1. Determinants of Grade 12 Pass Rates in the post-Apartheid South African Schooling System Haroon Bhorat and Morne Oosthuizen Determinants of Grade 12 Pass Rates in the post-Apartheid South African Schooling System Haroon Bhorat and Morne Oosthuizen Development Policy Research Unit, University of Cape Town E-mail: hbhorat@commerce.uct.ac.za Website: www.commerce.uct.ac.za/dpru/

2. Background & Approach “There is no blueprint for a model school that can be reproduced and handed out to policymakers, and such a blueprint is unlikely to be developed in the near future” (Hanushek,1995) “There is no blueprint for a model school that can be reproduced and handed out to policymakers, and such a blueprint is unlikely to be developed in the near future” (Hanushek,1995) Massive and Swift Fiscal Reallocation: Massive and Swift Fiscal Reallocation: 1980s: R1.00 spent on White pupils, while expenditure on each African pupils stood at 19c 1980s: R1.00 spent on White pupils, while expenditure on each African pupils stood at 19c 1997: R1.00 spent on African pupils, 71 cents per White pupil. 1997: R1.00 spent on African pupils, 71 cents per White pupil.

3. Background & Approach Achievement Production Function Approach. Achievement Production Function Approach. Specifically: How do the different covariates simultaneously impact on (school) average Grade 12 pass rates? Specifically: How do the different covariates simultaneously impact on (school) average Grade 12 pass rates? Data principally drawn from the SRN 2000, Matric Results 2000 and Census EA-level data for 2001. Data principally drawn from the SRN 2000, Matric Results 2000 and Census EA-level data for 2001.

4. Descriptive Statistics: Schools Variable/Former DepartmentAfricanWhiteTotal Matric Pass Rate 48.3995.1455.47 Pupil-Teacher Ratio 31.0423.7230.26 Schools with User Fees 0.200.920.31 Lowest Grade Offerred 7.886.347.6 Non-std. classroom:learner ratio 0.00180.00060.0017 Specialist classroom:learner ratio 0.0040.0280.008 Principal Office 0.7330.9840.749 Desks per learner 0.540.910.59 At least 1 library 0.340.910.42 Computer for Teaching & Learning 0.090.890.22 Telecommunications 0.721.000.78 Electricity for Lighting 0.710.980.73 Sports Facilities 0.620.960.65 Criminal Incident in Previous year 0.500.630.51 Sample Size 4,0195655,610

5. Descriptive Statistics: EAs Variable/Former Dept.AfricanWhiteTotal Share Rural 0.710.080.59 Mean Household Size 4.754.374.69 Children per Household 1.710.821.58 Share of Informal Housing 34.674.4229.6 Share of Households Without: Piped Water 29.091.6124.37 Electricity 41.895.8235.72 Telephone 87.1937.8278.37 Mean Years of Schooling for Adults 6.6210.857.28

6. Matric Pass Rates,2000

7. Econometric Approach Through ordinary least squared (OLS) estimation, we derive a sample mean Through ordinary least squared (OLS) estimation, we derive a sample mean The sample median can be derived through quantile regression approach by minimising the sum of the absolute residuals The sample median can be derived through quantile regression approach by minimising the sum of the absolute residuals Quantile Reg.: Estimation at different points in the conditional distribution of the dependent variable Quantile Reg.: Estimation at different points in the conditional distribution of the dependent variable

8. Estimation Difficulties School-level data, so no intra-classroom variation School-level data, so no intra-classroom variation Omitted Variable Bias [parental and teacher variables weak] Omitted Variable Bias [parental and teacher variables weak] Measurement Error [Quality of DoE datasets] Measurement Error [Quality of DoE datasets] Selection Bias [School drop-out rates] Selection Bias [School drop-out rates] Not true post-apartheid estimates Not true post-apartheid estimates

9. Production Function Results (1) Dependent Variable: Matric Pass Rate OLS Median [1][3] Pupil- Teacher ratio-0.0020.005 Independent3.011*1.826 Lowest grade-0.264*-0.284* Non-std. classroom:learner ratio -89.044*-134.539* Specialist classroom:learner ratio -14.393-22.578* Platoon School-1.442-2.768 Used for ABET-0.728-0.582 Principal’s Office-0.837-0.596 Accom. For Staff3.462*2.347* Tuckshop0.8400.843 Boards per classroom-0.075-0.079 Seats per learner0.346-0.050 Desks per learner1.455*1.094 Overhead p.l8.7473.714 Photocopier p.l124.346223.967

