The Historical Supplemental Poverty Measure

In broadest terms, the definition of poverty is whether a family has sufficient resources to meet their basic needs. Since the late 1960s, the Census Bureau has released U.S. poverty statistics as part of the Current Population Survey's Annual Social and Economic Supplement (CPS ASEC), beginning with the Official Poverty Measure (OPM) in 1969 and adding the Supplemental Poverty Measure (SPM) in 2011. The Center on Poverty and Social Policy (CPSP) at Columbia University's School of Social Work has extended the SPM retroactively to 1967 (the reference year for the 1968 ASEC) and collaborated with IPUMS to make this "Historical SPM" available via IPUMS CPS.

This user guide provides a brief outline of the way poverty is measured by the Census Bureau, the differences between the OPM and SPM, a quick introduction to Historical SPM, and a discussion of "anchored" poverty variables. The final section of the note offers users guidance on issues to be aware of, further resources, and citation information for these data.

Measuring Poverty in the United States

To determine if a family's resources can meet their needs, one must first define family, resources, and needs. The OPM and the SPM define each of these differently (see Table 1), which has implications for the representativeness of each poverty measure and the ease of constructing it. The OPM employs straightforward definitions that are easy to generate in a wide variety of data sources. This widespread availability comes at a cost - critiques of the OPM note that it does not capture resource sharing in non-family households, omits income from social safety net programs and burdens from childcare and healthcare costs, and that the threshold does not reflect modern household expenditures or geographic variation.

In response to these criticisms of the OPM, social scientists at the Census Bureau and the Bureau of Labor Statistics (BLS) developed the SPM as an alternative measure of poverty in the US. The SPM was designed to paint a more comprehensive picture of poverty by taking advantage of the new and more detailed data in the CPS ASEC to offer a more nuanced picture of a family composition; available resources, accounting for tax liabilities, credits, and in-kind transfers; and a definition of needs that accounts for changing consumer spending patterns and differences in housing prices between geographic locations. Researchers at CPSP developed methods to extend the methodology for creating the SPM back in time to 1967 (represented by survey year 1968). CPSP's Historical SPM uses data from the CPS ASEC and the Consumer Expenditure Survey (CE), estimating or imputing data where the historical data are inadequate or missing.

Table 1. Comparison of OPM, SPM, and Historical SPM (adapted from Fox et al. 2015)
OPM SPM Historical SPM
Family / poverty unit definition Persons related by blood, marriage, or adoption OPM family, plus
  • unmarried partners and their children
  • foster children under 22
  • unrelated children under 15
 Same as SPM family, with estimated or imputed values for missing data on:
  • unmarried partners and their children identified using adjusted POSSLQ method before 1995
  • foster children under 22 (unidentified before 1988)
Resources Pretax money (cash) income including cash welfare, SSI, Social Security, unemployment insurance, worker's compensation, and veteran's payments OPM resources, plus
  • in-kind benefits (SNAP, school lunch, WIC, housing subsidies, and LIHEAP)
  • tax credits
  • federal stimulus payments
but less
  • federal, state, and payroll tax liabilities
  • out-of-pocket medical expenses (MOOP)
  • work/childcare expenses, capped at the earnings of the lower earning spouse or partner
  • work expenses are estimated using the Survey of Income and Program Participation (SIPP)
  • child support paid
 Same as SPM, with estimated or imputed values for missing data
  • SNAP and school lunch imputed prior to 1980
  • WIC participation is imputed prior to 2001, and WIC value is imputed for all years
  • receipt of housing assistance imputed prior to 1975
  • taxes are estimated prior to 1980 using NBER's Taxsim program
  • childcare and MOOP expenses are imputed using CE data
  • work expenses are estimated using the Survey of Income and Program Participation (SIPP)
Poverty thresholds

Based on minimal food diet costs in 1963 multiplied by three, updated for inflation, primarily by the CPI-U

Adjusted for family size and structure, age of household head (above or below 65)

