IRS Working Paper, January 10, 2025, Not for quotation or citation.

IRS Working Paper, January 10, 2025, Not for quotation or citation.

More Than Meets the Eye:

The Comprehensive General Indirect Effect of Auditing Individual

Income Tax Returns ∗

Alan Plumley and Daniel Rodriguez

IRS Research, Applied Analytics, and Statistics

Jess Grana and Miguel Sarzosa

MITRE

The views expressed in this paper are those of the authors and do not necessarily represent the

views of the IRS or of MITRE.

January 10, 2025

Abstract

IRS audit rates have generally fallen for over a decade due to declining resources.

In addition to loss of direct revenue, decreased enforcement likely results in

increased noncompliance, as well. We contribute to a small literature on the

“comprehensive” indirect effects of IRS enforcement on voluntary compliance

across the general taxpayer population—mostly those who were not directly subject

to the enforcement. Using microdata from random audits conducted for research

purposes, we find that misreporting increases on certain tax return line items as

overall audit rates decline. As we might expect, these effects are more pronounced

for line items that are the typical target of audits rather than line items addressed

through the automated matching program. We translate effects on misreporting into

effects on tax revenues and compare revenues against enforcement costs. We find

that over the Tax Year 2006-2014 period, the overall average marginal return on

investment (ROI) of IRS individual tax enforcement was between 13:1 and 16:1,

including a direct average ROI of 3:1 and an indirect average marginal ROI

between 10:1 and 13:1. In other words, for this time period, the general indirect

effect was 3 to 4 times the direct revenue effect. These ROI estimates advance

current understanding of the IRS’s overall impact and can inform budgetary

discussions.

Keywords: Tax audits, Spillovers, General Indirect Effect, ROI of audits

JEL Codes: H23, H26

The authors would like to thank Will Boning and comments from participants at the IRS-Tax Policy Center Joint

Research Conference on Tax Administration (June 2023), the National Tax Association annual conference (Nov 2023),

the Georgetown Tax Law and Public Finance Workshop (Jan 2024), the Society of Government Economists annual

conference (April 2024), the IRS Research, Applied Analytics, and Statistics peer review (Oct 2024), the Department

of the Treasury Office of Tax Analysis and IRS Joint Statistical Research Program seminar (Oct 2024), and the

Government Accountability Office (December 2024). Alan Plumley email: alan.plumley@irs.gov. Daniel Rodriguez

email: daniel.rodriguez@irs.gov. Jess Grana email: cheny@mitre.org. Miguel Sarzosa email: msarzosa@mitre.org.

∗

IRS Working Paper, January 10, 2025, Not for quotation or citation.

1 Introduction

How much additional revenue could be generated if the enforcement budget for the Internal

Revenue Service (IRS) were increased by $X per year? The answer to that question is far from

simple; it depends on the size of the current budget and how it is allocated to enforcement,

services, IT investments, and other activities. It also depends on how the enforcement budget is

allocated to the various enforcement programs. One impact on overall revenue would come in

the form of increased direct enforcement revenue – additional tax collections resulting from the

enforcement action for the tax year being enforced. Moreover, it is likely that the direct effect

would be accompanied by some indirect revenue effects—whether due to a subsequent change in

compliance behavior among the specific taxpayers who were the subjects of the enforcement

(known as the “specific indirect effect”), and/or a spillover due to a change in compliance

behavior among taxpayers in the general population who were not the subjects of the

enforcement (known as the “general indirect effect”). Estimating the spillovers of enforcement,

and thus the full extent of the return to investing in tax enforcement, is not straightforward, but it

is extremely important. The IRS 2024 budget request to Congress is a testament to this. It cites a

return on investment (ROI) in terms of direct revenue but “does not include the indirect effects of

IRS enforcement activities on voluntary compliance” (IRS, 2024). This paper intends to fill this

gap.

There have been numerous attempts over the last 40 years to estimate the general indirect effect

of changes in IRS enforcement—particularly changes in audit coverage rates. These efforts fall

within two approaches: (1) “local network” models; and (2) “comprehensive” models. Local

network models attempt to demonstrate that a general indirect effect exists in a particular

context. For example, they estimate the general indirect effect within a given segment of the

population (e.g., sole proprietors) through a specific type of network (such as the network of

taxpayers who are clients of the same tax preparer) and according to a particular behavioral

mechanism (e.g., deterrence). Using well-defined networks supports strong identification

strategies whereby a treatment group (i.e., a network that had an audited member) is compared

against a similar but untreated group. A drawback of local network models is that they are

context-specific. 1 Their findings may not be generalizable outside of the specific context or

behavioral mechanism studied. Taxpayers presumably participate in multiple networks

simultaneously (e.g., employer networks, professional networks, community networks, etc.), and

it is unclear whether the separate impact of these networks are additive. Taxpayers undoubtedly

form their perceptions in a more subtle way based on all the factors in their environment.

Although local network models lend themselves to theoretical premises and practical

experimentation, such narrowly defined analyses do not directly translate into operational

applications such as budget justification. To achieve that, the estimated indirect effects should in

theory include effects arising: 1) from all IRS enforcement activities; 2) across the general

taxpayer population; and 3) across all possible (or as many as possible) networks of propagation.

Comprehensive models are better suited for these purposes as they are agnostic about the

For instance, studies like Boning et al. (2020), Badgley et al. (2021), and Chetty (2013) show large spillovers of

audits that spread through networks of different kinds. Others that explore more light-touch interventions like

mailing letters shaming delinquent tax filers find mixed or no evidence of an indirect effect (Meiselman (2018);

Perez-Truglia and Troiano (2018); Grana et al. (2022)). These mixed results indicate that context matters: the

existence and size of an indirect effect depend on the specific network or community studied or even on research

design choices.

1

IRS Working Paper, January 10, 2025, Not for quotation or citation.

mechanism(s) affecting taxpayer behavior and are generally not restricted to a narrow subset of

the population. However, their identification is less straightforward. They depend heavily on

being able to control for all the main drivers of behavior in addition to the enforcement activity

in question.

This paper estimates a comprehensive model of the impact of individual income tax audits on the

general population. It is motivated by the observation that, due to a steady decline in IRS budgets

over the last 12 or so years, overall individual income tax audit coverage rates (the percentages

of any given subpopulations that are audited) have declined substantially, going from 1% in 2008

to less than 0.6% in 2014 (see Figure 3 in the Appendix). Despite the overall decline, audit rates

did not fall uniformly across Examination Activity Codes—IRS’ groupings of taxpayers based on

Total Positive Income 2 (TPI) level, the filing of certain schedules, and the claiming of the Earned

Income Tax Credit (EITC). 3 For instance, while audit rates fell from 10.8% in 2008 to 2.6% in

2014 for those earning $1,000,000 or more (Activity Code 281), Figure 1 shows that audit rates

in other Activity Codes remained somewhat stable and, in some instances, they even increased at

various points in time.

Figure 1. Audit Rates by Activity Code

Descriptive evidence suggests the existence of co-movement between audit rates and

noncompliance—measured by the Net Misreporting Percentage (NMP) on tax after refundable

credits (TARC). 4 As an example, Figure 2 illustrates the audit coverage and misreporting rates

TPI is the sum of all positive amounts of income and excludes income losses, such as from investments.

See the lists of Activity Codes, their definitions, and their relative importance as percentage of the taxpayer

population in Table 7 of the Appendix.

4

The NMP is defined as the aggregate net amount misreported on a given line item across a group of returns divided

by the sum of the absolute values of the corresponding amounts that should have been reported. The absolute values

are used in the denominator to ensure that negative amounts do not distort the aggregates. These misreporting

statistics were compiled from data generated by audits of a stratified random sample of tax returns each year under

the IRS National Research Program (NRP).

