Mechanism Experiments and Policy Evaluations

摘要

Randomized controlled trials are increasingly used to evaluate policies. To cite one example, in 2002, the U.S. Department of Education founded the Institute of Education Sciences with a primary focus on running experiments; its annual budget is now about $700 million (U.S. Department of Education, 2010). This trend has been spurred by numerous independent groups that promote policy experimentation, including the Coalition for Evidence-Based Policy, the Campbell Collaboration (an international network of researchers hosted by the Norwegian Knowledge Centre for the Health Services), the Poverty Action Lab, and Innovations for Poverty Action. However, a vigorous debate has arisen about the value of experimental methods for informing policy (for example, Angrist and Pischke, 2009, 2010; Banerjee and Duflo, 2009; Deaton, 2010; Heckman, 2010; Imbens, 2010). We argue this debate has often been framed too narrowly as a choice between experimental and nonexperimental methods. An important distinction between experimental methods merits greater attention.