• About the Study
• Methodology
  • Instrument Development
  • Training
  • Agency Survey
  • Client Survey
  • Response Rates
  • Analysis Methods
  • Reporting Conventions
  • Degree of Comparability
    • Estimating Turnover Rates
    • Provider Screening
    • More Than One Visit
    • Longer Hours
    • Number of Participating Food Banks
    • Open Less Often
    • Small Providers
    • Summary
  • Number of A2H Clients Served
• Key Findings
• Who We Serve
• How We Serve
• Local Picture
• Snapshot Comparisons
• Media Center

Number of Clients Served by Small Providers (3.8.7)

During the 2001 survey, some food banks, especially those in rural areas or with agencies in remote areas, were burdened with sending teams of interviewers to very small providers that might have as few as two or three clients in a day. In 2005, we reduced the burden on food banks by excluding very small providers from the survey work and imputing values for them in the data. Our rationale was that it would reduce data collection burden and increase cooperation, and because, by definition, these providers were small and could be accounted for by an adjustment factor in computing the sampling weights, the analysis would not be significantly affected. In general, the criterion chosen was not to send interviewers to providers that served fewer than 10 clients a day.

During the analysis, we considered several methods for accounting for small providers in the estimates. One was to incorporate a design-consistent ratio adjustment factor directly into the weighting process. In this adjustment, the number of clients at small providers would have been accounted for by using the ratio of the projected number of clients at all providers to the projected count of clients at large providers (with the ratio developed by provider type within food bank using 2005 sampling data from the individual food banks). Other methods considered involved imputing the estimates for clients at small providers based on data collected in 2001, when interviewing was done at small providers. We performed estimates using both approaches and made judgments about the likely accuracy of the methods in discussions with the Technical Advisory Group for the project and with the MPR analysis team and sampling advisors. Based on these discussions, the method used was based on a variant of the imputation approach. This approach was decided by the Technical Advisory Group and the MPR analysis team to be the most accurate. This final approach was then implemented technically by including aratio adjustment in the weighting process (by provider type), with the ratio set to achieve the estimate determined in the analysis described above.