ARIZONA. I. Introduction to the State

Transcription

1 ARIZONA I. Introduction to the State A. Highlights Arizona s Department of Economic Security (DES) has several information systems that run on its large Hitachi mainframe. All clients in the department are trackable through the "High-Level Client Index" on the mainframe, which prevents duplicate cases from being created. Data linking for a welfare demonstration project and other large studies has shown that linking across systems is feasible. Many smaller research projects have also been done that link data across one or two systems in order to answer legislative or management requests. Linking is accomplished fairly easily because all the programs and data are in one department (DES) and on one mainframe, and the staff have experience working together. Getting data from the Arizona Health Care Cost Containment System is slightly more difficult for those reasons. Arizona is unusual in that it has the base wage record in the social services department (DES) rather than in a separate labor department, and the child support enforcement program is in DES rather than the Attorney General s Office. B. Demographics and Caseload Description In 1997, Arizona had an estimated population of over 4.5 million, ranking it twenty-first in size among the states. The median household income in the state in 1995 was $30,863 (in constant 1994 dollars), ranking it thirty-fifth among the states. In addition, 16.1% of the state s population in 1995 lived below the poverty line, making it the twelfth poorest population among the states. With 6.5% of the population receiving federal public aid in 1994, the state ranks twenty-seventh on that variable. There were 198,000 recipients of AFDC and 69,000 recipients of SSI in In 1995, the state had 480,000 Food Stamps recipients.

2 II. Data Systems for Social Programs A. Names of Data Systems AFDC/TANF Food Stamps Medicaid eligibility JOBS/JOBS successor Medicaid claims Child Protective Services Foster Care Child Support Child care subsidies Arizona Technical Eligibility Computer System (AZTECS) Arizona Technical Eligibility Computer System (AZTECS) Arizona Technical Eligibility Computer System (AZTECS) JOBS Automated System (JAS) Prepaid Medical Management Information System (PMMIS) Children s Information Library and Data System (CHILDS) Children s Information Library and Data System (CHILDS) Arizona Tracking and Location Automated System (ATLAS) Arizona Child Care Automated Tracking System (AzCATS) B. Description of Data Systems Arizona Technical Eligibility Computer System (AZTECS) Arizona s Department of Economic Security, Family Assistance Administration (FAA) maintains the AZTECS system, which contains data on AFDC/TANF, Food Stamps, and Medicaid eligibility. AZTECS is a relational database in the language "Natural," and it runs on the department s large Hitachi mainframe. The primary identifier for individuals on AZTECS is the "ISN," or Identifying System Number, also referred to as the "client ID." The same client ID is used for all three programs: AFDC/TANF, Food Stamps, and Medicaid. The ISN is the number also called the "high level client index" maintained by the department s Division of Technology Services (DTS). The high-level client index is a record of all cases ever known to the Department of Economic Security and is designed to prevent duplication of cases. It can be used to link client ID to SSN and name of individual. There is also a unique case identifier on AZTECS that is generated by the system itself. To find individuals in a particular case number, the "participation record" and the "case composition record" provide the clients IDs of all case members. Then the high-level client record provides the names of clients with those client IDs. For every individual in the case, AZTECS collects SSN, first name, last name, date of birth, sex, race, complete address (including zip code), county, and reservation (if applicable). All client data are overwritten, including date of birth, SSN, race, sex, and welfare reform codes. There are two fields for SSN; if a new SSN is added, the first one moves over, if a third is added, the first is lost. There has been a six-line field for documenting what the worker told the client

