Upgrade to Pro — share decks privately, control downloads, hide ads and more …

NU18

 NU18

October 18, 2018 presentation of pilot research results at Northeastern University on redlining, segregation, and health

Christopher Prener

October 18, 2018
Tweet

More Decks by Christopher Prener

Other Decks in Education

Transcript

  1. Detrimental Influences Redlining, Race, and Health Christopher 
 Prener, Ph.D.

    Assistant Professor
 of Sociology Saint Louis 
 University 10.18.2018
  2. Detrimental Influences Redlining, Race, and Health Christopher 
 Prener, Ph.D.

    Assistant Professor
 of Sociology Saint Louis 
 University 10.18.2018
  3. Acknowledgments Northeastern University Department of Sociology & Anthropology Especially Isabel

    Geisler, Kathrin Zippel, Ph.D., and Linda Blum, Ph.D. Student Contributors Nivi Biju Branson Fox Jack Norys Institute for Health Equity and 
 Social Justice Research Especially Alisa K. Lincoln, MPH, Ph.D.
  4. AGENDA 1. Preface 2. Background: Redlining & St. Louis 3.

    Data & Pilot Methods 4. Redlining & Asthma in St. Louis 5. Next Steps NORTHEASTERN UNIVERSITY | BOSTON, MA | 10.18.2018
  5. FOR THE PROBLEM OF THE TWENTIETH CENTURY IS THE PROBLEM

    OF COLOR LINE W.E.B. Du Bois The Souls of Black Folk
 (1903) Wikimedia Commons
  6. REDLINING A - “Best” B - “Still Desirable” C -

    “Definitely Declining” D - “Hazardous”
  7. REDLINING A - “Best” B - “Still Desirable” C -

    “Definitely Declining” D - “Hazardous” “‘homogeneous,’ and were in demand in ‘good times or bad’” (Hillier 2005:216) Hillier, Amy. 2005. “Residential Security Maps and Neighborhood Appraisals.” Social Science History 29(2):207-233.
  8. REDLINING A - “Best” B - “Still Desirable” C -

    “Definitely Declining” D - “Hazardous” “In St. Louis, the white middle class suburb of Ladue was colored green because…it had ’not a single foreigner or negro.’” (Rothstein 2017) Rothstein, Richard. 2017. The Color of Law. New York, NY: W.W. Norton & Co.
  9. REDLINING A - “Best” C - “Definitely Declining” D -

    “Hazardous” B - “Still Desirable” “‘They are like a 1935 automobile—still good, but not what the people are buying today who can afford a new one.’” (Hillier 2005:217) Hillier, Amy. 2005. “Residential Security Maps and Neighborhood Appraisals.” Social Science History 29(2):207-233.
  10. REDLINING A - “Best” B - “Still Desirable” D -

    “Hazardous” C - “Definitely Declining” “‘infiltration of a lower grade population’” (Hillier 2005:217) Hillier, Amy. 2005. “Residential Security Maps and Neighborhood Appraisals.” Social Science History 29(2):207-233.
  11. REDLINING A - “Best” B - “Still Desirable” C -

    “Definitely Declining” D - “Hazardous” Hillier, Amy. 2005. “Residential Security Maps and Neighborhood Appraisals.” Social Science History 29(2):207-233. “‘detrimental influences in a pronounced degree,’ and ‘undesirable population or an infiltration of it’” (Hillier 2005:217)
  12. REDLINING A - “Best” B - “Still Desirable” C -

    “Definitely Declining” D - “Hazardous” Rothstein, Richard. 2017. The Color of Law. New York, NY: W.W. Norton & Co. “Lincoln Terrace was colored red because ‘it had little or no value today…due to the colored element now controlling the district’” (Rothstein 2017)
  13. 2. BACKGROUND: REDLINING & ST. LOUIS A SOCIOLOGY OF REDLINING

    ▸ Growth in recent interest in dynamics related to redlining ▸ Hillier (2003; 2005a; 2005b; 2005c), Rothstein (2017), and Woods (2012) have all recently focused heavily on the history of the Home Owners’ Loan Corporation ▸ Most sociological focus (~30 recent studies) has been on redlining as a piece of history related to the development of segregated metros and cities in the mid-20th century ▸ National Community Reinvestment Coalition report (2018) ▸ Little direct use of redlining data, particularly across cities ▸ No direct study of how health has been affected by redlining
  14. 2. BACKGROUND: REDLINING & ST. LOUIS SEGREGATION & HEALTH ▸

    Have long recognized the segregation may impact communities’ health • Yankauer (1950) is one of the first examples of research in this vein ▸ Hypothesis has been that segregation impacts individuals through concentrated poverty and disadvantage (Collins and Williams 1999) ▸ Studies treat segregation primarily as a compositional factor in neighborhoods ▸ Literature has focused primarily on African American communities (Anderson 2016)
  15. 3. DATA & PILOT METHODS DATA SOURCES Mapping Inequality (University

    of Richmond) GIS data for HOLC redlining boundaries in ~150 cities; Redlining data
  16. 3. DATA & PILOT METHODS DATA SOURCES Mapping Inequality (University

    of Richmond) GIS data for HOLC redlining boundaries in ~150 cities; Redlining data 500 Cities Project (Centers for Disease Control) Census tract estimates for 28 health behaviors, outcomes, and prevention; CDC data
  17. 3. DATA & PILOT METHODS CDC DATA Preventative 
 Health

