Imputing 2020 Economic Impact Payments in the 2021 CPS ASEC*

Imputing 2020 Economic Impact Payments in the 2021 CPS ASEC



Adam Bee, U.S. Census Bureau

Charles Hokayem, U.S. Census Bureau

Daniel Lin, U.S. Census Bureau

SEHSD Working Paper No. FY 2021-18

September 2021

Abstract

In response to the COVID-19 pandemic, Congress passed legislation to aid individuals and families. This

legislation included the Coronavirus Aid, Relief, and Economic Security Act (CARES Act) in March

2020 and the Coronavirus Response and Relief Supplemental Appropriations Act (CRRSA Act) in

December 2020. Both the CARES Act and the CRRSA Act provided households with additional

resources in the form of stimulus payments. This paper presents a model of the two rounds of stimulus

payments received by households in 2020 based on combining survey responses from the 2021 Current

Population Survey Annual Social and Economic Supplement (CPS ASEC) with the CPS ASEC Tax

Model. We present the Production Model used in the 2021 CPS ASEC and in the 2020 Supplemental

Poverty Measure (SPM). We present two additional models with different underlying assumptions that

affect take-up rates and provide upper and lower bounds on the aggregate stimulus amount and the

number of individuals receiving payments. We also provide aggregates comparable to tabulations

released by the Internal Revenue Service Statistics of Income program and estimate the SPM rate under

each model.



This paper is released to inform interested parties of ongoing research and to encourage discussion of work in

progress. All errors are those of the authors. Any views expressed, including those related to statistical,

methodological, technical, or operational issues, are solely those of the authors and do not necessarily reflect the

official positions or policies of the U.S. Census Bureau. The Census Bureau reviewed this product for unauthorized

disclosure of confidential information and has approved the disclosure avoidance practices applied to this release.

CBDRB-FY21-POP001-0183 and CBDRB-FY21-POP001-208.

Background

Economic Impact Payments

In March 2020 the Coronavirus Aid, Relief, and Economic Security Act, also known as the

CARES Act, created a refundable tax credit for the 2020 tax year. The advance payment of this tax credit

was referred to as the First Economic Impact Payment (EIP 1). In December 2020 the Coronavirus

Response and Relief Supplemental Appropriations Act of 2021 created an additional refundable tax credit

for the 2020 tax year. The advance payment of this credit was referred to as the Second Economic Impact

Payment (EIP 2). The table on the following page (Economic Impact Payment Comparison) summarizes

features of each payment including eligibility criteria and amount.

We describe these features below.

People were eligible for EIP 1 if they were either a U.S. citizen or U.S. resident alien, and they

were not claimed as a dependent of another taxpayer in 2018 or 2019, and they had a Social Security

number (SSN) valid for employment.

Generally, married joint filers with adjusted gross income (AGI)

under $150,000, head of household filers with AGI under $112,500, and individual filers using any other

filing status with AGI under $75,000 were eligible for their full EIP 1 and EIP 2 payments.

The

maximum EIP 1 amount was $1,200 for single, head of household, and married separate filers and $2,400

for married joint filers, plus $500 per qualifying child. The maximum EIP 2 amount was $600 for single,

head of household, and married separate filers and $1,200 for married joint filers, plus $600 per

qualifying child. A qualifying child is defined by the Internal Revenue Service (IRS) as someone who is

the child, stepchild, eligible foster child, sibling, half sibling, step sibling, or descendant, who can be

claimed as a dependent, who is under age 17 at the end of the reference tax year,

is a U.S. citizen, U.S.

national, or U.S. resident alien, and has a valid SSN or taxpayer identification number. Both EIP 1 and

EIP 2 had a phaseout provision where the payments were reduced by 5 percent of the amount by which

Information for this table was compiled from these IRS websites: <https://www.irs.gov/newsroom/questions-and-

answers-about-the-first-economic-impact-payment-topic-a-eligibility>,

<https://www.irs.gov/newsroom/first-economic-impact-payment-questions-and-answers-topic-h-social-security-

railroad-retirement-and-department-of-veteran-affairs-benefit-recipients>, and

<https://www.irs.gov/coronavirus/second-eip-faqs#Eligibility>.

The Social Security Administration issues three types of Social Security cards: (1) an unrestricted card for U.S.

citizens and people lawfully admitted to the U.S. on a permanent basis; (2) “VALID FOR WORK ONLY WITH

DHS AUTHORIZATION” for people lawfully admitted to the U.S. on a temporary basis with authorization to

work; and (3) “NOT VALID FOR EMPLOYMENT” for people who are lawfully admitted to U.S. without work

authorization or people who need a Social Security number for federal benefits or services.

In December 2020 the COVID-related Tax Relief Act of 2020 increased the AGI phaseout amount for a Qualifying

Widow(er) from $75,000 to $150,000.

Our interpretation of the limited available information regarding eligibility suggests that children’s eligibility was

determined by the tax year of the return used by IRS. For example, a filer with a child who was aged 16 years in the

2018 tax return, and did not file a 2019 tax return, still received the extra amount for that qualifying child. However,

we uniformly model children’s eligibility based on their ages as of 2020.

the filer’s AGI exceeded the applicable thresholds. Qualifying income for EIP 1 could be either 2018 or

2019 income, depending on what IRS records showed at the time of payment.

Eligibility for EIP 2 was broadly similar. EIP 2 expanded eligibility to additional groups; these

expansions were retroactively applied to EIP 1, but tax filers needed to apply for the additional amounts

in tax year 2020 returns. Qualifying income for EIP 2 was only based on 2019 income. Eligibility criteria

are summarized in the table below.

In March 2021 the American Rescue Plan created the Third Economic Impact Payment (EIP 3)

which is a refundable tax credit for the 2021 tax year. Since EIP 3 is a 2021 tax credit, and the advance

payments of EIP 3 were made in 2021, and the Recovery Rebate Credit for EIP 3 is also for tax year

2021, the third EIP is not included in the models presented in this paper, which cover payments for only

2020.