2

3

IRS Working Paper, January 10, 2025, Not for quotation or citation.

for individuals in Activity Code 272 (i.e., those whose returns fall below $200,000 in TPI and

which are not accompanied by supplemental forms like Schedule C, E, F or Form 2106 5 and do

not claim the EITC). This group comprises over half of all individual tax returns. Figure 2 shows

an overall upward trend in noncompliance contemporaneous with a declining trend in the audit

coverage rates over these years, suggesting the presence of a general indirect effect among this

large group of taxpayers.

Figure 2. Audit Coverage and NMP Trends, TYs 2006-2014 for Taxpayers with TPI <$200k

and no EITC, Schedule C, E, F or Form 2106 (55.3% of the Population)

In addition to an overall compliance response to changes in audit coverage, we further

hypothesize that this effect varies by the visibility of income and other tax return line items.

Compliance is more likely to be affected on line items that are often targets of audits. Line items

subject to automated matching programs (which are not audits) may be less affected by changing

audit rates.

This paper adds to the literature on the indirect effect of audits by using alternative model

specifications and exploiting new individual microdata to capture noncompliance. We differ from

prior research in our econometric specification: instead of the contemporaneous audit rate, we

evaluate the effect of a lagged audit rate on compliance. Taxpayers do not have contemporaneous

knowledge of the audit rate since information disseminates with a lag. Moreover, the IRS’s

estimate of the risk that a given return is noncompliant is also not contemporaneous; it is based

on audit results of similar returns from prior years. The significance of these lags is that they

reduce endogeneity concerns arising from reverse causality. We also differ from prior work by

Schedules C and F are used to report nonfarm and farm sole proprietor income and expenses, respectively;

Schedule E is used to report income from rental real estate, royalties, partnerships, S corporations, estates, trusts, or

residual interests in real estate mortgage investment conduits; and Form 2106 is used to report employee business

expenses.

5

IRS Working Paper, January 10, 2025, Not for quotation or citation.

exploring how the compliance response differs across groups of line items based on how visible

the line item is to the IRS through third-party reporting.

Our findings largely confirm the hypothesis that individual misreporting responds to audit rate

changes differently across visibility groups of line items in ways consistent with the extent of the

visibility. We translate the impact of audit rates on the misreporting of income or offset amounts

to the impact on tax revenues. Then, comparing revenues against enforcement costs, we calculate

the overall return on investment (ROI) of audits of individual income tax returns. We find that,

on average, $1 spent on individual income tax audits generates about $3 of direct revenue and an

additional $10 to $13 of indirect revenue (roughly 3 to 4 times the direct revenue). Our findings

are within the range of magnitude estimated by a handful of prior studies and close to the

estimate put forward by the U.S. Treasury indicating that the indirect effect is three times the

direct effect (Department of the Treasury, 2019).

The paper is organized as follows: Section 2 reviews the relevant empirical literature and

provides theoretical motivation for this research; Section 3 describes our data; Section 4

summarizes our estimation methods; Section 5 presents our empirical results; and Section 6

concludes.

2 Background and Theoretical Motivation

The decision to declare taxes is made under uncertainty. That is because a taxpayer’s failure to

fully report their income does not automatically trigger punishment from tax authorities. If a

taxpayer underreports income, the reward of doing so will depend on whether or not they are

investigated by the authorities. If they are not investigated, they are better off underreporting

than declaring their full income. However, if they are investigated and the penalty for

underreporting is greater than its benefits, they are worse off. That is why in the classical

economic theory of tax compliance, rational (risk-averse) individuals maximize the expected

utility of the tax evasion gamble, purposefully comparing the expected monetary benefits of

gaming the tax system against the risky prospect of detection and punishment (Allingham and

Sandmo, 1972). A key parameter in this context is, of course, the probability of detection. A

well-established result in the classical economic theory of tax compliance is that an increase in

the probability of detection will always lead to more income being declared (Lopez-Luzuriaga

and Scartascini, 2019). That is because a higher probability of detection reduces the expected

payoff of underreporting.

Incidentally, this is the theoretical foundation for the existence of the general indirect effect of

audits that we explore in this paper—the effect of IRS contacts (such as audits) on those who are

mostly not contacted themselves. It is not the fact that the person is audited, but the chances of

someone getting audited that drive the change in tax reporting. Early empirical evidence supports

this result. Studies like Dubin and Wilde (1988), Dubin, Graetz and Wilde (1990), Tauchen,

Witte and Beron (1993), and Plumley (1996), which we refer to as measuring the

“comprehensive indirect effect”, find that higher aggregate (e.g., state or ZIP code level)

contemporaneous audit rates on the general population (as a proxy for audit probability) are

associated with greater tax compliance. For example, using state-level panel data, Dubin, Graetz

and Wilde (1990), Plumley (1996), and Dubin (2007) find that the comprehensive indirect effect

of audits is six, eleven, and nine times that of the direct effect, respectively. Dubin and Wilde

IRS Working Paper, January 10, 2025, Not for quotation or citation.

(1988) and Grana et al. (2022) use zip-code level panel data and find mixed evidence of an

indirect effect, varying across taxpayer subpopulations and audit categories (see Table 1).

Table 1: Findings from Prior Studies on the Comprehensive General Indirect Effect

Ratio of Indirect to Direct Revenue (not ROI)

2:1 (high-income taxpayers only)

Tauchen, Witte, and Beron (1993)

6:1

Dubin, Graetz and Wilde (1990)

9:1

Dubin (2007)

11:1

Plumley (1996)

Mixed evidence

Dubin and Wilde (1988) and Grana et al. (2022)

However, contemporaneous audit rates are not public knowledge. So, if national audit rates are

abstract or distant from the day-to-day concerns of individual taxpayers, how can they have a

significant impact on tax reporting behavior? Taxpayers must build perceptions about them from

partial information gathered through various channels. The analyses of some of those channels

build a complementary and larger body of knowledge within the general indirect effects

literature. This sub-strain of the literature incorporates “local network” models that focus on a

single context and channel of information transmission.

One of those channels is tax preparers. Professional tax preparers closely monitor national audit

rates to better advise their clients. If audit rates are high, tax preparers may be more diligent in

ensuring compliance and advising clients to avoid aggressive tax positions. This professional

guidance influences taxpayers' behavior, even if they are not directly aware of the audit statistics

(Keppler, Mazur and Nagin, 1991). Boning et al. (2020) and Badgley et al. (2021) show that

professional tax preparers also catalyze a network effect on tax reporting. They find that

taxpayers who share tax preparers with IRS-visited/audited taxpayers tend to report more income

to the tax authorities. 6 That may be because the tax preparer becomes aware firsthand of the

possibility of misreporting detection and transfers that information to their other clients who

update their perceptions about the detection probability they face.

A similar channel through which aggregate audit rates can inform individual’s perceptions on

their chances of detection, akin to that of tax preparers, comprises taxpayers’ social networks. As

with tax preparers, this channel relies on making the taxpayer aware that she could have been

audited (or not) as their peers, family members, or colleagues have (or have not) been. This

channel’s effects capture responses driven by information about enforcement spread through the

network by word of mouth. As shown by Chetty et al. (2013) when documenting the geographic

variation in the take-up of the EITC, the knowledge generated by word of mouth can lead to

significant heterogeneity in behavior adoption. This mechanism is made explicit by Alstadsæter

et al. (2019) and Drago et al. (2020), who document that taxpayers affect each other’s decisions

about tax avoidance. In particular, Drago et al. (2020) find that neighbors of those who received

a letter addressing their tax reporting are more likely to switch from evasion to compliance than

households living in neighborhoods where no one received such a letter. These findings highlight

the fact that individuals form beliefs about their own detection probability using even fragmented

Similarly on the corporate taxation side, Bohne and Nimczik (2018) find that tax avoidance behaviors follow

managers and tax experts as they transfer between firms. Pomeranz (2015) finds that after a firm is audited, tax

compliance also improves among that firm’s suppliers.

