Changing Levels of School Autonomy as a Mediator of the Private School Effect Across Countries

• In Event: Poster Session: Cross-national research using large-scale international data

Abstract

OBJECTIVE
This paper aims to explore a mechanism that can be manipulated by policymakers in terms of explaining differences in student performance across public and private schools internationally. The paper explores whether the level of autonomy given to teachers or principals mediates the private school effect.

CONCEPTUAL FRAMEWORK
Many school privatization supporters claim school autonomy contributes to the superior performance of private schools as compared to public schools (Chubb & Moe, 1990; Jimenez & Sawada, 1998). School autonomy or School Based Management has been proposed as one of the few school level factors that can be altered to contribute to improved educational outcomes (Wobmann 2003, Hoxby & Rockoff 2004, Clark 2009, Galiani & Schargrodsky 2002, Naper 2010, World Bank, 2007). On of the reasons given by many school autonomy advocates is that school autonomy leads to greater teacher satisfaction, retention and less absenteeism which contributes to a positive school climate and improved student performance (Lee et al, 1991). This paper examines the relationship between public and private schools to see if autonomy is a mediating variable for superior performance of private schools. In addition this paper examines the principal arguments of why this might take place.

The first research question is does school autonomy appear to mediate the private school effect internationally? The main research question of interest is to what extent does a change in the level of school autonomy in public and private schools act as a mediator of the private school effect? The change in relative power of teachers, principals, and outside school actors is also examined in order to gain insight into whether within-school autonomy or principal autonomy is more strongly related to student performance in secondary schools internationally.

ANALYTICAL METHODS
The mediation analysis is conducted using the methods from Barron and Kenny (1986). According to Barron and Kenny three routes of influence must be checked before a final analysis with the mediating variable included in an equation of the dependent and independent variables of interest is conducted. This study uses school type (public, government-dependent private, independent private) as the independent variable (X), the measures of school autonomy as the mediating variable (M), and math (and reading) performance as the dependent variable (Y).

Autonomy is split into autonomy in resource allocation and autonomy in curricula and assessments. These are split up to reflect the differences found in previous research between resource related autonomy versus academic autonomy. Both variables are developed using the school questionnaire from a series of questions for which the principal is asked to report whether “principals”, “teachers”, a “school governing board”, a “regional or local education authority”, or a “national education authority” has considerable responsibility for each of a series of choices.

DATA
All countries that participated in the school and student questionnaire of the PISA exam in 2009 and 2012 are included in the analysis (additional models are run with 2006 PISA data, but the sample size drops when requiring participation in all three exams/surveys). Schools are clustered by school type (public, private government dependent, and private government independent). A multiple cross-section design is used, with multi-level modeling to account for the nested structure of the data (additional analyses are run with a country level fixed effect). Longitudinal models are also run with the unit of analysis being public, private subsidized, and private un-subsidized schools within countries. For these models each unit of analysis represents the average of the schools in that category from each year of the study. This is done similarly to schools being given the characteristics of their students in a given year. Analysis is done with lagged autonomy scores and with change scores.

RESULTS
Initial findings show that school level curricular autonomy as well as some aspects of teacher and principal involvement in school resource and curricular decisions mediate the private school effect on student level math and reading performance in countries participating in PISA 2009 and PISA 2012. While autonomy may not completely explain the private school effect, it is a factor that can be manipulated.

SIGNIFICANCE
These results have important policy implications, as the level of autonomy given to schools is an aspect of education policy that can be altered by policymakers. Other factors that scholars have found to explain the private school effect, such as factors related to student selection (SES, peer effects, etc) are not as easily manipulated through policy levers as teacher and principal autonomy. This work contributes to the field of comparative education by exploring mechanisms for differences in international school performance.

Author

• Ngaire Honey, Vanderbilt University