This report presents the results of an experiment aimed at comparing the effectiveness of private vendors and public child support agencies in locating absent parents. The experiment was one feature of the Colorado Model Office Project, a multi-faceted demonstration and evaluation project funded by the Federal Office of Child Support Enforcement aimed at improving child support performance. Two Colorado counties participated in this privatized location experiment: Mesa and Archuleta. Both counties negotiated contracts with two private vendors that specialize in child support work and credit bureau reporting, Lockheed IMS and TMR. A system of data collection and follow-up was implemented to monitor the performance of the private contractors and the county child support agencies and to gauge whether the privatization of the location function leads to superior outcomes. Cases that were worked were compared to a sample of comparable cases that were not worked.

Many states are moving to privatize child support enforcement services that have traditionally been delivered by the public sector. Based upon a recent report by the General Accounting Office (1996), fifteen states have turned to full-service privatization of selected local child support enforcement offices as a way to improve performance and handle growing caseloads. In some states, privatization is a response to state restrictions on hiring additional public employees. Still another reason to privatize a child support function is to enlist the services of agencies and individuals who are potentially more adept at certain procedures and/or have access to better technology and resources. This was the rationale for contracting with private vendors for "locate only" services in the Model Office Project experiment.

The research on the effectiveness of privatizing child support services is sparse (GAO, 1996). Three comparisons of performance between fully privatized offices and public child support programs suggest that privatized offices perform "at least as well as or, in some instances, better." For example, according to the GAO study, differences in location rates for public versus privately run offices in Arizona and Virginia were not statistically significant. In Tennessee's Twentieth Judicial District, however, performance may have improved since outcomes remained comparable after privatization while serving a more challenging caseload with a much higher percentage of AFDC cases.

Clearly, it will take more research over a longer period of time to make more definitive conclusions about privatization of child support services and whether performance is comparable, better or worse for public versus private offices. Colorado's experiment with privatized location services represents a contribution to that effort.

Mesa County negotiated contracts with two vendors to assist the agency with location activities. Archuleta County worked with one vendor. Appendix A contains a copy of the contract for services developed with the vendors, Lockheed IMS and TMR.

To assess the relative effectiveness of each vendor and compare privatized locate efforts with location conducted by county-level child support workers in Mesa and Archuleta counties, we randomly assigned all cases requiring location work to the two vendors and the in-office staff. In Mesa County, each vendor and the in-office staff were assigned 1,500 cases, respectively. In Archuleta County, the vendor was given 80 cases. A random assignment procedure was use to make all case allocations to ensure that each group of cases contained roughly equal numbers of establishment versus enforcement cases, AFDC versus non-AFDC cases, old cases versus new ones, and cases involving self-employed and employed obligors. All the cases selected to be in the experiment required location activity in order to proceed with order establishment or child support enforcement. They were drawn from the pool of cases in Category 3 (locate and enforce), Category 5 (locate and establish a support order), and Category 7 (locate and establish paternity/support).

In Mesa County, the contract provided enough funds for each vendor to produce 200 verified locations. In order to assess the effectiveness of the vendor in handling enforcement and establishment cases, each vendor was required to produce a maximum of 100 locations for enforcement cases and 100 locations for establishment cases. In Archuleta County, the vendor was permitted to achieve a maximum of 50 locations. The vendors were given printed copies of various ACSES screens to assist them with their location searches. To the extent it was available, vendors and county staff had the following information to work with about the parent to be located: name, social security number, date of birth, current residence and phone number, previous residences, current employer and employer phone number. The following information about the custodial parent was also provided: name, social security number and date of birth.

Each vendor and in-office staff used a variety of techniques to locate absent parents: checking directory assistance, phoning credit bureaus and creditors, phoning residences, employers and neighbors, and mailing letters to residences, employers and neighbors. Vendors and in-office staff kept track of their successful location efforts by noting this on a report form designed by the evaluation contractor. The form captured key information on the types of steps taken to locate the obligor and the outcome of those efforts. Thus, for each case, the vendor and/or in-office staff noted whether new location information was found, the nature of located assets, the method of verification used and the types of verification sources consulted.

