Mecklenburg County Linked Database

By Human Services Planning and Evaluation, Mecklenburg County,
Department of Social Services

Prior to the ASPE linked database grant, Mecklenburg County had a local Automated Income Maintenance System (AIMS) and access to the State Eligibility Information System, Food Stamps Information System, and State Employment Programs Information System.  These databases allowed us to gather statistics on families in Work First, North Carolina’s Temporary Assistance for Needy Families (TANF) program, but these databases could only tell us about caseloads in a given month.  Now we have longitudinal data that allows us to track our Work First families from January 1995 through the present.  The longitudinal TANF data can be joined to various other State data such as Employment Programs, Food Stamps, and Employment Security Commission files.  These linkages provide a more comprehensive view of clients’ circumstances.

Capability Achieved

The longitudinal, linked database consists of multiple components.  These components come mostly from state social services information systems.  Below is a table that shows these components, and their primary functions:

Prior to the grant, none of the data were available in longitudinal form.  In addition to the capability of tracking TANF and Food Stamps recipients over time, we now have databases that were previously unavailable to us, even in cross-sectional form, such as the ESC database and the purchase of care file.  Also unavailable to us in longitudinal form before the grant were Employment Services activities and employment data, time limit data, and child-only case data.

Possible Analyses with the Longitudinal Database

• Changing size of Work First caseload over time.
• Comparisons of rates of exit and recidivism among various subgroups (e.g. rates of exit and recidivism for whites versus non-whites, men versus women)
• Comparisons of quarterly earnings among various subgroups (e.g. earnings of TANF recipients who participated in Employment Services versus those who did not)
• Comparisons of long-term "core" TANF recipients to "successful customers" who leave TANF and do not return.
• Evaluation of the number of former TANF recipients who are currently working.
• Evaluation of consistency of employment after leaving TANF (i.e. how many quarters did customers work when they left WF?)
• Number of present and former TANF recipients who also receive food stamps.
• Number of TANF recipients who have used all months against the state 24- or federal 60-month time limits.
• Number of TANF recipients who are likely to use all of their 24-month time limits in the next 6 months.
• Descriptive profile of child-only cases.
• Changes in the average duration in Work First after the implementation of 24-month time limits.  (i.e. was the average duration longer before time limits?)
• Factors explaining variation in duration in Work First.
• Factors explaining variation in post-TANF employment earnings.
• Comparison of scheduled hours of employment services activities to completed hours of employment services activities.
• Analysis of industries that hire TANF recipients and earnings within these industries.
• Effects of sanctions on duration in Work First.
• Differences in duration in Work First for two-parent cases versus single-parent cases.
• Geographic concentrations of Work First customers in the county.

Implementation Issues

Mecklenburg County’s partnership with the Jordan Institute for Families at the UNC Chapel Hill School of Social Work facilitated the data collection process.  The Jordan Institute has been working with extracts from state administrative systems for nearly a decade under contracts with the state to evaluate social services programs.  Examples of these contracts are the Evaluation of North Carolina Job Opportunities and Basic Skills program (the predecessor of Work First Employment Services), Management Assistance for the Work First Program, and Evaluation of Family to Family: Reconstructing Family Foster Care.  State data was used to create longitudinal files for all of these projects, demonstrating that the Jordan Institute and the state of North Carolina have an established history of sharing information as early as 1990.

In this contract with the state Division of Social Services, a confidentiality agreement states that the Jordan Institute is bound to protect the confidentiality of any and all data to which they receive access.  The Jordan Institute was permitted to provide Mecklenburg County extracts of the data through an informal agreement with the contract manager at the Division.  Therefore, a foundation for sharing Work First data, Food Stamps data, and Employment Security Commission Earnings data had already been laid at the time the grant phase began.  Mecklenburg County had no direct relationship with the state on matters of data sharing.

Attempts were made to collect data on subsidized childcare, child support collections, and Medicaid.  Unfortunately, these data are currently unavailable to us.  Mecklenburg County has a partial extract of subsidized childcare data.  However, the data is only available for six months, and has not yielded any meaningful analyses.  The reason for the delay in obtaining the data in full is that Mecklenburg County’s direct partners at the Jordan Institute do no extract the childcare data themselves.  Another party at the Jordan Institute has been extracting the data for separate purchase of care research.  Because the data is being extracted for purposes other than research for Mecklenburg County, we have only been able to obtain a sample of purchase of care data.  Once the party has completed constructing, cleaning, and working with the database, it will be given to our partners, who, in turn, will give it to us.  A more complete childcare database is being constructed, and should be available next month.

Data from child support collections also remains unavailable to us, largely due to confidentiality concerns.  However, there is interest at the state level in obtaining this data.  We believe that eventually Child Support Enforcement will be persuaded by state officials to share the requested data.

Medicaid information is also currently unavailable.  We originally believed that this information would be included in the TANF case file.  However, there is no indicator as to which families receive Medicaid once they leave the Work First program.  The Jordan Institute now has data from the state Medicaid system and is working to create a longitudinal Medicaid database.  This information should be available within the next year.

