Theory, Measurement, and Analysis
Measuring Comprehensive Community Initiative Outcomes Using Data Available for Small Areas
Claudia Coulton and Robinson Hollister
Comprehensive community initiatives (CCIs), typically carried out within relatively small, geographically bounded communities, need information about those communities and their residents in order to plan and evaluate their work. Although extensive demographic, economic, and social indicators are available for the nation and other larger geographic units, neighborhood-level information is seldom produced routinely. Furthermore, because CCIs are comprehensive, they need information about a broad range of outcomes, tracked over multiple points in time. For all these reasons, it is desirable that CCIs draw upon existing data sources to derive information about small geographic areas.
Our purpose is to describe the many and varied kinds of data sources that have the potential to produce small area information for CCIs. The focus is on obtaining and using administrative, survey, and census data that are already being collected for other purposes and converting them into information for CCIs. We also address the important cautions and limitations of small area analysis and the application of data collected for other purposes to the measurement of community change. We attempt to cast a wide net by considering many domains that are important to CCIs, including housing, economic development, safety and security, education, service reform, and community building.
Advantages and Disadvantages of Available Data
The advantages of using available data to measure change are several. First, a retrospective baseline can be created because the measures use data that have already been collected. Second, communities can be compared with one another, since many data sources cover the entire city, district, or county within which the CCI target community is located. Third, geographic information system (GIS) technology makes it practical to manipulate data and build up to the desired units of geography through aggregation. Fourth, the data can be subjected not only to traditional time-trend and comparative analyses but also to spatial and ecological analyses.
There are, of course, disadvantages as well. Data collected for purposes other than the particular evaluation or planning needs of CCIs may only approximate the concepts of interest. Also, because the available data sources vary considerably in their accuracy, care must be taken to adjust for or avoid data elements that are vulnerable to well-known errors. Finally, most available data do not capture the important social processes that go on among community residents and within and between community organizations. These important outcomes of CCIs require special measures and original data collection activities.
Establishing Geographic Boundaries
Although "community," as used in the term "comprehensive community initiative," is a social unit, a CCI must establish clearly demarcated geographic boundaries to acquire and use available data. Geographically bounded communities are often thought of as neighborhoods (Chaskin, 1995). Neighborhood definition is not always an easy task, however, especially when CCI participants disagree about boundaries or when the target area encompasses several neighborhoods.
Researchers have traditionally used census geography for the purpose of data aggregation, with census tracts and block groups serving as proxies for neighborhoods. (A census tract is a geographic area containing between 2,500 and 8,000 residents. A block group is a number of contiguous blocks within a census tract designed to contain about 250-550 housing units. Both designations are established by the U.S. Bureau of the Census.) Research on resident perceptions has shown that residents seldom agree on neighborhood boundaries, but that the average size of their perceived neighborhood is somewhat larger than their block group but smaller than their census tract (Coulton, Korbin, Chan, Su, and Wang, 1997). For convenience and comparability, CCIs often link boundary definitions to census geography, although GIS technology can support resident-defined boundaries as well.
The use of available data sources may present some limitations on geographic definitions. For example, although it is always desirable to obtain data at the smallest geographic unit available, for confidentiality purposes some data sources contain only census codes or administrative districts rather than individual addresses. This can reduce the flexibility of the CCI to set and change boundaries and still make relevant calculations.
If measures built upon available data are to be sensitive indicators of change, it is also important that the geographic boundaries used for data aggregation are commensurate with the real targets of the CCI at a particular point in time. If, for example, the CCI is working on different outcomes or is at different stages in particular sections of the neighborhood, the data should be aggregated so as to capture those differences.
There is a long tradition of using data collected for administrative purposes to produce social and economic indicators (Rossi, 1972; Annie E. Casey Foundation, 1997). The emphasis on outcomes and accountability in many social programs has raised additional interest in such information (Schorr, 1994). Most administrative agencies now have computerized record systems, and the advent of GIS technology makes it feasible to calculate indicators for smaller areas from those large data bases.
Numerous sources and types of data from administrative agencies can be used to produce measures useful for CCIs. Most data bases are maintained by local agencies, but a few state and federal data bases can be used for small area measures. Because of the local nature of much of the data, the descriptions in this chapter may not match exactly what is available in a particular locale. Although the list of data sources described here is long, it is not exhaustive. As shown in the table (table 2), the sources are grouped into six categories: housing, economy, safety and security, education, health, social services, and community resources and involvement. These categories reflect the primary uses to which the data have been put, although most sources are applicable in several categories when combined with other information.
Housing-Related Data Sources
Many CCIs are interested in improving housing in their communities. Data from a variety of sources can be used to develop indicators of housing stock, conditions, and markets in small areas. Available information covers housing characteristics, condition and quality, construction and demolition, and financing.
Home Mortgage Disclosure Act (HMDA) Information
The Home Mortgage Disclosure Act (HMDA), enacted in 1975 and implemented by the Federal Reserve Board, requires covered institutions to compile and disclose data about loan applications they receive and home purchase and home improvement loans they originate or purchase during each calendar year. Institutions required to file HMDA data include commercial banks, savings and loans, credit unions, and mortgage companies that meet specific criteria.
The data are maintained in the institution’s Loan Application Register (LAR). Each LAR record contains loan/application information such as type, purpose, amount, and action taken. Each record also contains some applicant and co-applicant characteristics, such as race, gender, and gross annual income. Information about the property location, such as the census tract, county, metropolitan statistical area (MSA), and state is also recorded. In addition, each record includes a few variables about the census tract, such as population, number and percent of minority population, median income, and number of owner-occupied units.
Another record, the Transmittal Sheet (TS), contains information about each financial institution, including name, address, parent company name and address, and tax identification number. The LAR and TS data can be linked by using a respondent identification and agency code found on both files.
These data, available on reel, cartridge, and CD-ROM from the Federal Reserve Bank’s Federal Financial Institutions Examination Council (FFIEC), are issued annually, with each year’s data released in the fall of the following year. An order form can be downloaded from http://www.ffiec.gov/hmda. Additional information about the data, such as reporting criteria and background information, can also be found on the web site.
Several small area indicators can be developed from the HMDA data. The total numbers of loans applied for, approved, denied, or withdrawn can be determined, along with reasons for denial. The purpose of the loan/application (to purchase, improve, or refinance a home) is available, as is the type (conventional, FHA, VA, or FmHA). Demographic characteristics of loan applicants and co-applicants are also available. From this information, evaluators can compute approval and denial rates for the small area based on race, gender, and income of applicants; approval and denial rates by financial institution; financial institutions’ shares of lending in a particular area or to a particular group; and the lending patterns of specific financial institutions.
The data have been used by fair housing groups to examine lending patterns in communities and to test for evidence of discriminatory practices. Public officials have analyzed the performance of financial institutions in meeting the housing credit needs of their communities. The economic stability of a neighborhood can also be assessed by computing trends in commercial lending and conventional mortgage activity.
Local Property Tax Data
A wide variety of information about every parcel of property is collected and maintained by the local auditor or assessor office for the purpose of levying taxes. These records contain three types of information: tax billing records, characteristics data, and deed transfer data. The tax billing record includes parcel number, parcel size, owner name and address, land and building assessed values, property class, land use codes, gross taxes, special assessments, and delinquency status. The characteristics data include parcel number, number of rooms, year built, lot size, land use code, and roof type. The deed transfer data includes information about property sales and transfers, names of buyer and seller, address of property, sales amount, date of sale, and deed type. The tax billing and characteristics data are available annually, and the deed transfer data are available monthly. All three types can be linked by parcel number.
Although these are public records, obtaining them in useable formats for analysis can be difficult in some regions. Some local offices provide the data electronically, while others may not have the resources to do so.
Some of the same property information is contained in a commercial software product known as MetroScan, published for approximately 130 counties nationwide by Transamerica Information Management. Primarily intended for use by realtors, the MetroScan data base includes the census tract and block code for each parcel, as well as some school district information, and gives the user the ability to print county and street maps. The MetroScan data base can be searched by a number of variables, including tract, street, and property type. Although the data can be exported in several formats, the user can export only 5,000 records at a time. Although cumbersome for an area with a large number of properties, this process can be quite convenient for a more modest neighborhood. Information on MetroScan can be obtained at http://www.transamerica.com.
Small area indicators that can be developed from real property data include the market and assessed values of homes, median and average sales prices, volume of property sales and transfers, number of sales by deed type (such as sheriff, trustee, or warranty), and number of tax delinquent properties. All these indicators can be computed by property class (residential, commercial, or industrial), land use (single family, commercial warehouse, manufacturing plant, etc.), and geographic area.
The data have been used to study small area trends in housing values and the market for real estate. Other studies have examined the pattern of population movement by linking deed transfers for sale and purchase, thus tracking where individuals are buying and selling property. These and other indicators have been used together to assess neighborhood stability and condition.
Building and Demolition Permits
Building and demolition permits—intended to ensure that zoning requirements, fire and structural standards, and other building standards are met—are collected and maintained within a municipality’s building or housing department. Each permit includes the name of the owner, address of the property, parcel number, written description of work to be done, codes identifying work to be done, permit use class (such as commercial or residential), estimated cost of work to be done, and permit issue and expiration dates. In addition, the permit shows the name and registration number of the contractor performing the work, inspection date, building inspector, and permit fees. The permit information is public record and can be obtained in the appropriate city department, although its availability in electronic format varies by city.
Small area indicators that can be developed from this data include the number of permits by type (such as new construction or external rehabilitation) and by use class (such as commercial or residential). By calculating values associated with building permits by type and geographic area, it is possible to determine, for example, the level of investment being made in residential rehabilitation or new housing construction. Neighborhood groups have used the information to track housing demolition and, in conjunction with other housing indicators, to assess neighborhood stability and condition.
