Background & Approach

Determinants of Grade 12 Determinants of Grade 12 Pass Rates in the post-Pass Rates in the post-

Apartheid Apartheid South African South African Schooling SystemSchooling System

Haroon Bhorat and Morne Haroon Bhorat and Morne OosthuizenOosthuizen

Development Policy Research Unit, University of Cape Town

E-mail: [email protected]: www.commerce.uct.ac.za/dpru/

Background & ApproachBackground & Approach ““There is no blueprint for a model school that There is no blueprint for a model school that

can be reproduced and handed out to can be reproduced and handed out to policymakers, and such a blueprint is unlikely policymakers, and such a blueprint is unlikely to be developed in the near future”to be developed in the near future” (Hanushek,1995)(Hanushek,1995)

Massive and Swift Fiscal Reallocation:Massive and Swift Fiscal Reallocation:

1980s:1980s: R1.00R1.00 spent on White pupils, while spent on White pupils, while expenditure on each African pupils stood at expenditure on each African pupils stood at 19c19c

1997: R1.00 spent on African pupils, 71 cents per 1997: R1.00 spent on African pupils, 71 cents per White pupil.White pupil.

Background & ApproachBackground & Approach

Achievement Production Function Achievement Production Function Approach.Approach.

Specifically: How do the different Specifically: How do the different covariates simultaneously impact on covariates simultaneously impact on (school) average Grade 12 pass rates?(school) average Grade 12 pass rates?

Data principally drawn from the SRN Data principally drawn from the SRN 2000, Matric Results 2000 and 2000, Matric Results 2000 and Census EA-level data for 2001.Census EA-level data for 2001.

Descriptive Statistics: SchoolsDescriptive Statistics: SchoolsVariable/Former Department African White TotalMatric Pass Rate 48.39 95.14 55.47Pupil-Teacher Ratio 31.04 23.72 30.26Schools with User Fees 0.20 0.92 0.31Lowest Grade Offerred 7.88 6.34 7.6Non-std. classroom:learner ratio 0.0018 0.0006 0.0017Specialist classroom:learner ratio 0.004 0.028 0.008Principal Office 0.733 0.984 0.749Desks per learner 0.54 0.91 0.59At least 1 library 0.34 0.91 0.42Computer for Teaching & Learning 0.09 0.89 0.22Telecommunications 0.72 1.00 0.78Electricity for Lighting 0.71 0.98 0.73Sports Facilities 0.62 0.96 0.65Criminal Incident in Previous year 0.50 0.63 0.51Sample Size 4,019 565 5,610

Descriptive Statistics: Descriptive Statistics: EAsEAs

Variable/Former Dept.African

White Total

Share Rural 0.71 0.08 0.59Mean Household Size 4.75 4.37 4.69Children per Household 1.71 0.82 1.58Share of Informal Housing 34.67 4.42 29.6Share of Households Without: Piped Water 29.09 1.61 24.37Electricity 41.89 5.82 35.72Telephone 87.19 37.82 78.37Mean Years of Schooling for Adults 6.62 10.85 7.28

Matric Pass Rates,2000Matric Pass Rates,2000

0.0

5.1

.15

.2kd

ensi

ty m

r1_2

000p

ass

0.0

05.0

1.0

15.0

2kd

ensi

ty m

r1_2

000p

ass

0 20 40 60 80 100x...

African Col/Ind New Total White

Figure 1: Matric Pass Rateby Former Department

Econometric ApproachEconometric Approach Through ordinary least squared (OLS) Through ordinary least squared (OLS)

estimation, we derive a sample mean estimation, we derive a sample mean The sample median can be derived The sample median can be derived

through through quantile regression quantile regression approach approach by minimising the sum of by minimising the sum of the absolute residualsthe absolute residuals

Quantile Reg.: Estimation at different Quantile Reg.: Estimation at different points in the conditional distribution of points in the conditional distribution of the dependent variablethe dependent variable

Estimation DifficultiesEstimation Difficulties School-level data, so no intra-classroom School-level data, so no intra-classroom

variationvariation Omitted Variable Bias [parental and Omitted Variable Bias [parental and

teacher variables weak]teacher variables weak] Measurement Error [Quality of DoE Measurement Error [Quality of DoE

datasets]datasets] Selection Bias [School drop-out rates]Selection Bias [School drop-out rates] Not true post-Not true post-apartheid apartheid estimatesestimates

Production Function Results Production Function Results (1)(1)

Dependent Variable: Matric Pass RateOLS Median[1] [3]

Pupil- Teacher ratio -0.002 0.005Independent 3.011* 1.826Lowest 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 -2.768Used for ABET -0.728 -0.582Principal’s Office -0.837 -0.596Accom. For Staff 3.462* 2.347*Tuckshop 0.840 0.843Boards per classroom -0.075 -0.079Seats per learner 0.346 -0.050Desks per learner 1.455* 1.094Overhead p.l 8.747 3.714Photocopier p.l 124.346 223.967

Production Function Results Production Function Results (1) (1) contd.contd.

