Education and Country Growth Models

Education and country growth models

by

Martin Anders Gustafsson

Dissertation presented for the degree of Doctor of Philosophy (Economics) at Stellenbosch University

Supervisor: Prof Servaas van der Berg

April 2014

Declaration

By submitting this thesis electronically, I declare that the entirety of the work containedtherein is my own, original work, that I am the sole author thereof (save to the extentexplicitly otherwise stated), that reproduction and publication thereof by StellenboschUniversity will not infringe any third party rights and that I have not previously in itsentirety or in part submitted it for obtaining any qualification.

Martin Anders Gustafsson

April 2014

Abstract

The over-arching concern of the three parts of the dissertation is how economics can and should influence education policymaking, the emphasis on the economics side being models of country development and the contribution made by human capital.

Part I begins with a review of economic growth theory. How educational performance and country development have been measured is then discussed, with considerable attention going towards conceptual and measurement complexities associated with the latter. An approach is presented for expanding the number of countries whose educational quality can be compared, by expanding the number of linkable testing programmes. This approach, which above all allows for the inclusion of more African and Latin American countries, is one of the key contributions made by the dissertation to the existing body of knowledge. Three existing empirical growth models are examined, including work by Hanushek and Woessman on the relationship between educational quality and income. Part I ends with a discussion on how the economics literature can best be packaged to influence education policymaking. A ‘growth simulator’ tool in Excel for informing the policy discourse is presented. The production of this tool includes establishing empirically a feasible improvement trajectory for educational quality that policymakers can use and some analysis of how linguistic fractionalisation in a country evolves over time. This tool can be considered a further key output of the dissertation. A basic model for relating educational quality, via income growth, to teacher pay, is presented.

Part II offers an analysis of UNESCO country-level data on enrolment and spending going back to 1970, with a view to establishing historical patterns that can inform education planners, particularly those in developing countries, on how budgets and enrolment expansion should be distributed across the levels of the education system. The analysis presented in Part II represents a novel way of using existing country-level data and can be seen as an important step towards filling a gap experienced by education policymakers, namely the paucity of empirical evidence that can guide decisions around the prioritisation of education levels. Part II moreover arrives at a few empirical findings, including the finding that enrolment and spending patterns have been systematically different in countries with faster economic growth and the finding that historical per student spending at the secondary level appears to play a larger role in development than was previously thought.

Part III contrasts the available economic advice for education policymakers with what policymakers actually appear to believe in. The focus falls, in particular, on four developing countries: South Africa, Brazil, Chile and China. A few areas where economists could explore the data to a greater degree or communicate available findings differently, in the interests of better education policies, are identified. Part III partly serves as a demonstration of how comparisons between education systems can be better oriented towards providing advice to education policymakers on questions relating to efficiency and equity.

Opsomming

Die oorkoepelende fokus van die drie gedeeltes van die verhandeling is hoe die studie van ekonomie beleid in die onderwyssektor kan en moet beïnvloed. Veral belangrik is modelle van die ekonomiese groei van lande en die rol van menslike kapitaal in hierdie modelle.

Die eerste gedeelte van die verhandeling bied ʼn oorsig van die teorie rakende ekonomiese groei. Hoe onderwysprestasie en nasionale ontwikkeling gemeet word, word dan bespreek, met ʼn sterk fokus op die konseptuele en tegniese kompleksiteit van laasgenoemde. ʼn Metode word aangebied waardeur meer lande se onderwysgehalte vergelyk kan word, deur middel van die koppeling van data van ʼn groter aantal toetsprogramme. Hierdie metode, wat veral die insluiting van meer lande uit Afrika en Latyn-Amerika toelaat, is een van die kernbydraes van die verhandeling tot die bestaande korpus van kennis. Drie bestaande empiriese modelle van ekonomiese groei word geanaliseer, insluitende die werk van Hanushek en Woessman oor die verhouding tussen onderwysgehalte en inkomste. Die eerste gedeelte sluit af met ʼn bespreking oor hoe die ekonomiese literatuur optimaal aangebied kan word om beleidmaking in die onderwys te beïnvloed. ʼn Groei-simulasie hulpmiddel in Excel wat die beleidsdiskoers kan vergemaklik word aangebied en verduidelik. Die ontwikkeling van hierdie gereedskap maak dit moontlik om op ʼn empiriese basis ʼn moontlike trajek vir die verbetering van onderwysgehalte te bepaal, wat vir beleidsmakers nuttig kan wees, sowel as ʼn ontleding van hoe linguïstiese verbrokkeling in ʼn land histories kan ontwikkel. Hierdie gereedskap kan as ʼn verdere sleutelproduk van die verhandeling beskou work. ʼn Basiese model van hoe onderwysgehalte en die inkomste van onderwysers deur middel van ekonomiese groei gekoppel is, word ook aangebied.

