"Everything in Moderation: Using Proximal and Distal Measures to Forecast the Long-term Impacts of Math Interventions"
Poster Title: Everything in Moderation: Using Proximal and Distal Measures to Forecast the Long-term Impacts of Math Interventions
Presenter: Daniela Alvarez-Vargas
Poster Advisor: Drew Bailey
Research Specialization: Human Development in Context
Interventionists often justify short-term intervention targets on the basis of their potential for long-term effects. Past attempts have overestimated or underestimated these outcomes. In the present study we use data from a randomized control trial of a first-grade math intervention. We show how omitted variable bias and over-alignment bias from the use of measures proximal to the intervention contribute to the over-estimation of long-term treatment impacts, while under-alignment bias from the use of measures distal to the intervention contributes to the under-estimation of long-term treatment impacts. We identify some promising and some very biased methods for forecasting treatment impacts. In particular, we find that using proximal measures with small impacts and distal measures with large impacts may yield realistic forecasts.