Based on five years of consumer expenditure data on food, clothing, shelter, and utilities for all families with two children multiplied by 1.2 to account for other necessities

Adjusted for

  • family size and structure (adult-only, single or dual parent households and number of children)
  • housing tenure (renters, owners with mortgages, owners without mortgages)
  • differences in rental costs by geographic location1
 Same as SPM, with estimated or imputed values for missing data
  • based on fewer than five years of consumer expenditure data prior to 1984
  • mortgage status is imputed for all years
  • home ownership status is imputed prior to 1975
  • geographic adjustments are based on rent data from multiple sources1

1 Geographic adjustment factors are based on five-year estimates of median gross rents from the American Community Survey (ACS). Adjustments are made for metropolitan statistical areas, smaller cities, and nonmetropolitan areas. The Historical SPM uses the same method with the most granular data available in each year; precision (particularly in early years) reflects uneven availability of source data. For further reading, see the appendix of Nolan et al. 2016.

Family Definitions

The official Census definition of a family - which is used in the OPM - is "a group of two people or more (one of whom is the householder) related by birth, marriage, or adoption." The SPM family also includes unmarried cohabiting partners and their children, foster children under the age of 22, and any unrelated children under the age of 15 who share resources and expenses together.

Historical SPM Families

In calculating the Historical SPM, CPSP uses the Census Bureau's SPM definition of family unit (identified by SPMFAMUNIT) where possible. The CPS began identifying foster children in 1988 and unmarried cohabiting couples in 1995, initially identifying householders and their unmarried partners and then, beginning in 2007, any self-reported cohabitors. In prior years, these household members would have been classified either as "partner/roommate" or "other nonrelative." CPSP uses the adjusted POSSLQ method to capture unmarried partners in years before 1995. There is no established method for identifying foster children in the CPS before 1988. Therefore, in the Historical SPM, foster children may not be incorporated into families that they otherwise would have been if they were identifiable in the pre-1988 data.

Family Resources

OPM resources include a family's total pre-tax cash income from all sources, including wages, retirement, interest and dividends, social security, and veteran's benefits (FTOTVAL). SPM resources (SPMTOTRES) include the same cash income as the OPM and add the estimated value of food and nutrition program benefits (SPMLUNCH, SPMWIC, SPMSNAP), housing and utility supplements (SPMCAPHOUS, SPMHEAT), and tax credits such as the Earned Income Tax Credit (EITC) and Child Tax Credit (CTC) (SPMEITC). Resources are then adjusted downward to reflect the cost of federal (SPMFEDTAXAC), state (SPMSTTAX), and payroll taxes (SPMFICA), child support paid (SPMCHSUP), childcare and unreimbursed work expenses (SPMCAPXPNS), and out-of-pocket medical costs (SPMMEDXPNS).

Historical SPM Resources

The Historical SPM derived by CPSP uses the same basic methodology as the Census Bureau's calculation of resources in the SPM, but it must infer or impute values where the necessary variables for estimating SPM resources were not collected. This is particularly true for the early years of the CPS ASEC. Table 1 contains general information about which variables are imputed. Detailed information about how each variable in the Historical SPM was constructed is available in Fox et al. (2015).

Poverty Thresholds

The composition of the family unit underpins the construction of the poverty threshold in both the OPM and the SPM. The threshold for the OPM is pulled directly from a matrix of 48 possible values based on the number of adults and children in the household and the age of the householder. Each year, the original OPM thresholds (set in 1963) are updated for inflation based on the annual Consumer Price Index for All Urban Consumers (CPI-U). By contrast, SPM thresholds are calculated using a rolling five-year average of the out-of-pocket cost of food, clothing, shelter, and utilities (FCSU) as a baseline. The specific methodology changed in 2020 (2019 data is available using both the initial and the revised methodology). The 2005-2019 methodology averages FCSU for the 30th to 36th percentile of consumers as measured by the CE and multiplied by 1.2 to account for other miscellaneous household expenses. The 2020 methodology no longer counts telephone services as part of utilities; these are listed separately and both internet and select in-kind transfers (LIHEAP, NSLP, WIC, and rental assistance) were added (FCSUti). There are three separate baseline thresholds based on housing tenure (renters, owners with/without a mortgage; see SPMMORT), which are adjusted further to take into account family composition (SPMEQSCALE) and geographic differences in the cost of living (SPMGEOADJ).

