Women, Work, and Caring for Aging Parents: Modelling Difference By Education in the Psid

Background & Purpose:

The largest source of long-term care services for aging Americans is unpaid care from family. The majority of unpaid family caregivers (61 percent) are employed and must balance earning income with providing unpaid labor for care. How does caring for an aging parent affect women’s labor force outcomes? And do effects vary by socioeconomic status (SES)? Evidence to date is mixed, with relatively little attention given to SES. Some recent investigations use samples where all respondents are near or over retirement age; however, we know a key difference in mortality by SES is earlier disability onset among low-income communities. Including young potential caregivers is necessary to examine the relationship between caregiving and labor force participation by SES.

In this paper, I build on the existing literature to give more attention to socioeconomic status by (1) using a sample of women of all ages in the 2013 Panel Study of Income Dynamics (N=5,296); (2) testing for a moderating effect by education.

Methods:

I model effects of employment on the extensive margin (whether or not employed at all) and on the intensive margin (hours worked and hourly wage rate), using logit and OLS regressions, controlling for demographic factors. Since employment is likely endogenous to caregiving (for instance, women with low labor force participation may be more likely to begin caring), I also produce an approximate causal estimate of the effect of caregiving through a two-stage least squares (2LS) instrumental variable model to instrument for hours of caregiving with parental health and distance to parent. All analyses are done without and without interaction terms by education to test for moderating effects.

Results:

Preliminary results find caregivers are as likely to be employed and work the same number of hours as non-caregivers. Employed caregivers make about $1 less per hour that non-caregivers (p<.05), and employed intense caregivers make about $2 less per hour (p<.001). This pattern holds by education, except among low-education caregivers over 51—who are less likely to be employed, but when employed work many more hours. Preliminary causal estimates find caregiving results in higher employment overall (even among women caring at least 10 hours per week), but lower employment for women with low education. (Note these results are preliminary because I have received approval but have not yet included restricted geographic data from the PSID with a more precise measure of distance between aging parent and adult daughter in the instrumental variable model.)

Conclusions & Implications:

Understanding the relationship between caregiving and labor force participation, especially for economically vulnerable populations, is crucial as we set the long-term care policy agenda for the coming decades. Here, the null finding of no difference in hours worked is noteworthy, since it means hours spent caregiving must be traded off with hours of time use other than paid employment. The causal estimates of a positive effect on hours worked from parental caregiving suggests we should be paying attention to how the financial strain of eldercare may be inducing additional paid labor.