Our analyses start by documenting the evolution of state-level Medicaid eligi- bility. Online Appendix Figure A.2 shows the time path of eligibility from 1975 to 1988 for each of the fifty states. It is immediately clear that states exhibit differ- ent expansion patterns, and that these patterns fall into two dominant types: many states experienced small, gradual increases in eligibility throughout the period (e.g., California and Rhode Island), while others exhibit flat, low levels of prenatal cov- erage during the 1970s that are later punctuated by a large, abrupt, increase (e.g., South Carolina and Maine). This is exactly what the discussion in Section II should lead us to expect. Specifically, in early years, states that initially embraced strict eligibility criteria and did not avail themselves of the state options, exhibit low levels of eligibility. In later years, we observe discrete jumps in these states that correspond to a large, voluntary state-level expansion or one of the later federally mandated expansions. In contrast, states that took early advantage of the existing state options to cover low-income pregnant women have relatively high levels of eligibility from the beginning of the study period. As a result, these states exhibit smaller, more gradual increases that are the state by year controls we implement in our model, described in detail below (column 4), and weight by the number of births in each cell (column 5). 15 An earlier version of this paper used this approach. That version is available upon request. 16 Information on eligibility prior to 1975 is very spotty, therefore we begin our sample period in 1975. VOL. 113 NO. 1 EAST ET AL.: EARLY LIFE EXPOSURE TO MEDICAID 109 less pronounced in the wake of the subsequent state and federal changes. These two different eligibility trajectories will be demonstrated visually in the next subsection. B. Classification of Treatment and Control States Using online Appendix Figure A.2 as a guide, we organize states into "treated" states (those that experienced a sharp jump in eligibility) and "control" states (those for whom eligibility trended smoothly). Our event study is based on differences in the evolution of prenatal eligibility across these two groups. The treatment group consists of the 28 states in which eligibility was stagnant for a minimum of the five years between 1975 and 1979, and that later experienced a large positive shock. We focus on shocks that took place between 1980 and 1985 to ensure that we have at least four "postevent" years over which to observe the expansions' effects. We include the 22 remaining states as "control" states to help identify secular changes that were separate from the Medicaid expansions.17 Figure 2 shows which states we classify as control states, which states we classify as treatment states, and the timing of the expansion for the treatment states. Given that "treatment" requires a large, abrupt increase in eligibility, it is unsurprising that treated states are disproportionately located in the South and Midwest; Medicaid was less generous in the pre-expansion period in these states, and transfer policies have historically been less generous in these parts of the country as well.18 Figure 3 panel A provides further insights into the sources of our identifying variation by combining all of the treated states together, and documenting how the changes in eligibility break down by source. The sources are shown additively from most to least restrictive (shown from lowest to highest lines on the graph), and include changes in eligibility driven by changes in AFDC rules (solid blue line), targeted expansions occurring via state options (long dashed red line), targeted expansions occurring under federal mandates (short dashed green line), and broader income-based expansions (dash-dotted yellow line). The last category is concen- trated at the end of the period, and reflects optional state expansions that included pregnant women with family incomes up to the poverty line.19 It is easy to see that prior to the 1980s, Medicaid eligibility in treated states was largely determined by state AFDC rules, and that these states were less likely to 17 We did not have a specific algorithm to define treatment and control states or event time. Rather, each of the four authors individually classified states based on online Appendix Figure A.2, and then we decided on a consistent classification as a group. We relied on measures of simulated eligibility, rather than actual eligibility, to ensure that changes in eligibility are driven by policy changes and not by demographic or economic shifts, and there were few classification discrepancies across the authors. States with large changes outside the 1980-1985 period are included in the control group, but the few states that fall into this category do not affect our results: we show that estimates are similar when we drop the control states from the analysis. 18 Prior work has found that racism played a role in generating a less generous safety net (Lee and Roemer 2006), and, online Appendix Table A.2 shows that treated states have slightly higher Black populations than control states. We multiply all proportions by 100 for ease of presentation. There are a few small differences in other demo- graphic and economic characteristics across treatment and control states. Our analyses include state fixed effects to account for any time-invariant differences, and we show that our estimates are robust to including a large number of state-year varying economic and policy controls, and region-year fixed effects that control for common shocks to outcomes within regions over time. 19 Under the Omnibus Budget Reconciliation Act of 1986, states were allowed to expand Medicaid to preg- nant women with family incomes below 100 percent of the FPL starting in April 1987. See additional discussion in online Appendix A.I, along with more detailed breakouts of the different eligibility paths in online Appendix Figures A.3 and A.4. 110 THE AMERICAN ECONOMIC REVIEW JANUARY 2023 Figure 2. Treatment and Control States Notes: See text for further details. Not pictured is Alaska (1982 expander) and Hawaii (control). Arizona is omitted due to the late start date of their Medicaid program. take up state options than the control states. Beginning in the early 1980s, a few treatment states increased eligibility through the state options. This was followed by a sizable bump under the federal mandates. Between 1975 and 1988, prenatal eligibility roughly doubled in treatment states, from just under 9 percent to 18 per- cent. In contrast, Figure 3 panel B shows that control states started off with higher eligibility, because they had already implemented the state options, and that eligi- bility gradually increased over time as more options were adopted. As a result, the federal mandates had smaller effects on eligibility. Not only is the change in eligibility greater for treatment states than control states, but the women affected are more disadvantaged due to the more stringent baseline eligibility rules for these states. To demonstrate this, we compare the char- acteristics of women who gained eligibility over this period in both groups of states in online Appendix Table A.3.20 The women in treated states are noticeably more disadvantaged in terms of their family income and poverty rate than women in con- trol states. Notice that nearly one-half of the women gaining eligibility in treatment states have family incomes below 50 percent of the FPL, versus only 13 percent in the control states. Women in the treatment states are also more likely to be single and non-White. Finally, Figure 4 demonstrates that among treated states there is also variation in when the large jumps occur. States are grouped together by the initial year of the abrupt increase (the "expansion year"), and eligibility is plotted relative to the year prior to expansion, with the number of states expanding in each year shown in parentheses. While the most common expansion years are 1982 (11 states) and 1985 20 Using the CPS, we calculate mean characteristics for the women in treatment states who are eligible during the postperiod (defined using each state's specific treatment timing), but who were not eligible under the rules that were in place during the last pretreatment year. We calculate mean characteristics for women in control states who are eligible during 1982-1988 but were not eligible under the rules in place in 1981. Control 1980 expanders 1981 expanders 1982 expanders 1983 expanders 1984 expanders 1985 expanders Excluded VOL. 113 NO. 1 EAST ET AL.: EARLY LIFE EXPOSURE TO MEDICAID 111 Panel A. Treated states 20 15 10 5 1975 1980 1985 Panel B. Control states 25 20 15 10 5 1975 1980 1985 Figure 3. Prenatal Eligibility by Source and Treatment Status, 1975-1988 Notes: Authors' calculation from the CPS and Medicaid eligibility rules. Each line represents Medicaid eligibility through each of the state options and federal mandates. Specifically, we construct this figure by calculating the per- cent of women who would be eligible under each pathway in an additive fashion that reflects the order of eligibil- ity pathways from lowest to highest pathway on the figure. These pathways are discussed in more detail in online Appendix A. All estimates are reported in percents. (7 states), when federal policy provided states with new options to expand cover- age (1982), and mandated expansion of coverage to more women (1985), there are also states that experience large expansions in every other year between 1980 and 1985.21 In our main event study framework, we rely on the differential timing of expansions across states, as well as differences between treated and control states, to identify the effects of Medicaid separately from time trends. Please summarize this