The results in the third column indicate that the market price of domestic services, while not significantly associated with weekday time, has a substantial and consistently positive relationship with his and her housework time on weekend days. This relationship is statistically significant at the 5% level for men in the UK (at the 12% level for men in France). The magnitude of the impact suggests that a 10% increase in the market price of domestic services increases his weekend time on housework by about 10 minutes in both countries. Though not estimated with as much precision, the magnitude of the effect is only a little less (7–9.5 minutes) for women on weekend days. The fact that the time devoted to household production at the weekend is more sensitive to market prices than the time devoted on weekdays is in accordance with our hypothesis that maids are likely a closer substitute for tasks that can be deferred/accumulated rather than tasks that must be performed every day. That the relation is more significant for men than for women suggests that men are more likely to be burdened with these tasks when maids services is more expensive—while perhaps women perform them anyway. That we find any significant association given our rough measure of the price of maid service is rather remarkable. As regards non‐labour income, only for women in France on weekend days is the presence of non‐labour income associated with a statistically significant (p‐value

As a major contribution of this analysis is its incorporation of alternative input prices in the analysis of demand for time inputs to home production, some discussion of the benefits of doing so is warranted. Not surprisingly given that his opportunity cost of time and the cost of maids services are not significant determinants in her time‐use equations, adding measures of these input prices does little to improve the fit of these equations. Adding her opportunity cost of time and the cost of maid service does, however, substantially improve the fit of his time‐use equations in both countries. R‐squared increases by between 33% (France) and 43% (UK) in the case of his weekend time, and by between 12% and 18% in the case of his weekday time. This is true including all the other covariates in both specifications.

All told, relatively few of the other covariates are statistically significant. The general lack of significance in the male housework equations is not new in the literature (see Hersch and Stratton 1997; Friedberg and Webb 2007). By contrast, household composition (as measured by the presence of children of various ages and the presence of other adults) is highly statistically significantly associated with her weekday time in both countries, and to a lesser extent with her weekend time in the UK. The magnitude of the effect of children is larger in the UK, perhaps because of the more widespread enrolment of young children in childcare and the longer school day in France. In addition, these results indicate that French women spend much more time on housework than their British peers, and that French but not British households with children are significantly more likely to have maids services. Cohabiting couples are less likely to have a dishwasher in both countries, and cohabiting women in both countries perform less housework on weekdays. Cohabiting women in France also report significantly less housework at the weekend, while cohabiting men in France report significantly more time on housework on weekdays. This cohabitation effect may be evidence that cohabiting women have more bargaining power in France.

Finally, this formulation allows us to estimate cross‐equation correlations between the unobservables (see Table 4). As discussed earlier, all these correlation terms will tend to be positive if household preference regarding domestic services is an important component of the residual. Although households that are more likely to have maids services for unobserved reasons are also significantly more likely to have a dishwasher for unobserved reasons, there are more significant negative than positive correlations, suggesting that the residuals are likely not primarily driven by common household preferences. Several alternative explanations receive some support.