Feasibility of Conducting A Child Support EnforcementCustomer Satisfaction Survey

The Child Support Enforcement (CSE) program was established in 1975 through Title IV-D of the Social Security Act to ensure that noncustodial or absent parents provide adequate support for their children when the children live in another household with a custodial parent, a caretaker relative, or a foster parent. For many low-income households, child support can make the difference between requiring public assistance and remaining self-sufficient. All states operate CSE programs to locate noncustodial parents; establish paternity; obtain or establish support orders; and collect, distribute, and enforce support orders. The federal government through the Office of Child Support Enforcement (OCSE) funds 66 percent of the administrative and operating costs of the state CSE programs (GAO, October 1996).

Until recently, CSE programs have focused on the provision of services to the custodial parent (usually the mother), and have viewed the noncustodial parent (usually the father) solely as a payor or debtor. This perspective is gradually changing: the 1995 OCSE strategic plan stresses customer service and emphasizes that both the custodial and the noncustodial parents are customers (GAO, October 1996).

In mid-1995, OCSE established the national CSE Customer Satisfaction Work Group, whose members are representatives of state and local IV-D agencies. In 1997, OCSE funded an investigation of the feasibility of conducting a customer satisfaction survey for CSE services. Administering a nationally representative CSE customer service survey would allow OCSE to determine how satisfied customers are with the quality of CSE services, the way services are delivered, the attitudes of CSE workers toward their clients (both custodial and noncustodial parents), and to learn how the support received (or paid) affects the lives of both custodial and noncustodial parents and their children. Such information could help OCSE and state agencies assess the quality of service delivery, determine areas where improvement is needed, and assess customer satisfaction over time with changes in service delivery. By breaking out the data by client characteristics (e.g., race/ethnicity, public assistance status), the survey findings could be used to design service improvements to target the needs of specific client groups.

This report presents the results of an investigation conducted by The Urban Institute and its subcontractor, Mathematica Policy Research, Inc., into the feasibility of conducting a national representative customer satisfaction survey of CSE parents.

In conducting this study, we first defined CSE customer satisfaction and then reviewed the relevant literature on customer satisfaction surveys. Developed in conjunction with the CSE Customer Satisfaction Work Group, our working definition of CSE customer satisfaction reads as follows:

The satisfaction of custodial parents, noncustodial parents, and caretaker relatives with the quality of the services provided by CSE staff (such as responsiveness, timeliness, courteousness, and clarity of written and spoken communication) and of service outcomes (such as the amount of child support actually received on behalf of the child and the timeliness of the payment), and their perception of the effect of child support on the child's well-being.

We examined CSE customer satisfaction survey activities and the factors that could affect the implementation of such surveys in five states (Ohio, Missouri, Tennessee, Washington, and Massachusetts) and one county (Hennepin County, Minnesota) -- sites recommended by the CSE Customer Satisfaction Work Group. The site visit and data collection protocol collected data from the sites on how they had implemented their customer satisfaction surveys, and about characteristics of CSE services and customers that could affect the feasibility of conducting a nationally representative survey of CSE customers. The Appendix contains a copy of the site visit protocol. We collected this information in October and November 1997 through in-person visits to Washington and Ohio, and telephone interviews with the staff at the other sites. During the site visits, we interviewed the IV-D Director or Assistant Director as well as staff associated with customer satisfaction survey activities or with the automated CSE database. The telephone interviews were conducted with the state or county IV-D Directors in Missouri, Tennessee, Massachusetts, and Hennepin County, Minnesota. In Hennepin County, an additional interview was conducted with the chair of the Hennepin County customer survey group. Following the review of state activities and issues, we prepared a Working Paper on Sampling Strategy (Cox, 1997).

Five of the six sites have conducted, or are in the process of conducting, a CSE customer satisfaction survey. Tennessee reported never having conducted a customer satisfaction survey. Two sites (Massachusetts and Ohio) were conducting a first-time statewide CSE customer satisfaction survey. The other three sites had conducted local (i.e., in specific counties or field offices) customer satisfaction surveys in recent years. Our review of these survey efforts determined that states and counties do not at present have much experience with, or capacity for, conducting customer satisfaction surveys. All but one of the surveys were conducted in-house.

