Psychological Factors, the Choice of a Tax Preparer, and Tax

Agency decision

Ask Donna

What actually matters in this document.

Text

Compliance*

James Alm, Tulane University

jalm@tulane.edu

Jubo Yan, Nanyang University

yanjubo@ntu.edu.sg

William D. Schulze, Cornell University

wds3@cornell.edu

Melissa Vigil, Internal Revenue Service

melissa.vigil@irs.gov

Carrie von Bose, Public Company Auditing Oversight Board

carrievonbose@gmail.com

Running Head: Psychological Factors, the Choice of a Tax Preparer, and Tax Compliance

* This research was funded by the Internal Revenue Service (IRS) (TIRNO-14-Z-00017). The

views expressed are those of the authors and do not necessarily reflect the opinions of the IRS or

of any researchers working within the IRS. We thank the IRS Office of Knowledge Development and

Application, especially Kim Bloomquist, Alan Plumley, and John Guyton, for helpful comments. Please

address all correspondence to James Alm as the corresponding author: James Alm, Department of

Economics, Tilton Hall, Tulane University, New Orleans, LA 70118 USA (jalm@tulane.edu).

Psychological Factors, the Choice of a Tax Preparer, and Tax

Compliance

Abstract:

We use laboratory experiments to examine factors influencing taxpayer choice of a tax preparer

and the subsequent reporting behavior. We find that individuals in this environment

simultaneously choose a preparer and their compliance based in part on factors predicted by

rational choice theory. However, we also find that psychological factors play a central role in this

setting: participants prefer tax preparers who are “credentialed,” even when the cost is high or

the credential has no impact on outcomes; participants fear an audit, regardless of its likelihood;

participants often choose high-cost preparers even when they are fully compliant; and many

participants forego substantial expected earnings rather than underreport income.

Key Words: tax compliance; tax preparer; experimental economics; rational choice theory;

behavioral economics.

JEL Classifications: H2, H26, C91.

I. INTRODUCTION

The tax preparation process can be quite complicated, and many taxpayers find it

worthwhile to hire a tax professional to complete the tax-filing process on their behalf. The U.S.

Internal Revenue Service (IRS) estimates that the vast majority of all individual tax returns filed

were prepared either using automated software/web applications or with the assistance of a

professional tax preparation service. 1 As discussed in detail later, there is some research that has

examined the demographic characteristics and life events associated with choosing to use a tax

preparer, and there is also some work on the subsequent impact on taxpayer compliance of the

decision to use a tax preparer. However, despite the many insights from this work, it is plagued

by data problems, especially the inability to quantify both the specific factors that determine tax

preparer choice and to identify the impact of this decision on taxpayer compliance.

In this paper, we present evidence from laboratory experiments that examines two basic

questions. 2 First, what tax preparer characteristics are most important to taxpayers in their

decision to use a tax preparer? Second, how does this choice affect taxpayer compliance? For

both questions, we focus on the possible impact of psychological factors (or behavioral

economics considerations), as well as rational choice considerations, on these decisions.

The most recent data from the IRS Statistics of Income for 2018 indicates that 56.3 percent of all individual tax

returns were prepared with the help of a paid tax preparer, while another 36 percent were completed using tax

preparation software. See https://www.irs.gov/statistics/soi-tax-stats-individual-tax-statistics. These percentages

have stayed roughly constant for the last decade. See also the IRS webpage devoted to various types of information

about, and for, tax preparers, at https://www.irs.gov/tax-professionals. Additional background information on

taxpayer use of tax preparers is provided by U.S. Government Accountability Office (2003).

There is a long tradition of using laboratory experiments to examine tax compliance behavior. As discussed later,

laboratory experiments seem particularly well-suited for the study of many aspects of the taxpayer reporting

decision, despite potential concerns about their external validity. For some early experimental studies, see Friedland,

Maital, and Rutenberg (1978), Spicer and Becker (1980), and Beck, Davis, and Jung (1991); for more recent

examples, see Austin, Bobek, and LaMothe (2019), Young (2020), and Kasper and Alm (2022). For a detailed

discussion of the methodology of laboratory tax compliance experiments, along with a survey of many of the results,

see Alm (2019).

Answers to these questions have implications not only for tax compliance scenarios but

also for financial and even medical decision-making. There are many situations in which

consumers must make choices with uncertain outcomes and in which these choices are

moderated by experts (e.g., tax preparers, financial advisors, or medical doctors). 3 Some

components of risk are inherent to the decision itself, but risk and cost also vary based on

attributes and actions of the service provider, and individuals often choose a provider with

imperfect information about his or her performance. As a result, the decision to use a specific

service provider involves beliefs about the provider's performance and the ways in which this

performance relate to the expected outcome.

In the specific tax compliance context that we examine in the laboratory, taxpayers may

hire a tax preparer for several possible reasons, including a desire to save time, minimize the

amount of taxes owed, or ensure that their tax return is completed correctly in order to avoid an

audit. Our experiment presents detailed information to the taxpayer about potential tradeoffs

between these reasons in an environment in which individuals choose their level of audit risk by

selecting a specific automated tax preparer and also by choosing the amount of income to report

(or to underreport).

By presenting participants in the experiments with detailed information about preparer

characteristics and observing their choices, we are therefore able to make inferences about

taxpayer preferences on the roles of psychological factors (e.g., emotion) versus rational choice

considerations (e.g., reason) in the choice of tax preparers, and we are also able to make

inferences about the roles of emotion versus reason in participants' subsequent reporting

For example, see Giacobbe and Segal (1996), Grable and Joo (2001), Bechwati (2011), Hanna (2011), Robb,

Babiarz, and Woodyard (2012), and Cummings and James (2014).

decisions. These choices thereby provide insights into how taxpayers evaluate potential tax

preparers and also how taxpayers make their reporting decision. This information also provides

important practical guidance that informs how much and what kind of information should be

provided to taxpayers about tax preparers while simultaneously improving compliance.

To design our experiment, we first used focus groups to determine which tax preparer

characteristics were relevant in taxpayer choice of tax preparers. We then took these focus

groups results into the laboratory. Our experimental design required subjects facing a

complicated tax compliance decision to select from various types of tax preparers, where the

possible tax preparer types incorporated those preparer characteristics that our focus groups

revealed were most relevant in the taxpayer choice. Along with the subject's choice of tax

preparer, each subject also made his or her tax reporting decision over multiple and independent

rounds. After the completion of these rounds, we administered an online questionnaire to solicit

information on real-world tax experiences, demographic characteristics, risk aversion, and

social-value orientation.

We find that individuals simultaneously choose their tax preparer type and their

compliance level based on the expected tax savings generated by the tax preparer, the expected

tax savings generated by underreporting income, the expected penalties, and the cost of

preparation, just as rational choice theory would suggest. However, our results also suggest that

psychological factors play an important role in taxpayer decisions: taxpayers prefer tax preparers

who are “credentialed”, even though a tax preparer’s credentials had no impact on actual

outcomes and even when the cost of a credentialed tax preparer is high; taxpayers are very risk

averse and they fear the mere possibility of an audit, regardless of its likelihood; taxpayers often

choose high-cost preparers with low or zero probabilities of audit even when they report all of

their income; and many participants are willing to forego substantial expected earnings rather

than underreport income. The possibility of avoiding any kind of an audit seems especially

important to taxpayers in their choice of a tax preparer, perhaps because of the fear of an audit or

the guilt associated with an audit that uncovers less than full compliance. In short, both

psychological and rational choice factors seem to play important roles in taxpayer decisions

when choosing tax preparers and reporting income.

II. SOME RELEVANT LITERATURE

There are several strands of literature that are relevant to our analysis. One strand

examines what factors affect a taxpayer ’s choice of a tax preparer. Another strand – and in some

sense a prior, strand examines what motivates a taxpayer to pay (or not to pay) his or her legally

due tax liabilities. We discuss both strands, starting with the taxpayer’s compliance decision.

The Taxpayer’s Compliance Decision

The standard theoretical model used in nearly all research on tax compliance is based on

the rational choice model of Allingham and Sandmo (1972), as derived from the economics-ofcrime model of Becker (1968). Here a rational individual is viewed as maximizing the expected

utility of the tax evasion gamble, weighing the benefits of successful cheating against the risky

prospect of detection and punishment, and the individual pays taxes because he or she is afraid of

getting caught and penalized if he or she does not report all income. This “portfolio” approach

gives the plausible and productive result that compliance depends upon audit rates and fine rates,

with reported income increasing with an increase in either the audit rate or the penalty rate.

Indeed, the central point of this approach is that an individual pays taxes because – and only

because – of this fear of detection and punishment. 4

However, it seems clear to many observers that compliance cannot be explained entirely

by such purely financial considerations, especially those generated by the level of enforcement. 5

The percentage of individual income tax returns that are subject to a thorough tax audit is

generally quite small in most countries, almost always well less than 1 percent of all returns.

Similarly, the penalty on even fraudulent evasion seldom exceeds more than the amount of

unpaid taxes, and these penalties are infrequently imposed; civil penalties on non-fraudulent

evasion are even smaller. A purely economic analysis of the evasion gamble based on rational

choice models suggests that most rational individuals should either underreport income not

subject to source-withholding or overclaim deductions not subject to independent verification

because it is extremely unlikely that such cheating will be caught and penalized. However, even

in the least compliant countries, evasion seldom rises to levels predicted by a purely economic

analysis, and in fact there are often substantial numbers of individuals who apparently pay all (or

most) of their taxes all (or most) of the time, regardless of the financial incentives they face from

the enforcement regime. The low levels of compliance predicted by the economics-of-crime

approach are simply not observed. Indeed, the puzzle of tax compliance behavior may well be

why people pay taxes, not why they evade them. 6

For useful recent surveys of the compliance literature, see Sandmo (2012), Alm (2012, 2019), and Slemrod (2019).

For example, see Andreoni, Erard, and Feinstein (1998), Torgler (2007), and Kirchler (2007), among many others.

There are reasons why this analysis somewhat overstates the problem with the standard rational choice model,

given the presence of such factors as third-party information, source-withholding, targeted audits, and various

“audit-type activities” (e.g., line matching and information requests). Even so, there is little doubt that in many

settings the chances of detection and punishment are slight. Especially in circumstances in which third-party sources

of information and employer source-withholding are limited, the chances that an individual who does not report

truthfully will be caught and penalized are quite limited.

In sum, the standard rational choice model of tax compliance has generated important,

plausible, and relevant insights. Even so, the model has some well-recognized deficiencies,

especially its conclusion that enforcement is the sole factor that motivates compliance. In

addition, some of its predictions are counterintuitive and in fact inconsistent with actual

evidence, such as the prediction that an increase in the tax rate will actually increase reported

income. These concerns suggest that either the compliance decision must be affected by other

factors or it must be affected in ways not captured by the standard approach.

In large part because of these concerns, there have been numerous efforts to extend the

basic rational choice model of tax compliance. These efforts have taken two basic forms. Some

of these theoretical extensions have occurred within the basic economics-of-crime approach,

thereby keeping a reliance on rational choice considerations but adding elements that make the

model more realistic. These extensions include adding such factors as: source-withholding; an

individual labor supply decision; multiple individual strategies for reporting; alternative penalty,

tax, and tax withholding functions; complexity and the associated uncertainty about tax liability;

the receipt of government services; positive (individual) rewards for honesty (e.g. eligibility for a

lottery if found to be compliant); audit selection rules that utilize information from tax returns to

determine whom to audit; and, importantly for our purposes, the use of paid preparers. 7

These extensions add necessary realism to the basic model. Even so, they leave

enforcement as the main factor that motivates compliance, and they also often yield

counterintuitive predictions, especially low levels of predicted compliance that are seldom

observed.

Again, see Sandmo (2012), Alm (2012, 2019), and Slemrod (2019) for detailed discussions.

Importantly, there has also been much work to expand the basic rational choice model of

tax compliance by introducing aspects of behavior considered by psychology, as discussed under

the broad rubric of “behavioral economics”. These extensions recognize that individuals do not

always behave according to the standard assumptions of the neoclassical model of human

behavior: that individuals are rational, that they are motivated only by their self-interested desire

to maximize their own individual welfare, and that they have unlimited willpower. Instead, as

emphasized by Rabin (1998), Kahneman (2011), and Congdon, Kling, and Mullainathan (2011),

individuals often deviate from these assumptions in two broad (and sometimes overlapping)

dimensions: imperfect individual optimization (stemming from, say, limited computation

abilities or bounded self-control) and non-standard preferences (like other-regarding

preferences). The former area emphasizes individual behavior; the latter focusses more on group

considerations.

For example, much of the individual behavior that diverges from neoclassical predictions

involve some form of frame dependence, in which an individual’s decision depends upon how

the choice is presented. Frame dependence is typically related to some psychological

predisposition or some cognitive limitation of the individual. Many individuals react much

differently to gains than to equal-but-opposite valued losses; they often misperceive the true

costs and benefits of their actions; they may not be able to make all of the computations implied

by standard optimization given, say, limits on time or cognitive abilities; and they may be

motivated by a wide range of factors, including self-interest (narrowly defined) but also by

notions that arise more from group considerations, as discussed later. Also important here is

individual behavior under uncertainty, and there are now various formalizations of non-expected

utility theory that have been applied to individual choices, especially those based upon the work

of Kahneman and Tversky (1979, 1984) and Tversky and Kahneman (1974, 1981) via their

prospect theory.

The other strand of behavioral economics focusses more on group considerations, often

summarized as social interactions theory. There is abundant evidence that individuals are

influenced by the social context in which, and the process by which, decisions are made. There is

also much evidence that they are motivated not simply by self-interest but also by group notions

like social norms, social capital, social customs, appeals to patriotism or conscience, or feelings

of fairness, altruism, reciprocity, empathy, sympathy, trust, guilt, shame, morality, and

alienation, all of which depend upon the individual’s interactions with a larger group. These

group considerations also affect individual behavior in significant ways.

For some specific applications of non-expected utility theory to tax evasion, see the

detailed discussions in Hashimzade, Myles, and Tran-Nam (2013), and Alm (2019). There are

also many applications of social interactions theory to tax evasion, as surveyed by Torgler

(2007), Kirchler (2007), and Alm (2019). All of these models considerably complicate the

analysis of taxpayer behavior. However, they also often generate predicted levels of compliance

that far better approximate observed levels.

