53
32
Mean for the dependent variable = 3.55, standard deviation = 2.51, skewness=0.11, and the Shapiro-Wilk
statistic=0.95, indicating that normality is reasonably well approximated by these data. All variables are
coded in the same manner as they were for the previous regression table. Multicollinearity is low to
moderate as shown by the condition index, which is 15.4. Pairwise correlation coefficients are all less than
0.5. Heteroskedasticity is not in evidence, as measured by White’s general test
33
Multicollinearity is low to moderate for the regression model (Condition index=17.4), and
heteroskedasticity is not in evidence. Individual-level variables have been coded as before.
34
This variable reflects the omnibus effect of caste status, and it is calculated by assigning ranks (from 1 to
5) to each caste related to its position within the caste hierarchy. This assignment is somewhat arbitrary, no
doubt, however, the results did not change when we worked, instead, with caste dummies.
35
Dummy variable, coded as follows: 1 if the village is located at gram panchayat HQ, zero otherwise.
36
Coded as before: MP=0, Rajasthan = 1.
37
The mean for the dependent variable = 59.6, standard deviation=21.9, skewness= -0.65, and the
Shapiro-Wilk statistic=0.95.
Convention