A Decade of Centralized School Choice Admission in Chile: Achievements & Challenges

Abstract

Chile's Sistema de Admisión Escolar (SAE) replaced a fragmented, often discretionary admissions process with a transparent, strategy-proof deferred acceptance mechanism that spans both public and subsidized private schools. Using nationwide administrative records, household surveys, and multiple randomized and quasi-experimental evaluations, the project traces how ten years of SAE affected match quality, access to higher value-added schools, and perceptions of fairness, with especially large gains for disadvantaged students. At the same time, the evidence shows that behavioral frictions—short and risky application lists, limited awareness of nearby options, and biased beliefs about quality and price—still constrain outcomes. Information tools such as real-time risk warnings, personalized recommendation reports, and the MIME explorer meaningfully improve applications and reduce non-assignment risks but also create congestion and spillovers that must be managed in general equilibrium. The paper concludes by analyzing the new Anótate en la Lista aftermarket platform and distilling governance lessons for algorithmic transparency and incremental mechanism improvements.

Citation & BibTeX

Christopher A. Neilson, "A Decade of Centralized School Choice Admission in Chile: Achievements & Challenges", Work in progress, 2025.

Project summary

Chile’s Sistema de Admisión Escolar (SAE) replaced a fragmented, often discretionary admissions process with a transparent, strategy-proof deferred acceptance mechanism that spans both public and subsidized private schools. Using nationwide administrative records, household surveys, and multiple randomized and quasi-experimental evaluations, the project traces how ten years of SAE affected match quality, access to higher value-added schools, and perceptions of fairness, with especially large gains for disadvantaged students. At the same time, the evidence shows that behavioral frictions—short and risky application lists, limited awareness of nearby options, and biased beliefs about quality and price—still constrain outcomes. Information tools such as real-time risk warnings, personalized recommendation reports, and the MIME explorer meaningfully improve applications and reduce non-assignment risks but also create congestion and spillovers that must be managed in general equilibrium. The paper concludes by analyzing the new Anótate en la Lista aftermarket platform and distilling governance lessons for algorithmic transparency and incremental mechanism improvements.

Extended abstract

Chile’s Sistema de Admisión Escolar (SAE) is one of the world’s most ambitious national-scale deployments of a centralized, strategy-proof school admissions platform. Introduced in 2016 and rolled out nationwide by 2020, the system replaced a decentralized, discretionary process with a unified application and assignment mechanism based on Deferred Acceptance, layered with legal priorities (including dynamic sibling rules) and full public transparency. This paper synthesizes a decade of design, implementation, experimentation, and operational learning with three aims: to document what centralized digital admission can achieve, to diagnose persistent frictions that limit its performance, and to outline pragmatic policy adjustments—especially in governance and the aftermarket—that can sustain legitimacy and improve welfare.

The paper first traces the staggered rollout and design evolution. Early pilots allowed the government to iterate on institutional features that were novel in their operationalization, notably real-time monitoring dashboards, transparent priority hierarchies, and sibling priorities that adapt to multi-child applications. Despite strong theoretical underpinnings, initial cohorts revealed stubborn behavioral barriers. Families often overestimated admission probabilities at selective schools, submitted short preference lists, and under-explored nearby high-quality options. These frictions are consequential in a system where safety relies on adequate list depth and where distance both shapes priorities and affects educational access.

To address these barriers, the platform progressively integrated an “informational layer” of tools that support better decisions while preserving strategy-proofness: (i) non-assignment risk warnings triggered by application profiles and local capacity-demand conditions; (ii) personalized school suggestions and a school explorer interface that surface proximate, program-appropriate alternatives; and (iii) targeted communications that nudge earlier and safer search behavior. The paper reviews evidence from randomized trials and regression discontinuity designs that evaluated these tools at both the individual and market level. Risk alerts increased list length and diversification, reduced the probability of non-assignment, and raised access to higher-quality placements without detectable negative spillovers on untreated students under typical market conditions. Recommendation tools shifted choices toward feasible, better matches that families had overlooked, particularly for students with special needs or those eligible for technical-professional tracks. These effects were strongest for applicants with the largest initial belief errors, indicating that platform guidance can partially correct biased priors and high search costs.

