Neighborhoods and Health: Building Evidence for Local Policy:

Part 2
Cross-Site Analysis

[ Main page of Report | Contents of Report]

Contents

• Section 7 - Approach and Methodology
• Section 8 - Context Analysis
• Section 9 - Neighborhood Health Trends
• Section 10 - Ecological Correlations

Endnote

Section 7

APPROACH AND METHODOLOGY

This section begins with a description of the framework, drawn from relevant literature, that drives the work of our cross-site analysis. It notes what research of this kind can and cannot accomplish. It then describes the way we have structured the types of neighborhood conditions that we suspect influence health outcomes and, accordingly, the research hypotheses we will test. The section next describes the specific data--of different types and at different geographic levels--which we utilize in the research. Finally it presents the hypotheses themselves and reviews our methods of analysis in more detail. Specifics on analytic methods (bivariate and multivariate analysis) will be discussed in section 10.

FRAMEWORK

Specific Purposes

The notion than conditions of the social and physical environment in urban neighborhoods affect the lives of the residents has been the subject of speculation and scholarly research since the 1920s (Burgess 1925; Park 1929, 1936). This tradition of ecological research has demonstrated that neighborhood conditions matter, but clearly they are not the only things that matter.

Socioeconomic characteristics and behaviors of individuals and families, for example, also have an extremely important effect on their well-being, whatever their neighborhood of residence. In fact, the field offers stern warnings to researchers to avoid the "ecological fallacy" (i.e., concluding because there is a close statistical relationship between some neighborhood conditions and changing outcomes that the one is the cause of the other, without recognizing the role of other variables (e.g., family characteristics) that are not included directly in the analysis). Nonetheless, it is generally conceded that neighborhood conditions do have some effects independent of the influence of individual or family characteristics (Ellen and Turner 1997).

The limitations involved in obtaining individual and family level data are an important overall constraint on what this research could accomplish. Administrative data files with health indicators, such as those maintained by the NNIP partners, seldom contain (or at least permit agencies to release) detailed descriptive information on individuals and their families. Without such data, we cannot perform the type of multilevel analysis needed to explain more fully relationships between changes in neighborhood-level variables and health outcomes.

Still, there are other purposes of studying these relationships, and they can be of substantial practical value.⁽¹⁷⁾ Ecological analyses can be used effectively in identifying potential problems and developing hypotheses about what is causing them, if not for drawing definitive conclusions on causation. Most basically, health agencies need to know about trends in the extent to which (and where) health problems are spatially concentrated if they are to develop efficient strategies for prevention and care. Knowing trends in the relationships between specific neighborhood conditions and health outcomes can provide valuable hints as to likely changes to spatial patterns in the future and, thereby, to specific programmatic opportunities those shifts may imply.

Categories: Overview

In this analysis, we have grouped the conditions likely to influence health outcomes (our independent variables) into three basic categories, as suggested by our review of the literature. First, we divide neighborhood-level conditions into two groups: socioeconomic conditions (e.g., income and poverty levels) that indirectly influence health, and direct pathways that can have more direct effects (e.g., environmental hazards or crime).

Neighborhood Conditions: Socio-economic Factors

As noted, our first set of indicators tries to capture the demographic and socioeconomic status characteristics of the neighborhood, and there is little doubt of important associations with health. In the most comprehensive review of the literature we identified, Ellen, Mijanovich, and Dillman (2001) recognize a broad consensus that "residents of socially and economically deprived communities experience worse health outcomes on average than those living in more prosperous areas . . . suffer from higher rates of heart disease, respiratory ailments and overall mortality." Yet more seriously, their review suggests to them that:

"…neighborhoods may primarily influence health in two ways: first, through relatively short-term influences on behaviors, attitudes, and health care utilization, thereby affecting health conditions that are most immediately responsive to such influences; and second, through a longer term process of "weathering," whereby the accumulated stress, lower environmental quality, and limited resources of poorer communities, experienced over many years, erodes the health of residents in ways that make them more vulnerable to mortality to any given disease (Geronimus 1992)."

There have been numerous studies on linkages between neighborhood socioeconomic conditions and varying types of health outcomes. Researchers have found correlations between socioeconomic status (SES) and indicators of health-related behavior, such as smoking (Kleinschmidt, Hills, and Elliot 1995) and physical activity and body-mass index (Robert 1999). Relationships between low-SES neighborhoods and mental health problems have been examined by Aneshensel and Sucoff (1996)--a study of adolescents in Los Angeles County --and Katz, Kling, and Liebman (2000)--psychological benefits of moving to better neighborhoods in the Moving to Opportunity Program.

Perhaps the most work of this type has been done on associations between neighborhood socioeconomic characteristics and birth-related outcomes, including low-birth weight births (Collins and David 1990; Duncan and Laren 1990; O'Campo et al. 1997) and infant mortality (Collins and David 1992; Coulton and Pandy 1992; Guest, Almgren, and Hussey 1998). Research has also evidenced the association between distressed neighborhood conditions and lower early prenatal care rates among African-American mothers (Perloff and Jaffee 1999). Finally, Robert (1998) and Marmot et al. (1998) have found relationships between lower neighborhood SES measures and more chronic health problems (self-rated) of adults.

For this research, we have subdivided indicators in this broader category into three subgroups. The first is demographic, covering data on race and age composition. The second, and probably most significant, is economic. To represent the economic circumstances in a neighborhood, we use several variables, including average household income, the unemployment rate, and the overall poverty rate. The third subgroup includes measures related to social risks, such as the share of adults with no high school degree, the share of households receiving public assistance, and the share of families with children that are headed by females.

Neighborhood Conditions: Direct Pathways

Ellen, Mijanovich, and Dillman (2001) identify four pathways through which neighborhood can affect health more directly: (1) neighborhood institutions and resources; (2) stresses in the physical environment; (3) stresses in the social environment; and (4) neighborhood-based networks and norms.

The first of these refers to the availability of neighborhood institutions and assets, including health care facilities, grocery stores, and reliable transit, which can affect individuals' ability to access health care, healthy food, or other health supports in their neighborhood or elsewhere. The second pathway refers to physical conditions or contaminants that directly affect people's health, such as environmental hazards or lead paint. The third touches on the mental or psychological stress from neighborhood conditions such as crime. The last refers to the social networks, support systems, and community expectations of a neighborhood that can help promote healthy behaviors.

The contextual data available to us cannot measure all of these mechanisms directly, but they can act as imperfect proxies for the underlying processes for three of them. For the first of the pathways identified, neighborhood institutions and resources, we do not have data from any source that we believe can serve as adequate proxies.

However, we can access a set of indicators on housing quality that should act as a proxy for physical stressors in the neighborhood, dealing with the age of housing (older housing is more likely to have problems with lead paint, poor heating and plumbing systems, and run-down structures), the extent of overcrowding (overcrowded housing has been associated with less sanitary conditions and the spread of disease), and measures related to home values. Use of the latter is responsive to hedonic price modeling theory, which says that a home price represents a bundle of characteristics that have a particular value in the housing market. We include average home values and values of home purchase mortgage loans as measures of overall quality of the housing in the neighborhood.

Next, we use data on crime rates to represent social stressors in residents' lives. Data from local systems are available on total, violent, and property crime rates in the five cities. Based on the work of Zapata et al. (1992) and others, we would expect the violent crime rate to be a greater stressor than property crime and thus to have a stronger correlation with health outcomes.

The final pathway, neighborhood-based networks and norms, is difficult to quantify. To investigate proxies for this concept, we selected several indicators of turnover and mobility within tracts. They are based on the assumption that in areas where people move around a lot, there is less opportunity to develop meaningful connections with neighbors or be integrated into neighborhood social networks. In this analysis, we used higher rates of renter-occupied housing, vacancies, and home purchase mortgages as proxies for weaker social ties. They will also have a greater proportion of people who lived in a different house in 1995, and a rapidly changing (growing or declining) population. Finally, we suggest that a greater number of home improvement mortgages is a positive sign of stability, indicating households who are committed to staying in their home and area.

Putting all of this together, we will be reporting on relationships between health indicators and independent variables in five major categories: (1) socioeconomic conditions; (2) physical stressors; (3) social stressors; and (4) social networks

DATA SOURCES AND DEFINITIONS

We now discuss the data assembled for this analysis. Basically, there were two types of sources: (1) the data systems operated by the participating local partners in NNIP and (2) national datasets with information at the census tract level. In our discussion of the former, we distinguish between data on health outcomes and those on contextual variables. We also note two types of national datasets: decennial census data, as presented in the Neighborhood Change Database (NCDB), and the Home Mortgage Disclosure Act data files.

Geographic Definitions

Most of the analysis in this report pertains to the "central city/county" in each study site; the central area within each metropolis for which we were able to obtain health-related data (see discussion below). For one of our sites, this was the central county, Cuyahoga County in metropolitan Cleveland. For the others, we use central city boundaries as our reference area, although it should be noted that the central cities are also counties in two of the sites: Denver County in metropolitan Denver and Marion County in metropolitan Indianapolis.

In some sections, we also present data for the metropolitan areas in full: the central city/county plus a number of additional surrounding counties. In these cases we use the 2000 boundaries of the Primary Metropolitan Statistical Area or the Metropolitan Statistical Area as defined by the federal government (Office of Management and Budget).

Data Systems Operated by Local NNIP Partners

Before we discuss individual variables, it should be helpful to say a few things about the data systems operated by NNIP's local partners in general. All 20 of them have built or are building advanced GIS information systems with integrated, recurrently updated information on neighborhood conditions in their cities. This is a capacity that did not exist in any U.S. city in the 1980s. The breakthrough became possible because (1) most administrative records of government agencies (for example, on crimes or births) are now computerized; and (2) inexpensive GIS software now exists that can match the thousands of addresses in these records to point locations, and then add up area totals for small geographic areas (such as blocks or census tracts).

The indicators in their systems cover topics such as births, deaths, crime, health status, educational performance, public assistance, and property conditions. Operating under long-term data-sharing agreements with the public agencies that create the base records, they recurrently obtain new data, integrate them in their systems, and make them available to a variety of users for a variety of purposes. Their accomplishment demonstrates that, while never easy, it is quite possible today to overcome the past resistance of major public agencies to sharing their data in this way.⁽¹⁸⁾

Important for this research is what is known about the quality of their data. Urban Institute staff were generally familiar with their data holdings and the procedures they follow to ensure quality before this project began, and we found out more during the project about the specific data files sent to us for this work. All sites do follow regular procedures to check and clean the files they receive from public agencies, and the individual indicators they calculate and disseminate are documented (see web sites listed in annex A for the selected partners).

An advantage for us was that the specific data files we needed from them for this effort were among the highest quality in their systems: files on (1) vital statistics; (2) reported crimes; and (3) Aid to Families with Dependent Children/Temporary Assistance for Needy Families (AFDC/TANF) recipiency. For all of these, particularly the first two, the original data providers work under fairly tight guidelines as to data quality and use definitions that are reasonably standard for national reporting.

Local Data - Health Indicators

Virtually all NNIP partners obtain detailed vital statistics data (births and deaths) at the census tract level on an annual basis. A few maintain other health-related indicators (e.g., from records on immunizations and hospital admissions), but these indicators were not useable for our cross-site analysis because they were not uniformly available, let alone uniformly defined, in all five sites. Therefore, we relied solely on vital statistics data to develop health indicators for this component of the research. The selected measures are as follows:

Maternal and infant health indicators

• Percentage of low birth weight births of all births (low birth weight defined as less than 2,500 grams)
• Percentage of births where mother received early prenatal care (in first trimester)
• Teen birth rate (number of births to teens aged 15 to 19 per 1,000 women of that age group)

Mortality indicators

• Infant mortality rate (deaths of infants 0 to 12 months old per 1,000 live births in the same year)
• Age-adjusted mortality rate (total deaths per 100,000 population that would have occurred assuming local death rates by age category and the national percentage distribution of population in the same categories)⁽¹⁹⁾

We obtained all data needed to construct these indicators by census tract for all of the individual years from 1990 through 2000 for which the partners maintained the information (see table 7.1). Our work was generally guided by advice on constructing indicators based on vital records provided by Coulton (1998).

All data were provided to us by age and race/ethnicity categories within tracts. As to the latter, local designations in Denver, Oakland, and Providence were combined to form five categories in our analysis: Hispanic plus four groups of non-Hispanics, white, black, Asian, and other. Data from Cleveland and Indianapolis did not allow us to separate Hispanics from non-Hispanics within races, so the data for those sites are not comparable. While this must be kept in mind, we do not believe it much affects our inter-site comparisons, since there are comparatively few Hispanics in those two sites.

For the teen birth rate and the age-adjusted mortality rate, we constructed denominators from census data. To do so, we interpolated between 1990 and 2000 numbers in the relevant age categories in each tract (straight-line method) to create annual estimates to correspond to the years of the vital statistics data. Although we know change was not likely to be exactly uniform over the 1990s, this procedure seems generally reasonable and has the benefit of being consistent across sites.

