Scholarly Comments on Academic Economics

Response to “The Moving to Opportunity Experiment: What Do Heterogeneous Estimates of the Effect of Moving Imply About Causes?”

by , , and

Read this article

Access statistics
1,983 article downloads
4,061 complete issue downloads
Total: 6,044


Robert Kaestner (2020) presents a critique of how our work on the Moving to Opportunity (MTO) experiment (Chetty, Hendren, and Katz 2016) should be interpreted. His primary point is that our empirical analysis does not establish that reducing neighborhood poverty rates will improve children’s outcomes in adulthood. We found that moving to a lower-poverty neighborhood increases earnings in adulthood—an empirical result whose validity Kaestner (2020) does not dispute. We agree that one cannot conclude from this result that neighborhood poverty rates are the mechanism through which neighborhoods impact children’s long-run outcomes. However, we made this point in our 2016 article, where we wrote that the treatment effects of moving to lower-poverty neighborhoods should “be interpreted as the effect of changing a bundle of neighborhood attributes rather than any one feature of neighborhood environments.” Kaestner further notes that our point estimates by city are statistically less precise than the pooled estimates we report in our baseline specification, reproducing tables published in our paper. The fact that smaller samples yield larger standard errors is to be expected, but the estimates by site are not statistically distinguishable from the point estimates that pool across sites. Hence, there is no clear evidence for (or against) site-level heterogeneity. Finally, Kaestner analyzes heterogeneity across areas as a method for generating variation in poverty rates. This approach suffers from the very problem Kaestner highlights in his critique: there are other factors that vary across places beyond just their poverty rates. We conclude—as we did in Chetty, Hendren, and Katz (2016)—that moving to lower-poverty areas improves children’s outcomes in the long run, but the MTO experiment does not isolate the causal effect of any one neighborhood characteristic (such as poverty rates or peer effects) on upward mobility.