. 2014 Feb 19;178(1):101–124. doi: 10.1111/rssa.12048

Table 3.

Specification and robustness checks for the baseline effects of health coverage on mortality^†

	Results for under-5 mortality rate, IV-2SLS, (1)	Results for female mortality rate (adult), IV-2SLS, (2)	Results for male mortality rate (adult), IV-2SLS, (3)
Government health spending
Baseline	−13.193	−2.583	−2.210
Baseline	(0.018)	(0.050)	(0.025)
(a) With first and second lags (sum of all coefficients)	−7.860	−1.377	−1.966
(a) With first and second lags (sum of all coefficients)	(0.034)	(0.212)	(0.043)
(b) Excluding outliers	−16.291	−2.877	−2.166
(b) Excluding outliers	(0.022)	(0.022)	(0.046)
(c) Excluding years before 1999	−8.341	−2.163	−2.303
(c) Excluding years before 1999	(0.026)	(0.050)	(0.015)
OOP health spending
Baseline		−23.385	−15.545
Baseline		(0.040)	(0.016)
(a) With first and second lags (sum of all coefficients)		−25.693	−14.928
(a) With first and second lags (sum of all coefficients)		(0.085)	(0.048)
(b) Excluding outliers		−14.022	−12.475
(b) Excluding outliers		(0.012)	(0.006)
(c) Excluding years before 1999		−22.731	−15.066
(c) Excluding years before 1999		(0.045)	(0.018)
(d) OOP as a share of total health spending		34.196	38.934
(d) OOP as a share of total health spending		(0.031)	(0.012)
Immunization coverage
Baseline	−2.203	−9.841	−7.858
Baseline	(0.073)	(0.030)	(0.020)
(a) With first and second lags (sum of all coefficients)	−0.807	−13.419	−9.870
(a) With first and second lags (sum of all coefficients)	(0.503)	(0.052)	(0.043)
(b) Excluding outliers	−1.807	−6.447	−5.947
(b) Excluding outliers	(0.212)	(0.027)	(0.016)
(c) Excluding years before 1999	−1.808	−7.968	−5.993
(c) Excluding years before 1999	(0.054)	(0.043)	(0.037)

^†

For each health coverage indicator, the first row shows the statistically significant two-stage least squares (method IV-2SLS) coefficients found in the baseline models (presented in Table2), followed by the corresponding coefficients estimated in each specification and robustness test. Test (a) adds the first and second lags of the coverage indicators and the estimates refer to the sum of the contemporaneous and lagged effects. Test (b) excludes 41 outlying observations. Test (c) excludes the period 1995–1998 from the estimations. Test (d) includes as coverage measures only OOP health expenditure as a share of the total (instead of OOP spending per capita), total health expenditure and the rate of immunization. All regressions also control for GDP per capita, the primary education enrolment rate, the share of population aged 0–14 years, the share of population aged over 65 years and country and year fixed effects. p-values (in parentheses under the coefficients) are from two-sided t-tests with standard errors robust to arbitrary heteroscedasticity and auto-correlation.