Independent variables

Dependent variable


Family planning
(1)

Antenatal care
(2)

DTP3
(3)

Measles
(4)

Tuberculosis
(5)

Sanitation
(6)

ln Beds
(7)

ln Physic
(8)


Lag (IFFT)

−0.299 (0.157)

−0.340^{a} (0.089)

− 0.611^{c} (0.007)

− 0.426^{b} (0.021)

0.106 (0.350)

− 0.254 (0.389)

− 0.001 (0.934)

−0.020 (0.278)

ln GDPpc

7.40^{c} (0.000)

10.96^{c} (0.000)

10.95^{c} (0.000)

10.66^{c} (0.000)

6.05^{c} (0.000)

7.81^{c} (0.000)

0.07^{c} (0.000)

−0.04 (0.287)

RESET test (pvalue)

4.76 (0.033)

1.54 (0.219)

3.87 (0.053)

4.16 (0.045)

2.00 (0.163)

56.62 (0.000)

14.03 (0.000)

7.20 (0.009)

KP rk LM (pvalue)

46.70 (0.000)

11.08 (0.004)

12.30 (0.002)

12.30 (0.002)

36.65 (0.000)

14.18 (0.001)

11.90 (0.003)

12.08 (0.002)

KP rk F (SY 10% max. IV size)

88,671.96 (19.93)

9.68 (13.43)

16.48 (19.93)

16.48 (19.93)

79,885.44 (19.93)

81.77 (19.93)

48.94 (13.43)

87.86 (19.93)

J statistic (pvalue)

1.11 (0.291)

0.09 (0.764)

0.03 (0.850)

0.04 (0.835)

0.32 (0.569)

0.26 (0.608)

0.17 (0.685)

0.05 (0.817)

# Countries

72

64

72

72

59

72

63

66

 Estimated coefficients and statistics, pvalues in brackets. ^{a} Denotes coefficient statistically significant at the 10% level, ^{b} at the 5% level, and ^{c} at the 1% level. Standard errors and covariance are heteroscedasticityconsistent. RESET is the regression equation specification error (heteroscedasticrobust) test designed to test for missing (excluded) regressors. It also has great power to detect nonlinearities in the model. Thus, rejection of the null hypothesis could be due to either a nonlinearity or an omitted explanatory variable. The model is identified if the heteroscedasticityrobust KleinbergenPaap rk LM statistic for underidentification test (KP rk LM) rejects the null hypothesis. If the robust KleinbergenPaap rk Wald F statistic for weak identification test (KP rk F) is greater than the Stock and Yogo [60] critical value (SY 10% max. IV relative size), then the null hypothesis of weak instruments can be rejected. The null hypothesis of the robust Hansen’s J statistic is that the instruments are uncorrelated with the error term and that excluded instruments are correctly excluded from the estimated equation. The excluded instruments used in the estimations for the instrumented variables are as follows (lagged values of variables (lag) refer to averages of variables for the period 2002–2007):
  Column (1). Instrumented: ln GDPpc. Excluded instruments: lag (Urban), lag (ln GDPpc)
  Column (2): Instrumented: lag (IFFT), ln GDPpc. Excluded instruments: lag (CPI), lag (IFF/Population), lag (ln GDPpc)
  Column (3): Instrumented: lag (IFFT). Excluded instruments: lag (CPI), lag (IFF/Population)
  Column (4): Instrumented: lag (IFFT). Excluded instruments: lag (CPI), lag (IFF/Population)
  Column (5): Instrumented: ln GDPpc. Excluded instruments: Literacy, lag (ln GDPpc)
  Column (6): Instrumented: lag (IFFT). Excluded instruments: lag (CPI), lag (IFF/GDP)
  Column (7): Instrumented: lag (IFFT), ln GDPpc. Excluded instruments: lag (CPI), lag (IFF/GDP), lag (ln GDPpc).
  Column (8): Instrumented: lag (IFFT). Excluded instruments: lag (CPI), lag (IFF/GDP)