Summary of Completed Research and Analysis

One of the major pieces of research completed with Mecklenburg County’s longitudinal database was a descriptive profile of both heavily dependent customers and customers who left Work First and never returned.  These two customer groups were compared on various attributes such as demographics, employability, earnings, employment industries, and usage of other service programs such as food stamps and employment services.  The results answered questions often asked by local program administrators:  "What do our hardest to serve customers look like?" and "Who are the Work First success stories?"

Other key analyses of Work First customers include tracking of reported employment earnings for various exit cohorts, rates of exit and recidivism for selected entry and exit cohorts, differences in rates of exit and recidivism by race, and a descriptive analysis of child only cases in a given month.

The data has also been used for operational purposes at Mecklenburg Social Services.  Work First administrators, for example, have requested lists of names of TANF recipients not yet served by Work First Employment Services who may need additional support from community agencies.  The database has also been used to create samples for contracted research projects such as surveys and focus groups with Work First participants.  These contract projects will advise Social Services on how to improve employment services programs and how to market Work First employment services to customers.

Finally, county maps have been produced using Geographic Information Systems (GIS) and addresses of TANF recipients in the linked database.  These maps advise Social Services where pockets of poverty are located and where staff and other resources are most needed in Mecklenburg County.

There are also several projects currently in progress using the longitudinal database.  One major research project being conducted uses cross-tabulations and life tables to uncover relationships between various case characteristics and duration of cases in Work First.  A cross-tabular analysis and possibly a regression analysis will also be conducted to learn more about variation in post-TANF employment earnings.  Potential independent variables are case characteristics at point of entry, such as case head demographics, family composition, work history, education, and previous welfare attachment.  Programmatic effects on cases such as participation in Employment Services, sanctions, and employment earnings during the program will also be analyzed.

Another project in progress is a "mini-link" between the linked TANF database and data from other human services agencies in Mecklenburg County such as homeless shelters, food pantries, and crisis assistance groups.  The purpose of this project is to determine whether and how often TANF customers used these other services before, during, and after their participation in Work First.  The knowledge to be gained from this link is whether or not TANF customers are facing new difficulties due to pressures of Work First time limits.  This information will not be permanently integrated into the main longitudinal database.  Instead, a smaller database using portions of the main longitudinal database will be created.

Another "mini-link" is also in progress with the local sheriff’s office.  The purpose of this study is to create a profile of youthful offenders incarcerated in county jails.  This profile will help law enforcement officials better manage the population and provide appropriate intervention services while those offenders are in custody.  A list of names of 300 youthful offenders has been released to Social Services.  These offenders have been linked to the longitudinal database to determine whether or not they were part of a TANF case and how long they received TANF.  This project will benefit Social Services by demonstrating the breadth of social problems that young TANF recipients face.

Description of Products

Mecklenburg County has reported findings from the linked data in various arenas and formats.  In August 1998, Mecklenburg planners attended a conference for the National Association of Welfare Research and Statistics in Chicago.  Helen Lipman, Director of Planning and Evaluation, participated in a panel discussion on state data-linking grants at this conference.  The crux of her presentation was how differences in data measurement yield differences in reported results.  That is, longitudinal data sometimes show different results than cross-sectional data.  Interpretation of the data is dependent upon the format.

There have also been presentations at the local level on the topic of "successful" and heavily dependent customers in Melenburg County.  "Profiles of Self-Sufficiency and Welfare Dependence from the Longitudinal Database" was presented at a Brown Bag Luncheon for Applied Social Researchers in Mecklenburg County.  The same presentation was also made to the Department of Social Services subcommittee of the Human Services Council of Mecklenburg County.  Members of this group advise the Human Services Council, which, in turn, advises the local Board of County Commissioners about budgets and results in local human services programs.

Several reports using the longitudinal data have been sent to officials at the state level.  Analyses of post-welfare earnings of Work First participants were sent to Governor James Hunt, Secretary of Human Resources David Bruton, Representative Beverly Earle, and Senator Dan Clodfelter to keep them informed of progress in our county.  The governor recently wrote a letter to the Director of Social Services, Richard W. Jacobsen, Jr., thanking him for the information.  The governor also stated that he would distribute the report to others on his staff involved in Work First projects.  Reports regarding rates of exit and recidivism in Work First in total and by race group were also sent to officials at the state.  Mecklenburg County has made a priority of keeping those at the state level informed about Work First in our area.

Future Expectations

Mecklenburg County has maintained its working relationship with the Jordan Institute for Families.  They continue to provide us updated extracts from various state information systems.  They also help us to analyze and interpret our results.  As the Jordan Institute continues to build longitudinal files for the state and counties, they will be supplementing their databases with information from other data systems such as subsidized childcare, child support enforcement, and Medicaid.  Jordan Institute’s food stamp database will also undergo some improvements.  Mecklenburg County will eventually receive all of this information from them.  The agreement between Mecklenburg County and the Jordan Institute to continue our partnership is informal and somewhat open-ended.  As long as the Jordan Institute is providing data for the state and other counties, Mecklenburg will benefit.  County funds have maintained the planner position created as part of the data-linking grant.  The Jordan Institute has absorbed costs associated with generating and distributing data.