We acquired the data in early 2002, after multiple conversations with staff in our partner organizations to clarify variable definitions, file specifications, and steps taken to clean the data. Where possible, we checked county-level totals and rates in the data provided to us by the partners with comparable measures for the same counties independently reported by the original local data providers to the National Center for Health Statistics. We found no unreasonable differences.

The "Rare Events" Issue

The last and most challenging step in data development was to consider how to reliably depict indicators derived from rare events (e.g., the number of low-birth weight births or infant deaths in one census tract in any given year). As Buescher (1997) explains:

Most health care professionals are aware that estimates based on a random sample of a population are subject to error due to sampling variability. Fewer people are aware that rates and percentages based on a full population are also estimates subject to error. Random error may be substantial when the measure such as a rate or percentage, has a small number of events in the numerator. . . . A rate observed in a single year can be considered as a sample or estimate of the true underlying rate.

Buescher recommends computing a confidence interval (the interval within which we would expect the "true" rate to fall a certain percentage of the time) around the proportion or rate to help decide the size of the numerator needed for the analysis at hand. There is no hard and fast rule, but a numerator of 20 cases is often considered as an absolute minimum, since for any smaller number a 95 percent confidence interval will be wider than the rate itself. With this in mind, the numbers for infant mortality at the census tract level were found to be too small for use in this research.

To address this issue in neighborhood analysis, the data often have to be grouped to obtain larger numerators. There are two ways to do that. First, add the data for several years together for one census tract (e.g., present information for a multiyear period such as 1991-1993). Second, add the data for several tracts together for one year (e.g., present information for a new neighborhood aggregation of several tracts). Neither approach is better than the other is in general, and very small numbers may require that both approaches be used. The choice should be made based on the purposes of the analysis at hand (i.e., depending on whether you care more about a high level of geographic detail or a more finely grained examination of change over time).

Table 7.2 shows characteristics of the distribution of numerator sizes for different variables in our five sites. The data are shown first for census tracts and then for neighborhood clusters (clusters of adjacent tracts that each of the cities uses for planning purposes). These cluster definitions may be very useful for local planning and action, when policymakers and community members have a common understanding of the different areas. Unfortunately, they vary a great deal in size and how they were defined, making cross-site comparisons difficult. For example, the city of Cleveland has 224 tracts and 36 neighborhood clusters, for an average of 6.2 tracts per cluster, whereas Denver has only 1.8 tracts per cluster.

For total births and births to mothers with early prenatal care, a comparatively small share of the tracts had fewer than 20 events during the three-year period 1998-2000, whereas for low-birth weight births and births to teens, the share below 20 is very high. Switching to neighborhood clusters reduces all shares below 20 significantly, although the shares are still fairly high for low-birth weight births and births to teens in all cities except Cleveland.

It is one thing to use tract-level data in a regression (as we do in section 5), but it is quite another to present exact rates in a table form for individual tracts or even neighborhoods at these levels. Tract-level tables with three-year data might make sense for total births and births to mothers with early prenatal care, if a special symbol instead of a number were given for tracts with very small numerators. However, switching to the neighborhood level would certainly be advisable for rates of low-birth weight births and births to teens, and strong cautions would need to be stated.

In this research we employed both approaches at different stages. We avoid these problems in our analysis of time trends in section 4 by presenting data only for higher aggregations of time and geography. We group the data for three-year periods and present results for only two geographic aggregations of tracts in each city: (1) all high-poverty tracts and (2) all other tracts. In the bivariate and multivariate analyses (as will be explained more fully in section 5), we use tract-level data grouped for three-year periods. Each three-year tract average is an observation in the regression, and we believe the large number of observations (more than 8,000) and the multivariate methodology offset any random year-to-year variation.

Local Contextual Variables

The bulk of the demographic and contextual data needed for the cross-site analysis could be obtained from the census (see discussion below), but we felt that it would helpful to have two types of year-by-year information from local partners' data systems in addition.

The first is reported crime. We obtained tract-level data for Part I crimes (as uniformly defined by the FBI) from all sites. The reporting of Part I crimes is guided by the standards set by the Uniform Crime Reporting (UCR) program. The UCR provides a nationwide view of crime based on the submission of statistics by law enforcement agencies throughout the country. Part I crime consists of eight offenses--murder, forcible rape, robbery, aggravated assault, arson, burglary, larceny-theft, and motor vehicle theft. Murder, rape, robbery, and aggravated assault are crimes against persons. Arson, burglary, larceny-theft, and auto theft are crimes against property.

Data are provided for the following summary indicators for years as noted in table 7.1:⁽²⁰⁾

• Violent crimes per 1,000 population
• Property crimes per 1,000 population
• Total Part I crimes per 1,000 population

The second topic is federal welfare (AFDC/TANF) recipiency. In this case, Oakland is the only one of our sites that has been unable to obtain data on the topic. For the others, three provided individual record data and one provided household data. We have calculated the following indicators:⁽²¹⁾

• Individuals receiving AFDC/TANF as a percentage of total population (Cleveland, Denver, and Indianapolis)
• AFDC/TANF cases as a percentage of total households (Providence)

NNIP partners maintain other contextual indicators that could have potentially been useful for this analysis. For instance, Denver, Oakland, and Providence have more extensive education data, while Cleveland and Providence have more information about property conditions. Because we did not have data on these topics uniformly across at least three or four sites, however, we decided not to try to incorporate a wider range of variables in this work.

National Data Files: The Census and the Neighborhood Change Database

Before NNIP-type data systems were set up, tract-level data from the U.S. censuses were virtually the only nationally comparable indicators of neighborhood conditions in America. For our type of analysis, however, the data made available directly by the U.S. Bureau of the Census have a problem: all is not held constant from one census to the next. Some variable definitions and, more important, about 35 percent of tract boundaries change between censuses, so the data are not directly comparable over time.

To remedy this, the Rockefeller Foundation funded the Urban Institute and GeoLytics, Inc., to go back over the definitions and the data (using block data where possible to make adjustments) and achieve comparability. The result was the Neighborhood Change Data Base (NCDB), the only database that contains nationwide census data at the tract level with tract boundaries and variables that are consistently defined across the four U.S. censuses from 1970 through 2000.⁽²²⁾

In this research, we rely on the NCDB as a primary source of data for measuring 1990-2000 conditions and trends in most of the topical categories introduced earlier in this section. At the neighborhood level, they include data on demographic, economic, social, and housing characteristics and neighborhood stability. The NCDB is also our source for measures of economic health and segregation at the metropolitan level.

Since the sites' year-by-year data on health and contextual indicators are based on tract boundaries as defined for the 1990 census, we conduct all of these analyses using 1990 tract boundaries (i.e., weights developed in the NCDB were used to enable us to present 2000 census data for 1990-defined tracts).

National Data Files: Home Mortgage Disclosure Act

The Federal Reserve annually releases files on home mortgage applications by census tract nationwide, as required by the Home Mortgage Disclosure Act (HMDA).⁽²³⁾The files contain records on individual applications, including the census tract identifier, loan amount, race and income of applicant, purpose of loan (purchase/home improvement, owner/renter occupied), and whether the application was approved or denied. To prepare for this analysis, we compiled the individual loan data for originated mortgages (those both approved by the banks and accepted by the borrower) from 1995 through 1999 by tract. We then constructed indicators for the number and average value of loans, broken down by those used for home purchase and those used for home improvement. These indicators act as proxies for neighborhood investment and home values.

RESEARCH HYPOTHESES

Static Hypotheses

Considering our potential range of independent variables within the categories noted earlier, we have formed the following hypotheses about these associations at any point in time.

Socioeconomic hypotheses

• Census tracts with a majority nonwhite population and higher levels of immigrants will have higher levels of mortality and poor maternal and infant health outcomes than majority white census tracts.
• Low-income census tracts, as measured by poverty rate and average income, will be associated with poorer scores on the mortality and the maternal and infant health measures than higher-income tracts.
• Census tracts with higher overall social risks (e.g., lower education and employment levels, higher rates of welfare or public assistance recipiency) will have poorer scores on the mortality and the maternal and infant health measures than those with lower risks.

Physical stressors

• Census tracts with poor housing quality, as measured by age of the housing, overcrowded units, and low home values, will have higher levels of mortality and poorer maternal and infant health outcomes than stronger and more stable tracts.

Social stressors

• Census tracts with high total, violent, or property crime rates will have poorer scores on the mortality and the maternal and infant health measures than safer tracts.

Social networks

•  Census tracts with less stable populations, as measured by renter occupancy, vacancy rate, and mobility rate, will have higher levels of mortality and poorer maternal and infant health outcomes than stronger and more stable tracts. Tracts with less change in total or minority population or a higher rate of home improvement or refinancing loans will have better mortality and birth outcomes.

Dynamic Relationships

The discussion so far has focused on relationships between indicators at a point in time. Since we had indicators over a period of several years, we were able to examine how these relationships changed over time. We tested all of the hypotheses above to see if the associations remained constant throughout our analysis period. The time periods examined depended on the data source. We looked at 1990 and 2000 census data, 1995-2000 HMDA data, and varied dates for local contextual variables. Generally, we hypothesized that the relationships held over the 1990s; that is, as the level of one indicator that was negatively related to health (such as the crime rate) rose, the health outcomes worsened.

However, there are three cases in which we think that the relationships will continue to be in the same direction but will be weaker at the end of the decade than at the beginning. First, we know that racial disparities, while still considerable, are decreasing for some maternal and health outcomes, so we suspect that the relationship between high-minority tracts and poor health outcomes may be somewhat reduced (Keppel 2002). Second, Medicaid, Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), and other public programs have been expanded over the past decade, enabling more pregnant women to access prenatal care. Low-income tracts should therefore be less correlated (although still positively) with the percentage of births to mothers receiving late or no prenatal care.

Finally, we believe two factors have altered the relationship between low-birth weight births and the income level of the tract. The better care that low-income women are receiving will reduce the number of low-birth weight births and allow more of the remaining low-birth weight babies to survive. On the higher end of the income scales, more affluent women are delaying having children, thereby increasing their chances of having a low-birth weight baby. As with the first two, we believe the relationship will remain significant and positive, but at a lesser magnitude than earlier in the decade.

In sum, our hypotheses about changing relationships between health and demographic variables are as follows:

• The correlation between high-minority tracts and poor birth and mortality outcomes will remain positive, but has decreased over the 1990s.
• The correlation between low-income tracts and births with late or no prenatal care will remain positive, but has decreased over the 1990s.
• The correlation between low-income tracts and high rates of low-birth weight births will remain positive, but has decreased over the 1990s.

METHODOLOGY

Tables, Graphs, and Mapping Analysis

We have been working with a large amount of information, and our challenge has been to develop statistical and graphic methods to test these hypotheses and to cogently display the key findings. We begin our analysis in sections 8 and 9 with a series of tables and graphs. Section 8 tells the story of the changing context of urban change in the 1990s in our five sites. Tables there cover comparative metropolitan characteristics, and then go on to contrast demographic and nonhealth conditions and trends in high-poverty neighborhoods with those in other urban neighborhoods over the decade.

Section 9 begins by similarly contrasting health conditions and trends in high-poverty tracts with other tracts in the five cities. Differences between high-minority tracts and low-minority tracts within each city are also discussed. Using these divisions, we examine the extent to which these noncontiguous groupings of tracts are helpful in revealing health disparities among different types of neighborhoods. From the tabular information, we identify interesting stories to depict in graphic format. We include line charts illustrating the change over time and differences between types of neighborhoods across cities.

Also in section 9, we present illustrative maps of the most informative indicator values and the change in levels over the time. High-poverty areas (tracts with greater than 30 percent poverty) are indicated by a dot pattern laid over colors that indicate patterns for the health indicators.

Correlation and Regression Analyses

In section 10, we calculate the correlation coefficients among the health and contextual variables across time. These matrices allow us to examine the magnitude and direction of the connections among variables. With our longitudinal data files, we examine how some of these relationships have changed over the 1990s--whether they are growing stronger or weaker. These results will be used to confirm or reject our hypotheses and to suggest the most meaningful variables to map together. Full correlation matrices are included in annex C.

We use multivariate analysis to analyze health outcomes (e.g., low birth weight rates) as dependent variables. Independent variables will include some area characteristics as control variables (e.g., sociodemographic characteristics), as well as the community characteristics that are being investigated as possible causes for the health problem (such as crime rates, housing quality, or other factors).

Again, as discussed above, relationships that are statistically significant will not definitively identify causal relationships but rather will be used in discussions with communities as ways to develop hypotheses regarding what could be contributing to an identified community problem.