The sites in general had no prior experience in designing and implementing customer surveys. Problems encountered in conducting the surveys included the use of nonrepresentative samples and lack of procedures for calculating response rates--thus making it difficult to use the findings to make inferences about the overall state or county CSE populations from which the samples were drawn. The known response rates averaged 28 percent for custodial parents and 12 percent for noncustodial parents.

The review was very instructive in identifying the following characteristics of the state CSE systems that need to be considered in designing a representative national customer satisfaction survey:

As yet, not all states have a centralized CSE database. The 1988 Family Support Act required each state to develop a statewide automated data system for conducting and monitoring all aspects of child support enforcement activities. Originally October 1, 1995 was the deadline set for states to have their automated data system in place and certified as meeting federal guidelines (GAO, June 1997). Later the deadline was extended to October 1, 1997. In December 1997, the OCSE Office of State Systems reported that only 21 states were certified. In four states, the review report was either being written or was awaiting further documentation. An additional 11 states had requested or were scheduling reviews prior to April 1998. This is an important issue for the proposed national customer satisfaction survey. Lack of centralized and automated systems does not prevent drawing representative state samples, but does make it considerably more expensive.

The CSE universe can be construed as consisting of cases or clients. There is an important distinction here between the terms case and client. The federal government uses the following definition of a Title IV-D case: "A parent (mother, father, or putative father) who is now or eventually may be obligated under law for the support of a child or children receiving services under the Title IV-D program (OSCE, 1997). Each case consists of several clients: usually the mother, the father, and one or more children. Typically, one parent is the payor and the other parent is the payee. However, in some cases both parents are payors and must pay child support, usually on behalf of a child in foster care.

It is our understanding that in states with centralized databases, all individuals currently are, or will eventually be, linked to cases. However, when there is no automated statewide database, such linkage is not always possible. For sampling purposes, this linkage is important, particularly when a parent is associated with more than one case. This happens, for instance, if a noncustodial father is providing support for the children of two different mothers. He will be associated with two cases: the first case will consist of this father, the first mother, and their joint child(ren); the second case will consist of the same father, the second mother, and their joint child(ren). If a woman has children by three different men who owe her child support, she will be linked to three different cases.

For sampling purposes, it is important to confirm that certified state central registries will have only one database record for each individual, even though individuals may be linked to multiple cases.

Differences exist across states in the number and type of non-public assistance parents who receive CSE services. Child support orders may be established by the courts or through judicial or administrative action of a CSE agency. All orders established by a CSE agency are de facto CSE IV-D cases. When the child support order is established through the courts and the family has never received public assistance, CSE services are not necessarily provided. In some states (for instance, Ohio), all court-established support orders include a mandatory application to CSE for services. In other states, CSE services for non-public assistance support orders are voluntary and are provided only if one or both parents apply for these services. What percent of non-public assistance cases voluntarily apply for CSE services is unknown. Although there are no hard data to confirm this, it is generally believed that judges are more likely to refer low-income parents to CSE agencies. Higher-income families are more likely to use private attorneys and less likely to be referred to CSE agencies by the courts. This variation across states in the composition of the non-public assistance caseload could have an effect on the across-state comparability of data collected from a national survey.

Most, if not all, CSE databases include "arrears-only" cases. Generally speaking, any debt of more than $500 owed by the custodial parent to the state is not "forgiven" when the child dies or is emancipated or even when the custodial parent dies. If the debt is owed to the custodial parent, he or she may or may not choose to waive the debt. This means that state databases include "arrears-only" cases. We found differences among sites in how, whether, and when they decide that a debt is uncollectible and therefore close the case. In some states, "arrears-only" cases may remain open almost indefinitely; in other states, such cases are closed fairly soon and can be reopened as needed (for instance, in the unlikely case that the debtor wins the lottery).

Each state's database appears to be configured differently. This means that sampling statisticians would have to work with each state to define and build a sampling frame. It also appears that in some states the case is in the name of the custodial parent who is viewed as the key service recipient, while in other states the case is in the name of the noncustodial parent who is the focus of wage withholding and other support enforcement efforts. Some states also include in the same database cases that receive services under specific state regulations but are not counted as IV-D cases by the federal government (e.g., "spousal support-only" or "payment-services-only" cases, the latter being cases in which non-public assistance parents have asked the state to collect and distribute the child support, but have not applied for any of the other CSE services, such as enforcement).