Die tweede gedeelte van die verhandeling bied ʼn ontleding van UNESCO se nasionale statistieke van lande oor skoolinskrywings en onderwysuitgawes vanaf 1970, met die oog op die identifikasie van belangrike historiese tendense vir onderwysbeplanners, veral in ontwikkelende lande. Die fokus hier is veral op hoe begrotings en inskrywings ideaal oor die verskillende vlakke van die onderwysstelsel versprei behoort te wees. Die ontleding in die tweede gedeelte verteenwoordig ʼn innoverende manier om die bestaande nasionale statistieke te gebruik en kan beskou word as ʼn belangrike stap om ʼn gaping te vul wat deur beleidsmakers in die onderwys ondervind word, naamlik die gebrek aan empiriese gegewens vir besluite oor prioritisering tussen onderwysvlakke. Die tweede gedeelte bied ook verskeie empiriese bevindinge, soos dat die tendense rakende inskrywings en besteding per student sistematies tussen lande met vinniger ekonomiese groei en ander lande verskil, asook dat historiese besteding per student op die sekondêre vlak blykbaar ʼn groter invloed op ontwikkeling het as wat vroeër gedink is.

Die derde gedeelte van die verhandeling vergelyk die advies wat die ekonomiese literatuur aan beleidmakers in die onderwys bied met wat beleidmakers self blykbaar glo. Die fokus val op veral vier ontwikkelende lande: Suid-Afrika, Brasilië, Chili en China. Gebiede word bespreek waar ekonome in die belang van beter onderwysbeleid tot ʼn groter mate data kan analiseer of bevindings op beter maniere kan kommunikeer. Die derde gedeelte kan beskou word as ʼn demonstrasie van hoe vergelykings tussen verskeie onderwysstelsels beter georiënteer kan word om vir die beleidmaker in die onderwys advies te verskaf rakende kwessies van doeltreffendheid en gelykheid.

CONTENTS

The dissertation consists of three parts, listed below. Each part has its own table of contents, figures and tables. A conclusion for the dissertation as a whole and an integrated bibliography appear at the end of Part III.

Part I: An evaluation of existing cross-country models and some quantitative applications of relevance to education planning...... 6

Part II: Modelling optimal prioritisation across levels of education in the country development process...... 183

Part III: How policymakers in developing countries deal with internal and external efficiency issues...... 251

Education and country growth models

Part I

An evaluation of existing cross-country models and some quantitative applications of relevance to education planning

CONTENTS FOR PART I

1INTRODUCTION

2AN OVERVIEW OF KEY THEORETICAL MODELS

2.1Introduction

2.2Solow’s neoclassical model

2.3The augmented Solow model

2.4The AK model

2.5The Lucas model

2.6Product variety model

2.7Schumpeterian model of Nelson and Phelps

2.8Schumpeterian model with creative destruction

2.9Schumpeterian model with two levels of human capital

2.10Conclusion

3MEASURING HUMAN CAPITAL AND COUNTRY DEVELOPMENT

3.1Introduction

3.2Indicators of the quantity of education

3.3Indicators of schooling quality

3.4Economic growth indicators and their alternatives

3.5Growth versus sustainable development

3.6Conclusion

4A SELECTION OF EMPIRICAL GROWTH MODELS

4.1Introduction

4.2The augmented Solow model of Mankiw, Romer and Weil

4.3The BACE model with iterative variable selection

4.4Hanushek and Woessman’s model

4.5Conclusion

5AUGMENTATIONS TO INFORM EDUCATION POLICYMAKERS

5.1Introduction

5.2A problematic research-policy nexus

5.3Feasible education improvement targets

5.4Magnitude of educational and non-education under-performance

5.5The relative magnitude of education’s growth effect

5.6Educational improvement and teacher pay

5.7Conclusion

6CONCLUSIONS FOR PART I

Appendix A

Appendix B

Appendix C

Figures and tables

Figure 1: Solow’s neoclassical model

Figure 2: Solow’s neoclassical model in the long run

Figure 3: MRW’s augmented Solow model

Figure 4: MRW’s augmented Solow model in the long run

Figure 5: The AK model

Figure 6: The Lucas model

Figure 7: Product variety model

Figure 8: Nelson and Phelps

Figure 9: Schumpeterian with creative destruction

Figure 10: Schumpeterian with two levels of HC (I)

