Table 3. Baseline covariate values and the difference between intervention and control districts in the trend between 2001 and 2009, Cambodia.

Covariate	Mean valuea of covariate in 2005		Pc (for association between change in covariate over time and introduction of voucher scheme)
	Intervention districtsb (n = 22)	Control districts (n = 55)
Child
Child is male	0.482	0.503	0.728
Mother's age at childbirth ≤ 20 years	0.092	0.101	0.646
Mother's age at childbirth 21–34 years	0.730	0.731	0.889
Short interval from previous childbirth	0.134	0.120	0.177
First born	0.300	0.282	0.389
Birth order > 4	0.214	0.211	0.302
Mother
Mother's age at first marriage, years	19	19	0.665
Mother has used modern contraception	0.244	0.279	0.253
Mother completed primary education	0.599	0.529	0.909
Mother completed secondary education	0.164	0.165	0.722
Household
In poorest quintile by wealth indexd	0.294	0.340	0.645
In second poorest quintile by wealth indexd	0.229	0.224	0.326
In third poorest quintile by wealth indexd	0.184	0.148	0.842
In second richest quintile by wealth indexd	0.156	0.144	0.398
Age of household head, years	39	40	0.708
Male household head	0.837	0.853	0.766
District
Urban	0.189	0.189	0.776
Has health equity fund	0.091	0.517	0.260
Has government fee waiver scheme	0	0	0.197
Has any performance-based financing	0.108	0.406	0.556
No. of observations	651	1950	NA

NA, not applicable.

^a All covariates were dummy variables set to 1 if the respective description was satisfied and 0 otherwise, except for the mother's age at first marriage and the age of the household head.

^b Intervention districts were those that had implemented a voucher scheme by 2010. No district had a voucher scheme in place in 2005.

^c P-value for the null hypothesis that the change in the mean of each covariate between 2001 and 2009 was not associated with the introduction of a voucher scheme were derived using t tests. This involved regressing each covariate on a set of birth period and district fixed effects, as well as on an indicator for the operation of a voucher scheme within the period and district, and testing whether the coefficient for the voucher scheme indicator was zero.

^d The wealth index was devised by principal component analysis of a set of household assets and dwelling characteristics.

Note: Data were obtained from the 2005 and 2010 Cambodian Demographic and Health Surveys.