How Entrenched Is the Spatial Structure of Inequality in Cities? Evidence from the Integration of Census and Housing Data for Denver from 1940 to 2016

Abstract

How entrenched is the spatial structure of inequality in cities? Although recent discussions provide conflicting answers to this question, the absence of long-term, longitudinal neighborhood data curtails direct examination of the issue. Focusing on the city of Denver, we develop a new strategy for analyzing neighborhood dynamics from 1940 to the present day. Our analysis of these data reveals surprising persistence in the income rank of neighborhoods between 1940 and 2016, which appears to be driven by the enduring position of white, upper-income places at the top of the neighborhood hierarchy. When low-income neighborhoods do rise in income rank, we find that change tends to be spatially concentrated in specific areas of the city and accelerates during broader historical episodes of urban change. We conclude that neighborhood inequality in Denver has endured over long periods of time and through substantial shifts in the wider urban landscape. Key Words: gentrification, GIS, inequality, neighborhoods, spatial demography.

城市的不平等空间结构有多么根深蒂固？近期对这个问题的讨论似乎得出相互矛盾的结论。此外，由于缺乏长期的纵向社区数据，也限制了以直接方法验证这个问题。我们以丹佛为例开发出一种新方法，用用户分析 1940 年至今的社区动态数据。对这些数据的分析显示出：从 1940 年至 2016 年间，社区收入排名保持了惊人的持续性，基本上都是由社区中上层的高收入白人群体所驱动。我们还发现，尽管也有低收入社区排名上升的情况，但变化在空间上集中在城市中的某些具体区域，在更广义的城市变化历史阶段中，会出现加速的现象。我们得出的结论是，丹佛的社区不平等现象一直长期持续存在，但在更广义的城市范围内经历了实质性的变化。关键词：中产阶层化、地理信息系统、不平等、领域、空间人口统计学。

¿Qué tan arraigada es la estructura espacial de la desigualdad en las ciudades? Si bien las discusiones recientes dan respuestas conflictivas a esta pregunta, la ausencia de datos longitudinales y a largo plazo a nivel de barrio limita el examen directo de este asunto. Enfocándonos en la ciudad de Denver, nosotros desarrollamos una nueva estrategia para analizar la dinámica del barrio desde 1940 hasta el día de hoy. Nuestro análisis de estos datos revela una sorprendente persistencia en el rango de ingresos de los barrios entre 1940 y 2016, que parece orientada por la posición duradera de los lugares de blancos con ingresos altos en el tope de la jerarquía vecinal. Cuando las barriadas de ingreso bajo logran ascender en el rango de ingresos, hallamos que el cambio tiende a estar concentrado espacialmente en áreas específicas de la ciudad y se acelera durante episodios históricos más amplios de cambio urbano. Concluimos que la desigualdad vecinal de Denver ha perdurado durante largos períodos de tiempo y a través de cambios sustanciales en el paisaje urbano más amplio. Palabras clave: barrios, demografía espacial, desigualdad, gentrificación, SIG.

Acknowledgments

We sincerely thank Ling Bian (editor) and two anonymous reviewers for their feedback and assistance with this research. For their insightful and constructive comments, we also recognize Elizabeth Roberta, Johannes Uhl, Seth Speilman, Peter Catron, Katrin Anacker, and participants at the annual meetings of the Population Association of America (Denver, 2018) and the Social Science History Association (Phoenix, 2018). We also thank the participants of the Geographical Perspectives on Inequality series at the Annual Meeting of the American Association of Geographers.

Funding

This research has benefited from research, administrative, and computing support provided by the University of Colorado Population Center (Project 2P2CHD066613-06), funded by the Eunice Kennedy Shriver National Institute of Child Health and Human Development. This content is the sole responsibility of the authors and does not necessarily represent the official views of the University of Colorado, CUPC, or NIH. We have also received support from the University of Colorado Boulder. The authors were provided access to the Zillow Transaction and Assessment Dataset (ZTRAX) through a data use agreement between the University of Colorado Boulder and Zillow, Inc. We gratefully acknowledge support by Zillow, Inc. More information on accessing the data can be found at http://www.zillow.com/ztrax. The results and opinions are those of the author(s) and do not reflect the position of Zillow Group. Funding for this work was provided through the Humans, Disasters, and the Built Environment program of the National Science Foundation, Award Number 1924670 to the University of Colorado Boulder.

Supplemental Materials

Supplemental data for this article can be accessed on the publisher's site.

Notes

1 In any given year, for example, the ranking of the highest income tract will be 1, the lowest income tract will be 0, and the median tract will be ranked at approximately 0.5.

2 These numbers are calculated from IPUMS (Ruggles et al. 2017).

3 In some cities, this return to the city has co-occurred with growing suburban poverty (Kneebone and Garr 2010; Cooke and Denton 2015).

4 Although it is challenging (perhaps impossible) to precisely demarcate a neighborhood (see Spielman and Logan 2013), census tracts are designed to be coherent spatial units and are thus suitable for our purposes.

