An agent-based model of school choice with information asymmetries

ABSTRACT

Going from a neighbourhood-based to a choice-based system of school selection can have positive effects on enrolment in higher achievement schools, increasing average student achievement. We develop an Agent-Based model (ABM) that simulates students’ decisions on a heterogeneous agents’ framework with information asymmetries between income levels, allowing to simulate school choice policies and determine their impact on school enrolment and average student achievement. We use data from Santiago schools to initialise the model and study the impact of a discrete information signal of school achievement, as a policy implemented in 2010 in Chile called traffic lights.

KEYWORDS:

• Agent-based model
• school choice
• public policy
• heterogeneous agents

Acknowledgments

We are thankful for useful discussions with Juan Pablo Couyoumdjian, Daniel Lerner, Constanza Gomez, Victor Martinez and Zimu Wan. We are also grateful to Andrés Marticorena for excellent research assistance. Any error is the authors’ fault alone.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.

Notes

1. The simulated environment has no boundaries to avoid differences in behaviour from the agents located near the edges. This way an agent located near the edge can cross through the edge to the other side of the simulated environment. A boundless setup is different from Maroulis et al. (2014), which represents the Chicago geography in a grid of dimensions 18 × 18.

2. In Maroulis et al.’s (2014) model, the numbers of students and schools are fixed; we allow user input for the values of these variables.

3. We set the number of periods that a student has to attend school to 10. We can think of the total period as pre, primary and middle school combined. The length of the period can be adjusted depending on location or the type of education studied in the model settings.

4. Román and Perticará (2011) and Contreras et al. (2010) argue that low performing schools have not been forced out of the market by increased competition in the Chilean educational market.

5. At setup, to start the model, most students are not part of the new cohort, but are created with a random value for the attribute years-in-school, drawn from a uniform distribution of natural numbers from 1 to 9, some will graduate in the first period, and some will be in school for 9 more years, but none of them will have 10 years remaining as the new cohort.

6. The model is like the Chilean education system, in which there are private schools that charge tuition and private-voucher schools that do not charge tuition. Private-voucher schools are privately managed but publicly funded. In the model, all public schools behave the same way, so they represent both public schools and private-voucher schools.

7. The difference between high and low-income agents besides income, is that those that are high-income have perfect information about school achievement and those who are low-income have no information (unless the traffic lights is in effect). To give all agents the ability to choose with perfect information, one can simply set the parameter “percent of agents as high-income” to 100%. The share of high-income students can also be understood as the percentage of choosers, as in Maroulis et al. (2014).

8. The model and all data used in this paper is available for download as supplemental material.

9. The value of 1768 schools in Santiago comes from the Mineduc (2018) database considering a balanced panel. It only includes schools that have complete enrolment data between 2006 and 2016.

10. By truncating the distribution in zero we are saying that zero is the lowest achievement possible a school might have, and by the way the agents’ utility function is defined, such a school would give the minimum possible utility.

11. The data in Figure 3 does not consider the share of students that do not attend school. If it did, the share of students that attend low-achievement schools would be slightly smaller.

12. The concept of priority student was created in 2008 by the government of Chile with law number 20,248 of Subvención escolar preferencial (priority school subsidy).

Additional information

Funding

This work was supported by Universidad del Desarrollo with funding from Concurso Interfacultades.