6

IRS Working Paper, January 10, 2025, Not for quotation or citation.

information about aggregate audit rates, such as observing a neighbor receiving a tax-related

letter, profoundly shaping personal perceptions of audit risk. This suggests that individuals

extrapolate from these isolated instances to infer broader enforcement patterns, thus integrating

these observations into their understanding of the likelihood of being audited themselves, which

in turn influences their tax compliance decisions.

A similar reasoning can be used to understand how media coverage of specific audits and audit

rates can inform a person’s perception of audit risk, especially if those audits occur to people

with similar characteristics to them. One of those characteristics can be the location where the

audited people live. Tauchen, Witte, and Beron (1993) use audit rate variation at IRS-office level

on microdata from the IRS Taxpayer Compliance Measurement Program (TCMP) 7 to find that

local audit rates stimulate individual compliance. 8 They estimate that the indirect effect of audits

is twice the size of the direct effect. The reasoning for using office-level audit rates stems from

the fact that back in the 1970s, the period the paper analyzes, IRS district offices conducted the

audits and had different staffing levels. Thus, the number of audits they could complete varied

across district offices. As a result, some taxpayers were audited or not audited because they filed

in districts that were over- or under-staffed in relation to other districts. Therefore, the local audit

rate was informative to local taxpayers building their belief about the detection probability they

would face. As the IRS budget shrank over time and catalyzed by IRS restructuring in 1998,

auditing responsibility shifted from district offices to a centralized system heavily reliant on

correspondence audits. Hence, since the 2000s taxpayers’ audit rate references are largely

national instead of local.

3 Data

Our methodology relies on modeling individual level compliance as a function of IRS audit

rates, while controlling for other drivers of compliance. Our primary compliance measure is

derived from National Research Program (NRP) microdata. NRP selects a stratified random

sample of individual income tax returns for examination for a given tax year. Because the NRP

sample is designed to be representative of the population, audits through the NRP examine

taxpayers who might not have been examined under normal operational audit procedures. These

audits potentially encompass the whole tax return, as opposed to targeting specific areas of

noncompliance, as in operational audits. The program provides useful information about

noncompliance among the general population and the insights it reveals are used to update

operational audit selection procedures, improve resource allocation, and provide estimates of the

tax gap (IRS, 2022).

We interpret the behavior of the individuals in the NRP sample as being representative of similar

taxpayers in the general population. However, we are interested in the aggregate audit rate faced

by the segment of the population represented by the NRP taxpayer—not the audit probability of

TCMP, a precursor to IRS’s NRP, contained detailed information on compliance (resulting from detailed audits) for

a stratified random sample from the population.

8

On the corporate side, Hoopes, Mescall and Pitman (2012) take a similar approach and find that doubling the audit

rate increases effective tax rates by 7 percent. Notably, they survey corporate tax executives and find that many take

note of historical audit rates.

7

IRS Working Paper, January 10, 2025, Not for quotation or citation.

the taxpayer in the NRP sample. Audit rates are constructed by aggregating IRS enforcement

data according to the audit categories employed by both NRP and operational audits.

3.1 Dependent Variables

We select all returns audited through the NRP for TYs 2006-2014. 9 For each return, we use the

reported amounts and NRP-corrected amounts of certain line items. Our primary outcome

variable is the net misreported amount (NMA), a concept used throughout tax gap studies (IRS,

2022). It is calculated for a given set of line items as the difference between the correct amounts

and reported amounts for each return. We calculate six measures of NMA based on categories of

line items at the return level that span different types of income and offsets. For income and tax

categories, NMA is calculated as Corrected Amount – Reported Amount, and positive NMA

values indicate understatements of tax. For offset categories (e.g., offsets to income, such as

deductions, and offsets to tax, such as credits), NMA is calculated as Reported Amount –

Corrected Amount, so that positive NMA values again indicate understatements of tax.

For each return, we compute the NMA for six groups of tax return line items based on how

visible they are to the IRS. Four of the line-item groups relate to different types of income

(Visibility Groups 1-4), while the remaining two groups combine offsets to income (Visibility

Group 5) or offsets to tax (Visibility Group 6). We define visibility as the degree to which

income or offsets are subject to withholding and/or third-party information reporting.

Compliance on income reporting varies with the “visibility” of the income. Income subject to

little or no information, such as sole proprietor income, makes up the largest portion of the

underreporting tax gap (IRS, 2022).

Figure 3: Underreporting of Income as a Function of its Visibility to the IRS

Source: Internal Revenue Service (2022)

2015 NRP data was released at the time of the writing of this report, and we are adding these data to our sample in

ongoing work.

9

IRS Working Paper, January 10, 2025, Not for quotation or citation.

Visibility Group 1 is the income category subject to the most information reporting and

withholding while Visibility Group 4 is subject to the least. We hypothesize that compliance on

certain line items may be more responsive to IRS audit rates than others. For example, rising

audit rates may induce taxpayers to more accurately report line items that would be typically

targeted by an audit – items that have substantial, limited or even low visibility. It is unclear

whether taxpayers change their compliance behavior on high visibility line items that are usually

handled by automated document matching programs rather than by audits. It is also unclear

whether taxpayers change compliance on items with no information reporting since such income

can be difficult to validate through audits. In our analysis, we evaluate each NMA measure as the

dependent variable in separate analyses.

Each of the six NMA measures are relevant only to certain taxpayers, depending on their tax

situation. For each visibility group regression, we remove taxpayers who report zero amount and

have zero true (corrected) amount on any of the line items in the visibility group. This ensures

that a zero NMA value corresponds to compliance behavior and not to irrelevance of line items

for the given taxpayer.

Table 2. Visibility Group Definitions

Visibility

Group

1

Category

Line Items Included

Income

Wages & Salaries

2

Income

3

Income

4

Income

Pensions and annuities, unemployment compensation,

dividend income, interest income, state income tax

refunds, and taxable social security

Partnerships/S corp. income, capital gains, and

alimony income

Nonfarm proprietor income, other income, rents and

royalties, farm income, and form 4797 income

5

Offsets to

income

Offsets to

tax

6

Adjustments, deductions, and exemptions

Refundable and nonrefundable credits

Visibility

High: subject to substantial

information reporting and

withholding

Substantial: subject to substantial

information reporting

Limited: subject to some

information reporting

Low: subject to little or no

information reporting

Mixed: subject to varying

amounts of information reporting

Mixed: subject to varying

amounts of information reporting

3.2 Independent Variables

3.2.1 Audit Rates

The primary regressors of interest are audit rates. We construct the audit rate for a given tax year

from IRS enforcement data as the number of unique tax returns from that tax year that were

audited divided by the total number of unique returns filed for that year. We also create separate

audit rates for different groupings of taxpayers based on TPI level, the filing of certain schedules

(like Schedule C for nonfarm sole proprietors and Schedule F for farm sole proprietors) and

EITC claiming. These groupings of individual tax returns – known as “activity codes” – are

listed in Table 7 of the Appendix. As the third column of Table 7 shows, most of the taxpayer

population has modest annual income (below $200,000) and no active business income or

expenses (Activity Codes 272 and 273).

IRS Working Paper, January 10, 2025, Not for quotation or citation.

It is important to note that our dependent variable and other control variables are specified at the

return level, but our primary variable of interest – audit rates – is specified at the group level.

Each observation in our NRP sample is assigned the audit rate for that return’s activity code –

reflecting the likelihood that taxpayers are most responsive to audits of similarly situated

taxpayers (e.g., with similar types and amounts of income and offsets).

