Cost-benefit information closes aspiration gaps – if parents think their child is ready for college

ABSTRACT

Can information close socioeconomic gaps in parents’ postsecondary aspirations for their children? We administer a survey experiment to a nationally representative sample of U.S. parents, who are also asked whether their child is academically prepared for college. We inquire whether parents prefer their child to pursue a four-year degree, two-year degree, or no further education. Some parents are also randomly told the costs of college for an individual in their state and income bracket; local labor-market returns to a degree; or both costs and returns. Information closes socioeconomic aspiration gaps only if parents believe their child is ready for college.

KEYWORDS:

• College access
• parental college aspirations
• academic preparation
• socioeconomic gaps
• college costs and returns
• information
• parents

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Bleemer and Zafar (2018) do show that beliefs about the magnitude of the wage premium for earning a four-year degree for the general population seem to influence beliefs about the magnitude of the wage premium for their specific child.

2 See Chang and Krosnick (2009) for an analysis of the KN’s online survey strategy. For prior scholarly use of this survey, see Barrows et al. (2016), Chingos, Henderson, and West (2012), Lergetporer et al. (2018), and Schueler and West (2016).

3 As a robustness check, we estimated impacts with an alternate classification of the preparedness variable. We created a binary variable that took on the value of 1 if parents replied ‘very prepared’ or ‘somewhat prepared’ when asked to assess the level of preparedness of their child for college-level academic work. The results remained qualitatively unchanged. Due to loss of statistical power, some coefficients are less precisely estimated. However, the direction and magnitude of the results are consistent with those in the main analysis, and the statistical significance of several estimates is retained.

4 Our sample size of 1785 differs from the sample size of 1839 our summary statistics due to some missing data in the control variables. The loss of 54 observations is qualitatively trivial (2.9 percent), and in a formal check to see if attrition from the sample was systematically different across treatment conditions, we fail to reject the null hypothesis that attrition was the same across all treatment conditions.

5 Recall that parental estimates are compared to actual net costs of public higher education institutions in respondents’ state by respondents’ household income, and estimated returns are compared to actual returns in respondents’ commuting zone.

6 We conducted additional analyses to see if the magnitude of the error in their estimates of college costs and returns relative to the true values mediated the treatment impacts. In an additional analysis, we ran our main empirical model but included a continuous variable that indicated the magnitude of the error as well as a variable that interacted that term with the treatment condition. Coefficient estimates on these new terms were not statistically significant.

7 Given the marked differences in the institutional setting for higher education in Germany, we do not discuss the estimates provided by Lergetporer et al. (2019).

8 For example, one could ask parents how confident they are of their estimates of costs and returns prior to the survey experiment. It is hypothesized that the experiment would show lower levels of confidence on the part of lower than higher income parents and that the effect size of interventions would be larger for lower than higher income parents, if the child is perceived to be academically prepared.