3 about welfare reform. If a new worker changes this field, it is completely typed over, even if there was space left. However, recently a new Case Documentation screen was added to AZTECS for the purpose of recording communication to clients, and the information entered on that screen will be retained online for five years. Every month, most records over 13 months old are purged from the online system and archived on tape. A few files are purged to tape after four to six months, and some are kept two or three years, but the basic information about case composition is kept online for 13 months. Data from the beginning of the AZTECS system, early 1988, are kept on tape. Medicaid eligibility data was first recorded on the system in 1993, so the early years only contain AFDC and Food Stamps data. Public assistance data from the period prior to AZTECS are archived on microfiche, and date back to the 1970s. The microfiche archives are kept at the Family Assistance Administration systems section; however, the archived AZTECS tapes are stored in another location. The tapes contain most records, including participation and case composition records; thus, historical data can be recreated (back to start of tapes, early 1988) on each case or individual on AFDC by month. Some tapes have been purged because of space limitations, but grant amount, program code, case entry date, and address are all available extending back to FAA systems can get data off tapes by submitting an ad hoc request to DTS, and then DTS loads it for them. Also, a longitudinal file of case records from 1991 is available on tape (see below under Linked Databases, "Longitudinal AFDC File") The microfiche archives (1970s-1987) contain the action history and participation records but generally not case composition records. Although they are maintained on FAA systems, it is harder to reconstruct the history of AFDC receipt using these because they are less complete than the tapes. JOBS Automated System (JAS) Arizona s Department of Economic Security has a separate system for the JOBS program, called the JOBS Automated System (JAS), implemented in late The JAS system is a person s system, with its own seven-digit unique client identifier. JAS captures AZTECS information on program participants when a referral is made to JOBS. Identifying information from AZTECS that comes in the referral and is maintained in JAS includes client s ID (ISN) and case number, SSN, address (including zipcode), race, date of birth, first name, and last name. The client s sex is not included. Address is updated by JOBS workers. There is no information on the JOBS participant s family. Some client and demographic information is overwritten in JAS, and some is retained. For example, client s address is overwritten if changed. If a client starts a new JOBS plan, the dates from the old plan are overwritten. However, a client s employment information is not overwritten if they get a new job. There are several history screens in the client records as well. One history screen is "payment status," which includes information on payments for support services. This file would include a history of address as well. Other history screens include a client s program status and a client s participation in JOBS activities. These history files are

4 purged periodically and archived onto tapes. For example, JOBS participation for a client preceding the previous four months is on tape and can be pulled within a day if necessary. Prepaid Medical Management Information System (PMMIS) Arizona s Medicaid program is administered by a separate agency called the Arizona Health Care Cost Containment System (AHCCCS). AHCCCS maintains all information on the program, including eligibility and enrollment, encounters, claims, and reinsurance, on their Prepaid Medical Management Information System (PMMIS). PMMIS has been in place since early Records on the systems are stored and identified by the individual Medicaid recipient. To identify the individual, a unique AHCCCS number is used. The system also uses SSN, date of birth, first and last name, and sex to identify individuals. Race is kept, but not used in identifying. The public assistance client ID, such as the AFDC number, if given to AHCCCS by the Department of Economic Security, is also kept, but that number is not used for identification. There are some state-only Medicaid recipients in the system as well, so not all individuals in the system are in public assistance. Through individual records, household members can be determined; however, there is no tracking of households by the system. Children s Information Library and Data System (CHILDS) The Children s Information Library and Data System (CHILDS) has contained foster care, adoptions, and Child Protective Services program data since June Before that, CPS data were in the CPSCR (Child Protective Services Central Registry) system, and foster care data were in the Arizona Social Services Information and Statistical Tracking System (ASSISTS). Both of these used the high-level client index as the person s identifier. CHILDS creates its own unique client ID and case ID. It also captures SSN, name, address, and date of birth, but not with consistency. AZTECS information and the high-level client index number are not automatically recorded in the system, but they can be obtained through the system. "Wage System" in General Unemployment Insurance Development Effort (GUIDE) Rather than being housed in a separate department, the UI program is run by the department s Unemployment Insurance Benefits Unit. GUIDE is the processing system for the Unemployment Insurance program, and it consists of a software package that runs on the department s mainframe. One subset of GUIDE is the "Wage System," which contains UI earnings and employer records. The only identifier for individual earners in the system is their SSN. The Wage System does not use the high-level client index. Thirteen quarters of data per individual are kept active on the system. Each year, data are purged to tape, resulting in a tape with 2-3 million records. Ten years of data are kept on tape, so data on earnings extending back to 1988 are currently available. Wage data are available and readable on AZTECS. Wage data are also made available to outside agencies, such as the IRS and the Department of Revenue. Arizona Tracking and Location Automated System (ATLAS) The state s child support program is administered within DES, in their Division of Child Support Enforcement (DCSE). The Arizona Tracking and Location Automated System (ATLAS) has been used by DCSE since 1990 to manage and track child support payments and collections. It runs on the department s mainframe. In 1996, ATLAS entered "Phase II" when it became

