Estimating the Effectiveness of Health-Risk Communications with Propensity-Score Matching: Application to Arsenic Groundwater Contamination in Four US Locations

Table 5

Marginal effects from probit regressions on a model of point-of-use device adoption with observations weighted by the propensity to be informed by a friend about arsenic contamination of groundwater.

	Probit regression (with no propensity weights)		Probit regression with propensity weights
	Marg. eff.	S.E.	Marg. eff.	S.E.
	0.184	0.112	0.207**	0.098
INCOME/10,000	0.017*	0.009	0.024*	0.013
	0.053	0.067	0.273***	0.098
	−0.147*	0.082	−0.371***	0.118
	0.285***	0.082	0.293**	0.146
	0.287***	0.084	0.234	0.154
Number of observations	266		266
Adj-	0.113		0.198

Marginal effects of indicator variables calculated as a discrete change from 0 to 1.
The referent category is OK.
refer to statistical significance at less than or equal to 0.01, 0.05, and 0.1.