Telephone numbers for clients are often not available or up-to-date. Local sites invariably report not having much confidence in the validity of the telephone numbers listed in the database. CSE workers usually contact clients by mail rather than by telephone and therefore rarely have a need to check the accuracy of the telephone numbers. Moreover, it may be difficult to use criss-cross directories and similar techniques to determine the telephone number, as the mailing address may not be the person's home address but rather a post office box, or the address of an employer or relative. Thus, conducting a telephone survey of CSE clients (or a telephone follow-up to a mail survey) may be expensive because of the costs of obtaining correct telephone numbers.

While recognizing that certain difficulties may be encountered, we believe that it is feasible to conduct a nationally representative survey of CSE customers. Before OCSE implements a full-scale survey, or even a pilot survey, however, we recommend that it undertake a design study to create and pretest a data collection instrument and flesh out the sampling, data collection, and analysis plans. The results of the design phase should then be subjected to a rigorous pilot test. This test could be done as the precursor to the national survey or as the first step in its implementation.

The recommendations below are based on the assumption that OCSE is planning an ongoing customer survey to be conducted annually, or perhaps every other year, so that changes in client satisfaction can be measured over time and used to make program improvements. Further, these recommendations are given as starting points for testing and refinement during the proposed design and pilot phases.

We recommend no changes to the working definition of customer satisfaction developed early in this study. This definition addresses three aspects of customer satisfaction: a) satisfaction with client-worker interaction, whether in-person, by phone, or by mail; b) satisfaction with the support payment (e.g., its accuracy and timeliness); and c) satisfaction with the effect of CSE on the child. The first two aspects were successfully covered in the state and local surveys. The pilot test will have to determine whether it is feasible to ask parents about their perception of the effect of CSE services on children.

As parents and caretakers move through the CSE system, they may receive multiple services (i.e., location of the absent parent, paternity establishment, establishment of child support, collection and distribution of child support, and child support enforcement). They may also be associated with more than one case. In designing a CSE customer satisfaction survey, the following questions arise: Should clients be asked about the one service they are currently receiving, or about all services received in the past "X" months? Should parents who are involved in more than one case be interviewed about one or all cases?

Whether to ask respondents about more than one case. The level of satisfaction may vary significantly for each case to which the respondent is linked. For instance, a custodial mother may be extremely satisfied with the timely and accurate payments received from the father of one of her children, and extremely dissatisfied with the noncompliance of the father of another child and with the way the CSE worker is handling the case. We therefore believe that it is preferable to ask the individual about only one case. Asking the individual about a single case will also make it easier to link the survey data to specific case characteristics. Focus groups and other pretest mechanisms should be used to determine whether respondents can distinguish between cases.

Whether to ask respondents about multiple CSE services. We recommend asking respondents about all services received for the one case about which they are being interviewed. We suggest OCSE use focus groups to determine what time period should be covered by the questionnaire--probably the past three to six months. Longer time frames may lead to problems of recall.

For customer satisfaction surveys, the main modes of data collection are the mail survey and the telephone survey, or a combination of these two approaches. As a rule, we favor the somewhat more expensive telephone approach because it yields higher response rates, but in this case we recommend a mail survey supplemented by telephone follow-up. As mentioned above, it may not be feasible to obtain accurate telephone numbers for many of the parents. In addition, some low-income CSE parents, like other low-income persons, may not have telephone service. Thus, for the OCSE survey, the location costs are likely to make a telephone survey too expensive. We believe a mail survey, with telephone follow-up as needed, is more appropriate. The mailing addresses are more likely to be up-to-date (especially for parents in cases where the noncustodial parent is compliant).

The proposed sampling plan consists of two stages: the first stage involves sampling counties; the second stage involves sampling CSE clients within the selected counties.

