Teachers’ unions and industrial action in South African primary schools: Exploring their impacts on learning

Figures & data

Table 1. Union membership in the South African education sector, 2012.

	Province
	WC	EC	NC	FS	KZN	NW	GP	MP	LP	National
Total registered teacher union members	26 792	62 193	8 469	23 436	91 293	25 643	55 589	33 965	53 446	380 826
All education personnel^ in 2012	46 184	88 667	13 394	36 246	127 303	34 522	105 572	48 300	74 151	574 339
% of personnel who are union members	58.0	70.1	63.2	64.7	71.7	74.3	52.7	70.3	72.1	66.3
% of all education personnel who are SADTU members	28.0	51.8	43.5	38.2	44.8	53.8	27.8	53.3	58.9	44.1
% of teacher union members who are SADTU members	48.3	73.9	68.8	59.1	62.5	72.4	52.7	75.8	81.8	66.4

Notes: Teacher union members include educators and a small number of support staff which may be public servants or employed privately at the school level. ^Total educator numbers are from the education payroll in November 2012 and SGB/privately remunerated staff identified in the SNAP 2012 data of ordinary schools (35 461).

Source: PSCBC, ELRC.

Table 2. Teacher strike activity by school wealth, SEACMEQ III 2007.

	Teacher sample in the …
	poorest 75% of schools	wealthiest 25% of schools	in all schools
Teacher strike participation (mean) (absent for at least one day for strike)	0.798	0.569	0.734
	(0.019)	(0.042)	(0.018)
Total days teachers are absent for strike (mean)	13.270	4.308	10.771
	(0.387)	(0.841)	(0.385)
Total days teachers are absent for strike (median)	15	1	14
N (teachers)	844	314	1 158
N (schools)	297	95	392

Notes: Standard errors are in parentheses. Data are weighted.

Source: SEACMEQ III.

Figure 1. Cumulative % graph of teachers’ strike absenteeism by school wealth status, SEACMEQ III 2007.

Table 3. Estimating grade 6 test scores.

	Wealthiest 25% of schools
	1		2		3		4
Teacher Controls:	OLS	Student FE	OLS	Student FE	OLS	Student FE	OLS	Student FE
Strike participation indicator	0.029	0.067**	0.044	0.071**	0.041	0.038	0.015	0.024
	(0.050)	(0.029)	(0.049)	(0.029)	(0.045)	(0.031)	(0.046)	(0.034)
R-squared	0.366	–	0.37	–	0.38	–	0.39	–
Within R^2	–	0.051	–	0.076	–	0.081	–	0.103
F-stat (p-value)	68 (0.000)	55 (0.000)	61 (0.000)	25 (0.000)	52 (0.000)	14 (0.000)	48 (0.000)	15 (0.000)
Subject-student (N)	5 587	5 587	5 587	5 587	5 587	5 587	4 936	4 936
Clusters (N)	84	1 868	84	1 868	84	1 868	83	1 825
Estimated Bias (Equation 7)		−0.31		−0.15		–		–
Ratio^a		–		–		–		–
	Poorest 75% of schools
Strike participation indicator	0.034	−0.078***	0.0412	−0.065***	0.027	−0.103***	0.002	−0.100***
	(0.046)	(0.018)	(0.045)	(0.018)	(0.056)	(0.023)	(0.065)	(0.027)
R-squared	0.150	–	0.154	–	0.162	–	0.164	–
Within R-squared	–	0.022	–	0.038	–	0.042	–	0.050
F-stat (p-value)	25 (0.000)	67 (0.000)	22 (0.000)	26 (0.000)	1 (0.000)	17 (0.000)	18 (0.000)	18 (0.000)
Subject-student (N)	19 114	19 114	19 114	19 114	19 114	19 114	17 446	17 446
Clusters (N)	280	6 386	280	6 386	280	6 386	278	6 319
Estimated Altonji Bias (Equation 7)		–		−1.022		−0.740		−1.035
Ratio (Altonji)^a		–		0.063		0.138		0.097
Variables determined before unionisation	No	No	Yes	Yes	Yes	Yes	Yes	Yes
Variables determined after unionisation	No	No	No	No	Yes	Yes	Yes	Yes
Teacher test score (limited sample)	No	No	No	No	No	No	Yes	Yes

Notes: *p < 0.1, **p < 0.05, ***p < 0.001. All OLS regressions include additional student and school characteristics and provincial dummies. Standard errors are in parentheses. ^aThe ratio of the coefficient on the strike participation indicator and the estimated bias. The bias is only calculated for the estimations where strike effects were significant and where additional controls are included for teacher characteristics.

