Back to overview

Identification and Estimation of Treatment Effects in Discrete Response Models

English title Identification and Estimation of Treatment Effects in Discrete Response Models
Applicant Boes Stefan
Number 124191
Funding scheme Fellowships for prospective researchers
Research institution Institute for Quantitative Social Science Harvard University
Institution of higher education Institution abroad - IACH
Main discipline Economics
Start/End 01.04.2009 - 30.04.2010
Show all

Keywords (6)

nonparametric bounds; ordered discrete choice; policy evaluation; random utility models; Bayesian econometrics; Dirichlet process

Lay Summary (English)

Lay summary
Lead:The evaluation of interventions is one of the most important aspects of active policy-making. This project develops robust statistical methods for impact evaluation if the outcomes of interest are discrete, such as being employed or not, the number of doctor visits, or the highest education achieved.Background:The aim of this project is the development of statistical methods for impact evaluation, with particular focus on discrete outcomes. This may be interpreted as inferring the causal effect of an endogenous regressor (the treatment) on a discrete response (the outcome). The counterfactual model (what would have happened under alternative policies) is the common framework in which to evaluate interventions. The research identifies two immediate consequences of discrete as opposed to continuous data. First, individual treatment effects defined by differences in potential outcomes are not informative given ordinal or nominal measurement of responses. Second, identification strategies for continuous data do not immediately generalize to discrete data due to the potential non-existence of moments. Distributional treatment effects are more meaningful in this context. However, without imposing strong parametric assumptions, these effects are only set, not point, identified. Methods and main results:The project discusses weak monotonicity, and convexity and concavity assumptions, and how they contribute to the identification of causal effects. Two main conclusions can be drawn from the analyses. First, the assumptions may substantially narrow the identified set, although point identification is only achieved in special cases. Second, the assumptions establish a tight connection between theory and data, as they can be well motivated by models of individual decision making, but without imposing functional form or distributional assumptions (which are rarely justified by the underlying theory).
Direct link to Lay Summary Last update: 21.02.2013

Responsible applicant and co-applicants

Associated projects

Number Title Start Funding scheme
139984 Identifying the causal effects of education on labor market outcomes 01.05.2012 International short research visits
150155 Perception of the workplace after disability onset: A longitudinal analysis for Switzerland 01.02.2015 Project funding (Div. I-III)