Maison des Sciences Humaines
11, Porte des Sciences
L-4366 Esch-sur-Alzette / Belval
LISER- MSH 1st floor conference room
This course provides an introduction into the analysis of transition data (=duration models) and into the Timing of Events (ToE) method. Duration analysis draws heavily on statistical methods developed in industrial engineering where they are used to describe the useful lives of various machines and in the biomedical sciences to describe events such as the survival times of heart transplant recipients. In economics they are typically used to analyze the determinants of the duration of unemployment and employment spells, but also of the time that elapses until the realization of a specific event, such as the closing down of firms. The analysis of duration models is complicated by issues of incomplete observation (i.e. left and right censoring) and of dynamic sorting induced by unobserved heterogeneity. The course introduces provides an overview of the issues involved in the analysis of transition data.
Based on the ToE method researchers can estimate the causal impact of an event on the duration of spell that is occupied by an individual or firm. For instance, the method allows the researcher to identify the causal impact of a benefit sanction or of participation in a training program on the speed of transition from unemployment to employment. The advantage of the method is that, in contrast to the method of Instrumental Variables, it does not require exclusion restrictions to identify the causal impact of a treatment on the outcome of interest. But identification is not a free lunch. The course will clarify which assumptions are required for identification.
The course is structured as follows:
- Basic Concepts: hazard rate and survival function, discrete versus continuous time data.
- Unobserved heterogeneity
- Multivariate duration models
- Timing of Events method.
There will be no time for discussing the practical implementation of the method in software packages.
The student is assumed to know the principle of Maximum Likelihood estimation, ordinary least squares and basic statistics.
Bart Cockx is a Full Professor at Ghent University and Research Fellow at IRES (Université Catholique de Louvain), IZA-Bonn and CESifo-Munich. His main research interests are labour economics, active labour market policies, unemployment insurance, applied microeconometrics and education economics. He is teaching non-linear econometrics and causal inferences for program evaluation at Ghent University and at the Belgian Doctoral School of Economics. He has published in several journals among which we mention Journal of Labor Economics, The Economic Journal, The Review of Economics and Statistics, The Journal of the Royal Statistical Society: Series A (Statistics in Society), Journal of Public Economics, Journal of Health Economics, Journal of Applied Econometrics.