Sampling counties. Actually, our first choice is not to sample counties but rather to select from all states, if at all feasible. However, the extent of difficulty associated with gaining state cooperation is unclear. If convincing states to cooperate proves to be labor-intensive, a clustered sampling design would have to be considered. Using a clustered approach, one would start by randomly selecting counties from a nationwide list of counties. If, for instance, OCSE sampled 50 counties nationwide, these counties would probably be located in 20 to 30 states. This means that OCSE would need to obtain cooperation from far fewer states, thus reducing the costs of gaining cooperation and obtaining sampling frames. However, clustering would reduce the precision of survey estimates over that of a simple random sample collected from all states.⁽¹⁾

Note that we are not proposing that OCSE include all states in order to make state-level comparisons (the samples will be too small for that). Rather, the intent is to a) increase the precision of the national survey estimates, and b) allow states, if they so choose, to subsidize an enlarged state-specific sample for optional state-specific surveys.

Sampling individuals. Assuming that individuals will be asked about specific cases, each state's sampling frame will need to have multiple records for individuals involved in multiple cases. Our approach calls for two sampling frames for selecting individuals by case, the first frame consisting of noncustodial parents, and a second frame of custodial parents.

Before the sample is drawn, it will be necessary to a) remove from the sampling frame certain case types that OCSE determines to be outside the scope of the study, and b) decide whether to stratify the sample by domains of analytic interest to OCSE.

If OCSE determines that certain case types fall outside the scope of the study, these cases need to be removed from the sampling frame. The following cases appear to be candidates for exclusion from the survey:

We propose that the two samples be stratified by one or more domains of key analytic importance to OCSE , such as the public assistance status of the case to which the parent is linked,⁽²⁾ or the type of CSE services currently received by the parents. These domains must be definable based upon variables available in the database records of the sampled individuals. For separate estimation capability, it may be necessary to oversample for certain domains if the expected number of individuals in these strata would otherwise be small (for instance, the number of sampled individuals in paternity establishment cases). Note that the number of domains for which separate estimation capability is desired increases the size and therefore the cost of the sample. Our Working Paper on Sampling Strategy (Cox, 1997) presents the procedures for sampling individuals.

The sample size is dependent on the level of desired precision. It is essential that the sample be as representative as possible of the population from which it is drawn because the sample statistics will be used to make inferences about the corresponding population parameter. Because any individual statistic is not likely to be exactly the same as the population parameter, it is important to establish the range of values that contain the parameter. The confidence level establishes the range into which the population parameter is likely to fall. The Working Paper on Sampling Strategy (Cox, 1997) presents precision results based on the standard 95 percent confidence level. A 95 percent confidence level means that if repeated random samples were drawn from the same population, 95 percent of the confidence intervals would cover the population parameter of interest. If the level of confidence is decreased, the interval becomes wider. For customer satisfaction surveys, a lower confidence level, perhaps as low as 90 percent, might be sufficient, even though this would increase the probability of error. It has been argued that it is reasonable to accept the possibility of larger errors for customer surveys because the consequences are not as dire as they might be for other types of surveys (Hatry et al., 1998). The decision on which confidence level to use in designing the survey will depend on how OCSE plans to use the findings, and the risks associated with having less precision for the findings. These factors will have to be weighed against the higher costs associated with the larger sample size needed to obtain a greater level of precision.

It is not possible at this stage to specify the expected sample size. Typically, national surveys have sample sizes that range from 2,000 to 5,000. One of the objectives of the proposed design phase will be to determine what the sample size should be. Factors to be considered in determining sample size include a) the desired confidence level, and b) the need for comparisons across and within domains.

The design presented is based on the assumption that the various state privacy acts will allow the state to release names and other identifying information, such as telephone numbers and addresses, of sampled individuals to OCSE or its contractor for purposes of conducting a national customer satisfaction survey.

For a mail survey, the questionnaire needs to be brief and formatted in such a way that it ideally consists of only one or two pages. It is essential that the format be appealing to the users or they may not respond. The wording needs to be simple, the answer categories precise, the layout attractive, and the cover letter persuasive. The cover letter provides information on the purpose of the survey, encourages respondents to participate, and lets them know that the information they provide is confidential, that participation is voluntary, and that refusal to participate will in no way affect eligibility for services. For telephone interviews, this information needs to be read clearly to the respondent prior to asking any of the survey questions.

The existing questionnaire developed by OCSE, and pretested in several states (including Washington and Ohio), provides an excellent starting point for the development of the survey instrument. We suggest that the content, wording, and coverage of the questionnaire be further tested through focus groups and pretests. Such tests, for instance, can determine whether the wording of the questions or instructions present difficulties for persons who do not read well.

