Inequality Versus Inclusion in US Cities

The U.S. Department of Housing and Urban Development defines low-income as those making 80 percent or less of area median income. https://​www.​huduser.​gov/​portal/​datasets/​il.​html.​

In our analyses, we define people of color as any person identifying in US Census Bureau records as black or African American, American Indian and Alaska Native, Native Hawaiian and other Pacific Islander, other race two or more races, or Hispanic or Latino white. Because of the lack of data in 1980, the number of Hispanic or Latino whites in our racial education gap measure had to be imputed assuming equal distribution as in the overall population for percentage of non-Hispanic or Latino whites in that year. We undertake our analyses using this definition to be inclusive of different types of cities. In some cities, Hispanics outnumber African American residents and face barriers to inclusion that would be missed if we only examined racial disparities between non-Hispanic whites and African American residents. Similarly, in some cities, subpopulations of Asians outnumber other populations of color and face their own unique barriers to inclusion.

We use the US Census Bureau designation of income: “income received on a regular basis before payments for personal taxes, social security, union dues, Medicare deductions, etc”.

We use the US Census Bureau designation of poverty in which sets dollar income thresholds for each family size below which an individual within that family is considered “below the poverty line”.

We use an education indicator in our economic and racial inclusion indices in order to capture city-level service provision—a high score on these two measures would indicate that a city’s education system ensures quality education for all residents.

Ideally, we would also include a similar measure for the low-income population in our economic inclusion index. However, a decrease in the number of low-income residents could result from either increases in income or displacement. Individual-level data would be useful in getting at this issue, but is unavailable.

The rank-order information theory index is usually on a scale from 0 to 1, but we converted it to 0 to 100 for ease in interpretation and to make it more comparable to the racial segregation measure.

The most common cut point used to measure rent burden is 30%, but this cut point is not available for all years.

"Appendix" shows the inclusion and Gini index scores for every city in our sample in each year.

The correlations between economic inclusion and income inequality drop between 2000 and 2010, likely due to the Great Recession.

While city level fixed effects account for baseline differences across cities in the regression models for things such as baseline income levels, unemployment rates, and homeownership rates, this is not the case for the descriptive measures. As such, they should be interpreted with care.

To explore whether these relationships are different for cities that are increasing in inequality versus those that are decreasing, we split the sample into those that had a higher Gini coefficient in 2010 than in 1980, and those that had a lower Gini coefficient in 2010 than 1980 (results suppressed for brevity but available upon request of the author). Results hold for both cities within increasing Gini coefficients over time, and those with decreasing Gini coefficients over time.

Results suppressed for brevity but available upon request.