Historical SPM Thresholds

The construction of the Historical SPM thresholds relies on the CE data series and follows the construction of the SPM thresholds, but adjustments are made to account for intermittent administration of the CE. The annual CE series begins in 1980 and the first year for which the full five-year average is available is 1984. From 1980-1983, thresholds are calculated from one to four years of data. Prior to 1980, only two sets of CE surveys exist: one in 1960/1961 and one in 1972/1973. CPSP generates thresholds for 1967-1979 by estimating thresholds for 1961 and 1972/1973 and then interpolating data for other years using the rate of change in the CPI-U.

The methodology for developing a historical geographical adjustment factor is particularly complex. The basic formula is the median rent for a given geographical area divided by the average national rent. As there is no canonical source of rent information that spans the entire period of the Historical SPM, CPSP used a variety of different datasets to determine the median rents in each year. A breakdown of how CPSP constructed the geographic adjustment factor can be found in Nolan et al. (2016).

Anchored Poverty Variables

The Supplemental Poverty Measure is a “quasi-relative” measure of poverty, in that it applies a consistent metric to contemporaneous consumer spending (e.g., under the 2005-2019 methodology, this was based on a rolling five-year average of the 30th to 36th percentile of expenditures on food, clothing, shelter, and utilities, plus 20% more to cover other essentials). As household expenditures change, so too does the poverty threshold. Because it is quasi-relative measure, changes in the SPM poverty rate over time reflect both changes resulting from the resources that families have and differences in living standards across time.

To create a more absolute-style measure of the historical SPM, CPSP created anchored poverty variables. These use the SPM method to generate poverty thresholds for a single year (the anchor year), and then derive thresholds for other years by adjusting the anchor year thresholds for inflation only, similar to the OPM method. By anchoring the poverty thresholds to a specific year and carrying them forward and backward in time (adjusted only for inflation), this anchored SPM addresses the question, "what would poverty rates have been across this period when considered against the anchor year's living standards?"

CPSP has provided three variables for measuring poverty using an anchored SPM threshold, with 20122 being the anchor year.

More information on the calculation of these anchored variables can be found in Wimer et al. (2016).

2 These values are for reference year 2012, and are found in the data for 2013 (as indicated by the variable YEAR).

Steps for constructing the geographically-adjusted SPM thresholds and calculating the SPMGEOADJ variable

CPSP calculates Historical SPM thresholds and geographic adjustments using the same methodology that the Bureau of Labor Statistics (BLS) uses for the CPS ASEC. The following example illustrates the formula for generating the SPM threshold and geographic adjustment for a 2-adult, 2-child family of renters in the New York City metro area in 2023:

The 2012-anchored threshold uses the same formula with one additional step to adjust the baseline threshold for inflation:

Using the threshold and geographic adjustment variables, users can work backwards to generate the unadjusted threshold manually:

Citation

Publications and research reports based on the Historical SPM data generated by by CPSP and accessed via IPUMS CPS should use the following citation:

Wimer, Christopher, Liana Fox, Sophie Collyer, Irwin Garfinkel, Neeraj Kaushal, Jennifer Laird, Jaehyun Nam, Laura Nolan, Jessica Pac, Ryan Vinh, and Jane Waldfogel. 2024. Historical Supplemental Poverty Measure Data. New York: Center on Poverty and Social Policy, Columbia University.

