Designing Anti-Poverty Policy Experiments from Observed Mobility Patterns

About this event

We develop a data-driven method for designing the optimal policy experiment to increase the chances of escaping poverty. We collected original in-person survey data on nearly 1,000 individuals who grew up poor in Memphis, Tulsa, and New Orleans. The data provide an unusually comprehensive picture of childhood conditions and adult outcomes among individuals born poor. We complement this new dataset with nationally representative, longitudinal data from the NLSY collected in real time. Using standard descriptive methods (e.g., OLS) to motivate an intervention implicitly assumes that individual characteristics can be altered in whatever ways the data deem predictive. Such methods would recommend pathways out of poverty that rarely occur in the data, such as encouraging college attainment among children who experienced abuse. We replace this assumption with a set of more plausible assumptions about the cost of altering the characteristics of individuals. Under these assumptions, the optimal experiment mirrors how individuals escape poverty in the data. Applying this method, we find that educational attainment is the strongest correlate of upward mobility. However, several noncognitive traits and childhood experiences that receive far less attention in traditional economic analyses are nearly as important. Our results suggest that policies focused solely on schooling or place-based interventions are unlikely to be sufficient, and that effective strategies must also address psychological skills and early life environments.