• About the Study
  • Introduction
  • Sponsor Statement
  • Service Area Map
  • Download the Reports
  • Appendices
    • A: Simulating Methods
    • B: Details of Calculation
    • C: Food Stamp Eligibility
    • D: Sources
    • Technical Appendix
• Methodology
• Key Findings
• Who We Serve
• How We Serve
• Local Picture
• Snapshot Comparisons
• Media Center

Appendix A: Simulating 2005 Methods on 2001 Data

To evaluate further the effects of changes in methodology between the 2001 and 2005 studies, we have, to the extent possible, applied 2005 methods to the 2001 data to examine what results would have been found in 2001 if the new methods had been used.  This appendix summarizes the results.

To conduct the analysis, we began with the 2001 participant estimates as presented in the 2001 report and then simulated what would have happened to these estimates if certain of the 2005 methods had been employed.  For instance, to simulate the tighter ineligibility screening applied to providers in 2005, we increased the assumed rates of provider ineligibility in the 2001 data to bring these rates up to the observed 2005 levels.

Four components of the methodological changes were simulated:

1. Increased provider eligibility screening (which decreases the 2001 estimates)
2. Consideration of the fact that some 2001 providers were open longer than the on-site interviewing period (which increases the 2001 estimates)
3. Recognition of the fact that some providers are open less than once a week (which decreases the 2001 estimates)
4. Improvement in the analytic treatment of multiple interviewing days at the same provider (which decreases the 2001 estimate)

Note that in some instances the 2001 data were not sufficiently detailed to simulate the changed methodology fully.  It is also possible that not all changes have been recognized and included in the simulation.  For these reasons, the simulations should be regarded as likely underestimates of the full effects of the changes made in 2005.  They do, however, provide some indication of their approximate magnitude.
In simulating the changes, to facilitate estimating sampling variation, we have focused on the sampling units as they were defined in the studies in both years.  There are households for pantry users and adults using the provider’s services for both kitchens and pantries.

The results are shown in Table A.1.

The top panel of the table (the first three rows of numbers) shows how results just for participating food banks.  The bottom panel shows results after using the 1.5 and 1.23 “expansion factors” to convert to all A2H food banks in 2001 and 2005, respectively.

In general, after the simulations, the 2001 numbers remain higher than the 2005 estimates.  This is due, at least in part, to the difficulties of fully carrying out the simulation, as noted earlier.  However, in no instance is there a statistically significant difference between the adjusted 2001 estimate and the 2005 estimate.  In general, these results are supportive of the belief that the differences between the 2001 and the 2005 data are due largely to the methodological changes that were made.

TABLE A.1
SIMULATING 2005 METHODS ON 2001 DATA

	Original 2001 Estimates	2001 Estimates Adjusted for 2005 Methods	2005 Estimates
Participating food banks only
Pantry clients (households)	1,425,000	1,183,000	1,237,000
Kitchen clients (adults)	494,000	416,000	455,000
Pantry clients (adults)	190,000	151,000	181.000
All A2H food banks
Pantry clients (households)	2,138,000	1,774,000 (s.e. = 124,000)	1,525,000 (se = 48,000)
Kitchen clients (adults)	741,000	624,000 (s.e. = 75,000)	547,000 (s.e. = 30,000)
Pantry clients (adults)	285,000	226,000 (s.e. = 52,000)	223,000 (s.e. = 12,000)