The effect of improved varieties on PCE and poverty status
of the households is estimated with the nearest neighbor
(NNM) and kernel-based matching (KBM) models. 13 The
empirical results for all improved maize varieties and those
of hybrids in Chiapas and creolized in Oaxaca are given in Tables
5 and 6. The results in Table 5 show that the adoption of
improved maize varieties exerts a positive and significant impact
on the per capita expenditure in Chiapas. Specifically,
the NNM estimates suggest that the causal effect for all improved
maize varieties adoption on household welfare is about
MX$ 136 (US$ 14.6) in Chiapas and MX$ 173 (US$ 18.6) in
Oaxaca. This is the average difference in per capita expenditure
between similar pairs of households that belong to different
technological status. In terms of causal effects, the estimates
of the KBM appear to be similar to those of the NNM. The
NNM estimates on the probability of households falling below
the poverty line were 31% and 27% less than those of nonadopters
in Chiapas and Oaxaca, respectively.
The estimates for hybrid maize adoption in Chiapas and
creolized maize adoption in Oaxaca are also presented in
Tables 5 and 6, respectively, and they also show some interesting
results. The causal effect of adoption of maize hybrid on
the per capita expenditure in Chiapas is positive and statistically
significant and ranges between MX$ 116 (US$ 12.5)
and MX$ 123 (US$ 13). This significant increase in PCE
helped adopters reduce poverty levels by 38%. With regards
to the impact of creolized maize variety adoption in Oaxaca,
both NNM and KBM estimates also show that the adoption
of improved germplasm contributed positively to an improvement
in the welfare of households. In particular, PCE of
adopters was almost MX$ 200 higher than those of nonadopters,
while the poverty level was almost 20% lower than that of
nonadopters. These results suggest that the causal effect of hybrid
variety adoption on poverty reduction is greater in Chiapas
than the creolized variety in Oaxaca. The differential impact of increases in the per capita expenditure on poverty
reduction in the two regions is actually consistent with the results
of the poverty analysis presented in Table 4. The poverty
gap, which measures the mean distance below the poverty line
as a proportion of the poverty line, is much higher in Oaxaca
than in Chiapas, indicating that for the same income gains,
households in Chiapas are more likely to escape out of poverty
than their counterparts in Oaxaca.
Results from the sensitivity analysis on hidden bias, which
show the critical levels of gamma, C, at which the causal inference
of significant adoption impact may be questioned are also
presented in Tables 5 and 6. For example, the value of 1.55–
160 for improved varieties in Chiapas implies that if individuals
that have the same X-vector differ in their odds of adoption
by a factor of 55–60%, the significance of the adoption effect
on output may be questionable. 14 Balancing powers of the
estimations are ascertained by considering the reduction in
the mean absolute standardized bias between the matched
and unmatched models. The median absolute standardized
bias before and after matching are presented in the last column
of Table 7. The estimates show that standardized difference before
matching is in the range of 17% and 20%, while the
remaining standardized difference after randomization ranges
between 5% and 15%, with substantial bias reductions.