Abstract
This article examines the relationship between racial/ethnic composition and neighborhood economic change in a multilevel and longitudinal framework. I employ multilevel modeling to examine how neighborhood minority composition is associated with change in neighborhood relative economic status from 1970 to 2010 in the largest 100 metropolitan areas of the USA. In the multilevel framework, the empirical analysis shows that the shares of black and Hispanic residents are consistently negatively related to neighborhood economic gain even when metropolitan-level factors are taken into account. This study also finds that the negative effect of neighborhood minority composition on neighborhood economic gain is differentiated by deindustrialization and minority composition at the metropolitan level. In the longitudinal framework, the findings show that the negative effect of neighborhood minority composition on neighborhood economic gain has declined over time.
Acknowledgment
The author thanks Byungwon Woo for his insightful and critical comments on earlier versions of this article.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1 To avoid including neighborhoods with atypical changing patterns, this study excluded tracts that had a population of less than 200 and more than 50 per cent of the population living in group quarters (e.g., inmates in correctional facilities and students in dormitories) in all panels. This exclusion removed about 2000 tracts for the panels of the 1970s, 1980s, and 1990s and about 1000 tracts for the panel of the 2000s from the data-set.
2 As shown in Equations (3) and (4), I included the cross-level interaction term with deindustrialization at the metropolitan level only for percentage black at the neighborhood level. This is because studies suggest that deindustrialization affected mostly black neighborhoods (e.g., Wilson Citation1987, Galster et al., Citation1997). I also ran another model including the cross-level interaction term which interacts percentage Hispanic at the neighborhood level with deindustrialization at the metropolitan level, but found that the interaction term was not statistically significant.
3 While this study uses neighborhood conditions at the beginning of one decade as predictors of neighborhood change during the decade, neighborhood change during the decade can result in different conditions of the predictors in the following decades, thereby causing an endogeneity problem. The author recognizes the potential bias associated with the endogeneity issue and considers it in running empirical analyses.
4 Its mathematical expression is log (yij,t / yij,t − 1) where y is the relative ratio of neighborhood per capita income to the 100-MA average of per capita income, i is neighborhood, j is metropolitan area, and t is time. Thus, a positive value and a negative value of the dependent variable indicate that a neighborhood's economic status improved and declined, respectively, during one decade.
5 For example, the dependent variable values are not just 80 per cent for a neighborhood in a small metropolitan area and 120 per cent for a neighborhood in a large metropolitan area. Rather, the dependent variable values are a neighborhood economic change from 80 per cent of the 100-MA average in 1990 to 100 per cent of the 100-MA average in 2000 in a small metropolitan area and a neighborhood economic change from 120 per cent of the 100-MA average in 1990 to 110 per cent of the 100-MA average in 2000 in a large metropolitan area.
6 The interaction index was computed as:
7 The data about manufacturing jobs and total jobs by metropolitan area were obtained from the Bureau of Economic Analysis' Regional Information System. The mathematical expression of deindustrialization, for example, in the 1970s panel is: (Manufacturing jobs located in a MA in 1979/Total jobs in a MA in 1979)—(Manufacturing jobs located in a MA in 1988/Total jobs in a MA in 1988). One-year lagged variable (e.g., 1979–1988 changes for 1980–1989 neighborhood economic change values) was employed, following Galster & Mincy's (Citation1993) assumption that a one-year lag of metropolitan economic conditions will be reflected in neighborhood economic change next year.
8 This study does not distinguish non-Hispanic blacks from Hispanic blacks in the 1970s model because there was no distinction between non-Hispanic black and Hispanic black in Census 1970. It is a relatively minor issue in the analysis because Hispanic blacks made up only 5 per cent of total blacks in 1970 (US Census Bureau, Citation2008).
9 This study focuses on the effects of blacks and Hispanics as they are the most segregated and more segregated from non-Hispanic whites, respectively.
10 One might be concerned with using the tract with the highest population density as the city center, given that some of the city centers in the USA do not have many residents. According to Taubenböck et al. (Citation2013), however, there is no universal definition or method to define the city center such as the Central Business District (CBD). In addition, the author finds that the tracts with the highest population density are still close to the CBD or contains the CBD and population density is often used to define the CBDs (e.g., Rosenthal, Citation2008; Thurstain-Goodwin & Unwin, Citation2000). Therefore, the author does not think that the method used in this study to define the city center invalidates the analysis.
11 The manufacturing jobs data are missing for some of the 100 MAs in the Bureau of Economic Analysis' Regional Information System. Thus, the number of observations at the metropolitan level from the 1970s to 1990s is more likely 100.
12 Gentrification is more like to occur in minority neighborhoods in near the city center. To examine the gentrification effect varying by distance from the city and the share of minority people, I ran the models including interaction terms between distance from the city center and percentage black and Hispanic. But, the interaction terms were mostly not statistically significant.