Extended Data Table 2 Average treatment effect on the treated on the main outcomes with first stage

From: A randomized controlled trial on the effect of administrative burden and information costs on social inequalities in early childcare access in France

  1. First stage reports OLS estimates of offering support and actual support take-up. Average treatment effects on the treated are estimated jointly for both comparisons by instrumenting administrative support in each comparison sample through assignment to information and support (centred by the pairwise instrument propensity score), interacted with the comparison sample dummy and block-by-comparison fixed effects. Standard errors are cluster-heteroskedasticity robust and adjusted at the block level. They are reported in parentheses. We report exact p-values (p) based on two-sided t-tests, and Padjs (Padj). Padj and confidence intervals account for simultaneous inference using the Westfall method. Confidence intervals are reported between brackets. Each column jointly estimates the program’s effects using fully saturated stacked regressions. The average counterfactual indicates the average of the untreated compliers and is estimated by TSLS with (1-D)*Y as the outcome, and (1-D) as the treatment variable instrumented by the centred assignment. Joint significance test of null effect using the Chi-2 test and p-values is reported at the bottom of the table.
Back to article page