Our theoretical models of taxpayer compliance (and of taxpayer choice of a tax preparer)

use many elements of the rational choice model. Importantly, our models also use many elements

suggested by psychology via behavioral economics. These models are discussed in detail later.

The Taxpayer’s Choice of a Tax Preparer

As for the taxpayer’s decision to use a tax preparer, there is a large, and largely

descriptive, literature on what factors are associated with taxpayer choice of tax preparers. 8 This

work finds that there are specific demographic groups that are more likely to use tax preparer

services. People who are older, female, married, self-employed, have children, or identify as

Black or white are more likely to use a tax preparer or a financial planner. In addition, taxpayers

with higher incomes, more wealth, higher financial risk tolerance, and more financial knowledge

are also more likely to use a tax preparer or a financial planner. Taxpayers who have experienced

specific life events (e.g., losing a spouse or experiencing a drastic change in income) are also

more likely to use a tax preparer. Finally, taxpayers who have more tax forms to complete, who

need to file specific types of non-traditional tax forms, who face a higher risk of audit or

exposure to penalties, or who have greater uncertainty about their tax liability are more likely to

use a tax preparer or financial planner.

This literature also examines the motivation for choosing a tax preparer. Taxpayers

choose to use a tax preparer for several reasons, including a desire to obtain the maximum

refund, to properly comply with all tax regulations, and to save time. Time constraints and shifts

in the economy can also change people's willingness to pay for any type of professional service,

including a paid tax preparer. A desire to saving money or a desire for increased leisure time are

positively associated with a greater general willingness to pay for professional services.

See especially Long and Caudill (1987), Shavell (1988), Scotchmer (1989), Dubin et al. (1992), Christian, Gupta,

and Lin (1993), Ashley and Segal (1997), Erard (1997), Cloyd and Spilker (1999), Frischmann and Frees (1999),

Tan (1999), Guyton et al. (2005), Urban Institute (2005), Stephenson (2010), and Fleischman and Stephenson

(2012).

The decision to use a tax preparer is also related to the taxpayer's tax refund status and

the desire for aggressive or conservative tax reporting. 9 Taxpayers typically overpay to reduce

anxiety about being audited and to compensate for uncertainty about their true tax liability. They

also withhold more of their earnings so they can obtain the positive feeling of receiving a refund

check, which is more satisfying than retaining their original funds. Researchers call this

propensity to overpay taxes and avoid underpayments a “conservative framing stance” compared

to an “aggressive framing stance”, in which taxpayers prefer to pay as few taxes as possible and

risk underpayment. More experience with the tax process, higher tax liabilities, smaller refunds,

additional funds owed at filing, and more uncertainty are all negatively associated with

overpaying taxes and thus positively associated with taking an aggressive framing stance. Overly

conservative tax-reporting frames create a market for tax preparers, and people with overly

conservative tax-reporting frames are more likely to desire instantaneous refunds, to want to

reduce their overpayments from the previous year, and to seek out tax advice to reduce or

confirm their uncertainty.

Finally, there is work on the impact of tax preparer usage on subsequent taxpayer

compliance. 10 Theory suggests, and empirical work largely confirms, that the use of tax

preparers will tend to reduce many “unintentional” reporting errors, or those associated with tax

issues that are ambiguous and unclear. On these issues, the expertise of tax preparers seems to

reduce reporting errors that would otherwise lead to less compliance. However, empirical work

also finds that the use of tax preparers is associated with more noncompliance on tax issues

See Christian et al. (1994), Dusenbury (1994), Ayers, Kachelmeier, and Robinson (1999), Jackson et al. (2005),

Bobek, Hatfield, and Wentzel (2007), and Jackson and White (2008),

For theoretical analyses of some of the effects of tax preparers on compliance, see Shavell (1988), Scotchmer

(1989), and Klepper and Nagin (1991). For empirical analyses of these effects, see Long and Caudill (1987),

Klepper, Mazur, and Nagin (1991), Erard (1993, 1997), and Battaglini et al. (2020).

where the law is largely clear; that is, tax preparers tend to worsen the problem of deliberate

noncompliance. 11

All of this work is insightful and valuable. However, much of it is based on settings in

which it difficult to identify clearly both the causal impact of specific tax preparer characteristics

on taxpayer choices and the impact of tax preparer usage on compliance. The following sections

present our framework for examining these issues and then our experimental results. 12

III. EXPERIMENTAL DESIGN

Pre-experiment Focus Group Findings

In order to gather qualitative insights into the process of choosing a tax preparer and also

to determine which tax preparer characteristics should be included in the experiment, we first

conducted a series of four focus groups in which we recruited focus group participants and then

asked them to discuss their most desired characteristics of a tax preparer and their general

process of seeking out a tax preparer. Participants also completed specific decision-making tasks,

in which they were asked to choose between several hypothetical tax preparers. Four focus

groups with eight participants each were conducted: two took place in the Washington, D.C.,

metro area, and two took place in Ithaca, N.Y. All participants were members of the general

adult population who had previously used a paid tax preparer on at least one occasion.

There is also work that examines other issues, such as optimal enforcement policies in a world in which taxpayers

can choose tax preparers. For example, see Reinganum and Wilde (1991).

Note that one implication of taxpayer use of tax preparers is the likelihood that the probability that the taxpayer is

selected for audit becomes endogenous, dependent upon the information that the tax preparer conveys to the tax

authority on the tax return. For analyses of these “endogenous audit rules”, see Alm, Cronshaw, and McKee (1993),

Alm and McKee (2004), Clark, Friesen, and Muller (2004), Cason and Gangadharan (2006), and Gilpatric, Vossler,

and McKee (2011).

From the focus groups, we learned that many individuals choose a tax preparer based on

recommendations from friends, on the personal characteristics of the tax preparer, and on

performance-related variables. The focus groups determined that the tax preparer's audit rate,

credentials, level of customer satisfaction, form and schedule expertise, passing a compliance

check, passing an IRS background check, percent of errors on tax returns, cost, years of

experience, and location were the most useful pieces of information when selecting a paid tax

preparer. Overwhelmingly, the most important qualities that participants sought in tax preparers

were competency and trustworthiness. The desire to hire someone who would “get it right” was

mentioned often. Participants made linkages between the tax preparers' audit rate, Automated

Under Reporter (AUR) notices, percent errors, and compliance checks, expecting these attributes

to be consistent with each other. Indeed, many of the variables discussed in the focus groups

were based on variables contained in the IRS preparer registration database.

Consistent with findings from the field, many focus group participants originally began

using a tax preparer because of a change in tax circumstances or in anticipation of a particularly

complex tax year compared to previous years. Most participants found a tax preparer through

their social networks, either by hiring someone they knew or by asking for recommendations.

This familiarity helped them feel that they could trust the person with their financial information.

Most said that they did not use resources such as newspapers, websites, advertisements, or other

media, to help them choose a tax preparer. Participants expressed a preference for having more

information when working through the tax preparer choice task. Some participants expressed a

desire to do more research when choosing a tax preparer in the future, based on the variables

presented in the focus group.

Experimental Protocol and Design Features

We then designed the experiment to represent the elements of the tax preparer choice that

our focus groups revealed were the most relevant to their decisions. Specifically, our focus

groups indicated clearly that taxpayers mainly compare a more aggressive preparer who will help

them obtain a higher tax refund but possibly expose them to a higher audit risk, to a less

aggressive preparer who will help them obtain a lower tax refund but also will expose them to a

lower audit risk.

One of the potentially useful implications of this research is to suggest opportunities for

the IRS to provide additional information to aid taxpayers in their selection of a tax preparer.

This work can help inform the IRS in its consideration of balancing of taxpayers’ desire for

information on preparer quality and the need to protect preparer privacy.

The experiment consisted of several steps designed to mimic the tax-filing process of a

typical U.S. taxpayer. Participants were recruited using standard and accepted procedures. Upon

arriving at the computerized laboratory, participants earned income, they were faced with a

complicated tax reporting decision, and they chose a tax preparer to help them with this decision.

Throughout, participants interacted with the experiment through private computer terminals, and

all participants proceeded through the experiment choices concurrently. 13

At the beginning of the experiment, participants earned income by guessing the number

of gumballs in a jar. This guess determined each participant's “certain” income in each round of

the experiment, which ranged from 5,000 to 10,000 experimental dollars. In each round, the

participants were allocated an additional amount of “random” income, which ranged from 0 to

Demographic and other descriptive statistics on participants are available upon request.

5,000 experimental dollars. Each participant also received information about the number of

credits and deductions for which he or she was eligible, ranging from 0 to 5 credits and

deductions, which were randomly assigned in each round. The amount of income, credits, and

deductions was designed to vary from round to round, imitating the changes in individual

circumstances and also any changes in the tax code that affect tax liability from year to year. The

scenarios were designed to be somewhat more complex than those in a typical tax experiment, so

that participants would feel the need to use a tax preparer. A self-prepare option was not

presented to the participants in order to keep the focus on the choice of a tax preparer.

The experiment then proceeded in several steps:

1) In the first stage of each round, participants were presented with their specific tax

information. Their certain income was given along with the amount of taxes withheld.

Tax payments are automatically withheld from this income at a rate of 30 percent, similar

to the payroll taxes that are automatically deducted for most U.S. workers’ income.

Participants were also presented with their random income for the round (0–5,000

experimental dollars) and their number of credits (0–5) and deductions (0–5).

2) In the second stage of each round, participants were presented with a choice of four tax

preparers with varying attributes. Participants are informed that these tax preparers are

automated rather than controlled by other participants. For each tax preparer, participants

knew whether the tax preparer had credentials (defined in the experiment as having

passed a background check and being a licensed Certified Public Accountant), the audit

rate associated with the preparer, the average tax savings (defined as the expected

reduction in taxes due to credits and deductions) for the preparer’s clients, and the tax

preparer’s fee. Each of these terms was defined in a glossary that was appended to the

experiment instructions for the participants’ reference. The credentials were irrelevant to

the tax preparer’s performance, but the average tax savings and audit rate directly

affected the experiment outcome. Participants who chose a preparer with a higher value

of average tax savings (referred to hereafter as high-refund preparers) had their

deductions randomly valued between 500 and 900 experimental dollars and their credits

valued between 150 and 250 experimental dollars, while those who chose a preparer with

a lower value of average tax savings (referred to hereafter as low-refund preparers) had

their deductions randomly valued between 100 and 500 experimental dollars and their

credits valued between 50 and 150 experimental dollars. The participants’ probability of

being audited was based on the audit probability of the tax preparer they chose. The

participants were informed of the ways in which their choice of tax preparer would affect

their tax return and audit outcomes.

3) In the third stage of each round, participants reported their tax information to their tax

preparer. A text box was available for participants to enter their random income for that

period. They could enter any amount from zero to the full amount of their random income

(i.e., the program did not allow them to over-report income). Fields for certain income,

deductions, and credits were prefilled and could not be changed, in order to represent

real-world scenarios in which wage income and many types of credits and deductions are

usually well-documented, whereas other types of income (e.g., tips, freelance income, or

contract work) are often self-reported with little documentation required. Participants

were informed at the start of the experiment that, if they are audited in any round, all

unpaid taxes will be collected and a 100 percent penalty on unpaid taxes will be assessed.

4) In the fourth and final stage of each round, participants received their tax refund

information and found out their audit result. Any audit penalties were assessed, and

participants viewed their final net income for the round.

The experiment continued for ten rounds, comparable to tax periods. Upon completion of the

final round, an online questionnaire was administered in which participants answered several

questions about their real-world tax experiences, demographic characteristics, risk aversion, and

social-value orientation (the degree to which a person values the welfare of others relative to

their own welfare). After completing the questionnaire, participants collected their earnings in

cash. Earnings were based on a predefined exchange rate of experimental dollars to U.S. dollars.

The experimental instructions are included in Appendix (1), selected screen shots of the

decision screens are presented in Appendix (2), and the post-experiment questionnaire is

included in Appendix (3). The experiment was programmed using z-Tree software. In total, 22

sessions were conducted. Ten of these were conducted in the Fors Marsh Group Experimental

Economics Laboratory in Arlington, VA, and 12 were conducted in Cornell University’s Lab for

Experimental Economics and Decision Research in Ithaca, NY. All participants were members

of the general adult population, who either had used a paid tax preparer or had prepared their

own taxes at least once.

Experimental Treatments

The experimental design and procedures remained the same across all 22 sessions, as

summarized in Table 1. Four different treatments were implemented to explore the choice

tradeoffs. The values of the parameters across the four treatments included the choice of tax

preparers who were credentialed or not; tax preparers with audit probabilities of 0.00, 0.05, 0.20,

0.35, or 0.40; and tax preparers with tax preparation costs of $150, $200, $300, $400, $500,

$1200, or $1500. The average tax savings for each preparer was either $437.50 for low-refund

preparers or $937.50 for high-refund preparers. The parameter values used for each treatment are

given in the charts below. Note that Treatment 1 is designed to explore the impact of credentials

for higher and lower audit probabilities, and for high and low tax savings, with believable

preparation prices. Treatment 2 explores the impact of a zero audit rate in order to examine the

impact on cheating while varying the other parameters in a sensible way. Treatment 3

incorporates a dominated alternative (e.g., Option C dominates Option A) to test the attention of

the participants; Option C also is designed to make underreporting income more attractive and so

to test the degree to which respondents resist temptation and are willing to forgo earnings to be

honest. Options A and B in Treatment 4 are designed to further encourage underreporting

income by setting prices above tax savings while providing a zero audit rate, and participants

with preferences for honest reporting should choose either Option C or Option D. 14

IV. THEORETICAL CONSIDERATIONS

Note that we are not misleading participants by assuming that tax preparer credentials are irrelevant to tax

preparer performance. Although credentials do not affect tax preparer performance, the average tax savings and

audit rate directly affect the experimental outcome, and these aspects are affected by tax preparer credentials. This

assumption is consistent with real-world evidence. Indeed, there is substantial evidence that many tax preparers are

more likely to cheat than taxpayers would prefer, and there is no evidence that credentialed preparers are more or

less likely to cheat than those who are not credentialed.

