Covariate
Scenario 1
Scenario 2
Scenario 4
Scenario 5
Scenario 6
Relative importance (LMG) OLS estimate (clustered SE) Relative importance (LMG) OLS estimate (clustered SE) Relative importance (LMG) OLS estimate (clustered SE) Relative importance (LMG) OLS estimate (clustered SE) Relative importance (LMG) OLS estimate (clustered SE)
Teachers 34.3% 0.102*** (0.028) 23.3% 0.157*** (0.032) 20.6% 0.823*** (0.096) 2.1% 0.043 (0.034) 4.8% -0.039** (0.016)
Students 18.0% -0.002*** (0.001) 13.4% -0.004*** (0.001) 12.7% -0.016*** (0.003) 3.0% -0.000 (0.001) 3.5% 0.001* (0.000)
Over-age 0.3% 0.023 (0.020) 0.2% 0.054 (0.033) 0.3% -0.248*** (0.069) 1.8% -0.185*** (0.035) 12.8% -0.228*** (0.036)
No fee 2.2% -0.269 (0.588) 5.6% 0.329 (0.676) 0.4% -0.692 (1.516) 1.0% -1.120** (0.567) 5.0% 0.520* (0.272)
Grade (base: R) 0.4% NULL 0.1% NULL 2.1% NULL 0.6% NULL 1.3% NULL
Grade: 1 / 0.087* (0.053) / -0.060 (0.072) / -2.571*** (0.338) / -0.318*** (0.110) / -0.162*** (0.038)
Grade: 2 / 0.098 (0.062) / -0.060 (0.083) / -2.961*** (0.362) / -0.355*** (0.128) / -0.171*** (0.042)
Grade: 3 / 0.090 (0.068) / -0.069 (0.089) / -3.192*** (0.386) / -0.381*** (0.140) / -0.156*** (0.044)
Grade: 4 / 0.113 (0.103) / -0.147 (0.138) / -5.585*** (0.495) / -0.711*** (0.228) / -0.291*** (0.070)
Grade: 5 / 0.125 (0.104) / -0.146 (0.140) / -5.536*** (0.502) / -0.756*** (0.227) / -0.285*** (0.072)
Grade: 6 / 0.120 (0.104) / -0.169 (0.139) / -5.451*** (0.518) / -0.710*** (0.231) / -0.281*** (0.070)
Grade: 7 / 0.106 (0.107) / -0.175 (0.140) / -5.216*** (0.508) / -0.680*** (0.226) / -0.240*** (0.077)
Gender (base: female) 0.0% NULL 0.0% NULL 0.0% NULL 0.0% NULL 0.1% NULL
Gender: male / -0.005 (0.012) / 0.006 (0.015) / 0.040 (0.036) / 0.032*** (0.012) / 0.042*** (0.010)
Race (base: black) 5.8% NULL 8.0% NULL 4.5% NULL 8.3% NULL 16.8% NULL
Race: coloured / 0.101 (0.271) / -0.675* (0.382) / -4.505*** (0.924) / -2.160*** (0.540) / -1.226* (0.705)
Race: indian / 0.448 (0.454) / 0.711 (0.937) / 4.411 (3.179) / 0.142 (0.895) / -0.120 (0.150)
Race: other / -1.808*** (0.671) / -2.733*** (0.766) / -1.450 (2.541) / 1.311 (0.850) / 0.656* (0.378)
Race: white / -0.207 (0.334) / -0.755** (0.324) / 1.902*** (0.736) / -0.384 (0.785) / 0.186 (0.139)
Quintile (base: 1) 8.9% NULL 15.0% NULL 2.1% NULL 3.3% NULL 32.3% NULL
Quintile: 2 / 0.013 (0.249) / -0.799** (0.392) / -0.767 (0.820) / -0.617 (0.437) / 0.483* (0.275)
Quintile: 3 / -0.376 (0.244) / -0.314 (0.407) / -0.180 (0.766) / -0.821** (0.383) / -0.040 (0.282)
Quintile: 4 / -0.016 (0.344) / -0.892* (0.486) / -2.898** (1.184) / -1.151** (0.550) / 0.858** (0.347)
Quintile: 5 / -0.510 (0.639) / -1.574** (0.769) / -3.785** (1.892) / -1.473** (0.708) / 1.423*** (0.336)
Province (base: ec) 23.6% NULL 31.2% NULL 6.8% NULL 63.3% NULL NA NA
Province: fs / -0.241 (0.774) / -0.945 (0.993) / -2.929* (1.700) / -2.110** (0.880) / NA
Province: gt / -0.856 (0.647) / -0.816 (0.829) / -1.178 (1.696) / 2.250** (0.952) / NA
Province: kz / -1.024* (0.565) / -0.451 (0.778) / -0.893 (1.673) / 2.552*** (0.932) / NA
Province: lp / -0.665 (0.540) / -0.031 (0.867) / 2.431 (1.768) / 3.074*** (0.966) / NA
Province: mp / -0.630 (0.870) / 0.693 (1.055) / 1.887 (2.128) / 3.779*** (1.131) / NA
Province: nc / -0.995 (1.141) / -1.112 (1.411) / -4.440** (2.256) / -5.835*** (1.028) / NA
Province: nw / -1.351* (0.777) / -1.681* (0.990) / -0.587 (2.330) / -0.337 (1.099) / NA
ECS 6.4% -0.005 (0.008) 3.2% 0.018 (0.011) 50.5% 0.495*** (0.026) 16.6% 0.074*** (0.019) 23.4% 0.024*** (0.005)
Observations 551,707 551,707 551,707 551,707 551,707 551,707 551,707 551,707 551,707 551,707
R² 0.085 0.085 0.106 0.106 0.551 0.551 0.414 0.414 0.100 0.100
Adj. R² 0.085 0.085 0.106 0.106 0.551 0.551 0.414 0.414 0.100 0.100
Clusters (emis) 499 499 499 499 499 499 499 499 499 499
  • Relative importance (LMG) via calc.relimp with 10 bootstrap replicates (95% percentile CIs).
  • OLS coefficients with heteroskedasticity-consistent (HC1) SEs clustered at emis (reported as b(SE)).
  • Class size (ECS) model includes educators, learners, and school/learner covariates.
  • Scenario models control for class size, educators, learners, and the same covariates; prov excluded when DV is prov_class_size_reduc.
  • Categorical regressors (Province, Race, Quintile, Grade, Gender) are treated as factors; LMG is reported at the group level, OLS at factor-level coefficients.
  • Significance: *** p<0.01, ** p<0.05, * p<0.10.
  • Scenario 3 (classroom_reduc) excluded from all analyses due to high (90%+) missingness.
  • Province coefficients may be NA where province effects are collinear with province-driven structure, likely because provinces share a binding factor when educators are pooled provincially.