Running head: PRIVATE AND PUBLIC EDUCATION

This article investigates cross-national mathematics and science achievement differences between public and private schools. Using the TIMSS 2003 data, we empirically examine differences through a set of multilevel models that attempt to control for select student background factors. We also attempt to correct for selection bias using propensity score matching methods. A number of methodological issues including the treatment of missing data and the construction of a quality student background measure are also addressed. While our analysis generally supports previous findings of higher private school achievement, we have found that higher private school achievement is not uniform across educational systems or the content domains analyzed. This variation is significant in light of the blanket privatization policy currently promoted by large international organizations. Public and Private Schools 3 Private and Public Education: A Cross-National Exploration with TIMSS 2003 In the past several decades a range of policy actions have created a space for the promotion of private educational systems throughout the world. For example, as part of an education initiative, the World Bank, (incidentally, the world’s largest external funder of education) has created a collaborative of researchers to explore the benefits of public-private partnerships in education. As recently as June 2007, the World Bank sponsored a conference to discuss the benefits of the privatization of education. Case studies which codified the merits of private education were presented (World Bank, 2007). Papers presented at this conference as well as work done within the broader privatization project assume the dominant paradigm in most economic based educational research, which suggests that private educational institutions outperform their publicly funded peer institution cross-nationally. Further, when discussion moves to the global level of discourse, benefits to privatization often become the prevailing way of discussing educational reform. However, whether the benefits of privatization are robust is unclear. Further, it is uncertain whether the private school advantage is consistent around the world. Therefore, this paper aims to investigate the magnitude of the private school achievement advantage in and across a number of educational systems. Historically, Friedman (1955) positioned himself as an early proponent of private education, who argued that similar to private industry, free markets in the educational sector would lead to improved educational institutions. This line of argument continued unabated over the next 50 years. Institutions such as the World Bank and authors such as Tooley and Dixon (2006) continue Friedman’s legacy by attempting to demonstrate the benefits of private schools. Public and Private Schools 4 The World Bank’s conviction to the superior performance of private educational systems is one example of how the organization is attempting to introduce free market principles to educational governance. Internationally and, especially, within the United States, a number of organizations and institutes also spend significant time and resources advocating the benefits of a private school model. While the World Bank clearly has the largest (and therefore most influential) pocketbook, organizations such as the Cato Institute, the Adam Smith Institute and the Futures of Freedom Foundation all unconditionally support a broad based private school model nationally and around the world. By what criteria are researchers and funding organizations concluding a private school advantage? While measures of private school success typically vary, evidence of superior performance at the national and international level is often presented via differences in standardized assessment achievement (for example, Fuchs & Woessman, 2007; Woessman, 2003). Using these data allow for a low cost examination of benefits; however, previous findings have been based on what we believe are less optimal methods or a failure to fully consider the context. As such, the current paper seeks to investigate whether private schools have higher mathematics and science achievement than their public school counterparts on the 2003 Trends in International Mathematics and Science Study (TIMSS). TIMSS provides data resulting from a reliable international assessment of fourth and eighth grade students on the mathematics and science achievement. Using what we reason are the most appropriate and up to date methods, we aim to better understand the relationship between institution type and achievement in a number of educational systems. Public and Private Schools 5 Theoretical Debate Educational privatization notions are closely linked to neo-classical theories that promote free markets as the most efficient way of providing a product or service (education) to the customer (students and parents). Friedman’s (1955) theories on the privatization of education were and continue to be supported by researchers who contended that public educational institutions lacked incentives to improve educational systems. Among others, Chubb and Moe (1990) and Coleman (1997) argued that allowing school choice, mainly through the promotion of private schools, would improve educational markets. In 1995, forty years after his seminal work in education, Milton Friedman continued to posit, “the only way to make a major improvement in our educational system is through privatization to the point at which a substantial fraction of all educational services is rendered to individuals by private enterprises” (online). Aligned with basic economic principles of supply and demand, these privatization advocates argued that increased marketization of education would spur competition, which would ultimately lead to improved quality. Unfortunately, the debates surrounding the privatization of education are not straightforward. In particular, economists such as Friedman acknowledge that education results in some public good and therefore deserves funding from the public sector. However, the neoclassical approach to education typically views most benefits of education as individual. As such, compulsory public schooling violates this economic tenet. Further, it is argued that public schools have become an inefficient bureaucracy, accountable for very little and often with little or no incentive to improve. Finally, neo-classical economic theory argues if the commodity of education is allowed to enter the free market, providers will have an incentive to meet customers’ needs or risk sanction and closure. Public and Private Schools 6 In the United States and around the world, privatization is often packaged as a voucher system, whereby students choose a school within certain parameters and the government provides to the chosen institution an allotment for that student. While appeasing privatization advocates by giving students limited choices, voucher systems often manage to also placate detractors who argue that private school access is exclusive and limited to children of the wealthy. Vouchers are also seen as a way to preserve school autonomy while making public funding available (Buchanan & Tullock, 1962; Epple & Romano, 1998). Despite the popularity of voucher systems, benefits of a broad based privatization approach remain unclear (Witte, 1999; Manski, 1992). In the U.S., Rouse (1998) indicated a small benefit to school choice in mathematics achievement with mixed results in reading while other researchers found no significant achievement advantage for the school choice models (Bettinger, 2005; Levin, 1998; Mayer, Peterson, Myers, Tuttle, & William, 