Housing Code Enforcement Reports
Housing code enforcement information is available from a city’s building inspection or housing division. The housing code attempts to ensure the health and safety of the occupants of a building by setting rules for basic maintenance and upkeep. Reports of housing code violations typically include the address and owner of the property with the violation, parcel number, code being violated, inspector name, date the violation was cited, time period to comply, compliance date, and whether or not legal action was taken. In addition to code violations, the records show complaints of nuisances or problems that make a neighborhood unattractive or unsafe. The information is available to the public from the appropriate city department, although not every city can supply it in electronic format. The data are frequently updated.
Indicators that can be developed from these data include number and rate of properties with violations, violations by type (for example, faulty wiring or paint needed) and severity, and number and rate of nuisances (abandoned buildings and cars, garbage improperly stored, etc.) by geographic area. Compliance rates and violations that result in legal action can also be calculated. The data have been used, in conjunction with other housing indicators, to assess the stability and condition of neighborhoods and to document success in enforcing compliance with code violations.
HUD Information on Public and Subsidized Housing
The U.S. Department of Housing and Urban Development (HUD) collects and maintains a variety of information about public and subsidized housing units, most of it gathered at the local level and reported to HUD. These data are available for census tracts and individual housing projects.
The HUD information includes the name and address of the housing project, total number of subsidized housing units by type (such as public housing or Section 8), percent of units occupied, average rent, percent of tenants who moved during the last year, average stay in unit, average number of months on waiting list, average number of persons per unit, average income, percent of persons in different income categories, where the majority of tenants’ income comes from (such as welfare or wages), percent with assets over $5,000, average age of head of household, percent with disability, racial breakdown, average bedroom size, percent of residents by family type, percent overhoused (more bedrooms than people), and percent with utility allowance. Geographic information includes the zip code, latitude and longitude, census tract, and county. Additional tract-level information includes percent poor, percent minority, and percent households that are owner-occupied. This information is available through HUD’s HUDUSER data base by project site or census tract. The data can be downloaded for inclusion in data bases, spreadsheets, and statistical packages. More information is available at http://www.huduser.org/data.html.
Each local housing authority collects and maintain its own records regarding tenants and management of units. The locally maintained data bases may be more complete than information supplied to HUD and may provide more detail and flexibility than the national data set, which offers only aggregate information. The ease with which the information can be obtained may vary by housing authority.
Public housing data can be used to determine the economic status and mobility of public housing residents within the neighborhood. Combined with a total housing unit count from the census, it can be used to calculate the percentage of housing in the neighborhood that is publicly operated. The data have been used by a variety of researchers, evaluators, and housing administrators to profile the public and subsidized housing population and develop programs to assist tenants in moving from subsidized housing to home ownership.
Economic-Related Data Sources
Many administrative data sources contain information about the economic and job activity in an area, which can be useful in supporting CCI work toward economic development and related issues.
Employment Data Bases
Although departments of employment services are the primary sources of employment information in states, these agencies do not usually publish information for small areas. Nevertheless, several of their data bases may be useful to CCIs in measuring aspects of employment within neighborhoods.
ES202 information. According to federal mandate, states must collect reports related to unemployment insurance from every establishment that employs more than one worker. The ES202 data base, generated from these reports, therefore covers nearly all paid employees, although self-employed individuals, such as doctors, and family businesses with no paid employees are largely unrepresented. This is the only government data base that provides company-level information along with geographic location.
The data contained in the ES202 file includes establishment name, legal name, address, city, zip code, state, county code, standard industrial classification (SIC) code, ownership code (indicating public or private ownership), number of employees, and total quarterly wages. Total monthly employment is provided for each establishment. Two variables, the unemployment insurance (UI) number and the reporting unit number (RUN), uniquely identify a company. A date indicates when the UI number was established, and successor and predecessor UI numbers may enable the user to link companies that have changed ownership over time. (These variables are not required and therefore may not be reported.) Another variable indicates whether the company is a multi-unit employer (such as a bank with multiple branches) or a single-unit employer.
The data are available quarterly from the state’s bureau of employment services, with each quarter’s data available in the same quarter of the following year. The information is available electronically, but the format may vary by state.
Although they provide unique establishment-level information, the ES202 data pose some well-known problems (White, Zipp, and McMahon, 1990; Leete and Bania, 1995; Waits, Rex, and Melnick, 1997). First, address information may not be accurate, since some firms mistakenly report all employment at the address of a headquarters or of an accountant who completes the report rather than the address of the actual employment location. There are additional problems if the user wants to look at establishments over time. Each establishment has unique UI and RUN numbers to link quarter by quarter, yet these numbers change if an establishment changes ownership. If the predecessor or successor UI number is not reported, there may be some difficulty linking establishments.
The advantage of the ES202 data is the accuracy of the total employment reported by each firm. The federal mandate carries with it a tax liability that encourages accurate reporting of employment. In addition, the ES202 employment numbers at the county level have been found to mirror other official government measures of employment closely.
Federal and state rules restrict who may obtain these data and how they may be used, and some confidentiality rules apply. For example, if there are only two employers in a geographic area with a particular SIC code, or if one establishment accounts for more than 80 percent of the employment in a particular SIC code, the data must be suppressed.
The ES202 data can be used to calculate several small area indicators on employment. Employment, wages, and number of establishments can be calculated by SIC code and geographic area. Average payroll per employee by industry and geographic area can also be calculated. If data are analyzed over time, the number of business openings, closings, and relocations and associated employment and wages can be determined. Employment gains and losses by geographic area and SIC code can also be computed. The ES202 employment data have been used to measure employment change by SIC in small areas, develop estimates of employment in various geographic areas, determine location of specific types of employment, and estimate locations of expected job openings.
UI wage record. The unemployment insurance (UI) wage record is also available from each state’s bureau of employment services and collected under the same federal mandate as ES202. In addition to information about the employer, the wage record file contains specific information about the employee, including name, social security number, quarterly wages paid, and weeks worked. Employer information includes name, address, city, zip code, state, SIC code, and UI number. Like the ES202 data, UI wage records are available quarterly and are subject to some restrictions on obtaining and using the files.
The UI wage records cannot be used alone to develop small area indicators because they do not contain individual wage earners’ home addresses. Even so, some researchers have linked address-coded public assistance records to UI records to determine the employment experiences of welfare recipients within a geographic area. Links could be made to other agency records as well.
UI claimant file. The UI claimant file, collected under the same federal mandate as ES202 data and the UI wage record, contains specific information about individuals who have filed for unemployment compensation. The data include the claimant’s social security number, address, birth date, sex, and race. Additional information includes weekly benefit amount, average weekly wage, number of qualifying weeks, date of claim, date of separation from job, pay rate, and employer UI account number. These data are confidential, and state and federal regulations restrict who may obtain them and how they may be used. The file is available quarterly from the state’s bureau of employment services.
Small area indicators that can be developed from these data include percent of persons receiving benefits, average length of time on unemployment, and percent who have exhausted benefits. Demographic characteristics of unemployment compensation recipients can be also computed. The data have been used, in conjunction with other data sources, to target recipients who have exhausted benefits and provide access to job training programs.
Business directories can be helpful sources of employment information. Only a few directories are listed here, but most libraries maintain a catalogue of the directories available and the types of information provided by each. On their own, the directories may not be a complete source of employment information, but they can enhance the accuracy of addresses and establishments in a geographic area when used in conjunction with ES202 data (Carlson, 1995).
Cole’s Business Directories. A directory is published for each state by American Directory Publishing Company in Omaha, Nebraska, and distributed by Cole Publications. Directories are available in print and electronic (CD-ROM) formats. The data base has several drawbacks. Unlike ES202 files, it contains employment ranges only, not actual employment levels, and firms are under no legal requirement to be listed. In addition, the directory is updated continuously, making it difficult to establish a list of firms for a single point in time. The listing is based on telephone directories, and so firms not listed in the telephone book are unlikely to be included.
Various versions are sold as electronic "national yellow page" listings for businesses, primarily intended for other businesses marketing products to firms in particular industrial classes. Other electronic business directories include ProPhone Business Listing, PhoneDisc PowerFinder, and Select Phone Business Listings, each with similar shortcomings.
Dun and Bradstreet Indicators. Based on Dun and Bradstreet credit rating data, the Dun Market Indicators (DMI) file is limited to firms that actively seek out a credit rating record with Dun and Bradstreet. Smaller firms and those without a credit history are unlikely to be included. DMI is also likely to contain outdated data, since firms that close or move have no incentive to update their records. Additionally, firms may understate or overstate employment levels in an effort to improve credit ratings. DMI does not attempt to track employment for every establishment location, therefore employment counts may represent county or area totals for a given firm.
Harris Directory. The Harris Directory is published annually by Harris Publishing Company of Twinsburg, Ohio, and is available for selected Midwestern and Southern states. Most of the firms listed are manufacturing companies. Like DMI, the Harris Directory does not necessarily report employment by establishment location, therefore the employment count may represent a firm total. It may also contain some of the same problems as the other business directories listed.
Community Development Block Grant Information
The Community Development Block Grant (CDBG) program is funded by the Department of Housing and Urban Development and administered by communities that are receiving funding. This entitlement program provides annual grants to central cities of metropolitan statistical areas, other cities with populations of at least 50,000, and qualified urban counties with populations of at least 200,000. The purpose of the grants is to assist communities in carrying out a wide range of community development activities directed toward neighborhood revitalization, economic development, and the provision of improved community facilities and services. Specifically targeted are areas with high concentrations of low- and moderate-income residents. Each city allocates funding to projects it deems appropriate and consistent with HUD regulations and submits to HUD an annual report of funded activities. Activities include housing rehabilitation and new construction and improvements to or construction of public facilities, such as neighborhood centers, parks, streets, and health facilities.
The data are prepared and maintained by each community receiving the funding, usually by a local department of community development or planning. The data available from CDBG include name and address of the funding recipient, census tract and political boundary, description of the activity funded, activity codes, amount of funding, month and year of the activity, amount expended in a given period, geographic area served by the activity (census tract, political boundary, or city as a whole), and number of residents or households served. The racial and income characteristics of residents or households served by particular activities are also available. National objective codes indicate whether the area being served is considered low/moderate income, slum/blighted, or in urgent need of assistance.
