Weighting regressions by propensity scores

Regressions can be weighted by propensity scores in order to reduce bias. However, weighting is likely to increase random error in the estimates, and to bias the estimated standard errors downward, even when selection mechanisms are well understood. Moreover, in some cases, weighting will increase the bias in estimated causal parameters. If investigators have a good causal model, it seems better just to fit the model without weights. If the causal model is improperly specified, there can be significant problems in retrieving the situation by weighting, although weighting may help under some circumstances.     

Union-nonunion wage differentials revisited

A reverse regression method of estimating the union-nonunion wage differential is developed using a multiple indicator model. The method provides a test for the multiple indicator model’s validity and suggests that conventional estimation techniques should underestimate the union-nonunion differential. Empirical estimates show that the reverse regression estimates are larger than conventional estimates and that the multiple indicator model cannot be rejected. The author wishes to thank Robert J. Flanagan and H. G. Lewis for their valuable comments on earlier versions of this paper.     

Government wage differentials revisited

This paper updates studies done in the mid-1970s concerning the measurement of government wage differentials in the United States. Based on an analysis of two large micro data sets, we reached the following conclusions. First, as a group, government employees earned a small wage premium of approximately 2 to 3 percent during the 1970s and early 1980s. Second, the government wage premium was substantially higher for federal employees and for women employees. These findings are consistent with earlier studies. We thank Eanswythe Leicester and Ken Emery for their research assistance and Diane Kurtz for her clerical assistance.     

Gender differentials in poverty-mortality well-being

"This paper explores the application of a life table approach to the analysis of poverty [in the United States]. After reviewing trends in gender differentials in poverty and mortality, we investigate an indicator of poverty-mortality well-being. The rationale for this approach is that to some extent poverty and mortality are causally interrelated phenomena which are both fundamental to well-being. The empirical results indicate that females are expected to live more years in poverty than are males, but females are expected to live more years in nonpoverty as well. Although the gender differential in poverty rates has increased to the disadvantage of females, and although most of the greater poverty of females cannot be attributed to their greater longevity, the gender differential in poverty-mortality well-being has nonetheless not deteriorated to the disadvantage of women in recent decades."     

The minimum wage and productivity differentials

A firm’s ability to adjust its production process to economize on low-skilled labor when faced with a minimum wage increase will differ greatly depending on industry or occupation. For example, more capital-intensive means of cleaning hotel rooms or serving customers at restaurants may not be readily available without degrading service quality. In such situations, the productivity of labor is essentially capped, and firms have few options when the minimum wage increases. This simple observation has implications for studies that rely on microdata to examine the effects of minimum wage increases. If firms only increase prices in response to a minimum wage increase, employment effects are likely small. If the goal of the minimum wage is to redistribute income from firms and consumers to workers, minimum-wage increases targeted at industries and occupations where such rigidities result in an inelastic demand for labor may achieve the desired goal at a lower cost than across-the-board increases. However, such a scheme causes an inefficient allocation of labor and would be subjected to substantial political pressures that may lead to anomalous results. Additionally, it is unreasonable to conclude that policy makers have the necessary information to skillfully set the minimum wage. I thank Brian E. Chezum and Jeff Waddoups for helpful comments. All mistakes, of course, are my own.     

Mortality differentials among immigrant populations in Canada

Mortality among different groups of migrants to Canada is compared with that of native-born Canadians and with trends in the migrants' countries of origin, using data for the period 1981-1985. Results show that life expectancy at birth among migrants in Canada is generally higher than that of the populations at place of origin and destination. (SUMMARY IN FRE AND SPA)     

Gender and racial earnings differentials in academic labor markets

This study estimates earnings differentials across racial and gender groups among college faculty, and decomposes these differentials into the components attributable to differences in individual and institutional characteristics and that which remains unexplained. We find that white females earn approximately 4 percent less than white males; black males earn 7.4 percent more; black females earn one percent more; Hispanic males earn 2.7 percent more; Hispanic females earn 1.7 percent less; and Asian males earn 7.7 percent more than comparable white males. We also find a significant earnings penalty for being a naturalized citizen or noncitizen versus an U.S.-born citizen.     

Race differentials in obesity: the impact of place

This article reveals race differentials in obesity as both an individual- and neighborhood-level phenomena. Using neighborhood-level data from the 1990-1994 National Health Interview Survey, we find that neighborhoods characterized by high proportions of black residents have a greater prevalence of obesity than areas in which the majority of the residents are white. Using individual-level data, we also find that residents of neighborhoods in which at least one-quarter of the residents are black face a 13 percent increase in the odds of being obese compared to residents of other communities. The association between neighborhood racial composition and obesity is completely attenuated after including statistical controls for the poverty rate and obesity prevalence of respondents' neighborhoods. These findings support the underlying assumptions of both institutional and social models of neighborhood effects.     

