Amartya Sen set off a debate in development economics when he estimated that there are 100 million ‘missing women’ in the world, referring to the magnitude of female survival disadvantage due to unequal treatment in the intra-household allocation of
Yüklə 0,8 Mb.
|
- Bu səhifədəki naviqasiya:
- V. Econometric Model and Estimation Results
- Table 8: Panel Estimation for 0-6 Age Cohort Sex Ratio (Total)
Note: 1Percentage of children under 3 years classified as undernourished on three anthropometric indices of nutritional status, according to selected demographic characteristics, India, 1998-99. Source: National Family Health Survey-2, 2000, India. As the anthropometrics evidence on nutrition indicates that there is only slight difference between girls and boys in terms of nourishment, therefore the health-related intrahousehold variable is not considered for the model. The availability of health infrastructure is also a potential dependent variable for the analysis. As per NFHS second round data, only 35 per cent of the deliveries are institutional in India, it may not explain the gender bias in the mortality rates of girls more than boys in age cohort 0-6. The studies also noted that availability of health infrastructure might not reduce gender bias in health seeking behaviour, if parents are less likely to use such facilities for girls (Bardhan, 1988; Murthi et al, 1995).
where, ln jcrit = log of sex ratio of age cohort 0-6 ln flitit = log of female literacy rate (+7) ln flprit = log of female labourforce participation rate lnT flitit = log of total literacy rate (+7) lndegit = log of decadal economic growth rate = spatial dependence error correction term t = surrogate of omitted explanatory variables. The conditioning set includes two important variables that measure the economic value of women as well as the educational attainment. Educational attainment of the state captured through general literacy rate (the number of literate divided by the population above age seven) is a better measure of development that goes further than GDP per capita to capture the effectiveness of public policy stance in terms of development interventions3. However, we have controlled for decadal economic growth rates also in model specification. Prima facie, the relative chances of survival of the girl child should improve with rising literacy rates. The availability of health infrastructure and access to intrahousehold resources including healthcare proxied through nutritional disadvantage across gender are omitted as per the reasons mentioned in the earlier section. All the variables are taken in natural logarithms because the dependent variable is a ratio and therefore asymmetric around the reference value. With logarithmic transformation, a deviation from a reference point becomes equidistant in either direction (Fossett and Kiecolt, 1991). The regression coefficients are then elasticities of the juvenile sex ratio with respect to the explanatory variables. The data exploration across states revealed that not only does juvenile sex ratio worsen as one moves from Southern belt of India to Northern belt, but there is also a clustering of the sex ratio such that low sex ratio states tends to have similar neighbouring States. Econometrically, this problem of spatial dependence can be treated through two ways. Introduction of dummy variable for the Northern States may partially capture this pattern. Alternatively, the observed spatial dependence may be modeled directly. Measurement problems associated with the arbitrariness of geographic delineation and aggregation of States, spatial spillover effects, or omission of spatially correlated explanatory variables may cause spatial correlation in errors. This type of spatial dependence, called spatial error autocorrelation, makes ordinary least square (OLS) estimates inefficient. It can be corrected by adding an autoregressive process whereby the error term depends on its spatially lagged value and a random error term. In the context of Turkey, Berik and Bilginsoy, 2000 identified a spatial dependence, which indicates the presence of a pattern of interaction whereby neighbouring provinces influence each other’s behaviour. This type of spatial dependence is called mixed regressive spatial autoregressive model, where the spatial lag in the dependent variable renders OLS estimates biased and inconsistent (Anselin, 1988; Case 1991 cited by Berik and Bilginsoy, 2000). This functional relationship between the dependent variables across provinces is modeled by including the spatially lagged sex ratio (that is, the weighted average of the values of the sex ratio in neighbouring provinces) among the RHS variables of the regression equation. In the context of Turkey, Berik and Bilginsoy (2000) applied both spatial econometric models as they donot have theoretical priors on the type of dependence. But in the context of India, we opted out mixed regressive spatial autoregressive model because of the broad absence of influence of neighbouring provinces in the pattern of sex ratio except for few selective border units of provinces. The results of pooled least squares estimation, not disaggregated across rural and urban, are reported in Table 8. The models (ii) and (iv) ignores spatial dependence term and all models were estimated by fixed effects. All models are adjusted for White Heteroskedasticity consistent standard errors and covariance. The analysis showed that female labour force participation rate and female literacy rate are significant in determining the juvenile sex ratio, but inversely related. The economic growth rate is also found inversely related to juvenile sex ratio but insignificant. Table 8: Panel Estimation for 0-6 Age Cohort Sex Ratio (Total)
Yüklə 0,8 Mb. Dostları ilə paylaş:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||