Although the information is public record, the ease of obtaining and using it varies by community, as does its availability in electronic format. The data contain some ambiguities, particularly when the user is focusing on neighborhood analysis. For example, it is difficult to assign the financial benefit of some projects to a particular neighborhood, since activities may target multiple neighborhoods or a project based in one neighborhood may operate citywide. Such expenditures must either be apportioned across many neighborhoods or left out of neighborhood-level analyses.
Small area indicators that can be developed from these data include estimated CDBG funding by geographic area, estimated per capita funding by geographic area, types of activities being funded, and racial and income profiles of populations being assisted. The data have been used to study the impact of CDBG funding on residential rehabilitation and other community development activities in specific communities and to determine the investment being made in various CDBG activities.
Municipal Income Tax Records
Many states allow municipalities or counties to collect taxes on income earned by their residents and by nonresidents who work within their boundaries. These taxes are withheld by the employer and collected by the jurisdiction imposing the tax or by a central collection agency on its behalf. The information collected by the jurisdiction or collection agency includes the employer name and address, total amount of local taxes withheld from employees, taxes paid based on net profit of business, and federal identification number. Generally, the taxes are paid on a monthly or quarterly basis, although this varies by state. Confidentiality rules may make this information difficult to obtain in some localities. Typically, the tax collection agency produces aggregate data for selected neighborhoods in response to a special request.
Small area indicators that can be developed from these data include income tax generated by businesses in a geographic area. If appropriate confidentiality agreements can be crafted, it might be possible to link these data with other data sources, such as ES202 files, using federal identification numbers. This would allow additional indicators to be developed, such as income tax generated by particular industries (using SIC codes) and total wages and number of employees by geographic area and industry.
Training Program Records
Federally funded training programs generate data that can be used to determine levels of participation in these activities in local communities. Although the programs may undergo significant changes in the next few years as a result of block grants and welfare reform, their data bases are likely to continue to record the same basic information.
Job Opportunities and Basic Skills program information. From 1988 to 1997, Job Opportunities and Basic Skills (JOBS) was the federally mandated program aimed at helping families make the transition from welfare to self-support through job search, work experience, education, training, and other services. (It has since been superseded by the Temporary Assistance to Needy Families program.) JOBS was governed by federal and state regulations, and its data systems were operated primarily by state departments responsible for public assistance administration. Under welfare reform, states and local welfare-to-work initiatives are likely to maintain similar, perhaps improved, data bases.
The JOBS information is located in two files, one containing demographic information about public assistance recipients obtaining JOBS services and the other containing information about recipients’ activity histories in the JOBS program. The demographic information includes social security number, address (street, city, zip code, and county), race, sex, date of birth, educational level or school enrollment status, marital status, cash benefit from Aid to Families with Dependent Children, and length of time on assistance. Program information, updated monthly, includes a chronological history of an individual’s participation status, including eligibility, job assessment, job assignment, attendance record, exemptions (for long-term illness, very young children, etc.), failure to participate, sanctions, and employment history. The two files can be linked by social security number. Obtaining data may depend on the purpose for which it is sought, since the files contain confidential information about each JOBS participant.
Small area indicators that can be developed from the JOBS data include the average time between a job assessment and assignment to an activity, the participation rate among recipients, participation rates by JOBS activity, sanction rates, and employment rates of participants. Data from the JOBS program have been used to determine participation rates in welfare-to-work programs and, in conjunction with public assistance records, participation by long-term recipients in JOBS activities (Coulton, Verma, and Guo, 1996).
Job Training Partnership Act records. The Job Training Partnership Act of 1982 (JTPA) established the nation’s largest employment and training program for disadvantaged adults and youths facing serious barriers to employment. The goals of the program are to increase employment and earnings and reduce welfare dependence. Program participation is voluntary, but candidates must meet certain criteria to receive JTPA services. The program provides classroom vocational training, on-the-job training, job search assistance, and other related training services. It is administered by the states, with service delivery provided locally.
JTPA program data include participant information, such as social security number, birth date, sex, race, address, eligibility status, employment history and status, educational status, and participation in public assistance programs. Application date, termination date, and services provided are also recorded, as is information regarding the aptitude, ability, and skill level of each participant. Employment and public assistance status are recorded approximately three months after termination from the program. The data are collected and maintained by the state’s bureau of employment services and are available quarterly. Like the other employment related data files, restrictions apply regarding who may obtain these data and how they may be used.
Several small area indicators can be developed from JTPA data, including participation rates overall and by race and sex, percent of participants on welfare, and education and skills levels of participants. The data have been used to document participant demographics and outcomes, such as drop-out rates, employment rates, average wage when employed, and decreased reliance on welfare.
Data Sources Related to Community Safety and Security
Neighborhood safety and security is a major concern of almost all CCIs. Information regarding these issues can be found in several sources.
Municipal Police Records
Each police department maintains a record of each incident of crime reported in its jurisdiction. These records contain a significant amount of information about the crime, the victim, and, when available, the suspect or arrestee. The crime reports contain specific information, such as crime location, type of crime, time, date, weather conditions, and information about the arresting officer, including name and badge number. Types of crime include homicide, rape, aggravated assault, robbery, burglary, arson, auto theft, domestic violence, simple assault, menacing, and drug-related violations such as trafficking or possession. Some crime reports also contain a file of information about the victim, including race, sex, address, age, and date of birth.
Police departments also maintain an arrest data base. Included in each arrest report are address, race, sex, age, and date of birth of the arrestee. Information is also available about suspects, including geographic and demographic information and physical characteristics according to witness or victim descriptions. Information is also available regarding weapons used during incidents of crime. Crime reports can be linked to victim, arrest, and suspect reports using a report number.
Most police department data are available electronically and are released annually. Since this information is confidential and sensitive, the willingness of police departments to release these data may vary by jurisdiction.
Among the small area indicators that can be developed with these data are numbers and rates of crime by geographic area. Many researchers consider only serious crimes, called Part I crimes under the terminology of the Uniform Crime Report (UCR). Crimes can also be disaggregated by the race, sex, and gender of victims and assailants or by the victim-assailant relationship. For example, the reports can be used to identify crimes in which the victim and offender live in the same area or are of the same race. Weapon use by crime type can also be calculated. Crime data have been used to document crime levels in communities, to determine the need for violence prevention or community policing programs, and to understand possible causes and effects of crime.
Most police departments have adopted the UCR codes for crime reporting and standardization of offense definitions, which allow crime statistics to be compared across police jurisdictions. The UCR program compiles and maintains nationwide crime statistics, using data provided through the voluntary participation of local and state law enforcement agencies. The Federal Bureau of Investigation administers the program. Each year, the FBI releases a report entitled Crime in the United States, which contains data collected through the UCR program. The report provides data by state, county, and municipality (for cities and towns with 10,000 or more in population), including crime rates, number of crimes by type, and number of arrests by sex, race, and age. Crime in the United States can be found in libraries or purchased from the U.S. Government Printing Office. More information is available at http://www.fbi.gov.
911 System Data
Most 911 emergency systems are operated by county agencies, with 911 calls for emergency service processed by the appropriate police, fire, or emergency medical service department. Data available from 911 calls include the name, address, and phone number of the caller, although this information is not always complete because some callers are reluctant to identify themselves. Information about the emergency includes the exact location, date, and time of call, description of the emergency, whether an ambulance is required, and a priority and alarm level based on type of emergency. The time a call is received by the police, time of arrival at the scene, whether a contact is made, and a description of the result of the call are also available.
The data are organized by priority level, with 1 being the most serious and 4 being the least. The calls are also categorized according to whether the call indicates a crime against a person, an accident (involving, for example, hazardous waste), a danger to public safety (such as a bomb threat), a property crime, or general assistance (such as transporting a prisoner or assisting with a traffic stop). The data are available annually and may vary in format by jurisdiction.
Small area indicators that can be developed from these data include police response times and number of 911 calls by priority level. In addition, the calls can be categorized according to the description of the emergency. For example, the number of 911 calls indicating violent crime (such as homicide, robbery, or domestic violence), property crimes, or public safety issues can be calculated. Data about 911 calls can supplement other crime-related information. For example, some 911 calls are precipitated by incidents—such as altercations within households or minor disturbances—that do not result in crime reports. Data regarding response times can be useful to police departments and communities as a whole, since slow response times may indicate a need for more staffing during particular periods or in certain geographic areas.
Juvenile Court Records
The juvenile court handles cases of delinquency, unruliness, and dependency for all individuals under the age of eighteen. A record is maintained for each juvenile who enters the court system, including age, sex, race, date of birth, and census tract.
Offenses that come before the juvenile court include violent crimes, such as homicide and robbery; property crimes; drug violations; and less serious offenses, such as disorderly conduct, curfew violations, and truancy. Information collected by the court includes the location, date, and type of each offense, case number and type, source of the complaint (parent, school, etc.), judge, and disposition and disposition date. In addition, the records contain information regarding probation, such as probation officer, days on probation, and, where applicable, detention home location and release date.
The court maintains records of the addresses of offenders, victims, and offense locations, but researchers’ ability to obtain those data varies by court system. Access to demographic characteristics of victims also varies. Detailed information about individual victims and offenders, such as address, race, sex, and age, is confidential. The court may release an annual report aggregating data to the municipality or county level, but neighborhood indicators are seldom published.
Small area indicators that can be developed include delinquency rates, number and type of crimes committed by juveniles, and race, sex, and age of offenders. If victim information is obtained, the victim-offender relationship by sex, race, age, and geography can also be determined at an aggregate level. The data have been used to determine the level of juvenile crime and develop strategies and programs to reduce it. The federal Office of Juvenile Justice Prevention provides information at http://www.ncjrs.org/ojjdp/html/pubs.html.
