Does regulation induce homogenisation? An analysis of three voucher programmes in the United States

ABSTRACT

We examine whether voucher programme regulation reduces private school specialisation in the US. We employ school-and-year fixed-effects regression and examine individual private schools in Washington, DC, Indiana, and Louisiana as they transition into voucher programme environments. We leverage the Private School Universe Survey to observe how schools self-identify before and after switching into voucher environments. We find that upon switching into school voucher environments, private schools in all 3 locations are more likely to identify as less specialised than they were prior to entering the programmes. We find suggestive evidence that the homogenising effects may be stronger in more heavily regulated voucher programme environments, and that those schools in more lightly regulated environments are more likely to continue highlighting their specialised approach to education. These findings are examined within an institutional theory framework to understand the potential homogenising effect of regulations on the diversity of the private school market.

KEYWORDS:

• Private school
• school choice
• schooling supply
• institutional theory
• isomorphism

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Corey A. DeAngelis is an education policy analyst at the Cato Institute’s Center for Educational Freedom. He received his PhD in education policy from the University of Arkansas.

Lindsey M. Burke is the director of the Center for Education Policy at the Heritage Foundation and a PhD candidate in education policy and research methods at George Mason University.

ORCID

Corey DeAngelis http://orcid.org/0000-0003-4431-9489

Notes

1 As a robustness check, we use a random effects probit regression model as well. Since individual schools choose to identify as one of several different types of institutions, we also use a multinomial probit regression model as a robustness check. However, since our sample sizes are all too small to rely on maximum likelihood estimation, our base school and year fixed-effects regression model is preferred (Deke, 2014; Wooldridge, 2016). Nonetheless, our results tables also note when effects are robust to alternative specifications.