Some but not all state registries will indicate what the individual's primary language is, if other than English. During the design phase, OCSE will be able to determine into what other languages the questionnaire needs to be translated, using appropriate back-translation techniques. The decision whether to translate will depend on the size of the population for whom translation is necessary.

An adequate response rate is key to determining the confidence one has in survey findings. For the proposed OCSE survey, it should be recognized up front that certain potential respondents are for all intents and purposes unreachable: a) noncustodial parents in cases that are in "locate" status (on the assumption that if they cannot be located by CSE staff, they will also be nonlocatable for purposes of the survey); b) noncustodial parents in "good cause" cases where the state has given a safety waiver and is not collecting child support because of the possibility of abuse to the custodial parent and the children; and c) noncustodial parents who are institutionalized, for instance, in mental hospitals.⁽³⁾ These individuals might be determined ineligible for participating in the survey only after the sample was drawn. This approach ensures that OCSE will be able to report how many noncustodial parents fit these de facto "nonresponse" categories. However, we recommend interviewing the custodial parents associated with these cases.

For most social science research projects, a response rate of 75 percent or higher is desirable. We recognize that it may be difficult to obtain that high a response rate, especially for noncustodial parents. One outcome of the pilot will be to determine what response rate OCSE can expect from the various respondent groups, and for which subgroups it will be more difficult to obtain an acceptable rate. Further, it will tell us why some groups have lower rates: because of refusal to participate or because of our inability to contact them. For instance, for noncustodial parents who are not in compliance with their child support orders, there may be no up-to-date address in the database.⁽⁴⁾

We suggest the following strategies to increase the response rate. For the proposed mail survey, we suggest that the states inform their CSE client population of the planned survey through inserts in routine mailings to their clients and through an announcement on the KIDSline. For the selected respondents, we suggest reminder postcards, and a minimum of two subsequent mailings of the questionnaire to nonrespondents, followed up if possible by a telephone call. Other techniques to improve response to mail surveys should be adopted as well, such as use of stamps rather than machine-imprinted postage, personalized letters, and envelopes that request address corrections from the U.S. Postal Service. If the survey is instituted on a regular basis, we suggest that states inform new clients of the periodic survey.

The unit of analysis can be a) the case, b) an individual without any reference to a particular case, or c), as we suggest, an individual associated with a particular case. If the sampled individual is associated with multiple cases, there will be multiple records in the sampling frame for this individual. If selected for the sample, the individual would be asked about a specific case, and data from that specific case would be added to the survey data, thus allowing subsequent analytical breakouts.

This approach is predicated on the assumption that the individual parent will be able to distinguish between cases and can report on services received for a specific case. As mentioned earlier, the design phase can determine whether this approach is feasible or whether OCSE should ask respondents global questions about all cases in which they are involved.

We do not recommend pairing custodial and noncustodial parents associated with a joint case as we expect that for many cases it will not be possible to get responses from both parents. Some categories of noncustodial parents will be excluded from the data collection process (e.g., those who are institutionalized or whom the state has been unable to locate); other noncustodial parents, especially those who are noncompliant, may refuse to participate in the survey.

In the sampling section, we discussed the importance of identifying domains of special analytic interest to make sure that there are enough cases to make comparisons across domains (for instance, comparisons of current public assistance, former public assistance, and never public assistance cases). There are other comparisons that OCSE may want to make if it so happens that there are enough sampled units with these particular characteristics. Below we suggest breakouts based on a) respondent characteristics, b) case characteristics, and c) state characteristics.

Breakouts by respondent characteristics. The following two characteristics of the individual respondent are likely to be of interest to OCSE and the states: race or ethnicity, and sex. Breakouts by race and ethnicity may provide valuable feedback on the cultural appropriateness of written materials. Breakouts by sex may provide information on CSE workers' attitudes toward female rather than male noncustodial parents, and male rather than female custodial parents.

