IS NEIGHBORHOOD POVERTY A GOOD MARKER FOR NEIGHBORHOOD QUALITY?
A number of federal programs that invest significant resources in the nation's cities use the neighborhood poverty rate (i.e., the fraction of each neighborhood's residents with incomes below the federal poverty line) as an eligibility criterion. Curiously, some of these programs, such as the U.S. Department of Housing and Urban Development's (HUD) project-based housing voucher program, exclude neighborhoods with poverty rates that exceed a certain threshold (currently 20 percent), while others, such as the U.S. Department of Treasury's Low-Income Housing Tax Credit, target neighborhoods with large low-income populations. This inconsistency, in itself, suggests some confusion about the value of using neighborhood poverty or low income as a targeting criterion.
The purpose of this study was to examine the relationship between the neighborhood poverty rate and neighborhood quality in Baltimore. We addressed four questions: (1) Is there a relationship between neighborhood poverty rates and neighborhood quality? (2) What is the form of this relationship? (3) Is the trend in poverty rates over time a better marker of neighborhood quality than the rate at a point in time? (4) Are there mitigating factors that affect the relationship between neighborhood poverty and neighborhood quality? To our knowledge, this study provides the most comprehensive examination of the poverty-quality relationship at the neighborhood level that we are aware of.
The relevant literature offers little insight into the relationship between poverty rates and neighborhood quality. With little theory to guide our analysis, we conducted case studies of 25 Baltimore neighborhoods, examining both quantitative and qualitative data, to assess whether the neighborhood poverty rate influenced multiple measures of neighborhood quality. The neighborhoods vary by poverty rate and range from low-poverty (<.20), to middle-poverty (.20-.40), to high-poverty (>.40). The analysis examined roughly 90 measures of neighborhood quality covering eight domains: (1) demographics and socioeconomics; (2) physical environment; (3) social environment; (4) school quality; (5) crime and safety; (6) economic investment; (7) image; and (8) health. Sources of data include the last three decennial Censuses; local administrative sources; systematic on-site observations; interviews conducted with residents, arm's length experts, and business owners; and historical newspaper and internet research. The spatial variation in the neighborhood sample allows an examination of whether proximity between poverty and non-poverty neighborhoods has an effect on the relationship between the neighborhood poverty rate and neighborhood quality.
The first analysis sample includes five adjacent census tracts in southwest Baltimore: St. Joseph's (17 percent of residents living below the federal poverty line); Gwynns Falls/Carroll/South Hilton (Gwynns Falls) (23 percent poverty); Penrose/Franklin Square (Penrose) (28 percent poverty); Shipley Hill (32 percent poverty); and Carlton Ridge/Boyd Booth (Boyd Booth) (51 percent poverty). Few indicators for these neighborhoods exhibited a linear relationship either with the poverty rate in 2000 alone, or with the poverty rate over time. The lack of correlation between the neighborhood poverty rate and neighborhood quality was most evident in the middle-poverty neighborhoods. For many indicators, Shipley Hill (.32), the highest of the middle-poverty neighborhoods, ranked worse than the highest-poverty neighborhood, Boyd Booth (.51). In addition, the lowest of the middle-poverty neighborhoods, Gwynns Falls (.23), ranked more poorly on several negative quality indicators compared to the other middle-poverty neighborhoods, and Penrose (.28) often ranked better than St. Joseph's (.17) or Gwynns Falls (.23).
In assessing quality indicators, the involvement of community organizations and economic investment emerged as more informative about the neighborhood's quality than the poverty rate alone. The poverty trajectory was helpful for understanding the variations among the middle-poverty neighborhoods, while the underclass is a better marker of neighborhood quality for only the two highest-poverty neighborhoods. There was no evidence of a poverty threshold. Even though St. Joseph's (.17) exhibited high quality overall, we saw no evidence of a precipitous decline in quality indicators between the low-poverty neighborhood and the middle-poverty neighborhoods, as the threshold theory suggests.
The second analysis sample includes four adjacent neighborhoods with poverty rates exceeding 20 percent--Coldstream-Homestead-Montebello (CHM) (.25), East Baltimore Midway (Midway) (.26), Barclay (.35), and Better Waverly (B. Waverly) (.44)--and the not-adjacent Lower Hamilton (L. Hamilton) (.18) neighborhood, whose poverty rate falls below 20 percent. This spatial grouping was chosen to see whether the poverty level of adjacent neighborhoods bears on the relationship between neighborhood poverty and neighborhood quality. A linear relationship between poverty and quality was evident in only a few socioeconomic indicators, but the vast majority of indicators across multiple domains did not exhibit this pattern. Though the middle-poverty neighborhoods displayed worse quality than the low-poverty neighborhood, the high-poverty neighborhood consistently performed as well as, or better than, the middle-poverty neighborhoods. Other indicators such as school quality and some measures of economic investment showed no relationship with the poverty rate. Some quality indicators, such as median residential sales price, suggested a possible threshold or tipping point at 20 percent poverty beyond which quality declined markedly.
These five neighborhoods suggest that adjacency may be a mitigating factor affecting neighborhood quality. For example, a portion of the high-poverty neighborhood that is adjacent to higher quality neighborhoods exhibits higher quality than its poverty rate would lead us to expect. Although we expected the poverty trajectory to be a better marker of quality, the evidence for these five neighborhoods did not support this view. In addition, the four underclass measures did not appear to be better indicators of neighborhood quality for these five neighborhoods.