Source: SEACMEQ III, own estimates.

Table 4. Estimating grade 6 test scores using a continuous variable for strike absenteeism, poorest 75% of schools.

	Poorest 75%
Days absent for …	OLS	Student FE	Variable mean
Strike	−0.005	−0.005***	14.7
	(0.003)	(0.001)	(0.4)
Own illness	0.003	0.002*	3.0
	(0.002)	(0.001)	(0.4)
Funeral	0.001	−0.006**	1.0
	(0.006)	(0.002)	(0.1)
Official business	−0.006**	−0.010***	2.0
	(0.003)	(0.002)	(0.2)
R-squared	0.165	–
Within R-squared	–	0.05
F-stat	18	18
N student-subject	17 446	17 446
N clusters	278	6 319

Notes: *p < 0.1, **p < 0.05, ***p < 0.001. All OLS regressions include additional student and school characteristics and provincial dummies. Standard errors are in parentheses. Weighted estimates.

Source: SEACMEQ III, own estimates.

Table 5. Student fixed effect estimations of grade 6 test scores, sub-samples.

	Urban/rural status of the student’s school		Average academic achievement of students in three subject tests
	Rural	Urban	Quartile 1	Quartile 2	Quartile 3	Quartile 4
Strike participation indicator	−0.165***	0.027	−0.107*	−0.080*	−0.071*	0.028
	(0.044)	(0.024)	(0.055)	(0.044)	(0.039)	(0.036)
Within R-squared	0.078	0.038	0.017	0.083	0.070	0.084
F-stat (p-value)	20 (0.00)	11 (0.00)	2.4 (0.00)	15 (0.00)	8.3 (0.00)	15 (0.00)
N subject-students	10 290	12 092	5 737	5 820	5 689	5 136
N clusters	3 700	4 500	2 100	2 100	2 100	1 900

Notes: Specification follows model 4 in Table 3. Samples are not limited to the 75% poorest schools. *p < 0.1, **p < 0.05, ***p < 0.01.

Source: SEACMEQ III, own estimates.

Table 6. School days lost due to strikes/boycotts by school feeding.

	School days lost due to strikes/boycotts. Principal response.
Modal response of grade 3s to the question ‘I eat food from the feeding scheme at my school’	Mean	Std. Err.	Lower CI	Upper CI	Median	N schools
Everyday	17.30	0.18	16.94	17.66	18	1 434
Most of the day	17.82	0.51	16.83	18.81	20	157
Some of the days	17.76	0.43	16.92	18.61	20	225
Never	10.64	0.48	9.70	11.57	11	344
All schools	16.32	0.16	16.01	16.62	17	2 160

Notes: Data are weighted.

Source: Foundation Phase Systemic Evaluation 2007- a representative survey of grade 3 learners.

Table 7. Bias-adjusted effect sizes for the poorest 75% of schools, R_max and delta – Oster method.

Student FE model	N	Uncontrolled effect (se) [R^2]	Controlled effect (se) [R^2]	Bias adjusted β (δ = 1, Rmax = 1.3*R)	Bias adjusted β (δ = 1, Rmax = 2*R)	Bias adjusted β (δ = 1, Rmax = 3*R)	Bias adjusted β (δ = 1, Rmax = 4*R)	Rmax for β < 0	Value of δ if β = 0, Rmax = 1
2	19114	−0.078 (0.018) [0.022]	−0.065 (0.018) [0.031]	−0.050	−0.009	0.065	0.169	0.060	0.037
3	19114	−0.078 (0.018) [0.022]	−0.103 (0.023) [0.042]	−0.131	−0.433	0.108^a	0.041^a	0.360	0.330
4	17446	−0.107 (0.025) [0.031]	−0.101 (0.027) [0.050]	−0.094	−0.075	−0.034	0.045	0.180	0.133

Notes: See Table 3 for corresponding student fixed effect model. Uncontrolled and controlled effect both include the subject dummies as controls. Delta calculations where β = 0 are different across the Oster and AET methods. The bias-adjusted estimates are calculated using the psacalc command in Stata where ‘xtreg, fe’ is used before the psacalc command. ^aAn alternative solution that minimises the squared difference from the controlled estimate that assumes a change in the direction of the bias.

Source: SEACMEQ III, own estimates.