5 compliant with the Family Support Act of (Within the Department the two versions of ATLAS are called "ATLAS I" and "ATLAS II.") Implementation of welfare reform changes to the system was scheduled for late ATLAS is a individual-based system, although cases are defined in the system as the mother and father and their children. The person s identifier is a combination of name, date of birth, and SSN. ATLAS generates a unique case number. The addresses and SSN s of all members of a case are maintained in the system as well. The court docket number is also associated with the case, but that number has been found to not be unique because different counties can use the same numbers. The DCSE can use the high-level client index on the DES mainframe to determine if anyone associated with the case is know to any DES program. They can thus track down more information, particularly financial information, on noncustodial parents. Arizona Child Care Automated Tracking System (AzCATS) The Arizona Child Care Automated Tracking System (AzCATS) has been operating since April 1997 in the Department of Economic Security, and contains information on various child care subsidies and payments made on behalf of DES clients. Before AzCATS, child care information was kept on the Arizona Social Services Information and Statistical Tracking System (ASSISTS), where Child Protective Services data were also stored. AzCATS holds client records for both the parent who applies for child care assistance and the child in child care; there are no "cases." The primary identifier is the high-level client index of the department. If the client is known to AZTECS, there is also the first name, last name, middle initial, SSN, birth date, race, and sex. About 65% of child care clients are known to AZTECS. AzCATS also has address and county of clients, and zip code is the primary geographic identifier. III. Linked Databases Arizona Department of Economic Security has two units involved in research and statistical analysis using and linking administrative data, the Office of Evaluation and the FAA (Family Assistance Administration) Strategic Planning, Budgeting, and Management Information section. The Office of Evaluation is involved in long-term research and database development. The FAA planning and budgeting section responds to immediate requests from management, legislators, and advocacy groups for answers to specific questions and handles all federal reporting requirements. Some projects, such as the EMPOWER demonstration, involved both sections because of the requirements for both a long-term study and a cost neutrality report. Staff in both sections have extensive knowledge of the data systems and how to merge and use administrative data for analysis. Consequently, they are important liaisons to outside researchers, and they compile and maintain valuable data sets for welfare research. A. AFDC waiver demonstration project (Employing and Moving People Off Welfare and Encouraging Responsibility (EMPOWER)) The EMPOWER demonstration is managed by the Office of Evaluation, and the evaluation is contracted to Abt Associates. The FAA planning and budgeting section is responsible for assisting data extraction and calculating cost neutrality. Arizona State University is a subcontractor for a survey of clients.

