Treatment effects and the measurement of skills in a prototypical home visiting program

This paper evaluates the causal impacts of an early childhood home visiting program for which treatment is randomly assigned. We estimate multivariate latent skill profiles for individual children and compare treatments and controls. We identify average treatment effects of skills on performance in a variety of tasks. The program substantially improves child language and cognitive, fine motor, and social-emotional skills development. Impacts are especially strong in the most disadvantaged communities. We go beyond reporting treatment effects as unweighted sums of item scores. Instead, we examine how the program affects the latent skills generating item scores and how the program affects the mapping between skills and item scores. We find that enhancements in latent skills explain at least 90% of conventional unweighted treatment effects on language and cognitive tasks. The program enhances some components of the function mapping latent skills into item scores. This can be interpreted as a measure of enhanced productivity in using given bundles of skills to perform tasks. This source explains at most 10% of the average estimated treatment effects.