10. Production Function Results (1) contd. Dependent Variable: Matric Pass Rate OLS Median [1][3] Library2.634*2.330* Computer for teaching10.392*11.173* Computer for admin6.680*7.199* Phone3.028*3.889* Water Indoors-0.016-0.328 Electricity2.591*3.619* Sports facilities2.477*3.429* Crime Incident-2.663*-2.492* Col./Indian School13.039*16.208* White School26.904*29.159* New School1.908*3.258*

11. Production Function Results (1) contd. Dependent Variable: Matric Pass Rate OLS Median [1][3] Rural1.1621.611 Household Size0.1340.137 Children per hh.-1.067*-0.603 Adult mean yrs of schooling0.813*0.726* Poverty Index-0.182-0.800 Sample Size5014 Pseudo R 2 0.4400.301

12. Key Results Insignificance of pupil-teacher ratios in determining the mean or the median pass rate Insignificance of pupil-teacher ratios in determining the mean or the median pass rate Physical infrastructure: Almost all classroom resources insignificant in shaping pass rates. but very specific variables are significant, namely Physical infrastructure: Almost all classroom resources insignificant in shaping pass rates. but very specific variables are significant, namely Non-standard classrooms and Staff Accomm. Non-standard classrooms and Staff Accomm. Knowledge Infrastructure critical. Knowledge Infrastructure critical. Environm. factors e.g. crime, electricity & telecomm. Influential. Environm. factors e.g. crime, electricity & telecomm. Influential. Classification Dummies are critical and reflect composite of important omitted variables Classification Dummies are critical and reflect composite of important omitted variables Household Variables: Household Variables: Location & Asset Poverty insignificant Location & Asset Poverty insignificant Dependency Ratios & Adult Years of Schooling Significant Dependency Ratios & Adult Years of Schooling Significant

13. Production Function Results (2) Dependent Variable: Matric Pass Rate Percentile 10 th (21.4%) 25 th (33.6%) 75 th (77.3%) 90 th (96.7%) Pupil- Teacher ratio-0.0100.009-0.015-0.033* Independent-1.5531.6414.294*2.705 Lowest grade-0.166*-0.025-0.248*-0.256* Non-std. classroom:learner ratio -219.887*-107.299*-19.37944.605 Specialist classroom:learner ratio -8.166-10.782*-2.2516.828 Platoon School-3.091-4.275*2.1893.975* Used for ABET-1.008-0.337-0.399-1.330 Principal’s Office0.5510.800-2.496*-1.157 Accom. For Staff1.8901.320*2.455*2.352* Tuckshop3.613*0.672-0.603-1.129 Boards per classroom-0.011-0.085-0.1480.096 Seats per learner0.603-0.172-0.727*-0.277 Desks per learner1.0221.301*0.6310.801 Overhead p.l6.9585.3993.53522.054* Photocopier p.l-337.833*152.276529.648*151.735

14. Pupil-Teacher Ratios Revisited

15. Production Function Results (2) contd. Dependent Variable: Matric Pass Rate Percentile 10 th 25 th 75 th 90 th Library1.8122.568*1.588*1.260 Computer for teaching9.807*10.389*9.201*6.189* Computer for admin3.269*5.128*9.843*9.478* Phone1.6302.577*3.517*2.795* Water Indoors1.0070.165-0.3510.095 Electricity0.1463.254*3.499*1.594* Sports facilities1.3951.688*4.095*2.703* Crime Incident-1.967*-2.059*-1.525*-2.847* Col./Indian School19.460*18.568*11.037*5.536* White School47.229*41.621*16.967*9.660* New School1.238-0.7473.258*1.804

16. Production Function Results (2) contd. Dependent Variable: Matric Pass Rate Percentile 10 th 25 th 75 th 90 th Rural1.2570.2222.739*1.223 Household Size0.0920.0280.1320.355* Children per hh.-1.037-0.540-0.727-1.529* Adult mean yrs of schooling1.009*0.720*0.905*0.579* Poverty Index-0.6540.512-0.4830.495 Sample Size5014 Pseudo R 2 0.2370.2750.3030.220