6 The purpose of EMPOWER was to evaluate several waiver provisions, the major four being a benefit cap, the requirement that teen parents to live at home, 24 months of continuous Medicaid, and a 24-month time limit (pre-prwora). The demonstration project started in November 1995 and continues to the present, comparing control and experimental groups on these policies. The control group is not subject to the new TANF rules, except those that have no relation to the policies under study. Administrative data will be collected on the sample through July The EMPOWER data are being used by DES for Food Stamps cost neutrality and by Abt Associates for the impact analysis. The first major impact report was due December To date, there have been quarterly reports and an interim process study with some impact estimates. DES staff agreed that Abt should produce its outcomes before any other analysis is done by state or other researchers using this data. DES systems staff created the EMPOWER data set by downloading and merging data from different systems. The files are then maintained by the DES Office of Evaluation and Abt Associates. The EMPOWER data set was put together successfully because almost all the staff involved were in the same department, and because groups had worked together previously on extracting, linking, and analyzing data; thus, turf problems were minimized. The evaluation used an experimental design comparing 3,000 control to 3,000 experimental cases, all from four welfare office sites in the state. An initial sample of 1,500 control and 1,500 experimental was randomly selected, and an equal number were added by randomly selecting 20% of new cases each month until July Three welfare offices are larger, ethnically representative ones in the Phoenix area and the third is on a reservation, in order to represent accurately the state s population. Cases are kept in the sample unless they move away from those sites and never return to public assistance in any of the sites. Monthly data on the sample have been collected since November 1995, resulting in a longitudinal data set. Abt Associates receives monthly compilations about a month or two after the quarter is over. UI earnings data, however, are six months old. The data are organized "at the person level within a case," that is, it contains data by individual. However, some elements are case-driven. For example, the grant amount for the whole case is associated with each individual in the case. The smallest geographic identifier is the zip code. Linking is accomplished through the SSN and the high-level client index. Both numbers are maintained as well as the case ID. The JAS number is also maintained, but once used for linking it is no longer needed. The data set is all flat files housed on the DES Hitachi mainframe. Monthly data are extracted from the following programs: AFDC, Food Stamps, Medicaid eligibility, JOBS, child care, AHCCCS (Medicaid plan enrollees and history), Emergency Assistance, Unemployment Insurance, child support, and child welfare. AFDC/TANF data elements include the grant amount for each month, case size, household size, reasons for exit from aid, and highest grade or GED of head of case. Job training, work activity, and employment information are available on JAS for JOBS participants. Children known to Child Protective Services are identified in the sample, and child support establishments are taken from ATLAS.

7 B. Longitudinal AFDC File In the late 1980s, the Department of Economic Security decided to examine length of stay and recidivism in AFDC, and for this purpose began developing a longitudinal file on the mainframe with monthly AFDC case information. In building this data set, researchers found the data in the first few years were difficult to use, so the most useful data begins in July 1991 through the present. Between 1991 and 1995, the file was built with yearly extracts. Some data elements had twelve months of information, such as benefits, but for others, only the information for the month of the extract was reflected. From 1996 until the present, quarterly extracts are archived into this file. The case information is basically payment and demographic information, and with some limited client data from 1994 and JOBS and child care records for each case head of household are merged into the AFDC case record. Although the DES hoped to use this file for other studies and to link to other programs, it was created for and has only been used for the longitudinal study completed in May. The first several years of the longitudinal data have been archived in the tape library for the mainframe. The longitudinal study followed a cohort of AFDC recipients from 1992 though The cohort consisted of all cases in January 1991 that were either new to AFDC or had not been on AFDC in the six months prior. Researchers examined welfare spells as defined by a break of at least three months. Using a survival model, they looked at survival by various demographic variables and the effects of aid programs on survival. This study was possible because the longitudinal file was already in place, thanks to the commitment of the agency to archiving. The data on the sample are being kept by the Office of Evaluation and will be used for updating the study and as a pre-prwora baseline. The data on the cohort are in a flat file containing 2,450 records. The cohort consists of 2,215 AFDC single-parent cases and 235 unemployed two-parent cases. Each case record contains merged AFDC, JOBS, and child care data. Food Stamp, Medicaid, child welfare, and UI information are not included. The smallest geographic identifier is the welfare office, also called the "site code" in Arizona. Data elements include case information and demographic information on the head of household. The longitudinal data include grant amount, but for the purposes of the study this was recoded into a binary variable of whether a grant was received or not. JOBS information include participation, educational activities, and last grade completed. Child care disregard amount was available, and that was recoded into whether it was claimed or not and the total use of the disregard during a spell. Incidence of disability was only available if it could be inferred from other information, such as a JOBS exemption. C. Data Warehouse Federal funding was recently awarded to develop a cross-program departmentwide data warehouse that can be used to query administrative data on department programs. The Office of Evaluation still hopes to have a portion of the system accessible by January 1, The system is meant for ad hoc reporting and client tracking by program staff as well as research and evaluation by department staff and outside researchers. The warehouse will be an Oracle relational database housed in a RAID server. The warehouse itself will contain several "datamarts" pertaining to different programs, and program staff can then use Microsoft Access to query the datamarts. The warehouse itself will be regularly purged,