Economic Impact Pa

ment Comparison

Economic Impact Payment 1 (EIP 1) Economic Impact Payment 2 (EIP2)

Legislation/

Date

Coronavirus Aid, Relief, and Economic

Security Act (CARES Act)

March 2020

Coronavirus Response and Relief Supplemental

Appropriations Act (CRRSA Act)

December 2020

Maximum

payment

 $2,400 for married filing jointly

 $1,200 for single, head of household,

married filing separately, and

widowers

 $500 per qualifying child

 $1,200 for married filing jointly

 $600 for single, head of household,

married filing separately, and widowers

 $600 per qualifying child

Income

threshold

for full

payment

amount

 Married, filing jointly: $150,000

 Head of Household: $112,500

 Single or married filing separately:

$75,000

 Widowe

: $75,000

 Married, filing jointly: $150,000

 Head of Household: $112,500

 Single or married filing separately:

$75,000

 Widower: $150,000

Phaseout Payment reduced by 5% of AGI exceeding

threshold

Payment reduced by 5% of AGI exceeding

threshold

Eligibility Recipients must have a Social Security

number valid for employment. Qualifying

dependents are qualifying children under 17

at the end of 2019 tax year. Married joint

filers may be eligible for a partial credit when

only one spouse has a Social Security

number valid for employmen

Recipients must have a Social Security number

valid for employment. Qualifying dependents are

qualifying children under 17 at the end of 2019 tax

year. Married joint filers only need one spouse to

have SSN valid for employment.

Auto-

Payment

Recipients

 Tax year 2018 and 2019 filers.

 Social Security, Supplemental

Security, Veterans Affairs, and

Railroad Retirement recipients

 Tax year 2019 filers.

 Social Security, Supplemental Security,

Veterans Affairs, and Railroad Retirement

recipients

Source: U.S. Internal Revenue Service, <https://www.irs.gov/newsroom/questions-and-answers-about-the-first-

economic-impact-payment-topic-a-eligibility>,

<https://www.irs.gov/newsroom/first-economic-impact-payment-questions-and-answers-topic-h-social-security-

railroad-retirement-and-department-of-veteran-affairs-benefit-recipients>, and

<https://www.irs.gov/coronavirus/second-eip-faqs#Eligibility>.

Notes:

As updated in December 2020, expansion of the AGI threshold for widow/widower filers under EIP 2 was

retroactively applied to EIP 1. Affected filers needed to apply for the Recovery Rebate Credit to receive the full

amount.

As updated in December 2020, expansion of eligibility for some joint filers under EIP 2 was retroactively applied

to EIP 1. Affected filers needed to apply for Recovery Rebate Credit to receive the full amount.

2021 Current Population Survey Annual Social and Economic Supplement

To collect data on the Economic Impact Payments for 2020 the Census Bureau added new

questions about the receipt and amount of these payments to the 2021 Current Population Survey Annual

Social and Economic Supplement (CPS ASEC).

The questions asked, at the household level:

“Since April 1, 2020, have you or anyone in your household received a ‘stimulus

payment,’ that is the coronavirus (COVID-19) related Economic Impact Payment from

the Federal Government?”

“What was the amount of the stimulus payment(s) that you received for all the covered

adults and children since April 1, 2020?”

At the time of the questionnaire’s design, only the first payment had been announced. The

questions were intended to capture EIP 1 and a possible upcoming EIP 2.

However, because households

were only asked to report stimulus payments since April 1, 2020, with no end date, respondent

households may have interpreted the question in multiple ways.

Some households may have only reported EIP 1 and EIP 2, as intended. Some households may

have only reported stimulus received during the calendar year. Other households may report EIP 1 and

EIP 2, even if those were received during 2021. Due to the timing of the survey and the start of EIP 3

distribution, other households may have included EIP 3.

Specifically, the 2021 CPS ASEC was

conducted from February 2021 to April 2021. Part of this collection period, March 2021 and April 2021,

overlapped with the period that EIP 3 was beginning to be disbursed. To investigate the potential for

households including this third payment in their responses, we also model amounts for EIP 3. Among the

53.2 percent of households with positive modelled EIP 3 amounts and non-imputed survey responses,

22.0 percent gave amount responses exactly equal to their modelled sum of EIP 1 and EIP 2, while 13.0

percent provided amount responses equal to the predicted sum of all three payments.

Further, 33.0 percent of households either did not respond to the survey stimulus questions (or

had inconsistent responses) and thus required imputation. Due to these variations in the timing of receipt

For more information on confidentiality protection, methodology, sampling and nonsampling error, and

definitions, see <https://www2.census.gov/programs-surveys/cps/techdocs/cpsmar21.pdf>.

The conceptual scope of the Supplemental Poverty Measure (SPM) and the CPS ASEC tax model is to include tax

credits filed for the reference tax year (2020 in this case), which excludes EIP 3 because it is a credit for the 2021

tax year.

EIP 1 was distributed starting April 2020. EIP 2 was distributed starting late-December 2020 and early January

2021.

of EIP 2, potential confusion around EIP 3 timing, and high nonresponse rates to this question, we model

payments for EIP 1 and EIP 2 rather than solely rely on the CPS ASEC survey responses.

CPS ASEC Tax Model

Modeling each EIP amount requires three inputs for each tax unit in the household: tax filing

status, the number of qualifying children, and adjusted gross income. This section describes the process of

creating each input in the 2021 CPS ASEC tax model.

Tax filing status, number of dependents, and number of qualifying children: The CPS ASEC data

first must be transformed from person- and household-level units into tax units. The process for forming

tax units begins with the person file by creating three subsets: (1) one for all spouses; (2) one for all

children; and (3) one for the “others.” The first subset of all filers with spouses is created by looking for

any non-zero value for the spouse pointer variable. The second subset of all children is created by looking

for any non-zero value in the parent pointer variable. The remaining persons – with no values in spouse

pointer or parent pointer – form the third subset for the “others.”

