number of dependents, consistent with the data means for the larger IHS sample.28
Second, the gap between data means for saving is much larger than the gap between
medians depicted in Figure 6. Controlling for the effects of family size, lifecycle
phase and the net income of each partner using regression analysis generates profiles
that show a wider gap in household saving. The result is driven by a coefficient on the
net income of the female partner that is around twice that on primary net income,
indicating a much higher propensity to save from her net earnings.29 An additional
child is found to have a large negative effect, but because the two household groups
have close to the same number of children, family size explains very little of the
additional saving of the H2 household, evaluated at data means. A formal model of
the joint determination of female labour supply and household saving over the
lifecycle that generates profiles similar to those in Figure 5 can be found in Apps and
Rees (2003).30
Similar results are obtained for lifecycle spending on health and insurance across the
child-rearing phases. The profiles of private health insurance are depicted in Figures
6a and 6b. Figure 6a plots the percentage of households in each group, and in the full
sample, that purchase private insurance to cover hospital, medical and dental costs.
Figure 6b plots the corresponding data means for fees incurred by each group, and by
the full sample. The H1 and H2 profiles in both figures show a strong tendency to
track female labour supply in the early child-rearing years. Only 45.0 per cent of H1
households in phase 2 and 40.1 per cent in phase 3 purchase private health insurance
cover for hospital, medical and dental costs. The figures for the H2 household are
63.1 per cent in phase 2 and 60.5 per cent in phase 3.
28 For the smaller HES sample, data means for the number of children are: 1.98, 2.41 and 1.87 for the
H1 household and 1.61, 2.20 and 1.63 for the H2 household, respectively.
29 The result was found to hold across a wide range of model specifications. The models were
estimated on a sample that excluded the bottom five per cent of male net incomes, negative female
incomes, and the top 1 per cent of male and female net incomes, to remove the effect of outliers.
30 An essential feature of the model is that it takes account of the presence of two-adults in the
household. Much of the literature on saving behaviour treats the household as a single decision unit.
See, for example, Blundell et al. (1994) and the survey by Browning et al. (1996). In models of this
kind high wage H1 households are confused with much lower wage H2 households because, in effect,
the models fail to control for wage rates. The studies therefore miss the strong positive association
between female labour supply and household saving at a given wage level.
17