- Date:
- 2018-10-19
- Main contributors:
- Wing, Coady
- Summary:
- Classical applications of instrumental variables analysis are justified by structural models of behavior, and assumptions about the relationship between measured and unmeasured variables. Experimental and quasi-Experimental research designs present a partial alternative to structural modeling that is useful for answering certain types of research questions. It turns out that instrumental variables analysis can also help us make sense of several different research designs. This workshop will introduce the key assumptions involved in instrumental variables analysis from the perspective of research design. It will examine the way instrumental variables can play a role in the analysis of data from (i) classical randomized experiments, (ii) experiments that mix randomization and participant choice, and (iii) surveys that suffer from nonresponse. In each case, research designs justify some instrumental variable assumptions and not others. Examples and best practices for applied research will be discussed throughout.