The coroner determines the circumstances, manner, and cause of each violent, sudden, unusual, or unattended death and prepares a detailed report outlining the findings. Coroner’s reports contain confidential information about the victim and, when applicable, the assailant. Information about the victim includes age, sex, race, and address, along with any findings regarding drugs or alcohol in the victim’s system at the time of death. Information about the death includes place, date, time, day of the week, type of death (homicide, suicide, etc.), and mode of death (firearm, stabbing, etc.), and, where applicable, caliber of weapon. Information about the assailant includes sex, race, age, address, and previous offenses. The data are available from the local coroner’s office annually, but the format and rules for access may vary by jurisdiction. In some localities, for example, much of the report is not computerized.
Small area indicators that can be developed from these data include suicide and homicide rates, places of death, times and days when deaths occur, and how deaths occur. Information about victims and assailants, such as race and sex, and about victim-assailant relationships by age, race, sex, and geography can also be determined. Coroner’s reports are more detailed than police department crime reports and can be used in conjunction with those reports to enhance information regarding homicides and suicides.
Child Maltreatment Reports
Incidents of child abuse and/or neglect are reported to local child protection agencies, which investigate claims and determine whether abuse or neglect has occurred or is occurring. A record is maintained for each reported incident of abuse and/or neglect. The data available from reported cases of child abuse include location of alleged abuse and/or neglect, person making the report (teacher, doctor, victim, etc.), type of alleged abuse and/or neglect (sexual, physical, or emotional), and whether the allegation was substantiated, indicated, or unsubstantiated. Information about the victim is also available, including address, age, gender, and victim’s relationship to the perpetrator. Perpetrator information includes age and gender. The confidentiality and sensitivity of this information is an important issue, particularly at the address level. Obtaining these data depends on the purpose for which they are sought. Generally, the data are available annually and electronically.
Indicators that can be developed from these data include child abuse rates, types of abuse and/or neglect being reported, and numbers of substantiated and unsubstantiated incidents. The age and gender of victims can be determined, as can victims’ relationships with perpetrators of abuse and those reporting abuse. The data have been used to document child maltreatment and determine factors that may contribute to it.
The National Center on Child Abuse and Neglect, part of the U.S. Department of Health and Human Services, collects and maintains national and state-level information on abused and neglected children through its National Child Abuse and Neglect Data System (NCANDS), available at http://www.caliber3.calib.com/nccanch. The system does not release information regarding small areas. Participation by states is voluntary, but most states provide data to the system. The NCANDS annual report includes information regarding reporters of maltreatment; number of reports substantiated, indicated, and unsubstantiated; types of maltreatment; and perpetrator and victim characteristics, such as age, race, and gender.
Liquor License Records
In each state, the Department of Liquor Control is responsible for issuing permits to manufacture, sell, and distribute alcoholic beverages. Records of those permits are public information and include the name of the permit holder, address of permit location (street, city, zip code, county, and taxing district), and permit class (carryout beer only, wine only, etc.). The address of the outlet as listed on the permit is inaccurate in some cases, but the name of the permit holder may provide some clue to its true location.
A few small area indicators can be developed from these data, including the number of alcohol outlets in a geographic area, the number per capita, and the types of outlets, such as carryout stores or bars. Investigators can also determine the type of alcohol sold (beer, wine, hard liquor, or all three) and how late alcohol can be sold. Researchers have used these data to examine and document the relationship between the density of alcohol outlets and the level of violent crime in an area.
Education-Related Data Sources
Educational outcomes are important indicators of the well-being of a community and the functioning of its systems. CCIs often focus on children and young people since education is a fundamental component of development (Thorton, Love, and Meckstroth, 1994). Public school systems typically generate reports for the state or the community as a whole, but seldom produce data for small areas. Only limited data are available for private and parochial schools, which maintain separate and unique record systems.
Public School Records
Most public school districts maintain computerized files of individual student records. These records are confidential but, with proper protection agreements, can be used to develop measures for small areas. The files include each student’s address, school attended, school transfers or leavings, scores on standardized achievement and proficiency tests, attendance and disciplinary records, free lunch eligibility, and family status.
The data can be used to calculate attendance rates and average achievement for students by school or by neighborhood. School and residential mobility can be calculated by matching students’ records across years to determine movement. By matching records for a cohort of students—usually from the 8th grade onward—to determine those who graduate, school completion can be calculated. School entry records have also been used to document immunization status and school readiness. The nature of these records varies considerably across districts.
Head Start Records
Area Head Start agencies maintain records on children enrolled in Head Start and on individual Head Start programs. The files contain each child’s address, Head Start center location, enrollment date, and other selected family and child information. These records are confidential and their format and availability differ across local agencies.
Head Start records can be geo-coded and aggregated to calculate rates of Head Start enrollment in small areas and distances between children’s homes and the centers they attend. Matched with school enrollment files, Head Start records can be used to calculate preschool participation rates.
Health-Related Data Sources
Health is defined not merely as the absence of disease but as overall physical, mental, and social well-being. Although vast improvements have been made in the area of preventive health in the last 50 years, health indicators in many low income communities in the United States compare unfavorably with the overall high national standard of health (Geronimus, Bound, Waidmann, Hillemeier, and Burns, 1996). Well-established methods allow researchers to track many of those indicators using small area data (Gould, Mahajan, and Lucero, 1989).
The registration of births, deaths, fetal deaths, and other vital events is a state and local function. The civil laws of every state provide for a continuous, permanent, and compulsory vital registration system. The state vital statistics office issues certificates of live birth, fetal death, or death, either directly or through a local registrar, and typically compiles records of these events.
Birth information is available in two sections. The first section, known as the index portion, contains a unique birth certificate number, mother’s name, address, and other demographic and identifying information. Some localities include a census tract designation. The index portion is confidential and is released only upon approval of a special request justifying the need for such information. The other portion, called the statistical file, contains the birth certificate number along with such additional information as prenatal care, congenital anomalies, and birth weight. Some jurisdictions also include the census tract. The statistical portion is widely available for public health research. The death file consists of a unique death certificate number, name, social security number, and other indicators such as cause of death. The fetal death file has almost the same information as the birth file but also includes the cause of death.
Many small area indicators can be calculated from birth certificate data. Recorded birth weights can be analyzed to arrive at the number of low birth weight infants. Information about mothers’ prenatal care can be used to calculate the adequacy of prenatal care according to Kessner’s index, which considers number of gestation weeks, timing of entry into care, and total number of prenatal care visits (Kessner, Singer, Kalk, and Schlesinger, 1973). Death files can be analyzed to derive such information as leading causes of death and infant death rates, which can then be compared with local, state, and national standards. Excess mortality can be calculated by comparing age-specific deaths in the neighborhood with expected deaths based on a standard population (McCord and Freeman, 1990). Prenatal information can be analyzed from the fetal death file to identify small geographic areas of risk for such outcomes.
Most state and local governments publish vital statistics reports, some containing small area data. The state and local departments of vital statistics can be contacted for these reports. The National Center for Health Statistics (NCHS) publishes monthly and annual reports for the nation, states, counties, cities, and regions. Selected NCHS publications can be viewed at http://www.cdc.gov/nchswww/nchshome.htm.
Communicable Diseases Information
Physicians are required by law to report certain diseases, such as tuberculosis, syphilis, and AIDS, to local and state health officials. Records of these reports are maintained by departments of health, and in some cases are computerized. Health department data usually contain addresses, although for reasons of confidentiality these data can be released only if special justifications are made. Furthermore, small geographic areas may have incidences so small that no meaningful analysis can be carried out. Indicators that can be developed include incidence of disease in a particular area or within a specific population group.
Data about some communicable diseases are available from the World Health Organization at http://www.who.ch/programmes/emc/emc_home.htm. The Centers for Disease Control publishes information on communicable diseases in selected metropolitan areas at http://www.cdc.gov/publications.htm.
Emergency Medical Service Records
Emergency medical services are delivered with a sense of urgency to patients, such as accident or heart attack victims, in need of immediate attention. Information about patients transported via the public emergency medical service (EMS) system is recorded in 911 data. For an example of how these data are used in Durham County, North Carolina, see http://www.durhamems.com/Research.htm.
Many medical emergencies are treated in emergency rooms but do not appear as 911 calls. A more complete measure of these emergencies could come from emergency room records. In cities where local emergency rooms collaborate on an injury registry system, it is possible to calculate injury rates for small areas. Among the important indicators are rates of intentional and accidental injury by age group (Rivara, Calonge, and Thompson, 1989). Among young people in particular, injuries are a good indicator of health risk as well as social control in a community (Prothrow-Stith, 1991). Although hospitals in most cities do not maintain a common data base for emergency room visits, there is a considerable interest in injury surveillance (Centers for Disease Control, 1988). The availability of emergency service codes in the International Classification of Disease System makes it possible to establish data systems, and a growing number of communities are exploring such systems.
The immunization status of a population is considered an important measure of the adequacy of preventive health care, indicating not only the protection afforded by the vaccine itself but regular contact with a medical professional. Although states are not required to collect data on immunization, some states conduct surveys to estimate the number of children immunized, while some communities are experimenting with computerized immunization tracking systems. Local area school registration gives some indication of immunization status, since schools and day care centers are mandated to ask for proof of immunization before they admit students.
Some small area indicators can be developed from available data, such as the percentage of children without health insurance at the point of entry into school or who did not receive vaccinations at the appropriate age. The incidence of communicable diseases for which vaccines are available is another indicator of lack of immunization. More information is available from the Centers for Disease Control and Prevention at http://www.cdc.gov/nip/home.htm.
Claims filed by medical providers for services delivered under Medicaid—which provides health insurance coverage to low income individuals and families—may be a valuable source of data on health and medical services for populations in small geographic areas. Administration of the Medicaid program varies from state to state, and most states enter only a limited number of variables into the computerized system. Even so, available data usually includes provider description, classification of illness, procedure codes, service dates, and service charges. The recipient’s address, necessary for small area analysis, may need to be merged into the claim from an eligibility file. Researchers may need to file a special request justifying the need for the data, owing to confidentiality issues.