Rational Choice Theory Predictions (I): The Taxpayer’s Choice of a Tax Preparer

Our initial theory proceeds from the assumption that the participant is rational and

maximizes the expected value of the net earnings during the experiment session. For small

laboratory payoffs, risk aversion is inconsistent with rational choice, and deviations from the

predictions of expected value likely indicate the presence of one or more of the many behavioral

anomalies that have been demonstrated in controlled experiments (Rabin, 2000). Since there

were only two levels of tax savings used to describe tax preparers in the experiment, we

simplified the theoretical analysis by considering only a high- and low-refund tax preparer

choice; this assumption can be easily extended to the general case where additional tax preparers

are added (e.g., four possible tax preparers in each treatment), who vary by the probability of

audit, their cost of preparation, and their credentials. Recall that each participant’s certain income

in each round was subject to automatic tax withholding and reporting, and their random income

in each round was self-reported, meaning that each subject chose what portion of the additional

random income to report in each round.

Using the following notation

𝑡𝑡 = 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟

𝑦𝑦 = 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖

𝑅𝑅𝑡𝑡 = 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖,

0 ≤ 𝑅𝑅𝑡𝑡 ≤ 5000

𝐷𝐷𝑡𝑡 = 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑,

0 ≤ 𝐷𝐷𝑡𝑡 ≤ 5

𝐶𝐶𝑡𝑡 = 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,

0 ≤ 𝐶𝐶𝑡𝑡 ≤ 5

𝐻𝐻

𝑉𝑉𝐶𝐶𝐶𝐶

= 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑜𝑜𝑜𝑜 ℎ𝑖𝑖𝑖𝑖ℎ − 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,

𝐿𝐿

𝑉𝑉𝐶𝐶𝐶𝐶

= 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑜𝑜𝑜𝑜 𝑙𝑙𝑙𝑙𝑙𝑙 − 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,

𝐻𝐻

𝑉𝑉𝐷𝐷𝐷𝐷

= 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑜𝑜𝑜𝑜 ℎ𝑖𝑖𝑖𝑖ℎ − 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑,

𝐿𝐿

𝑉𝑉𝐷𝐷𝐷𝐷

= 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑜𝑜𝑜𝑜 𝑙𝑙𝑙𝑙𝑙𝑙 − 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑒𝑒𝑟𝑟 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑,

𝜇𝜇 = 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = 2,

𝐻𝐻

𝑉𝑉𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶

≤ 𝑉𝑉𝐶𝐶𝐶𝐶

≤ 𝑉𝑉𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶

𝐿𝐿

𝑉𝑉𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶

≤ 𝑉𝑉𝐶𝐶𝐶𝐶

≤ 𝑉𝑉𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶

𝐻𝐻

𝑉𝑉𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷

≤ 𝑉𝑉𝐷𝐷𝐷𝐷

≤ 𝑉𝑉𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷

𝐿𝐿

𝑉𝑉𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷

≤ 𝑉𝑉𝐷𝐷𝐷𝐷

≤ 𝑉𝑉𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷

(tax owed + 100% penalty)

𝑃𝑃𝐻𝐻 = 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑜𝑜𝑜𝑜 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑓𝑓𝑓𝑓𝑓𝑓 𝐻𝐻

𝑃𝑃𝐿𝐿 = 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑜𝑜𝑜𝑜 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑓𝑓𝑓𝑓𝑓𝑓 𝐿𝐿

𝐵𝐵𝐻𝐻 = 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 (𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐) 𝑓𝑓𝑓𝑓𝑓𝑓 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝐻𝐻

𝐵𝐵𝐿𝐿 = 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 (𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐) 𝑓𝑓𝑓𝑓𝑓𝑓 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝐿𝐿

𝐼𝐼𝑡𝑡 = 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖,

0 ≤ 𝐼𝐼𝑡𝑡 ≤ 𝑅𝑅𝑡𝑡

𝑊𝑊 = 𝑤𝑤𝑤𝑤𝑤𝑤ℎℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟

𝜏𝜏 = 𝑡𝑡𝑡𝑡𝑡𝑡 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟,

the choice variables for each participant in each round of the experiment are how much of the

random income to declare (𝐼𝐼𝑡𝑡 ) and which tax preparer to use (𝐻𝐻 or 𝐿𝐿). Using a bar above a

variable to indicate the mean (or expected value 𝐸𝐸𝐸𝐸) of the variable and ignoring for now any

possible costs to obtain deductions or credits, the expected values of choosing the high- and lowrefund tax preparer in a given round are:

(1) 𝐸𝐸𝐸𝐸𝐻𝐻𝐻𝐻 = 𝑦𝑦 + 𝑅𝑅𝑡𝑡 − 𝜏𝜏(𝑦𝑦 + 𝐼𝐼𝑡𝑡 ) +[𝜏𝜏 ∗ 𝐷𝐷𝑡𝑡 ∗ 𝑉𝑉𝐷𝐷𝐻𝐻 + 𝐶𝐶𝑡𝑡 ∗ 𝑉𝑉𝐶𝐶𝐻𝐻 −𝑃𝑃𝐻𝐻 {𝜇𝜇 ∗ 𝜏𝜏(𝑅𝑅𝑡𝑡 − 𝐼𝐼𝑡𝑡 )}] − 𝐵𝐵𝐻𝐻

(2) 𝐸𝐸𝐸𝐸𝐿𝐿𝐿𝐿 = 𝑦𝑦 + 𝑅𝑅𝑡𝑡 − 𝜏𝜏(𝑦𝑦 + 𝐼𝐼𝑡𝑡 ) +[𝜏𝜏 ∗ 𝐷𝐷𝑡𝑡 ∗ 𝑉𝑉𝐷𝐷𝐿𝐿 + 𝐶𝐶𝑡𝑡 ∗ 𝑉𝑉𝐶𝐶𝐿𝐿 −𝑃𝑃𝐿𝐿 {𝜇𝜇 ∗ 𝜏𝜏(𝑅𝑅𝑡𝑡 − 𝐼𝐼𝑡𝑡 )}] − 𝐵𝐵𝐿𝐿

where equation (1) shows the expected value if a high-refund tax preparer is chosen and equation

(2) shows the expected value for low-refund tax preparer. The first order conditions for 𝐼𝐼𝑡𝑡 for

each choice of tax preparer are:

(3)

𝜕𝜕𝐸𝐸𝐸𝐸𝐻𝐻𝐻𝐻

(4)

𝜕𝜕𝐸𝐸𝐸𝐸𝐿𝐿𝐿𝐿

𝜕𝜕𝐼𝐼𝑡𝑡

>0→𝐼𝐼 =𝑅𝑅

= −𝜏𝜏 + 𝑃𝑃𝐻𝐻 ∗ 𝜇𝜇 ∗ 𝜏𝜏{≤0→𝐼𝐼𝑡𝑡𝑡𝑡= 0𝑡𝑡

> 0→ 𝐼𝐼 =𝑅𝑅

= −𝜏𝜏 + 𝑃𝑃𝐿𝐿 ∗ 𝜇𝜇 ∗ 𝜏𝜏{≤ 0→ 𝐼𝐼𝑡𝑡𝑡𝑡= 0𝑡𝑡

Substituting the decisions from equations (3) and (4) regarding 𝐼𝐼𝑡𝑡 into equations (1) and (2), the

high-refund tax preparer will be chosen if 𝐸𝐸𝐸𝐸𝐻𝐻𝐻𝐻 > 𝐸𝐸𝐸𝐸𝐿𝐿𝐿𝐿 , and the low-refund tax preparer will be

chosen if the reverse is true.

Using the following parameter values

𝐿𝐿

𝜏𝜏 = .25 𝑃𝑃𝐻𝐻 = .2 𝑃𝑃𝐿𝐿 = .05 𝜇𝜇 = 2

𝐻𝐻

𝐿𝐿

𝐻𝐻

𝑉𝑉𝐷𝐷 = 300 𝑉𝑉𝐷𝐷 = 700 𝑉𝑉𝐶𝐶 = 100 𝑉𝑉𝐶𝐶 = 200

𝐵𝐵𝐻𝐻 = 300 𝐵𝐵𝐿𝐿 = 200 ,

the predicted choices for, say, Treatment 1 can be found by substituting these parameters into the

first-order conditions to calculate the optimal level of random income reported for each choice of

tax preparer, which yields:

(3)’

𝜕𝜕𝐸𝐸𝐸𝐸𝐻𝐻𝐻𝐻

(4)’

𝜕𝜕𝐸𝐸𝐸𝐸𝐿𝐿𝐿𝐿

𝜕𝜕𝐼𝐼𝑡𝑡

= −.25 + .2 ∗ 2 ∗ .25 = −.15 < 0

= −.25 + .05 ∗ 2 ∗ .25 = −.225 < 0.

Equations (3)’ and (4)’ imply that random income should, rationally, never be reported in

Treatment 1 (or 𝐼𝐼𝑡𝑡 = 0). 15 Thus, for Treatment 1, the choice between tax preparers should

theoretically assume 𝐼𝐼𝑡𝑡 = 0, regardless of which tax preparer is chosen. To determine which tax

preparer will be chosen, we simply compare the expected value of each tax preparer using

equations (1) and (2). Since the first parts of (1) and (2) are identical, we need only to examine

the difference in the square bracketed portion of each equation; if [ ]𝐻𝐻𝐻𝐻 – [ ]𝐿𝐿𝐿𝐿 > (𝐵𝐵𝐻𝐻 − 𝐵𝐵𝐿𝐿 ), tax

preparer 𝐻𝐻 will be chosen, and otherwise, preparer 𝐿𝐿 will be chosen. Because [ ]𝐻𝐻𝐻𝐻 = .25 ∗ 𝐷𝐷𝑡𝑡 ∗

700 + 𝐶𝐶𝑡𝑡 ∗ 200 − .2 {2 ∗ .25 ∗ 𝑅𝑅𝑡𝑡 } and

[ ]𝐿𝐿𝐿𝐿 = .25 ∗ 𝐷𝐷𝑡𝑡 ∗ 300 + 𝐶𝐶𝑡𝑡 ∗ 100 − .05 {2 ∗ .25 ∗ 𝑅𝑅𝑡𝑡 }, then [ ]𝐻𝐻𝐻𝐻 − [ ]𝐿𝐿𝐿𝐿 = .25 ∗ 𝐷𝐷𝑡𝑡 ∗ 400 +

𝐶𝐶𝑡𝑡 ∗ 100 − .15 {2 ∗ .25 ∗ 𝑅𝑅𝑡𝑡 }. So, if [100(𝐷𝐷𝑡𝑡 + 𝐶𝐶𝑡𝑡 )] − .075 ∗ 𝑅𝑅𝑡𝑡 > (𝐵𝐵𝐻𝐻 − 𝐵𝐵𝐿𝐿 ), then the

prediction is to choose preparer 𝐻𝐻; if[100(𝐷𝐷𝑡𝑡 + 𝐶𝐶𝑡𝑡 )] − .075 ∗ 𝑅𝑅𝑡𝑡 < (𝐵𝐵𝐻𝐻 − 𝐵𝐵𝐿𝐿 ), then the

prediction is to choose preparer 𝐿𝐿.

The same result holds for other treatments because the highest audit probability in any treatment is 0.40.

Thus, more deductions and credits and a lower random income increase the likelihood

that the participant will choose tax preparer 𝐻𝐻. This prediction makes intuitive sense. Since it is

always rational to declare zero random income, a high draw on random income and a low draw

on deductions and credits imply that the participant should select preparer 𝐿𝐿. Such a participant

would have more to lose from being audited compared to an individual with a lower amount of

random income. Higher random income creates a larger incentive to avoid being audited, while

higher credits and deductions create a larger incentive to choose the tax preparer with higher tax

savings for these items. Note that the expected or “average” rational choice outcome (for 𝐷𝐷𝑡𝑡 +

𝐶𝐶𝑡𝑡 = 5 and Rt = 2,500) yields a tax preparer choice condition for Treatment 1 of: 100(5) −

.075(2,500) = 312.50 > 100. This implies that the average participant will choose 𝐻𝐻, the

high-audit, high-refund tax preparer. However, it is also possible for participants to choose 𝐿𝐿 in

the rational choice model. For example, if a participant draws 𝐷𝐷𝑡𝑡 + 𝐶𝐶𝑡𝑡 = 2 and 𝑅𝑅𝑡𝑡 = 4,500, the

optimal choice is 𝐿𝐿 because 100(2) − .075(4,500) = −137.50 < 100. Other treatments can

be analyzed in a similar manner.

Rational Choice Theory Predictions (II): The Taxpayer’s Compliance Decision

It is clear that the compliance decision (i.e., how much random income to report) and the

tax preparer decision (i.e., which preparer to choose) are interrelated. Maximizing behavior

dictates that maximum noncompliance is the optimal behavior; that is, participants should report

zero random income for any audit rate less than or equal to 50 percent. Since each tax preparer in

this experiment has an audit rate below 50 percent, the rational choice theory would predict full

noncompliance (i.e., zero random income reported), regardless of the tax preparer choice. Given

that complete noncompliance is optimal, the rational taxpayer would then choose the tax preparer

whose combination of expected refund (which is related to the average tax savings, but varies

depending on the individual’s specific number of credits and deductions) and audit rate

maximizes the expected value, depending on particular circumstances of random income,

deductions, and credits. Note, however, that much previous literature finds that taxpayers often

choose to pay taxes even when it violates rational choice theory (Alm, McClelland, and Schulze,

1992, 1999; Davis, Hecht, and Perkins, 2003; Young, 2020). Factors such as social norms,

fairness, loss aversion, and patriotism can contribute to a desire to pay taxes, even though the

likelihood of an audit is low. If an individual decides to voluntarily comply with tax regulations

for reasons other than avoiding penalties, this will affect the optimal tax preparer choice.

Consider a taxpayer who will be completely compliant regardless of audit penalties (or

lack thereof). The only cost incurred by an audit in the experimental setting is that any unpaid

taxes are collected along with a 100 percent penalty. Thus, for a taxpayer who reports all income

correctly, an audit is costless. For fully compliant taxpayers, the audit probability should have no

effect on their tax preparer decision. Furthermore, a high audit probability poses no threat to

these taxpayers, so they should always choose the preparer who will allow them to claim the

highest expected net refund, even if this tax preparer also has a high audit probability. Also, the

potential gains from noncompliance are limited by the amount of random income because this is

the only value that can be misreported to the tax preparer. A taxpayer who has a relatively high

amount of random income and a relatively low number of credits and deductions has an

incentive to misreport this income (i.e., claiming zero random income) and to choose a preparer

with a low audit rate, to decrease the probability of an audit penalty. Likewise, a taxpayer who

has a relatively low amount of random income and a relatively high number of credits and

deductions still has an incentive to misreport his or her random income but also has less to lose

from an audit and much more to gain from a tax preparer who will get a higher value for each

credit and deduction.

