urban institute nonprofit social and economic policy research

The Impact of CDBG Spending on Urban Neighborhoods

Final Report

Read complete document: PDF


PrintPrint this page
Share:
Share on Facebook Share on Twitter Share on LinkedIn Share on Digg Share on Reddit
| Email this pageE-mail
Document date: August 01, 2002
Released online: August 01, 2002

Submitted To: U.S. Department of Housing and Urban Development. Contract No. C-OPC-18572 UI No. 06729-006-00.

The nonpartisan Urban Institute publishes studies, reports, and books on timely topics worthy of public consideration. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders.

Note: The Portable Document Format (PDF) of this report includes all tables and charts.


EXECUTIVE SUMMARY
THE IMPACT OF CDBG SPENDING ON URBAN NEIGHBORHOODS

Background of the Study

In 1992, the United States Congress passed the Government Performance and Results Act (GPRA), which was intended to increase the effectiveness and accountability of Federal programs by requiring agencies to measure the results of their program expenditures. Throughout the government, agencies are obliged to devise performance indicators, benchmarks and targets and apply these to the programs they administer. This research was intended to help the Department of Housing and Urban Development develop and test a variety of performance measures for its flagship urban improvement program—the Community Development Block Grant Program (CDBG).

The CDBG Program allocates Federal funding to States, cities and urban counties according to a formula based on population, poverty, age of the housing stock and other needs factors. Established in 1974, the CDBG program departed from earlier, categorical models of federal government support for urban redevelopment because it "entitled" cities and urban counties to a block of funds, to be spent at local option, but within broad guidelines established by Congress. Because the Congress viewed cities and counties as the best judges of their own community development priorities and the best designers of the best ways to pursue these priorities, the program has left almost all program decision-making up to local governments.

The program design allows HUD little influence over local choices of goals and strategies (although it requires HUD to exercise some oversight over local government capacity to administer community development programs). Nevertheless, GPRA obliges HUD to specify performance goals for all of the programs it administers, including CDBG. These goals can be found in HUD's five-year strategic and annual performance plans. Strategic Objective 4.2 of the Department's FY 2000 - 2006 Strategic Plan reflects one commonly pursued community development goal: "Disparities in well-being among neighborhoods are reduced." Many localities use CDBG funds to accomplish this goal, and to help determine whether this overall objective had been achieved, the FY 2001 Annual Performance Plan specified Outcome Indicator 4.2.1.7 - "Neighborhoods with substantial levels of CDBG investment will show improvements in such dimensions as household income, employment, business activity, homeownership and housing investment." This research aimed to test one reasonable approach to developing these and other indicators and using them to assess CDBG program performance.

Valid, reliable, and commonly accepted measures of neighborhood "improvement" or "substantial investment" are not easily derived. The dimensions of improvement specified in the Outcome Indicator are reasonable ones, but not the only ones that community development practitioners might adopt. In addition, the Department recognizes that in many instances, neighborhood improvement is the product of myriad inter-related factors, of which CDBG spending is only one. Community development practitioners understand that large-scale investments over a long period of time often are necessary to overcome decades of residential and commercial market decline. But some neighborhoods may respond much more readily than others to public investment; e.g., those that continue to hold some attraction for investors of private capital because of unique locational advantages or a stable cadre of moderate-income residents.

Recognizing that development of valid, reliable, and generally accepted performance indicators was not straightforward, and to help it meet its obligations under GPRA, HUD requested this study to examine the extent to which CDBG investments were correlated with outcomes measurable through generally-available data sources. Specifically, HUD requested that the study

  • Develop a methodology for determining "substantial" investment of CDBG funds;
  • Identify specific neighborhoods with substantial investments of CDBG resources between 1995 and 2000;
  • Develop a methodology to track changes in neighborhood characteristics over a similar time period as the investment; and,
  • Report on progress made in these neighborhoods.

Central to the request is that the study use readily available data, and that the methodology be replicable every two to three years.

Research Approach

We intended this research to accomplish four primary goals. The first goal was to develop a small number of readily available, generally accepted and easily replicable indicators of neighborhood quality of life suitable for an assessment of CDBG impacts. Our analysis strategy was to identify the few indicators that were valid, reliable, and routinely collected and to see if these were correlated with (or were related to) other indicators that were good measures of neighborhood quality, but were not readily available. If we found strong relationships between these two groups of variables, we would feel confident in using the former set as proxy indicators of neighborhood quality.