We collected additional background information on the cases that were located. We also collected background information for a non-treatment sample of 50 establishment and 50 enforcement cases assigned to each vendor in each county. These were cases that were not worked in a concerted manner by TMR, Lockheed or the Mesa County child support technicians. The information on case characteristics was generated from screen prints from the automated child support enforcement system (ACSES). We extracted many items about each case, including the following: case category, AFDC status, race and marital status of the obligor, length of time in the locate status and case category.

We measured the ultimate impact of this intervention by generating an ACSES extract for treatment and non-treatment cases approximately six months after the samples were generated. This consisted of all 828 cases in this experiment. At this follow-up time point, we noted the case category, whether paternity had been established, whether a child support order was in place and whether child support was being paid. Finally, we interviewed vendors and child support enforcement administrators about the utility of this intervention.

Appendix B contains a list of the data elements extracted from ACSES for the evaluation and a copy of the location report forms we developed for use by vendors and in-office staff.

TYPES OF CASES WORKED

TMR, Lockheed and Mesa County workers all worked the same types of cases: cases with a social security number for the absent parent, cases with some employment information on ACSES, and cases that had experienced at least one category change on ACSES in the past two years (see Table 1).

TMR worked these same types of cases in Archuleta County too. Indeed, 75 percent of the cases TMR worked in Archuleta showed movement on ACSES between 1993-1995. Clearly, all workers devote their efforts to the cases with the highest probability of success.

NUMBER OF CASES LOCATED

TMR, Lockheed and Mesa County child support workers each achieved locates in about 20 percent of the cases they were given to work. Each vendor was limited to locating 200 cases (100 enforcement and 100 establishment). Had they achieved a location for all 200 cases, this would have translated into a success rate of 26.7 percent. In Archuleta County, TMR generated non-duplicative location information for 16.5 percent of the 97 cases they were given to work.

TYPES OF CASES LOCATED

Both TMR and Lockheed reported better success with enforcement cases than with establishment cases. Both vendors also had better luck verifying employment in enforcement rather than in establishment cases. Residence verification patterns were more similar for enforcement and establishment cases with TMR and Lockheed both reporting new verified addresses for about 45-50 percent of both types of cases. Mesa county workers reported new addresses for about a third of the cases they worked. In Archuleta County, new employers were reportedly located for 19 percent of the cases and old employer information was verified in another 19 percent of the cases. TMR located new home addresses for absent parents in half of the cases.

TYPES OF LOCATE TECHNIQUES USED

Lockheed used directory assistance and telephone numbers provided on ACSES screens. They sent letters to RAs and APs. They also used "crisscross" and neighbor directories.

TMR used credit bureau records, relying on header information for establishment cases and full credit bureau reports for enforcement cases. They also phoned past employers and found employers and credit bureau reports to be their best locate resources.

The Mesa County child support enforcement unit relied heavily on mass locate efforts and made extensive use of document generation procedures to produce postmaster verification letters, employer verification letters, URESA status requests and out-of-state locate requests.

Despite the variety of location techniques used, overall performance by the three vendors was remarkably similar.

NATURE AND TYPES OF VERIFIED INFORMATION AND ASSETS

In Mesa County, new information was reported as verified in 99 percent of the located cases returned by TMR, 98 percent of the located cases returned by Lockheed and 78 percent of the located cases handled in-office by Mesa County workers.

Employers were reportedly located for 53 percent of the TMR cases with new information, 63 percent of the Lockheed cases with new information, and 42 percent of the in-office Mesa County cases with new information. In Archuleta County, TMR located new employers for 19 percent of the cases.

New addresses for absent parents were reported in 46 percent of TMR cases, 45 percent of Lockheed cases and 33 percent of in-office Mesa County cases. In Archuleta County, TMR located new home addresses for absent parents in half of the cases.