Overall, a taxpayer who intends to report all income should always choose the tax

preparer with the highest average net refund, but a taxpayer who intends to report zero or only a

portion of random income may find it beneficial to choose a tax preparer with lower average tax

savings with an associated lower audit probability. Also, the lower-audit-rate tax preparer is the

EV-maximizing choice for some amounts of noncompliance and some draws of random income,

deductions, and credits. However, full noncompliance is always EV-maximizing regardless of a

participant’s financial situation. Given that a taxpayer chooses full compliance, the tax preparer

with the higher audit rate and higher expected refund is the EV-maximizing choice.

Modifying the Rational Choice Model to Account for Psychological Factors

Results from pilot experiments at each location showed findings that were inconsistent

with the predictions of the rational choice model. Many participants reported all of their random

income, and a minority reported zero random income. Reporting all random income is consistent

with a linear objective function that produces a corner solution, but it is inconsistent with the

prediction that the rational decision is to declare zero random income. Also, many of the

participants chose the credentialed tax preparer, despite the fact that the credentialed preparer

had no impact on the outcome. Finally, very few individuals chose the high-refund tax preparers

in spite of the prediction that the high-refund tax preparer should be the dominant choice for

most draws of random income, credits, and deductions.

These findings suggest that there are specific factors relating to the tax context that may

affect behavior, outside of the rational choice model. For example, fear of being audited might

affect the taxpayer’s choice of tax preparer. Although audits only impose a financial cost in the

experimental setting, audits in the real world are likely to be stressful and time-consuming,

regardless of the individual’s actual tax compliance status. Thus, the desire to avoid an audit in

the real world might influence behavior in this experiment and lead to a behavioral anomaly due

to heuristic thinking. Also, individuals might be mistaken in their mental calculations regarding

the probability of an audit and the associated penalties, so that participants may fully report their

income yet still choose a low-refund, low-audit probability tax preparer. These participants were

clearly not maximizing their expected payout, regardless of their compliance choice.

Accordingly, we modify the rational choice model, allowing for the possibility of these

types of behavioral anomalies. Specifically, we incorporate variables for guilt from

underreporting income as well as fear of audit into the behavioral model, Additionally, we

include a variable indicating whether a tax preparer is credentialed to determine whether this

influenced taxpayer choices. 16

1. Guilt. The observed response pattern was consistent with the conjecture that many

participants did not underreport their random income because doing so would make them feel

dishonest. In most rounds, participants chose not to underreport their random income, even

though doing so was clearly the utility maximizing course of action when taking-into-account

only the financial incentives. Therefore, we infer that participants viewed underreporting as a

dishonest action and felt some guilt associated with it. We incorporate this conjecture into the

objective function by adding a term – 𝑔𝑔𝐿𝐿𝑡𝑡 , where 𝐿𝐿𝑡𝑡 = 0 if 𝑅𝑅𝑡𝑡 – 𝐼𝐼𝑡𝑡 = 0 and 𝐿𝐿𝑡𝑡 = 1, if 𝑅𝑅𝑡𝑡 – 𝐼𝐼𝑡𝑡 >

0, so 𝐿𝐿𝑡𝑡 denotes lying and 𝑔𝑔 denotes the psychological cost of guilt associated with

Many of these modifications are based on the work of Kahneman and Tversky (1979, 1984) and Tversky and

Kahneman (1974, 1981).

misrepresenting one’s income, expressed in terms of monetary value for the purposes of the

model. If participants do not have any aversion to misreporting their income other than the

financial cost of an audit, then the value of 𝑔𝑔 will be zero since there is no psychological cost.

2. Fear of Audit. The observed response pattern is also consistent with the conjecture that

many participants feared being audited even if they did not underreport income, despite the fact

that the audit in this experiment was private, automatic, and had no consequences in the absence

of cheating. This observed aversion to audits might be an irrational fear, possibly brought into

the laboratory by a heuristic developed from past experiences outside of the laboratory and

reinforced by a negative emotional response (Kahneman, Slovic, and Tversky, 1982). It can be

modeled by adding another term to the objective function equal to – 𝑓𝑓𝐴𝐴𝑖𝑖 , where 𝑓𝑓 denotes the

psychological cost of the participants’ fear of having any chance of being audited (expressed in

terms of monetary value), and 𝐴𝐴𝑖𝑖 is a dummy variable, such that 𝐴𝐴𝑖𝑖 = 0 if the probability of

audit associated with tax preparer 𝑖𝑖 is zero, 𝑃𝑃𝑖𝑖 = 0; and 𝐴𝐴𝑖𝑖 = 1 if the probability of audit

associated with tax preparer 𝑖𝑖 is positive, 𝑃𝑃𝑖𝑖 > 0. If participants do not have any emotional cost

associated with a positive audit probability, the coefficient 𝑓𝑓 will have a value of zero because all

financial costs have already been accounted for in the utility function.

3. Tax Preparer Credentials. It is also possible that, in spite of the lack of quantitative

support for choosing a credentialed tax preparer, this attribute affects choices, an example of

context (or framing) affecting decision-making. To incorporate this possibility, we add an

additional term 𝑉𝑉𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (𝐷𝐷𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,𝑖𝑖 ) to the objective function, where the first variable is the

psychological value of choosing a credentialed preparer and the second variable is a dummy

variable that takes on a value of one if credentialed and zero, if not.

Summary

In summary, we add the following psychological variables to the original list of rational

choice variables:

𝑔𝑔 = 𝑡𝑡ℎ𝑒𝑒 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝ℎ𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑜𝑜𝑜𝑜 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑤𝑤𝑤𝑤𝑤𝑤ℎ 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙

𝐿𝐿𝑡𝑡 = 𝑎𝑎𝑎𝑎 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑓𝑓𝑓𝑓𝑓𝑓 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢

𝑓𝑓 = 𝑡𝑡ℎ𝑒𝑒 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑜𝑜𝑜𝑜 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑑𝑑 𝑤𝑤𝑤𝑤𝑤𝑤ℎ 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎

𝐴𝐴𝑖𝑖 = 𝑎𝑎𝑎𝑎 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝

𝑉𝑉𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = 𝑡𝑡ℎ𝑒𝑒 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑜𝑜𝑜𝑜 𝑐𝑐ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑎𝑎 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑡𝑡𝑡𝑡𝑡𝑡 𝑝𝑝𝑝𝑝𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒

𝐷𝐷𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,𝑖𝑖 = 𝑎𝑎𝑎𝑎 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑓𝑓𝑓𝑓𝑓𝑓 𝑤𝑤ℎ𝑒𝑒𝑒𝑒ℎ𝑒𝑒𝑒𝑒 𝑡𝑡ℎ𝑒𝑒 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑖𝑖𝑖𝑖 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐

As discussed later, the random utility theoretical model that incorporates these psychological

on/off responses recognizes that there are actually eight possible choices that combine the choice

of declaring all random income or none with the choice of one of the four preparers, since the

choice of cheating on declared income is essentially a binary choice. 17 Including these

anomalies, the expected value or linear utility for a taxpayer choosing tax preparer 𝑖𝑖 and

reporting random income 𝐼𝐼𝑡𝑡 takes the form

(5)

𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = 𝑦𝑦 + 𝑅𝑅𝑡𝑡 − 𝜏𝜏(𝑦𝑦 + 𝐼𝐼𝑡𝑡 ) + �𝜏𝜏 ∗ 𝐷𝐷𝑡𝑡 ∗ 𝑉𝑉𝐷𝐷𝐻𝐻 + 𝐶𝐶𝑡𝑡 ∗ 𝑉𝑉𝐶𝐶𝐻𝐻 − 𝑃𝑃𝑖𝑖 {𝜇𝜇 ∗ 𝜏𝜏(𝑅𝑅𝑡𝑡 − 𝐼𝐼𝑡𝑡 )}�

+𝑉𝑉𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝐷𝐷𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,𝑖𝑖 − 𝑔𝑔(𝐿𝐿𝑡𝑡 ) − 𝑓𝑓( 𝐴𝐴𝑖𝑖 ) − 𝐵𝐵𝑖𝑖

or equivalently

(6)

𝐸𝐸𝐸𝐸𝑖𝑖𝑖𝑖 = [𝑦𝑦 + 𝑅𝑅𝑡𝑡 – 𝜏𝜏(𝑦𝑦 + 𝑅𝑅𝑡𝑡 )] + 𝜏𝜏(𝑅𝑅𝑡𝑡 − 𝐼𝐼𝑡𝑡 ) + �𝜏𝜏 ∗ 𝐷𝐷𝑡𝑡 ∗ 𝑉𝑉𝐷𝐷𝑖𝑖 + 𝐶𝐶𝑡𝑡 ∗ 𝑉𝑉𝐶𝐶𝑖𝑖 �

−𝜇𝜇 ∗ 𝜏𝜏(𝑃𝑃𝑖𝑖 (𝑅𝑅𝑡𝑡 − 𝐼𝐼𝑡𝑡 )) − 𝑔𝑔(𝐿𝐿𝑡𝑡 ) − 𝑓𝑓(𝐴𝐴𝑖𝑖 ) + 𝑉𝑉𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (𝐷𝐷𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,𝑖𝑖 ) − 𝐵𝐵𝑖𝑖

For further discussion, see Schulze and Wansink (2012). In a small percentage of rounds (14 percent), participants

declared only a portion of their random income. Our empirical analysis includes only those who reported 5 percent

or less (coded as noncompliant) and those who reported more than 95 percent (coded as compliant). One observation

was dropped because the participant received a random income draw of 0 experimental dollars for that round

(making the compliance choice moot), and 655 observations were dropped because the participant reported an

amount between 5 percent and 95 percent of random income.

We can therefore decompose the expected value of a given tax preparer and reporting decision

into the following relevant terms:

(a) [𝑦𝑦 + 𝑅𝑅𝑡𝑡 − 𝜏𝜏(𝑦𝑦 + 𝑅𝑅𝑡𝑡 )] = 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜

(b) 𝜏𝜏(𝑅𝑅𝑡𝑡 − 𝐼𝐼𝑡𝑡 ) = 𝑡𝑡𝑡𝑡𝑡𝑡 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑜𝑜𝑜𝑜 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖

(c) �𝜏𝜏 ∗ 𝐷𝐷𝑡𝑡 ∗ 𝑉𝑉𝐷𝐷𝑖𝑖 + 𝐶𝐶𝑡𝑡 ∗ 𝑉𝑉𝐶𝐶𝑖𝑖 � = 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑜𝑜𝑜𝑜 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑎𝑎 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝

(d) −𝜇𝜇 ∗ 𝜏𝜏(𝑃𝑃𝑖𝑖 (𝑅𝑅𝑡𝑡 − 𝐼𝐼𝑡𝑡 )) = 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑎𝑎𝑎𝑎 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢

(e) − 𝑔𝑔(𝐿𝐿𝑡𝑡 ) = 𝑡𝑡ℎ𝑒𝑒 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑜𝑜𝑜𝑜 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 𝑎𝑎𝑎𝑎𝑎𝑎𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑤𝑤𝑤𝑤𝑤𝑤ℎ 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢

(f) −𝑓𝑓(𝐴𝐴𝑖𝑖 ) = 𝑡𝑡ℎ𝑒𝑒 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑜𝑜𝑜𝑜 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑤𝑤𝑤𝑤𝑤𝑤ℎ 𝑡𝑡ℎ𝑒𝑒 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑜𝑜𝑜𝑜 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎

(g) 𝑉𝑉𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (𝐷𝐷𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,𝑖𝑖 ) = 𝑡𝑡ℎ𝑒𝑒 𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝑜𝑜𝑜𝑜 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑎𝑎 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑒𝑒𝑟𝑟 𝑤𝑤ℎ𝑜𝑜 𝑖𝑖𝑖𝑖 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐

(h) – 𝐵𝐵𝑖𝑖 = 𝑡𝑡ℎ𝑒𝑒 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑜𝑜𝑜𝑜 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑖𝑖

In the conditional logit analysis that follows, the probability of choosing each tax

preparer is based on the attributes defined above. There are two ways to organize the explanatory

variables. First, terms 𝑏𝑏, 𝑐𝑐, 𝑑𝑑, and ℎ can be combined to create a net tax savings variable, which

corresponds to the rational choice economic incentives. The behavioral variables 𝑒𝑒, 𝑓𝑓, and 𝑔𝑔 can

then also be included separately to test rational versus emotional factors in the choice of a tax

preparer. Second, psychologists argue that some rational choice factors might be more prominent

than others, so the factors 𝑏𝑏, 𝑐𝑐, 𝑑𝑑, and ℎ can also be incorporated individually to see if they have

statistically different coefficients, reflecting the prominence of some factors over others in the

decision process. We employ both approaches in our analysis.

It should be noted that the expected tax savings for each preparer in a round are equal to

(175𝐷𝐷𝑡𝑡 + 200𝐶𝐶𝑡𝑡 ) for the high-refund tax preparer and (75𝐷𝐷𝑡𝑡 + 100𝐶𝐶𝑡𝑡 ) for the low-refund tax

preparer. These formulas are kept constant across treatments. This expression is a close

approximation to the average refund amount that the IRS could calculate based on its records of

returns filed by each tax preparer. Presenting such a figure to taxpayers would likely have a

similar effect in the real world as in the experiment; that is, a taxpayer’s eligibility for certain

credits and deductions will affect the tax outcome in ways that may differ from the average

refund amount. The expected penalty—a function of the probability of audit multiplied by

penalty multiplied by random income—takes on values of 𝑃𝑃𝑖𝑖 ∗ 2 ∗ .25(𝑅𝑅𝑡𝑡 − 𝐼𝐼𝑡𝑡 ), and this

formula is also consistent across treatments (i.e., participants in all treatments have the same tax

rate and the same penalty for unpaid taxes).

Hypotheses

Based on results from previous literature, our focus group findings, and our theoretical

models, we suggest several main testable hypotheses:

Hypothesis 1: Individuals will be willing to “overpay” to avoid being audited, and they

will tend to choose preparers with a zero or low probability of audit even when it results in a

strictly higher cost to them.

Hypothesis 2: Most individuals will report all or most of their random income even

though it is not rational to do so.

Hypothesis 3: Individuals are influenced in their choice of a tax preparer by the tax

preparer’s credentials.

The hypothesis of most interest is Hypothesis 1. The common perception of tax preparer choice

is that taxpayers want a tax preparer who will utilize every possible tax reduction strategy, even

if it is somewhat questionable. An alternative perception is that people want to do everything

correctly in order to avoid an audit. Hypothesis 2 is also of interest, and it forms the basis for the

role of guilt and fear in our analysis. Hypothesis 3 tests the role of tax preparer credentials in tax

preparer choice, as suggested by the results of our focus groups.