Several small area indicators can be developed using these data, such as annualized rates of medical care utilization by type (emergency, inpatient, ambulatory, etc.), patient health status and age, and diagnosis or procedure. National statistical information, compiled by the Health Care Financing Administration, can be viewed at http://www.hcfa.gov/medicaid/mcaidpti.htm. For additional national information, see http://www.census.gov/ftp/pub/prod/1/gen/95statab/health.pdf.
Claims data have been used extensively in health services research, but the formats in which they have been provided are likely to change under managed care. Managed care providers may be required to submit encounter forms to states for their Medicaid enrollees, but some states may accept aggregate reports of services provided for population groupings. This would preclude address-based small area analysis.
Hospital Discharge Files
Hospital discharge files are maintained by some state hospital associations and government agencies. Many state hospital associations maintain and publish data on patient age, payer, clinical service, sex, length of stay, diagnostic related group (DRG), hospitals, beds, and admissions, aggregated at the zip code level. Patient-level data exist, but confidentiality rules govern their release. "Hospital Statistics," published by the American Hospital Association, provides some hospital information with address data.
Small area indicators can be developed from these data, including average cost by severity of diagnosis, number of inpatients and outpatients, incidence of the most prevalent preventable conditions per 1,000 population, average length of stay, and number of beds and hospitals. Utilization rates have also been compiled across geographic areas using discharge data (Wennberg, Freeman, and Culp, 1987). The National Association of Health Data Organizations (NAHDA) addresses some data-related issues at http://www.nahdo.org/index.html.
Social Services Data Sources
Social services are public and private programs rendered to individuals and families to improve their economic, social, physical, and mental well-being. They are relevant to CCI evaluation, both because service reform is an important objective of many initiatives and because service provision and utilization are useful indicators of the status of the population in a small geographic area.
Public Assistance Files
Public assistance programs, most of which operate under state and federal law but are delivered locally, supply various forms of cash and in-kind assistance to eligible persons who qualify under means testing criteria. These programs include Temporary Assistance for Needy Families (which replaces the federal Aid to Families with Dependent Children program), food stamps, Medicaid, emergency assistance, and local general assistance.
Data on public assistance benefits are available through state or local departments of human services. Computerized individual records, including name and address, case and recipient numbers, program participation, eligibility status, and benefit amount, are contained in monthly files. Records can be extracted for assistance units or for individual recipients. A few states maintain longitudinal records, but in most places these need to be created by merging monthly records to create individual histories.
Public assistance files are confidential and can be released only for valid purposes, with proper protection agreements in place. Some departments can supply monthly files geo-coded by census tract, rather than name and address, reducing confidentiality problems. However, without recipient identifiers, longitudinal or matched files cannot be created.
Monthly files can be used to calculate participation in various public assistance programs by neighborhood residents. Longitudinal files can be used to calculate rates of long-term and short-term welfare participation. When public assistance records are merged with UI wage records, rates of moving from welfare to work in small geographic areas can be calculated.
Subsidized Day Care Records
Child care programs operating under the Personal Responsibility and Work Opportunity Act maintain records on children receiving day care subsidies, their families, and day care providers. These data are confidential but may be available through state or local departments of human services for valid research purposes with proper protections. Records include name of parent and child, address, eligibility status, service hours per week, estimated cost per week, at-risk indicator, family size, income, education, and caretaker. The records are organized by month and include children whose families qualify for day care subsidies under public assistance or low income working status.
Subsidized day care participation of families in small geographic areas can be calculated, along with the amount of the subsidies and the types of providers chosen. The research and publications department of the U.S. Department of Health and Human Services offers statistics on day care indicators at http://aspe.os.dhhs.gov/GB/sec12.txt.
Day Care Licenses
Every state requires child care providers to meet certain training and staffing criteria in order to obtain a license. Data on day care providers may be obtained through the state or local licensing agency or from local child care resource and referral agencies. Confidentiality guidelines limit the release of these data, but information without addresses or other identifiers can ordinarily be released to users who justify the need and guarantee protection.
Records contain the name and address of the licensee, numbers of slots for infants, toddlers, preschoolers, and school age children, and some additional information. Indicators that can be developed include numbers, types, and locations of slots relative to employment locations and homes of welfare recipients and working poor residents. Some information about day care can be found at http://www.careguide.net/.
Child Welfare Records
Child welfare services include preventive services such as social support, investigation of reported child abuse or neglect, services for abused and neglected children, crisis intervention, and other related services. Child welfare records are maintained by county and state departments of child welfare or child protection. These records are confidential.
Although federal requirements stipulate that child welfare information be computerized, the data vary in their completeness and accuracy. Records may include name, address, income, school, education, religion, ethnicity, marital status, and other demographic information. Dates and status of child abuse and neglect reports, entry into and exit from custody, foster care, residential treatment, protective services, and special programs are also important pieces of information for analysis.
Small geographic area indicators can be developed using geo-coded records. The number of children in custody or in foster care can be calculated, along with rates of reported child maltreatment. Child welfare records can be merged with other agency records to examine relationships among services provided by various child-serving agencies and to track outcomes (Goerge, Van Voorhis, and Lee, 1994). Although such matching and merging is a challenging task, it has the potential to produce information useful to CCIs in targeting local service reform. More information on child welfare data bases can be found at http://aspe.os.dhhs.gov/hsp/cyp/chapin1.pdf.
Mental Health, Alcohol, and Drug Services Information
Mental health, alcohol, and other drug abuse services are delivered by a wide range of public, nonprofit, and for-profit organizations. Although no single agency maintains data on all programs, those operated under public authority may generate data useful for developing small area measures. Local service providers often maintain confidential, computerized records that contain the client’s name, address, and dates and types of service, including admissions to inpatient facilities or treatment centers. The availability and format of these records vary considerably from one agency to another.
Small area indicators that can be developed from the geo-coded records include numbers and rates of residents under treatment, demographic characteristics of patients, prior treatment histories, criminal justice histories, social services, addiction severity, duration of treatment episodes, key services received, program staffing, ownership, resource base, and costs. National information is available through the Substance Abuse and Mental Health Services Administration (SAMHSA) at http://www.samhsa.gov/.
Data Sources on Community Resources and Participation
Building or rebuilding community infrastructure, capacity, and participation are among the goals of many CCIs. These phenomena are difficult to capture using administrative data, but a few possible sources are listed here.
Voter registration and participation records, maintained by local boards of elections, are open to the public but vary in their format and accessibility. The number of registered voters may be reported by ward or other political jurisdiction, but addresses can be used to calculate the rates of participation and registration within the boundaries of a CCI. Voter records have been used to measure participation by neighborhood and ethnicity; for an example of research tracking voter participation by Latinos, see http://naid.sppsr.ucla.edu/southwest/test1.html.
Membership in neighborhood organizations is another indicator that may be useful to CCIs, since membership growth may be an indicator of rising social participation. Although organizational membership records vary in quality and format, those organizations that actively recruit, collect dues, or provide services to members are most likely to have up-to-date, computerized records. If individual addresses are available, these records can be geo-coded to obtain counts of members for small geographic areas.
As community assets, community organizations are of considerable interest to CCIs. Unfortunately, no single data base provides information on organizations operating within a small area, yet some partial listings may be useful. For example, addresses in computerized yellow pages can be geo-coded and mapped by neighborhood. Other possible sources include lists of libraries, available from the American Library Association; churches, available from a local inter-church council; neighborhood development corporations and neighborhood centers, available from a local economic development agency; and parks and playgrounds, available from the local parks department. Although these lists can provide geographic locations of community assets, information about the magnitude of their operations or contributions may need to be obtained directly from the organizations. Information about computerized yellow pages is available at http://www.nctweb.com/cds/selphone.html.
Public Transit Information
Public transportation may offer some important indicators for CCIs, since civic participation depends on access to key locations outside the neighborhood and convenient transit stops within the neighborhood, particularly for non-drivers such as elderly, impoverished, and disabled residents. To measure access, evaluators can use bus and train schedules to calculate travel times to key destinations from the CCI neighborhood if computerized information is not available from local transit authorities. Transit stops can be geo-coded, mapped, and used in calculating distances from residents’ homes to stations. These data can also aid in calculating average commuting time to areas of employment growth, a crucial factor in economic development (Coulton, Verma, and Guo, 1996). In addition, average commutes to service providers might be a useful indicator for service improvement. Some national information is available from the American Public Transit Association at http://www.apta.com/.
Automobile Registrations and Licenses
Other indicators of a neighborhood’s access to regional services and its general level of resources are auto registrations and licensed drivers. Records of auto registration and licenses are open to the public and can be obtained for a fee from the state’s bureau of motor vehicles. Some states use dynamic data bases, updated as changes are made, thus allowing researchers to obtain a "snapshot" of all available data at a given point in time. Data are available with addresses and zip codes by type and classification (commercial, passenger, etc.) of vehicle registered, which can be used to assign census tracts using existing geo-coding software. Indicators that can be developed include number of annual registrations, per capita automobiles owned, and average vehicles owned by a family.
Issues and Challenges in Using Administrative Data
Using administrative data to construct small area measures presents a series of challenges. How successfully these problems can be overcome depends upon the type and source of the data and specific local circumstances. Some general issues may or may not become barriers in particular locales or situations.
Many administrative data sources contain individual information that is protected either by law or custom. Unless the data base already contains census block or tract codes, the CCI evaluator must request the release of confidential information about individuals’ street addresses to conduct small area analysis. Even information aggregated to the level of the census tract may breech confidentiality if only a handful of cases fall into a particular category and could thus be identified.
