Session: Neighborhoods, Health and Selection
Room: Phillips 203
Time: Wed 08:30-10:00

Presenter: Zhenxiang Zhao (University of Illinois at Chicago. IHRP)

Discussant: Michael Grossman (City University of New York & NBER)

Abstract

This paper identifies the role of neighborhood characteristics on obesity using data from a randomized experiment, the Moving to Opportunity for Fair Housing Demonstration (MTO) study, to avoid the endogenous selection problem. The ever increasing prevalence rates of obesity have motivated a growing number of studies. However there is little consensus as to what are the primary causes of the recent rise in obesity. Physiologically, the cause of this increase in obesity is the excessive energy intake compared to energy expenditure, though it is unclear what factors are responsible for altering the energy intake/expenditure balance over the last three decades. The major impediment to identifying the causal effect of these factors is the lack of exogenous variation in these determinants of obesity. The MTO social experiment provides exogenous variation in place of residence and, consequently, exogenous variation in several potentially important environmental determinants of obesity, which can be used to identify the underlying causal mechanism of obesity.

In this paper, we adopted the MTO experimental data to test whether several widely-hypothesized environmental attributes are causes of obesity by linking the MTO Interim Evaluation data with numerous external data sources that provide information on environmental characteristics that the existing literature suggests as potential determinants of obesity, including food prices, restaurant and food store availability, physical activity facility availability, the prevalence of crime and population density. As a randomized experiment, MTO study provides neighborhood variations in those hypothesized determinants of obesity which are not systematically related to time-invariant personal characteristics that may affect obesity across treatment groups. In addition, the observed difference in weight status across treatment groups after randomization is of great importance as changed environmental factors might be factors that play a central role in making the difference.

Empirically, measures of these neighborhood contextual factors are added to the basic intention-to-treat model, and the differences in the treatment effects when these variables are and are not included provide indirect evidence of the causal effect of these factors on obesity. We find that the intention-to-treat effects change little after adding the contextual factors, indicating these widely-hypothesized environmental factors are unlikely to be causes of obesity. A null finding is of great significance as we have focused on widely cited causes of obesity and find no evidence that these factors matter for a group particularly affected by obesity.

Key Terms

Obesity, neighborhood contextual factors, random experiment

Authors:

Zhenxiang Zhao (University of Illinois at Chicago. IHRP) , Robert Kaestner (University of Illinois at Chicago. IGPA and Economics) and Xin Xu (University of Illinois at Chicago. IHRP)