[Go To Contents]

Section 8

CONTEXT ANALYSIS

INTRODUCTION

As emphasized earlier, the centerpiece of this project is the analysis of neighborhood-level health trends and correlations that will be presented in sections 9 and 10. This section presents data on conditions and trends in our various categories of independent variables that we believe add to our ability to interpret ecological relationships and their implications. We develop descriptive characterizations of the context of urban change under way in the 1990s as the health trends we study were emerging. We examine conditions at the metropolitan as well as neighborhood levels, for the five study sites and for the 100 largest U.S. metropolitan areas on average.

Current notions of trends in urban neighborhoods are largely formed by literature based on data from the 1980s. Most prominent have been the accounts of the deterioration of neighborhoods in the inner city and the increasing concentration of urban poverty in that decade (Jargowsky 1997; Wilson 1987). That story featured the decline of well-paying manufacturing jobs in the cities and the departure of the black middle class from urban ghettos, leaving little in the way of economic opportunity, role models, or supportive institutions for those left behind. Increasingly recognized as critical in this mix was the decrease in the share of all adults in these areas who had jobs (Wilson 1996). Much of the research on the relationships between neighborhood conditions and health (e.g., as reviewed by Ellen, Mijanovich, and Dillman 2001) was developed in that context, and it generally found deteriorating health going hand-in-hand with declines in other neighborhood conditions.

This section reviews data on what happened to these background conditions in the 1990s. It is structured in accord with the indicator categories presented earlier, although we felt it would be wise to characterize differences among the five study sites at the metropolitan level first before going into variations in neighborhood conditions. Thus, we begin with a review of changes and differences in metropolitan demography, economic strength, and spatial structure. We then examine trends in the categories established for neighborhood conditions: (1) socioeconomic conditions (demographic, economic, and social); (2) physical stressors; (3) social stressors; (4) social networks.

METROPOLITAN CONDITIONS AND TRENDS

Demographic and Economic Change

Table 8.1 shows that the total size of the metropolitan populations of the five study sites in 2000 did not vary dramatically. They ranged from 1.2 million (Providence) to 2.4 million (Oakland). All are among the 100 largest metropolitan areas in the country, and their average size, 1.9 million, is modestly above the 1.7 million average for that group.⁽²⁴⁾

Along other dimensions, however, the five are very different from one another.⁽²⁵⁾Among the most important are differences in the pace of population and employment growth. Population in three of the sites grew more rapidly than the 14 percent per decade rate experienced by the large metropolitan areas on average: Denver at 30 percent, and Indianapolis and Oakland at 16 and 15 percent, respectively. The other two sites experienced much slower population growth: Cleveland at 2 percent and Providence at 5 percent.⁽²⁶⁾

Employment changes followed the same general pattern. Cleveland and Providence have grown slowly, Oakland and Indianapolis are more in the middle range, and Denver's growth has been most rapid by far. Rapid growth creates tensions (e.g., difficulty in keeping up with infrastructure demands), but for examining the effects on health outcomes, the disadvantages of slow growth (e.g., difficulty of providing employment opportunities to low-income residents) are probably more important.

Differences in racial/ethnic composition are also pronounced. The 2000 census data indicate that the two Midwestern metropolises have comparatively large African-American populations (14 percent in Indianapolis, about the same as the average for the top 100, with a notably larger 19 percent in Cleveland) and much smaller nonblack minority shares (Hispanic, Asian, and other at 5 percent in each, compared with a 22 percent average for the top 100). Denver and Providence exhibit the reverse pattern, with higher nonblack minority shares (23 and 13 percent, respectively) and low black shares (6 and 5 percent). Oakland stands alone with large shares in both categories: a remarkable total minority share of 50 percent (13 percent black plus 37 percent nonblack).

Differences in other economic indicators on table 8.1 are generally as expected given what economic growth rates suggest about comparative economic health. Denver and Indianapolis had the lowest unemployment rates in 2000 (4 percent compared with 5 to 6 percent for the others). Two other measures show important improvements for all five sites over the 1990s. In all of them, the share of all adults with college degrees increased (from 20 to 24 percent on average), and average family income (constant 1999 dollars) increased significantly (by 13 percent on average). Still, there were notable differences. Denver and Oakland had shares with college degrees (34 to 35 percent) much above the rates for the other sites (23 to 26 percent). Income growth was most rapid in Denver (20 percent), in the middle range in Indianapolis and Oakland (14 and 15 percent, respectively) and slower in Cleveland and Providence (10 and 4 percent, respectively).

Interestingly, the poverty rate (the percentage of the population in poverty) did not vary by much across these sites in 2000 and had not changed by much over the preceding decade. As might be expected based on the economic indicators discussed above, the overall poverty rate in 2000 was lowest in Denver and Indianapolis (8 to 9 percent), but the others were not far behind (10 to 12 percent).

Metropolitan Spatial Structure

There is a substantial literature on the deleterious effects of racial segregation on society, recently including impacts on health (Ellen 2000). Similar concerns have been raised about segregation of the poor within the nation's urban regions--the spatial concentration of poverty. Jargowsky (1997) notes that in neighborhoods where very high proportions of the population are poor--

families have to cope not only with their own poverty, but also with the social isolation and economic deprivation of the hundreds, if not thousands of other families who live near them. The spatial concentration of poor people acts to magnify poverty and exacerbate its effects (p. 1).

A decade ago, researchers found that while it remained high, racial segregation (blacks vs. whites) generally abated somewhat in the 1980s, but that decade had seen an acceleration of what appeared to be an inexorable increase in the concentration of poverty. The news from the 2000 census in this regard is more positive. Black/white segregation continued to abate modestly, and there was a clear reduction in the concentrated poverty in most of the country.

Table 8.2 presents the data. The first lines show the changes in the concentration of poverty (measured as the share of the poor population in each area that live in high-poverty neighborhood - those with poverty rates of 30 percent or more). The remainder of the table presents values for commonly used indexes of segregation or dissimilarity. The dissimilarity index measures the evenness of a distribution, with values ranging from 0 (complete integration) to 100 (complete segregation). It can be interpreted as the percentage of a minority group's population that would have to move to achieve full integration (Massey and Denton 1988).

Several points are noteworthy about the data for 2000. First, the overall reduction in concentrated poverty was indeed significant. While, as noted above, the poverty rate itself did not change much in 100 largest metropolitan areas the 1990s, the share of the poor living in high poverty areas dropped from 31 percent to 26 percent. Second, however, there were important differences across cities. Cleveland has the most severe levels of segregation and poverty concentration by every measure on the chart. With respect to poverty concentration, Providence was second highest, but interestingly enough, with respect to black/white segregation it is now by far the lowest. Dissimilarity index values with respect to poverty for Denver, Indianapolis, and Oakland are similar, but Denver's current population share of poor in high-poverty tracts (8 percent) is much below the 13 to 15 percent level of the other two.

With respect to site-by-site changes over the decade, although the differences are generally modest, the data show improvements by every measure for every site, except for the concentration of poverty in Providence, which deepened considerably. The most sizeable improvements were registered by Cleveland (all measures), Denver (particularly with respect to poverty concentration), and Providence (with respect to black/white segregation).

SPATIAL PATTERNS

The maps in figures 8.1 through 8.10⁽²⁷⁾provide orientations to the spatial patterns in the central areas of the five study sites. All show 1990-defined census tract boundaries, and are at a uniform scale: 1 inch = 3 miles. The first map for each city shows 1990 poverty rates for all census tracts. In almost all cities, high-poverty areas tend to be located in

a ring surrounding or adjacent to the central business district. An exception is metropolitan Oakland, where in addition to a high-poverty cluster next to the central district there are a number of other poverty concentrations (e.g., on the coast and around the university in Berkeley and stretching along major highways to the southeast of central Oakland).

Figure 8.1
Cuyahoga County, OH. Poverty Rate 1990

Figure 8.2
Cuyahoga County, OH. Predominant Race 1990

Figure 8.3
Denver County, CO. Poverty Rate 1990

Figure 8.4
Denver County, CO. Predominant Race 1990

Figure 8.5
Marion County (Indianapolis), IN. Poverty Rate 1990

Figure 8.6
Marion County (Indianapolis), IN. Predominant Race 1990

Figure 8.7
Alameda County, CA. Poverty Rate 1990

Figure 8.8
Alameda County, CA. Predominant Race 1990

Figure 8.9
Providence County, RI. Poverty Rate 1990

Figure 8.10
Providence County, RI. Predominant Race 1990

The second map in each set shows the predominant racial/ethnic group in each tract in 1990 overlaid against a dot pattern showing the location of high-poverty tracts (the predominant group accounts for 60 percent or more of the population of the tract).

• In Cleveland, there is a significant, but not complete, overlap of predominantly African-American population and high poverty to the east and southeast of the city center. High-poverty areas farther west are predominantly white.
• Denver has a more complex pattern. Predominantly Hispanic tracts are almost all in the high-poverty category, forming something like a semicircle to the northwest of the city center. There is a cluster of predominantly black neighborhoods just below the northern border to the east of the center, about half of which are high-poverty. There are also a large number of racially mixed tracts (no group with 60 percent or more) generally adjacent to the areas of Hispanic and black concentration.
• The pattern in Indianapolis is simpler: a large concentration of predominantly African-American neighborhoods to the north of the center (about half of which are high-poverty), just a few racially mixed neighborhoods (spatially scattered), and all the rest predominantly white (eight tracts are also high-poverty).
• Oakland's pattern reflects what we have already said about its racial/ethnic diversity: by far the largest number of racially mixed tracts, with two major concentrations that are predominantly black (one in northwest Oakland and one about five miles southeast of the city center). Almost all of the tracts directly east of Oakland and Berkeley are predominantly white, sitting on the other side of the hills inland from those two cities.
• The Providence area has no tracts with a predominant race other than white, but it has several racially mixed tracts, significantly overlapping the high-poverty clusters north and especially south of the city center. (A number of the racially mixed tracts to the south of the city center shown on this map had become predominantly Hispanic by 2000.)

NEIGHBORHOOD CONDITIONS AND TRENDS

The remainder of this section presents data on the changing characteristics of different types of neighborhoods in the central city or county in each metropolitan area. As noted in section 2, these central areas are defined as Cuyahoga County in metropolitan Cleveland, and the cities of Denver (same as Denver County), Indianapolis (same as Marion County), Oakland, and Providence.

For each site, we contrast conditions in high-poverty neighborhoods (those with 1990 poverty rates of 30 percent or more) with those in all other neighborhoods. This approach avoids a serious interpretation issue that often arises when citywide averages are compared. The conditions in City A may look worse than in City B only because the former has a higher proportion of high-poverty neighborhoods, while in fact conditions in the high-poverty neighborhoods themselves are better in A than B. We are most interested in finding out whether, and to what extent, conditions in high-poverty neighborhoods have improved.

It is important to note that we purposefully chose to focus on neighborhoods (tracts) that were in the high-poverty category in 1990 rather than the largely overlapping group that were in that category as of 2000. If we had chosen the latter we would have missed changes for a large number of tracts whose poverty rates dropped below the 30 percent threshold in the 1990s, a story that is critical to understanding the dynamics of change in that decade. We recognize that this choice omits tracts whose fortunes went in the other direction (i.e., whose poverty rates increased to a point above 30 percent). There is a considerably smaller number in that category, however. Their story is also worth telling, but we judge that doing so here would have added complexity and the benefits would not have been worth the cost.

Table 8.3 shows that there are indeed variations in the shares of population that live in high-poverty neighborhoods in these sites; a range in 2000 from only 7 percent in Indianapolis up to 34 percent in Providence, with the other three sites in the 15 to 18 percent range. Note that these shares had not changed much at this city/county level, in contrast to more notable declines in high-poverty shares at the metropolitan level.

Population Growth and Composition

Table 8.3 also makes it clear that minorities are concentrated in high-poverty neighborhoods in these sites. Across the five in 2000, on average, minorities account for 77 percent of the population in high-poverty tracts vs. 42 percent in other tracts. As was the case at the metropolitan level, blacks are the dominant minority in Cleveland and Indianapolis, Hispanics are more important in Denver and Providence, and Oakland stands out as the most mixed racially. Minorities accounted for 94 percent of Oakland's high-poverty area population, compared with a low of 66 percent in Indianapolis.

Perhaps the most dramatic change during the 1990s, however, was the growth of Hispanics and other nonblack minorities. Except for Cleveland and the nonpoor sections of Indianapolis and Providence, black shares of total population actually declined. The share in the Hispanic and other category, in contrast, increased everywhere (poor and nonpoor neighborhoods in all cities), but particularly in Providence, Oakland, and Denver. In high-poverty neighborhoods, this Hispanic and other share went up from 38 to 54 percent in Providence, from 39 to 52 percent in Oakland, and from 50 to 56 percent in Denver.

Only Denver, Oakland, and Providence have sizeable shares of foreign-born population. The foreign born are concentrated in high-poverty areas in these three cities, although their shares increased markedly in both high- and low-poverty areas in the 1990s.