Figure 11: Schumpeterian with two levels of HC (II)

Figure 12: A swing towards innovation (I)

Figure 13: A swing towards innovation (II)

Figure 14: Programmes and bridge countries

Figure 15: Age-specific life expectancy

Figure 16: The derivation of the Bayesian posterior distribution

Figure 17: The BACE sum of squared errors

Figure 18: The progression of model sizeand the BACE SSE

Figure 19: The weighting of the BACE SSE

Figure 20: Effect of weights on mean model size

Figure 21: Effect of weights on posterior inclusion probability

Figure 22: Annual country score increases by level

Figure 23: Minimum years needed to improve

Figure 24: Minimum years needed to improve with actual examples

Figure 25: Growth trajectories in four countries

Figure 26: Relative sizes of development challenges

Figure 27: Malaria prevalence reduction I

Figure 28: Malaria prevalence reduction II

Figure 29: Linguistic fractionalisation in South Africa by age of adults

Figure 30: Contributing factors behind higher future growth

Figure 31: Present value of future changes

Figure 32: Relationship between relative teacher pay and GDP per capita

Figure 33: Evolution of the teacher pay ratio in South Africa I

Figure 34: Evolution of teacher pay in South Africa II

Figure 35: Future teacher pay in South Africa with quality improvements

Table 1: Summary of model features

Table 2: Reliability of measures of educational quantity

Table 3: Summary of the test data

Table 4: Programmes and number of bridge countries

Table 5: Variables for MRW model

Table 6: Regression outputs – replication of MRW

Table 7: Regression outputs – tests of convergence

Table 8: Regression outputs – MRW with quality of human capital

Table 9: Regression outputs – tests of convergence with educational quality data

Table 10: BACE regression outputs – without educational quality variables

Table 11: BACE regression outputs – effect of educational quality variables

Table 12: Variables reported on in BACE tables

Table 13: Data sources used for replicating H&W growth regressions

Table 14: Regression outputs – replication of H&W

Table 15: Regression outputs – with transformed explanatory variables

Table 16: Regression outputs – large range of explanatory variables

Table 17: Regression outputs – culture as an instrumental variable

Table 18: Regression outputs – 2nd stage with simple OLS

Table 19: Details on explanatory variables for growth

Table 20: Language use in South Africa amongst adults (2004)

Table 21: Adjusted country averages of pupil performance

Table 22: Feasible long-range improvement trend

ACRONYMS USED

The following are acronyms used across the three parts of the dissertation. Only acronyms that are used without an adjacent explanation within the text are listed here. For details on the ISO country codes used in many of the tables and graphs, see Table 21in Part I of the dissertation.

