Anna Malinovskaya has posted “Understanding the Native American Tribal ‘Disenrollment Epidemic’: An IV Approach” on SSRN.
Here is the abstract:
Recently, over 80 Native American tribes have banned or disenrolled members of their tribes or denied citizenship to eligible individuals. This phenomenon has received media attention nationwide, and even the term the “disenrollment epidemic” was coined. Many speculate that at least some of it is driven by political struggles over multi-million dollar revenues of tribal casinos, which are sometimes distributed in per capita payments to all tribal members. In this paper, we test whether gaming has been driving disenrollments, and since a tribe’s involvement in gaming might be endogenous, we employ an instrumental variable approach. In particular, we use machine learning to select an optimal subset of instruments for a Native American tribe operating a casino from the set of potential instruments all plausibly meeting the exclusion restriction and associated with the geographical characteristics of reservations, such as their proximity to an MSA, an interstate highway, or a border of a neighboring state with no brick and mortar casinos. We find that a tribe’s involvement in gaming leads to a large and statistically significant increase in the probability of the tribe experiencing a disenrollment episode.
This paper sought to understand if tribes’ involvement in the gaming industry, particularly wealth from per capita distribution of gaming revenues made possible by this involvement, has been the primary factor driving disenrollments and other types of dismemberment episodes. Both gaming and per capita distributions of gaming revenues are likely to be endogenous. Although we did not find an instrument for per capita distributions, we identified a set of instruments for gaming, and used them to understand, albeit indirectly, whether gaming has been driving disenrollments (the direct approach would be instrumenting for per capita distributions rather than gaming). Although this approach has its limitations
(as discussed in the Empirical Strategy section, it represents, to the best of our knowledge, the first attempt to identify a causal link between gaming and dismemberment in Indian tribes. Additionally, the consistency of results across our OLS and IV estimates, as well as across several sub-samples, is encouraging.
This research could be strengthened by instrumenting for per capita distribution of gaming revenues directly, though finding an appropriate instrument might be challenging. It could also be strengthened by finding instruments that would pass the Weak Instruments test for the full sample (327 tribes) that is likely to produce less biased IV estimates, or by using a different quasi-experimental empirical approach that would overcome the limitations associated with IVs when the sample is relatively small.