Search Results
Title: Three Essays on Semiparametric Models of Dynamic Discrete Choice, Program Evaluation, and the College Premium in the Eighties
Resulting in 1 citation.
| 1. |
Taber, Christopher Robert |
|
Three Essays on Semiparametric Models of Dynamic Discrete Choice, Program Evaluation, and the College Premium in the Eighties Ph.D. Dissertation, The University of Chicago, 1995 Cohort(s): NLSY79 Publisher: UMI - University Microfilms, Bell and Howell Information and Learning Keyword(s): Armed Forces Qualifications Test (AFQT); College Education; Earnings; Endogeneity; Heterogeneity; High School Completion/Graduates; High School Dropouts; Modeling; Schooling; Skills; Wage Gap The first chapter takes the first step towards semiparametric estimation of discrete choice dynamic programming models by establishing sufficient conditions for their identification. I develop a specification in which the distribution of the error terms is unrestricted. In addition I allow the agents' information sets to be heterogeneous where this private information may covary with the error terms in a very general way. I treat both a discrete choice version of the model and a semiparametric tobit version in which there exist endogenous random variables which are only observed conditional on the choices made. I show that the parameters in my model are identified with essentially no restrictions on the distribution of the error terms and on the information structure except that they are independent of the regressors. I also show that additional restrictions are necessary to assure identification of the full model and I provide two sets of conditions that suffice. The second chapter applies this model to schooling decision. Its goal is to distinguish whether the rising wage gap between college educated workers and other high school graduates during the nineteen eighties results primarily from an increase in the value of skills learned in college or from an increase in the value of skills typically possessed by college students prior to entering college (i.e. ability). I attempt to make this distinction in two ways. As a preliminary exercise, I use Armed Forces Qualifying Test (AFQT) scores in the National Longitudinal Survey of Youth (NLSY) to proxy for ability and examine how controlling for observed ability influences the trend in the college premium. Secondly, I control for unobserved ability by specifying a structural discrete choice dynamic programming model in which a student who is deciding whether to drop out of high school takes into account both the direct value of graduating high school and the value of the option to att end college. I estimate the model both fully parametrically and semiparametrically using nonparametric maximum likelihood. The econometric specification provides an interpretable framework for distinguishing between the two alternative hypotheses. The methodology uses only one step which improves efficiency by incorporating all of the information available from the longitudinal data. I summarize these results by documenting the change in the expected gain in earnings from attending college for those individuals who are indifferent about attending college. I find that including AFQT scores in the regressions has no influence on the rise in the college premium. However, controlling for unobserved ability eliminates it. The third chapter is written jointly with James Heckman and Jeffery Smith. It considers the evaluation of social programs using experimental data in the presence of dropouts. We begin with a popular estimator that produces estimates of the mean impact of treatment on the treated in experiments with dropouts. In experiments in which the dropouts receive none of the treatment prior to dropping out, this estimator should work very well. However, in cases where the dropouts may have received some of the treatment prior to leaving the experiment, the estimator may not work well. This paper addresses this concern and the issues it raises. We motivate the discussion with the recent experimental evaluation of the JTPA program which used this estimator even though most of the dropouts received some training prior to dropping out. (Abstract shortened by UMI.) |
|
|
Bibliography Citation
Taber, Christopher Robert. Three Essays on Semiparametric Models of Dynamic Discrete Choice, Program Evaluation, and the College Premium in the Eighties. Ph.D. Dissertation, The University of Chicago, 1995. |