Breakouts by case characteristics. OCSE and the states will undoubtedly want to know whether custodial and noncustodial parents associated with certain types of cases are more or less satisfied than others. These case characteristics include:

Characteristics of the states. Assuming an unclustered design was used, in which sampling occurred within all states, one could disaggregate on other dimensions such as geography or characteristics of the state CSE system. While it is unlikely that the sample size will allow state-level comparisons, it may be possible to make comparisons of states with like characteristics. Note, however, that as more disaggregation is insisted upon, the required sample sizes for the total survey increase. The following are possible ways that OCSE may want to group states in order to look at comparisons across states:

To collect data from more than nine individuals using the same data collection instrument requires clearance from the Office of Management and Budget (OMB). Although initial testing of the instrument may involve iterative testing of several versions of the instrument, it will eventually be necessary to conduct a larger scale pilot of the instrument, as well as of the data collection and sampling procedures. We therefore recommend that an OMB clearance package be developed for the design phase. This preliminary clearance would lay the groundwork for obtaining clearance for a full-scale implementation of the survey. Thus, an end product of the design phase would be an OMB clearance document that would include the survey procedures, the tested survey instrument, the sample design, and the analysis plan.

We strongly encourage OCSE to provide states with the technical assistance and support needed to allow them to piggyback on the national study and conduct a state-specific customer satisfaction survey. In all likelihood, the state samples will be too small to allow any state-specific analysis. However, paying for an expanded state sample will allow a state to collect customer satisfaction data at a fraction of what it would cost to conduct a survey "from scratch." It would also provide the state with a convenient, low-cost approach for doing quality work in surveying their customers, and a greater understanding of, and investment in, the findings of the national CSE customer survey. We suggest that such federal/state collaboration be further explored during the design phase.

The feasibility of the recommendations presented in Section II of this report, and their underlying assumptions, will need to be tested. Many decisions will depend on the needs of OCSE and other users. We therefore propose that OCSE conduct focus groups of internal as well as external customers to explore the ramifications of the various options. We propose that the design phase explore and test the following aspects of the proposed survey:

Questionnaire content and format. If OCSE decides to proceed with the mail survey, an important task during the design phase will be the further development of the existing OCSE draft customer satisfaction survey form. We expect questionnaire development to be an iterative process, consisting of focus groups, pretests, and review by CSE experts. Questions to be tested include the following:

Response rates. A first step after designing and pretesting the questionnaire wording, format, and content will be to determine the response rate for the various subgroups and then to find the most cost-effective method for increasing response rates. Because of the likelihood that state CSE databases will not have current telephone numbers for potential respondents and because of the higher costs associated with telephone surveys, we recommend that OCSE pilot test a mail survey protocol (with telephone follow-up, if judged economically feasible). The design phase will allow OCSE to test:

Obtaining samples from the states.The sampling plan proposed in Section II of this report is based on the assumptions that a) selected states will be willing to draw the specified sample of custodial and noncustodial parents, and b) that each state's centralized database will be configured in such a way that it will be possible to select custodial and noncustodial parents by case. These and other assumptions will have to be tested:

Sample size. The design phase will allow OCSE to determine the size of the sample needed for a national survey of CSE customers. As mentioned above, samples of 2,000 to 5,000 are fairly typical for national surveys. The size of the OCSE sample will depend on a number of factors: whether stratification is needed or not, the need for comparisons between and across domains, and the level of precision needed. These issues will need to be explored in depth through focus groups of users (internal as well as external to OCSE).

Cost. Undoubtedly, cost constraints will be an important consideration, and the costs associated with each option should be carefully addressed during the design phase. The two possible obstacles to a national representative CSE customer survey are a) cost, and b) response rates. The two are inexorably linked together. Although the cost will depend on many factors (including sample size, number and type of breakouts, length of questionnaire), a key factor will be the level of effort required to obtain reasonable response rates from all groups, even those difficult-to-reach ones, such as noncompliant custodial parents. The proposed design phase will allow OCSE to examine each of these factors and to make an informed cost estimate. At this stage, it would be premature to do so.

We believe that a proposed national customer survey of OCSE customers, specifically of custodial and noncustodial parents and caretaker relatives, is feasible. It is at this point , however, not possible to estimate the cost of such a survey. It will first be necessary to further explore several feasibility issues, such as the willingness and ability of sampled states to assist in the drawing of representative samples and the response rate of parents and caretakers. It is anticipated that a significant amount of effort will be required to obtain a reasonable response rate from certain subgroups of respondents, such as noncompliant noncustodial parents.