We then adjust those initial sets as follows. We make sure that the children and dependents subset

only includes qualifying children or dependents as the IRS defines them

– children aged 18 years or

under, children under 24 and enrolled in school, or adult children with a disability. Children that don’t fall

into those three categories are moved to the “others” subset. We similarly reassign anyone under age 15 in

the “others” subset (with no parent pointer) to the children subset.

The tax model only constructs single

dependents and does not consider married dependent filers. The number of qualifying children and

qualifying dependents as defined for various federal and state deduction and credits, including economic

impact payments, are calculated for each tax unit.

Dependents are then linked to tax units. The child subset (2) is attached to the tax unit of their

parent or parents, using the parent pointer variable. If the tax unit contains two spouses, then the filing

status of the unit is Married, Filing Jointly, and the spouse incomes are combined for the tax unit. The

main filer is determined by the family relationship variable. The reference person is the main filer, and the

spouse is attached to the main filer. We assume that all married couples file jointly, so we do not model

the Married, Filing Separately filing status. If the tax unit contains a single person that has dependents,

then the filing status is set to Head of Household. If the tax unit contains only a single person, then the

Since all inputs to the tax model are based on responses to the household survey, only resident children and

dependents are included.

Children under 15 and with no parent pointer are assigned to be a dependent of the householder.

filing status is set to Single. The survey data don’t allow us to identify married separate filers, so the filing

status in the ASEC Tax Model only consists of Single, Head of Household, and Married, Filing Jointly.

Adjusted gross income: Federal adjusted gross income (AGI) is computed as the sum of wages

and salaries, interest and dividend incomes, alimony income, business self-employment income, capital

gains, IRA income, pension income, rent income, farm self-employment income, taxable unemployment

compensations, and taxable social security benefits and subtracted by one-half of self-employment taxes,

self-employed health insurance, self-employed health savings, and IRA contributions.

Several 2020 tax-year-specific tax code changes were applied to the calculation of AGI, such as

the Unemployment Compensation Exclusion and the suspension of the charitable contribution limit. The

Unemployment Compensation Exclusion, which excluded up to $10,200, is subtracted from AGI.

Changes not applied for tax year 2020 were the Old-Age, Survivors, and Disability Insurance (OASDI)

tax deferral by employers, OASDI tax deferral by self-employed, and the $300 charitable contribution

deduction under standard deduction.

Filing requirements: In addition, the stimulus payments depend on the filing of tax returns, which

is also determined by the tax model. The tax model assumes a tax unit files a return if it meets at least one

of the following requirements: (1) income above a filing threshold determined by age and filing status; (2)

positive Earned Income Tax Credit (EITC); (3) positive self-employment income; (4) gross income less

than $0; (5) self-employment income less than $0; (6) positive Additional Child Tax Credit; (7) positive

self-employment income for either spouse; or (8) has total income above $2,000. Tax units that do not

satisfy any of these eight requirements are presumed nonfilers.

Stimulus Model Approaches

We describe three models for assigning economic impact payments: the Production Model (our

preferred model), an Upper Bound Model, and a Lower Bound Model. These models differ in the

underlying assumptions that affect the take-up rate of payments. All models implement the same rules for

calculating the payments based on tax filing status, qualifying children, and adjusted gross income from

the ASEC Tax Model. For single filers, if their AGI is less than or equal to $75,000, then they receive the

full amounts of EIP 1 and EIP 2: $1,200 and $600, respectively. For married joint filers, if their AGI is

Some AGI and itemized deduction components are imputed from the IRS Statistics of Income microdata.

Since the OASDI taxes must be repaid, the deferrals are not modeled. The $300 charitable contribution deduction

for standard deduction introduced under the CARES Act is not applied to AGI, because in the ASEC Tax Model,

charitable contributions are imputed from IRS Statistics of Income microdata. However, there are only records of

itemizing filers with charitable contributions, so the data lack information on charitable contributions for standard

deduction filers.

less than or equal to $150,000, then they receive the full amounts of EIP 1 and EIP 2: $2,400 plus $500

per qualifying child and $1,200 plus $600 per qualifying child, respectively. For head of household filers,

if their AGI is less than or equal to $112,500, then they receive the full amounts of EIP 1 and EIP 2:

$1,200 plus $500 per qualifying child and $600 plus $600 per qualifying child, respectively. The

payments are reduced by 5 percent of the amount by which the filer’s AGI exceeds the applicable

threshold. For example, for filers with no qualifying children, EIP 1 is reduced to $0 if their AGI is at

least $99,000 for single filers, $136,500 for head of household filers, and $198,000 for married joint filers

and EIP 2 is reduced to $0 if their AGI is at least $87,000 for single filers, $124,500 for head of

household filers, and $174,000 for married joint filers.

Production Model (preferred model)

The Production Model produces EIP amounts in the CPS ASEC tax file that are also used in the

Supplemental Poverty Measure. This model assumes all tax units passing the standard filing requirements

according to the ASEC Tax Model (called presumed filers) receive both EIP amounts. Tax units that do

not file according to the ASEC Tax Model (called presumed nonfilers) do not receive any EIP amount

with four exceptions. Presumed nonfilers reporting (or are imputed to receive) any Social Security,

Supplemental Security Income (SSI), or Veterans Affairs (VA) benefits in the CPS ASEC are assigned

both EIP amounts. These groups represent actual nonfilers who were automatically issued payments by

the IRS. Additionally, we give EIP amounts to presumed nonfilers who are in households that report

receiving (or imputed receiving) any EIP amount.

