Table 4.

Regression Discontinuity Design results for the impact ofComunidades Solidarias Rurales (CSR) on the proportion of births with skilled attendance at birth, comparing 2006 entry to 2007 entry, using local linear regression

				Bandwidth
		|Distance|  $<$ 11		|Distance|  $<$ 8		|Distance|  $<$ 5
Kernel	Rectangular	Triangular	Rectangular	Triangular	Rectangular	Triangular
Coefficient estimate	0.104	0.164	0.148	0.180	0.207	0.171
	(0.063)	(0.064)	(0.059)	(0.072)	(0.092)	(0.085)
Bootstrapped $p$ value	.128	.046	.028	.066	.112	.164
Number of Obs.	388	388	360	360	257	257
Number of Clusters	18	18	14	14	10	10

Note. Standard errors clustered by municipality in parentheses below coefficients. “ $p$ value, unadjusted” is the $p$ value associated with the standard error; the bootstrapped $p$ value is based on the cluster bootstrap‐t procedure described by Cameron, Gelbach, and Miller (2008), and is not prone to false rejection of the null hypothesis with small numbers of clusters, the unadjusted $p$ value is. Bandwidth refers to the Euclidean distance to the boundary between clusters, where observations (obs.) outside this distance are excluded from the sample. Control variables are listed in Table 1. Source: CSR evaluation first and second survey.