Non-public competition and public school performance: evidence from West Virginia

ABSTRACT

In this study, we investigate whether non-public school enrolment affects the performance of public school districts. If homeschooling and private schools act as competition, public school districts test scores should be positively associated with non-public enrolment. Using data on West Virginia county school districts, and controlling for endogeneity with an instrumental variables approach, we find that a one standard deviation increase in relative non-public enrolment in a county is associated with statistically significant improvements in public school district test scores. Our findings thus confirm that non-public enrolment and the competition it provides act to improve, rather than impede, public school performance.

KEYWORDS:

• Competition
• markets
• education
• homeschooling

JEL CLASSIFICATION:

• H52
• I28

Acknowledgements

Hall would like to acknowledge that a portion of this work was conducted while he was a Big XII Fellow with the Free Market Institute at Texas Tech University. He also would like to thank the Center for Free Enterprise at West Virginia University for general research and travel support.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ See, for example, Ravitch (2011).

² See, for example, Forster (2013).

³ In addition, there is choice across districts. In West Virginia, however, the size of county districts means that there is less traditional public school ‘choice’ when compared to states like Ohio where school districts are not required to be at the county level. Ohio has 611 school districts and 88 counties, while West Virginia has 55 counties and 55 school districts.

⁴ Competition has also been found to lead to other results such as higher public school teacher salaries (Vedder and Hall 2000). It should be noted that some studies do not find a positive relationship between competition and school performance (Simon and Lovrich 1996; Sander 1999; Geller, Sjoquist, and Walker 2006; Marlow 2010). More recent papers on the relationship between local school competition and student performance such as Ponzo (2011), Misra, Grimes, and Rogers (2012), and Thapa (2013) have also found positive results in very different circumstances and situations.

⁵ Public choice scholars have a long history of studying education and its effects. For example, Cebula (1977) looks at how attitudes towards spending affect migration, while Flowers (1975), Akin and Lea (1982), Denzau and Grier (1984), Burnell (1991), Marlow and Orzechowski (1996), Aaberge and Langørgen (2003), Holcombe and Kenny (2007), and Holcombe and Kenny (2008), use schools to look at various public choice aspects related to public expenditures such as the influence of unions and fiscal illusion.

⁶ Although WESTEST only covers grades 3–11, we are not able to deduce enrolment by grade for homeschooled students. As such, we are can only look at the aggregate effect of non-public completion on the district as a whole, rather than competition at different grade levels, which might be preferable.

⁷ Not all papers find a statistically significant relationship between these variables and test scores. For example, Hall (2007) actually finds a negative and statistically significant relationship between district expenditures per pupil and Ohio school district math scores. Hall and Vedder (2003) find a negative, but not statistically significant, relationship between class size and a number of different test scores at the district level in Ohio.

⁸ See, for example, Roby (2004).

⁹ While income and private school enrolment are positively correlated it is important to note that this leads to higher standard errors and a reduced likelihood of finding a significant relationship but does not change the estimated coefficients. Additional demographics available include the percentage of district residents with a bachelor’s degree and racial data. These data are highly collinear with mean family income but produce similar results when substituted for income in our regressions. In addition, Hall and Leeson (2010) discuss the difficulty in interpreting a variable such as non-white in the context of district level regressions.

¹⁰ 4.53 × 0.282 = 1.277.

¹¹ It should also be noted that when a student transfers to a private school or is homeschooled, total school spending per pupil usually rises as locally raised revenue does not change (although state revenue that is based on enrolment will be reduced). It is an empirical question whether public school districts are able to achieve cost savings from enrolment declines in the short-run given the lumpy nature of much of their costs, such as teachers or transportation.

¹² 7.93 × 0.325 = 2.57.

¹³ Reading = [4.53 × 0.332 = 1.50], social studies = [4.53 × 0.270 = 1.22], and science = [4.53 × 0.282 = 1.27].

¹⁴ Spending per student might not matter since some spending categories could have a positive relationship with student test scores, other categories might have a negative relationship. Such as spending on public school administrators, see Anderson, Shughart, and Tollison (1991).

¹⁵ 1.18 × 1.739 = 2.05.

¹⁶ For more on reducing truancy and improving school attendance see Reid (2013).

¹⁷ In many states in the northeast and midwest, county borders and school district borders are not congruous (Ross, Hall, and Resh 2014). This makes West Virginia, despite the number of counties only equaling 55, a great state to investigate the effect of private school competition given the correspondence between county-level data for instruments and current county school districts.

¹⁸ For example, see the arguments raised in opposition to a pilot program on public charter schools in West Virginia (Hodousek 2015).

Additional information

Funding

This work was supported by the Center for Free Enterprise, West Virginia University; Free Market Institute, Texas Tech University.