Presentation by Bilal Habib, an analyst in CBO’s Tax Analysis Division, at the Bureau of Labor Statistics' 2021 Consumption Symposium.

1. Presentation at the Bureau of Labor Statistics’
2021 Consumption Symposium
September 23, 2021
Bilal Habib
Tax Analysis Division
How CBO Adjusts for Underreporting of
Means-Tested Transfers in
Its Distributional Analyses
For more information about the conference, see www.bls.gov/cex/2021-consumption-symposium-agenda.htm.

2. 1
Background

3. 2
The Congressional Budget Office regularly produces a report on the distribution
of household income, means-tested transfers, and federal taxes.
For each report, the agency draws data on transfers from the Annual Social and
Economic Supplement of the Current Population Survey (CPS).
To analyze the distribution of means-tested transfers, CBO must have a
complete accounting of those transfers, with enough precision for quintile-level
distributional analysis.
Why Does CBO Adjust for Underreporting?

4. 3
CBO adjusts administrative totals to make them compatible with the way data are reported in the CPS. For details, see Bilal Habib, How CBO Adjusts for Survey Underreporting of
Transfer Income in Its Distributional Analyses, Working Paper 2018-07 (Congressional Budget Office, July 2018), www.cbo.gov/publication/54234. CPS = Current Population Survey;
SNAP = Supplemental Nutrition Assistance Program; SSI = Supplemental Security Income.
Underreporting of Means-Tested Transfers in the CPS
Underreporting of
transfers in the CPS has
increased over time.
Without adjustment,
CPS-based analyses
would understate
income growth at the
bottom of the distribution
and the role of transfers
in reducing income
inequality.
0
20
40
60
80
100
1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012 2015 2018
Percent
Housing Assistance
Medicaid
SSI
SNAP
Overreporting
Underreporting

5. 4
CBO adjusts administrative totals to make them compatible with the way data are reported in the Current Population Survey. For details, see Bilal Habib, How CBO Adjusts for Survey
Underreporting of Transfer Income in Its Distributional Analyses, Working Paper 2018-07 (Congressional Budget Office, July 2018), www.cbo.gov/publication/54234.
Size of Major Means-Tested Transfer Programs
Program
Total Benefits in 2018
(Billions of dollars)
Medicaid and the Children’s Health Insurance
Program (CHIP)
528
Supplemental Nutrition Assistance Program (SNAP) 58
Supplemental Security Income (SSI) 55
Housing Assistance (Includes rental assistance and
public housing)
41

6. 5
▪ Administrative matching: Offers near-perfect accounting, but administrative
microdata are not widely available.
▪ Rules-based simulation: Offers precise estimates at the micro level but
requires a significant research investment.
▪ Regression-based estimation (CBO’s approach): Offers tractability for
multiple programs over a long time series but is less precise at the micro level.
Common Approaches to Adjusting for Underreporting

7. 6
CBO’s Approach

8. 7
▪ Use administrative data to set targets that reflect the CPS sampling frame.
– Subtract program participants who are institutionalized.
– Convert average monthly or point-in-time receipt to “ever-on” receipt.
– Convert fiscal year spending levels to calendar year targets.
▪ Use CPS data on reporting units to estimate the probability of receipt for all units.
– Run probit regressions for each year and population subgroup.
– Independent variables include income level and composition, labor force
participation, demographics, and household characteristics.
▪ Impute transfer receipt on the basis of estimated probabilities until the target is
reached.
– An algorithm assigns receipt to units in proportion to their predicted
probability of receipt.
CBO’s Regression-Based Imputation Method

9. 8
CPS = Current Population Survey; FPL = federal poverty level; SNAP = Supplemental Nutrition Assistance Program.
Share of Households Receiving SNAP Benefits, by Income, 2018
Most imputed recipients
of SNAP benefits are at
lower FPL ratios.
0
10
20
30
40
50
60
70
80
0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425 450
Household Income as a Percentage of the Federal Poverty Level
CBO (Imputed Plus Reported)
CPS (Reported Only)
Percent

10. 9
After the total number of recipients matches the administrative targets for all subgroups
and programs, each unit is assigned a benefit amount.
▪ Supplemental Nutrition Assistance Program (SNAP) and Supplemental Security
Income (SSI): CBO derives the average benefit per recipient on the basis of income
and demographic groups and assigns those averages to new recipients.
▪ Medicaid: CBO derives the average cost to the government per recipient from
administrative data by eligibility category and assigns those averages to all recipients.
▪ Housing assistance: CBO estimates benefits for each reporting unit on the basis of
location, household size and structure, and fair market rents determined by the
Department of Housing and Urban Development. No additional households are imputed
to be recipients.
Assigning Benefit Amounts

