School efficiency in low and middle income countries: An analysis based on PISA for development learning survey
Introduction
Education, understood as a process of accumulation of human capital, can be hampered by inefficient factors that prevent reaching the maximum possible educational level for a given level of spending. Hence, efficiency on education provision, where schools make the best use of available inputs, had become a topic of intense debate among educational stakeholders (De Witte and López-Torres, 2017). To empirically contribute to this debate within the context of resource-constrained education systems is paramount, not only because of the few available resources which need to be efficiently administered, but also because over the last decade there has been an increasing educational access (Filmer et al., 2018) yielding new challenges in educational quality and equity. This is why the analysis of school efficiency and its comparison among lower and middle income countries (LMIC) is policy relevant. Ultimately, efficient and equitable education systems can lead to higher levels of economic growth and social cohesion (Woessmann, 2008).
Some studies (e.g., Almlund et al., 2011, Chernyshenko et al., 2018, Cordero et al., 2016) emphasise both the relevance of cognitive and non-cognitive dimensions of educational results. However, within the school efficiency literature, studies have been silent with regards to the efficiency of non-cognitive outcomes and their relationship with cognitive efficiency. Thus, it is necessary to incorporate the non-cognitive dimension in efficiency analysis, particularly in LMIC educational contexts, where the detrimental effects of school violence and weak learning environments on students’ outcomes are well documented and compounded by the added class heterogeneity brought via multigrade teaching (Evans and Yuan, 2018, Murillo and Román, 2011). Moreover, even though earlier research on developed countries (Grosskopf et al., 2017, Woessmann, 2007) finds a lack of evidence for a negative relationship between efficiency and equity, a policy relevant question is whether this also holds for LMIC education systems since their expansion would make schools to operate with lower student's inputs. Knowing if schools’ efficiency can be increased in parallel to narrowing the gaps among sub-populations in the distribution of educational outcomes is, therefore, crucial.
In this paper, using Data Envelopment Analysis (DEA) (Thanassoulis, 2001) which mimics an education production function where school inputs are transformed into outputs (Levin, 1974, Hanushek and Woessmann, 2011), we estimate the efficiency scores of 1180 schools in seven low and middle income countries from three different regions (Ecuador, Guatemala, Honduras and Paraguay from Latin America; Senegal and Zambia from sub-Saharan Africa; and Cambodia from the Southeast Asia) and the determinants of their efficiency levels. The analysis is based on the 2017 PISA for Development (PISA-D). This allows us to contribute to the school efficiency literature in different ways. First, we present new evidence on technical school efficiency and its drivers for less well-off education systems, thereby extending recent findings from developed countries on schools efficiency based on international learning surveys (e.g., Afonso and Aubyn, 2006, Agasisti and Zoido, 2018, Sutherland et al., 2009). Second, we estimate not only the standard efficiency frontier based on cognitive outputs but also a non-cognitive frontier. Third, we provide a detailed analysis assessing the efficiency-equity trade off based on a range of inequality indicators. Our analysis relies on recent bootstrap methods yielding bias-correct estimates of efficiency scores and their determinants within a two-stage DEA (Simar and Wilson, 2007, Badunenko and Tauchmann, 2019). Specifically, we explore the following questions:
- 1.
What is the degree of cognitive and non-cognitive efficiency for LMIC and their relationship? And does efficiency vary across countries?
- 2.
Which generic students, teachers, schools characteristics as well as policy factors are associated with both types of efficiency?
- 3.
What is the relationship of equity with efficiency for the whole sample of LMIC and between countries?
- 4.
Are generic factors associated with efficiency fundamentally different among the seven LMIC?
The remaining of the paper is organised as follows. The next section contains a brief background review on efficiency in education (more background is presented when outlining the data). In Section 3, we describe the data and the methods used. Section 4 presents the findings for the four research questions. Conclusions and policy implications are included in the last section.
Section snippets
Literature review
There is an extensive body of empirical work that studies the efficiency of education at different educational levels, but largely for developed countries based on international learning surveys (Afonso and Aubyn, 2006, Agasisti and Zoido, 2018, Cordero et al., 2017, Sutherland et al., 2009). An exhaustive and up to date review of the literature on efficiency in education can be found in De Witte and López-Torres (2017). The main methodology used by empirical research on school efficiency is
Data and sample
We use data from the 2017 PISA for Development (PISA-D), a learning survey which includes students from seven low and middle income countries/economies at grade 7 (15 years-old) or above. The countries included in the survey are: Cambodia, Ecuador, Guatemala, Honduras, Paraguay, Senegal and Zambia. The aim of PISA-D is to be more relevant to education systems with larger populations of poor and marginalised students which, in turn, is achieved through two distinctive elements (OECD, 2018).1
Research Question 1: efficiency scores for cognitive and non-cognitive outputs, and country differences
Table 2 presents the efficiency scores of schools based on the cognitive and non-cognitive DEA analysis. Different estimates are presented: technical efficiency (TE) standard estimates, a bootstrap bias-corrected version with 95% CI, and their bootstrap performance. To begin with, we focus on bootstrap bias-corrected estimates (TEBC) of scores for the whole sample (column 2, Panel A). Because the estimated mean efficiency score for the PISA-D sample is 0.80 and 0.78 for cognitive and
Conclusions and policy implications
An international cross-country comparative analysis of the determinants of efficiency across less developed economies is essential because of their unique educational challenges. This requires a comparable database fitted to the realities of poorer education systems. In this paper, we used the PISA-D dataset which allowed us to analyse the determinants of school efficiency at the secondary level in low- and middle-income countries (LMIC). Our analysis employed the DEA technique to come up with
Conflict of interest
No potential conflict of interest was reported by the authors.
Acknowledgements
We would like to thank the OECD for offering the PISA-D dataset online at: http://www.oecd.org/pisa/pisa-for-development/database/. We are grateful for helpful comments from two referees. This paper is dedicated to Boris Carpio.
References (51)
- et al.
Cross-country efficiency of secondary education provision: a semi-parametric analysis with non-discretionary inputs
Econ. Modell.
(2006) - et al.
Personality psychology and economics
Handbook of the Economics of Education, Vol. 4
(2011) - et al.
Efficiency and equity in private and public education: a nonparametric comparison
Eur. J. Oper. Res.
(2010) - et al.
Estimating an educational production function for five countries of Latin America on the basis of the pisa data
Econ. Educ. Rev.
(2013) - et al.
The economics of international differences in educational achievement
Handbook of the Economics of Education, Vol. 3
(2011) - et al.
Aspiration failure: a poverty trap for indigenous children in Peru?
World Dev.
(2015) - et al.
Student achievement and efficiency in Missouri schools and the no child left behind act
Econ. Educ. Rev.
(2006) - et al.
Assessment of efficiency in basic and secondary education in TUNISIA: A regional analysis
Int. J. Educ. Dev.
(2016) - et al.
Estimation and inference in two-stage, semi-parametric models of production processes
J. Econometr.
(2007) The efficiency of Italian secondary schools and the potential role of competition: a data envelopment analysis using OECD-PISA 2006 data
Educ. Econ.
(2013)