8 but the datamarts will be archived onto CD-ROM by the Office of Evaluation. Outside researchers will then be able to request data from the Office of Evaluation and receive a CD- ROM or ideally access the data through an internet account. The high-level client index, available for most clients in the department, provides an anonymous identifier. The population will initially be everyone in TANF, and will later be expanded to all DES clients, even all clients in the base-wage file (all employed persons in the state). Data on all programs, including AFDC, Food Stamps, Medicaid, JOBS, child welfare, child care, child support, and UI earnings, will be linked for department clients. Ideally the data will extend back to the late 1980s, and it will be at both the person and case levels. For TANF recipients, there will be data on benefits, sanctions, demographics, and employment. Data linking will be accomplished using the high-level client index and Social Security number. D. TANF Exit Study A DHHS funded study of recipients leaving cash assistance used administrative data gathered over 12 months from various systems. Two cohorts were studied. The federal government required following a sample of recipients who left assistance during the last quarter of In addition, DES also followed a sample of clients, approximately 10,000 who left assistance during the first quarter of 1998, which is when progressive sanctioning could have caused recipients to leave the program. Administrative and survey data were used to determine what had happened to these clients. AZTECS data provided information on return to cash assistance, use of Food Stamps, and use of Medicaid. The study also incorporated UI wage records, child care use, child welfare services, child support enforcement, and emergency services. The survey (n=800) filled in the gaps and validated administrative data by gathering more details about clients lives and needs after leaving the program. The Office of Evaluation staff linked flat files from each system to complete the analysis, rather than having systems staff do so. The survey will be completed by a contractor. A final report will be submitted in July The data gathered for this study will also be integrated into the data warehouse. E. Food Stamps Exit Study Abt Associates was awarded a grant from the USDA to follow those who left the Food Stamps program. The DES Office of Evaluation is providing them with the administrative data to complete this study. Administrative data will be collected for all of 1998 on three groups of recipients in two counties. All recipients in the two counties who fall into three categories will be selected for the study. The three categories are those who left Food Stamps during 1997 and who are able-bodied adults without dependents (ABAWDs), non-abawd who received cash aid (TANF) in the last 12 months, and non-abawds who did not have cash aid (TANF) in the last 12 months. DES projects that these groups will contain 2,000, 10,000, and 8,000 respectively, for a total research sample of 20,000. They hope to survey 400 of each of those for a survey of 1,200. Administrative data will include Food Stamps activity, cash aid receipt, Medicaid receipt, UI earnings, child abuse and neglect, child care subsidies and arrangements, child support payments, and JOBS activities. When the research sample is drawn, the Office of Evaluation will periodically run that file against the various administrative data systems to produce flat files with the data on those persons. They will send the flat files to Abt Associates on tape or diskette. The final report is due in December 1999.