    Care Health Outcomes 1. Arthritis 2. Current Asthma 3. High Blood Pressure 4. Cancer 5. High Cholesterol 6. Chronic Kidney Disease 7. COPD 8. Diabetes 9. Heart Disease 10.Mental Illness 11.Physical Health 12.Teeth Loss 13.Stroke Health Behaviors 1. Binge Drinking 2. Current Smoking 3. Physical Activity 4. Obesity 5. Sleep 1. Annual Checkup 2. Blood Pressure Medication 3. Colorectal Cancer Screen 4. Core Preventative Services
 for Men 5. Core Preventative Services
 for Women 6. Cholesterol Screening 7. Dental Visit 8. Health Insurance 9. Mammography 10.Pap Smear Test
  18. 3. DATA & PILOT METHODS DATA SOURCES Mapping Inequality (University

    of Richmond) GIS data for HOLC redlining boundaries in ~150 cities; Redlining data 500 Cities Project (Centers for Disease Control) Census tract estimates for 28 health behaviors, outcomes, and prevention; CDC data 2012-2016 American Community Survey (U.S. Census Bureau) Census tract estimates for demographic measures (race, poverty, etc.); ACS data Tiger/Line Shapefiles (U.S. Census Bureau) GIS data for census tract boundaries and other features
  19. 3. DATA & PILOT METHODS DATA SET CONSTRUCTION Redlining GIS

    Data Census Tract Boundary GIS Data Area (m2) of “C” and 
 “D” Grade per tract CDC Data ACS Data Tract-level data set with redlining,
 health, and demographic measures
  20. 3. DATA & PILOT METHODS DATA SET CONSTRUCTION Redlining GIS

    Data Census Tract Boundary GIS Data Area (m2) of “C” and 
 “D” Grade per tract CDC Data ACS Data Tract-level data set with redlining,
 health, and demographic measures Code for project specific calculations generalized in an R package redHealth Open science approach
  21. 3. DATA & PILOT METHODS DATA SET CONSTRUCTION Redlining GIS

    Data Census Tract Boundary GIS Data Area (m2) of “C” and 
 “D” Grade per tract CDC Data ACS Data Tract-level data set with redlining,
 health, and demographic measures Reproducible API calls for downloading data instead of “point-and-click” Open science approach
  22. 3. DATA & PILOT METHODS DATA SET CONSTRUCTION Redlining GIS

    Data Census Tract Boundary GIS Data Area (m2) of “C” and 
 “D” Grade per tract CDC Data ACS Data Tract-level data set with redlining,
 health, and demographic measures Repackaged CDC data in “tidy” format available in an R package cityHealth Open science approach
  23. 3. DATA & PILOT METHODS DATA SET CONSTRUCTION Redlining GIS

    Data Census Tract Boundary GIS Data Area (m2) of “C” and 
 “D” Grade per tract CDC Data ACS Data Tract-level data set with redlining,
 health, and demographic measures Full process implemented with open source tools (principally R) Open science approach
  24. 3. DATA & PILOT METHODS DATA SET CONSTRUCTION Redlining GIS

    Data Census Tract Boundary GIS Data Area (m2) of “C” and 
 “D” Grade per tract CDC Data ACS Data Tract-level data set with redlining,
 health, and demographic measures Full process documented publicly on GitHub and the Open Science Framework Open science approach
  25. 3. DATA & PILOT METHODS DATA SET CONSTRUCTION Redlining GIS

    Data Census Tract Boundary GIS Data Area (m2) of “C” and 
 “D” Grade per tract CDC Data ACS Data Tract-level data set with redlining,
 health, and demographic measures Analytic data set available for download via GitHub and citable via Figshare Open science approach
  26. Segregation calculations and mapping: 1. What is the geography of

    segregation in St. Louis? OLS and spatial regression: 2. What is the relationship between redlining and contemporary
 health outcomes? 3. DATA & PILOT METHODS DATA ANALYSIS Tract-level data set with redlining,
 health, and demographic measures
  27. New R package tidyseg with modern toolkit for segregation calculations

    Open science approach Segregation calculations and mapping: 1. What is the geography of segregation in St. Louis? OLS and spatial regression: 2. What is the relationship between redlining and contemporary
 health outcomes? 3. DATA & PILOT METHODS DATA ANALYSES Tract-level data set with redlining,
 health, and demographic measures
  28. Segregation calculations and mapping: 1. What is the geography of

    segregation in St. Louis? OLS and spatial regression: 2. What is the relationship between redlining and contemporary
 health outcomes? 3. DATA & PILOT METHODS DATA ANALYSES Tract-level data set with redlining,
 health, and demographic measures Models fit using the PySAL spatial analysis library for Python Open science approach
  29. 4. REDLINING & ASTHMA IN ST. LOUIS DESCRIPTIVE STATISTICS Variables