Dependent Variable: Matric Pass Rate

OLS Median

[1] [3]Library 2.634* 2.330*Computer for teaching 10.392* 11.173*Computer for admin 6.680* 7.199*Phone 3.028* 3.889*Water Indoors -0.016 -0.328Electricity 2.591* 3.619*Sports facilities 2.477* 3.429*Crime Incident -2.663* -2.492*Col./Indian School 13.039* 16.208*White School 26.904* 29.159*New School 1.908* 3.258*

Production Function Results Production Function Results (1) (1) contd.contd.


OLS Median

[1] [3]

Rural 1.162 1.611

Household Size 0.134 0.137

Children per hh. -1.067* -0.603

Adult mean yrs of schooling 0.813* 0.726*

Poverty Index -0.182 -0.800

Sample Size 5014 5014

Pseudo R2 0.440 0.301

Key ResultsKey Results Insignificance of pupil-teacher ratios in Insignificance of pupil-teacher ratios in

determining the mean or the median pass ratedetermining the mean or the median pass rate Physical infrastructure: Almost all classroom Physical infrastructure: Almost all classroom

resources insignificant in shaping pass rates. resources insignificant in shaping pass rates. butbut very specific variables are significant, namelyvery 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 & Environm. factors e.g. crime, electricity &

telecomm. Influential. telecomm. Influential. Classification Dummies are critical and reflect Classification Dummies are critical and reflect

composite of important omitted variablescomposite of important omitted variables Household Variables: Household Variables:

Location & Asset Poverty insignificantLocation & Asset Poverty insignificant Dependency Ratios & Adult Years of Schooling SignificantDependency Ratios & Adult Years of Schooling Significant

Production Function Results Production Function Results (2)(2)


Percentile10th

(21.4%)25th

(33.6%)75th

(77.3%) 90th

(96.7%) Pupil- Teacher ratio -0.010 0.009 -0.015 -0.033*Independent -1.553 1.641 4.294* 2.705Lowest grade -0.166* -0.025 -0.248* -0.256*Non-std. classroom:learner ratio -219.887* -107.299* -19.379 44.605Specialist classroom:learner

ratio -8.166 -10.782* -2.251 6.828Platoon School -3.091 -4.275* 2.189 3.975*Used for ABET -1.008 -0.337 -0.399 -1.330Principal’s Office 0.551 0.800 -2.496* -1.157Accom. For Staff 1.890 1.320* 2.455* 2.352*Tuckshop 3.613* 0.672 -0.603 -1.129Boards per classroom -0.011 -0.085 -0.148 0.096Seats per learner 0.603 -0.172 -0.727* -0.277Desks per learner 1.022 1.301* 0.631 0.801Overhead p.l 6.958 5.399 3.535 22.054*Photocopier p.l -337.833* 152.276 529.648* 151.735

Pupil-Teacher Ratios Pupil-Teacher Ratios RevisitedRevisited

-.1-.0

50

.05

Coe

ffici

ent o

n P

upil-

Teac

her R

atio

0 .2 .4 .6 .8 1Quantiles

Quantiles OLS

Coefficient Estimates of Pupil-Teacher Ratio Impact on Grade 12 Pass Rates

Production Function Results Production Function Results (2) c(2) contd.ontd.


Percentile

10th 25th 75th 90th

Library 1.812 2.568* 1.588* 1.260

Computer for teaching 9.807* 10.389* 9.201* 6.189*

Computer for admin 3.269* 5.128* 9.843* 9.478*

Phone 1.630 2.577* 3.517* 2.795*

Water Indoors 1.007 0.165 -0.351 0.095

Electricity 0.146 3.254* 3.499* 1.594*

Sports facilities 1.395 1.688* 4.095* 2.703*

Crime Incident -1.967* -2.059* -1.525* -2.847*

Col./Indian School 19.460* 18.568* 11.037* 5.536*

White School 47.229* 41.621* 16.967* 9.660*

New School 1.238 -0.747 3.258* 1.804

Production Function Results Production Function Results (2) c(2) contd.ontd.


Percentile

10th 25th 75th 90th

Rural 1.257 0.222 2.739* 1.223

Household Size 0.092 0.028 0.132 0.355*

Children per hh. -1.037 -0.540 -0.727 -1.529*

Adult mean yrs of schooling 1.009* 0.720* 0.905* 0.579*

Poverty Index -0.654 0.512 -0.483 0.495

Sample Size 5014 5014 5014 5014

Pseudo R2 0.237 0.275 0.303 0.220

Adult SchoolingAdult Schooling.6

.81

1.2

0 .2 .4 .6 .8 1y

quantiles OLS

Quantile and OLS Coefficients on Adult Schooling

Five Composite Results Five Composite Results 1.1. The The Pupil-Teacher RatioPupil-Teacher Ratio is insignificant in is insignificant in

explaining the performance of all schools – barring explaining the performance of all schools – barring those in the 80th percentile upward.those in the 80th percentile upward.