TMR reported finding assets such as cars or bank accounts in about a third of the cases for which they produced new locate information. This probably reflects the fact that only TMR routinely did credit bureau checks for cases needing location. In Archuleta County, TMR located assets in only six percent of the cases.

IMPACT OF LOCATION ON CASE PROGRESS

Locating cases definitely led to category improvements (see Table 2). Six months after a locate had been achieved, 41 percent of TMR cases had improved by at least one category, 53 percent of Lockheed cases had improved by one category and 49 percent of cases worked by in-office Mesa County staff had improved. Only about 20 percent of cases in the control group that were not located experienced this level of improvement.

In a similar vein, about half of the cases located by each vendor in Mesa County stayed in the same category. In the Mesa County control group, at least three-quarters of the cases stayed in the same category. In Archuleta County, 71 percent of the located cases showed no category change after six months.

Cases that showed positive category movement were most likely to be distinguished by recent category change (within the past year). Enforcement cases were also more likely to show category changes than establishment cases. AFDC and non-AFDC cases were equally likely to show improvement (see Table 3).

IMPACT OF LOCATION ON CHILD SUPPORT PAYMENT

Across all three vendors, approximately 20 percent of worked cases moved into a paying category six months after the locate effort. For TMR, the percentage moving into category 1 was 22 percent. For Lockheed, the percentage moving into category 1 was 16 percent. For Mesa County, the percentage moving into category 1 was 23 percent (see Table 4). For unworked cases in a randomly drawn control group, only 2-9 percent of cases moved into a paying category (see Table 5). In Archuleta County, none of the cases worked by TMR moved into the paying category.

REACTIONS OF CHILD SUPPORT ENFORCEMENT AGENCY ADMINISTRATORS

Child support administrators were disappointed with the results of the privatized location effort. They had hoped that private vendors would know special locate techniques or have access to better locate resources and that this would be reflected in superior performance patterns. They were surprised to learn that performance patterns for privatized and in-office staff were virtually identical. Nevertheless, they felt that the privatized locate experiment had stimulated Mesa County child support workers to compete with Lockheed and TMR. Workers paid extra attention to the locate function and achieved more locates than they would have without the experiment. Thus, they felt that the experiment was helpful. Administrators in Mesa County also felt that privatized location efforts may be useful ways to augment the mass case processing techniques used by CSE unit workers, particularly for enforcement cases that require more individualized attention and follow up.

Administrators in Archuleta County, on the other hand, found little value in TMR's location efforts. They cite several reasons to explain why they believe an outside vendor can do so little. One is that the Unit is very aggressive in its in-house location efforts and enjoys certain advantages that an outside vendor can not duplicate. Unit workers in Archuleta County make a point of handling intake cases when they are fresh. Although the director estimates that the average case takes 45 minutes to an hour to locate, workers might spend up to half a day on the location function with a new case.

Workers also use a variety of local resources that have a proven track record of effectiveness. For example, workers typically contact the local sheriff's department which provides driver's license information from a national, online system. Unit workers follow this with a credit bureau check. As a result of their early action and aggressive use of local resources, the director of Archuleta County's Child Support Enforcement Unit estimates that the Unit achieves a 95 percent success rate with locates on new cases.

REACTIONS OF PRIVATE VENDORS

Private locate services would like to work more closely with child support agencies and believe that they bring useful skills to the job including the ability to focus on the locate function without the interruptions of case work. To improve their success rate, they recommend working more closely with agencies to develop intake procedures, obtaining fresher locate cases from agencies on a quarterly basis, developing relationships with technicians familiar with cases to probe for location "leads," and obtaining online access to ACSES.

