Do Modern Preschool Programs Suffer from Declining Effectiveness?

This article reviews evidence from recent randomized controlled trials and lottery studies, comparing them with older, small-scale interventions. Across studies, modern preschool programs produce much smaller short-term gains (about 0.21 SD vs. 0.45 SD in earlier programs) and show faster fadeout, with impacts shrinking by 62% within a year.

The Policy Scientist's Perspective

This article addresses an important policy question: why the measured impacts of preschool have weakened. The topic is broadly significant because policymakers continue to rely on evidence from earlier, small-scale demonstrations when justifying large public expenditures. The synthesis draws on credible data, including RCTs and lottery designs, which strengthens causal claims and enhances generalizability. Although largely a review, it advances the literature by integrating recent high-quality studies into a coherent, timely explanation of declining effects.

Full Citation and Link to Article

Whitaker, A. A., Burchinal, M., Jenkins, J. M., Bailey, D. H., Watts, T. W., Duncan, G. J., Hart, E. R., & Peisner-Feinberg, E. (2025). Why are preschool programs becoming less effective? Journal of Policy Analysis and Management, 1–11. https://doi.org/10.1002/pam.70031

Central Research Question

The article investigates why modern large-scale preschool programs generate substantially weaker and, in some cases, negative long-term effects compared with the classic early childhood interventions of the mid-20th century. The authors seek to determine whether these diminished effects stem from changes in children’s counterfactual environments, shifts in preschool instructional practices, variability in program quality, or differences in scale and implementation. The central question is therefore explanatory: what mechanisms account for the clear decline in measured program impacts over time, despite expanded public investment and stronger policy interest in preschool as a tool for reducing inequality?

Previous Literature

Early landmark interventions—most notably Perry Preschool and the Abecedarian Project—established the foundation of the field. Conducted as small-scale randomized trials, these programs yielded large short-term skill gains and substantial long-term improvements in education, earnings, and health. Their reported effect sizes (often around 0.45 SD at preschool exit) became benchmarks for subsequent policy arguments. Later research also suggested persistence of impacts into adulthood, reinforcing the belief that early childhood programs could generate high social returns.

However, more recent rigorous studies reveal a sharply different pattern. Contemporary evaluations using randomized controlled trials or lottery-based admissions show smaller initial effects and faster fadeout. The Head Start Impact Study, for example, reported modest early gains that disappeared by third grade. Evaluations in Boston identified positive long-term outcomes in high school completion and college-going but little evidence of sustained academic skill advantages. Tennessee’s statewide pre-K RCT produced the starkest divergence: positive entry-level academic gains that not only faded but turned negative through elementary and middle school. Lottery studies from Georgia show a similar trajectory. Meta-analytic work further documents a clear temporal pattern: programs launched after 2000 yield substantially smaller average effects than those launched between 1960 and 1999.

The literature also debates the contributing factors. Some scholars emphasize deteriorating program quality during scale-up, though evidence is mixed. Others invoke the sustaining environments hypothesis, which posits that low-quality K–12 schooling erodes early gains, yet meta-analytic reviews offer limited support. A more recent line of argument highlights improvements in children’s home environments and safety-net supports, which may reduce the marginal value of preschool. The article positions itself within this broader literature by synthesizing these competing explanations and reconsidering long-held assumptions about early childhood impacts.

Data

The article draws on multiple data sources, all secondary but high quality. First, it uses findings from randomized controlled trials and lottery-based evaluations of modern preschool programs, including Head Start, Boston pre-K, Tennessee pre-K, and Georgia’s state program. These evaluations incorporate large administrative datasets, standardized assessments, behavioral indicators, and longer-term educational records.

Second, the article uses meta-analytic evidence from the MERF dataset (Meta-Analysis of Educational RCTs with Follow-up). This dataset includes 18 rigorously evaluated early childhood interventions, with harmonized effect sizes across assessment waves. The MERF data allow direct comparison of older and newer programs, controlling for research design, timing, and sample characteristics.

