Instrumental Variables

kwai chang caine: From a single action, you draw an entire universe.

Kung Fu, Season 1, Episode 1

Our Path

Statistical control through regression may fail to produce convincing estimates of causal

effects. Luckily, other paths lead to other things equal. Just as in randomized trials, the forces of nature, including human nature, sometimes manipulate treatment in a manner that obviates the need for controls. Such forces are rarely the only source of variation in treatment, but this is an obstacle easily surmounted. The instrumental variables (IV) method harnesses partial or incomplete random assignment, whether naturally occuring or generated by researchers. We illustrate this important idea three ways. The first evaluates an American education innovation—charter schools—with an elementary IV analysis that exploits randomized school admissions lotteries. A second IV application, examining the question of how best to respond to domestic violence, shows how IV can be used to analyze field experiments in which the subjects randomly assigned to treatment are free to opt out. The third application explores the long-run effects of growing up in a larger or smaller family. This application illustrates two-stage least squares (2SLS), an elaboration on the IV method and one of our most powerful tools.

Leave a reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>