These additional payments could be considered

roughly analogous to the nonfiling households without Social Security, SSI, or VA payments who had to

use the IRS online nonfiler tool in order to receive payment. Given the possible misinterpretation of the

questionnaire, we thus incorporate the reported receipt while discarding the reported amount. This model

assumes individuals receiving less payments during 2020 than they should have will reclaim the

difference in the 2020 Recovery Rebate Credit. Theoretically, taken together these requirements yield an

EIP take-up rate of less than 100 percent, because there may be some presumed nonfilers who don’t

satisfy the above requirements but are still eligible.

Upper Bound Model

The Upper Bound Model assigns both EIP amounts to all presumed nonfilers, as well as filers.

Recipients still must qualify with respect to modelled AGI, filing status, and number of qualifying

children, as determined by the ASEC Tax Model, however. This optimistic model assumes a 100-percent

take-up rate, meaning all individuals eligible for both payments receive them. It also assumes individuals

The stimulus response variable used is an edited survey variable which includes imputed values and is available

on the internal ASEC file. The modeled stimulus variable (EIP_CRD) is available on the public-use file.

not receiving their full eligible payments during 2020 will claim the difference in the 2020 Recovery

Rebate Credit.

Lower Bound Model

The Lower Bound Model addresses a limitation of the ASEC Tax Model which potentially

overestimates the number of single filers due to the nature of the ASEC survey and limitations on the

construction of tax units. Members of the same tax filing unit in official IRS tax returns may be split up in

the tax model if they are not directly related or if they are not living in the same household. When they are

split up, the individuals who may have been dependents in a tax unit become separate individual single

filers. This process increases the number of single filers, and consequently, the aggregate EIP amount

because a qualifying child adds a lower payment amount than a single filer, and nonqualifying dependents

do not receive any EIPs. A misclassification of a single filer as a dependent will lead to a combined EIP

amount of either $0 or up to $1,100 if the dependent is a qualifying child; whereas a single filer will

receive a combined EIP amount up to $1,800.

To address this potential overestimate of aggregate EIP, the Lower Bound Model starts with the

Production Model and targets the aggregate number of First Economic Impact Payments made to single

filers, as tabulated by IRS.

Since the AGI thresholds for maximum EIP 1 and EIP 2 are the same, all EIP

2 recipients are theoretically also EIP 1 recipients. But the number of EIP 2 recipients is lower in the

aggregate since EIP 2 phases out more quickly with respect to AGI than does EIP 1. Therefore, starting

with the Production Model, we randomly reassign $0 to EIP 1 and EIP 2 to filers with a positive EIP 1

until the estimated number of EIP 1 payments matches the IRS aggregate. Like the Production Model,

this process yields an EIP take-up rate that is theoretically less than 100 percent. Furthermore, it should

yield the lowest take-up rate of the three models. Aside from the people randomly assigned $0, this model

also assumes individuals not receiving their full eligible payments during 2020 will claim the difference

in the 2020 Recovery Rebate Credit.

Aggregate EIP statistics can be found at <https://www.irs.gov/statistics/soi-tax-stats-coronavirus-aid-relief-and-

economic-security-act-cares-act-statistics>. At the time of writing, this site tabulates the First EIP as of January

2021 and the Second EIP as of February 2021.

Modeling Results

This section presents the results of the three modeling approaches along with the Supplemental

Poverty rate for each model. We provide comparisons by adjusted gross income and tax filing status. In

addition, we provide robustness checks by checking the use of adjusted gross income from the ASEC Tax

Model against tax records and the sensitivity of tax year income used for calculating payments.

Table 1 displays the ratio of each model’s output relative to IRS benchmark estimates released by

the IRS Statistics of Income (SOI) program.

For each model, the table shows the ratio for Number of

Payments (Count), Aggregate Payments (Total Amount), and Average Payment (Mean Amount) overall

and by aggregate gross income categories. The first set of estimates come from the Production Model (our

preferred estimates) while the second and third set of estimates come from the Upper Bound and Lower

Bound Models, respectively. Finally, the last set of estimates in the table are from the ASEC survey.

Since the survey question is asked at the household level, some choices must be made to allow

comparability. The table uses the AGI of the survey householder and combines payments of all tax units

in the household. It is possible for other filers within the household to receive their own payments, even

when the survey householder does not qualify. The underlying estimates for each ratio and the associated

standard errors are available in Appendix Table 1. The table also reports the take-up rate for each model.

Table 2 follows the same structure as Table 1 but shows the ratio by tax filing status: (1) Single or

qualifying widow(er); (2) Married filing joint or separately; and (3) Head of household. The underlying

estimates for each ratio and the associated standard errors are available in Appendix Table 2.

Tables 3 and 4 show simple robustness checks related to using adjusted gross income from the

ASEC Tax Model (Table 3) and actual IRS 1040 AGI across consecutive years used for calculating

payments (2020 vs. 2019 and 2019 vs. 2018 in Table 4).

Table 3 reports a simple regression of actual

adjusted gross income from the linked Form 1040 records on modeled adjusted gross income from the

ASEC Tax Model for primary filers who can be linked to their tax record. This table illustrates the fit of

this simple AGI model. Table 4 shows a cross tabulation of differences in adjusted gross income across

consecutive years and the consequent differences in economic impact payments. Since income from 2019

or 2018 can be used for determining EIP 1, and the ASEC Tax Model uses 2020 income, this table shows

IRS SOI statistics can be found at <https://www.irs.gov/statistics/soi-tax-stats-coronavirus-aid-relief-and-

economic-security-act-cares-act-statistics>. This site reports the First EIP as of January 2021 and Second EIP as of

February 2021.

These actual IRS 1040 AGI values come from individual linked IRS Form 1040 records provided to the Census

Bureau under Title 26. This federal tax information is approved for limited statistical evaluations and thus is

unavailable for survey production. For additional documentation of the linked IRS Form 1040 records, see

<https://www.census.gov/content/dam/Census/library/working-papers/2019/demo/sehsd-wp2019-36.pdf>.

the sensitivity of the payment amount to the year of income used. Panel A shows the sensitivity of using

2020 vs. 2019 income while Panel B shows the sensitivity of using 2019 or 2018 income. Each panel

shows what happens if the income is greater than, equal to, or less than the prior year and the

corresponding effect on the payment amount (equal, greater, or less). Panel A shows that 90.2 percent of

filers have EIP amounts that do not change using 2020 vs. 2019 income.

