considerably more common than abortion related causes, at around 34 and 3 per 100,000 fertile-aged women respectively. Rates of morbidity in the puerperium are again higher, at 110 per 100,000 fertile-aged women.13 On average, each municipality prescribes 21 EC pills per year, though larger municipalities prescribe many more, with maximum disbursements of 1029. The number of pills refused is considerably lower, at around 2.4 per municipality on average, though once again we note that there are municipalities who refuse many requests, with a maximum of 914 per year. 4.2 Binary Treatment Measures Capturing EC Roll-out In Figure 3 we present main event study specifications following equation 1, reporting 90 and (as standard) 95% confidence intervals. These document how rates of morbidities respond around the dates of first reported EC pill adoption in each municipality. Here we consider a baseline model without time-varying controls, and discuss a number of alternative specifications in section 4.6. Figure 3a displays the event study for rates of haemorrhage early in pregnancy, with typically quite flat pre-reform coefficient estimates. While there is a single pre-period in which a downwards blip is observed, in general these leads surround the dotted red line indicating 0 pre-reform impact. In the post-reform period, in the case of haemorrhage, while we observe a gradual reduction in rates of haemorrhage (reaching a high of a 15–20 per 100,000 unit reduction), these are not estimated with 13As we discuss in section 4.6, we will consider this in placebo test. 13 sufficient precision to reject a zero impact at typical levels of statistical significance, potentially given that there is considerable heterogeneity in actual EC usage, as discussed later. In the case of abortion morbidity (Figure 3b), we once again observe estimates quite tightly clustered around zero in the pre-reform period, suggesting that there were no prevailing differences between early and later adopters, and once again observe a reduction after the adoption of the EC pill. In this case, this reduction is immediately statistically significant, at around -1.3 fewer cases of abortion morbidity per 100,000 women, growing to reach an impact of up to 5 fewer cases of abortion morbidity on average, though as in the case of haemorrhage morbidity the standard errors grow as the estimated effect size increases. It is important to note however, that with the case of the first year postadoption, significant impacts reflect changes in rates of abortion morbidity in municipalities that have adopted the EC pill compared with municipalities that have not yet adopted. 4.3 Combining Binary Treatments with Intensity of EC Pill Usage To examine the intensity of use of the EC pill and the impact of total disbursements on health, we first estimate the DD specification described in equation 2 where our dependent variable is now the number of pills disbursed per 1,000 women. These models are presented in Table 2. In panel A, we estimate that for each additional pill disbursed per 1,000 women, rates of haemorrhage fall by 0.439 cases per 100,000 in the full population of women. We additionally consider rates by age, observing that these results are quite transversal across age groups, and are statistically distinguishable from 0 for ages 20-24, 25-29 and 35-39. In section 4.5 we put the magnitude of these effects in context. In Panel B we present DD estimates examining intensity of pill usage and the impacts on abortion related morbidity. In this case, unlike in binary models, we do not observe a large or consistent reduction in rates of abortion with increased usage of the pill. Indeed, the majority of the estimates suggest an impact of less than 0.1 cases of abortion specific morbidity per 14 100,000 women. The DD estimate for all women in column 1 suggests a small positive estimate of 0.029 cases, though not statistically distinguishable from zero. Given that single coefficient DD models potentially suffer from bias in situations such as the case in Chile where treatment occurs in different areas at different times14 we combine the binary model of EC rollout with measures of EC pill usage intensity in an interacted event study specification, displaying results in Figure 4. Each set of coefficients are estimated in a single event study specification, and in each case, the baseline period is 1 year prior to the moment of reform implementation. In Figure 4a, we document that high-intensity pill municipalities have the largest reduction in haemorrhage morbidity following the introduction of the EC pill. Over time, a significant reduction emerges in rates of haemorrhage, and this is driven specifically by this high-intensity group. While medium and low-intensity municipalities also have a reduction in rates of morbidity following the reform (with no similar divergence in the pre-reform period), these are never statistically significant. In general here when documenting heterogeneity by intensity of EC pill disbursements we note that estimates are considerably less precise. Thus, in the case of haemorrhage, while a clear reduction is observed from year 2 post-adoption onwards, this only becomes significant at 95% 5 years post-adoption. We replicate these results for abortion related morbidity in Figure 4b. In this case, and in line with the fact that there are fewer cases of abortion