We are unable to display your institutional affiliation without JavaScript turned on.
Browse Book and Journal Content on Project MUSE

Find using OpenURL

Comments and Discussion

From: Brookings Papers on Economic Activity
Fall 2012
pp. 251-264 | 10.1353/eca.2012.0021

In lieu of an abstract, here is a brief excerpt of the content:

Comment by Steven J. Davis

U.S. employment rates drifted down from 2000 to 2007, fell precipitously in the wake of the financial crisis and recession of 2008-09, and showed little or no sign of recovery as of this writing in late 2012. These developments fully erased the large employment rate gains achieved in the 1980s and 1990s. They constitute a very serious setback in economic performance, with long-term negative repercussions for human capital, real wages, living standards, and tax revenue net of government transfers.

In this paper Robert Moffitt focuses on employment rate changes from 1999 to 2007 and from 1989 to 1999. The selection of time periods reflects a desire to examine developments between business cycle peaks so as to highlight longer-term forces. Table 1 of the paper shows widespread employment rate declines across age-education groups from 1999 to 2007 for men, with steeper declines for the younger and less educated. The only exception to the downward drift among men is the essentially flat employment rates for the college educated aged 25 and older. Women aged 16-24 and 25-39 also show notable employment rate declines from 1999 to 2007, especially among the less educated. In contrast, women aged 55-64 experienced sizable employment rate gains across education groups. On the whole, men experienced larger employment rate declines than women from 1999 to 2007, but the decline relative to pre-1999 trends was greater for women.

After documenting these facts, the paper turns to simple empirical models that relate employment rate changes to real weekly earnings, nonlabor income, marital status, and family structure. This part of the paper produces few robust conclusions about the proximate determinants of employment rate changes. As it turns out, the chief empirical results are highly sensitive to the wage measure and, for women, to basic aspects of the regression specification.

The specification sensitivity for women comes through clearly in a comparison of tables 6 and 9. When the same regression specification is imposed on all women, the fitted empirical model accounts for none of the overall drop in female employment rates between 1999 and 2007 (table 6). However, letting the regression coefficients vary by marital status and, for unmarried women, by the presence of children yields a very different picture. The predicted changes implied by the flexible specification account well for the actual employment rate changes for married women and for unmarried women without children over 1989-99 and over 1999-2007 (table 9).

The performance of the flexible specification for unmarried women with children is harder to assess. Moffitt writes that the flexible model "does a poor job of explaining the small decline in employment for unmarried women with children, however, predicting instead an increase of some magnitude." But it is unclear whether the differences between the actual and the predicted changes for this group are statistically significant. Both quantities in these comparisons are subject to sampling variability. The samples for unmarried women with children are smaller than for the other groups in table 9 (personal communication with the author). The table reports a large standard error on the estimated wage elasticity for this group, and a footnote to the discussion of table 5 suggests that the reported standard errors are too small in any event. In sum, the flexible specification appears to perform well in accounting for changes in women's employment rates from 1989 to 1999 and from 1999 to 2007—with the possible exception of unmarried women, for whom the evidence is not very informative.

These results hold when using the wage measure featured in the main text. A key challenge in constructing the wage measure is how to handle nonworking persons, for whom there is no contemporaneous wage observation. Ignoring this issue would lead to potentially biased wage measures— and biased estimates of employment responses—because changes over time in observed wages may not accurately reflect changes over time in market opportunities for nonworking persons or for all persons. To address this issue, the main text uses the predicted wage for employed persons from a wage equation that incorporates a selection correction for employment status. The selection correction is identified by excluding nonlabor...

You must be logged in through an institution that subscribes to this journal or book to access the full text.


Shibboleth authentication is only available to registered institutions.

Project MUSE

For subscribing associations only.