The second methodological decision we made about the audit rate variable was to specify a twoyear lag of audit rate in the regressions. The choice to lag the audit rate arises from the natural

delay in enforcement processing time. Figure 4 provides an example of the distribution of audit

start and audit closure dates relative to the filing year of the audited return, for two categories of

audit. For many audit categories, an audit begins 2-3 years and closes 2-4 years after the filing

year of the audited return. For example, a return for income earned in TY2010 would be filed in

spring 2011. If selected for audit, the taxpayer might be notified in late 2012. In spring 2013, the

taxpayer will file the TY2012 return. Thus, the audit rate pertaining to TY2010 returns is the

most recent information the taxpayer will have on IRS enforcement levels when filing the

TY2012 return – motivating a two-year lag on audit rate in our regressions.

Figure 4: Distribution of Audit Start and Closure for Two Categories of Audit

3.2.2

Control Variables

For each NRP return, our control variables are constructed from tax characteristics that may help

explain compliance behavior. These include filing status (whether the taxpayer filed as Married

Filing Jointly), the total exemptions claimed by the taxpayer, the presence of wage income, the

claiming of the child tax credit, whether the taxpayer itemized deductions, whether mortgage

interest was deducted, an indicator for taxpayers over 65 years of age, whether the taxpayer used

a paid preparer, and an indicator for electronic filing. We base these variables on the taxpayer’s

reported information on their return.

We also control for the correct amount on the return corresponding to the NMA variable of

interest. For example, when Visibility Group 6 (credits) NMA is the dependent variable, we

include the correct amount of credits as the regressor. This construction allows us to model

changes in NMA that arise from compliance behavior and not from changes in the underlying

true tax, income, or offsets.

IRS Working Paper, January 10, 2025, Not for quotation or citation.

3.3 Data Summary

Figure 5 summarizes sample size by activity code. We remove outliers by trimming the bottom

and top five percent from the distribution of total reported income in each activity code, since

there are outliers in terms of high income and negative income. This trimming affects the entire

NRP sample, regardless of visibility group. Within each visibility group regression, we remove

observations with negative NMAs. Negative NMAs imply overstating of income or

underclaiming of offsets, and in this paper, we focus on focus on noncompliance in the other

direction (which is more common). Except for Activity Code 271, our sample includes at least

4,000 returns for each activity code during TYs 2006-2014.

Figure 5. Counts of NRP Returns Before and After Trimming (TYs 2006-2014)

Figure 7 summarizes the aggregate NMA over time by visibility group. The total NMA for each

visibility group is calculated by weighting each return-level NMA in our NRP sample (using

NRP sampling weights) and summing across all returns. The largest source of noncompliance is

from Visibility Group 4, income line items with little or no information reporting (such as

nonfarm proprietor income and rents and royalties income). Aggregate NMA in this group fell

and then increased over time. The totals for Visibility Groups 3 and 5 fell and plateaued

somewhat.

IRS Working Paper, January 10, 2025, Not for quotation or citation.

Figure 7. Aggregate NMA* over Time, by Visibility Group (Weighted)

* The NMA for groups 1-4 represent understated income, while the NMA for group 5 represents overstatements of

income offsets and the NMA for group 6 represents overstatements of tax credits.

Figure 8 disaggregates NMA totals by activity code. Certain types of taxpayers are more likely to

have certain types of income and offsets and are thus more likely to contribute to NMA on those

items. For example, Activity Code 270 makes up a large portion of misreporting on credits

(Visibility Group 6) but a much smaller portion of misreporting on partnership/S corporation

income, capital gains and alimony income (Visibility Group 3). Activity Codes 279-281, despite

comprising only 3.7 percent of the population (per Table 7), contribute almost 25 percent of

misreporting on Visibility Group 3 income. Activity Code 272, which includes over 55 percent of

the population, contributes the largest portion of misreporting in Visibility Groups 1 and 2 but

much less for 3 and 4.

Figure 8. Aggregate NMA by Activity Code (Weighted)

Table 3 summarizes the dependent and independent variables in our model (excluding audit

rates) by Tax Year. These summary statistics apply to our trimmed data, and observations are

IRS Working Paper, January 10, 2025, Not for quotation or citation.

weighted by NRP sampling weights. Dollar-denominated variables (NMAs and Correct

Amounts) are adjusted to 2018 dollars. For the average return in our sample, NMA drops slightly

then increases during this time for most visibility groups. Correct amounts of Visibility Group 1,

3 and 4 income also drop slightly then increase during this time. Commensurate with decreasing

marriage rates and our aging population, the proportion of NRP taxpayers filing as Single/other

status increases somewhat, as does the proportion of taxpayers over 65. Variables declining

during this time are the proportion of taxpayers with wage income, claiming a child tax credit,

itemizing, and deducting mortgage interest. The use of a paid preparer fell over time, while

electronic filing rose dramatically until 2012 then slightly declined.

IRS Working Paper, January 10, 2025, Not for quotation or citation.

Table 3. Weighted Average Statistics for NRP Sample by Tax Year

Variable

2006

2007

2008

2009

2010

2011

2012

2013

2014

Dependent Variable (NMA)

Visibility Group 1

$157

$175

$189

$107

$140

$78

$144

$57

$143

Visibility Group 2

$416

$363

$347

$478

$462

$300

$395

$387

$416

Visibility Group 3

$1,194

$1,283

$695

$688

$663

$600

$723

$717

$938

Visibility Group 4

$3,617

$3,048

$2,762

$2,544

$2,720

$2,379

$2,827

$3,247

$3,156

Visibility Group 5

$893

$1,561

$1,370

$1,367

$1,395

$1,404

$1,225

$1,250

$1,443

Visibility Group 6

$330

$355

$369

$487

$545

$552

$451

$457

$447

Independent Variables

Correct Amount

Visibility Group 1

$52,400

$52,745

$50,424

$49,944

$48,948

$47,822

$50,503

$49,492

$50,665

Visibility Group 2

$9,715

$10,367

$9,745

$9,836

$10,191

$9,972

$9,728

$9,697

$10,141

Visibility Group 3

$10,290

$10,028

$6,477

$5,009

$6,135

$6,507

$8,382

$8,008

$9,623

Visibility Group 4

$11,556

$10,538

$9,613

$8,680

$9,770

$9,467

$11,211

$11,308

$11,846

Visibility Group 5

$18,495

$18,005

$17,431

$17,023

$16,528

$15,965

$16,250

$15,990

$15,803

Visibility Group 6

$1,091

$1,059

$1,221

$1,316

$1,213

$1,126

$1,116

$1,100

$1,149

Filing Status

Single/other

57%

58%

57%

57%

59%

60%

60%

61%

60%

Married filing jointly

43%

42%

43%

43%

41%

40%

40%

39%

40%

Total Exemptions

0 or NA

2%

1%

2%

2%

2%

2%

2%

2%

2%

1

32%

31%

31%

33%

33%

33%

34%

34%

35%

2

32%

33%

31%

32%

32%

32%

31%

32%

28%

3

17%

17%

17%

15%

16%

15%

15%

15%

16%

4

12%

11%

12%

12%

11%

11%

12%

11%

12%

5+

6%

7%

7%

6%

6%

6%

7%

6%

7%

Had wage income

85%

85%

85%

85%

84%

83%

85%

83%

83%

Claimed child tax credit

24%

23%

23%

21%

22%

19%

19%

19%

19%

Itemized

46%

46%

41%

39%

41%

40%

40%

39%

38%

Deducted mortgage interest

36%

37%

33%

31%

32%

30%

30%

29%

27%

Over 65

12%

13%

14%

14%

13%

13%

14%

15%

15%

Used paid preparer

66%

66%

65%

62%

63%

62%

62%

62%

59%

Filed electronically

50%

65%

71%

73%

80%

84%

84%

70%

70%

Note: These summary statistics apply to our trimmed NRP sample. Statistics are weighted by NRP sampling weights. Means are displayed for NMAs and Correct

amounts, while proportions are displayed for all other variables. Dollar-denominated variables are expressed in terms of 2018 dollars.

IRS Working Paper, January 10, 2025, Not for quotation or citation.