In addition to supplements provided by the Jordan Institute, Mecklenburg County Social Services will work with its Mecklenburg County Resolves II partners to share appropriate data and information.  Members of Mecklenburg Resolves who are potential data sharing partners for Mecklenburg County are the City of Charlotte, the Charlotte Housing Authority, the Charlotte Chamber of Commerce, Mecklenburg Ministries, Central Piedmont Community College, Charlotte-Mecklenburg Schools, Carolinas HealthCare System, Presbyterian HealthCare, and the United Way.

The database in its present state will allow for continued analysis of the TANF population.  As previously mentioned, a project is underway which proposes to explain differences in duration in Work First as well as differences in post-welfare employment earnings.  This project will tell us something about why these differences exist rather than merely what the differences are.  The analysis promises to be useful in program planning, predicting outcomes for customers, and targeting resources and efforts.

Appendix: Copies of Products

Appendix I. Profiles of Self-Sufficiency and Welfare Dependence from the Longitudinal Database

This report is a descriptive analysis of heavily dependent TANF recipients who were in the program for 30 months or more, as well as successful customers who left TANF for 18 months or more and have not returned.  The two customer groups are compared on various aspects such as demographics, employment earnings, employment industries, and usage of various social services such as food stamps and employment services.

Appendix II. Third Quarter 1998 Earnings Data for Selected Cohorts

The first analysis in this report shows the percentage of selected Work First exit cohorts who are working as of the third quarter of 1998 (our most recent quarter of data).  It also shows median employment earnings for these cohorts.  The second analysis shows the percentage of each cohort who worked a given number of quarters after exit.  This information tells us about consistency of employment after exiting Work First.  About half of each cohort worked fairly consistently, working in all or almost all quarters after exit.

• Sixty percent of customers in all of the selected cohorts had reported earnings in QTR398.
• For all customers in the selected cohorts with ESC reported earnings in QTR398, median earnings were around $2,800.

Appendix III. Rates of Exit and Recidivism for Selected Work First Cohorts

The following report provides information on exit and recidivism rates for nine entry cohorts and nine exit cohorts from 1995 – 1997.  Results show that exit rates did not change much over the three-year period.  For all nine entry cohorts, about half of customers left at about six months into the program.  Patterns of recidivism for the nine exit cohorts are not as clear.  Typically, only about a quarter of the exit cohorts returned to Work First within six months of leaving the program.  For two of the 1995 exit cohorts, about half of the customers had returned to Work First within 30 months of leaving.

Appendix IV. Racial Differences in Exit Rates

Cross-sectional data on Work First caseloads in Mecklenburg County show that reduction of the white caseload is happening much faster than reduction of the non-white caseload.  Life table analyses were conducted for 22 entry cohorts over a four-year period (1995-1998) to lend support to the hypotheses of differing rates of exit by race.  The results were mixed.  For each year examined from 1995-1997, three of the entry cohorts showed significant differences in rate of exit by race, and three cohorts showed no significant differences.  In 1998, only one of the four entry cohorts showed significant differences by race.

Appendix V. Racial Differences in Recidivism Rates

Racial differences in rates of return to Work First were analyzed for 18 exit cohorts from 1995-1997.  The proportion of each cohort who returned to the program is recorded in the following tables at six-month intervals by race.  In 1995, all exit cohorts, with the exception of June, showed similar exit rates between whites and non-whites.  By 1996, differences in recidivism rates for whites and non-whites become somewhat more pronounced, with a greater proportion of non-whites returning to the program.  In 1997, every cohort showed a significant difference in recidivism rates between whites and non-whites.  However, the 1997 analysis is based on only 12 months of data.

Appendix VI. Child-Only Cases in Mecklenburg County September 1998

The following is a descriptive analysis of the child-only caseload in September 1998.  The report includes information on demographics, case composition, living arrangements, geographic location of residence, cash assistance duration and amount, and child welfare involvement.

Appendix VII. TANF Recipients with Months against the Clock since August 1996

This report gives the number of active TANF recipients with months against the two-year time limit.  It indicates how many clients have only a few months before their time limits run out, and also the number of clients who have used all 24 months of the limit.

• In Mecklenburg County, 3,768 TANF recipients show one or more months on the two-year time clock.
• The most common number of months on the time clock is one month. Almost one fifth (18%) of TANF recipients with any time on the two-year clock have only one month.
• About half (47%) of the people with time against the 24-month limit have only three months or less, and almost three-quarters (72%) have 6 months or less.
• Only 13 people (0.3%) have used all 24 months of their allotted time on cash assistance. By the end of 1998, 56 additional people (1.6%) could potentially reach the two-year limit (See months 20-23 in Table 1).

Appendix VIII. Map: TANF Recipients in Mecklenburg County by Police District (August 1998)

Longitudinal TANF data were used to geographically plot residences of TANF cases active in August 1998.  The county was divided into police districts, and the red markers indicate the dispersion of TANF cases within the police districts.  The map shows that most TANF cases in August 1998 were concentrated in the central part of the county, which is the center of the city of Charlotte.