Demand Modeling, Forecasting, and Counterfactuals, Part I

Abstract

There are relatively few systematic comparisons of the ex ante counterfactual predictions from structural models to what occurs ex post. This paper uses a large-scale policy change in Boston in 2014 to investigate the performance of discrete choice models of demand compared to simpler alternatives. In 2013, Boston Public Schools (BPS) proposed alternative zone configurations in their school choice plan, each of which alters the set of schools participants are allowed to rank. Pathak Shi (2013) estimated discrete choice models of demand using families' historical choices and these demand models were used to forecast the outcomes under alternative plans. BPS, the school committee, and the public used these forecasts to compare alternatives and eventually adopt a new plan for Spring 2014. This paper updates the forecasts using the most recently available historical data on participants' submitted preferences and also makes forecasts based on an alternative statistical model not based a random utility foundation. We describe our analysis plan, the methodology, and the target forecast outcomes. Our ex ante forecasts eliminate any scope for post-analysis bias because they are made before new preferences are submitted. Part II will use newly submitted preference data to evaluate these forecasts and assess the strengths and limitations of discrete choice models of demand in our context.