4 Methods

Our baseline specification models taxpayer i’s compliance in tax year t as a function of IRS

enforcement and other drivers of compliance: 10

𝑁𝑁𝑁𝑁𝑁𝑁𝑖𝑖𝑖𝑖 = 𝛽𝛽0 + 𝛽𝛽1 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑔𝑔,𝑡𝑡−2 + 𝛽𝛽2 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖𝑖𝑖 +

𝜷𝜷𝑻𝑻𝑻𝑻𝑻𝑻𝑻𝑻𝑻𝑻𝑻𝑻𝑻𝑻𝑻𝑻 𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝒊𝒊𝒊𝒊 + 𝜹𝜹𝑨𝑨𝑨𝑨𝑨𝑨𝑨𝑨𝑨𝑨𝑨𝑨𝑨𝑨𝑨𝑨 𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝒈𝒈 + 𝜀𝜀𝑖𝑖𝑖𝑖

(1)

We run a separate regression for each visibility group. Return-level NMA on those line items is

our main dependent variable. Since there is skewness in NMA, we winsorize NMAs to the 95th

percentile. Audit rate is the primary variable of interest. As discussed previously, each taxpayer is

assigned the audit rate for their activity code group g for the tax year in question. We lag the

audit rate by two years to reflect the delay in enforcement processing time. We hypothesize that

𝛽𝛽1 will be negative—a decrease in audit rates should lead to an increase in noncompliance.

Importantly, we control for the correct amount that should have been reported on the line items

in question, for each visibility group; this is presumably the most important determinant of what

is actually reported, and therefore the NMA. Additional taxpayer control variables refer to the

variables described in Section 3.2.2. We include fixed effects for activity code. These capture

time-invariant determinants of compliance that are unique to each activity code, unrelated to

audit rate changes. We do not include tax year fixed effects in our regressions due to our reliance

on variation over time to identify the audit rate effects. 11 Finally, all regressions are weighted by

NRP sampling weights.

Our econometric approach is most similar to Tauchen, Witte, and Beron (1993) and Hoopes,

Mescall, and Pitman (2012), who evaluate the effect of aggregate audit rates on compliance at

the micro level (while controlling for auditor assessed income or proxies thereof). One difference

from their approach is that we use lagged audit rates instead of contemporaneous ones. While a

contemporaneous audit rate reflects audit probability for the return being filed, it is unlikely that

the taxpayer knows the contemporaneous audit rate or their audit probability until the audit cycle

for that year has completed. Rather, they are more likely to be aware of historical audit rates. To

the extent that audit rates change over time (which they have), contemporaneous audit rates are

not a suitable replacement for historical ones.

Another departure from Tauchen, et al. (1993) and Hoopes, et al. (2012) is in the treatment of the

audit rates econometrically. They use an instrumental variable approach, but we don’t for two

reasons. First, lagged audit rates do not suffer from reverse causality, as taxpayers cannot

influence past audit rates through current reporting behavior and IRS cannot influence past

compliance behavior through current audits. Second, audit rates have generally declined across

the board at varying rates due to declining resources and shifts in allocation (but not in response

to improved compliance), thereby creating a natural experiment for evaluating the causal effect

of audit rates.

Since NRP samples are independent each year, our data are pooled cross-sections rather than panel/longitudinal.

Our model controls for tax law changes through the correct amount, but it does not control for any tax policy

changes that are specific to certain taxpayer groups, such as through the inclusion of activity code-tax year fixed

effects. Such effects would be collinear with our audit rate variables, which do not vary within an activity code and

tax year. In future work, we hope to include variables capturing known policy changes for certain activity codes.

10

11

IRS Working Paper, January 10, 2025, Not for quotation or citation.

5 Results

In this section, we present the results of estimating Equation (1), focusing on the main findings

related to the audit rate variable. We then translate the estimated impacts on line-item reporting

into impacts on revenue using a tax calculator. Finally, we combine revenue with cost data to

calculate the final return on investment of IRS enforcement during this time period.

5.1 Regression Results

Table 4 presents our regression results. The number of observations for each regression varies

due to the trimming of negative NMAs and “irrelevant” taxpayers (zero reported and zero true

amount) for each visibility group. Smaller sample sizes affect statistical power and may be

responsible for the lack of statistical significance on the audit rate variable for Visibility Group 3.

Audit rates have the expected negative effect on noncompliance for all visibility groups except

for Group 1, which shows a small positive effect (not statistically significant). 12 This aligns with

our hypothesis that the effect of audit rate variation is likely small or zero, given this line item is

mostly validated by automated matching programs. For Visibility Group 2, a one percentage

point increase in audit rates decreases noncompliance on a return by $139. This is a modest but

statistically significant effect. This group includes taxpayers such as retirees with

pensions/annuities income and taxpayers between jobs receiving unemployment income.

For Visibility Group 3, a one percentage point increase in audit rates decreases noncompliance

on a return by $694, but this effect is not statistically significant. However, this regression was

conducted on the smallest sample size. This income group includes partnership/S corporation

income, capital gains, and alimony income—sources of income with some limited information

reporting. These types of income are often the targets of audits, and it is likely that the lack of

statistical significance arises from sample size issues rather than from no meaningful effect.

For Visibility Group 4, a one percentage point increase in audit rates decreases noncompliance

on a return by $806 – the largest effect across all visibility groups and is statistically significant.

Income in this group is subject to very little information reporting – such as nonfarm proprietor

income, rents and royalties, farm income, and form 4797 income. This is also the visibility group

with the largest amount of noncompliance (see Figure 3) and thus more room for improvement in

compliance if audit rates were to rise.

Finally, audit rates have the expected effect on adjustments, deductions, exemptions, and credits

(Visibility Groups 5 and 6). A one percentage point increase in audit rates decreases

noncompliance on adjustments, deductions, and exemptions by $80 per return (not statistically

significant) and on refundable and nonrefundable credits by $64 per return (statistically

significant).

The result for Visibility Group 1 is consistent with a separate analysis we conducted using Automated

Underreporter (AUR) data (results are not included here for conciseness) among a sample of tax returns taken from

the entire population. AUR matches third-party information documents sent to the IRS with what taxpayers report on

their tax returns. This screens for noncompliance on line items with substantial information reporting, such as

wages and salaries. We construct a measure of NMA based on AUR-corrected line items. While NRP-adjusted NMA

is available only for NRP audits, AUR-adjusted NMA is available for all taxpayers using third-party information

documents. This approach allows us to evaluate a sample of taxpayers outside the standard NRP population for this

analysis.

12

IRS Working Paper, January 10, 2025, Not for quotation or citation.

Table 4. Regression Results†

Dependent Variable: NMA

Visibility

Group 1

Visibility

Group 2

Visibility

Group 3

Visibility

Group 4

Visibility

Group 5

Visibility

Group 6

14.07

-139.06 ***

-694.26

46.65

53.32

510.80

-806.28 *

-80.58

-63.09 ***

488.40

103.29

10.53

0.001 ***

0.008 ***

0.003 ***

0.016 ***

-0.024 ***

-0.0004 ***

0.000

0.000

0.000

0.000

0.001

0.000

Total Exemptions 1

-82.66

4.27

1,758.35

5,013.89 **

222.30

-44.04

Total Exemptions 2

112.69

138.38

1,223.25

2,031.73

-208.98 *

87.31

3,155.59 **

6,420.07 ***

119.70

148.56

1,429.04

2,166.77

-273.54 **

98.29

3,903.88 **

6,279.30 ***

126.39

157.52

1,530.01

2,252.24

-299.91 **

168.35

3,387.64 **

6,912.06 ***

133.92

164.74

1,599.40

2,338.20

-158.58

345.84 **

4,818.52 ***

7,831.20 ***

141.66

173.54

1,689.29

2,435.90

339.74

101.47

218.58 ***

-988.15 **

-3,383.41 ***

285.10 **

67.89 ***

Audit Rate (2 Year Lag)

