2. Is it preferable to use the actual wage rate for units with part-time employment and scale up their potential earnings to full time or to use an imputed wage rate for them as well?

3. How does one build into the estimates derived from imputation an appropriate variability based on the error term of the estimation model for those units that require imputation?

4. How should one estimate the capacity for those who have retired or are elderly and have not had a history of earnings at an earlier age?

Furthermore, if earnings capacity were fully measurable and brought into the measurement of poverty, then other analytic issues would be raised. For example, by introducing leisure time as a commodity that is purchased with the available resources of time and money, there is then a need to take account of the fact that those with a high wage rate face a relatively high price for that commodity. Until it is clear how to estimate the capacity to earn with greater precision and consistency than is now the case, an earnings capacity definition of resources should not be the basis of the poverty measure. Even when enough is known about how to integrate time and money resources in the measurement of poverty, it will also be necessary to consider how that introduction might alter the level that is set as the threshold for poverty. It would not be reasonable to simply add the value of some or all nonmarket time without considering how that modification on the resource side should affect the level of the threshold.

There is at present no feasible way to improve the measurement of poverty by incorporating the time allocation of families. We encourage further research that might yield a better solution in the near future, but we see no way adequately to address this perplexing issue now. The earnings capacity estimate of available income, suggested by Haveman and colleagues, and the wage rate usage as suggested above in the context of Vickery's analytic figure, both address the issue, but they are not warranted as a replacement for the current strategy of estimating income directly. Although there are important contributions in the literature regarding how Americans actually spend their nonmarket time (e.g., see Juster and Stafford, 1985; Robinson, 1977; and Walker and Woods, 1976), and analytically how to understand its allocation (e.g., Becker, 1965), we know of no implementable solution to the concern addressed here.

Thus, many concerns about the treatment or nontreatment of time are unresolved. One of these concerns is that some families are probably considered to be impoverished that could spend enough time working for pay to earn enough to get themselves out of poverty but do not do so. At the other