17. Adult Schooling

18. Five Composite Results 1. The Pupil-Teacher Ratio is insignificant in explaining the performance of all schools – barring those in the 80th percentile upward. 2. Relative unimportance of Physical Classroom Resources. Boards, desks and seats have low explanatory power. 1. Caveat: Non-std. classrooms do matter 3. Knowledge Infrastructure, and access to services and utilities critical in explaining relative performance. 4. Teacher and Parental Characteristics matter. 1. significant results for onsite staff accommodation; adult years of schooling and the classification dummies 5. A core group of poorly resourced, rural-based high-performing former Homeland Schools – requires closer analysis

19. Production Function Results (3) Dependent Variable: Matric Pass Rate Inter-Quantile Range 90 th - 10 th 90 th - 50 th 50 th - 10 th Pupil- Teacher ratio-0.023-0.0380.015 Independent4.2570.8783.379 Lowest grade-0.090.028-0.118 Non-std. classroom:learner ratio 264.493**179.14485.348 Specialist classroom:learner ratio 14.99429.406-14.411 Platoon School7.066**6.743**0.323 Used for ABET-0.323-0.7480.426 Principal’s Office-1.708-0.561-1.146 Accom. For Staff0.4620.0060.456 Tuckshop-4.741*-1.972**-2.770** Boards per learner0.1070.175-0.068 Seats per learner-0.88-0.227-0.653 Desks per learner-0.22-0.2930.072 Overhead p.l15.09618.34-3.243 Photocopier p.l489.568***-72.232561.800*

20. Production Function Results (3) Dependent Variable: Matric Pass Rate Inter-Quantile Range 90 th - 10 th 90 th - 50 th 50 th - 10 th Library-0.552-1.0690.517 Computer for teaching-3.618***-4.984*1.366 Computer for admin6.209*2.283.929** Phone1.165-1.0942.259*** Water Indoors-0.9120.49-1.402 Electricity1.448-2.0253.474 Sports facilities1.308-0.7272.035*** Crime Incident-0.88-0.355-0.525 Col./Indian School-13.924*-10.671*-3.253 White School-37.569*-19.499*-18.071* New School0.567-1.4542.021 Unspecified Sch.5.223-4.6449.868***

21. Production Function Results (3) Dependent Variable: Matric Pass Rate Inter-Quantile Range 90 th - 10 th 90 th - 50 th 50 th - 10 th Rural-0.033-0.3880.354 Household Size0.2630.2180.046 Children per hh.-0.491-0.9260.435 Adult mean yrs of schooling-0.43-0.147-0.284 Poverty Index1.1491.295-0.146 Constant53.097*37.318*15.780* High Quantile R 2 0.22 0.301 Low Quantile R 2 0.2370.3010.237

22. Key Relative Performance Results Inherited Socio-Economic Factors are Insignificant i.t.o Relative Performance Inherited Socio-Economic Factors are Insignificant i.t.o Relative Performance P-T Ratio is Insignificant in Explaining Relative Performance P-T Ratio is Insignificant in Explaining Relative Performance Non-Std. Classrooms & Platoons increase dispersion in pass rates Non-Std. Classrooms & Platoons increase dispersion in pass rates Administrative efficiency and knowledge infrastructure matter for reducing the performance gap Administrative efficiency and knowledge infrastructure matter for reducing the performance gap Classification Dummies: Very Strong Effect Classification Dummies: Very Strong Effect

23. [Early]Policy Suggestions Know Which Portion of the Performance Spectrum you want to Influence, as Determinants are Different Know Which Portion of the Performance Spectrum you want to Influence, as Determinants are Different Do not Invest in Reducing Classroom Size Do not Invest in Reducing Classroom Size Invest in Learning Infrastructure….but Invest wisely! Invest in Learning Infrastructure….but Invest wisely! Investment in Knowledge Infrastructure will reap rewards Investment in Knowledge Infrastructure will reap rewards Try and Better Understand what is going on in Former Homeland Schools (Natural Experiments?) Try and Better Understand what is going on in Former Homeland Schools (Natural Experiments?) Importance of Classification Dummies Suggests that many teacher, pupil characteristics are critical, and we need to understand what components impact on pass rates. Importance of Classification Dummies Suggests that many teacher, pupil characteristics are critical, and we need to understand what components impact on pass rates. Socio-Economic Status Not As Critical as may have been assumed. Socio-Economic Status Not As Critical as may have been assumed.