Except for Louisiana, where we observe a noticeable spike in both overall new enrollments and ELE-based new enrollments around the adoption of the policy, we see little descriptive evidence from the new enrollment analysis that ELE increased the number of children added to coverage (although among children who were added, we observe sizable differences in the distribution of age by enrollment pathway, suggesting that ELE may have been successful at enrolling some traditionally hard-to-reach older children). We likewise see little evidence of differences in retention between ELE and non-ELE new enrollees, again except for Louisiana, where roughly half of all ELE new enrollees exited coverage within 12 months due to a temporary policy requiring them to use their Medicaid card within 12 months. However, because the descriptive analysis can provide no estimate of the counterfactual—that is, what enrollment in ELE states would have been without the policy—it may significantly underestimate the effects of the policy on coverage. Moreover, because the descriptive analysis focused only on new enrollments tied directly to ELE, it could not account for enrollment gains arising indirectly from the policy, nor could it fully assess those affected by ELE renewal processes. These limitations are relevant for several states, including Alabama and Louisiana (both of which have adopted ELE for renewals) and New Jersey (where ELE has led to more extensive use of tax information, a “spillover” that cannot be identified in our data).
In contrast to these descriptive findings, the multivariate analysis, which combines aggregate total enrollment data for both ELE and non-ELE states, finds significant evidence of an ELE effect on total enrollment. Drawing on a range of models, the analysis finds that ELE increased total Medicaid enrollment by an estimated 5.5 percent and Medicaid and CHIP enrollment combined by an estimated 4.2 percent. Findings for Medicaid were notably robust, showing little variation in magnitude and remaining statistically significant across many sensitivity tests. However, as with all impact studies using regression methods to estimate the counterfactual, our multivariate findings might be biased from unobserved factors that differ between ELE and non-ELE states over time.
Taken together, these first-year findings suggest that ELE can be an effective tool for enrolling children into coverage. However, because evidence that ELE significantly increased enrollment is stronger in the multivariate findings than in the descriptive findings, it will be critical to revisit both analyses in the evaluation’s second year, as more post-ELE enrollment data become available. The case studies will further support these second-year analyses, permitting a more thorough understanding of other simplifications and the trends in enrollment under ELE.