One element of the Colorado Model Office Project involved the retention of private vendors to perform location services for the Mesa and Archuleta County Divisions of Child Support Enforcement. It was hoped that private vendors would have access to state-of-the-art resources and use a variety of specialized techniques to locate absent obligors and that this would make them more effective than in-office staff. To assess their performance and compare it with performance by in-office staff, cases needing location were randomly assigned to vendors and in-office workers. Vendors had access to existing location information available for each case on the ACSES. Vendors and county staff kept track of their location efforts, the sources they consulted, and the nature of the information they verified. To assess what would have happened to these cases in the absence of concerted location efforts by vendors and in-office staff, a control group was generated for each vendor consisting of comparable cases that were not systematically worked. Finally, all cases in the experiment were studied six to eight months after the selection process to gauge their child support status and their standing with respect to child support orders and payment.

The analysis revealed that performance patterns were comparable for the private vendors and in-office staff. All three vendors tended to focus their attention on the same types of cases. Thus, staff at Lockheed, TMR and the child support agencies chose to work cases for which there was: (1) a social security number for the absent parent, (2) some employment information for the absent parent on the ACSES, and (3) some recent changes in child support category on the ACSES.

Among cases that were worked, all three vendors were equally likely to verify a new home and employer address for the absent parent. The vendors reported using different sources to generate verifications. For example, TMR used credit bureaus for 93 percent of the cases and directory assistance for 51 percent, while Lockheed used credit bureaus for only 19 percent and directory assistance in 51 percent of the cases. To contrast, in-office staff in Mesa County rarely reported using either. Despite these differences in method, all three achieved similar performance patterns.

The locate efforts of private vendors and in-office staff definitely paid off. Six to eight months after they were worked approximately 20 percent of the cases located by vendors were in a paying child support status. Only 2-9 percent of the cases in the control group advanced to a payment status in the absence of a concerted location effort. Overall, about half of the cases worked by a private vendor or in-office staff showed improvement in case status while only about twenty percent of cases in the control group exhibited improvement.

For the private vendors, as with in-office staff, establishment cases rather than enforcement cases were more apt to register positive movement as a result of location activity, as were cases with recent movement. AFDC status made no difference in case progress.

These findings parallel those reported by the General Accounting Office (1996) in its recent report on child support performance in privatized and public offices. The Colorado experiment, like the GAO study, suggests that private offices perform "at least as well" as public ones. Our evaluation did not consider cost-effectiveness. Thus, we have no way of knowing whether private vendors achieved their comparable results at lower, higher or similar cost.

Nor do we have a way of gauging the level of location activity maintained by a public child support office in the absence of an experiment where public and private vendors essentially compete with one another. The director of the Mesa County Child Support Division readily concedes that the experiment stimulated her workers to focus on location activity in a concerted manner. During the experiment, her staff initiated Friday afternoon "location parties," and paid extra attention to the locate function. As a result, they achieved more locates than they would have without the experiment. If staff routinely accord less attention to location activities, their performance may be better reflected in the patterns for cases in the control groups.

Clearly public workers can do as well as private vendors when they both try. The question is whether in-office staff will make the extra effort in the normal course of events. One of the strengths that private vendors bring to the job is their ability to focus on the locate function without the interruptions of case work. For this reason, child support administrators may well consider the use of privatized location vendors to supplement the mass case processing activities they pursue, and be assured that the results will be at least as good as those achieved by the public agency when workers are fully motivated.

In very small counties like Archuleta, the benefits of privatization may be less acutely felt. Perhaps because case volumes are more modest, child support staff in small offices are able to aggressively pursue locate activities with new intake cases. Because they enjoy close relationships with local sheriff's departments and other agencies, county staff may also be able to access information that is not available to outsiders including private vendors. These factors may make privatization less compelling in small, rural counties.

It will clearly take more evaluation efforts to weigh the costs and benefits of full and partial privatization activities in the child support arena. Colorado's project provides a methodology to evaluate future privatization experiments. The report represents an important contribution to a literature that is still in its infancy. Finally, the results provide useful guidance to child support programs that are considering privatization options to address resource constraints and performance demands.

Government Accounting Office, "Early Results on Comparability of Privatized and Public Offices," Report to the Chairman, Committee on the Budget, House of Representatives, December, 1996.