maids services
A second key input is maids services. Both surveys contain information on purchased housework services. The French questionnaire asks whether the household regularly purchases domestic help/services, and if so how much time is purchased each week.a The British questionnaire asks separately about paid help for ‘cleaning, tidying up’ and ‘ironing’ over the past four weeks. There are then additional questions about who provides the help,a how often, and for how long. However, given the small number of households that purchase maid services, we focus on whether or not a maid is hired. Less than 8% of all households purchased such services.

To evaluate the accuracy of these sample measures, we examined consumption surveys. A similar sample (couples aged 20–59) from the 2000 Household Consumption Surveys for France (Enquête Budget des Familles) indicated that a comparable 8.6% of households purchased domestic help. Flipo et al. (2007), in an analysis of maids services in France, also find that only a small minority purchase such assistance.a The UK Family Expenditure Survey 2000–1 indicates that about 18% of couple households purchased household help within the last two weeks, but these figures include window cleaners. Including information on window cleaners from the UKTUS, 30% of our sample purchased a similar bundle of services within the last four weeks. As window cleaning services are likely not purchased more than monthly, these figures, while more difficult to compare than the French data, do provide support for our sample statistics. Few households have maid service.

Appliances
Both surveys also ask respondents about appliances. Specifically, there is information from each survey on whether there is a dishwasher or a clothes washer in the household. Use of a dishwasher is likely to affect the time devoted to washing up dishes, while use of a washing machine is likely to affect the time spent doing laundry. In each country, over 98% of couple households have a clothes washer. With such little variation in outcomes, we do not model this input. By contrast, 42% of our British sample and 57% of our French sample reported having a dishwasher. Thus we model in our analysis the availability of a dishwasher.12
Table 1 presents summary statistics for the dependent variables used in this analysis. In general, men report less time on housework than women, and more time is spent on weekend days than weekdays. British men report more housework time than French men, while, perhaps to offset this, French women report more housework time than British women.

Table 1. Inputs to Domestic Production by Country
(Times in minutes) UK France
Full sample No maid Maid Full sample No maid Maid
His time on weekdays 14.8 14.9 12.6 11.2 11.2 11.2
His time on weekend days 25.5 24.8 35.3 21.9 21.6 25.3
Her time on weekdays 78.5 79.9 59.0 103.9 107.4 62.3
Her time on weekend days 91.0 91.1 88.6 113.5 115.6 89.2
Hire a maid 6.6% 0.0% 100.0% 7.7% 0.0% 100.0%
Have a dishwasher 42.1% 39.4% 80.2% 56.6% 53.8% 90.2%
Number of observations 1295 1209 86 2924 2699 225
Looking at similar summary statistics for the sample with and without maid service, we see that in households with a maid, men report a little more housework time on average on weekend days (4–10 minutes), while women report substantially less time on weekdays (20–45 minutes) and, at least in France, less time also on weekend days. Those hiring a maid are also substantially more likely to have a dishwasher. Of course, these figures do not control for other household characteristics as our multivariate analysis will.

Prices
In the specification that follows, the variables of greatest interest from an economic perspective are the prices. Chief among these are the opportunity costs of each partner’s time. A price for maids services and one for appliances are also included.

To construct measures of the opportunity cost of time for each partner, we follow standard practice and impute wages much like Kalenkoski et al. (2009). All persons aged 20–59 who are not in school and who provide personal (and partner) data on education and potential experience as well as household data on non‐labour income receipt are included in the sample used to impute wages. This sample is not restricted based on the availability of time diary data or the presence of a partner. In total, 2571 (4141) men and 3015 (4560) women are included in the British (French) wage analysis. Hourly earnings for non‐self‐employed workers are calculated, and standard Heckman sample selection methods are employed to control for non‐participation separately by gender.13 The wage regressions contain a standard set of controls for education, potential experience, region, marital status, minority status, home access to a computer and, in the case of the UK, the local unemployment rate. Selected characteristics of the respondent (a quadratic in age, disability status), as well as characteristics of the household (number of other adults, presence of children of various ages, non‐labour income dummy) and the partner (age, education, experience, health, minority status), as well as season, are used to identify employment. These equations are then used to predict unconditional wages for everyone in the final sample: wage workers, the self‐employed, the unemployed and those out of the labour force. Further details regarding the wage imputations are available on request from the authors.14
These imputed wage measures constitute estimates of each individual’s value/cost of time, and as such we enter them as regressors in the demand equations. As estimates, they introduce error. We employ robust estimation techniques and bootstrap the standard errors of the econometric model in order to obtain more robust standard errors.

The natural log of the regional median hourly pay for domestic services is used to proxy for the price of maids services. The median price of domestic services in France is obtained from the French Labour Force Survey 1998, which contains gross wages of workers in this specific industry, as well as information on regional variation.15 The British data on domestic service prices are obtained from the British Annual Survey of Hours and Earnings as conducted by the Office for National Statistics. These data provide annual information by region on median gross hourly pay for elementary occupations in sales and service, a category that the British Quarterly Labour Force Survey indicates is primarily (greater than 50%) ‘cleaners, domestics’.16 In total there are 21 regions in France and 11 in the UK. As the UK data span two years, we have 22 distinct values for the UK. The fact that this crude price measure shows relatively little variation within the sample will act to increase the standard error associated with these coefficient estimates and reduce the probability that the price of maid service has a statistically significant effect on any of the inputs to domestic production.

The ‘price’ associated with having access to an appliance—here a dishwasher—is substantially different in type than the price of any of the time inputs. Appliances are capital investments that are not ‘used up’ in a single period, as is labour. A capital appliance would ideally be purchased only if the present discounted value of that appliance were greater than its present discounted cost over its expected lifetime. A lower market price would obviously lower its cost and, all else equal, would make households more likely to have the appliance. However, once purchased, or once a residence with such an appliance is acquired, the appliance’s price becomes a sunk cost.17 The relevant opportunity cost is then rather the cost of operating that appliance (and potentially many other appliances). Thus we employ information on the log of the average regional price per kilowatt hour of electricity as our measure of the price of appliances. Electricity prices in France do not vary by region, so this variable drops out of the French analysis.