We therefore suggest a preliminary design phase to further explore and test our recommendations and the assumptions on which they are based. At the conclusion of the design phase, OCSE should have all the needed design elements in place for a national pilot test of all data collection instruments, sampling procedures, and analyses.

References

Cox, B.G. (1997, December). Feasibility of conducting a child support enforcement customer satisfaction survey. A working paper on sampling strategy. Paper submitted to the Office of Child Support Enforcement, DHHS, under contract no. HHS-100-95-0021.

Hatry, H., Marcotte, J., van Houten, T. & Weiss, C. (1998). Customer surveys for agency managers: What agency managers need to know. Washington, DC: The Urban Institute.

Office of Child Support Enforcement, Administration for children and Families, U.S. Department of Health and Human Services (1997, November 14). Definitions to use in completing OSCE-157, Child Support Enforcement Program Annual Data Report. Washington, DC.

Office of Child Support Enforcement, Administration for Children and Families, U.S. Department of Health and Human Services (1996, March). Evaluation of child support guidelines. Volume I. Findings and conclusions. Washington, DC.

U.S. Government Accounting Office (1997, June). Child Support Enforcement: Strong leadership required to maximize benefits of automated systems. Report to congressional requesters. GAO/AIMD-97-72.

U.S. Government Accounting Office (1996, October). Child Support Enforcement: Reorienting management toward achieving better program results. Chapter Report. GAO/HEHS/GGD-97-14

The intent of the site visit is to learn more about current or previous state, or county-sponsored customer satisfaction surveys. The states for this study (Massachusetts, Minnesota, Missouri, Ohio, Tennessee, and Washington) were selected because of their customer satisfaction initiatives. The lessons learned by these sites may provide useful information regarding the feasibility of conducting a national CSE satisfaction survey. Since these surveys were conducted several years ago, it is possible that problems that existed then (for instance, difficulties in getting a representative sample) have since been resolved as a result of increased automation or more frequent updating of records. The interviews will therefore cover not only the earlier customer satisfaction survey, but also changes that have been made since then in local CSE administration that could help or hinder the design of a customer satisfaction survey.

The following questions and topics will be covered during the site visits and telephone calls to selected states. We will be collecting data on site in Washington and Ohio and through phone interviews from Massachusetts, Minnesota, Missouri, and Tennessee.

Questions that may affect the feasibility (and costs) of selecting a representative national sample of CSE parents:

^1. For the pilot test, it does not seem cost-effective to develop sampling procedures for states without a centralized registry. A full-scale CSE customer satisfaction survey probably will not be implemented for another year or so. By then, virtually every state might be expected to have an automated, centralized CSE data system available for use in sample selection for the full scale survey. (It is not essential that the state's CSE data system be fully certified.) However, automated, centralized data systems are needed to minimize operational costs and control data quality. The accuracy of the automation assumption should be reassessed periodically by noting the speed with which the states which are not as yet certified complete the various stages of certification. If necessary, of course, it is feasible, although more expensive, to develop sampling strategies for states without a centralized database.

^2.The federal government uses three categories to group CSE cases (i.e., cases entitled to CSE services under Title IV-D of the Social Security Act) by public assistance status. It distinguishes between current assistance cases, former assistance cases, and never assistance cases.

· Current assistance cases are cases where the children are recipients of Temporary Assistance for Needy Families (TANF) under Title IV-A of the Social Security Act, or entitled to Foster Care maintenance payments under Title IV-E of the Social Security Act.

· Former assistance cases are cases where the children formerly received Title IV-A (AFDC or TANF) or IV-E Foster Care services. The state and local surveys examined for this feasibility study categorized these cases as non-public assistance cases.

· Never assistance cases are cases where the children are receiving services under the Title IV-D program, but are not currently eligible for and have not previously received assistance under Titles IV-A or IV-E of the Social Security Act. Included are cases where the children are receiving state (not Title IV-E) foster care services or Medicaid-only cases. The latter are cases where the children are determined eligible for, or are receiving, Medicaid under Title XIX of the Social Security Act, but are not current or former recipients of aid under Title IV-A or IV-E of the Act.

^3. Note, however, that for the pilot test we suggest testing the response rate of individuals who are incarcerated.

^4. If it proves very expensive to survey particular subgroups due to difficulty in locating them or obtaining cooperation, these groups might be sampled at a lower rate to provide a more cost-efficient sample