Appendix to chapter 3

Statistical Analysis to Explain Intercountry Variations in Total Fertility Rates

The available data, for thirty - eight countries, were used to test a number of the country-level findings concerning the determinants of high TFRs. A first set of tests was undertaken using cross country data, looking at relationships in the variation of variables across countries. The results were consistent with the analysis presented in Chapter 3, but questions remained concerning the robustness of the results and the statistical fit. A second set of tests was therefore carried out on a much larger set of data which included both time series and cross-country data. The results are reported here, along with any differences with the findings from the first set of tests. The results of both sets of tests are highly consistent (The data and the methodology used are discussed in the Supplement to this volume.)

For these tests, TFRs are hypothesized to be related to the independent variables as follows: positively to infant mortality (the higher the expected loss of infants, the more births are desired to ensure sufficient survivors); negatively to female school enrollment (better educated women want fewer children); negatively to food security (the greater the food security, the lower the need for children to provide farm labor); positively to cultivable land per person (the more cultivated land per person, the greater the need for family labor to work it); positively to the rate of deforestation (the higher the rate of deforestation, the greater the need for child labor to help fetch wood and water); and negatively to urbanization (urbanization lowers the TFR).

These hypotheses are tested by means of statistical regression with the TFR as the dependent variable. The independent variables are all lagged one year. The lag structure is arbitrary; several were tried, but the statistical fit did not improve. The Supplement presents results of tests for combinations of different countries and different data. Only the sign of the coefficient, the range of t-statistics and the range of significance levels in the various tests are reported below. Since the methodology used does not permit the value of the coefficient to be readily interpreted, it is not reported.

Adjusted R squared = 0.44 to 0.46

The coefficients for female school enrollment' area cultivated per person, infant mortality, and the degree of urbanization are statistically significant at above the 90 percent level, with a 2-tail significance test of 10 percent or less (line 2-tail significance test indicates the probability of the coefficient actually being zero Hence, a 2-tail test of 2.2 percent for the rate of female school enrollment indicates a 22 percent probability that the coefficient is zero or a 97.8 percent probability that it is not zero.) Although the relationship between deforestation and the TFR is positive, as hypothesized, the statistical tests do not suggest significance. The coefficient for calorie supply has the wrong sign and is insignificant.

In the tests with single-year cross-country data, the results were essentiaIly the same, except that deforestation was also significantly related to the TFR, and the coefficient for calorie supply was positive, as hypothesized, but insignificant.

These findings suggest that the TFR is lower as female primary school enrollment is higher The greater the area cultivated per person, the higher the TFR. The higher the infant mortality rate, the higher the TFR. The greater the rate of urbanization, the lower the TFR The positive association between the rate of deforestation and the TFR has ambiguous significance This may be because the rate of deforestation is a poor proxy for the rate of degradation of the rural environment which includes soil and water degradation. Or the hypothesis itself maybe incorrect If further analysis establishes the significance of this relationship, it suggests that greater demand for child labor is associated with environmental deterioration (more labor needed to obtain wood and water and to produce food as the productivity of farm land declines due to deforestation) The relationship between nutrition and the TFR is even more ambiguous, and the hypothesis could not be supported statistically. At very low levels of nutrition, improving calorie intake may increase fertility and, hence, the TFR. Or there may be no relationship Better data are needed to resolve this.