Data Variables and Sources
Information in this dataset is provided in an Excel spreadsheet, consisting of four worksheets:
- Notes—describes coding conventions and data sources,
- Flag—identifies States and years in which program benefit data have been imputed by ERS (described below),
- State—provides State-level data, and
- County—provides county-level data. In general, the same variables are available on both the State- and county-level files.
This dataset draws from a number of different sources and represents estimates of true underlying values. Data for subsequent years will be added to the dataset as updates are provided by the Census Bureau and the Bureau of Economic Analysis.
STATE_CD: Two-digit U.S. Postal Service code identifying State.
STFIPS: Two-digit Federal Information Processing Standards (FIPS) code identifying State.
County-level file only:
FIPS: Federal Information Processing Standards code.
FIPSTXT: FIPS code in text format with leading zeros.
COFIPS: County code portion of total FIPS code (last three digits), not unique across States.
CODEF: Source of county definition: BEA_ONLY for Bureau of Economic Analysis, CEN_ONLY for Census Bureau, and CEN_BEA when both Census Bureau and BEA use the same definition.
CONAME: County name and two-digit State identifier.
URBCODE: 2003 Rural-Urban Continuum Code per ERS. For more information about the rural-urban code, see 2003 Rural-Urban Continuum Codes.
METRO: Metro/nonmetro code per U. S. Census Bureau. For the metropolitan/non-metropolitan codes used in the data set, see Rural Definitions: Data Documentation and Methods.
SNAP participant counts
Participation counts at the county level are provided primarily by the Small Area Income & Poverty Estimates program of the U.S. Census Bureau. According to the Census Bureau's website, the Census Bureau obtains raw data from the Food and Nutrition Service (FNS), USDA. For most States, the Census Bureau uses counts of participants for the month of July, although some States can provide data only for other reference periods.
For State-level figures, the Census Bureau calculates the number of participants as a 12-month average running from July to June. Since FY 2007, State-level participation counts in the data set have been based on administrative records maintained by the Food and Nutrition Service (FNS). Thus, the sum of participation counts for all counties in a State (from Census Bureau) may not equal the State total (from FNS) due to different data sources and definitions. FNS data for participant counts are more accurate than estimates from the Census Bureau, but FNS data are not uniformly available at the county level.
PRGNUMxx: Estimated number of active SNAP/Food Stamp Program (FSP) participants. The last two digits (xx) indicate the year for the data. Data are available for 1989, 1993, 1995, and every year from 1997 through 2012 for States and through 2010 for counties. Some county-level data from FNS are available for more recent fiscal years.
County-level data for most years are provided by the Small Area Income & Poverty Estimates (SAIPE) section of the U.S. Census Bureau. The original data come from the Food and Nutrition Service (FNS), USDA.
Participant counts are usually for the month of July. The Census Bureau controls the county-level values to their State-level value. When the original data cover an area larger than a single county, the Census Bureau uses a measure of poverty in the multiple areas to apportion the SNAP/FSP caseload numbers across the multiple areas.
State-level data through 2006 represent a 12-month average, with a 6-month delay. For instance, the values presented for 1995 are based on the monthly average between July 1995 and June 1996. Since 2007, the data are based on the Federal fiscal year; for example, the values for 2012 are based on the monthly average between October 2011 and September 2012.
Data fields for years prior to the introduction of the Food Stamp Program in a particular county or State are coded as "-9995." If the county's number of FSP participants was fewer than 50, the field is coded as "-9998." See "Coding Conventions" below for more details.
PRGxx_yy: Percentage change in the number of SNAP/FSP participants between the two years represented by "xx" and "yy" in the variable name. Five variables that are included—PRG89-93, PRG93-00, PRG00-05, PRG05-07, and PRG07-11—correspond to periods of major caseload increases, decreases, or stability.
Users may construct their own change variables using estimates for any combination of years. However, data are missing or not valid for some counties in some years, and the convention is to code any derived variable based on a variable with missing or invalid data as "-9993". See "Coding Conventions" below for more details.
SNAP participant benefits
Estimates of total annual benefits issued by area are provided by the Regional Economic Information System, Bureau of Economic Analysis of the U.S. Department of Commerce. State and county estimates are based on county tabulations of the distributed benefits from FNS, USDA.
PRGBENxx: Estimated total annual amount of SNAP/FSP benefits (in thousands) issued to State or county residents in a year. The last two digits indicate the calendar year for the data. Data are available for every calendar year from 1969 through 2010.
To obtain an estimate of average total benefits issued to all State or county residents during a month, multiply the data value by 1,000 and divide by 12.
These data are provided by the Regional Economic Accounts Directorate of the Bureau of Economic Analysis (BEA) of the U.S. Department of Commerce. The data measure the value of program benefits issued to qualifying low-income households to supplement their ability to purchase food. The State estimates are based on USDA's tabulations of the value of distributed benefits. County estimates are either based on payment data from the various State departments of social services or imputed.
State-level data through 2006 represent a 12-month average, with a 6-month delay. For instance, the values presented for 1995 are based on the monthly average between July 1995 and June 1996. Since 2007, the data are based on the Federal fiscal year; for example, the values for 2010 are based on the monthly average between October 2009 and September 2010.
When county data for a given year are not available, BEA imputes the county figures by applying the statewide percentage change in benefits that year to each county's benefit data from the prior year. States and years for which benefit data were imputed by BEA are identified by the variables BENFLGxx (see below).
