On Friday, January 17, Shane presented on synthetic control methods, a technique he used in his Ph.D. research. When designing experiments to determine the effect of a treatment, researchers would ideally like to compare a group that receives the treatment—the treated group—with a group that does not receive the treatment but is otherwise identical in every way—the control group. However, in many real world situations, it is often impossible to find an entity for comparison that is perfectly like the treated group. Synthetic control methods allow researchers to approximate a control group by using a weighted average of other entities that may share some characteristics with the treated group but not all. Since the synthetic control is not subjected to the treatment, researchers can assume any differences post-treatment between the treated group and the synthetic control are attributable to the treatment of interest. To use synthetic controls in Stata, see the synth package.
Shane applied synthetic control methods to study the effect of a large influx of money to the Wyoming school system in the 2006-2007 school year. He created a synthetic control group from the weighted average of a number of other states’ school systems and found that this capital increase did not appear to improve Wyoming students’ graduation rates or math aptitude scores.