Finally, Table 5 shows the Supplemental Poverty Measure (SPM) rate for each model. The table

shows the SPM rate for all people along with the rate by select characteristics (age, gender, type of unit,

race, and education). The table also shows the difference in the SPM rate between the Production Model

and the Upper and Lower Bound Models.

Discussion of Differences Between CPS ASEC Models and IRS Benchmarks

While the modeling results presented in this paper are shown relative to IRS SOI aggregates,

there are reasons for differences between the two sources which include universe differences and

assumptions underlying the ASEC Tax Model. The sample universe for the CPS consists of civilian,

noninstitutionalized population of the United States. Individuals in group quarters such as prisons, long-

term care hospitals, and nursing homes, are not eligible to be interviewed in the CPS. The CPS only

collects data in the 50 states and the District of Columbia. It includes military personnel who live either

off post or who live on post with civilian family members. It does not include military personnel living on

a base without civilian family members. The sample universe for the SOI aggregates differs from the CPS

universe. It is larger than the CPS, including the institutionalized population, those living in group

quarters, the military (both living with civilians and on base without civilians), and individuals living

abroad.

The EIP amounts produced by the models follow several assumptions. CPS ASEC lacks

information on SSN, dependents who are not resident in the household or don’t have a parent pointer in

the survey, and filing status for married separate filers, all of which are known by IRS. The model also

does not incorporate 2018 or 2019 income, which is what the actual EIP 1 and EIP 2 were mostly based

on. Due to these data constraints, EIP 1 and EIP 2 in the models use 2020 income and assume that filers

receive the maximum amount of EIPs they qualify for. This definition is closer to EIP 1 and EIP 2 plus

the 2020 Recovery Rebate Credit, which addresses the differences in what tax filers should have gotten.

The most recent SOI statistics only cover payments made through January 2021 for the First EIP and

February 2021 for the Second EIP, so these statistics likely do not include the 2020 Recovery Rebate

The 2020 Recovery Rebate Credit allows an individual to file for an additional payment during the 2021 tax filing

season to make up for the difference between what the individual should have received and actually received in

2020.

Credit, while the models we present implicitly include the credit. The SOI statistics will likely increase

when they are updated to reflect the credit. The ASEC Tax Model only models EIPs for single, head of

household, and married joint filers within a household.

Another source of difference between the models in this paper and the SOI statistics is the

definition of adjusted gross income regarding the treatment of unemployment compensation. Under the

American Rescue Plan enacted March 11, 2021, individuals with adjusted gross income of less than

$150,000 can exclude up to $10,200 of unemployment compensation paid in 2020 from adjusted gross

income when filing taxes. The models in this paper incorporate this new exclusion when calculating

adjusted gross income. The SOI statistics do not incorporate this exclusion because they are based on

2018 and 2019 tax returns.

Conclusion

This paper presents three models of the Economic Impact Payments (EIP), or stimulus payments,

authorized by the CARES Act in March 2020 and the CRSSA Act in December 2020. Our preferred

model, the Production Model, is based on combining survey responses from the 2021 CPS ASEC with the

CPS ASEC Tax Model. The EIP estimates from this model are incorporated in the 2020 Supplemental

Poverty Measure. The two additional models vary underlying assumptions that impact the EIP take-up

rate and thus give upper and lower bounds on the aggregate stimulus amount and the number of

individuals receiving payments. Data users interested in the modeled sum of both payments from the

Production Model should use the EIP_CRD variable available on the 2021 CPS ASEC public-use file.

These modeled stimulus payments are reflected in many Census Bureau reports characterizing income

and poverty during 2020, including Income and Poverty in the United States: 2020, The Supplemental

Poverty Measure: 2020, and a recent America Counts story, “The Impact of Stimulus Payments on

Supplemental Poverty.”

The 2021 CPS ASEC public-use file can be found at <https://www.census.gov/data/datasets/time-

series/demo/cps/cps-asec.html>.

Table 1. ASEC/SOI ratios of first- and second-round aggregate payments by categories of adjusted gross income.

Count Amount

Mean

Amount Count Amount

Mean

Amount Count Amount

Mean

Amount Count Amount

Mean

Amount

Overall 0.97 1.02 1.05 0.99 1.04 1.05 0.86 0.95 1.10 0.65 0.77 1.18

Take-Up Rate (%) 98.0 100.0 86.6 66.0

Negative or zero 0.93 1.09 0.86 1.06 1.25 0.86 0.77 0.94 0.89 0.48 0.62 0.96

$1 under $10,000 1.06 1.21 1.14 1.06 1.21 1.14 0.91 1.07 1.18 0.64 0.75 1.17

$10,000 under $20,000 0.74 0.76 1.02 0.74 0.76 1.02 0.64 0.68 1.06 0.39 0.45 1.16

$20,000 under $30,000 0.86 0.87 1.00 0.86 0.87 1.00 0.74 0.78 1.05 0.44 0.54 1.23

$30,000 under $40,000 0.92 0.94 1.02 0.92 0.94 1.02 0.80 0.85 1.07 0.52 0.61 1.17

$40,000 under $50,000 0.94 0.97 1.03 0.94 0.97 1.03 0.82 0.88 1.08 0.60 0.69 1.15

$50,000 under $60,000 1.07 1.07 1.00 1.07 1.07 1.00 0.93 0.97 1.05 0.69 0.78 1.13

$60,000 under $75,000 1.08 1.10 1.03 1.08 1.10 1.03 0.97 1.04 1.07 0.77 0.84 1.09

$75,000 under $100,000 1.07 1.04 0.97 1.07 1.04 0.97 0.98 1.01 1.03 0.81 0.90 1.10

$100,000 under $200,000 1.15 1.16 1.01 1.15 1.16 1.01 1.15 1.16 1.01 1.09 1.07 0.98

$200,000 or more 1.50 1.52 1.01 1.50 1.52 1.01 1.50 1.52 1.01 13.00 56.77 4.37

Lower Bound Model Survey Reported

Source: Internal Revenue Service Statistics of Income (IRS SOI) tables (irs.gov/statistics/soi-tax-stats-coronavirus-aid-relief-and-economic-security-act-cares-act-statistics),

U.S. Census Bureau 2021 Current Population Survey Annual Social and Economic Supplement (CPS ASEC).

