Incorporating Neighborhood Choice in a Model of Neighborhood Effects on Income

Abstract

Studies of neighborhood effects often attempt to identify causal effects of neighborhood characteristics on individual outcomes, such as income, education, employment, and health. However, selection looms large in this line of research, and it has been argued that estimates of neighborhood effects are biased because people nonrandomly select into neighborhoods based on their preferences, income, and the availability of alternative housing. We propose a two-step framework to disentangle selection processes in the relationship between neighborhood deprivation and earnings. We model neighborhood selection using a conditional logit model, from which we derive correction terms. Driven by the recognition that most households prefer certain types of neighborhoods rather than specific areas, we employ a principle components analysis to reduce these terms into eight correction components. We use these to adjust parameter estimates from a model of subsequent neighborhood effects on individual income for the unequal probability that a household chooses to live in a particular type of neighborhood. We apply this technique to administrative data from the Netherlands. After we adjust for the differential sorting of households into certain types of neighborhoods, the effect of neighborhood income on individual income diminishes but remains significant. These results further emphasize that researchers need to be attuned to the role of selection bias when assessing the role of neighborhood effects on individual outcomes. Perhaps more importantly, the persistent effect of neighborhood deprivation on subsequent earnings suggests that neighborhood effects reflect more than the shared characteristics of neighborhood residents: place of residence partially determines economic well-being.

Status Aversion, Attraction and Discrepancy as Drivers of Neighborhood Selection

“…They are often expressed in terms of transition matrices linking categories of origin and destination neighborhoods by some indicator of interest (e.g., Clark et al 2014). 4 Kim, Pagliara, and Preston (2005) (Kim et al 2005;Lee and Waddell 2010;van Ham et al 2018).…”

“…4 Using this approach, many studies have found that households are more likely to be attracted to neighborhoods where higher percentages of their own income and/or racial-ethnic group are present (cf. Boschman and van Ham 2015;Hedman 2013;Hedman et al 2011;Ioannides andZabel 2003, 2008;Kim et al 2005;Lee and Waddell 2010;Sermons and Koppelman 2001;Spring et al 2017;Steele et al 2016;van Ham et al 2018). 5 Unfortunately, none of the aforementioned works measures multiple dimensions of neighborhoods' income composition, confining themselves to measures like poverty rates and median family incomes.…”

Does Social Context Affect Poverty? The Role of Religious Congregations

“…From the multinomial logit model, we derive the linear probabilities reflecting the likelihood that an individual belong to a certain type of congregation (Van Ham and al., 2018). Following Ioannides and Zabel (2008), we transform these linear probabilities to generate correction terms akin to the inverse Mills ratios popularized by Heckman's (1979) two-stage regression framework.…”

“…We use data from the U.S. Congregational Life Survey and a multilevel approach to evaluate congregation effects. The problem with estimating congregation effects on poverty is that people are nonrandomly allocated to congregations: people select into groups based on their preferences and resources (Van Ham and al., 2018). That is, people tend to move to congregations that match their preferences.…”

“…Ham et al (2018) also estimate the second-step model substituting individual characteristics for the correction components derived from the selection model.…”