Adjusting a Matching Analysis for Many Covariates in an Educational Evaluation. DRAFT

As interest in data, data-gathering techniques, and storage capacity have increased, so has the availability of data for social researchers. In particular, The No Child Left Behind Act of 2001 mandated the collection and recording of large amounts of school-level data. In an observational evaluation of an educational intervention, these newly-available data may be quite valuable. Observational studies are susceptible to bias from confounders; more data, on more potential confounding variables, may help reduce bias in causal estimates. However, as the number of variables in a statistical model grows relative to the number of