    Mean SD Range Asthma 0.11 0.02 0.08-0.15 C/D Grade* 0.41 0.39 0.00-1.00 Dissimilarity* 0.00 0.014 -0.03-0.04 Poverty 0.28 0.14 0.04-0.64 Medicaid 0.25 0.15 0.02-0.60 TANF** 0.04 0.03 0.01-0.19 SNAP** 0.29 0.17 0.02-0.67 Renters** 0.57 0.18 0.16-0.99 Uninsured 0.20 0.07 0.08-0.33 * - neighborhood, ** - households; otherwise proportion of individuals
  30. 4. REDLINING & ASTHMA IN ST. LOUIS FULL MODELS Variables

    OLS Spatial Error (GMM) C/D Grade -0.0005 (0.001)*** -0.0005 (0.001)*** Dissimilarity -0.325 (0.053)*** -0.292 (0.057)*** Poverty 0.017 (0.008)*** 0.016 (0.008)*** Medicaid -0.009 (0.008)*** -0.004 (0.009)*** TANF -0.014 (0.020)*** -0.013 (0.017)*** SNAP 0.030 (0.007)*** 0.026 (0.007)*** Renters -0.005 (0.003)*** -0.007 (0.003)*** Uninsured 0.133 (0.017)*** 0.139 (0.023)*** Lambda 0.447 (0.109)*** Constant 0.078 (0.002)*** 0.077 (0.003)*** Pseudo R2 0.947*** 0.950***
  31. 4. REDLINING & ASTHMA IN ST. LOUIS STEPWISE MODELS Variables

    OLS 1 OLS 2 OLS 3 C/D Grade 0.016 (0.004)*** 0.005 (0.003)†** -0.0004 (0.001)*** Dissimilarity -1.153 (0.073)*** -0.376 (0.056)*** Poverty 0.014 (0.007)*** Uninsured 0.172 (0.014)*** Constant 0.105 (0.003)*** 0.110 (0.001)*** 0.074 (0.002)*** Adjusted R2 0.098*** 0.737*** 0.937*** Log Likelihood 276.484*** 342.199*** 418.647*** AIC -548.968*** -678.399*** -827.294***
  32. 4. REDLINING & ASTHMA IN ST. LOUIS FULL MODELS Variables

    OLS Spatial Error (GMM) C/D Grade -0.0005 (0.001)*** -0.0005 (0.001)*** Dissimilarity -0.325 (0.053)*** -0.292 (0.057)*** Poverty 0.017 (0.008)*** 0.016 (0.008)*** Medicaid -0.009 (0.008)*** -0.004 (0.009)*** TANF -0.014 (0.020)*** -0.013 (0.017)*** SNAP 0.030 (0.007)*** 0.026 (0.007)*** Renters -0.005 (0.003)*** -0.007 (0.003)*** Uninsured 0.133 (0.017)*** 0.139 (0.023)*** Lambda 0.447 (0.109)*** Constant 0.078 (0.002)*** 0.077 (0.003)*** Pseudo R2 0.947*** 0.950***
  33. 4. REDLINING & HEALTH IN ST. LOUIS DISCUSSION ▸ Impact

    of redlining explained away by segregation and poverty ▸ Robust spatial findings that neighborhood-level asthma rates impacted significantly by segregation, poverty, and un-insurance rates ▸ Limitations • These models are not ideal for proportional outcomes • Models would benefit from regular geographical features (such as square kilometer grids) • Measurement error in both ACS and CDC data
  34. NEXT STEPS Analyses Methods 1. Census tract level measures of

    segregation 2. Development of tidyseg 3. Effect sizes for spatial regression models • Not a part of the PySAL library 4. Measuring effects across sparse data • Meta analysis is one possibility Funding 1. Initial grants applications a. National Institutions of Health LRP b. Internal funding mechanism 2. Longterm funding a. National Science Foundation? b. National Institutes of Health? 1. What are the geographies of segregation across the 150 cities in the Mapping Inequality data set? 2. What are the contemporary relationships between redlining and segregation? 3. What health outcomes are most salient for measuring the effect of redlining on contemporary population health? 4. What is the contemporary relationship between these health outcomes and historical redlining boundaries?
  35. Raw data, code, and docs archived on OSF
 . 10.17605/OSF.IO/KXWP5

    github.com/prenerlab/redHealth github.com/slu-openGIS/tidyseg Slides available via SpeakerDeck
 speakerdeck.com/chrisprener/nu18 Raw data, code available via GitHub
 github.com/prenerlab/redliningPilot [email protected]
 chris-prener.github.io
 LEARN MORE THANKS FOR COMING! Analysis data is archived on Figshare
  10.6084/m9.figshare.7223621 , : @chrisprener Project software github.com/slu-openGIS/cityHealth