As to overall population dynamics, high-poverty areas in Cleveland (Cuyahoga) underwent a 9 percent population loss in the 1990s, compared with no change in other parts of the county, and high-poverty neighborhoods in Indianapolis lost 10 percent, compared with a gain of 10 percent in the rest of the city. In contrast, high-poverty neighborhoods in the other three cities gained population (ranging from 7 percent in Providence to 20 percent in Denver) and did so at rates generally comparable to those for the rest of their cities.

Economic and Social Conditions

Table 8.4 identifies six indicators of economic and social well-being. The data for 2000 show that, as expected, conditions remained significantly more problematic in high-poverty areas than other neighborhoods with respect to every indicator in every one of our five sites. However, the data also show that in the 1990s, conditions in high-poverty areas had improved in 28 of 30 possible cases (six indicators times five sites), and the problem gap between high-poverty and other neighborhoods had diminished in 27 of the 30.

Among the sites, the high-poverty neighborhoods in Cleveland evidenced either the worst or next to the worst scores on five of the six measures, and Oakland was worst or next to the worst on four. Denver's high-poverty areas registered the least problematic conditions on five of the six measures.

• Poverty rates in high-poverty tracts declined on average from 39 percent to 34 percent over the decade (year 2000 range from 37 percent in Cleveland and Providence to 29 percent in Denver). The average poverty rate in high-poverty tracts for the five cities dropped from 3.4 to 2.7 times that for other tracts.
• Unemployment rates in high-poverty tracts dropped from 15 percent to 13 percent on average (year 2000 range from 16 percent in Cleveland and Oakland to 10 percent in Denver). The high-poverty average dropped from 2.4 to 2.2 times that for other areas.
• Employment rates (employed as percentage of population age 16 and over) in high-poverty tracts increased on average from 44 percent to 48 percent over the decade (year 2000 range from 43 percent in Indianapolis to 57 percent in Denver). The high-poverty average increased from 0.74 to 0.78 times that for other tracts.
• Adults without high school diplomas in high-poverty tracts dropped from 47 percent to 41 percent on average (year 2000 range from 45 percent in Oakland to 36 percent in Indianapolis). The high-poverty average dropped from 2.10 to 2.06 times that for other areas.

• Female-headed shares of all families with children in high-poverty tracts dropped from 56 percent to 50 percent on average (year 2000 range from 62 percent in Cleveland to 40 percent in Denver). The high-poverty average dropped from 1.9 to 1.6 times that for other areas.
• Shares of households receiving public assistance in high-poverty tracts declined on average from 27 to 13 percent over the decade (year 2000 range from 18 percent in Oakland to 6 percent in Denver). The high-poverty average dropped from 3.6 to 3.0 times that for other tracts.

Two indicators related to public assistance have been used in the cross-site analysis. The first one (referred to in the above paragraph and in table 8.4) is a census measure that reflects the percentage of households who said they received public assistance in April 2000. This is a fairly comprehensive measure defined to include income from AFDC/TANF, Supplemental Security Income (SSI), and General Assistance. It is a self-reported measure, however, and researchers have found that it often understates participation rates determined from administrative records. The second measure, AFDC/TANF recipiency, is obtained from administrative records and is an indicator of share of all population or households of an area participating in this specific program at a point-in-time.⁽²⁸⁾

NNIP partners generally have some information from administrative records on AFDC/TANF recipiency, but Cleveland is the only one of the five sites that has a consistently defined time series going back to the early 1990s (figure 8.11). The pattern in the Cleveland data tells the same basic story as the census public assistance measure. Recipiency rates decline dramatically over the decade, and they do so more rapidly in high-poverty neighborhoods than elsewhere. For high-poverty areas, for example, Cleveland's administrative data (percent of population receiving AFDC/TANF) show a rate of 31 percent in 1992, going down to 14 percent in 2000, whereas the census measure (percentage of households receiving public assistance) was 35 percent in 1990, going down to 17 percent in 2000. The AFDC/TANF measures show the path of change annually over the decade. Clearly, the decline in Cleveland started long before the passage of the welfare reform law in 1996, and may reflect earlier welfare reform efforts taking place in the state. The census measure for Cleveland showed the 2000 public assistance recipiency rate in high-poverty neighborhoods on average to be 5.7 times that of other neighborhoods, whereas the AFDC/TANF rate in high-poverty neighborhoods was 7.0 times that for other areas in that year.

Some consistently defined data for TANF recipiency based on population are available for two other sites near the end of the decade. The 1998-2000 TANF recipiency rates for Indianapolis were 8 percent on average for the high-poverty neighborhoods (exactly the same as the for census rate in 2000) and a 2 percent average for other neighborhoods (compared with 3 percent according to the census). The Denver TANF data do not match the census figures quite as well. They show a 2 percent recipiency rate for high-poverty neighborhoods in 1999 (compared with 6 percent by the census measure in 2000), and 1 percent for other neighborhoods (compared with 3 percent by the 2000 census measure).

For Providence, the welfare indicator available for this analysis was the percentage of households that received AFDC/TANF (not the percentage of population used in other sites), so it is not directly comparable with the other sites. In all three years available (1996, 1998, and 2000), the AFDC/TANF household rates for high-poverty areas are about 1.8 times the rates for nonpoor tracts. The high-poverty areas fell from 19 percent of households receiving AFDC/TANF in 1996 to 16 percent in 2000, with the nonpoor area rates dropping by the same amount (from 11 percent in 1996 to 9 percent in 2000). The census public assistance numbers are very similar (15 percent for high poverty areas and 8 percent for nonpoor areas).

Neighborhood Physical Conditions

Various types of environmental and other physical problems in a neighborhood can contribute directly to health problems. We do not have data on environmental hazards, but we do have some indirect indicators of neighborhood physical conditions that may be of influence (table 8.5).

The first indicator is the percentage of all occupied housing units that are overcrowded (with more than one person per room). Overcrowding rates are very low in America, a fact that is reflected in the data on table 8.5, but in all cases, overcrowding is substantially more prevalent in high-poverty areas than in other types of neighborhoods. The rates have also been increasing recently, particularly in regions with rapidly growing Hispanic populations. Indeed, the table shows that the three sites where that is occurring (Denver, Oakland, and Providence) have overcrowding rates that are both higher and increasing more rapidly than the other two sites. Overcrowding even in high-poverty tracts in Cleveland and Indianapolis is quite low (4 to 5 percent) and did not increase over the 1990s.

The remaining indicators on table 8.5 refer to the physical quality of housing. The first of these is the age of housing: specifically, the share of an area's housing units in 1990 that had been built 30 years or more before. A higher share of housing in high-poverty areas is in this category than is found in higher-income areas in all sites except Providence.

The other two indicators, average value of owner-occupied homes (as reported on the census) and average value of home mortgages originated, are more current. The first measure represents the self-reported value of all single-family owner-occupied homes, while the second measures the mortgage loan amount borrowed for home purchases. However, the two corroborate each other on some interesting circumstances and trends. First, in both cases, the values almost everywhere are lower in high-poverty areas (on average in 2000/2001, high-poverty area home values were 67 percent of those in other neighborhoods, and mortgage values were 76 percent of those in other neighborhoods).⁽²⁹⁾

Second, however, both measures had increased more rapidly in the high-poverty neighborhoods than they had elsewhere in all of these cities (except Providence) in the 1990s. In these four cities on average, home values increased by 56 percent over the decade in the high-poverty areas compared with 31 percent in other areas.

These data all reinforce our notion of the differences between the two older sites in the Midwest (Cleveland and Indianapolis--2000 average home values of $70,000 to $73,000 in high-poverty areas) and the other three that have been experiencing more growth in general and more immigration in particular. The high-poverty neighborhood averages were $157,000 in Oakland, $149,000 in Denver, and $106,000 in Providence.

Providence is an outlier by these measures. In that site only, home values in the 1990s were higher in high-poverty neighborhoods than in the rest of the city, and values declined over the decade in all areas.⁽³⁰⁾

Neighborhood Conditions - Social Stressors

As noted in section 7, the main indicators in this area relate to crime in the neighborhood, particularly violent crime. Figure 8.12 shows rates of reported violent crime for all five cities in the 1990s, and figure 8.13 does the same for reported property crime (in each, the top panels show the average rates for high-poverty tracts and the lower panels show them for other tracts). The series cover the entire decade in Cleveland and Denver, but consistently defined data are available only for eight years (1992-2000) in Indianapolis and four years (1996-2000) in Oakland. Only one year of data was available for Providence census tracts, so it is not included in the figures.

The levels are very different from each other. The annual averages across years for violent crime in high-poverty neighborhoods range from 8 per thousand population (Providence) to 34 per thousand (Indianapolis). In lower poverty neighborhoods, the averages range from 2 percent (Cleveland) to 16 percent (Oakland). For property crimes in high-poverty neighborhoods, annual averages across years range from 39 per thousand population (Oakland) to 101 per thousand (Indianapolis). In low-poverty neighborhoods, the averages range from 14 percent (Cleveland) to 66 percent (Oakland).

The graphs evidence one surprising result. In Oakland, for both violent and property crimes, the rates in low-poverty neighborhoods are consistently well above those in high-poverty neighborhoods. We questioned this outcome, but after checking, we found no reason to doubt the data, although we have found no explanation for it so far.

For the other four cities, however, the results are as expected. Crime rates are always higher in high-poverty neighborhoods on average than in the other parts of these cities. For violent crime, the gap is greatest in Cleveland, where the high-poverty average is 8.9 times that for other areas, and smallest in Providence (1.2). For property crimes, the same cities mark the extremes. The high-poverty average in Cleveland is 4.9 times that for other areas, while in Providence it is again 1.2 times higher.

One other finding is consistent across all four cities with trend data, for both types of areas within cities and for both types of crimes: Reported crime rates declined significantly in the 1990s. There were notable variations in the speed of the decline, however. The most rapid declines were in Oakland, with violent crime dropping at 14 to 16 percent per year and property crime at 12 percent per year in both types of areas (over the four-year period for which data are available). The slowest declines were in Indianapolis, with violent crime declining annually by 2 percent and property crime by 3 to 4 percent (eight years of data, again with little difference between high- and low-poverty areas).

Denver is the only city for which annual rates of decline were consistently higher in high-poverty areas than in the rest of the city: 5.8 vs. 2.2 percent for violent crime, and 10.2 vs. 4.1 percent for property crime (data span a 10-year period).

Neighborhood Stability and Social Networks

All of the measures on table 8.6 are crude proxies for the characteristics Ellen et al (2002) were interested in for this category, but they do offer additional insight into differences within and between cities. Overall, they suggest that high-poverty neighborhoods are notably less stable than other areas in these cities and that the degree to which this is so did not change much over the 1990s.

• The share of households that had moved at least once in the five years before the 2000 census was higher in high-poverty neighborhoods than in the rest of the city in all sites; average 56 percent vs. 49 percent (same numbers as in 1990). Rates in the rapidly growing cities were higher than in the others: 63 percent in Denver's high-poverty neighborhoods vs. 46 percent in Cleveland's.
• An area made up of homeowners is generally considered to be more stable than one dominated by renters. On average across sites in 2000, renters accounted for 70 percent of all households in high-poverty neighborhoods vs. 45 percent elsewhere. Variation across sites was not marked, nor was change in these relationships over the decade.
• Low rental vacancy rates indicate tight housing markets and, normally, problems with housing affordability for low-income groups. Here, there are more differences across places and time. In the two older Midwest cities (Cleveland and Indianapolis), vacant rentals are much more prevalent in high-poverty areas, and this relationship remained much the same from 1990 to 2000. In the other cities, rental markets in high-poverty neighborhoods got much tighter in the 1990s, and their vacancy rates at the end of the decade were close to those in other parts of the city.

The final measure on the table is the number of home purchase mortgages in each area per 1,000 units. Again, Cleveland and Indianapolis (and in this case, Providence) evidence a pattern typical of older cities: much lower rates in high-poverty areas than in the rest of the city, where markets are more active. However, in both types of areas in all cities, this measure was going up significantly in the 1990s. On average, comparing the 1995/96 rate with the one for 2000/01, it increased from 14 to 27 in high-poverty areas and from 29 to 39 elsewhere.

SUMMARY

Whereas trends for America's cities in the 1980s seemed almost uniformly distressing, the 1990s tell a very different story. City populations generally increased more (or declined less) than they did in the 1980s, and the changes were not all confined to the better parts of town. Population declines were stemmed in a number of high-poverty areas, and growth returned to some (Kingsley and Pettit 2002).

It appears that there was a notable overall deconcentration of poverty in most, although not all, urban regions. In the 100 largest metropolitan areas, the share of the poor living in high-poverty areas declined markedly. Also, a number of other studies have shown that key social indicators (e.g., crime rates) improved substantially in many cities during the latter part of the decade, and we find similar trends in other census indicators in this analysis.