Notes: Each cell represents the ratio of the CPS ASEC tax-modeled sum of the first and second rounds of Economic Impact Payments to its corresponding IRS SOI benchmark.

Underlying levels are listed in Appendix Table 1. SOI counts are the maximum of first- and second-round EIP aggregate counts, as the union of recipients is not separately

tabulated. SOI totals include payments not based on tax returns, which is more comparable to the totals from the ASEC Tax Model. The “negative or zero” AGI category for

SOI was modified to include payments to individuals without tax returns. Ratios may differ slightly from those implied by Appendix Table 1 due to use of unrounded SOI

values.

Production Model Upper Bound Model

Table 2. ASEC/SOI ratios of first- and second-round aggregate payments by filing status.

Count Amount

Mean

Amount

Count Amount

Mean

Amount

Count Amount

Mean

Amount

Count Amount

Mean

Amount

Overall

0.97 1.02 1.05 0.99 1.04 1.05 0.86 0.95 1.10 0.65 0.77 1.18

Single or qualifying widow(er)

1.26 1.26 0.99 1.26 1.26 0.99 1.00 0.99 0.99 0.66 0.79 1.21

Married filing joint or separately

1.14 1.16 1.02 1.14 1.16 1.02 1.14 1.16 1.02 1.04 0.97 0.94

Head of household 0.72 0.81 1.12 0.72 0.81 1.12 0.72 0.81 1.12 0.50 0.59 1.18

Production Model Upper Bound Model Lower Bound Model Survey Reported

Source: Internal Revenue Service Statistics of Income (IRS SOI) tables (irs.gov/statistics/soi-tax-stats-coronavirus-aid-relief-and-economic-security-act-cares-act-statistics), U.S.

Census Bureau 2021 Current Population Survey Annual Social and Economic Supplement (CPS ASEC).

Notes: Each cell represents the ratio of the CPS ASEC tax-modeled sum of the first and second rounds of Economic Impact Payments to its corresponding IRS SOI benchmark.

Underlying levels are listed in Appendix Table 2. SOI counts are the maximum of first- and second-round EIP aggregate counts, as the union of recipients is not separately

tabulated. Ratios may differ slightly from those implied by Appendix Table 2 due to use of unrounded SOI values.

The overall values for the Upper Bound Model include payments received by nonfilers.

The CPS ASEC Tax Model assigns single filer status such that it is most comparable to the sum of SOI single filers and SOI qualifying widow(er) filers.

The CPS ASEC Tax Model assigns married filing jointly such that counts are most comparable to the sum of SOI married joint and one-half of SOI married separate filers, and

amounts are most comparable to the sum of SOI married joint and separate filers.

(1) (2) (3) (4) (5) (6) (7) (8)

ty2020 ty2019 ty2020 ty2019 ty2020 ty2019 ty2020 ty2019

Constant 48,890*** 49,430*** 16,260*** 16,910*** 17,480*** 17,830*** 19,490*** 21,640***

(4,048) (2,362) (269) (272) (265) (262) (532) (507)

AGI/10k 5,128*** 5,234*** 6,388*** 5,754*** 6,447*** 5,968*** 2,624*** 889***

(383) (374) (54) (54) (51) (51) (395) (375)

AGI/10k squared -- -- -- -- -- -- 815*** 967***

(77) (72)

AGI/10k cubed -- -- -- -- -- -- -43*** -49***

(4) (4)

R-squared 0.006 0.022 0.467 0.377 0.488 0.416 0.472 0.383

N 53,000 63,000 33,000 38,000 36,000 41,000 33,000 38,000

Source: U.S. Census Bureau, 2021 Current Population Survey Annual Social and Economic Supplement (CPS ASEC) individually linked to tax year 2019-

2020 Internal Revenue Service (IRS) Form 1040 records.

Notes: The dependent variable is AGI from individually linked IRS Form 1040 records, from tax year 2020 or tax year 2019, linked to primary filers. The

main independent variable is AGI as modeled in the CPS ASEC tax model, based on responses to the 2021 CPS ASEC regarding income received during

the 2020 calendar year and household characteristics at the time of interview. Form 1040 records are linked via Protected Identification Key of the primary

filer. Self-employment is reported to the CPS. AGI sample restrictions are applied to both the IRS amounts and the CPS ASEC-modeled amounts. Standard

errors are in parentheses.