Third, the authors incorporate national demographic, economic, and policy data that document improvements in family environments: increased maternal education, reduced family size, expanded safety-net programs, increased childcare participation, and significant reductions in child mortality. These data come from IPUMS, federal expenditure reports, NCES, CDC mortality datasets, and federal crime statistics.

Collectively, the data sources are strong. The reliance on RCTs, lotteries, and meta-analysis enhances internal validity and causal interpretability. The national demographic and expenditure trends provide contextual breadth, supporting claims about changes in children’s counterfactual environments. The findings, while grounded in U.S. settings, are potentially applicable to other advanced economies with similar improvements in children’s early environments.

Methods

The article is not an original empirical analysis; rather, it is a conceptual and synthetic review that incorporates results from RCTs, lotteries, and meta-analyses. Its methodological contribution lies in how it integrates these sources to identify consistent patterns across time. The MERF data are particularly central because they use meta-analytic weighting (1/SE²) and random-effects modeling to estimate average effect trajectories for older and newer programs. This allows for credible comparisons of initial effect sizes and fadeout rates.

The authors also use descriptive trend analysis, examining national-level changes in socioeconomic and child welfare indicators. These descriptive comparisons are not causal, but they are used to support the argument that children’s baseline conditions have improved substantially, reducing the marginal effect of preschool.

Importantly, the article places greater weight on studies using causal inference methods—randomization or lottery-based assignment—rather than multivariate regression or observational designs. This aligns with best practices in program evaluation and increases confidence in the reported patterns. While the article does not conduct new statistical modeling, it summarizes studies that employ rigorous methods and synthesizes them in a logically structured way.

Findings/Size Effects

The authors report several key findings.

First, modern preschool programs produce much smaller initial effects. Earlier programs launched between 1960 and 1999 averaged about 0.45 SD gains at the end of treatment. Modern programs launched between 2000 and 2011 averaged roughly 0.21 SD—less than half the earlier effect size.

Second, fadeout occurs more rapidly in recent programs. Older programs retained a substantial share of their initial impact, with about 62% loss at six to twelve months. Newer programs lost nearly all early gains within the same window, averaging a decline of over 60% and approaching zero or negative effects by elementary grades.

Third, in some cases, impacts turn negative. The Tennessee RCT found that early academic advantages reversed by later grades, with attenders performing worse than non-attenders in both academic and behavioral domains. The Georgia lottery study produced a similar pattern.

Fourth, improved counterfactual environments offer the most compelling explanation. Children not attending preschool now benefit from better parental education levels, safer home environments, more generous safety-net programs, better nutrition, increased access to center-based care, and substantially lower child mortality. These improvements mean that the difference between treatment and control conditions is far smaller than in the mid-20th century.

Fifth, alternative explanations are less convincing. Variation in preschool quality is real, but quality indices do not consistently predict fadeout or negative effects. K–12 school quality does not reliably explain impact patterns; in fact, higher-quality elementary schools sometimes accelerate fadeout by helping control-group students catch up.

Sixth, changes in instructional focus may contribute. Early landmark programs emphasized language-rich interactions, scaffolding, and caregiver-child relationships. Modern programs focus more heavily on early academic instruction, particularly literacy and numeracy. The authors argue that the current emphasis may generate short-term test score gains that quickly dissipate when elementary instruction covers similar content.

Conclusion

The article concludes that declining preschool impacts reflect fundamental shifts in children’s early environments rather than simple program failure. Today’s children—especially those from low-income families—enter preschool healthier, better supported, and more engaged in early learning environments than previous generations. As a result, the marginal effect of preschool has fallen. This interpretation challenges common assumptions and underscores the need for new research designs, longer-term follow-ups, and innovative program models that provide distinctive developmental value rather than duplicating services children already receive.

The authors argue that policymakers and researchers must recalibrate expectations. Classic demonstration programs set unrealistic benchmarks for modern large-scale systems. To improve effectiveness, future research should isolate the mechanisms that produce durable long-term gains and test program models that complement, rather than mirror, the improvements already present in children’s lives.