2002). Internationally, findings are also equivocal (McEwan & Carnoy, 2000; West, 2001, Angrist et al., 2002; Bradley & Taylor, 2002). Finally, Miron and Nelson (2002), argue that more research about the effects of these types of school models is needed. Factors, such as the socio-economic status (SES) of students and the community in which they learn as well as the student background, may account for higher private school performance relative to their public peers. While there exists a wealth of research comparing the outcomes of private and public schools, it is the intention of this paper to analyze the differences at the national level across several nations. A paucity of studies in this particular area exists in spite of heavy private school promotion internationally by large organizations that typically fund these types of reforms. The following includes a short summary of studies that have addressed the difference between public and private schools. Public and Private Schools 7 Existing Research The literature concerning public and private school effectiveness is extensive and vastly international. Inspiring a line of research and discussion in the United States, Coleman, Hoffer, and Kilgore (1981) argued that a positive private school effect was evident in United States high school level data. However, Coleman, Hoffer, and Kilgore’s work was soon criticized by a number of authors to include Goldberger and Cain (1982), who argued that the assessment methods used fell “below the minimum standards for social scientific research” (p. 103). In part, what can be viewed as a response to critics of their initial study, Coleman and Hoffer (1987), along with others such as Chubb and Moe (1990) and Bryk, Lee, and Holland (1993) provided further evidence that private institutions (mostly in the form of Catholic schools) outperform their public counterparts. Internationally, a number of studies using a variety of methods and data attempted to test the private effect. For example, in Indonesia, Bedi and Garg (2000) and in India, Kingdon (1996) found positive private school effects on labor market earnings. However, in a study of relatively new private schools in Tanzania, Lassibille and Tan (2001) found that public students outperformed their private peers in terms of national assessment achievement. Moreover, Dronkers (2001) reviewed differences between privately managed religious schools and their public counter parts in several European countries, with findings mostly in favor of religious private schools. Finally, Somers, McEwan, and Willms (2004) use multi-level models to test the private school effect in a number of Latin American countries. The authors found no consistent or strong effect for private schools when controlling for socioeconomic status (SES) and peer effects. While the above research informed the present project, our focus is international and LSA evidence based in that it attempts to examine the difference in educational attainment in public and privately managed schools when using the same assessment cross nationally. Public and Private Schools 8 To date, a small number of studies have attempted to use international LSA data to examine the difference in student performance in both public and private schools. In two studies, Woessmann (2003) and Fuchs and Woessman (2007), utilize TIMSS and the Programme for International Student Assessment (PISA) data, respectively, in an attempt to explain international variation in student performance in both assessments. While the focus of both studies is not explicitly concentrated on the difference in educational attainment between private and public institutions, both studies find higher private achievement. Woessmann claimed that “students in countries with larger shares of enrolment in privately managed educational institutions scored statistically significantly higher in both mathematics and science” (p. 149). The 2007 study found that publicly funded, privately managed schools outperform their publicly funded, publicly managed and privately funded, privately managed counterparts. Surprisingly, Woessman (2003) contends that “there is no significant variation in many institutional features within a single country” (p. 120). As a result, the author aggregates all data to the international level. This tendency may introduce a host of problems including a shift of meaning. Vandenberghe and Robin (2004) utilized the PISA data to show that for three of nine included educational systems, private schools outperform public schools in a number of subjects. While relatively sophisticated methods were used to control for selection bias (a possibly significant problem in survey analysis research), the hierarchical structure of the data is ignored, rendering the significance of findings questionable. Using multilevel modeling to account for the clustered structure of the data, Lubienski and Lubienski (2006) used the National Assessment of Educational Progress (NAEP) data to show that after controlling for demographic differences, there appears to be little to no statistically significant benefit to private education on standardized mathematics test scores when Public and Private Schools 9 compared to public systems in the United States. Also using NAEP results, Braun, Jenkins and Grigg (2006) found that after controlling for select student and school variables, no differences exist between public schools and private schools (undifferentiated by possible religious affiliation) on 8 grade mathematics achievement. It is our intent to extend this approach and adjust for home background and community SES levels while using techniques that account for the unique structure and attributes of educational LSA data. In particular, we fit multilevel models to account for the structure of the data. Further, we incorporate a method used in Vandenberghe and Robin (2004) in an attempt to correct for systematic differences between public and private school students. Specifically, we implement a propensity score matching analysis (Rosenbaum & Rubin, 1983). With the resultant matched data, we fit ordinary leastsquares regression models with robust standard errors to compare against the multilevel models. Research Question At the international level a dominant paradigm is emerging, which assumes that private educational institutions are superior to their public counterparts. However, as recent studies have shown there is a continued need to examine the hypothesized advantage in light of a variety of factors. Our overall research question is as follows: for educational systems that indicated a public-private distinction in the TIMSS international data base, is there a significant difference between public and private schools in terms of mathematics and science achievement after accounting for school SES and select student background variables? To address this question, a number of methodological research issues arose in this study including: what is the best possible background index that can be created from the data at hand? Also, what is the most appropriate method for accounting for missing data in the TIMSS data set? Additionally, a policy relevant research question is discussed: To what extent do educational systems differ in their results? The Public and Private Schools 10 theoretical debate and policy controversy surrounding the benefits of private educational systems are ongoing and constantly evolving. As such, the debates should be informed by and updated with the latest and most appropriate data and methodological tools. Using the TIMSS 2003 data and up to date multilevel modeling software and selection bias methods, we empirically investigate the above research questions.