*** p < 0.01, ** p < 0.05, * p < 0.10.

Table 3. Regressing linked IRS Form 1040 tax year 2019 and 2020 adjusted gross income (AGI) on CPS ASEC-modeled tax year

2020 AGI.

Linear

Linear, restricted to AGI

$1-$150k

Linear, restricted to AGI

$1-$200k

Cubic, restricted to AGI

$1-$150k

Panel A. 2020 compared to 2019 Panel B. 2019 compared to 2018

Cell proportions and their standard errors Cell proportions and their standard errors

AGI

EIP using 2020

AGI is equal to EIP

using 2019 AGI

EIP using 2020

AGI is greater than

EIP using 2019

AGI

EIP using 2020

AGI is less than EIP

using 2019 AGI

Total

AGI

EIP using 2019

AGI is equal to EIP

using 2018 AGI

EIP using 2019

AGI is greater than

EIP using 2018

AGI

EIP using 2019

AGI is less than EIP

using 2018 AGI

Total

0.5 - - 0.5 0.6 - - 0.6

(0.04) - - (0.04) (0.04) - - (0.04)

54.2 - 6.0 60.2 56.7 - 6.3 63.0

(0.25) - (0.12) (0.28) (0.25) - (0.12) (0.28)

35.5 3.8 - 39.3 32.6 3.8 - 36.5

(0.26) (0.11) - (0.28) (0.24) (0.10) - (0.26)

Total

90.2 3.8 6.0

Total

89.9 3.8 6.3

(0.36) (0.11) (0.12) (0.35) (0.10) (0.12)

Table 4. Differences in adjusted gross income (AGI) across consecutive years of IRS Form 1040 records, and their consequent differences in Economic Impact

Payments (EIP).

EIP EIP

2020 AGI is equal

to 2019 AGI

2019 AGI is equal

to 2018 AGI

2020 AGI is greater

than 2019 AGI

2019 AGI is greater

than 2018 AGI

2020 AGI is less

than 2019 AGI

2019 AGI is less

than 2018 AGI

Source: U.S. Census Bureau, 2021 Current Population Survey Annual Social and Economic Supplement (CPS ASEC) individually linked to tax year 2018-2020 Internal Revenue Service (IRS)

Form 1040 records.

Notes: Form 1040 records are linked via Protected Identification Key of the primary filer. Lower AGI generally corresponds to higher EIP amounts. Economic Impact Payment amounts are imputed

from linked Form 1040 AGI, with filing status and dependents as modeled from the CPS ASEC Tax Model. The only variable changing across years is AGI. Samples are restricted to persons who are

primary filers in both of each pair of consecutive years. High proportions of equal EIP amounts are generally due to a filing unit having sufficiently low AGI to have received a full, unphased amount

based on either year's AGI value. "Less than" and "greater than" categories include differences as small as $1.

Table 5. Supplemental Poverty Measure for each Economic Impact Payment model.

Percent MOE Percent MOE Percent MOE

All people 9.1 0.2 9.0 0.2 0.1*** 9.4 0.2 -0.2***

Sex

Male 8.6 0.3 8.5 0.3 0.1*** 8.9 0.3 -0.2***

Female 9.6 0.3 9.5 0.3 0.1*** 9.9 0.3 -0.3***

Age

Under 18 years 9.7 0.4 9.5 0.4 0.2*** 9.8 0.4 -0.1***

18 to 64 years 8.8 0.3 8.7 0.3 0.1*** 9.1 0.3 -0.3***

65 years and older 9.5 0.4 9.4 0.4 Z *** 9.8 0.4 -0.3***

Type of Unit

Married couple unit 5.0 0.2 5.0 0.2 Z ** 5.0 0.2 Z *

Cohabiting partner unit 8.5 0.8 8.2 0.8 0.3*** 9.1 0.8 -0.6***

Female reference person unit 18.2 1.0 17.9 0.9 0.3*** 18.7 1.0 -0.5***

Male reference person unit 11.7 1.2 11.4 1.2 0.3*** 12.0 1.2 -0.3***

Unrelated individuals 17.5 0.5 17.5 0.5 Z * 18.3 0.5 -0.8***

Race and Hispanic Origin

White 8.1 0.2 8.0 0.2 0.1*** 8.3 0.2 -0.2***

White, not Hispanic 6.5 0.2 6.5 0.2 Z *** 6.7 0.3 -0.2***

Black 14.6 0.8 14.4 0.8 0.3*** 15.1 0.8 -0.5***

Asian 8.8 1.0 8.7 1.0 0.1 8.9 1.0 -0.1***

Hispanic (any race) 14.0 0.7 13.7 0.7 0.2*** 14.2 0.7 -0.3***

Educational Attainment

Total, aged 25 and older 8.4 0.2 8.3 0.2 0.1*** 8.7 0.2 -0.3***

No high school diploma 20.3 0.9 20.1 0.9 0.2*** 21.1 0.9 -0.8***

High school, no college 11.1 0.4 11.0 0.4 0.1*** 11.4 0.4 -0.3***

Some college, no degree 7.3 0.3 7.3 0.3 0.1*** 7.6 0.3 -0.3***

Bachelor's degree or higher 4.3 0.2 4.3 0.2 Z * 4.4 0.3 -0.1***

Production Model Upper Bound Model

Production Model -

Upper Bound Model

Source: U.S. Census Bureau, 2021 Current Population Survey Annual Social and Economic Supplement (CPS ASEC).

Notes: *** p < 0.01, ** p < 0.05, * p < 0.10. The margin of error (MOE) is a measure of an estimate’s variability. The larger the MOE in relation to the size of the

estimate, the less reliable the estimate. This number, when added to and subtracted from the estimate, forms the 90-percent confidence interval. The MOEs shown in

this table are based on standard errors calculated using replicate weights. Z represents statistically significant differences that round to zero.

Production Model - Lower

Bound Model

Lower Bound Model

Appendix Table 1. Estimates and standard errors for each model by adjusted gross income.

