School choice, institutional environments and the precarisation of teachers’ working conditions in not-for-profit charter schools

Figures & data

Figure 1. Charter schools under analysis in Bogota.

Table 1. Complete data set per type of school.

Year	Charter	Private	Public	Total
2005	22	1,406	322	1,750
2006	22	1,471	323	1,816
2007	22	1,523	330	1,875
2008	22	1,528	347	1,897
2009	22	1,578	350	1,950
2010	22	1,613	335	1,971
2011	22	1,618	344	1,984
2012	22	1,611	333	1,966
2013	22	1,636	332	1,991
N	198	13,986	3,016	17,200

Table 2. Summary of statistics per schools observed.

Variables	Obs.	Mean	Std. dev.
Log of mean teacher’s salary per school	17,200	13.59	.56
Total number of teachers	45,536	26.29	31.74
Average teaching workload per school (classes per teacher)	17,200	54.01	64.17
Average teaching workload per teacher (ratio)	17,200	2.43	3.54
School’s socioeconomic status	17,200	2.61	1.13

Data source: DANE

Figure 2. Timeline of charter schools’ introduction and other labour reforms in Colombia.

Graph 1. Yearly average evolution of teacher’s monthly average salary per type of school in pounds.

Graph 2. Yearly average evolution of general workload (teaching units) per school and type of school.

Graph 3. Yearly average evolution of individual teacher’s ratio of workload (teaching units) per school and type of school.

Table 3. Difference-in-Differences outcomes with different estimation techniques for Y = log of teacher’s salary.

Dep: Log of mean teacher’s salary per school	Model 1OLS	Model 2Simple Diff-in-Diff	Model 3Diff-in-Diff with covariates	Model 4PSM-Kernel Diff-Diff with covariates
Total	.57** (.036)	0.085** (0.04)	0.096** (0.040)	0.038 (0.029)
Obs.	3,729	3741	3729	2925
R-sq	0.28**	.17**	0.29**	.20**
Male	.48** (.05)	0.10* (0.05)	0.041 (0.05)	0.067* (0.036)
Obs.	2,495	2,495	2495	1955
R-sq.	0.19**	0.14**	.20**	.19**
Female	.593** (.033)	0.106** (0.04)	0.128** (0.038)	0.060** (0.029)
Obs.	3,675	3704	3675	3675
R-sq.	.31**	.18**	.31**	.20**

** p<.05, * p<.1

Note: Our dependent variable is logged, and thus a point increase in vector X multiplies the expected value of Y by $e^{\beta }$. For more details please see Benoit and Van den Poel (2017). This can be interpreted as increasing percentages for small values of β, following eβ ≈ 1 + β${\mathrm{e}}^{\mathrm{\beta }}\approx 1+\mathrm{\beta }$. This is valid for all models.

Table 4. Difference-in-Differences outcomes with different estimation techniques for LW = ratio of teaching workload.

Average workload per teacher (ratio)	Model 5 OLS	Model 6 Simple Diff-in-Diff	Model 7 Diff-in-Diff with covariates	Model 8PSM-Kernel Diff-Diff with covariates
Total	-.67 (.45)	0.73** (.32)	−0.39 (1.1)	0.3 (.43)
Obs.	3,729	3,741	3,729	2,927
R-sq.	0.18**	.01**	.18**	.01**
Male	.543* (.28)	1.482** (0.33)	0.606** (0.28)	0.984** (0.4)
Obs.	2,495	2,514	2,495	1,965
R-sq.	.31**	.005**	.32**	.07**
Female	-.43 (.39)	.85** (.34)	−0.129 (0.3)	0.72** (0.36)
Obs.	3,675	3,704	3,675	2,289
R-sq.	.20**	.01**	.02**	.01**

** p<.05, * p<.1

Table 5. Balancing test: DID with covariates vs Propensity Score Matches with Kernel DID – model 4 vs model 3.

Variables	DID				PSM-DID
Coefficient difference	Total	Fem.	Male		Total	Fem.	Male
Log of mean teacher’s salary per school	0.038	0.025**	0.029	0.020		13.39	−0.008
Total number of teachers	8.915**	6.091**	1.465*	4.04		23.4	1.244
Average teaching workload per school	15.0**	10.1**	2.417	7.00		41.1	2.734
Average teaching workload per teacher (ratio)	−0.01	0.008	0.101	−0.008		1.98	0.121
Socioeconomic status	−0.709**	−0.708**	−0.631**	−0.133**		2.19**	−0.158**

** p<.05, * p<.1, Ho: treatment and control groups do not differ significantly in following covariates at the base level.

Table 6. Balancing test: DID with covariates vs Propensity Score Matches with Kernel DID – model 8 vs model 7.

Variables	DID			PSM-DID
Coefficient difference	Total	Fem.	Male	Total	Fem.	Male
Log of mean teacher’s salary per school	0.038	0.025	0.029	0.002	0.016	0.005
Total number of teachers	8.915**	6.091**	1.465*	3.732	2.637	1.375
Average teaching workload per school	15.0**	10.129**	2.41	6.521	4.380	2.560
Average teaching workload per teacher (ratio)	−0.01	0.008	0.10	−0.015	−0.023	−0.011
Socioeconomic status	−0.709**	−0.708**	−0.63**	−0.140**	−0.136**	−0.160**

** p<.05, * p<.1, Ho: treatment and control groups do not differ significantly in following covariates at the base level.