Correct Amount

Total Exemptions 3

Total Exemptions 4

Total Exemptions 5+

Wage Income

Used paid preparer

F Statistic

Degrees of Freedom

536.63 ***

299.75

1,860.49 ***

319.44

2,271.76 ***

100.16

609.14 ***

100.83

704.55 ***

482.67

633.11

133.99

19.52

-2,958.06 ***

-313.95 **

-110.79 ***

49.96

59.51

676.53

819.15

125.75

13.20

-139.96 *

-156.35 ***

-656.25

2,399.92 ***

2,869.74 ***

3.10

73.59

60.44

550.10

886.61

157.75

22.67

30.23

192.29 ***

881.68

-2,028.66 **

-741.34 ***

-51.24 **

72.57

59.68

549.61

888.76

160.72

22.51

-182.10 **

361.39 *** -1,232.10 **

-4,639.70 ***

-187.20

52.60 **

76.122

55.306

510.240

759.367

147.722

23.007

132.68 ***

-4.05

747.70 *

957.22 *

-280.53 ***

69.49 ***

36.91

38.86

422.47

573.40

88.84

11.12

81.36 **

-90.49 **

-1,099.77 ***

-1,616.72 ***

75.57

7.15

40.48

39.19

382.35

535.43

93.87

11.98

63.30

32.21

-1,711.42 **

-12.72

-1,590.65 ***

-410.54 ***

54.97

66.04

829.94

896.88

132.89

15.35

246.38 *

365.29 **

5,425.60 ***

3,883.68

-155.02

227.57 **

147.27

185.29

1,841.59

2,373.13

360.45

102.18

91,569

83,897

55,908

77,393

118,991

64,190

N

N

N

N

N

N

0.001

4.072 ***

91,479

0.027

92.130 ***

83,847

0.009

20.887 ***

55,864

0.04

126.071 ***

77,331

0.022

104.359 ***

118,867

0.047

122.403 ***

64,132

Tax Year Fixed effect

Adjusted R2

1,981.16 ***

-882.18

Married-Joint Status

Observations

99.60

54.40

Filed electronically

Constant

282.88

-262.04 ***

Deducted mortgage int.

Over 65

99.14

308.48 ***

41.23

Claimed child tax credit

Itemized

264.20

1,553.30 ***

† Standard errors on second line.

Statistical significance: *** 1%

** 5%

* 10%

IRS Working Paper, January 10, 2025, Not for quotation or citation.

5.2 Translating Changes in Line-Item Misreporting into Changes in Revenue

The coefficients on the audit rate variable in Table 4 describe the impact of a change in audit rate

on dollars of misreporting (i.e., NMA). We translate the impact on reporting compliance into the

impact on tax revenue. Mechanically, this first involves taking the change in dollars of

misreporting for the entire visibility group (derived from the regression coefficient and the actual

change in audit rate) and allocating these changes to individual line items within the visibility

group. This allocation was done in proportion to how the detected NMAs were distributed across

line items within the visibility group on the original return – reflecting the assumption that the

rate of change in misreporting is the same for each line item in the category. Further, we ensure

these allocations are subject to the tax rules governing each line item. This process is especially

important for offset line items, which often are subject to different limitations than other items in

the same visibility group.

Table 5 illustrates how a hypothetical audit rate decline affects a hypothetical tax return.

Columns 5 and 6 show the detected amount of NMA (from the NRP audit) and the reported

amount from the NRP return. These “actuals” are the implied result of an audit rate decline two

years prior (in this example). In columns 3 and 4, we calculate the counterfactual amount

reported and the corresponding NMA had the audit rate not declined. The last column shows the

difference between the actual and the counterfactual amounts – this is the impact on this return of

the decline in audit rate.

For example, no NMA was detected on wages and salaries for the hypothetical return in Table 5

– so the counterfactual NMA remains zero due to our allocation rules. However, there was $150

of misreporting detected on interest and dividend income. This detected amount was the result of

an audit rate decline in this example – so the counterfactual misreported amount ($100) is lower.

Likewise, the counterfactual misreported amounts are lower for all line items that had a detected

NMA on this hypothetical return. Lower NMAs in turn result in higher counterfactual income

and lower offsets.

Once NMA changes are allocated to individual line items, we feed the counterfactual tax return

through a tax calculator to determine the tax liability that would have been reported on the NRP

return had the audit not changed. The bottom right box (in yellow) shows the overall impact on

tax after refundable credits (TARC) – this taxpayer would have paid $552 more in TARC had

audit rates not declined two years prior. Finally, we apply this approach to each NRP return and

apply NRP weights to calculate population-level revenue impacts of the audit rate changes.

IRS Working Paper, January 10, 2025, Not for quotation or citation.

Table 5. Illustrative Impact of a Hypothetical Audit Rate Decline on Tax Paid by a

Hypothetical Taxpayer

Visibility Line Item

$ Reported

NMA

w/o decline w/o decline

6

Tax

Offsets

5

Income

Offsets

4

Low / No

3

Limited

2

Substantial

1 High Wages & Salaries

$60,000

Pensions & annuities

Unemployment compensation

Interest & dividend income

$2,500

State income tax refunds

$500

Taxable social security benefits

Partnership / S corp. income

Trust income

Capital gains

$3,000

Alimony income

$100

Nonfarm proprietor income

$70,000

Farm income

Rents & royalties

$50,000

Form 4797 & Other income

Total Income

$186,100

Adjustments

Exemptions

$8,000

Deductions

$20,000

Tentative tax

$31,515

Nonrefundable credits

$2,600

Detected

NMA

Observed

Return

$0

$0

$60,000

$0

$100

$0

$150

$0

$2,450

$500

-$50

$0

$160

$400

$10,000

$200

$500

$11,000

$2,960

$0

$69,000

-$40

-$100

-$1,000

$4,545

$5,000

$49,545

-$455

$15,205

$16,850

$184,455

-$1,645

$0

$3,000

$5,098

$100

$0

$3,150

$5,600

$150

$8,000

$20,150

$31,013

$2,650

$0

$150

-$502

$50

$5,198

$5,750

$28,363

-$552

∆

Refundable credits

Tax after refundable credits

(TARC)

$28,915

5.3 Calculating Return on Investment

The final step of our analysis is to calculate return on investment (ROI). We combine the revenue

estimates from the prior section with data on enforcement costs. We use IRS records to calculate

the cost of audits corresponding to the audit rates used in Equation (1). We include costs

associated with the Exam, Appeals, Counsel, and Collection functions. It is important to note that

aggregate audit costs generally move in the same direction as audit rates, with a few exceptions

that likely arise from productivity changes (such as from a different mix of auditor experience or

levels year over year). We remove these handful of year-activity code observations where this is

the case.

Table 6 summarizes the direct ROI, general indirect ROI, and combined ROI for four groupings

of taxpayers based on TPI. Direct ROI is calculated from audit records and includes only the

additional tax actually paid as a result of the audit for the tax year that was audited. We see that

$1 of enforcement cost during this 2006-2014 time period generated $3.30 of direct revenue on

average and almost $9 when applied to audits of taxpayers earning $400k and above. The general

IRS Working Paper, January 10, 2025, Not for quotation or citation.

indirect ROI shown is calculated from this paper’s analysis and shows the population-level

impact of a dollar of auditing cost. We provide a range of general indirect ROIs depending on

whether we use all point estimates from Table 4 (high end) or only statistically significant

estimates (low end). $1 of enforcement cost generates around $10-$13 of general indirect

revenue, with larger impacts on taxpayers earning between $200k-$400k. 13 Finally, combined

ROI shows the total impact of a dollar of enforcement. $1 of enforcement costs generates, on

average, roughly $13-16 of total revenue when considering direct and indirect effects. (Note that

the variation in these ROIs across the TPI ranges is not directly applicable to IRS resource

allocation decisions, which should be made on the basis of the cost-effectiveness of the next

enforcement case. In contrast, the direct ROIs here are averages (total revenue divided by total

cost) and the indirect ROIs are average marginals (the change in revenue divided by the change

in cost. Nonetheless it seems likely that taking indirect effects into account would change the

mix of enforcement allocations to the various categories.)

