英语论文网

53 percent are planning to use contraceptives in the future. Again, it is interesting to note that 58 percent of non-users who had a boy plan to use contraceptives in the future, while 43 percent of non-users who had a girl will use contraceptives in the future.

This discussion indicates a high awareness about family planning methods among slum dwellers and willingness to adopt such methods. However, we argue that while slum-dwellers realize the economic benefits of limiting family size, they are also motivated by the benefits of investing in male children as a means of assuring for the future. If our hypothesis is valid then the probability of using contraceptives will increase as number of sons and daughters increase, but the increase will be greater for additional sons. That is, if we run an econometric model, the odds ratio for boys will be greater than odds ratio for girls.

经济计量分析-4. Econometric Analysis

As mentioned earlier, the econometric analysis is based on an ordered logit model with the respondent assumed to face three choices - no use, traditional/folkloric method and modern method. The result of the Brant test (the χ2 value, with 14 degrees of freedom, is 56.08, with a probability value < 0.00) indicates that the null hypothesis of proportional odds assumption is rejected. Allowing parameters to vary enables an examination of the variations in impact of each variable at different choice levels. This provides more sophisticated information set for policy making. Both the partial proportional odds model and the generalized ordered logit model - which may be interpreted as a set of logistic equations (Williams 2006) - are estimated, and the Likelihood Ratio (LR) test used to choose the most parsimonious (in the sense of least restrictive) model specification. The value of the LR ratio is 20.65 (with a probability value of 0.02), so that the null hypothesis (the partial proportional odds model is nested in the variable parameter model) may be accepted. This implies that the former model is not too restrictive and is appropriate in the present case.

Table 1: Results of Generalized Ordered Logit of Contraceptive Use by Slum-dwellers
The final model, therefore, is a generalized ordered logit where the coefficients of age, its square, the language dummy and number of living sons is allowed to vary.

It can be seen that coefficients of both number of living sons and daughters are significant and have high odd ratios. Comparison of the effects of birth of a boy and girl child reveals that number of living boys has a higher odd ratio, indicating that birth of a son has a greater impact on adoption of contraceptive than birth of a girl. This is consistent with the son preference observed in developing countries (Roy et al. 2008, Jayaraman et al. 2009, Saha and Bairagi 2007) and our research hypothesis.

Quite a few of the other control variables are significant. Muslims have a lower probability of using contraceptives than BCH (Iyer 2002, Bhatt and Xavier 2005, James and Nair 2005, Kulkarni and Alagarajan 2005), while UCH and All Others have the same levels of contraceptive use. Inter-state migrants from Bihar, Jharkhand, Uttar Pradesh are expected to be more conservative than the local population. Language (used as a proxy for their different cultures) exerts a significant effect on the decision to use a contraceptive,