Table 6.

Effects accounting for pretreatment match quality

	Seek care	Public facility	OOP medical expenses	Private drug seller
Panel A: Weighted average by the logarithm of the inverse of the RMSPE
Estimated effect	0.007	0.016	−2.177	−0.062
Panel B: Restricting the sample to high‐quality matches
Number of perfect matches	9	19	13	28
National effect weighted by district size	0.042	0.024	−2.085	−0.080

Note. Perfect‐quality matches means that RMSPE < 0.01 for binary outcomes and <0.001 for ln(OOP). RMSE = root mean squared prediction error.