The analysis also examines medium-run effects and system-wide externalities. While the immediate gains include fewer non-placements and improved first-round outcomes, the long-run impacts on test scores are modest and heterogeneous, reflecting the complex mapping from initial assignment to learning within a dynamic schooling system. Importantly, the interventions’ congestion effects remain limited when designed with calibrated thresholds and tiered guidance, though the paper cautions that market-sensitive tuning is essential to avoid compression at the same margins. Complementary descriptive evidence shows that the digital centralization reduced administrative burden and transaction costs and increased transparency, even as it shifted capacity management and public expectations in new ways.

A central finding is that the core algorithm and priority hierarchy have changed little since national adoption, despite a decade of learning about preferences, mobility, and heterogeneity. The paper argues for institutionalizing “algorithmic learning” to match the pace of informational innovation. Concretely, it recommends a governance stance that (i) publishes accessible documentation of assignment logic and priorities; (ii) establishes a participatory review process—including civil society and independent experts—for evaluating variants (e.g., distance weights, tie-breaking, sibling rules); and (iii) pilots scoped algorithmic adjustments with ex-ante criteria and ex-post, public evaluation. Such governance not only aligns with transparency and legitimacy but also enables evidence-based updates when conditions change.

The paper highlights several design lessons with immediate operational relevance. First, information must arrive earlier and be tailored: outreach should connect to families in pre-primary systems and those already active in the aftermarket, foregrounding local options and programs for students with permanent special needs. Second, families with multiple children need tools that make joint outcomes salient. The platform should expose assignment “tuples,” present probabilities for joint versus separate applications, and guide when to couple or decouple siblings given local market conditions—drawing on stable, well-understood adjustments from matching markets with couples. Third, risk warnings should be offered more broadly and calibrated to exposure in tiers, with explicit alerts for interactions between features (e.g., sibling coupling and origin-school fallback) that can unintentionally raise non-assignment risk. Fourth, because distance shapes both priorities and recommendations, registration flows should verify addresses with a correction step and provide a documented pathway for pending moves; better geodata improves recommendations and reduces downstream errors. Fifth, platform controls should respect expressed preferences by allowing applicants to disable features that reduce expected welfare in their circumstances—especially origin-school fallback and forced sibling coupling when an older sibling anchors at the origin. Finally, targeted, capped capacity adjustments can be used sparingly where scarcity is acute to reduce non-assignments and increase top-rank placements without compromising transparency. A concrete, rule-based adjustment to co-assign same-grade siblings (“twins”) just after the main match can deliver meaningful welfare gains with negligible spillovers.

A persistent challenge arises in the unregulated or loosely regulated “aftermarket,” where families frequently seek transfers outside the formal match. Chile’s formalization of the aftermarket through Anótate en la Lista is a crucial step, reframing admissions as a continuous process rather than a single event. The paper argues for strengthening rules and monitoring in this phase to align incentives, sustain transparency, and limit churn that can erode both equity and public trust.

Together, the evidence shows that centralized digital assignment can improve fairness, predictability, and access—especially when complemented by behavioral tools that help families make strategy-proof choices that are also well informed. The combination of real-time monitoring, calibrated warnings, and personalized recommendations demonstrably improves application quality and safety, with manageable system-wide externalities. Yet to consolidate these gains for the next decade, Chile must couple its informational innovation with structured algorithmic learning and robust aftermarket governance. Publishing the algorithm’s logic, creating participatory channels for proposing and piloting scoped variants, and evaluating them transparently can sustain legitimacy as demographics and preferences evolve.

The paper contributes a decade-long, policy-embedded perspective on how to make centralized school choice work in practice: preserving the theoretical virtues of strategy-proofness and stability while embracing iterative, evidence-based improvements in the tools that shape behavior, the data that inform priorities, and the governance that maintains public confidence. Limitations include reliance on administrative outcomes that imperfectly capture learning and welfare, and the need for more systematic evaluation of algorithmic variants. Future work should deepen market-level assessments, expand participatory simulators to democratize understanding of trade-offs, and integrate continuous feedback loops between research and policy. Chile’s experience offers a reference model: centralized algorithms are necessary but not sufficient; lasting success hinges on adaptive design, transparent governance, and sustained engagement with the citizens whom the system serves.

  • Coauthors:
  • Published: Work in progress
  • Date: 2025-11-16
  • PDF