9 F. High Performance Bonus Project The Office of Evaluation will analyze administrative data in order to obtain the federal "high performance bonus" in their TANF program. For all persons 18 years or older who are receiving a cash grant, data on job entry, job retention, and earnings will be collected and be used to measure "success in the workplace." To accomplish this, a merged AZTECS and JAS file will be run against the UI records for two points in time, FFY 1997 and FFY The Office of Evaluation staff will extract and merge the data. G. Federal Reporting Extracts The FAA planning and budgeting section prepares the quarterly extracts of monthly administrative data required for federal reporting. A flat file of linked TANF, JOBS, child welfare, and child support data is produced by the DES programmers and sent to planning and budgeting, where it is then converted into the format required by the federal government. Child care data will be added eventually as well. The linking is done using SSN and the high-level client index. However, the file produced for federal reporting contains only the case number; it is impossible to identify individuals in the cases. Four quarters of data have been compiled to date, including July 1997 through June These same extracts have been sent to Washington University in St. Louis for a study of the impact of TANF on Arizona tribes and reservations. Washington University has a data-sharing agreement with DES and has obtained a release from all 17 tribes to examine this data. The TANF cases among all tribes totals approximately 12,000. The format of these extracts has proved difficult for researchers to use and merge with other files; therefore, DES planning and budgeting staff are considering doing specialized runs of data to answer the university s research questions rather than sending them the federal extracts. H. Smaller Linking Projects The following are four examples of linking done by the FAA planning and budgeting staff to answer simple questions about programs and clients. This type of linking is accomplished fairly easily because the programs and data are in one department (DES) and on one mainframe, and the staff have experience working together. Getting Medicaid data from the Arizona Health Care Cost Containment System is slightly more difficult for those reasons. Self-sufficiency study In the mid-1990s, a deputy director in DES was interested in knowing whether clients leaving AFDC found employment. FAA planning and budgeting staff found in AZTECS all AFDC cases closed in one month, then ran that sample against UI earnings records for two years after that month. This resulted in a longitudinal data set of the earnings of about 6,000 cases. Linking is accomplished by SSN. JOBS recidivism study For the DES strategic plan, the success of JOBS recipients is calculated for every calender year, starting with For each year, the FAA planning and budgeting staff calculates the number of JOBS clients who, within two months of starting the program, find employment and leave cash assistance and do not return for at least six months. This requires a match of JOBS data and AFDC participation data of AZTECS for about 800-1,000 persons each year. Linking is done

10 primarily by SSN, and the high-level clients index is used if SSN fails. Child care study Also in the mid-1990s, the FAA planning and budgeting unit assisted the child care division by running a list of child care subsidy recipients (off ASSISTS) against the AFDC rolls in AZTECS. The child care division needed to know how many of their recipients were on AFDC for their own performance monitoring. Four quarters were analyzed for approximately 1000 cases. Welfare data matched with foster care data Within the last few years, the Foster Care division has needed to report twice to the legislature on the number of children in foster care who also received AFDC/TANF. The Foster Care division gave a flat file to the FAA planning and budgeting staff, who then extracted data from AZTECS for those cases and gave that extract back to the Foster Care division for analysis. The SSN was not readily available in the foster care database (CHILDS); therefore, the high-level client index was used. IV. Contacts Name: Agency: Address: Phone: Fax: Expertise: Name: Agency: Address: Phone: Fax: Expertise: Name: Agency: Address: Phone: Fax: Expertise: Name: Agency: Address: Phone: Fax: Expertise: Aldona Vaitkus Strategic Planning, Budget and Management Information Section, Family Assistance Administration, Arizona Department of Economic Security 1789 West Jefferson St., Site Code 960A Phoenix, AZ / / Data Analysis: AFDC/TANF, Food Stamps, and Medicaid Frank Raia Office of Evaluation, Arizona Department of Economic Security Phoenix, AZ 602/ / Research/Evaluation of DES programs David Stensland Systems Section, Family Assistance Administration, Arizona Department of Economic Security 3443 North Central Ave., Suite 600 Phoenix, AZ / Ext / AZTECS system Judith Sparks-Craft JOBS Administration, Arizona Department of Economic Security 3443 North Central Ave., Suite 601 Phoenix, AZ / Ext / JOBS Automated System

11