Administrative agencies can enter into confidentiality protection agreements with researchers who have a valid purpose for using the data to develop community measures. The researchers must follow standard methods for guarding data with identifiers, assuring that only necessary and secure personnel have access to the data and guaranteeing that confidential data will not be released. Researchers affiliated with institutions with federally approved institutional review boards should have their confidentiality protection methods reviewed by those bodies. Some agencies have well-established guidelines for releasing confidential data, while others have little experience in this area. However, with the exception of a few agencies that are strictly prohibited by law from releasing confidential data, most agencies can release data if the researchers are made agents of the agency and agree to abide by agency rules.
Community measures based on confidential data must be calculated for areas large enough to avoid revealing individual identities. This issue arises for categories based on rare events or small groups. For example, employment in a particular industry in a small area could be concentrated in one or two identifiable firms. If so, such figures could not be released.
The decision to release confidential information requires consideration of the risk-benefit ratio. The agency must judge the reputation of the research institution and its expertise in protecting human subjects to determine the risk that an inadvertent breech of confidentiality will occur. It should also weigh the benefit to itself and the community of making small area data available. Many agencies do not have the internal resources to look carefully at their data by small area. Thus, the benefits of releasing the data (under strong and binding confidentiality protections) for geographically based analysis is often significant if the agency is assured that information will be provided to them in return.
Small area information produced from administrative records suffers from four types of accuracy problems: inaccuracies in the records themselves, bias in reporting, small numbers in particular categories, and distortions related to averaging.
Inaccuracies can be a problem in any data base. Users of administrative data should check with the supplying agency about each data element and make a judgment about its accuracy and possible sources of inaccuracy. Many researchers have found that the most accurate data elements tend to be those that are essential to the agency’s work or subject to quality control. Thus, for example, public assistance payroll records stemming from the issuance of checks are more likely to be accurate than intake information that has no bearing on eligibility, such as educational attainment. Especially important to small area analysis is the accuracy of addresses for particular events that are being analyzed. Unfortunately, agency data bases often overwrite original addresses with address changes, thus eliminating the location current at the time of the event.
A second problem is reporting bias, which can influence the accuracy of records that are generated only when particular events are reported. For example, crimes are known to be underreported to the police (O’Brien, 1985), and law enforcement jurisdictions differ in their response to crime reports (Sherman, 1989). These two factors can affect whether a crime record is generated and how the crime is classified. Child abuse and neglect reports are vulnerable to similar problems (O’Toole, Turbett, and Nalpeka, 1983).
Third, accuracy can be affected when the number of particular events in an area is small. Infant deaths, for example, occur in very small numbers in a CCI during a given year. A change in even one death can raise the infant death rate markedly without reflecting a true change in health status of the population or quality of health services. The literature on sample size and accuracy suggests that, for rare events, evaluators can use multi-year averages or group neighborhoods together to achieve numbers from which estimates can be generated with confidence (Lemeshow, Hosmer, Klar, and Lwanga, 1990).
Population estimates for small areas are a final source of inaccuracy in developing measures from administrative data. Many indicators—such as crime incidence—are calculated as rates in order to make them comparable across small areas that differ in size. Also, although the decennial census is considered the best count of the population, its numbers are quickly outdated. Population estimates can be used for the years between censuses, but there are well-known problems with accurately estimating the population in small areas ( Smith and Cody, 1994). Thus, many measures are made even more inaccurate by errors in both the numerator and denominator.
Data Extraction and Management
Administrative records come in many different formats. Most easily useable by CCIs are those that have already been aggregated to the relevant units of geography. For example, many local health departments routinely produce counts of births, deaths, infant deaths, and other useful statistics by census tract. Most data, however, have not been converted to this format. Typically, administrative data files contain individual-level records, with information on a single individual appearing in multiple records across multiple files. Some files are extremely large and contain cases and records irrelevant to small area analysis. Considerable work often goes into understanding the file formats, extracting the relevant records, geo-coding the addresses, and aggregating the data to the required units of geography.
A complicating factor in using administrative data is assuring that the correct records have been extracted for the desired measure. Decisions must be made about the "window" of time to be considered, whether the unit is persons or events, whether to count all cases or new cases only, and how to handle duplicates. For example, a child maltreatment report is an event that involves one or more children. In a given year, the same child may be reported multiple times, or a single event may yield several reports. Child maltreatment cases may be carried as open records in the agency data base over several months or years. Such data make possible several different measures for a small geographic area, including the total number of maltreatment reports in a year, individual children reported as maltreated at least once in a year, maltreatment cases served by the agency at a point in time during the year, and maltreated children ever served during the year by the agency. Researchers need to be clear about exactly how their calculations are made and what the resulting measures mean.
Matched and Longitudinal Files
Administrative data are often organized by month, quarter, or year. Most data bases are event driven, generating a record, for example, when a person is eligible for a program, a payment is made, a deed transfers, or a child is born. However, CCIs may require some measures that reflect that these events happen over time to an individual, building, firm, or some other entity. To develop this type of measure, longitudinal records must be created by matching events across separate records using a constant identifier such as a parcel number or a case number.
Although longitudinal measures require considerably more effort to process than individual records, they are frequently more reflective of important outcomes. For example, a CCI that was less interested in reducing public assistance use than in eliminating long-term dependency might want to calculate the number of long-term welfare recipients in a small area by matching monthly eligibility files for each individual over a number of years.
A similar challenge occurs when measures require that data be matched across multiple agencies or multiple data sets from a single agency. For example, a CCI may be interested in assuring that preschool children of mothers moving from welfare to work are still able to take advantage of Head Start programs in the neighborhood. A match would need to be made between public assistance, employment, and Head Start records to monitor progress on this outcome.
Matching across multiple agencies may require the use of probabilistic matching procedures when there is no universal or reliable individual identifier (Jaro, 1995; Newcombe, 1988). For example, social security numbers are erroneous or not available in many data files, but names, birth dates, addresses, and other identifiers can be used to improve the accuracy of matching. Matching across data bases has the potential to create sensitive and refined measures, which could be useful in capturing the synergistic effects of CCIs (Goerge, Van Voorhis, and Lee, 1994).
Census Products, Surveys and Related Sources
CCI evaluators can also find useful information in data generated by the U.S. Bureau of the Census and by special surveys and censuses on economic, housing, and other issues relevant to community regeneration.
U.S. Census of Housing and Population Data
In its decennial census, the U.S. Bureau of the Census strives for complete coverage of housing stock and population and provides data down to the level of census tracts for almost all characteristics measured and to the level of block groups for many characteristics. The major drawback of census data is that they are gathered only once every ten years.
The census asks two levels of questions: the short form (sometimes referred to as the "100 percent"), which contains 7 basic population and 7 housing questions pertaining to each person and housing unit; and the long form (sometimes referred to as the "sample"), which covers 26 population and 19 housing items, answered by approximately one in six households (17 percent) nationally. Census population data are available in two major forms: summary tape files and public use microdata samples. Information on the data files can be found at http://www.census.gov.
Summary Tape Files (STFs)
Summary tape files provide cross-tabulations of characteristics identified for specified geographic areas, usually in finer detail than printed census reports. These machine readable collections of summary statistics are available on computer tape; certain STFs are also available on microfiche or CD-ROM.
The STF3A is probably the most widely used file in this group because it contains long form responses regarding employment, income, trip to work, duration of residence, and housing characteristics, tabulated to the census tract level and sometimes to the block group level. It should be remembered, however, that the sample of approximately 17 percent of households is subject to some level of sampling error.
STF1 files. This collection of 100 percent, or short form, summary statistics covers congressional districts (101st Congress), counties, county subdivisions and places, census tracts and block numbering areas (BNAs, roughly equivalent to census tracts in areas where no tracts have been designated), and block groups and is available on microfiche and CD-ROM. STF1B, which covers counties, county subdivisions and places, census tracts and BNAs, block groups, and blocks, also covers metropolitan areas and urbanized areas. It is available on microfiche and CD-ROM, with only partial data for blocks. STF1C covers counties, places, and (in selected states) county subdivisions of population greater than 10,000 and metropolitan areas and urbanized areas. It is available on CD-ROM. STF1D covers congressional districts, counties, and places and (in selected sates) county subdivisions of population greater than 10,000.
STF2 files. These contain a more detailed collection of 100 percent summary statistics. STF2A covers counties, places of population greater than 10,000, and census tracts and BNAs. STF2B covers counties, places of population greater than 1,000, and county subdivisions. STF2C covers counties, places, and (in selected states) county subdivisions of population of greater than 10,000, along with metropolitan areas and urbanized areas and all county subdivisions of New England metropolitan areas.
STF3 files. These files contain a less detailed collection of sample, or long form, summary statistics. STF3A covers counties, county subdivisions and places, census tracts and BNAs, and block groups and is available on microfiche and CD-ROM. STF3B, covering five-digit zip codes, is available on CD-ROM. STF3C covers counties, places, and (in selected states) county subdivisions of population greater than 10,000, and metropolitan areas and urbanized areas. It is available on CD-ROM. STF 3D covers congressional districts, counties, and places and (in selected states) county subdivisions of population greater than 10,000.
STF4 files. This collection includes a more detailed set of sample summary statistics. STF4A covers counties, places with population greater than 10,000, and census tracts and BNAs. STF4B covers counties, places, and (in selected states) county subdivisions of population greater than 2,500; it also covers all county subdivisions in New England metropolitan areas. STF4C covers counties, places, and (in selected states) county subdivisions of population greater than 10,000, and metropolitan areas and urbanized areas.
Public Use Microdata Sample (PUMS)
These files contain individual microdata taken from the 1990 census long form samples. Each record includes essentially all the 1990 census data collected about each person in a sample household and the characteristics of the housing unit. Unlike the STF files, which include fixed cross-tabulations for a given geographic area, these files enable users to prepare their own cross-tabulations or conduct multivariate analysis. In order to preserve confidentiality, locations are indicated in PUMS designated areas, identified as areas containing at least 100,000 population and not crossing state lines. The data are available in two files, one containing a 5 percent sample of housing units where location can be groups of counties, a single county, a place, or a grouping of census tracts, and the other containing a 1 percent sample of housing units with location indicated by metropolitan area or other large area, the boundaries of which may cross state lines.