Overall, however, it seems likely that the 1990s will be known more for diversity of outcomes than uniform improvement, with many cities and neighborhoods within cities still declining as others have begun to turn around. It is clear that racial and ethnic diversity have increased substantially in urban America, and these changes will also surely have had effects on determinants of health at the neighborhood level.

In addition, it is important to remember that the reference data of the recent decennial census (April 2000) was near the peak of the economic boom that began in the mid-1990s. Circumstances may well have deteriorated since then. Nonetheless, even given that prospect, the review in this section represents a marked contrast to the almost universally bleak assessments that emanated from reviews of trends in American cities a decade ago.

[Go To Contents]

Section 9

NEIGHBORHOOD HEALTH TRENDS

INTRODUCTION

The previous section reviewed the characteristics of the regions surrounding our five study sites, as well as key demographic, social, and economic conditions and trends of the sites themselves. In this section, we independently examine trends in the health related variables that are available for this analysis: the birth and mortality indicators derived from vital records files maintained by the NNIP partners in the sites. Specifically, we examine trends for five indicators: teen birth rates, rates of early prenatal care, rates of low-birth weight births, infant mortality rates, and age-adjusted mortality rates. As in the previous section, we also contrast conditions and trends in high-poverty tracts (poverty rates of 30 percent or more in 1990) with those in nonpoor tracts in each site.

As noted in section 7, prior studies have shown that health-related problems measured by these rates are generally more severe in high-poverty neighborhoods than in nonpoor areas, but these studies have typically covered only one city and dealt with different time periods. Our data allow us to go farther and examine variations in the extent of these gaps and how they have shifted over the same time period in several different cities. Also, data have already been published to show that conditions as measured by these indicators improved in many American cities in the 1990s. This analysis is the first, however, to quantify and compare the extent of the improvements between poor and nonpoor neighborhoods within cities and between cities.

TRENDS FOR KEY INDICATORS

For this analysis, to address the rare events issue noted earlier, we averaged three years of data from 1990 to 2000 to smooth out the annual variations that could occur due to small numbers of events. For simplicity's sake, the text will refer to rates by the start and end points of the data. For example, the 1990/1992 rate discussed below refers to the rate obtained by averaging 1990, 1991, and 1992 data. For three of the cities, the data were available for the full time period from 1990 to 2000. As stated in section 7, there were two exceptions. First, Providence had no mortality data and birth data only from 1995 to 2000. Second, at the time of this analysis, Oakland only had mortality up until 1999.

Births and birth rates

For each of the cities, the high-poverty areas account for very different shares of all births in 1998/2000--from a low of 9 percent in Indianapolis to 37 percent in Providence. With the exception of Providence, birth rates in the non-poor areas of our cities declined in the 1990s. The birth rates in high poverty neighborhoods also fell, but at 2 to 4 times faster than the non-poor rates.

To provide context for our analysis of trends for the five indicators, particularly those related to maternal and infant outcomes, it should be helpful to review the characteristics of the births in each city overall. As discussed in the previous section, these study areas vary greatly in population size, and this pattern carries over in the number of births (table 9.1). However, the changes in births from 1990 to 2000 did not always track with the trends in population. In the high-poverty tracts in Cleveland and Indianapolis, the number of births dropped three times faster than the general population. The percentage rise in births in high-poverty areas in Denver and Providence generally tracks the increasing total population. Oakland has the most unusual pattern--moderate growth in population with large decreases in the number of births.

The analysis to follow discusses the aggregate indicators for high-poverty tracts and the nonpoor tracts. In 1998/2000, for each of the cities, the high-poverty areas account for very different shares of all births--from a low of 9 percent in Indianapolis to 37 percent in Providence. As expected from the racial change described in section 8, the racial/ethnic composition of births altered markedly over the decade in most of the cities. The share of Hispanic births increased in all of the cities, though it still remained low in Cleveland and Indianapolis. In Denver and Oakland, there was a corresponding loss of share for births to black mothers.

For the four cities with data for the full decade, birth rates in high-poverty areas in all the cities were higher than in the nonpoor areas throughout the decade (figures 9.1a and 9.1b). Of the high-poverty areas, Denver ended the decade with the highest rate (23 births per 1,000 population), and Providence ended with the lowest (16 births per 1,000 population).

With the exception of Providence, birth rates (births per 1,000 population) in the nonpoor areas of our cities declined in the 1990s (see figure 9.1a). The birth rates in high-poverty neighborhoods also fell, and at rates two to four times faster than in the nonpoor areas. Even Providence, with birth rate increases in the nonpoor areas, experienced a slight drop in its high-poverty areas from 1995/1997 to 1998/2000. While the rates in high-poverty areas were consistently higher than in the nonpoor areas, the patterns of change generally resulted in much smaller differentials between the two types of areas by 1998/2000.

Teen birth rates

Teen birth rates fell in both the poor and nonpoor areas in four cities in the 1990s, with the most substantial decreases in both types of areas in Oakland. Even with the decreases, considerable disparities between poor and nonpoor neighborhoods remain in Cleveland, Denver, and Indianapolis.

As shown in figure 9.2a, only the nonpoor areas of Denver and Oakland had 1990/1992 teen birth rates far above the national average of six births per 100 girls aged 15 to 19. Starting from this high level, the rates in Oakland's nonpoor areas showed a strong decline, falling twice as fast as the national average (see annex table C.15 for details). The nonpoor areas in the other cities also saw declines, but at a much slower rate. By the end of the decade, only the teen birth rate for the nonpoor areas in Denver (7.5) remained well above the national average.

In 1990/1992, the teen birth rates in the high-poverty areas for four of our cities were two to three times the national average (figure 9.2b). In Oakland, the teen birth rate in the poor areas 1990/1992 was 14 percent, lowest of the rates in that year. It fell 6 percentage points over the decade, with the majority of the gains in the first half of the decade. Figure 9.3 shows the low rates at the end of the decade, with rates in poor areas very similar to rates in nonpoor areas.

The high-poverty areas in Indianapolis had the highest rate in 1990/1992 (19 percent), but they also showed the most improvement--dropping 7 points to end at the second highest rate. This progress cut the difference between poor and nonpoor areas in half--from 12 points at the beginning of the decade to 6 points at its end. In Cleveland's high-poverty tracts, the teen birth rate fell midway between the sites and had a sharp decline like that of Oakland and Indianapolis. However, Cleveland still had the widest disparity in rates, with the 1998/2000 rate in high-poverty neighborhoods (10 percent) triple the rate of the low-poverty ones. The reduction in teen birth rates was primarily due to reductions in births to black teens. The African-American teen birth rates dropped from 2 to 9 points in high-poverty areas, while Hispanic teen birth rates stayed the same in Cleveland, increased in Indianapolis, and fell only 1 to 3 points in the remaining three cities (see annex tables C.17 to C.18 for details).

High-poverty areas in Denver (with very high rates) and Providence (with very low rates) did not experience reductions as large as the cities discussed above. Denver had the highest overall teen birth rate for most of the 10 years, surpassing Indianapolis early in the decade. Figure 9.4 clearly shows the overlap of the extreme teen birth rates with high-poverty areas in the western half of the city. This is also the predominantly Hispanic area--in 1998/2000 almost one in five Hispanic teen girls in Denver became mothers. Providence also had a very small drop over the 1995/1997 to 1998/2000 time period. This trend is not as troubling as in Denver, since the rates in Providence are remarkably low for all races and the time period covered is shorter. The Providence teen birth rates in the high-poverty areas were at the U.S. average, and the difference between the poor and nonpoor areas was less than 1 percent.

Figure 9.3
Alameda County, CA. Teen Birth Rates 1998-2000

Figure 9.4
Denver County, CO. Teen Birth Rates 1998-2000

Early prenatal care

Except for Providence, poor and nonpoor areas in all cities showed improvements in prenatal care rates. Indianapolis and Oakland stand out, with impressive expansion of early prenatal care to high-poverty areas.

Nationally, great gains were made in providing prenatal care to mothers. In 2001, 83 percent of pregnant women received prenatal care in the first trimester of pregnancy, up 7 percentage points from 1990/1992. At the beginning of the decade in nonpoor areas, only the Cleveland figure was significantly above the U.S. rate of 76 percent (figure 9.5a). For high-poverty areas, the rates in Cleveland and Oakland approached the national rate in 1990/1992, with the other three cities much farther behind (figure 9.5b). From these starting points, the improvements seen in the U.S. average are not evident in all of our sites.

In three of our cities, the change in levels of prenatal care may well be linked to specific program initiatives occurring during the 1990s. Beginning in 1991, Oakland was a demonstration site for the Healthy Start initiative, a federal program aimed at reducing infant mortality rates and generally improving maternal and infant health in at-risk communities.⁽³¹⁾ In the city's high-poverty tracts, the prenatal care rate improved remarkably over the 1990s, moving up more than 13 percentage points in high-poverty areas to end at 85 percent--the highest rate of all the cities. Figure 9.6 shows that the declines spread across the city. The gap in rates between poor and nonpoor neighborhoods was reduced to 3 percentage points by 2000. In addition to being spatially dispersed, the increases occurred for all races. The 9-point increase in the Hispanic early prenatal care rate is particularly impressive since those rates for Hispanics fell in the other four cities (annex table C.18).

Figure 9.5a:
Early Prenatal Care Rates in Non-poor Tracts

Figure 9.5b:
Early Prenatal Care Rates in High Poverty Tracts

Figure 9.6
Alameda County, CA. Change in Early Prenatal Care Rates 1990-2000

In Indianapolis, the Campaign for Healthy Babies was formed in 1989 as a public-private partnership to develop resources and strategies to reduce infant mortality. It stressed the need to increase the percentage of women receiving adequate prenatal care. The campaign organizers placed particular geographic focus on areas with the worst infant mortality rates. The official campaign ended in 1992, although the Marion Health and Hospital Association continued with a smaller scale effort. During the period of the campaign, early prenatal care rates improved for poor neighborhoods by 6 percentage points. The high-poverty area rates then leveled off for the remainder of the decade to end at 66 percent. This was still 13 points below the rates in nonpoor areas, though progress was made in closing the gap. Figure 9.7 displays the moderate and large declines in most of the high-poverty tracts, with more mixed results outside the core city.

Cleveland/Cuyahoga County began with the highest rate for nonpoor areas and continued to improve over decade, ending at 90 percent. From 1990/1992, the rate for high-poverty tracts went up 4 percentage points to reach 77 percent (the second highest level).⁽³²⁾ African-American rates showed the most progress, up 8 points in both high- and low-poverty areas. Black mothers still fare better in non-poor tracts, with an early prenatal care rate 6 points higher than in high-poverty tracts (annex table C.17). In conversations with local experts, we learned of two program initiatives that could be linked to the improvements. In one, Ohio made a strong effort in the late 1990s to increase access to Medicaid coverage, particularly for pregnant women and children. Second, Cleveland has been a Healthy Start site, and is cited as one of its "success stories." ⁽³³⁾ Cleveland's comprehensive program targets neighborhoods with high rates of infant deaths, and has built-in mechanisms for community involvement in the programs.

In the remaining two cities, local sources cited the increasing Hispanic and immigrant populations as the most likely drivers of the change in prenatal care rates. In Denver's nonpoor areas, early prenatal care declines in the beginning of the decade were reversed to reach a high of 80 percent in 1996/1998. However, falling rates in 1997/1999 and 1998/2000 eroded the progress. The high-poverty areas followed a similar trend at a lower level, beginning at 60 percent in 1990/1992 and ending at 63 percent in 1998/2000. In addition to the racial and ethnic changes, our local sources mentioned two other trends as possible contributing factors to the decrease in early initiation of prenatal care. First, cutbacks in Medicaid reimbursement caused

Figure 9.7
Indianapolis, IN. Changes in Early Orenatal Care 1990-2000

most private providers to either drop out of the Medicaid program or refuse to accept any additional Medicaid patients. The result was fewer health facilities for Medicaid patients and longer waiting lists at public clinics. Second, immigrant women may use more nontraditional health providers either from preference or because they fear that the use of public services or systems may result in exposure or deportation.

While we do not have data on early 1990s trends for Providence, toward the end of the decade it had the lowest rate of early prenatal care in nonpoor areas of all the cities, with only three-quarters of women receiving early care. The reasons for the relatively low level are not completely clear, but local sources speculated on two potential explanations. Like Denver, Providence had a growing immigrant population during the 1990s, who may be less likely to use the formal health care system for reasons described above. Local sources also spoke of the dismantling of the state-level public health system, resulting in an increased reliance on often-overburdened nonprofit providers.

Low-birth weight births

In three of the cities (Cleveland, Denver, and Oakland), rates in high-poverty neighborhoods fell over the decade, while the national trend of low birth weight rates was generally flat. The rates in both poor and nonpoor areas in Indianapolis and Providence, in contrast, are on the upswing.

