Table 7.

Effects of user fee removal using individual‐level data

	(1) Simple DiD model			(2) Matching and DiD (MDiD)
	Diff (T‐C) baseline	Diff (T‐C) follow‐up	DiD	Diff (T‐C) baseline	Diff (T‐C) follow‐up	DiD
Panel A: % seeking care
Estimated effect	−0.020	0.026	0.046	0.023	0.059***	0.036
SE	(0.025)	(0.034)	(0.034)	(0.023)	(0.04)	(0.045)
N	10,295	7,841	18,136	9,711	6,859	16,570
Panel B: % choosing government or mission provider
Estimated effect	0.036	0.110***	0.075 ***	0.026	0.072***	0.045 *
SE	(0.024)	(0.026)	(0.024)	(0.023)	(0.016)	(0.026)
N	5,975	4,817	10,792	5,685	4,259	9,944
Panel C: Ln(oop)
Estimated effect	−0.771**	−2.755***	−1.985 ***	−0.195	−2.141***	−1.946 ***
% change	−53.8%	−93.7%	−86.3%	−17.7%	−88.2%	−85.7%
SE	(0.370)	(0.467)	(0.367)	(0.161)	(0.165)	(0.374)
N	8,620	6,806	15,426	8,144	6,016	14,170
Panel D: % buying drugs in the private sector
Estimated effect	−0.139***	−0.168***	−0.029	−0.052	−0.077**	−0.025
SE	(0.052)	(0.050)	(0.038)	(0.046)	(0.036)	(0.043)
N	8,589	6,791	15,380	8,127	6,003	14,130

Note. Effects of the policy are reported in bold and percentage change are reported in italics. The propensity score was estimated using the same covariates at the individual level than the ones used in the synthetic control. Survey sampling weights are used. SE clustered at the district level in bracket. Estimated presented in (2) are based on propensity score matching using Epanechnikov kernel weights. DiD = difference‐in‐differences

^* Statistically significant at the 1% statistical significance level.

^** Statistically significant at the 5% significance level.

^*** Statistically significant at the 10% significance level.