Finally, we calculate the implied revenue loss from the audit rate declines observed from 2009 to

2012. Although this decline resulted in a $211M savings in enforcement costs, it led to an

estimated loss of almost $2.2B in voluntary tax revenue from 2011 to 2014. 14

Table 6. Return on Investment of IRS Individual Income Tax Audits, Tax Years 2006-2014

Return Total Positive Income

< $100K

$100K to under $200K

$200K to under $400K

$400K and over

All Groups

Direct ROI

2.0

2.8

3.1

8.9

3.3

General Indirect ROI

5.7 - 6.2

9.4 - 14.1

15.0 - 22.3

7.8 - 12.4

9.6 - 13.1

Combined ROI

7.7 - 8.2

12.2 - 16.9

18.1 - 25.4

16.7 - 21.3

12.9 - 16.4

6 Discussion

While most research on the impact of IRS enforcement on overall tax compliance evaluates

specific local networks, this paper contributes to a small literature on the “comprehensive”

general indirect effects of IRS enforcement. We aim to capture the effects on the entire taxpayer

population of all IRS individual income tax audits, regardless of the channels through which the

impacts propagate throughout the population. As such, these effects are relevant for IRS budget

justification, which currently cites the ROI of enforcement on direct revenue and does not

quantify overall indirect effects (IRS, 2024).

We advance understanding of the nature and magnitude of comprehensive indirect effects by

implementing several novel or rarely used approaches. Ours is one of the few papers in this area

Note that our ROI numerator uses tax amounts based on NMAs as recommended by the NRP auditors, while the

ROI denominator is the full life-cycle cost of the audits (i.e., not just the Examination cost, but also the cost of any

Appeals, Chief Counsel, and Collection activity to assess and collect the tax due). This “apples vs. oranges” ratio

yields a lower bound compared with an alternative of using only the Exam cost in the denominator. Alternatively, if

we projected the recommended amount to corresponding dollars collected, the ROI would go down, but it wouldn't

take into account changes in undetected NMAs, which are not observable. So, our ROI definition seems to reflect

the best available balance of being conservative yet realistic.

13

The net average marginal ROI of 10.3 is less than 13.1 because of offsetting increases in audit rates in some

activity codes and years.

14

IRS Working Paper, January 10, 2025, Not for quotation or citation.

to use microdata. This allows for more nuanced modeling of taxpayer behavior and the ability to

control for return-level characteristics. Departing from prior papers, we use lagged audit rates to

proxy for knowledge of IRS enforcement levels. While audit rates for the tax year at hand reflect

the true aggregate probability of audit, taxpayers (and their accountants) can plausibly know only

past audit rates. Additionally, using lagged audit rates solves the reverse causality (endogeneity)

problem; an earlier audit rate is not impacted by this year’s compliance, for example.

We find that the indirect effect of audits varies across tax return line items. The effect is larger

for items subject to less third-party information reporting and for items with large existing

noncompliance. These results are intuitive. High visibility line items such as wages and salaries

are screened by automated underreporter (document matching) programs, and misreporting on

these line items may be less sensitive to audit rates per se. On the other hand, misreporting on

line items not validated by simple document matching should be more responsive to the

enforcement actions, such as audits, that focus on those line items.

Our top-level finding is that IRS audits of individual income tax returns had a combined ROI of

13:1 to 16:1 during the 2006 to 2014 Tax Years. Put another way, the general indirect effect was

3 to 4 times larger than the direct effect. This is in line with prior studies (see Table 1) and

slightly on the lower end of the range of prior estimates. These results can be used to understand

how historical IRS budget cuts have impacted voluntary compliance and how new IRS funding

(such as through the Inflation Reduction Act) present an opportunity to reverse that trend.

6.1 Limitations and Future Research

It is important to remember that this study is focused solely on the reporting noncompliance

behavior detected on individual income tax returns, which is the largest component of IRS tax

gap estimates (IRS, 2022); it does not address the nonfiling or underpayment components of the

tax gap, nor does it encompass other types of tax. Because of this focus, the only IRS

enforcement considered so far has been audits of timely filed individual income tax returns.

Another limitation of this research is that NRP audits may not detect all noncompliance among

taxpayers with high and unreported income. This will impact the accuracy of our dependent

variable. Prior research has attempted to shed light on previously undetected offshore accounts

and passthrough income (Guyton et al., 2021) but has not explored its relation to changes in

compliance over time.

Moreover, our estimates relate just to the specific time period studied and may not be directly

generalizable to the present. This is because the relationship between audit rates and taxpayer

behavior in the general population seems to be highly dependent on things like: the distribution

of audit resources across the various categories of tax returns; the distribution of income,

deductions, and tax credits across tax returns; the extent to which other factors influence

taxpayer behavior; and the tax law in place in a given year. Although it is likely that the general

indirect effect today is similar to what we have estimated for the 2006 to 2014 time period, our

estimates are not a universal constant.

There are several near-term extensions we plan to address. We plan to deepen the theoretical

motivation for the audit rate variable and potentially change its specification to improve causal

linkages and introduce more variation. This could be done, for example, by deriving audit rates

for population sub-strata beyond Activity Code. We also hope to increase statistical power

IRS Working Paper, January 10, 2025, Not for quotation or citation.

through other means. NRP samples are limited in size (and have been declining in recent years),

affecting our ability to derive precise estimates. A potential alternative to using NRP data directly

is to impute compliance measures from NRP to the universe of tax returns. Although this would

greatly improve sample size, proper validation would need to be conducted to ensure compliance

imputations are reliable.

Finally, the ultimate goal of this research is to support IRS budget justifications by estimating the

ROI of all IRS activities. IRS service, outreach, education, and IT investments plausibly have an

impact on compliance, as well. These IRS services help taxpayers become more informed and

better equipped to report and pay their taxes correctly at the outset. To account for this, we hope

to incorporate into future iterations of this work measures such as IRS website hits and level of

service. Although we focus on individual taxpayers in this paper, prior research indicates that

corporations track IRS enforcement activities in their accounting practices (Hoopes, Mescall, and

Pitman, 2012). Estimating the indirect effect of enforcement on corporate voluntary compliance

is another area of future work.

IRS Working Paper, January 10, 2025, Not for quotation or citation.

7 References

Allingham and Sandmo 1972. Income Tax Evasion: A Theoretical Analysis. Journal of Public

Economics. 1:323-338.

Alstadsæter, Annette, Wojciech Kopczuk and Kjetil Telle. 2019. “Social Networks and Tax

Avoidance: Evidence from a Well-Defined Norwegian Tax Shelter.” International tax and

Public Finance 26: 1291-1328. https://doi.org/10.1007/s10797-019-09568-3

Badgley, Ellen, Kyle Furlong, Lucia Lykke, Leigh Nicholl and Alan Plumley. 2021. “Audit

Contagion? Investigating the General Indirect Effect of Audits Through Tax Preparer

Networks.” The IRS Research Bulletin: Proceedings of the 2021 IRS/TPC Research

Conference. Publication 1500 (Rev. 5-2022)

Bergolo, Marcelo, Rodrigo Ceni, Guillermo Cruces, Matias Giaccobasso, and Ricardo PerezTruglia. 2018. Policy Implications of Suboptimal Choice: Theory and evidence. AEA Papers

and Proceedings. 108:83-87.