Figure 9.8a shows that the earlier rates for low-birth weight births in nonpoor areas fell into two clusters: one at or below the national average of 7 percent of all births at a weight of less than 2,500 grams (Cleveland, Indianapolis, Providence) and another above 9 percent (Denver, Oakland). The nonpoor rates in Cleveland and Denver mirrored the national trend with slight increases, but Indianapolis and Providence rates (though only from 1995) rose significantly. Of the nonpoor areas, only Oakland saw decreases over the decade, moving into the cluster around the national average.

The earlier rates of low-birth weight births in high-poverty areas ranged from a low of 10.8 percent in Oakland to a high of 14.7 percent in Cleveland (see figure 9.8b). In three of the cities (Cleveland, Denver, and Oakland), the rates declined, contrary to the national trend. The rates went up in Indianapolis and Providence by 0.7 and 1.4 percentage points, respectively.⁽³⁴⁾

Figure 9.8a.
Low Birth Weight Rates in Non-poor Tracts

Figure 9.8b.
Low Birth Weight Rates in High Poverty Tracts

While the low birth weight rates are generally higher in high-poverty areas, the size of the disparity varied significantly by city. Cleveland had the largest disparity in 1990/1992 (7 points), but has cut the difference in half over the past 10 years. Figure 9.9 illustrates the additional understanding of conditions that maps can provide. The highest levels of low birth weight are not prevalent across all poverty areas, but are concentrated in the eastern side of the central city (the predominantly African-American section). Denver began the decade with less inequity than Cleveland, but it too has reduced the gaps by more than 50 percent.

Despite differing rates and trends, Oakland and Providence are similar in that they have very small gaps between high- and low-poverty areas by this measure--about 1 percentage point. Figure 9.10 demonstrates that the greater equity does not necessarily imply greater well-being for mothers in Providence. The low-birth weight rates of several, mostly nonpoor tracts are in the higher ranges in the late 1990's.

Over time and across sites, the African-American low-birth weight rates were generally twice the Hispanic rates. Like the prenatal care rates, the black low-birth weight rates were better in nonpoor than poor areas, but the neighborhood seems to make less of a difference in this indicator. For Hispanics, hardly any difference in rates exists between poor and nonpoor tracts (see annex tables C.17 and C.18).

Infant mortality rates

Infant mortality rates (deaths of infants age 0-12 months per 1,000 live births in that year⁽³⁵⁾) dropped in both the poor and nonpoor tracts in four of the cities, generally at a faster pace than the nation (data were not available for Providence). However, the 1998/2000 rates in the high-poverty neighborhoods in Cleveland and Indianapolis were still double the national rate.

Because of the smaller number of events when examining infant deaths, the trends in infant mortality rates are more erratic than the birth and total mortality indicators. Nonetheless, the trends in all the low-poverty areas and in the high-poverty areas of three of the cities appear stable enough to warrant some conclusions.⁽³⁶⁾

Figure 9.9.
Cuyahoga County, OH. Low Birthweight Rates 1998-2000

Figure 9.10.
Providence, RI. Low Birthweight Rates 1998-2000

The infant mortality rate is defined here as the number of deaths of infants 0-12 months as a percentage of total live births in the same year. For the nonpoor areas in 1998/2000, raw rates ranged from a rate of 6.3 infant deaths per 1000 live births (Denver) to a rate of 9.1 (Indianapolis). The rates dropped 2 to 4 points over the decade, in line with the decrease in the national rate. By 1998/2000, only the rate for the Indianapolis nonpoor areas was substantially above the national rate of 7 infant deaths per 1,000 births.

In 1990/1992, the infant mortality rates in high-poverty areas were two to three times the national rate, but all cities made progress in the 1990s. Denver's high-poverty areas experienced the greatest improvements in infant mortality rates, down almost 4 infant deaths per 1000 births from 1990/1992 to 1998/2000. The 1998/2000 rate even approaches the national average. The high-poverty areas in Cleveland showed the highest rates through most of the decade, ending at 16 infant deaths per 1,000 births. The infant mortality rate in high-poverty tracts in Indianapolis was generally lower than in Cleveland, but still well above the national average of 7.5. Like Denver, both Indianapolis and Cleveland made advances in the 1990s, with respective rate declines of 3.8 and 5.4.

Age-adjusted mortality rates

The trends in age-adjusted mortality rates for nonpoor areas were inconsistent for the four sites, with two cities showing small declines and two showing virtually no change. With the exception of Indianapolis, the high-poverty areas saw some decrease in age-adjusted death rates.

By using age-adjusted death rates⁽³⁷⁾ instead of crude death rates, this analysis can compare rates across time and place while controlling for the age distribution of each area. In this way, a city's rate will not be higher just because more elderly people reside there.

In the nonpoor areas of our four analysis cities, the 1990/1992 age-adjusted death rates ranged from 810 deaths per 100,000 population in Indianapolis to a high of 960 in Oakland, with only Oakland above the national rate of 930 (figure 9.12a). While the national rates declined slightly from 1990/1992 to 1998/2000, the trends in the cities were less consistent. In general, they increased in the first half of the decade, decreased for the next couple of years, and then turned back upward. Oakland's nonpoor areas are the exception, with a steady decline for most of the decade. For all the nonpoor areas, the trends resulted in very little change in rates from start to finish.

Figure 9.11a:
Infant Mortality in Non-poor Tracts

Figure 9.11b:
Infant Mortality in High Poverty Tracts

The rates in the high-poverty areas are all higher than the rates for nonpoor areas. The high-poverty rates were clustered together in 1990/1992--from 1,160 deaths per 100,000 in Oakland to 1,280 in Cleveland--but paths diverged over the 10 years. By the end of the decade, the high-poverty rates in Oakland and Denver slipped 137 and 151 deaths per 100,000, respectively. Oakland's rate ended the lowest at 1,000 deaths per 100,000 population. The high-poverty death rate in Indianapolis increased somewhat--up 71 from 1990/1992 to the highest rate of 1,280 per 100,000 population in 1998/2000. The Cleveland pattern is the least consistent--flat until 1993/1995, decreasing to 1996/1998, and then moving up again. The rate ends the decade near the starting point.

Cleveland and Indianapolis end the decade with gaps of 360 and 450, respectively, between the poor and nonpoor rates. These rates represent improvement for Cleveland and a setback for Indianapolis. Figure 9.13 shows the high-mortality-rate areas in Indianapolis clustered in and around the city's poorest areas. In addition to having lower rates for high-poverty areas than nonpoor ones, Oakland and Denver also have smaller differentials at the end of the decade in mortality rates between those two types of areas--100 and 270 deaths per 100,000, respectively.

Figure 9.13.
Indianapolis, IN. Age Ajusted Death Rates 1998-2000

SUMMARY AND IMPLICATIONS

The trends we have reviewed in this section are complex. However, the most important findings can be summarized under three points as follows:

1. Gaps between high-poverty neighborhoods and others by the indicators we reviewed were indeed substantial in the early 1990s, with health-related problems in high-poverty neighborhoods more severe for almost all indicators in all cities. However, the extent of the gaps varied. The differences in low-birth weight and mortality rates were much more pronounced in Cleveland and Indianapolis (where African-Americans are the dominant minority) than in the more racially diverse cities. However, for early prenatal care rates and teen birth rates, the disparities in Denver rose to the levels of Cleveland and Indianapolis.
2. In almost all cities, the 1990s saw notable improvements in the maternal and infant health indicators we have examined, in both the high-poverty and the nonpoor neighborhoods, parallel to the findings about contextual conditions in section 8. In fact, the rates of improvement in the health-related indicators were generally faster in the high-poverty neighborhoods than in the other parts of these cities. Nonetheless, these differences were not enough to eliminate the gaps between these two types of areas by the end of the decade.
3. Still, there were important variations in the rates of improvement. In some cases, it appears on the surface that the change was influenced largely by the city's racial composition. For example, the teen birth rate for African Americans dropped faster than for Hispanics, so high-poverty areas that were predominantly African America, such as those in Cleveland, experienced more rapid declines. In other cases, it is hard to explain the differences without taking programmatic efforts into account. For example, Oakland, which had a highly regarded Healthy Start initiative in the 1990s, experienced a rate of improvement in prenatal care in its high-poverty areas much above those in the other sites.

Our findings so far are suggestive. Section 8 showed that a broad range of neighborhood conditions, grouped in the explanatory categories of Ellen et al (2001), generally improved in the high poverty areas and the other parts of our study sites in the 1990s. This section showed similar trends and relationships for health related indicators. We have also identified some interesting variations from the general trends and have offered a few speculations about possible causes. None of this pins down relationships between the variables in a reliable way, however. To do that, we need to conduct rigorous bivariate and multivariate correlation analysis of these indicators in a more spatially disaggregated form. This is the task of the next section.

[Go To Contents]

Section 10

ECOLOGICAL CORRELATIONS

INTRODUCTION

As noted, this section formally analyzes the relationships between the contextual indicators (selections from the neighborhood conditions introduced in section 8) and health-related indicators (reviewed in section 9). We first test the hypotheses set out in section 7 by examining the bivariate correlations of contextual characteristics with four of the health indicators from section 9: teen birth rates, early prenatal care rates, low birth weight birth rates, and age-adjusted death rates.⁽³⁸⁾ This analysis is based on three-year averages for these indicators as in the last section, but the indicators are calculated for individual census tracts as the geographic unit of analysis instead of the high-poverty/nonpoor groupings used previously. We review the results for all the cities together and for each city separately.

We next move to multivariate analysis with a select set of contextual variables to test the independent relationship between a selected set of contextual indicators and the four health indicators listed above. The multivariate analysis reveals three things: (1) the independent association between the dependent health variable and the tract condition; (2) how much of the variation of each health indicator is explained by the five racial and socioeconomic factors versus other conditions in the city and time period; and (3) whether some of the key shifts in health indicators are statistically significant or due to random fluctuation of small numbers of events.

BIVARIATE CORRELATION ANALYSIS

The bivariate correlation analysis is organized around the hypotheses that were introduced in section 7. Under each static hypothesis, we discuss the aggregate and site-specific results from the data. As noted, we use the three-year rolling average approach to calculate values for the health variables, but we calculated them for all individual census tracts in all study sites. Each overlapping three-year average health indicator rate for each census tract is an observation. This method yielded a total of over 8,400 observations across the five sites.⁽³⁹⁾

For the contextual indicators derived from the census, we interpolated annual estimates for all tracts based on 1990 to 2000 trends. The census indicator for a given year was related to the value of the health indicators as of the mid-point of the three-year period employed for them. For example, each 1991/1993 health indicator value is paired with the 1992 estimated poverty rate, percentage female-headed households, etc. To test our dynamic hypotheses, we compare the correlation coefficients for two time periods: 1990 to 1995 and 1995 to 2000.

STATIC HYPOTHESES

Socioeconomic conditions

Hypothesis: Census tracts with a majority non-white population and higher levels of immigrants will have higher levels of mortality and poor maternal and infant health outcomes than majority white census tracts.

Our first hypothesis addresses race, ethnicity, and nativity. As shown in table 10.1, the aggregate results are all significant in the expected directions for the overall percentage minority and the African-American and Hispanic percentages of the population. However, when looking at the individual city correlations, the share of the population that is Hispanic significantly correlates for low-birth weight rates only in Cleveland. The remaining results are mostly consistent for all cities and indicators, with a few notable exceptions. (For the complete list of city-specific correlations, see annex table C.23.) In Providence, the relationships between low-birth weight rates and percentages minority and African-American are slightly positive but not significantly correlated. In Oakland, teen births and the percentage African-American were not significantly associated.

Relationships with the percentage foreign born are more complicated. Overall, high percentages of foreign-born population are associated with worse prenatal care and teen birth rate outcomes as expected, but with better low-birth-rates.⁽⁴⁰⁾ The correlation with age-adjusted death rates was not significant. On closer examination of the city-specific correlations,

the relationships with health indicators are split. Cleveland and Indianapolis show better outcomes in tracts with higher immigrant levels, but the immigrant areas in the three more Latino cities have worse birth and death indicators.

Hypothesis: Low-income census tracts, as measured by poverty rate, median income, and public assistance and AFDC/TANF recipiency, will be associated with poorer scores on the mortality and the maternal and infant health measures than higher income tracts.

The next group of characteristics describes the economic conditions of the neighborhood. As many studies have found personal economic status to be linked with birth and mortality outcomes, it is not surprising that the aggregate correlations show the same relationships at the census tract level. These correlations are overwhelmingly strong and consistent across cities and measures. As with the racial variables, low birth weight in Providence and teen birth rates in Oakland are the two exceptions.

Hypothesis: Census tracts with higher social risk factors, as measured by lower education, low employment rates, and higher shares of female-headed families, will have worse scores on the mortality and the maternal and infant health measures than those with lower values.

