Who benefits from increased service utilisation? Examining the distributional effects of payment for performance in Tanzania

Table 8 Effect of P4P on service utilisation by subgroups –using the non–linear logit model

Subgrouping variables	Institutional deliveries			Uptake of IPT2
	Average subgroup effect		Differential effect test (p-value)	Average subgroup effect		Differential effect test (p-value)
	N	(dy/dx)	Differential effect test (p-value)	N	(dy/dx)	Differential effect test (p-value)
Marital status
Married	3385	9.2^a	(p = 0.503)	3253	9.2^a	(p = 0.935)
Unmarried	1338	13.3^a		1481	9.8^c
Maternal age
Younger below the median age	2361	11.2^a	(p = 0.492)	2336	9.2^b	(p = 0.830)
Older above the median age	2325	9.1^a		2421	9.5^b
Education
Some education	4021	10.9^a	(p = 0.070)	3877	8.6^a	(p = 0.793)
No education/illiterate	900	9.1		816	16.5^c
Occupation
Farmer	2638	13.4^a	(p = 0.590)	2396	16.0^a	(p = 0.149)
Non-farmer	2126	7.5^b		2295	5.3
Religion
Muslim	3991	10.8^a	(p = 0.497)	3614	9.7^a	(p = 0.554)
Non-Muslim	980	5.6		1061	7.8
Parity/births
One birth	1180	15.2^a	(p = 0.122)	1476	9.9^c	(p = 0.939)
Two or more births	3436	9.3^a		3247	10.0^a
Household size by members
Small size (< 5)	2381	7.3^b	(p = 0.320)	2464	7.6^c	(p = 0.903)
Large size (≥5)	2299	12.8^a		2281	9.1^b
Health insurance
Insured	171	−20.7	(p = 0.012)	315	18.3	(p = 0.900)
Uninsured	4820	11.1^a		4328	10.1^a
Household wealth status
Tercile 1 (poorest)	1656	13.4^b	(p = 0.894)	1508	13.2^b	(p = 0.145)
Tercile 2 (middle)	1528	12.7^a	(p = 0.737)	1539	17.1^a	(p = 0.106)
Tercile 3 (least poor)	1066	8.2^b	Reference	1599	2.4	Reference
Place of residence
Rural district	4387	11.3^a	(p = 0.152)	3851	11.2^a	(p = 0.268)
Urban district	787	1.6		906	1.7

Non-linear logit model with FE, covariates, clustering at HF level; Logit with FE cuts down the sample size; dy/dx is the estimated partial P4P effect on a specific subgroup in terms of marginal effect after controlling for a year dummy, facility-fixed effects, and individual and household-level covariates (age, education, occupation, religion, marital status, parity, health insurance status, household size, and household wealth status); Each cell for dy/dx and differential effect reports the result from a separate regression; Differential effect test is a t-test of the null that the coefficient on the three-way interaction between the P4P effect and subgrouping indicator is zero; ^a denotes significance at 1%, ^bat 5%, and ^c at 10% level

ISSN: 1475-9276