V. EXPERIMENTAL RESULTS

Descriptive Statistics

Table 2 gives the average values of several participant characteristics, including

experiment earnings and reported earnings, demographic characteristics (e.g., gender, age, and

race), and information about past tax experiences.

Main Results

In all treatments, participants typically reported all of their random income, and the

majority also chose a tax preparer with a low probability of audit, even if that tax preparer had a

lower level of tax savings or a higher price. In spite of the fact that these participants did not

misreport their income and would not be subject to any penalties if audited, they showed an

overwhelming preference for tax preparers with low audit rates. In other words, even though

audits are completely painless, private, automatic, and instantaneous in the computerized lab

experiment, participants were still strongly motivated to avoid audits. This phenomenon is

inconsistent with the predictions of the rational choice model but in line with Hypothesis 1. In

particular, most participants reported all their random income, whereas only a few reported zero

random income, which is consistent with a linear objective function that produces a corner

solution but inconsistent with the prediction that the maximizing solution is to declare zero

random income. Finally, relatively few individuals chose the high-refund tax preparer, even

though the rational choice prediction is that the high-refund tax preparer should be the dominant

choice contingent on reporting all income.

The initial data collection sessions use the parameter values for the first treatment. These

data allow a check on the prediction that participants should choose either to report all of their

random income or none of their random income (e.g., a corner solution), as well the prediction

that all participants should declare zero income. Figure 1 shows the fraction of random income

reported in Treatment 1 (or the “compliance percentage”). These results are inconsistent with the

rational choice prediction that no random income should be reported. Indeed, a majority of

participants reported all of their random income in Treatment 1. Similarly, subjects in

Treatments 2-4 also showed a tendency toward reporting all random income. Figure 2 shows the

fraction of random income reported across all four treatments, again supporting a corner solution

and suggesting a strong inclination to report all income and to not cheat on their reported

income. All of these results are consistent with past studies showing that participants in tax

experiments comply more than rational choice would suggest (Alm, McClelland, and Schulze,

1992, 1999), as suggested by Hypothesis 2.

These results are also consistent with a corner solution that characterizes linearity in the

participants’ objective functions with respect to declared income. The vast majority of subjects

chose either to report 100 percent of their random income or to report 0 percent of their random

income. 18 This is consistent with Rabin’s (2000) arguments that risk aversion as proposed in

expected utility theory is impossible for the small stakes used in laboratory experiments.

Therefore, for the purposes of analyzing the choice between tax preparers, we model the decision

problem as a choice between eight different options in all treatments of the experiment: the

choice of tax preparer and, for each tax preparer choice, whether to report or not to report their

random income for that round. Although only four options were explicitly displayed on their

Participants reported either less than 5 percent or more than 95 percent of their random income in 85 percent of

decisions. In the remaining decisions, the percentages reported were nearly uniformly distributed among possible

percentages. Thus, these decisions were treated as random and were dropped for the purposes of our econometric

analysis.

computer screen, participants were actually choosing among the eight options shown in Table 3.

The eight different options in the econometric analysis are generated by creating an indicator

variable for the reporting decision and combining this with the preparer choice. Participants who

reported 95 percent or more of their random income in a given period were coded as reporting

100 percent, whereas participants who reported 5 percent or less of their random income in a

given period were coded as reporting 0 percent. All other choices were dropped from the analysis

because the declared income for these individuals was essentially uniformly distributed between

5 percent and 95 percent, suggesting that these individuals were randomly selecting the amount

to declare.

The distributions of options that were chosen in each treatment are shown in Figures 3 to

6, which clearly demonstrate that subjects respond to the different parameter values in each

treatment specification. An interesting pattern that emerged is that, in all treatments except

Treatment 4, most participants chose to truthfully report random income regardless of the chosen

tax preparer. Even in Treatments 3 and 4, which incorporated tax preparer options with a zero

probability audit rate, many participants who chose the zero probability option still reported all

of their random income. If a subject only maximizes income from the experiment, he or she

should never report any random income if he or she knows there is zero probability of being

audited. 19 This further confirms the argument we made in the theoretical section; that is, people

consider other factors in addition to monetary payoffs when they choose a tax preparer and

decide the amount of random income to report.

This argument is a “lower bound” on rationality defined in the rational choice theory section. Based on our

theoretical model, taxpayers should not report any of their random income if the expected value of not reporting it

exceeds the expected value of reporting it.

Treatment 3 is of particular interest for two reasons. First, participants were apparently

paying careful attention to the options since Option C was deliberately designed to dominate

Option A, and only 3 percent of the choices were for Option A while around 74 percent were for

Option C. Second, the rational choice incentives strongly support choice Option C0, or declaring

nothing, since this choice had zero probability of audit. However, slightly more than half of the

participants who chose Option C chose C1 rather than C0, deciding to report all their income.

This result suggests that the emotion of guilt or the desire to be honest almost perfectly offset the

financial incentive to cheat. In this situation, the random utility model would predict that roughly

half of the participants choosing Option C would choose C0 and half would choose C1.

Estimation Results

To investigate how different tax preparer characteristics affected the subjects’ tax

preparer choices, we conducted a regression analysis of tax preparer choice, using the

conditional logit model. 20 The conditional logit model allows us to incorporate characteristics of

the choice alternatives instead of or in addition to the characteristics of the individual making the

choice. This differs from a multinomial logit model, which only considers the characteristics of

the individual making the choice. The conditional logit model estimates the probability 𝑃𝑃𝑖𝑖𝑖𝑖 that

an individual 𝑖𝑖 chooses tax preparer 𝑗𝑗 (or in this case the combined preparer-compliance options)

as a function of the characteristics of the individual, represented by 𝑋𝑋𝑖𝑖 , and the characteristics of

the preparer, represented by 𝑍𝑍𝑖𝑖𝑖𝑖 , or:

For a more complete discussion of the conditional logit model and examples of its application, see Duncan and

Hoffman (1988).

𝐽𝐽

𝑃𝑃𝑖𝑖𝑖𝑖 = � 𝑒𝑒𝑒𝑒𝑒𝑒�𝑋𝑋𝑖𝑖 𝛽𝛽𝑗𝑗 + 𝑍𝑍𝑖𝑖𝑖𝑖 𝛼𝛼�/𝑒𝑒𝑒𝑒𝑒𝑒(𝑋𝑋𝑖𝑖 𝛽𝛽𝑘𝑘 + 𝑍𝑍𝑖𝑖𝑖𝑖 𝛼𝛼)

𝑘𝑘=1

It is important to note that, by using a conditional logit model, we specify our econometric model

as if there were one representative agent from whom we could make repeated observations.

Individual heterogeneity is, of course, a concern for all discrete choice modeling. However, the

sampling across different treatments is not relevant to this assumption. The purpose of the

estimation is not to detect any treatment effect but to estimate how people responded to different

characteristics of the tax preparer options.

We estimated several conditional logit models for the choice of tax preparer. The first

model is directly derived from the theoretical model in which we assume subjects considered the

total monetary benefit of each choice and the emotional costs and benefits associated with the

behavioral variables. Thus, in the first regression, we included one variable named Net Saving,

which is the sum of all benefits from the chosen tax preparer net of the cost (price) of the tax

preparer. We also included three dummy variables (Fear, Guilt, and Credential) to study whether

and how people responded to nonmonetary emotional factors. Fear takes the value of 1

whenever the probability of being audited is any positive amount; otherwise, it is 0. Guilt takes

the value of 1 whenever a subject chooses to underreport the random income (by reporting less

than 5 percent of it); otherwise, it is 0. One can think of Guilt as an additional cost incurred for

choosing noncompliance, and, although it is a psychological cost, it is represented by a monetary

equivalent. This cost is only incurred when the subject is noncompliant, regardless of the tax

preparer chosen. Credential takes the value of 1 if the preparer was certified as passing a

background check and as passing a CPA examination; otherwise, it is 0.

These regression results are reported in Column 1 of Table 4. 21 Signs of all coefficients

are intuitive and expected. Increasing the Net Saving of a tax preparer increases the likelihood

that a taxpayer would choose the option. The coefficient on Fear is negative and significant,

showing that people were less likely to choose a tax preparer with a positive audit probability

because they were averse to being audited. Guilt has a similar effect and decreases the likelihood

that taxpayers would underreport their random income. Having Credential increases the

likelihood of a particular tax preparer being chosen. All of these coefficients are significant at the

0.1 percent level. The last result is consistent with Hypothesis 3 on the role of credentials in the

choice of a tax preparer.

To account for possible heterogeneity in the effects between men and women, we

included an interaction term for gender (or a dummy variable equal to 1 if the subject is Female

and 0 otherwise) with both the Fear and Guilt variables (Column 2). 22 The coefficients and their

significance levels are quite similar to the first regression, but we can infer from the second

regression that females are more influenced by Guilt than males, with the effect of Guilt on

women negative and significant at the 1 percent level. The effect of Fear on women’s probability

of choosing a given preparer is negative but not statistically significant.

In additional regressions, we deviated from our theoretical model by treating monetary

benefits and costs associated with each tax preparer as characteristics that might be given

Given the large number of variables included in our regression analysis, one may worry about the possible issue

of multicollinearity. Because the consequence of multicollinearity is usually inflated standard errors, we are not

particularly concerned because most of our estimates from the conditional logit models have high precision levels

(small standard errors and highly significant coefficients). Nonetheless, we provide the variance inflation factor

(VIF) for all estimated standard errors and the regression models in the brackets below the standard errors in all

tables of estimation results. Note that the VIF is meant to detect possible multicollinearity issues in linear models, so

the reported VIF numbers are indicative of little multicollinearity.

Female was not available for eight participants, so observations associated with these participants were dropped

for the second set of regressions.

different weights by taxpayers, even though they are all directly comparable financial gains or

losses. There are then four different explanatory variables: Expected Tax Saving, Underreporting

Saving, Expected Penalty, and Price. Theoretically, subjects could have calculated all of these

values using the information that was given in their experimental instructions. We no longer use

the average savings presented in the table of tax preparer characteristics shown to participants

because subjects still need to calculate their expected return based on their own credits and

deductions. 23

Regression results with the disaggregated elements of economic expected value included

are shown in Columns 3 and 4 of Table 4. The effects of Fear of audit and Guilt from cheating,

as well as their corresponding gender heterogeneous effects, are largely unaffected. An F-test for

the joint hypothesis that the four disaggregated explanatory variables share the same coefficient

is rejected (p = 0.00), which suggests that participants treat different types of tax savings and

costs differently, as implied by the possibility of prominence effects and of mental accounting

(or thinking of different types of monetary gain or loss as having different value) (Thaler, 1999).

For the monetary incentive variables (Expected Tax Saving, Underreporting Saving,

Expected Penalty, Price), all have the expected sign, and all are highly significant, with the

largest impact coming from Price. Thus, using one net saving variable to represent all monetary

incentives might be inappropriate. This also suggests that individuals engage in mental

accounting. Again, these results are consistent with Hypothesis 3.

To explore the effects of time and repeated decisions, we then divided the data into

decisions made in periods 1–5 and decisions made in periods 1–10. These results are shown in

One can also think of this as interacting personal financial characteristics (e.g., random income, credits, and

deductions) with tax preparer types. It is only in this case that the interaction term is what matters; that is, it would

be meaningless for one to choose a preparer without considering his or her own tax situation.

Table 5, where Columns 1 and 3 show regression results for decisions in periods 1–5, and

Columns 2 and 4 show regression results for decisions in periods 6–10. 24 These results show that

the coefficients on Fear, Guilt, and Credential all decrease substantially between the beginning

and end of the experiment, which suggests that psychological influences on decision-making lose

potency over time and repeated exposure.

In additional regressions, we explored the relationship between Fear and Guilt. Although

we already incorporated the expected penalty as part of the net savings calculation, it might be

that these emotional responses interact with each other. Participants might have felt more afraid

of an audit if they had failed to report income, in which case they knew that an audit would have

had a bad outcome. Alternatively, they might have felt guiltier if there was a positive probability

of an audit because it meant their guilt could be revealed, albeit by a computerized audit. In

short, Guilt and Fear may magnify each other and make each one more salient than they would

be on their own. These results are shown in Table 6. Indeed, as shown there, including an

interaction term shows that Guilt and Fear together influence the tax preparer decision in

addition to the effect that each one has separately.

VI. CONCLUSIONS: IMPLICATIONS FOR POLICY

Our results suggest that standard monetary incentives influence an individual’s choice of

a tax preparer and the individual’s choice of reported income, as suggested by rational choice

theory (e.g., reason). Even so, we also find many results that are consistent with the important

The number of observations differs between rounds 1–5 and rounds 6–10 because we dropped observations in

which the participant reported more than 5 percent but less than 95 percent of random income. Such observations

accounted for less than 15 percent of total decisions. Since they are not distributed equally across rounds, there were

an unequal number of included observations for each round.

role of psychological factors (e.g., emotion). For example, individuals tend to report either all or

none of their random income, often reporting all of their income even when the probability of an

audit is low or zero. Individuals choosing a tax preparer strongly prefer a preparer who will help

them avoid being audited, which holds even when the cost of the tax preparer is high and when

there is a low chance of an audit and a low penalty even if there is an audit. In fact, individuals

often choose a tax preparer who is competent and qualified, even if it comes at a higher cost. The

presence of a positive audit probability has a negative effect on the probability of a preparer

being chosen, an effect that is in addition to the expected penalty resulting from an audit.

Individuals are especially eager to avoid any kind of an audit, even when an audit is unlikely or

nonpunitive: the fear of being audited and the guilt associated with failing to report income are

both strong motivators in tax preparer and compliance decisions, and these psychological factors

actually seem to dominate rational decision-making in tax preparer and compliance choices.

Overall, we conclude that the taxpayers prefer to fully report their income and to avoid being

audited, and these preferences appear to play a large role in the choice of a tax preparer.

Of course, one must remember that our results stem largely from laboratory experiments.

