School reorganization reforms: the case of multi-site schools in Colombia

ABSTRACT

To contribute to the debate about the reorganization of public-school supply, we analyze a reform introduced in Colombia in 2001 that merged independent schools into a single institution. Specifically, we estimate differences in achievement, dropout rates, teacher characteristics, and technological infrastructure between sites belonging to multi-site schools and single-site schools. Our results suggest that there are no differences in test scores, but lower dropout rates in multi-site schools. Also, we found that sites belonging to multi-site schools, especially in larger networks (more than six sites) and far away from the main site of the school, have younger teachers with lower salaries who are more likely to have a temporary contract. Finally, we provide suggestive evidence that more isolated sites from large-sized schools have less access to the internet and a computer room than single-site schools. These results provide evidence for reforms to increase cooperation between sites.

KEYWORDS:

• School reorganization
• Colombia
• educational achievement

Acknowledgements

We thank the Ministry of Education in Colombia for providing the data. We are also thankful to the Inter-American Development Bank (IDB) for funding the research presented in this paper. We would also like to thank participants of the IDB-Education BBL for providing valuable feedback.

Disclosure statement

The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. The authors have no conflicts of interests or financial and material interests in the results. All errors remain our own.

Notes on contributors

Gregory M. Elacqua is currently the Principal Economist in the Education Division in the Social Sector Department at the Inter-American Development Bank in Washington DC. He was previously the Director of the Public Policy Institute at the School of Business and Economics at the Universidad Diego Portales (UDP) in Chile. His research focuses on education policy, the economics of education, school accountability, teacher policy, school choice, and the political economy of the educational system. He has been advisor to three Ministers of Education in Chile. He has also served as an advisor to a member of the Education Committee in the Chilean Senate. Elacqua has been involved in the design of the education reforms signed into law in recent years and currently under congressional discussion in different countries in Latin America. He has also consulted with UNESCO, the World Bank, and several foundations and governments on education policy. Elacqua was also a member of the PISA 2015 Questionnaire Expert Group. He holds a PhD in Public Policy from Princeton University.

Fabio Sánchez is a Full Professor at the School of Economics in Universidad de los Andes and was the Director of the Research Center of the same school. Fabio holds a Master and PhD in Economics from Rutgers University. His research concentrates on origins and history of the Colombia armed conflict and its consequences on social development, impact evaluation of educational programs, and decentralization and efficiency of local governments. He has directed more than 40 projects funded by ESRC/DFID, IADB, IDRC, London School of Economics, Tinker Foundation, Spencer Foundation, Colciencias, Ford Foundation, Global Development Network, USAID, World Bank, and several institutions from the Colombian Government. He also was a Tinker Visiting Professor in the Institute of Latin American Studies at Colombia University.

Humberto A. Santos is currently a consultant in the Education Division in the Social Sector Department at the Inter-American Development Bank in Washington DC. He was previously a researcher in the Public Policy Institute at University Diego Portales, research associate of the Center for Comparative Politics of Education at the same university, and researcher in the Studies Department at the Chilean Ministry of Planning (MIDEPLAN). Humberto received his master’s degree in economics from University of Chile. His current research focuses on analysis of local education markets, regulations and accountability in the educational system, and school finance in education.

Notes

1 For more information about microcentros, see https://rural.mineduc.cl/el-microcentro/

2 Giordano (2008) introduced a typology of different models, based on the relations between the schools in the network and the way in which the clusters are created, distinguishing between bottom-up and top-down, voluntary versus mandatory, selective versus universal, high- versus low-intensity cooperation approaches, financially autonomous versus financially supported, and between clusters that serve primarily as tools for external control versus those that promote internal development.

3 For example, Law 115 of 1994 (General Education Law) states that schools must have the infrastructure and support of the pedagogical activity to offer at least preschool and basic education in the same school. It also indicates that if they do not offer the complete educational cycle, they should establish agreements with other schools, to guarantee the continuity of their students. For details about former attempts to reorganize the public supply, see Corpeducación (2014).

4 In Colombia, the school system is divided into three levels: básica primaria (basic primary; Grades 1–6), básica secundaria (basic secondary; Grades 6–9), and educación media (secondary; Grades 10-11). The equivalent in the International Standard Classification of Education (ISCED) classification is primary, lower secondary, and upper secondary, respectively.

5 In order to simplify the exposition of the policy and the presentation of the results, we use the word “network” or “cluster” to refer to the schools made up of various schools that were previously independent schools. However, it is important to clarify that in the case of Colombia, unlike experiences in other countries, the formation of these networks is not voluntary but established by the territorial entities that administer the schools.

6 In Colombia, education is managed by Certified Territorial Entities (Entidades Territoriales Certificadas or ETC) that can be Departments (Departamentos or subnational level) or Municipalities (Municipios or local level) (Law 715 of 2001). Departments and districts were certified automatically. The education in municipalities that are not certified is managed by Departments (Brutti, 2016).

7 Unlike public schools, almost all private schools operate as a single schooling unit.

8 In Colombia, the Sistema General de Participaciones (General Participations System; SGP) are the resources transferred from the national government to Certified Territorial Entities and the rules governing them. SGP quality free-of-charge is a central government transfer, created in 2008 to replace the resources that public schools charged to parents. These resources are transferred directly to schools to invest in all spending categories, except staff. School principals decide how to allocate these resources, which represented 2% of total public spending in education in 2016. Also, schools can receive money coming from the ETC’s own resources. In 2016, these resources represented 9% of total spending in education (Álvarez et al., 2018). Finally, schools can receive money from transfers and programs funded by the Ministry of Education and implemented by territorial entities, but there is no information about the criteria used to allocate those resources.

