Page 50 - Azerbaijan State University of Economics

Mbu Daniel Tambi , Peter Arung Etat: Implications of Access to Portable Water
For Child Health Production in Cameroon

In rural Cameroon, parents always do their best to give their sick child protected
water; however with regards to children in good health, they don’t care much on the
source of water. Thus, if household access to portable water has a positive impact on
child health, then we would expect the Probit estimate of  to be biased upward.
1

As observed with the case of rural Cameroon, the prime difficulty of the two-way
causality that comes in the effect on household access to portable water and child
health may cause the classical endogeneity problem. To avoid the strong likelihood
of this endogeneity bias, confounded by the problem of variables that is missing in
empirical data, we use a two stage least squares (2SLS) estimation approach. Thus,
the first-stage equation in this approach is:

APW  += i 1  RV + i 1 (2)

i
1
i

Whereby RV is simply rainfall variation (use as an instrument to solve for the
i
endogeneity problem); it should be noted that the 2SLS model captures the causal
effect of household access to portable water for those households whose children’s
health is affected by rainfall variation. Importantly, though APW is ordinal, 2SLS
i
estimates of can be interpreted as estimating the average marginal effect of a unit
1
increase in APW for children whose health is affected by the variation in rainfall.
i
Before presenting the 2SLS estimates, we present a reduced form analysis of
household access to portable water; here we would expect to observe mothers with
death children or children suffering from diarrhea diseases to be affected by variation
in rainfall. The reason is because children suffering from diarrhea diseases are
negatively affected by poor household access to portable water. The 2SLS estimation
allows us to scale the probit marginal effects into the effects on an increase in our
ordinal household access to portable water measure (Wooldridge, 1997).

As indicated earlier, we use variation in rainfall as an instrument to overcome this
endogeneity problem between household access to portable water and child health
outcomes which cannot be adequately controlled for by observable characteristics.
Assuming that variation in rainfall is a valid instrument, we will use the ivprobit
model (probit model controlling for endogeneity) which better respects the binary
nature of child health (child death attributed to diarrhea diseases) as represented by
the following two equations:

CDD =   +  APW +
*
2 i 2 i i 2 (3)

50