NEIGHBORHOODS AND HEALTH Building Evidence for Local Policy

NEIGHBORHOODS AND HEALTH: Building Evidence for Local Policy Cross-Site Analysis Kathy Pettit and Tom Kingsley The Urban Institute

NNIP u Collaborative effort - Urban Institute and local partners in twenty cities u Partners operate local information systems - u Recurrently updated neighborhood data Multiple topics-data sources Success from 3 kinds of innovations 1. Technological 2. Institutional 3. Using information for change u Competitive procurement this study - 9 proposals, 5 selected

Research questions Neighborhood conditions and health outcomes u Basic relationships health, demographic and contextual variables - Multiple cities, consistent definitions - Variations across sites u How relationships change over time - Long time series almost all sites - Specifically – 1990 s trends in high-poverty areas u Feasibility – tract level analysis

5 sites - major differences Metro area population growth - Slow in Cleveland Providence - Rapid in Denver, moderate Indianapolis, Oakland

Racial/ethnic contrasts - Cleveland, Indianapolis: high share black - Denver, Providence: rapid Hispanic growth - Oakland: minority-majority expanding

Concentrated Poverty u Devastating effects, inner-city neighborhoods - Loss of the middle-class, institutions and role models - Concentration “magnifies poverty and exacerbates its effects. ” u Concentration - seemingly inexorable trend - u Major increases in 1970 s and 1980 s share of poor in high-poverty tracts What happened in the 1990 s?

Nationally – important reduction in concentrated poverty - Share of poor: down in high categories up in middle categories

5 sites, overall poverty rate - change not dramatic, 1990 s

Concen. poverty: bigger changes - major drops, Cleveland, Denver - modest declines, Indianapolis, Oakland - increase in Providence

% adults without HS degree Improvements in high-poverty neighborhoods – all sites

% adults without HS degree but major gaps remain re rest of city

% households rec. pub. assistance Major declines, high-poverty neighborhoods – all sites

% households rec. pub. assistance again, major gaps remain

Rental vacancy rates Tight housing markets exacerbate other problems

Research approach Neighborhood conditions and health outcomes u Scan of Birth and Mortality Outcomes - Diversity of levels and trends within & between cities u Examine relationships between health and neighborhood conditions - Bivariate and Multivariate analyses

Maternal and Infant Health Indicators u Pct. low-birth weight infants u Pct. of births to mothers with early prenatal care u Teen birth rates (age 15 -19) u Age-adjusted mortality rate u Infant mortality rate

Gaps between poor/non-poor vary u u Confirmed that gaps exist between highpoverty and non-poor tracts BUT gaps vary by city and indicator - Gaps most consistent for prenatal care - Gaps generally lower in Oakland, Providence (smaller geographies)

Pct. low birth weight births, 1998/2000 Large gaps seen in black/white cities

Teen Birth Rates in Denver

Some progress in health outcomes across all neighborhoods u Improvements in both high-poverty and non-poor neighborhoods u BUT generally not enough to close gaps u Direction and speed of change varies by city and indicator - Oakland generally saw most improvements - Teen birth, infant mortality rates dropped in all - Some troubling signs in Providence

Age-Adjusted Death Rates Begin together, but paths diverge High poverty census tracts

Oakland: Change in Early Prenatal Care Rates

Lessons from Introductory Scan u Improvements are possible u Racial/ethnic differences - Change in teen births - Low birth rates u Policy and programs can have impact - Healthy Start - Campaign for Healthy Babies

Pathways of Neighborhood Effects on Health (Ellen et al. 2001) u Physical Stressors u Social Stressors u Neighborhood-based Networks & Norms u Neighborhood Institutions & Resources

Static Research Hypotheses Poorer birth & mortality outcomes will be associated with: - higher levels of minority and immigrant population - higher shares of low-income households - higher levels of social risk factors - poor housing quality - higher crime rates - more mobile and transient population

Dynamic Research Hypotheses Some relationships will remain positive, but have decreased in strength over the 1990 s - high minority tracts and poor birth and mortality outcomes - lower-income tracts and lower prenatal care rates - lower-income tracts and higher low-birth weight rates

Socioeconomic Links Hold True u Poorer birth & mortality outcomes will be associated with: - higher levels of minority and immigrant population » True except foreign-born with low-birth weight & mortality rates - higher shares of low-income households » True for all measures (poverty, income, TANF, public assistance) - higher levels of social risk factors » True for all measures (No HS, female-headed HH with children, unemployed)

Physical, Social Stressors Matter Social Network Proxies Mixed u Poorer birth & mortality outcomes will be associated with: - poor housing quality » True for all measures (older housing, overcrowded, home values and mortgage amounts) - higher crime rates » Not true for prenatal care, strongest with teen birth rates - more mobile and transient population » True for renter-occupancy & vacancy, mobility » Not true for change in population

Mixed results in link between non-white areas & poor health outcomes u Some relationships will remain positive, but have decreased in strength over the 1990’s: - High minority tracts and poor birth and mortality outcomes » True for low birth weight » Not true for prenatal care and pct. Hispanic » Not true for teen birth (much higher correlations) or mortality rates (about the same)

Again, mixed results in link between lowincome areas & poor health outcomes u Some relationships will remain positive, but have decreased in strength over the 1990’s: - lower-income tracts and lower prenatal care rates » True but only slight drop in relationship - lower-income tracts and higher low-birth weight rates » True for all measures of income

Multivariate Regression u To test independent relationships of birth/mortality outcomes with neighborhood context variables u To test how much variation the neighborhood context variables explained u To test shifts over time for significance

Regression Model u. Neighborhood Indicators - Pct. African-American, Pct. Hispanic Average Family Income Pct. population not employed Pct pop > age 5 who moved in past 5 years u. Dummy Variables - City, year, and interaction of city & year

Regression Summary u Regression strongest for prenatal care rates - R-Squared: . 45 (teen births) to. 77 (prenatal care) u Neighborhood variables did explain some variation - Pct. population not employed had highest coefficient u Conditions specific to time and place matter - not just due to racial or income change