Randall J. Olsen

The Effects of Salaries and Opportunity Costs on Length of Stay in Teaching Evidence from North Carolina

This paper shows that teachers who are paid more stay longer in teaching, that teachers with high opportunity costs, as measured by test scores and subject specialties, stay in teaching less long than other teachers do, and that salaries influence duration less for teachers with high test scores than for teachers with lower scores. The research is based on a new longitudinal dataset providing information on the career histories of 13,890 North Carolina teachers. The empirical work uses a generalized least squares estimation technique that accommodates censored observations, time-varying covariates, and fixed effects.

The Effects of Salaries and Opportunity Costs on Duration in Teaching: Evidence from Michigan

This paper demonstrates that salaries and opportunity costs have important influences on how long teachers stay in teaching. The empirical work is based on a new longitudinal dataset providing information on the careers of 7800 Michigan public school teachers. The generalized least squares estimation technique accommodates censored observations, time-varying covariates, and fixed effects. One implication of the result is that predictions from the current generation of teacher supply and demand models are likely to be misleading. Copyright 1989 by MIT Press.

You can go home again: Evidence from longitudinal data

In this paper we analyze the economic and demographic factors that influence return migration, focusing on generation 1.5 immigrants. Using longitudinal data from the 1979 youth cohort of the National Longitudinal Surveys (NLSY79), we track residential histories of young immigrants to the United States and analyze the covariates associated with return migration to their home country. Overall, return migration appears to respond to economic incentives, as well as to cultural and linguistic ties to the United States and the home country. We find no role for welfare magnets in the decision to return, but we learn that welfare participation leads to lower probability of return migration. Finally, we see no evidence of a skill bias in return migration, where skill is measured by performance on the Armed Forces Qualifying Test.

Comparing Public and Private Schools: The Puzzling Role of Selectivity Bias

Recent articles using the same data and variants of the same estimation technique report conflicting estimates of the relative quality of education provided by public and Catholic schools. This article explains the reasons for the conflict. In so doing, the article clarifies the assumptions underlying the increasingly popular two-step methods for controlling selectivity bias and highlights the hazards of using these methods when the assumptions are not satisfied. The article also illustrates an alternative method for detecting the presence of selectivity bias.

Let Them Eat Cake: A Note on Comparing Alternative Models of the Demand for Medical Care

This article comments on the Rand econometric approach to estimating and testing models of the demand for medical care as typified in Duan et al. (1983). The Rand methodology is contrasted with more mainstream econometric approaches to similar empirical problems as developed in the sample selection literature. It is shown that despite Rands claim, their approach requires some fairly unusual assumptions on the model joint error distribution and functional form. Alternatively, their model likelihood functions may be interpreted as being nested in more general sample selection model likelihood functions with easily testable parameter restrictions.

Endogenous Covariates in Duration Models and the Effect of Adolescent Childbirth on Schooling

This paper uses a hazards model with endogenous covariates to estimate the effects of family background, employment opportunity, and childbearing upon the school dropout of youths from low-income households. Family background and employment opportunity are found to be significant, but when the effect of childbearing is properly estimated, it is not significant. The authors conclude that childbearing and dropout are jointly determined, but that when this is properly accounted for, childbearing appears to exert little effect upon dropout.

EVALUATION OF THE OLSEN TECHNIQUE FOR ESTIMATING THE FERTILITY RESPONSE TO CHILD MORTALITY

In a previous issue of this journal, Olsen proposed a technique for quantifying the fertility response to child mortality. To estimate the extent of child replacement, one needs data only on the number of children ever born and the number of child deaths for each woman. The technique involves first running a regression of the number of births on the number of deaths and then correcting the regression coefficient in order to obtain a consistent estimate of replacement. Here we evaluate the performance of the technique by seeing how well it works on a simulated set of reproductive histories for which we know the true extent of replacement. In passing, we derive an extension of the technique to handle the situation in which replacement strategies are heterogeneous. We conclude that the technique performs very well, especially in those cases where the stochastic structure of the data can be diagnosed.

Employment opportunity can decrease adolescent childbearing within the underclass

Abstract Data from the evaluation of the Youth Incentive Entitlement Pilot Projects are used to test for a negative effect of economic opportunity upon adolescent childbearing among black youths from low-income households. We find that economic opportunity has a significant effect upon fertility, but that the direct effect of fertility upon school dropout is insignificant. The observed connection between fertility and dropout appears to reflect unmeasured factors for adolescent women which drive both variables. Methodological and policy implications of this finding are discussed.

Conception intervals and the substitution of fertility over time.

This paper applies the waiting-time regression methods of Olsen and Wolpin (1983) to an analysis of fertility. A utility maximizing model is set up and used to provide some guidance for an empirical analysis. The data are from an experimental guaranteed job program, the Youth Incentive Entitlement Pilot Project, aimed at young women 16 to 20 years old, from poverty-level families, and not yet high school graduates. The waiting-time regression method of estimation permits the youth in question to be used as her own control revealing how eligibility for the jobs program changes the durations of periods between live-birth conceptions. 3890 women surveyed had 1 birth, 429 had 2, 112 had 3, 26 had 4, and 7 had 5. Without this person specific control described here, the most important factors affecting fertility are number of siblings (negative effect), labor market attachment by parents, especially the father, and the presence of the natural father. With the person specific control, the results predicted from economic theory do emerge: even adolescent and young women consider the economic consequences of fertility reflected in effects of fertility when wages are high in favor of fertility with lower wages. Post program effects (taking place after youths lose eligibility for the program) are a rather rapid making up for foregone fertility, reducing likelihood of net reductions of total fertility.

Respondent Attrition Versus Data Attrition and Their Reduction

Panel attrition, or the loss of recruited members from a panel of respondents from whom longitudinal data are being collected, poses a serious risk to the validity of the results produced by panel data. Panel attrition is costly both monetarily and in terms of the effects on data quality and this chapter argues that there may be some benefit in shifting focus from minimizing respondent attrition from a panel to minimizing data attrition from the panel dataset. Using examples from important panels such as the Educational Longitudinal Survey (ELS) and the National Longitudinal Survey (NLS), it also proposes a number of strategies for minimizing the negative effects of panel attrition on longitudinal datasets, including targeted monetary incentives and long-term recontact strategies.