5 Other tract databases are available, but the quality of the LTDB appears to be at least as good as these other products (Logan, Stults, and Xu 2016).

6 Because the data from ACS are based on estimates from sample data, the Census Bureau provides margins of error. In this analysis, we use the primary estimate from the ACS. To allay concerns that error in the ACS might distort our conclusion, however, we reran all analyses with no ACS data with the end year as the 2010 decennial census of the United States. These alternate results are almost identical to those that we obtained with the ACS (see supplemental materials).

7 These maps can be accessed in the National Archives Catalog.

8 We undertook our address matching within the statistical software R using the “ggmap” package and the Data Science Toolkit spatial data source, itself based on OpenStreetMap data (Kahle and Wickham 2015).

9 Because the 1970 census data do not contain detailed information on the Hispanic population, we undertake a simple linear interpolation for 1970 based on the 1940 and 1980 margins. This issue also highlights the fact that racial categories such as non-Hispanic white have only appeared in the census since 1980 (Omi and Winant 2014). We found no difference, though, when we used the individual records of the 1940 census to test the sensitivity of our results to using an all-black or non-Hispanic black classification.

10 We use the Denver Civic Center as the Denver CBD.

11 We derived these measures by sequentially subsetting the original ZTRAX data and generating raster layers of the building characteristics at different time points (Leyk and Uhl 2018b). The retrospective ZTRAX data underestimate the presence of historical buildings, and we continue to work on developing products to adjust for data missingness due to rebuilding.

12 This correlation is even stronger for absolute income levels and is roughly +0.61 from 1940 to 2016.

13 The correlation between the rank of tracts in 1940 and 1980 is +0.70 and is +0.50 for 1980 and 2016.

14 We include the 1940 income rank control so that our 1970 regression is comparable to other years. When the dependent variable is the 1940 or 1970 income rank, we do not include a control for the income rank of the tract in that same year. Our estimates from 1980 are not seriously affected by the inclusion of the 1940 income rank control.

15 In Figure S.6 (supplemental materials), we present the black and Hispanic estimates separately. The results are very similar whether we use a composite or separate measure of the black and Hispanic population shares.

16 In the supplemental materials, we show the conclusion to be generally similar whether or not we standardize the variables in this way (Table S.6).

17 Because the actual income levels of the Hispanic and black populations also vary over time, one might be concerned that racialized differences in income could be the sole driver of the trends we observe over specific place-based racial status effects. In the supplemental materials, we estimate a model with time-varying, group-specific income controls and find little change in our results (Table S.2).

18 See Lawton (2019) for a recent perspective and review of gentrification and uneven development.

19 These efforts could be further aided by using the ZTRAX data as a basis from which to undertake dasymetric refined interpolation to improve the accuracy of historical neighborhood estimates (Ruther, Leyk, and Buttenfield 2015; Zoraghein and Leyk 2019).

Additional information

Funding

This research has benefited from research, administrative, and computing support provided by the University of Colorado Population Center (Project 2P2CHD066613-06), funded by the Eunice Kennedy Shriver National Institute of Child Health and Human Development. This content is the sole responsibility of the authors and does not necessarily represent the official views of the University of Colorado, CUPC or NIH. We have also received support from the University of Colorado Boulder. The authors were provided access to the Zillow Transaction and Assessment Dataset (ZTRAX) through a data use agreement between the University of Colorado Boulder and Zillow, Inc. We gratefully acknowledge support by Zillow, Inc. More information on accessing the data can be found at http://www.zillow.com/ztrax. The results and opinions are those of the author(s) and do not reflect the position of Zillow Group. Funding for this work was provided through the Humans, Disasters, and the Built Environment program of the National Science Foundation, Award Number 1924670 to the University of Colorado Boulder.

Notes on contributors

Dylan Shane Connor

DYLAN SHANE CONNOR is an Assistant Professor at the School of Geographical Sciences and Urban Planning at Arizona State University, Tempe, AZ 85287. E-mail: [email protected]. His research focuses on geographical inequality, spatial demography, and immigration.

Myron P. Gutmann

MYRON P. GUTMANN is Professor of History and Director of the Institute of Behavioral Science at the University of Colorado Boulder, Boulder, CO 80309. E-mail: [email protected]. His research interests include historical populations and their relationship with the environment, health, and the economy.

Angela R. Cunningham

ANGELA R. CUNNINGHAM is a PhD alumna of the Geography Department at the University of Colorado Boulder, Boulder, CO 80309. E-mail: [email protected]. Her research interests include population geography, spatial history, and the coconstitutive relationships of place and individual life courses.

Kerri Keller Clement

KERRI KELLER CLEMENT is a PhD Candidate in History at the University of Colorado Boulder, Boulder, CO 80309. E-mail: [email protected]. Her research interests include Indigenous people's historic relationships with borders, disease, environments, and animals.

Stefan Leyk

STEFAN LEYK is Associate Professor of Geography at the University of Colorado Boulder, Boulder, CO 80309. E-mail: [email protected]. His research interests include uncertainty in the geospatial sciences and spatiotemporal analysis for the study of socioenvironmental systems.