlation between fertility expectations and labor force participation results from their common antecedent causes rather than a direct causal link.

Theory Testing in a World of Constrained Research Design

Because censored sampling is often unavoidable in much sociological data analysis, computationally simple corrections of censoring bias would be useful. Heckmans correction is simple to compute, widely used, and proven asymptotically correct under certain assumptions, but its limitations in practical situations are not well known in sociology. Here, we overview prior criticisms of Heckmans estimator, and we consider the case in which its normality assumptions are satisfied, censoring rates are high, and sample sizes are small. Results of 14,400 analyses of computer-generated simulation data suggest that Heckmans method performs well under certain circumstances, but that it very frequently worsens estimates, especially under conditions that are likely to be present in sociological data. Thus, the technique is probably not a general cure for censoring bias in sociology, except perhaps where strong theory permits certain strong assumptions. We reconsider censored sampling correction strategies in the context of statistical analysis as a theory-building tool, with emphasis on research strategy in the presence of irremediable censoring bias.

Occupations, Labor Markets and the Process of Wage Attainment

A key problem in sociology, as in economics, is explaining why some workers earn more money than others. Sociological models of earnings have stressed the role of a workers occupation and have tended to ignore the conditions of the labor market in which he finds work. Economic models have stressed labor market functioning at the expense of considering the role of a workers occupation in determining his wages. In this paper, I attempt to combine sociological models of earnings with (a) economic models of earnings and (b) concepts and findings from the sociology of occupations and professions. I argue theoretically and empirically that some similar conclusions about the processes governing individual earnings attainment can be drawn by examining occupations in terms of labor markets and by analysis of labor markets from the standpoint of occupations. These conclusions are: (a) that labor markets tend to be fragmented along occupational lines, (b) that the processes governing wage attainment vary from one occupation to another and (c) that occupational differences in these processes can be predicted from and explained in terms of the forces which lead to occupational segmentation of labor markets. I discuss some useful implications of my analyses for the study of the relationship between worker age and worker earnings, and I perform some empirical and theoretical analyses of occupational differences in the age-wage relationship. Data are drawn from the U.S. Censuses of 1960 and 1970 and from publications of the U.S. Bureau of Labor Statistics.

Local labor markets, children and labor force participation of wives

Most research on married women’s labor force participation relates characteristics of individual women to their probability of labor force participation. Some studies relate characteristics of geographic areas to average labor force participation rates in those areas, although these aggregate level analyses are usually gross tests of ideas about individuallevel processes. Here we take a quintessentially sociological perspective and seek to understand how characteristics of geographic areas structure the relationship between properties of individual women and their probabilities oflabor force participation. Our analysis has two steps. In step one, we fit individual-level probit models of married women’s probability of labor force participation. A separate model is fitted in each of 409 areas using 1970 Census data, and the relationship between individual characteristics and labor force participation is found to vary substantially across areas. In step two, we attempt to explain areal variation in the effects of women’s children on their labor force participation. We hypothesize that the effect of children on their mothers’ labor force participation is a function ofthe cost and availability of childcare, and of the “convenience” of jobs for working mothers in the places where the mothers live. Measures of childcare cost, childcare availability and job convenience are developed. Weighted least squares analyses of probit coefficients from the first stage are, in general, very consistent with our findings, and suggest that the approach taken in this paper is likely to be a fruitful one for future studies.

Educational Continuation by College Graduates

This article resonsiders the apparently paradoxical absence of back ground SES effects on school continuation by college graduates. The approach is distinguished by its focus on specific graduate programs, in particular MBA programs, and by a concentration on different types of attitudes that appear relevant to graduate school matriculation. The article examines the impacts of job values and direct attention to conative attitudes toward graduate school, which are inferred from prior behavior. Results are consistent with earlier findings of SES effects on total years of schooling, yet also explain the absence of parental SES effects on schooling beyond college.

Age fertility expectations and plans for employment

Fertility, female laborforce participation, and the relationship between them are key subjects in a number of theoretical and applied areas of sociology. Because sex role norms and the widespread use of birth control devices have given American women much control over their fertility and substantial choice in their labor force activity (or inactivity), understanding the development and interrelationship of labor force participation plans and fertility expectations assumes great importance in understanding actual labor force participation and actualfertility. As a step toward understanding this development, we describe and attempt to explain the effect of womens age on the relationship between their labor force participation plans and their fertility expectations. Using data from a national sample of young women aged 19 to 29 in 1973 (N=3,589), we find a strong, linear relationship (r= -.96) between womens age and the effect of their plans for labor force participation on the number of children that they expect to bear in their lifetime. An explanation of this finding (called the Learning Hypothesis) is advanced which survives tests against several plausible alternative hypotheses. Policy implications and productive paths for future research are discussed.

It's about time and gender: spousal employment and health.

This article considers the effect of husbands’ and wives’ hours of work on each others’ health. Theoretical analysis focuses on gendering of health‐related behavior, the time needed to promote a spouses salubrious behavior, and the effects of work hours on the availability of time for nonwork activities. Empirical analyses are based on 1986 and 1989 longitudinal U.S. data. Fewer than 40 hours of work per week by wives has no effect on husbands’ health, but more than 40 hours has substantial negative effect. Long work hours by husbands are not detrimental to wives’ health. Wives’ work hours shows no effect on their own health, but husbands’ work hours show strong positive effect on their own health. Methodological issues are considered.

Education, Occupation, and Wage Differences Between White and Black Men

Past research has indicated repeatedly that black men receive lower wages than white males working in the same occupation. Past findings have also suggested that these within-occupation race differences in mens success in convertin their years of schooling into dollars of earnings. This paper reexamines the role of returns of schooling in producing differences in earnings of white and black men who are employed in the same occupation. The role of schooling into dollars of earnings. This paper reexamines the role of returns of schooling in producing differences in earnings of white and black men who are employed in the same occupation. The role of schooling in determining wages of all workers is also considered. A hypothesis suggesting that years of schooling and years of labor-force experience have joint nonadditive effects on earnings is formulated, tested, and supported by several regression analyses. A measure of race differences in wage returns to schooling based on partial derivatives is computed from separate regression analyses of earnings of white and black males in each of 62 detailed occupation categories. These occupation categories. These occupation categories subtend 79% of the black male labor force (and 68% of the total labor force) which reported its occupation in the 1960 census of population. The results surprisingly indicate that within-occupation race differences in wage returns to schooling are not large enough to cause substantial race differences in pay for incumbents of those occupations employing the vast majority of black men in the labor force. This finding suggests that racial differences in the quality of schooling are largely irrelevant to within-occupation race differences in earnings.

Retirement and Death in Office of U.S. Supreme Court Justices

We construct demographic models of retirement and death in office of U.S. Supreme Court justices, a group that has gained demographic notice, evaded demographic analysis, and is said to diverge from expected retirement patterns. Models build on prior multistate labor force status studies, and data permit an unusually clear distinction between voluntary and “induced” retirement. Using data on every justice from 1789 through 2006, with robust, cluster-corrected, discrete-time, censored, event-history methods, we (1) estimate retirement effects of pension eligibility, age, health, and tenure on the timing of justices’ retirements and deaths in office, (2) resolve decades of debate over the politicized departure hypothesis that justices tend to alter the timing of their retirements for the political benefit or detriment of the incumbent president, (3) reconsider the nature of rationality in retirement decisions, and (4) consider the relevance of organizational conditions as well as personal circumstances to retirement decisions. Methodological issues are addressed.