2.2. Relative unimportance of Relative unimportance of Physical Classroom Physical Classroom ResourcesResources. Boards, desks and seats have low . Boards, desks and seats have low explanatory power.explanatory power.

1.1. Caveat: Non-std. classrooms do matterCaveat: Non-std. classrooms do matter3.3. Knowledge InfrastructureKnowledge Infrastructure, and access to services , and access to services

and utilities critical in explaining relative and utilities critical in explaining relative performance. performance.

4.4. Teacher and Parental CharacteristicsTeacher and Parental Characteristics matter. matter. 1.1. significant results for onsite staff significant results for onsite staff

accommodation; adult years of schooling and the accommodation; adult years of schooling and the classification dummiesclassification dummies

5.5. A core group of poorly resourced, rural-based high-performing A core group of poorly resourced, rural-based high-performing former Homeland Schools – requires closer analysisformer Homeland Schools – requires closer analysis

Production Function Production Function Results (3)Results (3)


Inter-Quantile Range

90th - 10th 90th - 50th 50th - 10th

Pupil- Teacher ratio -0.023 -0.038 0.015Independent 4.257 0.878 3.379Lowest grade -0.09 0.028 -0.118Non-std. classroom:learner

ratio 264.493** 179.144 85.348Specialist classroom:learner

ratio 14.994 29.406 -14.411Platoon School 7.066** 6.743** 0.323Used for ABET -0.323 -0.748 0.426Principal’s Office -1.708 -0.561 -1.146Accom. For Staff 0.462 0.006 0.456Tuckshop -4.741* -1.972** -2.770**Boards per learner 0.107 0.175 -0.068Seats per learner -0.88 -0.227 -0.653Desks per learner -0.22 -0.293 0.072Overhead p.l 15.096 18.34 -3.243

Photocopier p.l489.568**

* -72.232 561.800*




90th - 10th 90th - 50th 50th - 10th

Library -0.552 -1.069 0.517Computer for teaching -3.618*** -4.984* 1.366Computer for admin 6.209* 2.28 3.929**Phone 1.165 -1.094 2.259***Water Indoors -0.912 0.49 -1.402Electricity 1.448 -2.025 3.474Sports facilities 1.308 -0.727 2.035***Crime Incident -0.88 -0.355 -0.525Col./Indian School -13.924* -10.671* -3.253White School -37.569* -19.499* -18.071*New School 0.567 -1.454 2.021Unspecified Sch. 5.223 -4.644 9.868***




90th - 10th 90th - 50th 50th - 10th

Rural -0.033 -0.388 0.354Household Size 0.263 0.218 0.046Children per hh. -0.491 -0.926 0.435Adult mean yrs of schooling -0.43 -0.147 -0.284Poverty Index 1.149 1.295 -0.146Constant 53.097* 37.318* 15.780*High Quantile R2 0.22 0.22 0.301Low Quantile R2 0.237 0.301 0.237

Key Relative Key Relative Performance Results Performance Results

Inherited Socio-Economic Factors are Inherited Socio-Economic Factors are Insignificant i.t.o Relative PerformanceInsignificant i.t.o Relative Performance

P-T Ratio is Insignificant in Explaining P-T Ratio is Insignificant in Explaining Relative PerformanceRelative Performance

Non-Std. Classrooms & Platoons increase Non-Std. Classrooms & Platoons increase dispersion in pass ratesdispersion in pass rates

Administrative efficiency and knowledge Administrative efficiency and knowledge infrastructure matter for reducing the infrastructure matter for reducing the performance gapperformance gap

Classification Dummies: Very Strong EffectClassification Dummies: Very Strong Effect

[Early][Early] Policy Policy SuggestionsSuggestions

Know Which Portion of the Performance Spectrum you Know Which Portion of the Performance Spectrum you want to Influence, as Determinants are Differentwant to Influence, as Determinants are Different

Do not Invest in Reducing Classroom SizeDo 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 Investment in Knowledge Infrastructure will reap

rewardsrewards Try and Better Understand what is going on in Former Try and Better Understand what is going on in Former

Homeland Schools (Natural Experiments?)Homeland Schools (Natural Experiments?) Importance of Classification Dummies Suggests that Importance of Classification Dummies Suggests that

many teacher, pupil characteristics are critical, and we many teacher, pupil characteristics are critical, and we need to understand what components impact on pass need to understand what components impact on pass rates.rates.

Socio-Economic Status Not As Critical as may have been Socio-Economic Status Not As Critical as may have been assumed.assumed.

Documents

Background & Approach