Additional Data Sets Based on the Decennial Census
Some specialized data sets developed from 1990 census data are also available.
County-to-County Migration File. Based on the 1990 census, this file gives summary descriptions of intrastate and significant interstate county-to-county migration streams, including counties of origin and destination and characteristics of members of the migration stream. It is available on CD-ROM.
Commuting Zones and Labor Market Areas. Developed from 1990 journey-to-work data, these units include 741 commuting zones (CZs), delineated for all U.S. counties and county equivalents, and 394 labor market areas (LMAs), aggregated from the commuting zones to meet the Bureau of Census criterion of 100,000 population minimum. CZs and LMAs can be used as geographic boundaries and combined with other data sources to track changes over time in economic activity. A CD-ROM and information about the methodology are available at http://www.lapop.lse.edu.
1990 Census Transportation Planning Package. This set of special tabulations of 1990 census data is tailored to meet the needs of transportation planners. It has two elements (statewide and urban) and three parts (residence, workplace, and journey-to-work). An urban element data set was created for each metropolitan planning organization across the country according to formats specified locally (for example, a customized geographic area or standard census tracts and block groups). It includes tabulations by area of work and area of residence, mode of travel to work, type of work, and time to get to work. It is available on CD-ROM through the U.S Department of Transportation at http://www.bts.gov.
American Housing Survey Information
The American Housing Survey (AHS) collects data on housing, including apartments, single-family homes, mobile homes, vacant housing units, household characteristics, income, housing and neighborhood quality, housing costs, equipment and fuels, size of housing unit, and recent movers. National data are collected every other year, with a sample on average of 55,000 homes. Data from each of 47 selected metropolitan areas are collected about every four years, with an average of 12 metropolitan areas included each year. The sample in each metropolitan area is 2,500 or more homes.
The AHS returns to the same housing units year after year to gather data, adding new units identified from permit listings. The data therefore provides a picture of household flows through a constant representative set of housing units. Because of sample size limits, however, valid estimates are possible only down to geographic units of 100,000 or more (like the PUMS data), even in metropolitan areas covered by the area surveys. Even so, by providing an indication of how housing stock and residents are changing in the broader context, the survey can be valuable to evaluators interested in smaller areas within the 47 designated metropolitan areas. When using these data to examine income or poverty of residents, evaluators should note that the AHS has historically under-reported income and over-reported poverty compared with the Current Population Survey, and that both surveys tend to under-report income and over-report poverty compared with tax returns and national income accounts. Information on the survey can be found at http://www.census.gov/hhes/housing.
1992 U.S. Economic Census Data
Gathered every five years since 1967, the U.S. Economic Census covers specific industry types: retail trade; wholesale trade; service industries; construction industries; manufacturers; mineral industries; financial, insurance, and real estate industries; transportation; communication; and utilities. Reports on the various industries give number of establishments, number of employees, revenue, and payroll, as well as some industry-specific information. Most are available on CD-ROM and can be converted into data files. However, confidentiality concerns dictate a fairly high level of geographic aggregation, usually by metropolitan area or metropolitan area and county and occasionally by zip code.
County Business Patterns Series
This annual series reports the number of establishments by employment size, industry group (agriculture, forestry, and fishing; mining; construction; manufacturing; transportation and utilities; wholesale trade; and retail; finance, insurance, and real estate), and payroll. The geographic unit is counties, but zip code data are also available. County Business Patterns are based primarily on administrative records and reports from current surveys, unlike the 1992 Economic Census, which is based on responses from individual establishments. These data are available on CD-ROM. Information is available at http:/www.census.gov/epcd/cbp.
Consolidated Federal Funds Report, Volume 1, County Areas
Consolidated Federal Funds Report (CFFR) data are obtained from federal government agencies and published in an annual series covering federal expenditures or obligations in the following categories: grants, salaries and wages, procurement contracts, direct payments to individuals, other direct payments, direct loans, guaranteed or insured loans, and insurance. Dollar amounts represent either actual expenditures or obligations. For more information on CFFR data, see http://www.census.gov/govs/www/cffr.html.
County and City Data Book
This source contains data from 1987 through 1992, including more than 220 data items for states and counties, almost 200 data items for cities, and 33 items for places of 2,500 population or more. Items include age, money and personal income, population, education, health care and human services, housing ownership and value, births, deaths, poverty, local government finance, employment, business, banking, climate, elections, and social programs. It is available on CD-ROM. Information on the data is available at http://www.census.gov/stat_abstract/ccdb.html.
Regional Economic Information System
This series, updated periodically, presents annual estimates of local area economic data for states, counties, and metropolitan areas for the years 1969-94. Statistics in the data base include personal income and earning variables, full and part employment variables, transfer payment variables, and farm income and expenses variables. Some breakdowns are given by industry, or SIC code. The data estimates are derived from a number of sources, many of which are used for the National Income and Product Accounts estimates. For example, earnings and employment estimates are derived mostly from the Bureau of Labor Statistics ES-202 series. Earnings are estimated by both place of work and place of residence. The data are available on CD-ROM. For further information, see http://www.bea.doc.gov or http://www.lib.virginia.edu/soscsci/reis/reis1.html.
Federal-State Cooperative Program for Population Estimates
This program generates several series of population estimates, including annual estimates of county populations and biannual estimates of city or place populations. Each set starts from 1990 census figures, then estimates components of change in population from birth and death records, domestic migration (based on federal income tax returns), international migration (based on statements to the Immigration and Naturalization Service by legal immigrants and refugees), and undocumented immigrants (based on 1990 census figures on recent foreign born immigrants to specified places). Because federal program allocations are based on these estimates, cities and places often challenge the estimates, and periodic updates reflect those challenges and the reconciliation process specified by law. Although the estimates are carefully made, caution should be used in applying them to small areas and especially to subgroups by age, race, and sex. For example, one study using these estimates as the numerator in a calculation of nonwhite teen birthrates at the county level found a rate in excess of 100 percent in 7 percent of the county-years in the sample (Kane and Staiger, 1996). The estimates can be downloaded from http://www.census.gov/population/www/estimates/popest.html.
Small Area Income and Poverty Estimates
The Bureau of the Census recently initiated a program to update selected income and poverty estimates at the state and country levels. Estimates are made of median household income, per capita income, number of persons below the poverty level, number of children under age 5 below the poverty level, number of children 5 to 17 years old below the poverty level, and number of persons age 65 and over below the poverty level. At a later stage, the project may attempt estimates of children in poverty at the school district level. A committee of the National Academy of Sciences is monitoring this project and advising on methodology. State and county estimates for 1993 are available at http://www.census.gov/hhes/www/saipe/saipe93/ftp93.html.
U.S. Department of Education Data
Some information about local school districts and schoolchildren is compiled by the federal Department of Education and made available to the public.
The School District Data Book
The School District Data Book (SDDB) is an education database and information system that contains the most extensive available set of data on children, their households, and the nation’s school systems. The SDDB includes demographic and cartographic CD-ROMs and provides up to 200,000 data items for school districts, including detailed information from the 1990 census school district special tabulation for states, counties, and districts. Its mapping features enable users to view maps of all individual school districts in the nation. Information is available at http://www.ed.gov/NCES/surveys/SDDB/introd.htm.
The Common Core Data
The Common Core Data (CCD) is a comprehensive, annual, national statistical database of all public elementary and secondary schools and school districts, which contains data that are comparable across all states. CCD presents three categories of information: general descriptive information on schools and school districts; data on students and staff; and fiscal data. The general descriptive information includes name, address, phone number, and type of locale; information on students and staff includes demographic characteristics; and fiscal information covers revenues and current expenditures.
The CCD is made up of a set of five surveys sent to state education departments. Most data are obtained from administrative records maintained by the state education agencies. Statistical information is collected annually from public elementary and secondary schools (approximately 87,000), public school districts (approximately 16,000), and the 50 states, the District of Columbia, and outlying areas. For further information see http://www.ed.gov/NCES/ccd/index.html.
Geographical Information Systems
A geographical information system (GIS) is a computer program that allows the user to organize and analyze information geographically. Nearly all the information discussed in this paper is about specific places and can be tracked to specific addresses, census tracts, wards, or neighborhoods. The major use of GIS programs is to take these kinds of data and create visually striking and informative thematic maps, some of which brand or shade areas of a territory along a single variable (such as percent of the population over age 65), while others highlight specific addresses or events (such as crime incidents).
GIS programs are also useful in analyzing data. Mapping programs allow the user to overlay one thematic map on top of another, showing, for instance, the relationship between concentration of minority population and delivery of public services. GIS programs can also provide descriptive statistics about geographical areas.
All GIS programs plot geographical data. To do so, however, GIS programs require electronic base maps, including street and address ranges, census tracts, blocks and block groups, and other common geographic areas such as wards or zip codes. There are a number of ways to get these base maps, either by purchasing the maps from a private vendor (which can be expensive) or converting the U.S. Census Bureau TIGER files into maps that can be used by a specific GIS program. The TIGER files are available at many libraries and universities, but converting them to a format that can be read by a GIS program requires additional software.
There are several GIS software programs available for the microcomputer. These include ArcInfo, ArcView, Mapinfo, and MapQuest. The two most accessible to new users are Mapinfo (available at 1-800-FASTMAP, or www.mapinfo.com) and ArcView. Both allow researchers to use geographic information, geo-code and manipulate data, and create thematic maps for presentation.
Once a user becomes familiar with these programs, they are very easy to work with, although it is important to add that ease of use neither assures accuracy nor prevents the creation of distorted maps. The accuracy of geo-coding, for instance, is dependent upon the quality of the street and address ranges of the base maps being used. It is worth noting that the TIGER files have significant gaps, particularly in quickly developing areas. Geo-coding based on TIGER files will likely produce a high number of addresses that cannot be accurately assigned X and Y coordinates. Also, many GIS programs have programmed a number of defaults in the creation of thematic maps, which can easily lead to the creation of maps that provide a distorted picture of the data.
