|Women’s labor force participation has skyrocketed in the past half-decade. With this shift, the demand for child care has also risen dramatically. Women increasingly need child care during work hours, whether it be from a relative, a family day-care or child care center. From an economic perspective, paid child care adds to the cost of employment. Thus child care cost may be a disincentive to employment especially affecting women as primary care-givers in the home.
Current welfare policy in the United States includes child care subsidies as a crucial tool intended to facilitate the employment of low-income mothers. Since the 1996 welfare reform, employment is a required goal for welfare participants, making child care subsidies ever more important to the extent that they may encourage or even enable employment.
Many past studies have tried to measure the effect of the price of child care on employment. However, subsidies are in many ways less effective than a reduction in the price of all child care. Not many people know about them, many people who apply are waitlisted or rationed out of the system even if they are eligible, there are bureaucratic delays, and there is stigma associated with taking up a subsidy. To find out whether or not subsidies increase employment in reality, I chose to examine the effect of actual subsidy receipt on employment.
Although the theoretical effect of subsidies on employment is clear, a causal effect is difficult to discern in the data. The difficulty lies in that women are often required to be employed to receive child care subsidies in the first place, and women who are likely to work may also be likely to get subsidies due to any number of unobserved personal characteristics, such as energy or intelligence.
Using data from the National Survey of America’s Families of 2002, I applied a number of econometric models to explore the relationship between subsidies and employment. I found that obtaining a child care subsidy increases employment by 28% using a two-stage least squares model, 42% using a bivariate probit model, and 23% using propensity score matching, though the result was not significant for the first model. These results confirm for the most part a positive relationship that may be causal.