The diverse effects of private competitors on public service performance: evidence from New Jersey’s school system

Abstract

An important question underlying government policies that aim to introduce competition from the private sector into public service markets, is whether the presence of privately managed service providers affects the performance of public providers. On the one side, neoclassical economic theory would predict that competition from the private sector has a positive effect on public service performance. However, on the other side, institutional theories would predict a negative, or no effect at all. Using a 9-year panel data set of New Jersey schools, we find some evidence for both models. In most cases, the presence of privately managed charter schools is associated with an increase in public school performance, but the expansion of charter schools is associated with a decrease in performance. Our findings suggest that the effects of private competitors on public service performance are diverse and that multiple mechanisms can be at play simultaneously.

Notes

1 The tests include the New Jersey Assessment of Skills and Knowledge (NJASK), the High School Proficiency Assessment (HSPA), and the Biology End-of-Course exam. Before the development of the NJASK for grades 3-8, New Jersey Department of Education tests students through the Elementary School Proficiency Assessment (ESPA) and the Grade Eight Proficiency Assessment (GEPA).

2 The NJDoE changed the requirements for the annual reports in 2011. Some items used for our empirical analysis (i.e., teacher’s quality and attendance rate) had not been reported after 2012. Thus, we only focus on the consistent data of public-school report cards which are available for the years 2002-2011.

3 To measure school performance, this study focuses on passing rates categorized into three levels: 1) advanced proficiency, 2) proficiency and 3) partially proficiency.

4 See Fernández‐Gutiérrez, James, and Jilke (2017) for a similar approach.

5 The values of HHI are calculated as ${\mathit{HHI}}_{it}=\frac{1-{\sum }_{s=1}^n{R}_{\mathit{sit}}^2}{1-\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$n$}\right.};$ where $R_{\mathit{sit}}$ is the fraction of enrollments of a traditional public school $s$ in a school district $i$ in an academic year $t,$ and $n$ indicates the numbers of traditional public schools within each school district.

6 After the academic year of 2011-2012, the reporting requirements for schools changed and some variables in our empirical model are not available. Nevertheless, when estimating our final models with the extended dataset (yet, without controlling for teacher quality, student-faculty ratio and attendance rate), the estimates including the additional three years of 2011-2014 are consistent in the effects of charter schools on the performance with our sample period of 2002-2011. Upon any request, the empirical estimates for the 12 years can be provided.

7 The variables for the numbers and the enrollments are converted to natural logarithm forms to resolve the skewness in their distribution.

8 Because of multicollinearity between charter school creation and our first two measures of charter school size, we cannot estimate the separate effects of charter creation and expansion simultaneously. As an alternative approach, the sub-sample enables us to compare the effects of charter school expansion on performance while ruling out differences across school districts that have and have not had any charter schools.

9 The system-GMM systematically estimates both the levels and the first difference equation as an alternative approach of the standard first difference GMM estimator (Arellano and Bond 1991).

10 The system-GMM regression results in Table A3 revealed that there is AR(1) but the GMM estimators do not follow AR(2). Furthermore, the joint validity of GMM instruments test shows no overidentified instruments, and the number of instruments (538) do not exceed the number of school districts (550). The test of exogeneity of examined instruments does not reject the null of exogeneity of instruments. Therefore, the GMM estimates can be used to robustness check the fixed-effects regression results.

11 Abbot districts are ‘disadvantaged’ school districts within New Jersey that were created in 1985 following the Abbott vs. Burke ruling which aims to ensure that its students received public funding in according to the State’s constitution, which resulted in 31 Abbott districts today (NJDoE, 2016).