B. Collections of Information Employing Statistical Methods
1. Respondent Universe and Sampling Methods
Most of the information collected through the CFSP and APSR is narrative in nature and does not lend itself to statistical methods. All State and Tribal respondents are required to submit the information in order to receive funding under the several programs covered by the information collect and therefore sampling does not apply to the universe of respondents for the data collection as a whole.
However, in order to reduce the cost burden, States may choose to use sampling for the collection of caseworker visit data which is one required piece of information the State must submit in order to receive funds under title IV-B, subpart 1.
The Children’s Bureau will work with States to ensure that they employ a sampling strategy for the States’ caseworker visit data collection that is statistically sound and that will yield accurate, reliable, and valid data as well as ease the reporting burden for larger States.
2. Procedures for the collection of information.
The Children's Bureau statisticians have developed a probability sampling plan based on attributes that States can opt to use to comply with the reporting requirements for caseworker visits. The sampling criteria include a confidence coefficient of .95, a tolerable sampling error of .05, and an estimated attribute proportion of .5 to maximize the sample size. The standard sample size formula for attributes when sampling from a finite population was used. The sample sizes will range from 370 to 390.
With the Children's Bureau sampling methodology, the sample is drawn from the Adoption and Foster Care Analysis and Reporting System (AFCARS,) to which each State is required to submit data on children in the foster care system. States’ submissions are made semi-annually. The AFCARS “A” period is from October 1 through March 31 and the “B” period is from April 1 through September 30.
The AFCARS A and B submissions would be used to create three sampling frame files; the children in care on September 30, the exits occurring during the AFCARS B period, and the exits occurring during the AFCARS A period. The Children's Bureau statisticians will provide the required sample size for each State based on the total number of children in these three files. The sample size will be allocated proportionately among the three sampling frames. For example, if the required sample size were 300 children and there were 10,000 children in the September 30th sampling frame and 5,000 children in each of the two exit sampling frames yielding a total of 20,000 children, then 50% (10,000/20,000) or 150 sample cases (plus a 10% oversample) would be selected from the September 30th frame and 75 (5,000/20,000) children (plus a 10% oversample) would be selected from each of the two exit frames.
The required data would be retrieved by State caseworkers from the case files of the sample foster care cases and the data would be entered on a spreadsheet developed by the Children’s Bureau statisticians. Once the data for all children are collected and entered onto the spreadsheet, aggregated and entered into the cells of the spreadsheet’s algorithms, the desired percentages will be computed. The only quality control aspect available to the Children’s Bureau would be to check the consistency of the data entered on the spreadsheet as we do not have access to the case records of the sample cases.
The State caseworkers would perform the following activities to obtain the necessary data to be entered on the spreadsheets. Each child from the September 30th sample would be tracked back to October 1st or the day he or she entered care if later than October 1st. Each child in the exit samples would also be tracked back to October 1st or the day he or she entered care if later than October 1st. In this way, every child served during the AFCARS AB periods would have a chance to be selected for the samples. Working backwards makes more sense than sampling on a monthly basis as this would often lead to complications. The Children's Bureau statisticians have developed the syntax that will generate the three sampling frames as well as indicating how many full months each sampled child was in care during the AFCARS AB period.
If any States that intend to utilize sampling to comply with this reporting requirement wish to adopt this sampling strategy, then they would have the option of selecting their own samples with the Children’s Bureau’s syntax, or have the Bureau’s statisticians select the samples after a State’s AFCARS B file is submitted.
If a State chooses to use its own sampling plan, the Children's Bureau statisticians must review it in order to ensure that it will yield results that can be generalized to the entire population.
Methods to Maximize Response Rates and Deal with Nonresponse.
This question is not applicable. All States are required to submit caseworker visit data for each child who meets the reporting requirement of having been in foster care for at least one full calendar month, therefore, non-response will not be an issue. Also, the sampling routine will yield data that can be generalized to the universe.
4. Tests of Procedures or Methods to be Undertaken.
The sampling methodology was tested by nine States. The sampling methodology was found to be workable and to yield reliable and accurate data that can be generalized to the universe.
5. Individuals Consulted on Statistical Aspects and Individuals Collection and /or Analyzing Data
John Gaudiosi, Mathematical Statistician
HHS, Administration for Children and Families,
Children’s Bureau
202-205-8625
| File Type | application/msword |
| Last Modified By | Department of Health and Human Services |
| File Modified | 2011-07-25 |
| File Created | 2011-07-25 |