Fourth Meeting Notes

What this is: The way Gretchen would proceed, if she were to start the project tomorrow

What this is NOT: The “NTA Way.” No such thing exists (yet).

Separating “Regular” NTA by Sex

1. Profiles based on external sources may be available by sex
   • Public and private education enrollment
   • Social Security
2. Profiles from surveys separable by sex
3. Macro controls not available by sex so assume adjustment factor is same for both sexes
4. Within-household allocations by sex
   • When age allocations are data driven, add sex in as one of the variables in addition to age
     • Private education
     • Private health
   • When allocations are based on equivalence scale, assume same scale for both sexes OR use some measure of difference in average consumption (Swedish team used consumer data?)

What would happen if we try to let this be data driven also? Might produce bad estimates because most of the variance would be determined by households in which both sexes were not represented in same rough age categories, i.e. would be driven by single mom or single data households? It wouldn’t be correct to apply those shares to two parent households. So, would need to include a separate household structure variable?

One complication: what about housing issue under a joint custody arrangement? Whichever parent was considered the "main custodian" would share a house with the children, reducing the adult share. The other gender parent would be considered to occupy a house along, resulting in higher share.

ALSO NEED TO INCLUDE THE ROLE OF HEADSHIP. PART OF EXAMINING NTA BY GENDER MUST INVOLVE RELAXING THE ONE HEAD ASSUMPTION BECAUSE THAT SO OFTEN GIVES OWNERSHIP OF ALL ASSETS TO MAN. USE "PROPORTIONAL HEADSHIP", BY INCOME OR BY EQUAL SHARES FOR ALL ADULTS IN HOUSEHOLD

Satellite Time Use NTA (assumes you have a time use survey, TUS)

1. Identify unpaid activities in TUS that are equivalent to activities included in the NIPA when they are paid
   • 3rd person criterion - could pay someone else to do it

     This is a very thorny issue. Some people would be willing to outsource an activity, but others would derive pleasure from it and wouldn't be willing to outsource. But in paid work, it might be true that I would work for less, but I don't have to because the market will pay me what I make. Even if I enjoy doing it, that doesn't lessen the value. Also, maybe my utility from performing the activity isn't as important as the fact that someone else can benefit from it.)

2. Impute a wage to those activities (method choice here is VERY important, so best to use both methods and publish range of estimates)
   • Opportunity Cost (upper bound?)
   • Replacement/Substitution Cost
     • By activity
     • Housekeeper wage (lower bound?)

       There are other methods in literature, but these seem most prevalent and possible given existing data constraints, i.e. the data constraints are to onerous to examine the value of production by the unpaid labor instead of just the value of the labor input.

       Big issue of pre-tax or post-tax wage. Argument for pre-tax is that it's closer to the value determined by the market, but this could exacerbate the issue of selection bias as in using the opportunity cost method. Example that in Sweden there are tax subsidies for paying for household services. The argument is that more monetization of activities means more tax revenue, but it is also an attempt to remove the marginal tax incentive for people NOT to outsource (unpaid labor has a sort of subsidy, compared to paid labor). Just like the fact that people eligible for pension benefits who are also at the edge of a higher tax bracket have an incentive to retire to avoid the tax. Choice depends on what issue we are trying to address! If it is cost of care, gross makes more sense. If it is gender differences, the post-tax wage gives more information about how people choose to perform the labor themselves or pay someone else to do it. Probably best to show both.

3. Calculate equivalent of YL profile but for production of unpaid labor
   • If TUS only identifies one activity for any unit of time, use that activity
   • If TUS includes multiple activities, consider all activities but allow only a single unit of time (otherwise you overvalue multitasking), divide evenly between activities
   • Common in TUS surveys to assign "primary" and "secondary" activity, let these activities be valued at 2/3 of unit time for primary and 1/3 of unit time for secondary (will probably need sensitivity analysis for this)
     • One hour of watching kids as primary while watching television as secondary is valued as 40 minutes of childcare and 20 minutes of leisure which wouldn’t be included in unpaid YL
     • One hour of primary watching kids while secondary cleaning is 40 minutes of childcare and 20 minutes of cleaning
4. Calculate equivalent of C profile but for consumption of services of unpaid labor
   • When activity specifies care recipient within the home, assign to age of recipient member (childare, eldercare)
   • When activity is consumed by household members as a group compare two methods:
     • EAC weights used to assign household total
     • Data driven (regression, iterative method) estimate would be better because makes no assumption that kids consume less of the product of household unpaid labor
   • When activity is consumed by specific individual outside of household, assign to that age (known age or if elderly parent of unknown age just make it caregiver's age + generation length)
   • When activity is consumed by non-specific individual outside of household distribute on equal per capita basis
5. Calculate equivalent of TFWI and TFWO based on C and YL of unpaid labor
   • Use same “unitary model” NTA uses for intrahousehold cash transfers (from each according to his surplus, to each according to his deficit)
6. Don’t have to worry about asset-based reallocations or government transfers of time because they don’t exist