The lab seems particularly well-suited for the study of many aspects of compliance. In particular,

the lab is able to generate direct measures of evasion under different settings in which there is

control over extraneous influences, it is relatively inexpensive, its results can be easily replicated,

and it has a high degree of “internal validity” (or identification of “cause and effect”). However,

laboratory experiments are sometimes viewed with suspicion. The most common criticism is that

the student subjects typically used in experiments may not be representative of taxpayers. As a

result, there is a concern that experimental results on policy innovations that rely upon student

subjects cannot generalize to the population; that is, the “external validity” of laboratory

experiments is sometimes questioned. 25 Given that our subject pool consisted of adults with

previous experiences in paying taxes and often using tax preparers, we believe that the concern

about subject pool effects is of lesser importance in our study.

Our results have several practical implications. From the prospective of the tax

administration, one potentially useful implication of this research is to suggest opportunities for

the IRS to provide additional information to aid taxpayers in their selection of a tax preparer.

This work can help inform the IRS in its consideration of balancing of the taxpayer’s desire for

information on preparer quality and the need to protect tax preparer privacy.

From the perspective of tax preparers, tax preparers would do well to advertise their strict

compliance standards and low average audit rates when marketing to new clients. Our results

show that the presence of an audit risk and the aversion to underreporting income are both strong

motivators in the choice of a tax preparer and in the compliance decision. In particular, many

participants in our study were willing to forgo monetary benefit to avoid an audit, even though

they correctly and fully reported their tax liability. Of course, audits in the real world are not

costless, as they were in the experiment; even individuals who have correctly reported all of their

taxes must still pay the cost of time and effort involved in complying with the auditor’s requests.

Even so, for individuals who are inclined toward compliance, information about a tax preparer’s

performance would help them choose a tax preparer who is most likely to follow the tax code

properly and help them minimize the probability of being audited, and drawing attention to the

fact that a tax preparer’s performance has an impact on audit probability will encourage

See Levitt and List (2007) for a general critique of laboratory experiments. For robust responses to this critique,

see especially Falk and Heckman (2009) and many of the papers in the volume edited by Frechette and Schotter

(2015). Also, see Alm, Bloomquist, and McKee (2015) for specific evidence on the external validity of tax

compliance experiments, who find that student and non-student behaviors are similar; see Choo, Fonseca, and Myles

(2016) for an alternative view on student versus non-student behaviors.

taxpayers to be more diligent in their choice of tax preparer. From the perspective of

policymakers, the IRS can certainly encourage the provision of this type of information.

Indeed, our experimental results suggest that individuals are willing to pay a premium for

a tax preparer with credentials (i.e., had passed an IRS background check and was a Certified

Public Accountant). This is particularly noteworthy because these credentials had no bearing on

financial outcomes in the experiment. The fact that participants are willing to pay more for a tax

preparer with credentials underscores the findings from the focus group that credentials are an

important characteristic in the tax preparer choice.

In fact, if it is the case that tax preparers generally facilitate the filing of noncompliant tax

returns, then our results suggest that this is not due to taxpayer demand. Future research by the

IRS and others should explore the reasons for noncompliance on the part of the tax preparer and

examine the interaction between taxpayer compliance and tax preparer compliance. The tax

preparer faces different incentives than the taxpayer and might experience guilt and fear

differently when performing a service on behalf of someone else. A secondary line of inquiry

could investigate whether noncompliance is related to cognitive load (i.e., is a result of mistakes

on the part of the taxpayer or tax preparer) or intentional misreporting.

REFERENCES

Allingham, Michael G., and Agnar Sandmo. 1972. Income tax evasion: A theoretical analysis.

Journal of Public Economics 1 (3-4): 323-338.

Alm, James. 2012. Measuring, explaining, and controlling tax evasion: Lessons from theory,

field studies, and experiments. International Tax and Public Finance 19 (1): 54-77.

Alm, James. 2019. What motivates tax compliance? Journal of Economic Surveys 33 (2): 353388.

Alm, James, Kim M. Bloomquist, and Michael McKee. 2015. On the external validity of

laboratory tax compliance experiments. Economic Inquiry 53 (2): 1170-1186.

Alm, James, Mark B. Cronshaw, and Michael McKee. 1993. Tax compliance with

endogenous audit selection rules. Kyklos 46 (1): 27-45.

Alm, James, and Matthias Kasper. 2022.

Alm, James and Michael McKee. 2004. Tax compliance as a coordination game. Journal

of Economic Behavior and Organization 54 (3): 297-312.

Alm, James, Gary H. McClelland, and William D. Schulze. 1992. Why do people pay taxes?

Journal of Public Economics 48 (1): 21-38.

Alm, James, Gary H. McClelland, and William D. Schulze. 1999. Changing the social norm of

compliance by voting. Kyklos 52 (2): 141-171.

Andreoni, James, Brian Erard, and Jonathan Feinstein. 1998. Tax compliance. The Journal of

Economic Literature 36 (2): 818-860.

Ashley, Terry, and Mark A. Segal. 1997. Paid tax preparer determinants extended and

reexamined. Public Finance Review 25 (3): 267-284.

Austin, Chelsea Rae, Donna D. Bobek, and Ethan G. La Mothe. 2019. The effect of temporary

changes and expectations on individuals’ decisions: Evidence from a tax compliance

setting. The Accounting Review 95 (3): 33-58.

Ayers, Benjamin C., Steven J. Kachelmeier, and John R. Robinson. 1999. Why do people give

interest-free loans to the government? An experimental study of interim tax payments.

The Journal of the American Taxation Association 21 (2): 55-74.

Battaglini, Marco, Luigi Guiso, Chiara Lacava, and Eleonora Patacchini. 2020. Tax professionals

and tax evasion. NBER Working Paper 25745. Cambridge, MA: National Bureau of

Economic Research.

Bechwati, Nada Nasr. 2011. Willingness to pay for professional services. Journal of Product &

Brand Management 20 (1): 75-83.

Becker, Gary S. 1968. Crime and punishment – An economic approach. The Journal of Political

Economy 76 (2): 169-217.

Beck, Paul J., Jon S. Davis, and Woon-Oh Jung. 1991. Experimental evidence on taxpayer

reporting behavior. The Accounting Review 66 (3): 535-558.

Bobek, Donna D., Richard C. Hatfield, and Kristin Wentzel. 2007. An investigation of why

taxpayers prefer refunds: A theory of planned behavior approach. The Journal of the

American Taxation Association 29 (1): 93-111.

Cason, Timothy N., and Lata Gangadharan. 2006. An experimental study of compliance and

leverage in auditing and regulatory enforcement. Economic Inquiry 44 (2): 352-366.

Christian, Charles W., Sanjay Gupta, and Suming Lin. 1993. Determinants of tax preparer usage:

Evidence from panel data. National Tax Journal 46 (4): 487-503.

Choo, C. Y. Lawrence, Miguel A. Fonseca, and Gareth D. Myles. 2016. Do students behave like

real taxpayers in the lab? Evidence from a real effort tax compliance experiment. Journal

of Economic Behavior & Organization 124: 102-114.

Christian, Charles W., Sanjay Gupta, Gar J. Weber, and Eugene Willis. 1994. The relationship

between the use of tax preparers and taxpayers’ prepayment position. The Journal of the

American Taxation Association 16 (1): 17-40.

Clark, Jeremy, Lana Friesen, and Andrew Muller. 2004. The good, the bad, and the regulator: An

experimental test of two conditional audit schemes. Economic Inquiry 42 (1): 69-87.

Cloyd, C. Bryan, and Brian C. Spilker. 1999. The influence of client preferences on tax

professionals’ search for judicial precedents, subsequent judgments, and

recommendations. The Accounting Review 74 (3): 299-322.

Congdon, William J., Jeffrey R. Kling, and Sendhil Mullainathan. 2011. Policy and Choice –

Public Finance through the Lens of Behavioral Economics. Washington, D.C.: The

Brookings Institution Press.

Copeland, Phyllis V., and Andrew D. Cuccia. 2002. Multiple determinants of framing referents

in tax reporting and compliance. Organizational Behavior and Human Decision

Processes 88 (1): 499-526.

Cummings, Benjamin F., and Russell N. James III. 2014. Factors associated with getting and

dropping financial advisors among older adults: Evidence from longitudinal data. Journal

of Financial Counseling and Planning 25 (2): 129-147.

Davis, Jon S., Gary Hecht, and Jon D. Perkins. 2003. Social behaviors, enforcement, and tax

compliance dynamics. The Accounting Review 78 (1): 39-69.

Dubin, Jeffrey A., Michael J. Graetz, Michael A. Udell, and Louis L. Wilde. 1992. The demand

for tax return preparation services. The Review of Economics and Statistics 74 (1): 75-82.

Duncan, Greg, and Saul Hoffman. 1988. Multinomial and conditional logit discrete-choice

models in demography. Demography 25 (3): 415-427.

Dusenbury, Richard. 1994. The effect of prepayment position on individual taxpayers’

preferences for risky tax-filing options. The Journal of the American Taxation

Association 16 (1): 1-16.

Erard, Brian. 1993. Taxation with representation: An analysis of the role of tax practitioners in

tax compliance. Journal of Public Economics 52 (1): 163-197.

Erard, Brian. 1997. Self-selection with measurement errors: A microeconometric analysis of the

decision to seek tax assistance and its implications for fax compliance. Journal of

Econometrics 81 (2): 319-356.

Falk, Armin, and James J. Heckman. 2009. Lab experiments are a major source of knowledge in

the social sciences. Science 326 (5952): 535-538.

Fischbacher, Urs. 2007. z-Tree: Zurich toolbox for ready-made economic experiments.

Experimental Economics 10 (2): 171-178.

Fischbacher, Urs, and Franziska Föllmi‐Heusi. 2013. Lies in disguise: An experimental study on

cheating. Journal of the European Economic Association 11 (3): 525-547.

Fleischman, Gary M., and Teresa Stephenson. 2012. Client variables associated with four key

determinants of demand for tax preparer services: An exploratory study. Accounting

Horizons 26 (3): 417-437.

Frechette, Guillaume R., and Andrew Schotter (eds.) 2015. The Methods of Modern

Experimental Economics. New York, NY: Oxford University Press.

Friedland, Nehemiah, Shlomo Maital, and Aryeh Rutenberg. 1978. A simulation study of income

tax evasion. Journal of Public Economics 10 (1): 107-116.

Frischmann, Peter J., and Edward W. Frees. 1999. Demand for services: Determinants of tax

preparation fees. The Journal of the American Taxation Association 21 (1): 1-21.

Giacobbe, Ralph W., and Madhav N. Segal. 1996. Correlates of professional services market(s):

An analysis of customer characteristics. Journal of Professional Services Marketing 13

(2): 17-32.

Gilpatric, Scott M., Christian A. Vossler, and Michael McKee. 2011. Regulatory enforcement

with competitive endogenous audit mechanisms. RAND Journal of Economics

42 (2): 292-312.

Grable, John E., and So-hyun Joo. 2001. A further examination of financial help-seeking

behavior. Financial Counseling and Planning 12 (1): 55-74.

Guyton, John L., Adam K. Korobow, Peter S. Lee, and Eric J. Toder. 2005. The effects of tax

software and paid preparers on compliance costs. National Tax Journal 3 (4): 439-448.

Hanna, Sherman D. 2011. The demand for financial planning services. Journal of Personal

Finance 10 (1): 36-62.

Hashimzade, Nigar, Gareth D. Myles, and Binh Tran-Nam. 2013. Applications of behavioural

economics to tax evasion. Journal of Economic Surveys 27 (5): 941–977.

Jackson, Scott B., and Richard C. Hatfield. 2005. A note on the relation between frames,

perceptions, and taxpayer behavior. Contemporary Accounting Research 22 (1): 145-164.

Jackson, Scott B., Paul A. Shoemaker, John A Barrick, and F. Greg Burton. 2005. Taxpayers'

prepayment positions and tax return preparation fees. Contemporary Accounting

Research 22 (2): 409-447.

Jackson, Scott B., and Richard A. White. 2008. The effect of tax refunds on taxpayers'

willingness to pay higher tax return preparation fees. Research in Accounting Regulation

20: 63-88.

Kahneman, Daniel. 2011. Thinking, Fast and Slow. New York, NY: Farrar, Straus and Giroux.

Kahneman, Daniel, and Amos Tversky. 1979. Prospect theory: An analysis of decision under

risk. Econometrica 47 (2): 263-291.

Kahneman, Daniel, and Amos Tversky. 1984. Choices, values, and frames. American

Psychologist 39 (4): 341-350.

Kahneman, Daniel, Paul Slovic, and Amos Tversky. 1982. Judgment Under Uncertainty:

Heuristics and Biases. Cambridge, MA: Cambridge University Press.

Kaplan, Steven E., Philip M.J. Reckers, Stephen G. West, and James C. Boyd. 1988. An

examination of tax reporting recommendations of professional tax preparers. Journal of

Economic Psychology 9 (4): 427-443.

Kasper, Matthias, and James Alm. 2022. Audits, audit effectiveness, and post-audit tax

compliance. Journal of Economic Behavior & Organization 195: 87-102.

Kirchler, Erich. 2007. The Economic Psychology of Tax Behaviour. Cambridge, UK: Cambridge

University Press.

Klepper, Steven, and Daniel Nagin. 1991. The role of tax practitioners in tax compliance. Policy

Sciences 22: 167-192.

Klepper, Steven, Mark Mazur, and Daniel Nagin. 1991. Expert intermediaries and legal

compliance: The case of tax preparers. Journal of Law and Economics 34 (1): 205-299.

Long, James E., and Steven B. Caudill. 1987. The usage and benefits of paid tax return

preparation. National Tax Journal 40 (1): 35-46.

Murphy, Kristina. 2004. Aggressive tax planning: Differentiating those playing the game from

those who don't. Journal of Economic Psychology 25 (3): 307-329.

Newberry, Kaye J., Philip M. J. Reckers, and Robert W. Wyndelts. 1993. An examination of tax

practitioner decisions: The role of preparer sanctions and framing effects associated with

client condition. Journal of Economic Psychology 14 (2): 439-452.

Rabin, Matthew. 1998. Psychology and economics. The Journal of Economic Literature 36 (1):

11–46.

Rabin, Matthew. 2000. Risk aversion and expected-utility theory: A calibration theorem.

Econometrica 68 (5): 1281-1292.

Reinganum, Jennifer F., and Louis L. Wilde. 1991. Equilibrium enforcement and compliance in

the presence of tax practitioners. Journal of Law, Economics, and Organization 7 (1):

163-181.

Robb, Cliff A., Patryk Babiarz, and Ann Woodyard. 2012. The demand for financial

professionals’ advice: The role of financial knowledge, satisfaction, and confidence.

