This research continues to apply newly developed statistical methods for life-history data to interesting questions in the economics of education and labor markets. These new methods help study the effects of unemployment insurance benefit levels on re-employment; gender, age, and race differences in re-employment behavior following job displacement; the retirement behavior of older individuals after job displacement; the effect of financial aid on the drop-out and graduation behavior of college students; whether delaying college enrollment after high school completion affects college completion; and the effect of advanced notification of an impending job loss on subsequent re-employment.
Many of these analyses involve policy simulations. Up to this point, the computational complexity of the problem has made computation of the standard errors of the policy simulation difficult. To ease computations, bootstrap methods are being developed to calculate the standard errors of these policy experiments.
Previous analyses of life history data have assumed that spell or duration data is continuous even though economic data is usually grouped, and duration data involves a number of ties. Because of this, statistical estimation techniques must be developed specifically for grouped duration data. Although such models have been developed to analyze single spell data, none are currently available for multi-spell duration data, and few have been developed for competing risks models.
Both the competing risks and multi-spell hazard models used in this research will be flexible-not only in the sense that they put no a priori functional form restrictions on the baseline hazard functions, but also in the sense that they allow for the effect of regressors on the hazard function to vary over time. In multiple spell hazard models, unobserved heterogeneity is accounted for using a multi-dimensional mass-point mixing distribution. The competing risks model accommodates dependent risks by allowing for (possibly correlated) unobserved heterogeneity to affect the latent marginal hazard functions. Again, a multi-dimensional mass-point mixing distribution is used to model such unobserved heterogeneity. Finally, the selectivity-corrected hazard and competing risks models employ multivariate mass-point mixing distributions.
99/10 |
"Tackling Endogeneity: Alternatives for Analysis of Women's Employment and Fertility," R. Connelly, D.S. DeGraff, D. Levison, and B.P. McCall, University of Minnesota Supercomputing Institute Research Report UMSI 99/10, February 1999. |
These estimation techniques are now being applied to continuing studies of the effect of financial aid on college drop-out and graduation behavior. This work is using administrative data from the University of Minnesota. A new cohort of data has been provided that better tracks the financial aid application behavior of students over time. A multi-spell hazard model is also being employed to study whether delaying college enrollment after high school completion affects college completion. Again, in this project, policy simulations are performed, and the bootstrap standard errors of those policy simulations are calculated.
|
|
URL: http://www.msi.umn.edu/about/publications/annualreport/ar2000/depts/CSOM/Ind_Rel/mccall.html |
|
| This page last modified on Friday, 30-May-2008 16:13:59 CDT | ||
| Please direct questions or problems to help@msi.umn.edu | ||
|
Website related questions or problems should be directed to
webmaster@msi.umn.edu
The Supercomputing Institute does not collect personal information on visitors to our website. For the University of Minnesota policy, see www.privacy.umn.edu. © 2001 by the Regents of the University of Minnesota |
||