EL NIÑO Y LA CIENCIA

Before investigation of the direct effect of the family planning program on fertility change, a hierarchical analysis of variance method and multiple classification analysis of the factors contributing to percentage of fertility difference between 1967-70 and 1976-79 are applied 1. The aims are first, to examine the quantum of fertility change differentials both before and after all the compositional differences in other variables have been adjusted, and second, to identify to what extent the magnitude of the effect of each predictor on fertility change may be explained by other predictors. For

^ All independent variables in this section are treated as categorical variables. Owing to the limitation o f the SPSSX package program which has very large memory requirements for running the A NO VA program, it is impossible to include all the ten independent variables with more than two categories for each predictor (SPSS Inc, 1988). However, to analyse the fertility change pattern, a smaller number o f independent variables with more categories for each predictor are also applied.

example, does the effect of percentage using contraception on fertility change either increase or diminish after adjusting other factors?

Table 5.1 shows the sum of squares of each variable after having adjusted for all the preceding variables. The sum of squares of the dependency ratio, for example, is adjusted for the proportion of urban residents and the sum of squares of the infant mortality rate is adjusted for dependency ratio and proportion of urban residents. The sum of squares of the proportion using contraception is adjusted for all other variables. The table shows that after following the order noted above, only five variables among the ten predictors significantly affect fertility change in the districts of Java. These predictors are the proportion of urban residents, the dependency ratio, the infant mortality rate, the proportion of females working as other than family workers, and the proportion of women aged 20 to 24 years who are never married. This set of variables explained 0.87 of the variance of the fertility change in districts in Java.

As expected, the first three variables, that is demographic variables, are the biggest contributors to the variance explained. The table shows that the proportion of women using contraception does not have a statistically significant effect on fertility change. This does not necessarily mean, though, that family planning practice has no effect on fertility change. As explained below there is substantial shared variance between the independent variables. In a hierarchical framework the effects of factors are assessed in this order of entry. In essence this means that variables entering first have the first chance of explaining variance in the dependent variable. In my hierarchical analysis use of contraception is entered last

Table 5.1: Result of hierarchical analysis of variance of fertility differences on selected variables.

Source of Sum of D P . F Contribution

variation Squares of

Demographic variables

Proportion urban residents Dependency ratio

Infant mortality rate

Ecological variables

Proportion irrigated paddy fields Proportion households non-nuclear

Educational variables

Proportion females finish primary school

Women’s economic activity variables

Female labour force participation rate Proportion females non family worker

Other variables

Proportion females 20-24 who are single Proportion using contraception

Summary Explained Residual Total 677.90 1 18.32C .094 2309.26 1 62.40C .321 1050.16 1 28.38C .146 1.85 1 .05 .000 12.02 1 .33 .002 7.39 1 .20 .002 73.74 1 1.99 .010 S175.84 1 4.75b .025 242.94 1 6.57b _.035 8.75 1 .24 .001 6233.68 55 962.21 26 7195.89 81 Notes:

b Significant at 5 per cent level of significance. c Significant at 1 per cent level of significance.

To see whether percentage of current users is an important predictor before and after adjustment of other variables, the eta and partial r (beta) for this variable are

compared1 (Table 5.2). The eta for percentage using contraception (0.52) is the third highest after infant mortality rate (0.62) and dependency ratio (0.60). However, after all other variables in the model are controlled, the strength of relationship between family planning practice and fertility change seems to disappear (the beta is 0.05). With reference to the analysis presented in Chapter 4, it is found that all variables included in the model in that chapter are significantly related to the use of contraception. It is worth noting that except for the proportion of females aged 20-24 years who are single, all predictors used in Chapter 4 are also used in this chapter. The weaker relationship between family planning practice and fertility change after controlling for other independent variables is probably because those variables are significantly related to the use of contraception.

The dependency ratio, which is a ratio of percentage of population aged under 15 years plus population aged 65 years and over to percentage of population aged 15-64 years, has the highest beta. This is caused by the fact that the nature of measurement itself is affected by fertility change. Therefore, it was decided not to include this variable in the further analysis of fertility change. When the dependency ratio is removed from the model it is found that the beta for the proportion using contraception is quite high: 0.15 (Table 5.3). This confirms the earlier statement that family planning practice has some effect on fertility change.

Consistent with the preceding tables, the table also shows that infant mortality rate is the most important predictor of fertility change in the districts. Both eta and beta are high for this demographic variable.

1 The eta statistic, which reveals the degree of influence of each factor without controlling for the influence of