In aggregate, all of the social risk factors--low levels of education and employment, and higher percentages of female-headed households with children--were significant and in the hypothesized direction. At the city level, the indicators were not always significantly related to low birth weight in Providence or teen birth rates in Oakland.

Physical stressors

Census tracts with poor housing quality, as measured by older housing, overcrowded units, and lower home values, will have higher levels of mortality and worse maternal and infant health outcomes than stronger housing markets.

For the next group of contextual variables, we use age of housing, overcrowded conditions, and home values as proxies for physical housing quality. Table 10.2 shows that indicators of better physical housing conditions are generally associated with better maternal and mortality outcomes as hypothesized, though the relationship is weaker than we saw with the socioeconomic indicators. Looking at city-by-city correlations (annex C.23) the only notable findings inconsistent with the hypothesis were that older housing was significantly related to better maternal and infant outcomes in Oakland and Providence.

Social stressors

Hypothesis: Census tracts with high total, violent, or property crime rates will have poorer scores on the mortality and the maternal and infant health measures than safer communities.

As in section 8, we used data on crime rates provided by our local partners as a proxy for social stressors in neighborhoods. Table 10.3 shows that census tracts with higher levels of crime generally have worse maternal and mortality outcomes as hypothesized, but that relationship did not hold for early prenatal care rates. The correlations are much stronger for teen birth rates and age-adjusted death rates than for low-birth weight rates. In Oakland, there is a remarkable 93 percent correlation between teen birth rates and violent crime rates, although the results for low-birth weight rates are insignificant. Although the aggregate calculation showed no association between crime and prenatal care rates, higher crime rates do have a substantial and negative association with prenatal care rates in Indianapolis.

Social networks

Hypothesis: Census tracts with less stable populations, as measured by renter occupancy, vacancy rate, and mobility rate, will have higher levels of mortality and worse maternal and infant health outcomes than stronger and more stable markets.

In this grouping, we select crude proxies for social networks, with the idea that less stable neighborhoods will have less opportunity for neighborhood cohesion than more stable ones. Overall, the correlations between the health indicators and our mobility variables confirm the hypothesis (see table 10.4). City-by-city results are not always consistent, however (annex C.3). The expected inverse relationship between renter occupancy and positive health outcomes holds generally, but the results for vacancy rate and percentage moved vary across cities and health indicators. Only the vacancy rates and percentage moved in Cleveland and Denver are consistently associated with worse health effects.

Hypothesis: Places with less change in total or minority population or a higher rate of home improvement or refinancing loans will have better mortality and birth outcomes.

The correlations for all three of these contextual indicators representing the social network category are in the opposite direction of the hypothesis, with more population change and fewer home improvement loans associated with better outcomes (see  table 10.4). Gentrification is one possible explanation for the better maternal outcomes associated with greater population change, though more research would need to be done to confirm this. The separate city correlations are often inconsistent for the percentage population change and the rate of home improvement, with some positive and some negative values. However, larger percentage change in the minority population relates to better indicators for all cities except Providence.

DYNAMIC HYPOTHESES

Hypothesis: The correlation between high-minority tracts and poor birth and mortality outcomes will remain positive, but will have decreased over the 1990s.

Consistent with the hypothesis, table 10.5 shows that the relationships between percent minority and early prenatal care levels and low birth weight rates weakened from the 1990-1995 to the 1995-2000 period but still continued to be significant and substantial. We see that for low-birth- weight rates, the strength of the relationships fell for both percentage African American and percentage Hispanic, while the declines in association with low early prenatal care rates occurred only for the percentage African American. For overall percentage minority and percentage African American, the correlation coefficients for age-adjusted deaths changed very little from the beginning to the end of the decade. There was a decrease, however, in the magnitude of the association with the percentage Hispanic.

Contrary to our hypotheses, however, in the late 1990s teen births appear to be more concentrated in minority (both black and Hispanic) areas than in the early 1990s. The pattern of much higher correlations occurs with the vast majority of our indicators and for all cities. After the tremendous progress seen over the decade in this indicator, we suspect that the places remaining with high teen births tend to be in the more segregated and distressed areas.

Hypothesis: The correlation between low-income tracts and births with late or no prenatal care will remain positive, but will have decreased over the 1990s. The correlation between low-income tracts and high rates of low-birth weight births will remain positive, but will have decreased over the 1990s.

Consistent with the hypothesis, lower early prenatal care rates are slightly less correlated with the income variables in 1990-1995 than in 1995-2000. More striking, the association between low-birth weight rates and tract economic conditions has diminished considerably over the 1990s. This is likely due to several interacting factors, including (1) advanced medical technology enabling an increasing number of low-birth weight infants born in high-poverty and nonpoor areas to survive, (2) the increase of low-birth weight infants due to women having children at later ages, and (3) reductions in low-birth weight infants in poor areas due to better and earlier prenatal care.

MULTIVARIATE ANALYSIS

Purposes

Once the bivariate analysis confirmed most of our hypothesized relationships, the next step was to develop a multivariate model. The multivariate regression allows us to go beyond the findings of the bivariate correlations in three ways. First, it enables us to look at the influences of several variables simultaneously. As stated before, many of the census tract contextual conditions are correlated with each other. Lower income tracts also tend to have higher shares of minority population and higher social risk factors, like female-headed families and welfare recipiency. Multivariate analysis allows us to identify the independent association between a health condition and each contextual variable, holding all other variables constant.

Second, in addition to estimating the strength and direction of the independent relationships, the model used in this analysis allows us to separate out how much of the variation among cities can be explained by underlying differences in the contextual variables. As we viewed trends in section 9, we sometimes speculated that rate or trend differences between cities were due to demographic and racial/ethnic differences. For example, we stated that Denver and Providence (the more Hispanic cities) might not be experiencing the same magnitude of teen birth reductions as Cleveland and Indianapolis (the more African-American cities). In this section, we will quantify how much of the rate differences are accounted for by the five contextual indicators we specify. The remaining unexplained difference relates to other factors particular to each city and not specified in our model, including health services, policy initiatives, or census tract characteristics we omitted.

The third benefit of the multivariate models is to test shifts over time in our dependent variables for statistical significance. We know from the discussion of rare events in section 7 that rates based on a small number of events can fluctuate widely from year to year, so any given annual rate may not represent the underlying true rate. Statistical tests allow us to sort out changes due to this "white noise" from true underlying trends. For example, in section 9, we saw that early prenatal care rates in Providence fell from 1997/1999 to 1998/2000. From looking at simple line graphs, it is impossible to tell whether this is a disturbing change due to lower shares of women receiving timely prenatal care or a random shift caused by few events that may well bounce back the next year. In this section, we identify key shifts in indicators and test whether they represent statistically significant changes or not.

Methodology

We implemented four regression models, each with one of our four health indicators as the continuous dependent variable. As with the bivariate analysis, each overlapping three-year average health indicator rate for each census tract is an observation. Each observation was weighted by the total number of births to account for the varying degree of precision in tracts of different sizes

As to the independent variables, we employed only a subset of the contextual indicators used in the bivariate analysis, primarily because most of the latter were correlated with one another and using them all would have biased the models. The final selection was based on a combination of the strength of correlation, how well the variable represented the concept, and the independence of the variable in relation to the others. Unfortunately, we had to exclude all of our indicators for physical conditions because they were all too closely correlated with the other factors.⁽⁴¹⁾ Crime rates were also omitted because they were not available for all of the tracts in the cities for all years. We finally selected five tract-level variables to serve as the independent variables in all four models: percentage African American, percentage Hispanic, average family income, percentage not employed, and percentage of population that moved in the past five years.

In addition to the five variables describing tract characteristics, we include three additional series of variables: city dummies, year dummies, and city-year interactions. The first set of city dummy variables controls for differences in the health-related rates solely due to conditions in the each city (other than the five we specified), while the second controls differences due to the time period of the rate. Each dummy variable is coded 1 to indicate the presence of specific attributes for a case and 0 to indicate their absence. So, for Denver observations the variable d_denv = 1; for observations in the other four cities, it would equal zero. All the possibilities within a set of dummy variables cannot be included in the model, since information about all but one of the dummies determines the value of the last category. For example, if you know the values of four of the city dummies are zero, you can figure out the observation is in the fifth city. This means the fifth city variable would not be independent of the other four, and bias the model. To account for this, one city and one year need to be left out of the series. For this analysis, we chose Cleveland and 1999 as the omitted choices.

The third set of variables consists of interaction variables between the city and year of the rate, calculated by multiplying the city and year dummy variables. These variables control for differences in the rates due to particular conditions in a city in a given time period.

The paragraphs below present the results for our four independent variables. Under each, we offer findings in three areas paralleling the three purposes of the multivariate analyses described above. First, we discuss the strength of the overall model, noting the R-squared and the level and significance of individual coefficients (table 10.6).

Second, we examine the overall explanatory power of the contextual variables using data in table 10.7. The values in the table present results for four cities in reference to how much they differ from the Cleveland value in 1999 (since as stated above Cleveland was the reference city and 1999 the reference year for the dummy variables in the model). These differences are averaged across all years for which data were available.

For each indicator, the first line ("average difference in city rates") is calculated by averaging the difference in a city's overall rates from Cleveland's rates. The second line under each indicator is the estimated difference in a city rate from the Cleveland rate that is due to differences in the five contextual characteristics. In section 9, for example, we referred to the fact that low birth weight rates are higher for more African-American areas like in Cleveland. The third line is the estimated difference in rates that is due to "unobserved characteristics" (i.e., not explained by the contextual variables), and that amount is important in interpreting results in each city.⁽⁴²⁾ These unobserved factors could be the characteristics of the individuals in the city or characteristics of the census tracts not included as independent variables in the model. They could also include influences of programs aimed at improving the particular health indicator (like Healthy Start) or barriers to healthy outcomes, such as lack of insurance or appropriate care facilities

Finally, we examine shifts in the indicator trends over the decade, charting year-by-year changes in the "due to unobserved characteristics" variable (using differences from the Cleveland 1999 values to standardize). This is also important to interpretation. For example, if the trend for teen birth rates in a city goes down on this chart it means that a total decline that might have been observed in section 8 was not totally due the contextual variables (e.g., changes in the race/income indicators) but was also partially explained by something else (e.g., local programs, broader changes in attitudes). We also note whether these trends are statistically significant. (Full data on these results is found in annex tables C.27-C.30.)

Teen Birth rates

Overall strength of model. In the first regression model, the teen birth rate for mothers age 15-19 is the dependent variable. The R-squared is moderately strong, with the independent variables accounting for 45 percent of the variation in the model. The percentage of the population not employed has the highest coefficient, with a 1 percentage point change relating to a 0.20 point change in teen birth rates (table 10.6). All of the city dummies are significant, indicating that the levels of teen birth rates vary by city characteristics not captured by the model. Interestingly, only the year dummy variables through 1994 are significant, reflecting unique conditions in those years. From section 9, we know that these were the years of the largest decreases in teen births. The only significant interaction variables are for three early years in Oakland, signaling the effect of unmeasured city conditions in Oakland in those particular years above and beyond the overall city and year influences.

Explanatory power of contextual variables. Over the 1990s, Denver's teen birth rate averaged about 4 points above the rate of Cuyahoga County (table 10.7). Summing of the regression coefficients for the city, year, and interaction dummies gives us the percentage points' difference that is not explained by the contextual characteristics included in the model. In this case, 2 percentage points' difference is explained by the contextual variables (percentage non-Hispanic black, percentage Hispanic, etc.) and another 2 percentage points are due to factors that are unique to Denver during this period that the model does not measure. For Indianapolis, the census tract characteristics would predict that the average teen birth rate over the decade would be about 0.8 points lower than Cleveland's rate, but, in fact, the rate was 1.4 points higher. This indicates that there are conditions in Indianapolis that increase the teen birth rate to 2.2 points above the predicted level.

Both Oakland and Providence show the opposite situation. The model predicts that with the contextual characteristics of their census tracts (both poorer and more Hispanic than Cleveland), their rates would both be above that of Cleveland. However, there must be some unobserved beneficial factors that bring the expected higher rates down for both cities (1.6 points lower in Oakland and 2.4 points lower in Providence).

Shifts in indicator trends. Figure 10.1 is the graphic illustration of the year-to-year change in Thus, the differences in the chart have already taken into account that Providence has generally lower incomes than Cleveland (Cuyahoga County), and that Denver has a greater share of Hispanic population, and so on. Again, the rate differences are all expressed in relation to the 1999 Cleveland rate (the reference city and year for the dummy variables). The figure looks very similar to the graph of teen birth rates from section 9, with all cities showing downward trends. We can see that Oakland's rate is declining at a faster pace. For the four cities for which we have complete data series (all except Providence), the changes from 1991 to 1999 are all highly significant (at the .001 level). The Providence trend is not statistically significant despite the fact that it parallels Oakland's decline.