Our second goal was to develop a definition of "substantial" CDBG investments in a neighborhood to allow development of performance standards that could be fairly applied to neighborhoods expected to show some neighborhood result. We aimed to create a definition grounded in analysis, avoiding arbitrary assignment of a performance standard pegged to expenditure levels or a statistical standard of relative spending across census tracts. We planned to do this by identifying CDBG investment thresholds, above which spending produces significantly greater improvements in neighborhood outcomes (Threshold levels are defined for different neighborhood and city socio-economic conditions).

Our third goal was to recommend alternative standards or benchmarks against which to assess the performance of neighborhoods that have received substantial levels of CDBG investments. We wanted to develop a set of standards tied to different city and neighborhood conditions because we should not expect that the same level of CDBG investment would have the same effect on neighborhood quality in a stable, moderately distressed neighborhood as would be needed in a severely blighted and worsening neighborhood.

Our fourth goal was to compare the study's results with local informant's understanding of the impact of CDBG on their neighborhoods in the late 1990's. This involved testing the reasonableness of our proposed categorization of neighborhoods or tracts into "out-performing" and "under-performing" with local officials and neighborhood representatives in four of the 17 cities including in this study.

Overall Results

In general, we found that larger CDBG investments are linked to improvements in neighborhood quality in the 17 cities studied for this project. Additionally, we found that two indicators - one reflecting residential mortgage lending activity and the other reflecting business and employment - are good proxy measures of some (but not all) dimensions of neighborhood quality. The data underlying these measures - median loan amount from Home Mortgage Disclosure Act data and number of businesses from Dun and Bradstreet - are readily available for all CDBG grantees, are inexpensive compared to other comparable sources of information, and are strongly related to aspects of neighborhood quality uncovered through extensive analysis of numerous other indicators.

Our finding of an overall relationship between CDBG spending and neighborhood quality improvements in the study sites is encouraging given the substantial gaps in our information about the effects of the CDBG program. But this initial study was not broad enough to conclusively prove that CDBG investments are positively correlated with specified measurable results. Among other issues, the study does not reflect a nationally representative sample of jurisdictions. It also does not account for the effects of other public investments, including earlier CDBG investments. Most neighborhoods receiving CDBG funding between 1994 and 1996 had been funded in earlier years, potentially including all of the years since program inception in 1974. We did not measure this spending, but the changes in neighborhood quality we observed could have resulted from this earlier spending in addition to the later spending we could measure. Moreover, rarely is CDBG spending the only public investment in neighborhoods, which could include other HUD programs (HOME, most notably), other Federal programs (Low-Income Housing Tax Credits, for example), and numerous sources of State, county, and local government programs to fund infrastructure and other investments and deliver public safety and other programs.

In the course of developing the performance measure described in this report, a number of decisions were made that might affect the results. For example, we used CDBG spending per poor resident as a measure of CDBG investment, thus tying CDBG spending to the size of the target population in each neighborhood. We could have adopted some other measure—for example, CDBG spending per low-and-moderate income person, or CDBG spending per capita -- that might have changed our results. We also excluded neighborhoods receiving less than the $86,737 average level of annual CDBG spending between 1994 and 1996 across the 17 cities. (This is roughly the price of a single renovated housing unit.) We could have adopted a more or less restrictive standard than this one, which also might have changed the results.

Our conversations with four cities included in our analysis yielded somewhat mixed results. Local informants were not able to resoundingly endorse or completely refute any of the proposed performance measures. In fact, local informants agreed with just 27 percent of our categorizations of neighborhood performance.

Although the aggregate verification results of the local site visits are mixed, they do reveal that the performance measures based on the median loan amount indicator are more likely to conform to the views of local practitioners than the performance measures that use the number of businesses in a tract. In other words, from the local informants' perspective, the median loan amount indicator does a better job overall of capturing the impact of the program than does the number of businesses indicator.

Conclusions and Recommendations

The analysis presented here is a good first step in identifying a relationship between CDBG spending and measurable improvements in neighborhood quality. The performance measures we developed have the considerable virtue of simplicity, ready availability, and intuitive plausibility. Moreover, the performance standards we developed require the analysis of only two variables—CDBG spending and one of two performance indicators (either median loan amount or number of businesses).

