Measuring the Effect of Quality Ratings on Child Care Choice

• In Event: 2-102 - Equal Access to Early Care and Education for All? A Look at Diverse Child Populations

Integrative Statement

A large body of research shows that the returns to high quality early childhood care and education can be substantial, especially for low-income children. Informational interventions, such as quality rating systems, are one promising tool for increasing the number of children in high quality early care. However, some recent research suggests that the quality measures that underpin ratings systems are not strongly related to parents' satisfaction with providers, (e.g. Bassok et al. 2018). A direct test of ratings is whether when ratings become available, households are more likely to choose the providers that receive high ratings. This uses a discrete choice model to estimate the effect of Minnesota's ParentAware Quality Rating and Improvement System on households' choice between available private child care providers.

We make two contributions to understanding child care choice. First, this is the first paper to fit a mixed multinomial logit model of provider choice to a statewide child care market. We model the relationship between choice of provider and the provider's characteristics, such as price, travel distance, and quality ratings. Second, the paper measures whether ParentAware is effective in encouraging more parents to choose centers that are rated as high quality.

Our study is based on unique longitudinal data on the Minnesota child care market during the period when ParentAware was introduced. We use Sojourner, Davis, and Lee's Minnesota Provider Panel, which includes information on enrollment, prices, and provider characteristics for licensed child care providers from 2012-2016. This is supplemented with administrative data from Minnesota CCAP, the state's largest child care subsidy program, which includes information on choice of provider by subsidy-receiving households. Finally, we use census ACS 5-year estimates to provide demographic information by location at the census block group level.

Our methodology uses the Berry, Levinsohn, and Pakes (1995) (“BLP”) approach for estimating a mixed logit demand model. Several features of this model, which has been frequently used in the economics literature to study product choice, make it particularly suitable for studying child care choice. First, the model allows interactions between provider characteristics and household characteristics. For example, an important challenge in studying child care choice is that many options that are technically “available” to low income families may be practically unaffordable for those families, and this model allows us to incorporate that relationship. Second, the model allows us to fit the model to statewide data while restricting households to choice sets that depend on their location. Third, the model allows the use of instrumental variables. This allows us to obtain credible estimates of the effect of price on choice of provider, which in turn allows us to obtain money-metric estimates of the value of ParentAware ratings.

Analysis are still ongoing, but preliminary results indicate that the ratings have a modest effect on child care choice. According to these preliminary results, ParentAware ratings are strongly associated with what parents care about in choosing a provider, and have a modest “treatment” effect when the ratings become available.

Authors

• Jonathan Borowsky, University of Minnesota
  Presenting Author

• Elizabeth Davis, University of Minnesota
  Non-Presenting Author

• Aaron Sojourner, University of Minnesota
  Non-Presenting Author