WORMS: IDENTIFYING IMPACTS ON EDUCATION AND HEALTH IN THE PRESENCE OF TREATMENT EXTERNALITIES, BY EDWARD MIGUEL AND MICHAEL KREMER Flashcards
(7 cards)
🧠 Flashcard 1: What core problem does the study address, and how does it innovate methodologically?
A:
Miguel and Kremer address the underestimation of the benefits of health interventions—specifically, school-based deworming—by highlighting treatment externalities. Prior studies randomized treatment at the individual level, ignoring the spillover effects on untreated peers. This study innovates by randomizing at the school level, allowing for the identification of both direct effects and externalities (within- and cross-school). This approach reveals that individual-level RCTs may systematically understate the returns to treatment when externalities are present.
💊 Flashcard 2: What were the main health and education outcomes of deworming treatment?
A:
The program significantly reduced moderate-to-heavy worm infections, dropping prevalence by up to 25 percentage points in treatment schools within a year. These reductions extended to untreated children in both treated and nearby untreated schools, evidencing strong positive externalities.
Educationally, school participation increased by 7.5 percentage points, equivalent to a 25% reduction in absenteeism. This effect was particularly pronounced among younger children, who also bore the heaviest infection burden. However, no significant improvement in test scores was detected, likely because the attendance gains were not large enough or sustained enough to translate into measurable cognitive improvements within the study window.
💸 Flashcard 3: How cost-effective was school-based deworming compared to other interventions?
A:
Deworming was found to be extraordinarily cost-effective, at only $3.50 per additional year of school participation, far cheaper than interventions like school uniforms or subsidies (e.g., $99/year). When factoring in health benefits, especially reductions in schistosomiasis, the program cost about $5 per Disability-Adjusted Life Year (DALY) averted, placing it among the most cost-effective health interventions for low-income countries.
Moreover, the majority of benefits came from externalities, meaning that individual incentives alone would not lead to socially optimal uptake, justifying public subsidies or even paying individuals to take the medicine.
🧪 Flashcard 4: What data and econometric strategies did the authors use to identify effects and externalities?
A:
The authors combined a cluster-randomized trial (75 Kenyan schools in phased deworming over 3 years) with detailed data:
Parasitological surveys (worm egg counts), School attendance records from unannounced NGO visits, Test scores, and Spatial GPS data to calculate proximity-based treatment density.
For analysis, they used intent-to-treat (ITT) estimation and models capturing how the number of treated pupils nearby (within 0–3km and 3–6km) affected outcomes in untreated schools. They also attempted to decompose within-school spillovers, though non-experimental methods had to be used here. Importantly, treatment effects were adjusted for spatial correlation and measurement error in treatment exposure.
🧭 Flashcard 5: What broader implications does the study offer for development economics and public health policy?
A:
The study underscores the need to rethink evaluation strategies in settings where treatment generates externalities. It challenges reliance on individual-level RCTs for infectious diseases and supports cluster-level randomization and spatial analysis.
Why It Matters:
Individual-level RCTs can underestimate or misrepresent treatment effectiveness in contexts with externalities.
Cluster randomization + spatial analysis give a more accurate picture in such settings by accounting for transmission dynamics and indirect effects.
Policy-wise, it reveals that tropical disease burdens like worms have sizable, under appreciated educational and economic consequences. By establishing a causal link from health to education, it strengthens the case for integrated health-education interventions. It also points to the importance of externalities in justifying public investment, offering a compelling economic rationale for universal free deworming and other public goods with spillover benefits.
🧩 Flashcard 6: In what ways is the study critical of past research, and how does it reflect critically on its own limitations?
A:
Miguel and Kremer critique earlier deworming studies for two major methodological flaws:
Randomization at the individual level, which fails to account for treatment externalities—leading to systematic underestimation of treatment effects, since untreated peers may benefit from reduced transmission.
Focus on cognitive test scores without examining broader and more immediate educational outcomes like school attendance, which may be more responsive to health interventions.
They argue that such studies, including those cited in the Cochrane review (Dickson et al., 2000), use designs that bias results toward the null, especially when health improvements among untreated children dilute observed differences.
Critically, the authors also reflect on their own limitations:
They cannot fully separate direct and within-school externality effects due to school-level (not pupil-level) randomization.
Test score improvements were not observed, which they attribute either to the insufficient duration or intensity of participation gains or the lack of improvement in anemia, a key mechanism linking deworming to cognition.
They acknowledge the difficulty of generalizing to settings with different worm burdens or transmission dynamics, and they caution that externality effects may not scale linearly.
📊 Flashcard 7: How did the study use data and spatial methods to identify cross-school externalities?
A:
Miguel and Kremer used a novel spatial econometric approach to identify cross-school externalities by leveraging random variation in local treatment intensity. After randomizing deworming across 75 schools in a phased rollout, they geocoded each school using GPS and computed the number of treated students within specific distance bands (0–3 km and 3–6 km) around each school.
They then used this local treatment density as an exogenous variable—since it was determined by random assignment—to measure spillover effects on untreated schools. This allowed them to estimate how infection rates and school participation changed not only within treatment schools but also in neighboring untreated schools.
This design circumvented identification bias from self-selection and addressed spatial correlation in disease transmission. It also provided lower-bound estimates of spillovers, since effects likely extended beyond 6 km, which they could not measure precisely due to study area size constraints.
