Randomization and Social Policy Evaluation

This paper considers the recent case for randomized social experimentation and contrasts it with older cases for social experimentation. The recent case eschews behavioral models, assumes that certain mean differences in outcomes are the parameters of interest to evaluators and assumes that randomization does not disrupt the social program being analyzed. Conditions under which program disruption effects are of no consequence are presented. Even in the absence of randomization bias, ideal experimental data cannot estimate median (other quantile) differences between treated and untreated persons without invoking supplementary statistical assumptions. The recent case for randomized experimentation does not address the choice of the appropriate stage in a multistage program at which randomization should be conducted. Evidence on randomization bias is presented.

You may purchase this paper on-line in .pdf format from SSRN.com ($5) for electronic delivery.

Machine-readable bibliographic record - MARC, RIS, BibTeX