Financial Services Review 21: 291-305.

Sandmo, Agnar. 2012. An evasive topic: Theorizing about the hidden economy. International Tax

and Public Finance 19 (1): 5-24.

Schulze, William D., and Brian Wansink. 2012. Toxics, Toyotas, and terrorism: The behavioral

economics of fear and stigma. Risk Analysis 32 (4): 678-694.

Scotchmer, Suzanne. 1989. The effect of tax advisors on tax compliance. In Jeffrey A. Roth and

John T. Scholz (eds.), Taxpayer Compliance – Volume 2: Social Science Perspectives.

Philadelphia, PA: University of Pennsylvania Press, 182-199.

Shavell, Steven. 1988. Legal advice about contemplated acts: The decision to obtain advice, its

social desirability, and protection of confidentially. Journal of Legal Studies 17 (1): 123150.

Slemrod, Joel. 2019. Tax compliance and enforcement. The Journal of Economic Literature 57

(4): 904-954.

Spicer, Michael W., and Lee A. Becker. 1980. Fiscal inequity and tax compliance: An

experimental approach. National Tax Journal 33 (2): 171-175.

Stephenson, Teresa. 2010. Measuring taxpayers’ motivation to hire tax preparers: The

development of a four-construct scale. Advances in Taxation 19 (1): 95-121.

Tan, Lin Mei. 1999. Taxpayers' preference for type of advice from tax practitioners: A

preliminary examination. Journal of Economic Psychology 20 (4): 431-447.

Thaler, Richard H., 1999. Mental accounting matters. Journal of Behavioral Decision Making 12

(3): 183-206.

Torgler, Benno. 2007. Tax Compliance and Tax Morale: A Theoretical and Empirical Analysis.

Cheltenham, UK: Edward Elgar Publishing.

Tversky, Amos, and Daniel Kahneman. 1974. Judgment under uncertainty: Heuristics and biases.

Science 185 (4157): 1124-1131.

Tversky, Amos, and Daniel Kahneman. 1981. The framing of decisions and the psychology

of choice. Science 211 (4481): 453-458.

Urban Institute. 2005. Paying the price? Low-income parents and the use of paid tax preparers.

Washington, D.C.: Urban Institute and New Federalism, National Survey of America’s

Families.

U.S. Government Accountability Office. 2003. Most Taxpayers Believe They Benefit from Paid

Preparers, but Oversight is a Challenge for IRS. Washington, D.C.: Government

Accountability Office.

Young, Donald. 2020. How social norms and social identification constrain aggressive reporting

behavior. The Accounting Review 96 (3): 449-478.

Table 1: Experimental Design and Procedures

Treatment 1

Tax Preparer Type

Credentials?

Yes

Audit Rate

0.20

0.05

0.20

0.05

Average Tax Savings

$937.50

$437.50

$937.50

$437.50

Price

$200

$300

$150

Treatment 2

Tax Preparer Type

Credentials?

Yes

Audit Rate

0.00

0.40

Average Tax Savings

$937.50

$437.50

$937.50

$437.50

Price

$500

$200

$150

Treatment 3

Tax Preparer Type

Credentials?

Yes

Audit Rate

0.00

0.35

0.00

0.40

Average Tax Savings

$937.50

$437.50

$937.50

$437.50

Price

$500

$200

$400

$150

Treatment 4

Tax Preparer Type

Credentials?

Yes

Audit Rate

0.00

0.35

0.40

Average Tax Savings

$937.50

$437.50

$937.50

$437.50

Price

$1,500

$1,200

$500

$150

Table 2: Participant Characteristics

Summary Statistics

All

Cornell

Experiment Variables

Fixed Income

9225.35

9260.99

Variable Income

2499.86

2483.80

Reported Variable Income

1651.28

1298.16

Demographic Characteristics

Female

63.4%

69.1%

Age

37.6

34.3

Hispanic

9.2%

7.6%

White

62.6%

70.0%

Black

21.5%

8.5%

Asian/Pacific Islander

20.3%

26.5%

HS Graduate

6.6%

6.7%

Some College

22.2%

30.0%

College Graduate

39.7%

35.4%

Postgraduate degree

31.2%

27.4%

Income less than $25,000

15.6%

20.6%

Income $25,000–$49,999

17.5%

Income $50,000–$74,999

24.8%

22.9%

Income $75,000–$99,999

12.8%

13.5%

Income $100,000–$149,999

20.1%

17.5%

Income Over $150,000

9.2%

8.1%

Employed

84.4%

77.6%

Married

30.7%

30.0%

Average Number of Children

0.66

0.57

Tax Experiences

Used National Tax Company

30.0%

27.4%

Used Local Tax Company

15.3%

18.4%

Used Individual Tax Preparer

27.8%

23.3%

Never Used Tax Preparer

26.9%

30.9%

Has Been Audited

12.5%

9.9%

Location

Fors Marsh

9185.82

2517.68

2043.04

57.2%

40.5

11.0%

54.5%

36.0%

13.5%

6.5%

13.5%

44.5%

35.5%

10.0%

17.5%

27.0%

12.0%

23.0%

10.5%

92.0%

31.5%

0.77

32.8%

11.9%

32.8%

22.4%

15.4%

Table 3: Preparer–Compliance Choices

Analysis Variable

Tax Preparer Choice

Preparer A

Preparer B

Preparer C

Preparer D

Income Percentage Reported

100%

Table 4: Conditional Logit Regression Coefficients: Determinants of Preparer Choice

Regression specification

Net Saving

Fear

(1)

0.000994***

(9.42e-05)

[1.07]

-0.642***

(0.101)

[2.00]

-1.082***

(0.0972)

[1.82]

0.814***

(0.124)

[2.15]

(2)

0.000998***

(9.43e-05)

[1.07]

-0.505***

(0.172)

[4.65]

-0.765***

(0.157)

[4.44]

0.815***

(0.124)

[2.15]

-0.238

(0.212)

[4.21]

-0.521**

(0.202)

[4.19]

(3)

-0.751***

(0.196)

[5.26]

Guilt

-0.804***

(0.167)

[7.31]

Credential

0.862***

(0.125)

[3.64]

Female X Fear

-0.272

(0.218)

[4.21]

Female X Guilt

-0.529***

(0.205)

[4.19]

Expected Tax Saving

0.000662***

0.000656***

(0.000142)

[3.24]

Underreporting Saving

0.00111***

0.00113***

(0.000119)

(0.000120)

[5.03]

[5.04]

Expected Penalty

-0.00118***

-0.00119***

(0.000198)

(0.000197)

[2.38]

Price (Cost)

-0.00166***

-0.00168***

(0.000248)

(0.000249)

[2.97]

29,176

29,096

29,176

29,096

Mean VIF

[1.76]

[3.45]

[3.52]

[4.25]

***

Notes: Robust standard errors clustered at subject level are reported in parentheses. denotes significance at 1%

level; ** denotes significance at 5% level. The variance inflation factors (VIF) are reported in brackets.

-0.903***

(0.142)

[2.61]

-1.121***

(0.118)

[4.73]

0.861***

(0.125)

[3.65]

(4)

Table 5: Conditional Logit Regression Coefficients of Preparer Choice: Rounds 1-5 versus Rounds 6-10

Regression specification

Net Saving

Fear

(1)

0.000943***

(0.000104)

[1.07]

-0.757***

(0.108)

[2.01]

-1.254***

(0.105)

[1.82]

0.899***

(0.131)

[2.14]

(2)

0.00105***

(0.000103)

[1.07]

-0.525***

(0.112)

[2.00]

-0.923***

(0.0995)

[1.82]

0.734***

(0.132)

[2.15]

(3)

(4)

-1.020***

-0.782***

(0.149)

(0.163)

[2.61]

Guilt

-1.271***

-0.988***

(0.140)

(0.135)

[4.66]

[4.81]

Credential

0.946***

0.779***

(0.132)

(0.133)

[3.63]

[3.66]

Expected Tax Saving

0.000698***

0.000632***

(0.000155)

(0.000156)

[3.23]

[3.25]

Underreporting Saving

0.00101***

0.00123***

(0.000152)

(0.000162)

[4.94]

[5.13]

Expected Penalty

-0.00107***

-0.00128***

(0.000202)

(0.000217)

[2.38]

Price (Cost)

-0.00160***

-0.00171***

(0.000277)

(0.000254)

[2.97]

14,392

14,784

14,392

14,784

Mean VIF

[1.76]

[3.49]

[3.55]

Notes: Robust standard errors clustered at subject level are reported in parentheses. *** denotes significance at 1%

level. The variance inflation factors (VIF) are reported in brackets.

Table 6: Conditional Logit Regression Coefficients of Preparer Choice: Including Fear and Guilt

Regression specification

Net Saving

Fear

(1)

0.000994***

(9.42e-05)

[1.07]

-0.642***

(0.101)

[2.00]

-1.082***

(0.0972)

[1.82]

(2)

0.000792***

(0.000102)

[1.37]

-0.471***

(0.119)

[3.63]

-1.365***

(0.143)

[3.06]

0.682***

(0.223)

[2.92]

0.799***

(0.123)

[2.70]

(3)

-0.815***

(0.151)

[3.84]

Guilt

-1.458***

(0.168)

[5.79]

Fear x Guilt

0.681***

(0.241)

[3.29]

Credential

0.814***

0.861***

(0.124)

(0.125)

[2.15]

[3.65]

[3.99]

Expected Tax Saving

0.000662***

0.000592***

(0.000142)

[3.24]

[3.28]

Underreporting Saving

0.00111***

0.000933***

(0.000119)

(0.000131)

[5.03]

[5.45]

Expected Penalty

-0.00118***

-0.000811***

(0.000198)

(0.000187)

[2.38]

[3.17]

Price (Cost)

-0.00166***

-0.00164***

(0.000248)

(0.000247)

[2.97]

[3.00]

29,176

Mean VIF

[1.76]

[2.73]

[3.52]

[3.98]

Notes: Robust standard errors clustered at subject level are reported in parentheses. *** denotes significance at 1%

level. The variance inflation factors (VIF) are reported in brackets.

-0.903***

(0.142)

[2.61]

-1.121***

(0.118)

[4.73]

(4)

Figure 1: Fraction of Random Income Reported in Treatment 1

Notes: The horizontal axis presents the reported income divided by true random income (the “compliance

percentage”); the vertical axis represents the percent of subjects with the relevant compliance percentage.

Figure 2: Fraction of Random Income Reported Across All Treatments

Notes: The horizontal axis presents the reported income divided by true random income (the “compliance

percentage”); the vertical axis represents the percent of subjects with the relevant compliance percentage.

Figure 3: Distribution of Chosen Extended Options, Treatment 1

Notes: The horizontal axis presents the possible tax preparer choices as defined in Table 3; the vertical axis

represents the percent of subjects who made the relevant tax preparer choice.

Figure 4: Distribution of Chosen Extended Options, Treatment 2

Notes: The horizontal axis presents the possible tax preparer choices as defined in Table 3; the vertical axis

represents the percent of subjects who made the relevant tax preparer choice.

Figure 5: Distribution of Chosen Extended Options, Treatment 3

Notes: The horizontal axis presents the possible tax preparer choices as defined in Table 3; the vertical axis

represents the percent of subjects who made the relevant tax preparer choice.

Figure 6: Distribution of Chosen Extended Options, Treatment 4

Notes: The horizontal axis presents the possible tax preparer choices as defined in Table 3; the vertical axis

represents the percent of subjects who made the relevant tax preparer choice.

APPENDIX (1): EXPERIMENTAL INSTRUCTIONS

Welcome to the Laboratory for Experimental Economics and Decision Research (Fors Marsh Group

Experimental Economics Laboratory). Note that deception is NOT allowed in economics experiments.

You will be compensated in cash for your participation at the end of the experiment. The amount you

receive is based on choices you make during the experiment. If you have any questions during the

experiment, please raise your hand and someone will come to assist you. Please do not speak out loud or

speak to the other participants.

In this experiment, you will be given the opportunity to earn money, and you will make choices that will

determine the amount of taxes collected on this money. Given the complexity of calculating your taxes,

you will be asked to choose a tax preparer to complete the filing process on your behalf. These tax

preparers are automated; however, they have characteristics of actual tax preparers and will affect the

amount of your tax refund and the probability of being audited as specified. Although this is only a

simulation of the tax reporting and preparation process, your actual earnings will be based on your

decisions. At the end of these instructions, we have included a glossary of tax-related terminology that

you are free to consult throughout the experiment.

You will receive income in each round, and the amount will be determined in two ways. First, your

certain income in all of the rounds will be determined by estimating the number of gumballs in the onequart jar at the front of the lab. If you estimate the number correctly, you will receive 10,000 experimental

dollars of certain income in each round. If you do not exactly estimate the number of gumballs, your

certain income will be reduced by 50 times your error in the number of gumballs in the jar. So, for

example, if your estimate is off by 10 gumballs, your certain income in EACH round would be 10,000 –

10x50 = 9,500 experimental dollars. You are guaranteed a minimum certain income of 5,000

experimental dollars even if your error is more than 100 gumballs. This income is similar to wage income

received from an employer, and taxes will be automatically withheld from this portion of your income at

the rate of 30 percent. The second component of your income is determined randomly at the beginning of

each round. This portion of your income will be between 0 and 5,000 experimental dollars, and any dollar

amount in this range is equally likely. Each person in the experiment will get a different random draw

from the computer in each round. This random component is meant to simulate the uncertainty most

people face in estimating their annual total income due to uncertainty over the size of possible income

from tips, freelance work, or other sources of income that are not reported to the tax agency by thirdparties or subject to withholding but are supposed to be reported as part of taxable income. The highest

income that you could make in a round is the sum of 10,000 experimental dollars in certain income, if you

exactly estimated the number of gumballs, plus an additional 5,000 in random income if you receive the

highest random income in a round.

In each round you will also be presented with other information regarding your particular tax situation.

You will be eligible for 0-5 tax deductions and 0-5 tax credits in each round. Deductions and credits are

randomly determined, and you have an equal chance of each possibility. A tax credit is an amount that is

subtracted from your total taxes owed, meaning that your tax liability is reduced by the amount of the

credit. Examples of tax credits are the American Opportunity Tax Credit for post-secondary tuition or the

Residential Renewable Energy Tax Credit for certain types of home energy systems. A tax deduction is

an amount that is subtracted from your taxable income, meaning that you do not owe taxes on that portion

of your income. Examples of tax deductions are interest paid on a home mortgage, charitable

contributions, or casualty and theft losses. Eligibility for tax deductions and credits varies by year

depending on changes in your life circumstances and changes in the tax code. To simulate the variation in

tax deductions and credits for which you may be eligible in a single year, the number of deductions and

credits will be randomly given in each round as described above. Additionally, the value of each credit

and deduction will be randomly determined within a certain range, which will depend on your choice of

tax preparer.

