Bureau of Labor Statistics > Consumer Expenditure Survey > Home

Summary of Methodology Changes

By Thesia I. Garner, Robert S. Martin, Brett Matsumoto, Scott Curtin [1]

The December 2024 release of the distribution of U.S. personal consumption expenditures using Consumer Expenditure (CE) Survey data incorporates several methodological updates to the methods used for the July 2024 and earlier releases which are outlined here. For more backgrounds on the methods generally, see the methodology summary.

Allocations and mapping

Health insurance: In prior versions, we allocated out-of-pocket and imputed premiums across up to four spending categories: medical equipment, nondurable goods, services, and net insurance. We now allocate across up to twelve categories: medical equipment, prescription drugs, other medical products, physician services, dental services, home health care, medical laboratories, specialty outpatient care, all other professional medical services, hospitals, nursing homes, and net health insurance. We combined data from the National Health Expenditure Accounts and the PCE bridge tables[2] to determine the allocation factors.
Food and beverages purchased for off-premises consumption: We still use the CE Interview variables that refer to global expenditures for food and nonalcoholic beverages and another for alcohol, but we allocate the consumer unit’s totals across 21 detailed PCE categories using their matched Diary observation’s spending shares.
Motor vehicle insurance: We now allocate motor vehicle insurance across four PCE categories: motor vehicle accessories and parts, motor vehicle repair services, hospital services, and net motor vehicle insurance, using data from the National Association of Insurance Commissioners (NAIC, 2024) and these categories relative PCE shares. With the addition of this allocation step, we removed the separate imputation for motor vehicle repair services, which similarly used insurance premiums.
Household insurance: the PCE for net household insurance includes the portion of homeowners insurance that is comparable a renters or condo owner policy. We include CE homeowner insurance premiums multiplied by an adjustment factor, similar to its treatment in the U.S. CPI.
Implicit rent for owner-occupied housing: Where available (2009 and forward) we adjust rental equivalence values for partial ownership and vacation rental usage outside the CU, using the same adjustment as in the CPI. We make a further downward adjustment for partial business use of the home.

Imputations for items not covered in the CE

Food produced and consumed on farms: we allocate PCE to households who report living on farms (whether or not they live inside a Core-based Statistical Area). For this set of households, we distribute the PCE amount proportionally to family size.
Net expenditures abroad: for goods and services purchased while abroad, we use passenger fares for foreign travel as an indicator. To represent deductions for nonresident spending (not included in the CE sample), we make neutral allocations by distributing the amounts proportionally to total PCE. This is how we also distribute spending by non-profit institutions serving households (NPISH).
Social Services and Religious Activities (other than childcare): These include social assistance, social advocacy and civic and social organizations, religious organizations’ services to households, and foundation and grantmaking and giving services to households which do not have close counterparts in the CE. We now make neutral allocations similar to the handling of NPISH and nonresident spending. Previously, some cash contributions were mapped to these categories, but these likely include transfers in addition to purchased services.
Financial services: We add imputations for commissions on securities purchases and sales, portfolio and investment advice services, and trust, fiduciary, and custody activities. Similar to McCully (2014), we treat this similarly to financial services furnished without payment and distribute PCE proportionally to the market value of securities held by the CU. The CE only collects this financial information in the fourth interview, which is missing for some consumer units due to attrition. To impute values for these missing units, we use a statistical matching procedure based on other observations with similar incomes, age, and race.

Imputing missing income

The matching procedure used to combine CE Interview and Diary data uses income before taxes to partition the samples and find similar observations. The CE only began imputing missing income in 2004. For earlier years, we adapt the multiple imputation procedure of Fisher, et al. (2014). For overlapping years, the authors find that their method replicates the CE’s distribution of income well. We apply it to the internal CE data (all Interview and Diary observations) and modify it so that consumer units who report bracketed ranges have imputed values which fall within these ranges.

Top-tail adjustment

As before, we adjust for potential underrepresentation and underreporting in the top tail of the distribution. We apply an adjustment based on draws from the Pareto distribution to the expenditures of consumer units in the top 5% of the expenditure distribution after imputations, but (prior to scaling to macro totals). We use a shape parameter of 2 and a scale parameter equal to the 95^th percentile of the expenditure distribution. The implied ratio between the Pareto draw and the consumer unit’s unadjusted expenditure total is used to adjust their expenditure for each category. In our revised methodology, we apply the random Pareto draws to consumer units in the top 5% in rank order ensuring that the ordering of households is preserved. We also base the adjustment on 1000 Pareto draws to reduce variance, taking the average by rank-order across draws. These revisions better align the procedure with Zwijnenburg et al., 2022. Finally, we do not adjust imputed categories which already sum to PCE totals by construction, namely imputed financial services.

Scaling to PCE totals

Previously, we scaled consumer unit PCE estimates to match totals by “major product” from NIPA Table 2.3.5. We now scale estimates to match totals for 150 detailed categories from NIPA Table 2.4.5U. At its most disaggregated, Table 2.4.5U contains 244 product categories, but we are not always able to match at the most detailed level. Some detailed categories have no counterpart in the CE (e.g., 22 categories for non-profit institutions serving households), and some are either more disaggregated in PCE (e.g., foreign vs domestic autos) or have spotty coverage in the CE due to sample size (e.g., pleasure aircraft).

Impact on results

Table 1 below shows how the shares of PCE by decile of equivalized PCE differ for 2022 between the July and the December release. The distribution is qualitatively similar, with the shares changing by less than one percentage point. The December estimated distribution is slightly more equal, with an equivalized Gini of 0.314 versus 0.325 in the July release.

Table 1: Share of PCE by Decile of Equivalized PCE in 2022 by Method Version
	0-10%	10-20%	20-30%	30-40%	40-50%	50-60%	60-70%	70-80%	80-90%	90-100%
December	3.8%	5.2%	6.2%	7.3%	7.9%	8.6%	9.7%	11.4%	13.5%	26.5%
July	3.6%	5.3%	6.1%	7.0%	7.7%	8.7%	9.6%	11.3%	13.4%	27.3%

References

Fisher, Jonathan D., David S. Johnson, and Timothy M. Smeeding. 2014. "Imputing income in the Consumer Expenditure Interview Survey." Monthly Labor Review. https://www.bls.gov/opub/mlr/2014/article/imputing-income-in-the-consumer-expenditure-interview-survey.htm.

McCully, Clinton P. 2014. "Integration of Micro and Macro Data on Consumer Income and Expenditures." In Measuring Economic Stability and Progress, by Dale W. Jorgenson, J. Steven Landefeld and Paul Schreyer (Eds.). University of Chicago Press.

Zwijnenburg, Jorrit, Joseph Grilli, and Pao Engelbrecht. 2022. "Pareto Tail Estimation in the Presence of Missing Rich in Compiling Distributional National Accounts." Paper prepared for the 37th IARIW General Conference, August 22-26, 2022. https://iariw.org/wp-content/uploads/2022/07/Jorret-Joseph-Pao-IARIW-2022.pdf.

[1] Division of Price and Index Number Research (Garner, Martin, Matsumoto), Division of Consumer Expenditure Surveys (Curtin), Bureau of Labor Statistics, 2 Massachusetts Ave., NE, Washington, DC 20212, USA. Emails: Garner.Thesia@bls.gov, Martin.Robert@bls.gov, Matsumoto.Brett@bls.gov, Curtin.Scott@bls.gov.

[2] See https://www.bea.gov/industry/industry-underlying-estimates.

Last Modified Date: December 9, 2024