Herbst, Chris M. Desouza, Kevin C. Alashri, Saud Kandala, Srinivasa Srivatsav Khullar, Mayank Bajaj, Vikash
IZA Discussion Papers 11741
This paper exploits novel data and empirical methods to examine parental preferences for child care. Specifically, we analyze consumer reviews of child care businesses posted on the website Yelp.com. A key advantage of Yelp is that it contains a large volume of unstructured information about a broad set of child care programs located in demographically and economically diverse communities. Thus our analysis relies on a combination of theory- and data-driven methodologies to organize and classify the characteristics of child care that are assessed by parents. We also use natural language processing techniques to examine the affect and psychological tones expressed in the reviews. Our main results are threefold. First, we find that consumers overall are highly satisfied with their child care provider, although those in higher-income markets are substantially more satisfied than their counterparts in lower-income markets. Second, the program characteristics most commonly evaluated by consumers relate to safety, quality of the learning environment, and child-teacher interactions. However, lower- and higher-income consumers evaluate different characteristics in their reviews. The former is more likely to comment on a program's practical features, such as its pricing and accessibility, while the latter is more likely to focus on the learning environment. Finally, we find that consumers in lower-income markets are more likely to display negative psychological tones such as anxiety and anger in their reviews, particularly when discussing the nature of their interactions with program managers and their child's interactions with teachers.
content analysis Yelp preferences early childhood education child care machine learning