The BEA imputations are subject to large errors because the procedure does not take into account different patterns of caseload growth (or decline) among counties within a State. ERS has used a different procedure to impute these annual benefit values. The procedure uses State-level data to compute an average benefit amount per participant per month. A county's estimated number of participants in a given year is then multiplied by the statewide average benefit (and multiplied by 12) to provide an imputed value of annual benefits issued to participants in that county. This imputation procedure assumes that average monthly benefit levels do not vary by county within a State, which is, of course, a simplification. Nevertheless, ERS's imputed values should be more useful than those provided by BEA for most research purposes. To reconstruct the BEA imputed value for any affected county, multiply the county's total benefits from the previous year by the statewide percentage change in benefits over the two years. Because BEA imputations often occur for several years in a row, the reconstruction of BEA's imputed values should begin with the earliest consecutive year for which benefits were imputed.
The ERS imputation procedure is also used to estimate benefit levels in geographic areas for which participant counts are available but benefits amounts are not (for example, counties marked as CEN_ONLY) and to correct BEA benefit estimates that lead to values of average monthly benefits per participant that clearly exceed program rules.
All benefit data and related variables are presented in nominal dollars and have not been adjusted for inflation.
In some States (notably Alaska, Hawaii, and Virginia), BEA data on benefits are for different geographic units than Census Bureau data on participant counts. To sum across county records to derive State or national totals and to avoid double-counting, use the variable CODEF to exclude either CEN_ONLY or BEA_ONLY records.
An additional modification to the PRGBENxx records is needed for Virginia. The Census Bureau treats many cities in Virginia as separate entities from their respective counties, providing caseload counts for both city and county. However, the BEA estimates of benefits issued are for the combined geographic area (independent city plus county). While benefits information for the combined area is retained on the file, pro-rated benefits (based on caseload counts) for the separate city and county records are also available on the file.
BENFLGxx: A flag indicating that the corresponding value of PRGBENxx—same year and location—was originally imputed by BEA and has been imputed by ERS using a different procedure. The variable equals 1 if the corresponding value is imputed and 0 if not. Program benefit data have not been checked for BEA imputations prior to 2001.
Area estimates of population counts are obtained from the U.S. Census Bureau's website which includes a description of the methods used to create the estimates.
POPxx: Estimated population count, with the year for the data indicated by the last two digits of the variable name. Data are provided for every year from 2000 through 2010. The reference date for estimates is July 1.
Data are provided by the Population Estimates Program of the U.S. Census Bureau. With each new release of July 1 estimates, the Census Bureau revises estimates for years back to the last decennial census. The updated estimates are available for 2010.
Area estimates of the number of persons in households with total income below 100 percent of the Federal poverty level (FPL) are provided by the Small Area Income & Poverty Estimates Program of the U.S. Census Bureau.
NUMPOVxx: Estimated total number of poor people, with the year for the data indicated by the last two digits of the variable name. Data are provided for 1999, 2000, 2006, 2007, and 2010.
Data are provided by the Small Area Income and Poverty Estimates (SAIPE) program of the U.S. Census Bureau. Estimates are based on models that combine survey data with population estimates and administrative records.
A number of variables in the file compare participant counts or program benefits to an underlying population group.
PRG_POPxx: The ratio of program participants to area population, with the last two digits of the variable name indicating the data year. Values are provided for 2000, and 2006-10.
PRG_POVxx: The ratio of program participants to area count of poor persons, with the last two digits of the variable name indicating the data year. Values are provided for 2000, 2006, 2007, and 2010.
Note that the value of this variable will be less than one to the extent that poor people are ineligible for the program for reasons other than income or that eligible persons elect not to participate in the program. The value may exceed one because the program's gross income limits are higher than 100 percent of the Federal poverty level.
AMB_POPxx: The ratio of total average monthly program benefits issued to all participants in an area to that area's total population count, with the last two digits of the variable name indicating the data year. Values are provided for 2000, 2006, 2008, and 2010.
AMB_POVxx: The ratio of total average monthly program benefits issued to all participants in an area to that area's count of poor persons, with the last two digits of the variable name indicating the data year. Values are provided for 2000 and 2006.
AMB_PARxx: The ratio of total average monthly program benefits issued to all participants in an area to that area's count of program participants, with the last two digits of the variable name indicating the data year. This variable estimates average monthly benefits per participant over the course of a year. Values are provided for all years from 2000 through 2011. Note: when benefit data have been imputed, the average monthly benefit per participant will be the same for all counties in the affected State.
Data values are not always available for every variable and every county. The following codes have been assigned for missing data in the original files, along with an explanation of the reason for which each code is used:
-9991: Data not yet available.
Null: No data, as indicated in the mapping application. This includes data missing for the following reasons:
-9992: Data not available for a county unit used by BEA (CODEF=BEA_ONLY).
-9993: Computed variable for which at least one underlying data value is missing.
-9994: Data not available for a county unit used by Census Bureau (CODEF=CEN_ONLY).
-9995: Data not defined for years prior to introduction of the Food Stamp Program (rollout of the FSP began in 1969 and was not fully nationwide until 1976).
-9996: Data missing due to data suppression by BEA or Census.
-9997: Data not available in specific year. Coded as (N) in original data files from BEA or Census.
-9998: Actual value less than 50. Coded as (L) in original data files from BEA or Census.
-9999: Data not available for reasons other than those listed above, including division by a true zero in a computed variable.
Economic Research Service (ERS), U.S. Department of Agriculture (USDA). Supplemental Nutrition Assistance Program (SNAP) Data System. http://www.ers.usda.gov/data-products/supplemental-nutrition-assistance-program-(snap)-data-system.aspx.