Early Prenatal Care

Overall strength of model and contextual relationships. The census tract-level early prenatal care rate was the dependent variable in the second model. The overall model has the most predictive power of all four models tested, with 77 percent of the variation in the rates explained by the independent variables. Our five contextual variables are all highly significant (table 10.6 and annex table C.25). The percentage Hispanic population and the percentage of the population not employed were the strongest of the set, with a 1 percent increase corresponding to a .23 and .22 percentage point respective decrease in early prenatal care rates.

Explanatory power of contextual variables. On average over the 1990s, Denver's early prenatal care rate was about 11 points below that of Cleveland (table 10.7). The explanation for the difference is equally split between the unobservable conditions in Denver during this time and the contextual variables we include in the model. In other words, with the demographic, economic, and social status of Denver's tracts in the 1990s, the model predicts that the early prenatal care rate would be about 5.4 points lower than Cuyahoga County's rate. But, in addition to this, other factors particular to Denver that are not measured in this model are associated with another 5.5 point drop in the average rate. In addition to omitted census tract attributes, some potential policy factors were already listed in section 9--fewer facilities accepting Medicaid and immigrants choosing nontraditional health providers (either because of preference or belief that those using public health services will risk deportation).

The differences in Providence appear similar to Denver, with about half of its lower rate (8.4 points) explained by our chosen contextual indicators and the other half (7.8) by other factors not specified in the model. Like Denver, Providence had a growing immigrant population during the 1990s, which may result in higher teen birth rates.

In Oakland, we see a different situation. The socioeconomic contextual indicators in Oakland should place its rate more than 7 points lower than Cuyahoga County, but other unmeasured circumstances there raise the expected rate by 6 percentage points, narrowing the gap in the overall rates to 1 percentage point. While not conclusive, this finding is consistent with the positive impact of the Oakland Healthy Start initiative described in section 7.

Indianapolis is the only city where the variation in the five contextual variables corresponds to a higher early prenatal care rate than Cleveland (+1.9 percentage points). The unmeasured factors, however, reduce the average rate by 8 points, more than offsetting the influence of the included tract attributes.

Shifts in indicator trends. Figure 10.2 displays the regression-adjusted differences in the early prenatal care rate among the cities. The chart shows the sharp increases in Oakland that we described in section 9, along with the increases in Indianapolis in the early decade. The decreases in Denver in 1997 and Providence in 1998 stand out amid the general improvement in prenatal care in the 1990s. With our model, we can perform a joint test of significance to see if these downturns are significant or just due to random variations. The regression shows that both changes are significant, at the 0.04 level for Providence and the .01 level for Denver. Given that the prenatal care rates in the high-poverty areas in these two cities were already lowest among the five cities, these further declines are troubling.

Low-Birth Weight Rates

Overall strength of model and contextual relationships. In the third model, our independent variables explained 56 percent of the variation in the low-birth weight rates. Our five contextual variables are all highly significant, but the coefficients are quite small (table 10.6 and annex table C.26.) The percentage of the population not employed had the highest correlation, with a 1 percent increase corresponding to a .09 percentage point decrease.

Explanatory power of contextual variables. The model predicts that the low-birth weight rate in Denver would be 2 points lower than the Cleveland rate, but the characteristics of Denver during this period raise the rate by 2.9 percentage points, placing the end rate above Cleveland's. Only in Denver does this portion of the difference increase the rate. In both Indianapolis and Providence, our specified contextual variables explain most of the difference between the city rates, with unspecified city influences lowering the rate slightly more. The power of this methodology is apparent when looking at the case of Oakland. Just looking at the difference in overall low birth weight rates between Oakland and Cleveland, it appears that the two are quite similar, only .2 percentage points different. However, decomposing the difference reveals that, as with the previous two models, the levels of the contextual factors in Oakland's should be associated with a rate worse than Cleveland's (0.8 points higher), but the city must have some other protective factors that compensate for the contextual conditions.

Significance of shifts in indicator trends. As stated in section 9, the trends in low-birth weight rates were the least consistent of all the maternal and infant health indicators, and the regression-adjusted means confirm this (figure 10.3). Both Denver (beginning in 1996) and Oakland (beginning in 1997) show statistically significant improvements in low-birth weight rates, at the .04 and .002 significance level, respectively. The increases in Indianapolis from 1995 and Providence in 1996 are also significant.

Age-Adjusted Death Rates per 100,000 Population

Overall strength of model and contextual relationships. The final model has the age-adjusted death rates as the dependent variable. The explanatory power of the independent variables is moderately strong, with a .45 R-squared value. This is the one model where not all of the tract descriptive variables are significant--the share of the census tract population who are Hispanic is not correlated with the age-adjusted death rates. Of the remaining three variables, the percentage of population not employed is the most highly associated, with a 1 percentage point increase linked with a 10.4 change in the death rates. The percentage of the population living in a different house five years ago has the second highest coefficient (8.0) (table 10.6 and annex table C.27.)

Explanatory power of contextual variables. The three cities show contrasting situations for the differences in age-adjusted death rates among cities. For Denver, almost all of the difference between its and Cleveland's rates can be explained by the contextual indicators. For Indianapolis, the census tract contextual indicators account for the majority of the lower death rate compared to Cleveland, but the unmeasured conditions do increase the total magnitude of the difference. Finally, Oakland's city-specific influence is much stronger than that of the census descriptors.

Significance of shifts in indicator trends. Three of the cities (Cleveland, Indianapolis, and Oakland) show upward trends in the differences between the regression-adjusted means of the age-adjusted death rates in the late 1990s (see figure 10.4). When tested, however, only the Cleveland trend is statistically significant.

SUMMARY

The following points summarize the key findings of this section:

1. Using the bivariate methodology, we find that most of our hypotheses about the relationships between neighborhood and health conditions proved correct, with a few occasions of site differences. The correlations confirmed that higher rates of minority population, lower socioeconomic status, and lower quality housing are correlated with lower early prenatal care rates and higher rates of low birth weights, teen births, and age-adjusted deaths.

2. Two of our hypotheses were not completely verified with the proxy measures we used. First, higher levels of social stressors (measured by crime rates) were significantly related only to higher rates of low birth weight, teen births, and age-adjusted deaths (not to early prenatal care rates). Second, the hypothesis about stronger social networks correlating with lower levels of mortality and better maternal and infant outcomes was confirmed by one set of proxy variables (rates of renter occupancy, rental vacancy, and mobility) but was related to worse health outcomes with the second set (change in total and minority population and rate of home improvement loans).

3. The multivariate analysis demonstrated that much of the variation among the health indicators is explained by five selected independent variables (percentage African-American, percentage Hispanic, average family income, percentage not employed, and percentage of population that moved in the past five years). The most predictive model was the one with early prenatal care rates as the dependent variable, though the remaining models also have substantial explanatory power (R-squares range from 0.45 to 0.77). Of the five independent variables in the model, the percentage of population that is not employed was the variable most highly correlated with three of the health indicators (with early prenatal care rates as the exception).

4. The models also show that a portion of the variation is not explained by the five selected contextual indicators but, rather, by conditions particular to the city and time period of the measure. For example, Oakland's rates for maternal, infant, and mortality outcomes were consistently better than the model predicts given the contextual conditions in Oakland's census tracts. Finally, we identify which of the trends represent significant changes versus random fluctuations. For example, the results of the model enable us to confirm that the early prenatal care rates in both Denver and Providence have fallen by a statistically significant amount in the recent years--going against positive trends in the United States and the other three cities.

Endnotes

17. For more complete discussion of the appropriate uses of descriptive studies in the health field, see Grimes and Schulz 2002.

18. NNIP has developed a handbook that explains the histories, philosophies, and operating methods and techniques of the original NNIP partners--see Building and Operating Neighborhood Indicators Systems: A Guidebook, (Kingsley 1999).

19.  The mortality rates are age-adjusted to the year 2000 standard population. As to the age groupings, we used a standard categorization: less than 1; 1 to 14; 15 to 24; 25 to 44; 45 to 64; and 65 and over.

20. Oakland's crime data are available only for police beat areas, but the staff of our partner institution there created tract-level estimates for the purposes of this study.

21 For all reported crime and AFDC/TANF indicators, population denominators for specific years are again estimated by interpolating between 1990 and 2000 census figures.

22. To find out about the structure and contents of the NCDB, visit http://www.urban.org/nnip/ncua/ncdb and http://www.geolytics.com.

23. For documentation on HMDA data files, visit http://ffiec.gov.

24. This selection includes the largest 100 Primary Metropolitan Statistical Areas (PMSAs) and Metropolitan Statistical Areas (MSAs) based on their 1990 populations. PMSAs are metropolitan subcomponents of our largest urban agglomerations, Consolidated Metropolitan Statistical Areas (CMSAs). MSAs are separate freestanding metropolitan areas. Since we ranked PMSAs and MSAs by size, some smaller PMSAs within CMSAs are not included. We also excluded suburban PMSAs that did not have large central cities within their own boundaries. For a complete listing of these areas and related population data, see Kingsley and Pettit 2002.

25. This diversity in metropolitan characteristics is simply a lucky outcome for this work, since as noted in section 1, we did not structure the selection process to achieve it.

26. Patterns of population change for the central cities of these metropolitan areas were similar. Population change in U.S. central cities was generally more positive (more growth, less decline) in the 1990s than it had been in the 1980s, and this was true for all of the study cities except Oakland (modest drop in growth rate). Denver's shift, from -5 percent to +19 percent, was one of the most striking turnarounds in the country. Even the central cities of the two sites experiencing sluggish metropolitan growth did better in the 1990s than they had in the 1980s. The population of Providence increased by 8 percent in the latter, compared to 2 percent in the former and, while Cleveland still lost population in the 1990s (-5 percent), this drop was a major improvement over its substantial loss rate of the previous decade (-12 percent).

27. The reasons for using 1990 data for poverty rates are explained in the section on Neighborhood Conditions and Trends below. Since we use the 1990 poverty rates, we use the 1990 racial and ethnic composition data as well for consistency.

28. For each year of data available, Cleveland, Denver, and Indianapolis AFDC/TANF data reflect the caseload in the month of June. Providence data reflects the month of December.

29. The exception is Providence due to one high-poverty tract surrounding Brown University. The college student population drives the poverty rate, but the surrounding neighborhood is otherwise affluent.

30. Table 8.6 shows that 78 percent of all households in high-poverty tracts in Providence are renters, and it is no doubt the renters who determine the high average poverty status of these neighborhoods.

31. Moren, Lorenzo, Barbara Devaney, Dexter Chu, Melissa and Seeley. Effect of Healthy Start on Infant Mortality and Birth Outcomes.

32. Trends in the late 1990s for Cuyahoga County are difficult to interpret due to data irregularities in 1997 through 1999. See Kids Count web site http://www.aecf.org/kidscount for details.

33. Thompson, Mildred. Community Involvement in the Federal Healthy Start Program. 2000. Policylink.

34. The unusual shape of the curve is due to small numbers of births in high-poverty areas causing large jumps in the rate.

35. Infant mortality rates are often calculated from files that link birth and death records to identify the number of deaths in a given cohort of births. Since we had only tract aggregates for this analysis, we calculate infant mortality by dividing the total number of infant deaths by the total number of live births in a tract. Census tract codes for Cleveland mortality data from 1990 to 1996 were truncated to four digits. We aggregated the Cleveland death data from 1997 to 2000 to the 4-digit tracts in order to compare rates across the full decade.

36. The exception is the high-poverty areas in Oakland, which had only 45 infant deaths from 1990 to 1992 and 25 from 1997 to 1999.

37. Age-adjusted mortality rates are equal to the total deaths per 100,000 population that would have occurred assuming local death rates by age category and the national percentage distribution of population in the same categories. Generally, mortality data are normalized for age and sex, but data by gender were not available for this particular analysis.

38. Infant mortality rates are not included because of very small numbers at the census tract level.

39. Census tract codes for Cleveland mortality data from 1990 to 1996 were truncated to four digits. We aggregated the Cleveland death data from 1997 to 2000 to the 4-digit tracts in order to compare rates across time.

40. This is consistent with previous research documented in Vega 2001 and Weigners 2001.

41. Average home value and average mortgage origination were 80 percent correlated with average family income, and percentage overcrowded was 60 percent correlated with the percentage of the population that is Hispanic.

42. The difference in rates due to unobserved characteristics calculated for each city and year by first adding the regression coefficients of city dummy, the year dummy, and the city-year interaction term. This value by city and year is then subtracted from the value for Cleveland (the reference site) in that year.

Where to?

Top of Page | Contents

Main Page of Report | Contents of Report

Home Pages:
Human Services Policy (HSP)
Assistant Secretary for Planning and Evaluation (ASPE)
U.S. Department of Health and Human Services (HHS)