As with any performance measure or set of measures, however, they are subject to endemic problems of data suitability, arbitrary specifications of standards, and inability to account for all factors that affect the relationship between community development investments and neighborhood outcomes. A follow-up research project could address some of these problems through the following modifications:

  • Inclusion of all entitlement grantees (and therefore, many more neighborhoods to analyze) and measurement of both CDBG spending and neighborhood change over a longer period of time. This analysis might result in a non-arbitrary cut-off for inclusion of neighborhoods into the performance system, rather than the above-average investment standard used here. This cut-off could be established through more sophisticated statistical techniques that would identify a point where CDBG investments produce accelerated improvements to neighborhood quality.
  • Continued improvements and upgrades to HUD's management systems to allow better tracking of CDBG expenditures. The Department already has plans to complete IDIS data cleaning and update of user protocols, ensuring more complete geographic coverage of the system. HUD also is improving the quality of the data it collects. (It should be noted that, by block grant standards, HUD's IDIS data system already is quite good; information on the community services block grant is paltry, by comparison.)
  • Increasing the numbers of neighborhoods that fall into each of the neighborhood categories constructed to yield more statistically significant relationships between CDBG expenditures and neighborhood quality indicators. This would allow construction of neighborhood-appropriate standards for many more classes of neighborhoods than we could produce in this research.
  • Inclusion expenditures under Federal HOME program, Low-Income Housing Tax Credit, and HOPE VI programs. Including other expenditures of community development funding in a neighborhood in addition to CDBG investments would constitute a more realistic (if still incomplete) measure of community development investments.

Even an enhanced performance measurement system would face problems in measurement and application, however. For example:

  • Any use of CDBG data will require adoption of decision-rules to allocate spending to neighborhoods, which will risk misallocation of spending to: (a) a single neighborhood when it benefits multiple census tracts, (b) multiple neighborhoods when it benefits a single tract, primarily, and (c) an entire tract when it benefits only a small portion within it.
  • No system would be able to take account of the local expenditures on infrastructure, police and fire protection, public education, or other municipal services that certainly contribute to neighborhood quality.
  • Only a far more complex and data-dependent system than constructed here could take account of the multiple objectives CDBG administrators pursue and which are not reflected in measures of neighborhood quality. Most problematic are investments intended to preserve or expand the supply of affordable housing in neighborhoods experiencing rapid increases in home prices and rents. In this example, CDBG investments are expected to help suppress increases in median loan amount—one of our best indicators of neighborhood quality.

In view of these limitations, perhaps the best way to think about the design and use of a performance measurement system such as that developed here would be as a tool to help communities interested in assessing their own community development performance.

Local administrators contacted for this study expressed considerable interest in the goals of the research. Although they would resist the application of a Federal standard that might entail sanctions for "poor" performance in relation to a specific statistical standard that limits the range of objectives for their block grant funds, they nevertheless would welcome a process of setting benchmarks by which they could assess their own progress in improving low-income neighborhoods. This is an area of public investment that has not, to our knowledge, ever developed such benchmarks. What are reasonable expectations for neighborhood change? How much investment is required to produce it, and under what circumstances? And where have neighborhoods performed better than expected and what can we learn about the strategies and supporting factors that produced this result? This research only begins to answer these questions, but we are convinced that it is a promising beginning.

Note: The Portable Document Format (PDF) of this report includes all tables and charts.


ACKNOWLEDGMENTS

We wish to thank the following members of the National Neighborhood Indicators Partnership for the use of data resources: The Boston Foundation; the Center on Urban Poverty and Social Change (Cleveland); United Way of Central Indiana/Community Service Council (Indianapolis); the Urban Strategies Council (Oakland); and the Providence Plan. We also wish to thank community development and city planning officials in Boston, Cleveland, Columbus, Houston, Milwaukee and Washington, DC, for their cooperation in the verification of our results. At the Urban Institute, Noah Sawyer, Seon Joo Lee and Diane Hendricks provided valuable assistance.



Topics/Tags: | Cities and Neighborhoods | Governing | Performance Measurement / Mgmt


Usage and reprints: Most publications may be downloaded free of charge from the web site and may be used and copies made for research, academic, policy or other non-commercial purposes. Proper attribution is required. Posting UI research papers on other websites is permitted subject to prior approval from the Urban Institute—contact publicaffairs@urban.org.

If you are unable to access or print the PDF document please contact us or call the Publications Office at (202) 261-5687.

Disclaimer: The nonpartisan Urban Institute publishes studies, reports, and books on timely topics worthy of public consideration. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders. Copyright of the written materials contained within the Urban Institute website is owned or controlled by the Urban Institute.

Email this Page