Estimates

Count Amount Count Amount Count Amount Count Amount Count Amount

Negative or zero 21.82 47.95 25.10 55.18 18.27 41.55 11.23 27.44 23.59 43.99

$1 under $10,000 21.94 51.36 21.94 51.36 18.76 45.63 13.24 31.76 20.67 42.45

$10,000 under $20,000 14.50 35.45 14.50 35.45 12.44 31.75 7.64 21.27 19.52 46.77

$20,000 under $30,000 15.54 38.48 15.54 38.48 13.35 34.54 7.89 23.87 17.99 44.37

$30,000 under $40,000 14.20 36.78 14.20 36.78 12.35 33.45 7.99 23.76 15.47 39.19

$40,000 under $50,000 11.47 30.78 11.47 30.78 9.98 28.08 7.28 21.86 12.20 31.84

$50,000 under $60,000 10.54 28.76 10.54 28.76 9.15 26.25 6.85 21.11 9.89 26.97

$60,000 under $75,000 12.46 37.76 12.46 37.76 11.21 35.50 8.89 28.56 11.57 34.20

$75,000 under $100,000 14.63 43.20 14.63 43.20 13.40 42.07 11.11 37.28 13.66 41.49

$100,000 under $200,000 19.56 71.41 19.56 71.41 19.56 71.41 18.56 65.92 16.99 61.34

$200,000 or more 0.62 0.41 0.62 0.41 0.62 0.41 5.36 15.34 0.41 0.27

Standard Errors

Count Amount Count Amount Count Amount Count Amount

Negative or zero 0.29 0.63 0.31 0.65 0.26 0.57 0.18 0.58

$1 under $10,000 0.25 0.60 0.25 0.60 0.22 0.56 0.21 0.63

$10,000 under $20,000 0.20 0.52 0.20 0.52 0.19 0.51 0.14 0.51

$20,000 under $30,000 0.20 0.50 0.20 0.50 0.18 0.49 0.14 0.60

$30,000 under $40,000 0.21 0.55 0.21 0.55 0.19 0.52 0.14 0.56

$40,000 under $50,000 0.18 0.47 0.18 0.47 0.17 0.46 0.14 0.50

$50,000 under $60,000 0.19 0.54 0.19 0.54 0.17 0.52 0.15 0.49

$60,000 under $75,000 0.19 0.65 0.19 0.65 0.18 0.64 0.16 0.64

$75,000 under $100,000 0.20 0.66 0.20 0.66 0.20 0.67 0.18 0.67

$100,000 under $200,000 0.22 0.91 0.22 0.91 0.22 0.91 0.23 1.09

$200,000 or more 0.04 0.04 0.04 0.04 0.04 0.04 0.12 0.50

Production Model Upper Bound Model Lower Bound Model Survey Reported IRS SOI

Source: U.S. Census Bureau, 2021 Current Population Survey Annual Social and Economic Supplement (CPS ASEC).

Notes: Each cell represents the standard error of the aggregate modeled sum of the first and second rounds of Economic Impact

Payments. Counts are in millions. Amounts are in billions of dollars.

Source: Internal Revenue Service Statistics of Income (IRS SOI) tables (irs.gov/statistics/soi-tax-stats-coronavirus-aid-relief-and-economic-security-act-cares-

act-statistics), U.S. Census Bureau 2021 Current Population Survey Annual Social and Economic Supplement (CPS ASEC).

Notes: Each cell represents the modeled aggregate sum of the first and second rounds of Economic Impact Payments. SOI counts are the maximum of first- and

second-round EIP aggregate counts, as the union of recipients is not separately tabulated. SOI totals include payments not based on tax returns, which is more

comparable to the totals from the ASEC Tax Model. The “negative or zero” AGI category for SOI was modified to include payments to individuals without tax

returns. Counts are in millions. Amounts are in billions of dollars. Ratios in Table 1 are based on unrounded SOI statistics.

Production Model Upper Bound Model Lower Bound Model Survey Reported

Appendix Table 2. Estimates and standard errors for each model by filing status.

Estimates

Count Amount Count Amount Count Amount Count Amount Count Amount

Sin

le or qualif

widow(er)

87.44 152.40 87.44 152.40 69.23 120.70 45.50 96.16 69.21 121.31

Married filin

oint or separatel

54.50 220.50 54.50 220.50 54.50 220.50 49.53 184.60 47.67 189.81

Head of household 15.35 49.45 15.35 49.45 15.35 49.45 10.61 36.00 21.36 61.31

Standard Errors

Count Amount Count Amount Count Amount Count Amount

Single or qualifying widow(er)

0.55 1.00 0.55 1.00 0.49 0.86 0.37 1.09

Married filing joint or separately

0.29 1.25 0.29 1.25 0.29 1.25 0.28 1.55

Head of household 0.19 0.63 0.19 0.63 0.19 0.63 0.16 0.64

Source: U.S. Census Bureau, 2021 Current Population Survey Annual Social and Economic Supplement (CPS ASEC).

Notes: Each cell represents the standard error of aggregate modeled sum of the first and second rounds of Economic Impact Payments to

its corresponding IRS SOI benchmark. Counts are in millions. Amounts are in billions of dollars.

The CPS ASEC Tax Model assigns single filer status such that it is most comparable to the sum of SOI single filers and SOI qualifying

widow(er) filers.

The CPS ASEC Tax Model assigns married filing jointly such that counts are most comparable to the sum of SOI married joint and one-

half of SOI married separate filers, and amounts are most comparable to the sum of SOI married joint and separate filers.

Source: Internal Revenue Service Statistics of Income (IRS SOI) tables (irs.gov/statistics/soi-tax-stats-coronavirus-aid-relief-and-economic-security-act-cares-act-

statistics), U.S. Census Bureau 2021 Current Population Survey Annual Social and Economic Supplement (CPS ASEC).

Notes: Each cell represents the modeled aggregate sum of the first and second rounds of Economic Impact Payments. SOI counts are the maximum of first- and second-

round EIP aggregate counts, as the union of recipients is not separately tabulated. Counts are in millions. Amounts are in billions of dollars. Ratios in Table 2 are based

on unrounded SOI statistics.

The CPS ASEC Tax Model assigns single filer status such that it is most comparable to the sum of SOI single filers and SOI qualifying widow(er) filers.

The CPS ASEC Tax Model assigns married filing jointly such that counts are most comparable to the sum of SOI married joint and one-half of SOI married separate

filers, and amounts are most comparable to the sum of SOI married joint and separate filers.

Production Model Upper Bound Model Lower Bound Model Survey Reported

Production Model Upper Bound Model Lower Bound Model Survey Reported IRS SOI