2SLS / Two-stage least squares
AIDS / Acquired immunodeficiency syndrome
BACE / Bayesian averaging of classical estimates
BAU / Business as usual
CGE / Computable general equilibrium
DHS / Demographic and Health Survey
GDP / Gross domestic product
GER / Gross enrolment ratio
GHI / Gross happiness index
GNI / Gross national income
GNP / Gross national product
HDI / Human development index
HIV / Human immunodeficiency virus
IALS / International Adult Literacy Survey
IAM / Integrated assessment model
ICT / Information and communication technology
IDEB / Index of Basic Education Development (from Portuguese Índice de Desenvolvimento da Educação Básica)
IEA / International Association for the Evaluation of Educational Achievement
IIASA / International Institute for Applied Systems Analysis
ILO / International Labour Organization
IPUMS / Integrated Public Use Microdata Series
ISO / International Organization for Standardization
IV / Instrumental variable
LFS / Labour Force Survey
MRW / Mankiw, Romer and Weil (1992)
NAEP / National Assessment of Educational Progress
OECD / Organisation for Economic Co-operation and Development
OLS / Ordinary least squares
PASEC / Programme for the Analysis of Educational Systems (from French Le Programme d’analyse des systèmes éducatifs de la CONFEMEN)
PIAAC / Programme for the International Assessment of Adult Competencies
PIRLS / Progress in International Reading Literacy Study
PISA / Programme for International Student Assessment
PPP / Purchasing power parity
RSS / Residual sum of squares
SACMEQ / Southern and Eastern Africa Consortium for Monitoring Educational Quality
SDM / Sala-i-Martin, Doppelhofer and Miller (2004)
SERCE / Second Regional Comparative Study (from Spanish Segundo Estudio Regional Comparativo y Explicativo)
SNA / System of National Accounts
SSE / Sum of squared errors
SSF / Stiglitz, Sen and Fitoussi (2009)
TIMSS / Trends in International Mathematics and Science Study
UIS / UNESCO Institute for Statistics
UK / United Kingdom
UN / United Nations
UNDP / United Nations Development Programme
UNESCO / United Nations Educational, Scientific and Cultural Organization
US / United States
USD / United States dollar
VAM / Vandenbussche, Aghion and Meghir (2006)
VID / Vienna Institute of Demography
WEF / World Economic Forum
WVS / World Values Survey
ZAR / South African rand

1INTRODUCTION

Key texts in the economic growth literature identify the economic growth process as a combination of imitation and innovation. Both are necessary, though less developed countries tend to depend more on imitation, as opposed to innovation. This dissertation can be thought of as a combination of imitation and innovation. The sections of the dissertation where imitation predominates take existing analyses and redo them, sometimes using more recent data, they compare the methodologies of existing analyses, and they test the degree to which existing models can assist in specific policymaking questions. The more innovative sections focus above all on using country-level data in new ways that can inform policymakers.

If there is an over-arching question that the three parts of the dissertation deal with it is perhaps the following: What is the utility of the economic growth literature and models using country-level data for education policymakers and to what extent have these policymakers made use of the available evidence? Part I of the dissertation is mainly a critical review of existing theory and empirical texts that could be of use to the policymakers. The various sections of Part I are introduced below. Part II addresses what appeared to be a gap in the literature, namely the need for greater empirical guidance with respect to the prioritisation of the various levels of the education system, in particular the secondary and post-school levels. Here, as in Part I, the focus is on models that use country-level data and implement some type of cross-country comparison. The justification for this focus is presented below. Part II is the more innovative part of the dissertation due to its emphasis on new ways of analysing the data.

Part III moves furthest from the over-arching question referred to above, yet the topics it covers are relevant for this question. The focus in Part III is on analysing the education policy directions of developing countries, specifically four fairly diverse developing countries from three continents, within an economic framework of internal and external efficiency. The aim of the analysis is to assess to what degree and how education policymakers have embraced the recommendations of economists, and what this means for the kind of conversation economists should be having with education policymakers. Part III is thus, like the two preceding parts of the dissertation, focussed on cross-country analysis, except in Part III this analysis uses not data points (as in the first two sections), but the institutional architecture and key dimensions of the education systems of the countries in question. To some extent Part III considers if and how policymakers have taken into account growth models in economics, but this is presented within a broader consideration of whether evidence and theory from the field of economics in general is influencing education policymaking. This broader approach seemed more interesting and feasible. The reception amongst policymakers to growth modelling specifically receives attention in Part I of the dissertation.

Turning to the sections of Part I, section 2 addresses the following question:What are the theoretical roots of economic models that link education to economic growth and how relevant is this theory for education planning? The focus here is on eight key theoretical models that have influenced the thinking and empirical modelling we see today. The models examined range from Solow’s (1956) neo-classical model to recent Schumpeterian creative destruction models. The models selected focus not only on how education and human capital contribute towards economic growth. They also deal with other contributing factors. Yet as will be seen, human capital became an increasingly central concern.Several of these models are from an era when a lack of data to a large degree forced economists to present theoretical (and mathematical) models with no or very little empirical evidence. There is some discussion of the historical developments, in particular the emergence of electronic data and computing,which helped economics adopt a more empirical approach, as well as the arguments that are still made for an economics discipline that is even better at using empirical analysis to test theory, partly so that policymakers can receive the advice they need.