The third analysis sample includes five neighborhoods where the two lowest-poverty neighborhoods, Frankford (.19) and Parkside (.24), are adjacent to one another, as are the two high-poverty neighborhoods, Darley Park (.35) and Broadway East (Bdwy. East) (.53). Cedonia (.27), the middle-poverty neighborhood, is located close to Frankford (.19) and Parkside (.24) but is not adjacent to either. Several measures of neighborhood quality have a generally linear relationship with the neighborhood poverty rate. These measures include assault rates, features of the social environment, and measures of neighborhood image. Other measures suggest a poverty rate threshold beyond which neighborhood quality declines dramatically. Examples include median residential sales prices, abandoned housing, non- two-parent households, and educational attainment. Income above $60,000, owner-occupied homes and economic investment did not suggest a relationship between neighborhood poverty and quality. The poverty rates of adjacent tracts may play some role in the relationship of poverty and quality in the study sites. The poverty trajectory over time provides insights into the pattern of quality measures in low-poverty Frankford (.19), where poverty has worsened significantly in the past 10 years, and middle-poverty Cedonia (.27), where poverty has remained stable over the past 20 years.
The fourth analysis sample is referred to as the North Avenue Mobility Corridor because it roughly approximates one of the migration pathways out of the city and into Baltimore County. It includes the neighborhoods of Walbrook (.19), Rosemont-Winchester (Winchester) (.23), West Forest Park (W.F. Park) (.27), Lower End of Reservoir Hill (Reservoir Hill) (.33), and Upton (.45). Our analysis revealed neither strong correlations between poverty rates and neighborhood quality, nor support for a 20 percent or 40 percent poverty threshold. In fact, although W.F. Park (.27) is a middle-poverty neighborhood, it mirrors the quality of the lowest-poverty neighborhood, Walbrook (.19). By contrast, Winchester (.23), another middle-poverty neighborhood, consistently mirrors the two highest-poverty neighborhoods. We found no evidence that race mitigates the relation between neighborhood poverty and neighborhood quality, but did observe a linear relationship between age and the poverty rate: as the proportion of residents under age 18 increases, the neighborhood poverty rates increase. The poverty rates of adjacent neighborhoods may also be influential in mitigating the effects of poverty on neighborhood quality. The location and physical attributes of the middle-poverty neighborhood W.F. Park (.27) may help explain the erratic pattern that characterizes the poverty-quality relationship in this neighborhood. Its location near the city's borders, its proximity to lower-poverty neighborhoods, and the presence of large forested areas within its boundaries may contribute to its relatively higher quality.
The fifth and final analysis sample includes the five neighborhoods that are non-adjacent and located throughout the western half of the city of Baltimore: Falstaff (.18), Cylburn (.21), Dickeyville (.26), Cherry Hill (.32), and Mt. Wynans (.42). We did not find a linear relationship between the neighborhood poverty rate and neighborhood quality. While the low-poverty neighborhood fared better than the high-poverty neighborhood on nearly all measures, there was significant variation among the middle-poverty neighborhoods. On some measures, middle-poverty neighborhoods demonstrated poorer neighborhood quality than the high-poverty neighborhood. Curiously, the middle-poverty neighborhood Dickeyville (.26) had equal, and sometimes better, quality than the low-poverty neighborhood.
We also did not find a threshold effect at either 20 percent poverty or at 40 percent poverty. While in one middle-poverty neighborhood, Cylburn (.21), measures of neighborhood quality declined significantly beyond the 20 percent threshold, consistent with the 20 percent poverty threshold hypothesis, it did not hold for another middle-poverty neighborhood, Dickeyville (.26), where quality was higher despite its higher poverty rate. And while the high-poverty neighborhood almost always had the poorest neighborhood quality, one middle-poverty neighborhood, Cylburn (.21), often had a similarly low level of neighborhood quality. This undercuts the 40 percent threshold hypothesis. Evidence on the poverty trajectory is mixed. In the three neighborhoods that have increasing poverty rates, several measures of quality have also declined. But in the two neighborhoods with declining poverty rates, there was no evidence that quality was improving.
To some extent, the presence of an underclass appears to be more closely related to neighborhood quality in these case study neighborhoods than the poverty rate alone. The existence of an underclass in the middle-poverty neighborhood, Cylburn (.21), might play a role in explaining why its quality is similar to the high-poverty neighborhood. Similarly, the absence of an underclass may be relevant to why the middle-poverty neighborhood, Dickeyville (.26), has similar quality to that of the low-poverty neighborhood. The analysis also suggests that the neighborhood quality in the middle-poverty neighborhood, Cylburn (.21), may be affected by high-poverty rates in adjacent neighborhoods.
The empirical evidence presented in this study indicates that neighborhood poverty is not a good marker of neighborhood quality. While the lowest-poverty neighborhoods almost always ranked higher on quality than the highest-poverty neighborhoods, the pattern for middle-poverty neighborhoods was erratic. There was also little support for a 20 percent poverty threshold, for the 20-year trend in poverty being a better indicator of quality than the poverty rate in 2000, or for the presence of an underclass representing a better marker of neighborhood quality.
Neither age nor race was a mitigating factor in the relationship between neighborhood poverty and neighborhood quality, though the lack of variation in these demographics limited our ability to test this effect. But the poverty rates of adjacent neighborhoods may plausibly influence the poverty-quality relationship in a few neighborhoods through the negative spillovers from high-poverty neighborhoods or the positive spillovers from those with low rates of poverty.
Recommendations for alternative eligibility criteria for government programs include systematic research to identify neighborhood attributes causally linked to improvement and success (or decline and failure), reconsidering census tracts as the representation of neighborhoods, accepting the possibility that multiple measures may be required to capture the elusive concept of neighborhood quality, and thinking about a role for localities in targeting government resources to neighborhoods.