Social mobility and the interpretation of social class mortality differentials

The discussion of health inequalities in Britain (e.g. in the Black Report) has been conducted largely on the basis of social class mortality differentials measured by achieved social class and not by social class of origin. It is shown in this paper that social class mortality differentials by achieved social class are not invariant to the rate of social mobility and that the use of them is likely to result in a biased measure of trends in health inequalities when the absolute rate of social mobility varies over time. It is further shown that if, as is likely, health status is a factor systematically affecting the probability for an individual of upward or downward social mobility, then an increase in the rate of social mobility may well result in constant or widening social class mortality differentials by achieved social class even if the differentials are narrowing when measured by social class of origin. It is claimed that this process may well explain why the observed social class mortality differentials, which are measured by achieved social class, have not fallen in Britain during the post-1945 period.     

State and local government wage differentials: An intrastate analysis

Previous estimates of state and local government wage differentials have been typically based on data aggregated across all states, and such aggregation may produce seriously misleading differential estimates. We estimate intrastate earnings differentials for the state and local sectors in Wisconsin, four other midwestern states, and two states outside the midwest. There is substantial variation in the differentials: aggregated differentials can be misleading. Our work also confirms that state and local government labor markets have reduced earnings dispersion and investigates the possibility that higher public sector earnings may attract an “over-qualified” work force. The authors thank the Wisconsin Policy Research Institute for financial support and members of the UWM economics seminar and Sammis White for detailed comments on earlier drafts.     

Union-Nonunion differentials and establishment size: Evidence from the NLSY

We provide new empirical evidence on union-nonunion differentials using the 1990 wave of the National Longitudinal Surveys of Youth (NLSY) data set which allows us to examine a broader set of fringe benefits than most other studies and provides a rich set of control variables. Our major finding is that the union effect decreases with establishment size for both components of the compensation structure, i.e., wages and fringe benefits. An earlier version of this paper was presented at the Southern Economic Association conference in Washington, DC in 1996. We thank Tom Hyclak and other session participants. We also acknowledge the financial support from Ade Howe Kent Fund. The usual disclaimer applies.     

Socio-economic and cultural differentials in contraceptive usage among Ghanaian women

Data from the Ghana Fertility Survey of 2001 married women in 1979-1980 were subjected to logistic regression to determine the factors influencing contraceptive use. In this Ghanaian sample only 22 women and no men were sterilized, 11% used an efficient contraceptive method and 8% were using an inefficient method. The most prevalent methods were abstinence by 6% and pill by 5%. The variables analyzed were birth cohort, age at 1st marriage, education, occupation, religion, ethnicity, rural/urban residence, northern/southern residence and number of children desired number of living children. All these factors were dichotomized, e.g., cohort: born before or after 1950. Factors positively significant for contraceptive use were younger women (20% more likely), married at age 20 or older (82% more), education (150% for any method, 67% for an efficient method), professional occupations, protestants, urban residence, southern residence, desire fewer children. Factors negatively associated with contraception were agricultural work (50% as likely), non-Christian religion, both traditional and Moslems (75%), desiring more children and living in the north. Unexpectedly, living in the northern undeveloped region was strongly linked with use of an efficient contraceptive. A factor without significant effect was ethnicity, Akan or non-Akan. These results were discussed with a general review of the literature on determinants of contraceptive use.     

Socioeconomic differentials in the transition to first cohabitation in Norway

Using Norwegian data this study examines the impact of individual as well as parental socioeconomic resources on the timing of the transition to first cohabitation. The analyses show that the entrance into a first cohabitation is positively affected by individual level of education, whereas school enrollment delays first cohabitation. Further, respondents whose parents have a secondary or tertiary education start their first cohabitation later than those with lower educated parents. There are, however, important gender differences in the association between socioeconomic variables and the timing of first cohabitation. First, there is a stronger positive association between level of education and timing of first cohabitation for women than is the case for men. The positive effect of income, on the other hand, is more pronounced for men. Lastly, the delaying effect of having a tertiary educated father is stronger for the female sub-sample.     

South-nonsouth differentials in National Labor Relations Board certification election outcomes

This paper investigates and evaluates the magnitude of regional differentials in outcomes in National Labor Relations Board elections. The data for this project include NLRB elections conducted during the period 1973–1978. The author wishes to thank Steven Hills, Michael Borus, Gil Nestel and Kezia Sproat of the Center for Human Resource Research for their comments on an earlier draft of this paper.     

The effect of type of union on member-nonmember wage differentials

The union-nonunion wage differential can be decomposed into bargaining and membership effects. While some analysts suggest that they are not separable and that bargaining power is a function of membership density, others argue that they are separable and that the former derives from monopoly power while the latter stems from socialization. Our results support the latter view. We derive estimates of bargaining and membership effects for workers covered by national, industrial, and craft union contracts as well as for all covered workers taken together. Since industrial and craft unions differ in structure and organization, we expect differences in the socialization effects among types of unions. It is clear from our results that union membership per se in each case gives a large positive wage advantage.