Revisiting Ethnic Differences in In-Person Learning During the 2021-22 School Year

ABSTRACT

During the 2020–21 school year, Black students were less likely to learn in person than white students. We examine whether this difference persisted as the pandemic progressed. We find that the rate of in-person learning increased in 2021–22 but remained lower for Black students compared to white students. While several factors helped explain observed racial differences in learning modality in 2020–21, only modality offerings continued to be an important explanatory factor in 2021–22 and a Black-white in-person learning gap persisted after controlling for offerings and other factors. These findings suggest a mismatch between Black families’ preferences and the options offered by their schools.

KEYWORDS:

• COVID-19
• parental choice
• racial gaps
• schooling modality

Acknowledgments

The project described in this article relies on data from survey(s) administered by the UAS, which is maintained by the Center for Economic and Social Research at the University of Southern California. The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of USC or UAS. The collection of the UAS COVID-19 tracking data is supported in part by the Bill & Melinda Gates Foundation and by grant U01AG054580 from the National Institute on Aging, and many others.

Disclosure statement

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

Data availability statement

The data that support the findings of this study are available with registration from the Understanding America Study at https://uasdata.usc.edu/.

Notes

1. The summer 2021 survey (UAS 348) was fielded from June 9^th to July 21^st, the fall 2021 survey (UAS 350) was fielded from September 23^rd to October 31^st, and the winter 2022 survey (UAS 351) was fielded from February 1^st to March 30^th.

2. https://uasdata.usc.edu/index.php

3. Importantly, the UAS research team provides internet access and hardware (e.g., tablets) to respondents who do not have computer hardware or internet access so all households in the panel may participate. Respondents receive compensation for their time spent answering questions at a rate of $20 per 30 minutes of interview time. The surveys are conducted both in English and Spanish.

4. While the UCA averages 7,000 respondents per survey wave, we limit our analysis to those respondents with school-aged children in the household with resulting sample sizes between 1,225 and 1,458 individual respondents. This proportion is comparable with the number of American households with school-aged children (https://www.census.gov/data/tables/2020/demo/families/cps-2020.html)

5. Our sample includes any adult living in a household with a child in K-12 schooling which may include extended family members or adult siblings. As a robustness check, we repeat our analyses for summer and fall samples restricted to parents of K-12 students as identified from the separate “My Household” survey in Appendix B.

6. For a small number of households (15) across all waves, there were multiple responses but no primary respondent. For these cases, we randomly select only one response per student in each household and wave and exclude the others from our analytic sample.

7. Throughout our analysis our sample size changes due to some missingness in survey responses. In Appendix C, we show that our findings are robust to these changes in sample composition by performing our analysis with each specification limited to the most restricted analytic sample. In additional analyses, available upon request from the authors, we find evidence that within-wave sample attrition between model specifications is largely uncorrelated with respondent race or ethnicity indicating that the key relationships we study here are likely unaffected by this attrition.

8. While we combine remote and hybrid learning for fall of 2021 into a single variable, we explore these as separate outcomes using a multinomial logit model following the logit specification described in the next section. Full results for this multinomial analysis are presented as average marginal effects in Appendix D.

9. Due to sample size limitations, we include non-Hispanic AIAN, Hawaiian or Pacific Islander, and mixed-race individuals in this other race category.

10. Detailed results from these factor analyses can be found in the technical Appendix A.

11. As this is constructed from one question, we are unable to build the measure using factor analysis. Instead, we include this as a continuous variable with lower values indicating lower levels of trust.

12. Comorbidities in the UCA survey include diabetes, cancer, heart disease, high blood pressure, asthma or a chronic lung disease, kidney disease, autoimmune disorders, and obesity.

13. When both surveys were fielded, individuals aged 12 and up were eligible for vaccination.

14. 45 respondents in the fall 2021 sample and 55 respondents in the winter 2022 sample indicated that their child attended a “virtual school” but did not differentiate between virtual schools operated by a public school district or charters. We exclude these individuals from our analysis.

15. We are unable to include these controls for the summer 2021 analysis.

16. https://uasdata.usc.edu/index.php

17. Note that weights aligned to the characteristics of U.S households with K–12 or higher education students are not provided in the UAS. Provided sample weights bring the sample in line with the U.S. adult population.

18. The summer survey did not ask respondents about their preference for remote or hybrid learning separately.

19. We performed similar robustness checks by interacting race and ethnicity indicators with our measures of public health trust, media trust, and COVID-19 related comorbidities but did not find meaningful patterns.

20. Full results are available upon request from the authors.