Despite the ease of use features of all available GIS programs, new users will frequently find that they have questions or unique problems. Fortunately, many colleges and universities teach courses on GIS mapping, and many have GIS labs. The staff of a geography department or GIS lab at a local university is a good place to ask for assistance and training. Indeed, it might be advisable to find out what GIS software is being used by local educational institutions before making a purchase.
Building Local Capacity for Small Area Information
CCIs are not the only members of their communities that can benefit from ready access to information about neighborhoods or other sub-city geographic units. Indeed, there is growing recognition of the need for this information, and growing capacity to use it, among government agencies, advocacy groups, planning councils, and neighborhood residents (Urban Institute, 1996; Sawicki and Craig, 1996).
Neighborhood information systems now exist in Cleveland, Providence, Denver, Oakland, Boston, Atlanta, and a few other cities. In each place, one or more local organizations has undertaken the development of neighborhood measures across a comprehensive range of topics and has agreed to assist local organizations in using that information to guide action. A common theme among these efforts is that neighborhood information is an essential element of community building. A community cannot truly create a responsible and responsive agenda for change without knowing a lot about itself.
Existing neighborhood information programs are run by local organizations that are either independent or part of universities or foundations. Each has built strong relationships with the agencies that provide the data, obtaining and processing information continuously, turning it into neighborhood measures or indicators, and returning geographic analyses to the agencies that supplied the data. The programs vary in how they disseminate the neighborhood information. Cleveland, for example, makes it available on line for census tracts, neighborhoods, and municipalities (Chow and Coulton, 1996). Other programs make the information available as part of a community planning or action agenda and to individual groups or organizations upon request.
Neighborhood information capacity is vital to CCIs because they can seldom afford to obtain and process large amounts of available data just for the few neighborhoods they target. There is an economy of scale when measures are created for all neighborhoods in a county or region. For example, finding birth certificates from a single neighborhood in the vital records file and calculating a low birthweight rate for that neighborhood would cost almost as much as conducting the same activity for all of the neighborhoods in a city.
The experience gained in setting up these neighborhood information capacities is now becoming available to other cities. The Urban Institute’s National Neighborhood Indicators Project is coordinating such an effort; for information, see http://www.urban.org. Examples of comprehensive neighborhood indicator programs are also available for Cleveland from Case Western Reserve’s Center on Urban Poverty and Social Change, at http://www.cwru.edu/CWRU/Dept/MSASS/poverty/cupsc.htm, or for Providence at http://www.providenceplan.org.
Using Small Area Information in CCIs
CCIs have unique needs for information for several reasons. First, they evolve over extended periods of time, so their information needs are ongoing and dynamic. Second, they are comprehensive, so their information needs range across sectors. Third, they attempt to change entire communities, so they need information pertaining to the communities’ residents, organizations, systems, physical conditions, social structures, and economies. Finally, they are action oriented, so that they need information that is timely but consumes only modest resources for data gathering and analysis.
Using available data is consistent with these unique information needs. Collected regularly and periodically and stored so that retrospective baseline information can be created, these data allow evaluators to examine trends and dynamics of change and support the creation and analysis of community change. Because the data are gathered for other purposes, they do not require the effort and expense of original data collection.
However, there is a danger that CCIs can drown in available data if the measures sought are not part of a carefully constructed theory of change. Available data have the potential to measure early, interim, and ultimate outcomes along a well-considered pathway. In this respect, it is essential that the time frame during which the available data are collected be carefully linked to the timing of change anticipated by the theory. Further, the measures will be much more meaningful if the theory specifies thresholds for the amount of change anticipated by the theory.
The ready availability of some but not other types of data can also lead to the light post fallacy—that is, looking for change where it is easiest to observe. CCIs generally anticipate many outcomes that cannot be measured with available data, especially those that relate to community perceptions or processes. Nevertheless, the available data sources reviewed in this chapter are key resources for the totality of information CCIs need for planning, action, and evaluation. Today’s information technology makes these data an important and practical tool, deserving serious methodological attention and further development.
This paper was prepared for the Aspen Institute’s Roundtable on Comprehensive Initiatives. The assistance of Lisa Nelson and Venky Chakravarthy is gratefully acknowledged.
Annie E. Casey Foundation. 1997. City Kids Count. Baltimore: Annie E. Casey Foundation.
Carlson, Virginia L. 1995. "Identifying Neighborhood Businesses: A Comparison of Business Listings." Economic Development Quarterly 9: 50-59.
Centers for Disease Control. 1988. "Public Health Surveillance of 1990 Injury Control Objectives for the Nation." Morbidity and Mortality Weekly Report 37: 27-68.
Chapin Hall Center for Children. 1996. Creation of a Community Information Infrastructure. Chicago.
Chaskin, Robert. 1995. Defining Neighborhoods: History, Theory and Practice. Chicago: Chapin Hall Center for Children.
Chow, J., and Claudia Coulton. 1996. "Strategic Use of a Community Data Base for Planning and Practice." Computers in Human Services 13: 57-72.
Coulton, Claudia, Nandita Verma, and Shenyang Guo. 1996. Time-Limited Welfare and the Employment Prospects of AFDC Recipients in Cuyahoga County. Cleveland: Case Western Reserve University.
Coulton, Claudia, Jill Korbin, Tsui Chan, Marilyn Su, and Edward Wang. 1997. Mapping Resident Perceptions of Neighborhood Boundaries. Cleveland: Case Western Reserve University.
Geronimus, Arline T., John Bound, Timothy A. Waidmann, Marianne M. Hillemeier, and P. Burns. 1996. "Excess Mortality among Blacks and Whites in the United States." New England Journal of Medicine 14: 491-98.
Goerge, Robert, John Van Voorhis, and B. J. Lee. 1994. "Illinois Longitudinal and Relational Child and Family Research Database." Social Science Computer Review 12: 351-65.
Gould, J. B., N. Mahajan, and M. Lucero. 1989. "Improving Perinatal Outcome through Data Management: The Design of Small Area Analysis Systems." Journal of Perinatal Medicine 16: 305-14.
Jaro, M. A. 1995. "Probabilistic Linkage of Large Public Health Data Files." Statistics in Medicine 14: 491-98.
Kane, Thomas, and Douglas Staiger. 1996. "Teen Motherhood and Abortion Access." Quarterly Journal of Economics: 478.
Kessner, D. M., J. Singer, C. E. Kalk, and S. Schlesinger. 1973. Infant Death: An Analysis by Maternal Risk and Health Care. Washington, DC: National Academy of Sciences.
Leete, Laura, and Neil Bania. 1995. Assessment of the Geographic Distribution and Skill Requirements of Jobs in the Cleveland-Akron Metropolitan Area. Cleveland: Case Western Reserve University.
Lemeshow, S., D. W. Hosmer, J. Klar, and S. K. Lwanga. 1990. Adequacy of Sample Size in Health Studies. New York: John Wiley and Sons.
McCord, C., and H. P. Freeman. 1990. "Excess Mortality in Harlem." New England Journal of Medicine 322: 173-77.
Newcombe, Howard B. 1988. Handbook of Record Linkage: Methods for Health and Statistical Studies, Administration, and Business. Oxford: Oxford University Press.
O’Brien, R. M. 1985. Crime and Victimization Data. Beverly Hills, CA: Sage Publications.
O’Toole, R., P. Turbett, and C. Nalpeka. 1983. "Theories, Professional Knowledge, and Diagnosis of Child Abuse, in The Dark Side of Families: Current Family Violence Research, eds. D. Finkelhor, R. Gelles, G. Hotaling, and M. Strauss. Newbury Park, CA: Sage Publications.
Prothrow-Stith, Deborah. 1991. Deadly Consequences. New York: Harper-Collins.
Rossi, P. H. 1972. "Community Social Indicators," in The Human Meaning of Social Change, eds. A. Campbell and P. E. Converse. New York: Russell Sage.
Rivara, F. P., N. Calonge, and R. S. Thompson. 1989. "Population-Based Study of Unintentional Injury Incidence and Impact during Childhood." American Journal of Public Health 79: 990-94.
Sawicki, David S., and W. J. Craig. 1996. "Neighborhood Indicators: A Review of the Literature and an Assessment of Conceptual and Methodological Issues." Journal of the American Planning Association 62: 240-53.
Schorr, Lisbeth. 1994. "The Case for Shifting to Results-Based Accountability with a Start-up List of Outcome Measures with Annotations." Cambridge, MA: Harvard Project on Effective Services.
Sherman, L. 1989. "Hot Spots of Predatory Crime: Routing Activities and the Criminology of Place." Criminology 27: 27-56.
Smith, S. K., and S. Cody. 1994. "Evaluating the Housing Unit Method." American Planning Association 60: 209-21.
Thornton, C., J. Love, A. Meckstroth. 1994. "Community-Level Measures for Assessing the Status of Children and Families." Paper presented at the annual research conference of the Association for Public Policy and Management, Chicago.
Waits, M. J., T. Rex, and R. Melnick. 1997. "Cluster Analysis: A New Tool for Understanding the Role of the Inner City in a Regional Economy." Paper presented at the Community Outreach Partnership Conference, U.S. Department of Housing and Urban Development, Phoenix.
Wennberg, John E., Jean L. Freeman, and William J. Culp. 1987. "Are Hospital Services Rationed in New Haven or Over-Utilized in Boston?" Lancet: 1185-88.
White, Sammis B., John F. Zipp, and William F. McMahon. 1990. "ES202: The Data Base for Local Employment Analysis." Economic Development Quarterly.
Urban Institute. 1996. Democratizing Information. Washington, DC: Urban Institute.
Back to New Approaches to Evaluating Community Initiatives index.
Copyright © 1999 by The Aspen Institute
Comments, questions or suggestions? E-mail firstname.lastname@example.org.
This page designed, hosted, and maintained by Change Communications.