9 From 2002, the teachers in the public sector are hired through a competitive public contest. Contests are organized separately in each ETC, and candidates must choose the one education authority they wish to apply to. The contest is based on a scoring system that establishes a ranking among applicants, which determines the order in which successful candidates will be allowed to choose their preferred vacancies (Brutti & Sánchez, 2016).

10 In practice, when there are no applicants or winners in the teaching hiring contest, ETCs must fill teacher positions with provisional appointments, often with teachers that do not pass the hiring contest exam. Recent studies show that temporary teachers are less effective than permanent ones (Ayala & Sánchez, 2017).

11 SABER is the standardized test applied in third, fifth, ninth, and 11th grades. The test evaluates all schools in the system, including public and private, both in urban and rural areas. It focuses on the basic competences that students must develop in the areas of Language and Mathematics. The Colombian Institute for the Evaluation of Education (ICFES) is the institution responsible for the organization and implementation of the tests. For more information, see http://www2.icfes.gov.co/instituciones-educativas-y-secretarias/acerca-de-las-evaluaciones/informacion-general

12 In the results section, we only present the estimations using the average monthly basic salary as a variable to characterize teachers in each site. But we also estimate the models using the distribution of teachers by age and type of contract. The results of these exercises are consistent with those presented below, and indicate that the sites in multi-site institutions, especially those farthest from the main site, have younger teachers who are more likely to have temporary contracts, compared with single-site schools. These results are available upon request.

13 Dependent variables and the full set of controls in X_sij are described in Table 4.

14 Angrist and Pischke (2008) show that regression can be motivated as a computational device for a sort of weighted matching estimator, and therefore the differences between regression and matching are unlikely to be of major empirical importance.

15 For more information, see https://www.dane.gov.co/index.php/estadisticas-por-tema/educacion/poblacion-escolarizada/educacion-formal

16 In Appendix 1, we present a table comparing the sample of sites used in the estimation for SABER 5 Language (Restricted) with the unrestricted sample. As expected, school sites in the restricted sample are more likely to belong to multi-site schools, are larger, are less likely to have only one teacher, have a higher percentage of students in traditional teaching models, and are more likely to function in a double-shift.

17 For fifth grade, 38% of school sites that have information on test scores do not have socioeconomic information on students. On the other hand, for ninth grade, only 18% lack this information. In Appendix 1, we present descriptive statistics comparing the sample used for the SABER 5 estimation (Table 5) with and without controlling for SES. The results show that there are no significant differences between both samples.

18 Although to the best of our knowledge, there is no specific restriction on the maximum distance to which a site can be from the main site, we consider 100 km the threshold to define that school site was not well georeferenced, because it is hard to assume that two sites located at such a distance likely belong to the same school. We tested a small sample of schools to check the location of the school (based on websites, maps of the municipality, and other sources) and compared it with the location associated with the coordinates. We found that a high percentage of these schools were not well georeferenced. We used the same argument for the distance to the main building of the municipality. Only 8% of school sites are discarded when these restrictions are imposed.

19 We calculate “distance as the crow flies”, that is, the shortest route between two points, without considering the geographical features or the available routes.

20 Given the existence of outliers in the distance to the main site, probably due to errors in georeferencing, the median distance is much shorter than the mean. The median distance to the municipality for independent sites is 5.8 km, 5.7 km for sites from medium-sized schools, and 8.7 km for sites from large-sized schools. On the other hand, the median distance to the main site for medium-sized schools is 1.7 km and 5.1 km for sites from large-sized schools.

21 The opposite of traditional is a flexible education model, which has specific educational materials and defined training processes. These models are designed to serve diverse populations, in situations of vulnerability, and who have difficulty accessing the regular model (e.g., Escuela Nueva [New School], Post-Primaria [Post-Primary], Telesecundaria [Virtual Secondary]). More information about these alternative models can be found in https://www.mineducacion.gov.co/1759/w3-propertyvalue-55270.html

22 In Colombia, there are two different pay scales for teachers in the public sector. Teachers that entered in the teacher career before 2002 are governed by Decree 2277 of 1979 and those hired after that year by Decree 1278 of 2002. The current pay scale is divided in three different grades, defined by the level of education of teachers. In each grade, there are four levels through which teachers increase as they gain experience and pass a voluntary competency assessments. There is a base monthly salary in each combination of grade and level. The old pay scale consists of 14 levels, each one having a higher monthly base salary. The education level of the teacher determines at what level they enter. Increases on the scale are determined by experience; every 3 years teachers go to the next level of the scale, which entails a salary increase of approximately 12% (Ome, 2013).

23 Tables 5–13 only present the coefficients of the type of school classified according to the number of sites and the interactions with the distance from the main site of the school and with the educational levels offered at the educational institution. The complete tables, with the full set of control variables, are available upon request.

24 The results in this section do not change substantively if the regressions are estimated only with the non-main sites of the multi-site schools and when we restrict the sample only to sites located outside of the capitals of the departments (32 municipalities) and Bogotá. They also do not change when we used propensity score matching to estimate the average treatment effect (ATT) of belonging to a multi-site school, using the same control variables in Tables 5–13 to estimate the propensity score and imposing a common support by dropping treatment observations whose pscore is higher than the maximum or less than the minimum pscore of the controls. These results are available upon request.