Bohne, Albrecht, and Jan Sebastian Nimczik. 2018. “Information Frictions and Learning

Dynamics: Evidence from Tax Avoidance in Ecuador.” Institute of Labor Economics

Discussion Paper Series. No. 11536. https://docs.iza.org/dp11536.pdf

Boning, William C., John Guyton, Ronald Hodge, and Joel Slemrod. 2020. “Heard It Through

the Grapevine: Direct and Network Effects of a Tax Enforcement Field Experiment.” Journal

of Public Economics 190: 104261. https://doi.org/10.1016/j.jpubeco.2020.104261

Chetty, Raj, John N. Friedman, and Emmanuel Saez. 2013. “Using Differences in Knowledge

Across Neighborhoods to Uncover the Impacts of the EITC on Earnings.” American

Economic Review 103(7): 2683-2721. https://www.jstor.org/stable/pdf/42920668.pdf

Department of the Treasury. 2019. “FY 2019 Budget in Brief: Internal Revenue Service, Program

Summary by Budget Activity.” https://home.treasury.gov/system/files/266/16.-IRS-FY-2019BIB-FY2019.pdf

Drago, Francesco, Friederike Mengel, and Christian Traxler. 2020. “Compliance behavior in

networks: Evidence from a field experiment.” American Economic Journal: Applied

Economics 12.2: 96-133.

Dubin, Jeffrey A., and Louis L. Wilde. 1988. “An empirical analysis of federal income tax

auditing and compliance.” National Tax Journal 41(1): 61-74.

https://www.jstor.org/stable/41788709

Dubin, Jeffrey A., Michael J. Graetz, and Louis L. Wilde. 1990. “The Effect of Audit Rates on

the Federal Individual Income Tax, 1977-1986.” National Tax Journal 43(4): 395–409.

https://www.jstor.org/stable/41788861

Dubin, Jeffrey A. 2007. “Criminal investigation enforcement activities and taxpayer

noncompliance.” Public Finance Review 35(4): 500-529.

https://authors.library.caltech.edu/99369/1/sswp1200.pdf

Erard, Brian (1992). “The Influence of Tax Audits on Reporting Behavior.” In Why People Pay

Taxes: Compliance and Enforcement, edited by Joel Slemrod.

IRS Working Paper, January 10, 2025, Not for quotation or citation.

Grana, Jess, Nelson Foster, Navya Kambalapally, India Lindsay, Betsy Lydon, Max McGill,

Alexander McGlothlin, and Brian Rhindress. 2022. “Indirect Effects Research and Support.”

MITRE Technical Report.

Grana, Jess, India Lindsay, Lucia Lykke, Max McGill, Alexander McGlothlin, Leigh Nicholl,

Alan Plumley. 2024. “The specific indirect effect of IRS audits.” International Tax and

Public Finance. https://doi.org/10.1007/s10797-024-09866-5

Guyton, John, Patrick Langetieg, Daniel Reck, Max Risch, and Gabriel Zucman. 2021. “Tax

evasion at the top of the income distribution: Theory and evidence.” National Bureau of

Economic Research, No. w28542.

https://www.nber.org/system/files/working_papers/w28542/w28542.pdf

Hoopes, Jeffrey L., Devan Mescall, and Jeffrey A. Pittman. 2012. “Do IRS audits deter corporate

tax avoidance?” The Accounting Review 87(5): 1603-1639.

https://www.jstor.org/stable/pdf/41721904.pdf

Internal Revenue Service. 2022. “Federal Tax Compliance Research: Tax Gap Estimates for Tax

Years 2014-2016.” Publication 1415 (Rev. 08-2022). https://www.irs.gov/pub/irspdf/p1415.pdf

Internal Revenue Service. 2024. “Fiscal Year 2025 Congressional Budget Justification & Annual

Performance Report and Plan.” Publication 4450 (Rev. 2-2024). https://www.irs.gov/pub/irspdf/p4450.pdf

Keppler, Mazur and Nagin. 1991. Expert Intermediaries and Legal Compliance: The Case of Tax

Preparers. The Journal of Law & Economics. 34(1): 205-229.

Lopez-Luzuriaga, Andrea, and Carlos Scartascini. 2019. Compliance Spillovers Across Taxes:

The role of penalties and detection. Journal of Economic Behavior and Organization.

164:518-534.

Meiselman, Ben S. 2018. “Ghostbusting in Detroit: Evidence on nonfilers from a controlled field

experiment.” Journal of Public Economics 158: 180-193.

https://doi.org/10.1016/j.jpubeco.2018.01.005

Perez-Truglia, Ricardo, and Ugo Troiano. 2018. “Shaming tax delinquents.” Journal of Public

Economics 167: 120-137. https://doi.org/10.1016/j.jpubeco.2018.09.008

Plumley, Alan. 1996. “The Determinants of Individual Income Tax Compliance.” Department of

the Treasury, Internal Revenue Service, Publication 1916 (Rev. 11-96).

Pomeranz, Dina. 2015. “No taxation without information: Deterrence and self-enforcement in the

value added tax.” American Economic Review 105(8): 2539-2569.

https://www.jstor.org/stable/pdf/43821348.pdf

Slemrod, Joel, Marsha Blumenthal, and Charles Christian. 2001. “Taxpayer response to an

increased probability of audit: evidence from a controlled experiment in Minnesota.” Journal

of Public Economics 79(3): 455-483. https://doi.org/10.1016/S0047-2727(99)00107-3

IRS Working Paper, January 10, 2025, Not for quotation or citation.

Tauchen, Helen V., Ann Dryden Witte, and Kurt J. Beron. 1993. “Tax Compliance: An

Investigation Using Individual Taxpayer Compliance Measurement Program (TCMP) Data.”

Journal of Quantitative Criminology 9: 177-202. https://www.jstor.org/stable/23365801

IRS Working Paper, January 10, 2025, Not for quotation or citation.

8 Appendix

Figure 4. Audit Coverage* Trend Among Individual Income Tax Returns, TYs 2006-2014

Note: Plots the overall decline in audit coverage among individual income tax returns during Tax Years (TYs) 20062014.* Coverage rate = (number of returns audited) / (total number of returns filed) for the tax year

IRS Working Paper, January 10, 2025, Not for quotation or citation.

Table 7. IRS Examination Activity Code Definitions

Activity Description

Code

Percent of

Population

Group

270

EITC present & TPI < $200,000 and Schedule C/F TGR <

$25,000 or EITC w/o Sch C/F (As of TY 2008)

17.1%

EITC

271

EITC present & TPI < $200,000 and Sch C/F TGR >

$24,999 (As of TY 2008)

1.2%

EITC

272

TPI < $200,000, no Sch C, E, F, or Form 2106 (As of TY

2008)

55.3%

Non-Business

Mid-Income

273

TPI < $200,000 and Sch E or Form 2106, no Sch C or F

(As of TY 2008)

10.8%

Non-Business

Mid-Income

274

Non-Farm Business w/ Sch C/F TGR < $25,000 and TPI <

$200,000 (As of TY 2008)

7.3%

Business

275

Non-Farm Business w/ Sch C/F TGR $25,000 - $99,999

and TPI < $200,000 (As of TY 2008)

2.1%

Business

276

Non-Farm Business w/ Sch C/F TGR $100,000 - $199,999

and TPI < $200,000 (As of TY 2008)

0.6%

Business

277

Non-Farm Business w/ Sch C/F TGR > $199,999 and TPI

< $200,000 (As of TY 2008)

0.5%

Business

278

Farm Business Not Classified Elsewhere and TPI <

$200,000 (As of TY 2008)

0.9%

Business

279

No Sch C or F and TPI > $199,999 and < $1,000,000 (As

of TY 2008)

2.4%

Non-Business

High-Income

280

Sch C or F present and TPI > $199,999 and < $1,000,000

(As of TY 2008)

1.0%

Business

281

TPI > $999,999 (As of TY 2008)

0.3%

Non-Business

High-Income