represents the ECCE effect on academic achievement. The ECCE effect for the unmatched sample is almost non-existent, demonstrating that it is impossible to observe any treatment effects if one does not control for pretreatment variables in order to reduce the selection bias.Following the estimates for the unmatched sample, Table 3 presents the difference in means between groups and standard error. These estimates show that ECCE is positively associated with academic gains on all three tests: math, reading, and social sciences. Children who attended ECCE scored on average 12.8 points(0.23 SD) in math, 10.7 points (0.19 SD) in reading and 9.4 points(0.18 SD) in social sciences higher than those who did not attend.The ECCE effects and their standard errors derived from the regression-adjusted model are shown in the last two columns of Table 3. Table 4 displays regression analysis estimates for ECCE effects on academic achievement, as well as estimates for all post-treatment variables. After controlling for post-treatment variables,as shown in Tables 3 and 4, children who participated in ECCEscored on average 11.9 points (0.21 SD) in math, 9.4 points (0.17SD) in reading, and 8.5 points (0.17 SD) in social sciences higherthan children who did not participate in ECCE programs. The dif-ferences between both analyses are small, which implies thatpost-treatment variables are playing a minor role in modifying theeffect ECCE programs have on children’s academic outcomes.