11. 10
For more details, see Sheila Zedlewski and Linda Giannarelli, TRIM: A Tool for Social Policy Analysis (Urban Institute, May 2015), https://tinyurl.com/TRIMurban2015 (PDF, 729 KB);
and Michael Davern and others, “A Partially Corrected Estimate of Medicaid Enrollment and Uninsurance: Results From an Imputational Model Developed off Linked Survey and
Administrative Data,” Journal of Economic and Social Measurement, vol. 34, no. 4 (December 2009), pp. 219–240, http://doi.org/10.3233/JEM-2009-0324.
▪ The method is straightforward to implement. For example, it doesn’t require a
team of researchers to code specific program rules as they evolve.
▪ It is easily scalable across multiple programs over a 40-year time series, and it
can be extended to other transfer programs, such as the National School Lunch
Program.
▪ The distributional results of CBO’s method are broadly similar at the quintile level
to those of the Urban Institute’s Transfer Income Model (which uses a rules-based
method) and of Davern and others (which relies on administrative matching).
Strengths of CBO’s Adjustment Method

12. 11
▪ As a regression-based method, CBO’s approach assumes that nonreporters
have the same characteristics as reporters.
▪ It does not account for false positives.
▪ Its ability to simulate complex policy scenarios is limited. However, it can be
augmented by inputs from other CBO models to estimate the effects of some
proposed rule changes to transfer programs.
Limitations of CBO’s Adjustment Method

13. 12
Effects on Aggregate
Distributional Outcomes

14. 13
All dollar amounts are in 2018 dollars. To calculate growth rates, CBO first converted all dollar amounts to 2018 dollars using the Bureau of Economic Analysis’s price index for
personal consumption expenditures. Shaded vertical bars indicate the duration of recessions. * = between zero and $500.
Average Income, Means-Tested Transfers, and Federal Taxes,
2018, and Cumulative Growth in Average Income, 1979 to 2018
Means-tested transfers
and federal taxes
significantly increase the
cumulative growth of
income after transfers and
taxes among households
in the lowest quintile.

15. 14
These results differ from those published in CBO’s Distribution of Household Income reports because they use data from the Current Population Survey (CPS) only, whereas those
reports use a combination of CPS data and data from the Internal Revenue Service’s Statistics of Income.
Income Inequality, 1979 to 2018
Adding imputed transfers
reduces income inequality
in CBO’s distributional
analyses.
The effects on inequality
of imputing transfers have
increased over time as
underreporting has
increased.
0.36
0.38
0.4
0.42
0.44
0.46
0.48
1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012 2015 2018
Gini Index
Income Before Transfers
and Taxes (Base income)
Base Income Plus Imputed
Means-Tested Transfers
Base Income Plus Reported
Means-Tested Transfers

16. 15
CBO released a working paper in 2018 describing its method for imputing means-tested
transfers. It included CPS extracts and Stata code to allow researchers to replicate CBO’s
results for 1979 through 2016.
In May 2020, CBO released a new version of the data and code with the following updates:
▪ It added imputations for 2017 and 2018,
▪ It added imputed values for housing assistance benefits,
▪ It improved estimates of Medicaid and CHIP benefits by accounting for variations in the
amount of time enrolled in the program,
▪ It updated recipient and benefit target values for 1979 through 2016, and
▪ It used GitHub as the repository.
In the coming months, CBO plans to release an update to that GitHub repository, including
imputations for 2019.
Public Release of CBO’s Means-Tested Transfer Imputations

17. 16
Congressional Budget Office, The Distribution of Household Income, 2018
(August 2021), www.cbo.gov/publication/57061.
Congressional Budget Office, “Means-Tested Transfer Imputations” (last updated
May 2020), https://github.com/US-CBO/means_tested_transfer_imputations.
Congressional Budget Office, Projected Changes in the Distribution of Household
Income, 2016 to 2021 (December 2019), www.cbo.gov/publication/55941.
Bilal Habib, How CBO Adjusts for Survey Underreporting of Transfer Income in Its
Distributional Analyses, Working Paper 2018-07 (Congressional Budget Office,
July 2018), www.cbo.gov/publication/54234.
Kevin Perese, CBO’s New Framework for Analyzing the Effects of Means-Tested
Transfers and Federal Taxes on the Distribution of Income, Working Paper 2017-09
(Congressional Budget Office, December 2017), www.cbo.gov/publication/53345.
Related Publications