Bibliography

Fox, Liana. 2017. Anchored and Relative: Supplemental Thresholds for the Supplemental Poverty Measure (SEHSD Working Paper # FY2017-120). U.S. Census Bureau. https://www.census.gov/content/dam/Census/newsroom/press-kits/2017/appam/Fox-APPAM-Anchored%20and%20Relative.pdf

Fox, Liana, Christopher Wimer, Irwin Garfinkel, Neeraj Kaushal, and Jane Waldfogel. 2015. "Waging War on Poverty: Poverty Trends Using a Historical Supplemental Poverty Measure." Journal of Policy Analysis and Management 34 (3): 567-92. doi:10.1002/pam.21833.

Nolan, Laura, et al. "A new method for measuring historical poverty trends: Incorporating geographic differences in the cost of living using the Supplemental Poverty Measure." Journal of Economic and Social Measurement 41.3 (2016): 237-264. doi:10.3233/JEM-160433.

Wimer, Christopher, Liana Fox, Irwin Garfinkel, Neeraj Kaushal, and Jane Waldfogel. 2016. "Progress on Poverty? New Estimates of Historical Trends Using an Anchored Supplemental Poverty Measure." Demography 53 (4). Demography: 1207-18. doi:10.1007/s13524-016-0485-7.

Further Reading

An Introduction to Poverty Measurement: a history of the official and supplemental poverty measures

How the Census Bureau Measures Poverty: information on the Official Poverty Measure

Research Poverty Thresholds: overview of the Bureau of Labor Statistics' methodology for calculating the SPM

Supplemental Poverty Measure: Technical Documentation (PDF)

User Guidance

R-CPI-U-RS

Two different data sources are used to create historical Consumer Price Indices to adjust the anchored variables for inflation. From 1967 to 1977 only, CPSP uses the R-CPI-U-RS produced by the Census Bureau (also known as the CPI-U-X1). From 1978 forward, CPSP uses the index produced by the Bureau of Labor Statistics. Users should be aware that the BLS periodically revises this index based on methodological changes in calculating the R-CPI-U-RS. The anchored variables for the 1967-2009 ASEC samples are provided by CPSP and reflect the R-CPI-U-RS as of December 15, 2023; these values are static and will not change in response to future BLS modifications. Anchored variables from 2010-onward are calculated directly by IPUMS using the annual mean for the reference year; these will be updated to reflect the most recent BLS index. The SPMTHRESHANC12 variable documentation includes details about the R-CPI-U-RS version used to create the anchored variables currently available via IPUMS CPS.


Unmatched and Duplicate Records

SPM variables are not available for persons identified in the CPS as residents of group quarters. In a subset of years (1968-1975, 1979-1987), there are a small number of duplicate records as well as records that cannot be matched between the IPUMS data and the CPSP data. The two tables below detail documented issues in the IPUMS data files.

Table 2. Unmatched records and data issues
Year Identifiers Data Issues
1968 SERIAL 12987 / LINENO 3 Age mismatch between CPSP and IPUMS; no SPM data
1969 SERIAL 45877 / LINENO 1 Age mismatch between CPSP and IPUMS; no SPM data
1971 SERIAL 10010 SPM values are not consistent across household members
1972 Sample size mismatch There are 333 duplicate records for 136 people in the CPSP data that do not exist in the IPUMS data, resulting in different sample sizes for CPSP and IPUMS. Users should be aware that while the duplicates do not exist in the IPUMS data, the sampling weights still reflect the CPSP sample size. We are working with CPSP to determine the source of the discrepancy.
1972 SERIAL 25427 / LINENO 2 Identified as group quarters resident in IPUMS data (NIU) but household resident in CPSP data; SPM data available
1972 SERIAL 35696 / LINENO 2 Sex mismatch between CPSP and IPUMS; duplicate record
1972 SERIAL 37635 / LINENO 1 & 2, SERIAL 45135 / LINENO 1 These records are in the IPUMS data but not in CPSP data; no SPM data
Table 3. Duplicate records in the IPUMS data
Year Number of duplicate records
1968 10
1969 6
1970 2
1971 10
1972 2
1973 18
1974 12
1975 14
1979 24
1980 12
1981 22
1982 14
1983 36
1984 38
1985 60
1986 30
1987 42

Back to Top