The tax rate in this experiment is 25 percent, and this will apply to all income earned minus the value of

any deductions. Any tax credits will reduce your total tax liability amount dollar for dollar. The

experiment involves four stages in each round. In Stage 1, you will be provided with your earnings and

tax information: amount of income subject to withholding, amount of taxes withheld, amount of other

income, and the number of actual deductions and tax credits you are eligible for in that round.

In Stage 2, because the required tax calculations for deductions and tax credits are complex and time

consuming, you are asked to choose a tax preparer from a list of four possible preparers. To help make

that choice, you will be provided with information about each tax preparer, including whether or not the

preparer is credentialed (if the preparer has passed a background check and is a Certified Public

Accountant (CPA)), what the audit rate is for tax returns completed by the preparer, the range of values of

credits and deductions for tax returns completed by the preparer, and the price charged by the tax

preparer. The audit rate gives the probability of being audited if you choose that preparer. If you are

audited, any unpaid taxes must be paid, along with a penalty equal to the amount of unpaid taxes. The

value of credits and deductions will vary based on your tax preparer. If you choose a tax preparer who has

a high value of credits and deductions, each deduction subtracts 500-900 from taxable income and each

credit subtracts 150-250 from taxes owed. If you choose a tax preparer who has a low value of credits and

deductions, each deduction subtracts 100-500 from taxable income and each credit subtracts 50-150 from

taxes owed. Each preparer has a given price which will be subtracted from your earnings for the round.

In Stage 3 you will provide information for your tax filing to the tax preparer you have chosen. Since the

IRS knows your certain income subject to withholding, that amount will be automatically entered.

However, you may report any amount of random income.

In Stage 4, after your tax return has been filed you will receive your earnings (certain plus random

experimental dollars) plus the refund amount as calculated by your chosen tax preparer. The price of the

tax preparer will be subtracted from your total earnings for the round. You will also find out whether you

have been audited. Your probability of audit will be based upon the audit rate of your chosen tax preparer

– for example, if you choose a tax preparer with a 5 percent audit rate, you will be randomly selected for

an audit with a probability of 5 percent. If you have been audited, your actual tax obligation will be

calculated and any unpaid taxes will be deducted from your earnings along with a 100 percent penalty on

unpaid taxes. What this means is that, if you are audited, for every lab dollar in unpaid taxes, you will

have to pay back the one lab dollar you owe in taxes and one additional lab dollar in penalty.

The first round of the experiment will be a practice round so you can see how the experiment works. The

number of gumballs used to calculate your practice earnings will be different than the actual number of

gumballs in the jar. Your choices in this practice round will not go toward your total earnings for the

experiment. After the practice round is completed, you will guess the number of gumballs again, and this

time your earnings will be based on the actual number of gumballs in the jar. The experiment will

continue for several rounds, and your earnings for the experiment will be based on your total earnings for

all the rounds after the practice round. At the end of the experiment, you will be given cash equal to $1

for every 2,700 (1,100 at Fors Marsh Group) experimental dollars you earn. Please raise your hand if you

have a question at any point.

Term

Definition

Audit

An examination by the tax authority of the financial information reported on a

person’s tax return to ensure that it is accurate.

The extra money owed due to any unpaid taxes discovered as a result of an

audit. In this experiment, the audit penalty is equal to 100% of any unpaid

taxes.

The probability of being audited. In this experiment, it ranges from 5% to 20%

depending on the tax preparer chosen.

A person’s official qualifications. In this experiment, a tax preparer with

credentials represents an individual who has passed an IRS background check

and is a Certified Public Accountant.

An amount of money that is subtracted from a person’s tax liability, meaning

that his or her taxes are reduced by this amount.

An amount of money that is subtracted from a person’s taxable income,

meaning that he or she does not have to pay taxes on this portion of income.

A person who helps to calculate your income tax obligation and to file an

income tax return with the tax authority on another person’s behalf in

exchange for a fee.

The percentage of taxable income that is owed in taxes, not including any tax

credits which will reduce the total tax liability. In this experiment, the tax rate

is 25%.

The portion of income on which the amount of income tax is based. It is

calculated by taking total income minus deductions.

The percentage of income that is retained from a person’s earnings and applied

toward his or her taxes. In this experiment, the withholding rate is 30%.

Audit Penalty

Audit Rate

Credentials

Credit

Deduction

Tax Preparer

Tax Rate

Taxable Income

Withholding Rate

APPENDIX (2): SELECTED SCREEN SHOTS

Taxpayer Screens, Baseline Treatment

Taxpayer Screens, Market Information Treatment

APPENDIX (3): POST-EXPERIMENTAL QUESTIONNAIRE

You will now be asked to complete a short survey that asks about your tax experiences,

personal preferences, and some additional background information about yourself. The

survey takes 10-20 minutes and your responses will be kept confidential. When you are

ready to begin, please click the next button to start the survey.

I. Tax Experiences

The following questions ask about your general tax experiences in the past.

Q1: In 2016 (or the most recent year in which you used a tax preparer), what kind of

business did you use to prepare your taxes?

Value

Value Label

National tax company (e.g., H&R Block, Jackson Hewitt, etc.)

Small business or local tax company

Individual tax preparer

I have never used a tax preparer

Q2: In 2016 (or the most recent year in which you used a tax preparer), what credentials did

your tax preparer hold? (Mark all that apply)

Value

Value Label

Attorney

Certified Public Accountant (CPA)

Enrolled Agent

Other

I don’t know

Q3: How many different tax preparers have you used in the past 5 years?

Value

Value Label

One

Two

Three

Four

Five

Q4: For the tax preparer you used in 2016 (or the most recent year in which you used a tax

preparer), what resource did you primarily use to select this tax preparer?

Value

Value Label

Newspapers, yellow pages, magazines, or other print media

Online review site (e.g., Yelp, Angie's List, etc.)

Friend, family member, or other personal connection

IRS.gov website

Tax company website

Q5: In the past 10 years, have you prepared income taxes for others? (Mark all that apply)

Value Value Label

I have never prepared taxes for anyone else

I have prepared taxes for friends or family as a favor

I have prepared taxes for others pro bono, as a volunteer

I have prepared taxes for others in exchange for payment as a part-time, freelance, or

seasonal job

I have prepared taxes for others as part of my full-time job

Q6: Have you ever been formally audited by the Internal Revenue Service (IRS)?

Value

Value Label

Yes

For the following statements, answer whether you strongly agree, agree, neither agree nor

disagree, disagree, or strongly disagree.

Q7. When I pay my taxes as required by the regulations, I do so…

(Mark one answer for each item)

Variable Name

Variable Text

Q7A

Because to me it’s obvious that this is what you do.

Q7B

To support the country and other citizens.

Q7C

Because I like to contribute to everyone’s good.

Q7D

Because for me it’s the natural thing to do.

Q7E

Because I regard it as my duty as a citizen.

Value

Value Label

Strongly agree

Agree

Neither agree nor disagree

Disagree

Strongly disagree

Q8. When I pay my taxes as required by the regulations, I do so…

(Mark one answer for each item)

Variable Name

Variable Text

Q8A

Because a great many tax audits are carried out.

Q8B

Because the IRS often carries out audits.

Q8C

Because I know that I will be audited.

Q8D

Because the punishments for tax evasion are very severe.

Q8E

Because I do not know how to evade taxes without attracting attention.

Value

Value Label

Strongly agree

Agree

Neither agree nor disagree

Disagree

Strongly disagree

The following questions ask you about a number of possible scenarios when filing your

taxes. Please answer how likely or unlikely you are to complete each scenario.

Q9A: You could take a detailed look at the tax regulations yourself to search for potential

savings. How likely would you be to take this detailed look at the tax regulations?

Value

Value Label

Very likely

Likely

Neither likely nor unlikely

Unlikely

Very unlikely

Q9B. You could install soundproof windows in your private dwelling and claim the resulting

cost as housing space reconstruction on your income tax return. This would have the effect

of reducing your tax burden. How likely would you be to carry out the housing space

reconstruction?

Value

Value Label

Very likely

Likely

Neither likely nor unlikely

Unlikely

Very unlikely

Q9C: You could attend a course that informs you about the current possibilities for making

claims against taxes. How likely would you be to attend such a course?

Value

Value Label

Very likely

Likely

Neither likely nor unlikely

Unlikely

Very unlikely

Q9D: You could buy low-value assets (e.g., PC, scanner, and other purchased equipment

with a value below $500) that you do not currently need for your company, so as to

decrease your taxable income. How likely would you be to purchase such equipment?

Value

Value Label

Very likely

Likely

Neither likely nor unlikely

Unlikely

Very unlikely

Q9E: You could deduct against taxes the training costs you incurred for your employees as

an allowable deduction for education and training. How likely is it that you would use the

allowable deduction for education and training?

Value

Value Label

Very likely

Likely

Neither likely nor unlikely

Unlikely

Very unlikely

Q10A: A customer paid in cash and did not require an invoice or receipt. You could

intentionally omit this income on your income tax return. How likely is it that you would omit

this income?

Value

Value Label

Very likely

Likely

Neither likely nor unlikely

Unlikely

Very unlikely

Q10B: You bought some of your goods privately. You could resell those goods later to

established customers and omit the profit from this sale on your income tax return. How

likely would you be to omit the profit from this sale on your income tax return?

Value

Value Label

Very likely

Likely

Neither likely nor unlikely

Unlikely

Very unlikely

Q10C: You could intentionally declare restaurant bills for meals you had with your friends as

business meals. How likely would you be to declare those restaurant bills as business

meals?

Value

Value Label

Very likely

Likely

Neither likely nor unlikely

Unlikely

Very unlikely

Q10D: You have been abroad to meet relatives and to have a short meeting with one of your

suppliers. Regardless of this, you could declare your expenses for the hotel and for the

meals you invited your relatives to as business travel and business meals. How likely would

you be to declare your expenses as business travel or business meals?

Value

Value Label

Very likely

Likely

Neither likely nor unlikely

Unlikely

Very unlikely

Q10E: Recently, you took part in a project in an acquaintance’s company. Now you could

conceal this taxable additional income on your income tax return. How likely is it that you

would conceal this additional income?

Value

Value Label

Very likely

Likely

Neither likely nor unlikely

Unlikely

Very unlikely

II. Risk Aversion

For each of the following questions, you are asked whether you would prefer to choose

lottery A or lottery B.

Q11A: Choose between lottery A and lottery B.

Value

Value Label

10% chance to receive $20; 90% chance to receive $16

10% chance to receive $40; 90% chance to receive $1

Q11B: Choose between lottery A and lottery B.

Value

Value Label

20% chance to receive $20; 80% chance to receive $16

20% chance to receive $40; 80% chance to receive $1

Q11C: Choose between lottery A and lottery B.

Value

Value Label

30% chance to receive $20; 70% chance to receive $16

30% chance to receive $40; 70% chance to receive $1

Q11D: Choose between lottery A and lottery B.

Value

Value Label

40% chance to receive $20; 60% chance to receive $16

40% chance to receive $40; 60% chance to receive $1

Q11E: Choose between lottery A and lottery B.

Value

Value Label

50% chance to receive $20; 50% chance to receive $16

50% chance to receive $40; 50% chance to receive $1

Q11F: Choose between lottery A and lottery B.

Value

Value Label

60% chance to receive $20; 40% chance to receive $16

60% chance to receive $40; 40% chance to receive $1

Q11G: Choose between lottery A and lottery B.

Value

Value Label

70% chance to receive $20; 30% chance to receive $16

70% chance to receive $40; 30% chance to receive $1

Q11H: Choose between lottery A and lottery B.

Value

Value Label

80% chance to receive $20; 20% chance to receive $16

80% chance to receive $40; 20% chance to receive $1

Q11I: Choose between lottery A and lottery B.

Value

Value Label

90% chance to receive $20; 10% chance to receive $16

90% chance to receive $40; 10% chance to receive $1

Q11J: Choose between lottery A and lottery B.

Value

Value Label

100% chance to receive $20; 0% chance to receive $16

100% chance to receive $40; 0% chance to receive $1

III. Social Value Orientation

For the following questions, imagine that you have been randomly paired with another

person, whom we will refer to as the other. This other person is someone you do not know

and both of you will remain mutually anonymous.

You will be making a hypothetical series of decisions about allocating money between you

and this other person. For each of the following questions, please indicate the distribution of

money to yourself and the other you prefer most by selecting the button below the payoff

allocations. You can make only one selection for each question. There are no right or wrong

answers.

Q12A: Please indicate the distribution of money to yourself and the other you prefer most.

You

receive

Other

receives

Q12B: Please indicate the distribution of money to yourself and the other you prefer most.

You

100

receive

Other

receives

Q12C: Please indicate the distribution of money to yourself and the other you prefer most.

You

receive

Other

100

receives

Q12D: Please indicate the distribution of money to yourself and the other you prefer most.

You

receive

Other

100

receives

Q12E: Please indicate the distribution of money to yourself and the other you prefer most.

You

100

receive

Other

100

receives

Q12F: Please indicate the distribution of money to yourself and the other you prefer most.

You

100

receive

Other

receives

IV. Demographics

The final section of this survey asks you for some additional information about yourself.

Q13. What is your birthday?

Q14. What is your gender?

Value

Value Label

Male

Female

Q15. Are you of Hispanic, Latino, or Spanish Origin?

Value

Value Label

No, not of Hispanic, Latino, or Spanish Origin

Yes, Mexican, Mexican American, Chicano

Yes, Puerto Rican

Yes, Cuban

Yes, Other Hispanic, Latino, or Spanish Origin

Q16. Please select all of the following that best describe your race.

Value

Value Label

White

Black or African American

Asian

American Indian or Alaska Native

Native Hawaiian or Other Pacific Islander

Q17 What is the highest degree or level of school that you have completed?

Value

Value Label

12 years or less of school

High school graduate — regular diploma

High school graduate — GED or alternative credential

Some college credit, but less than 1 year

1 or more years of college, no degree

Associate degree (e.g., AA, AS)

Bachelor’s degree (e.g., BA, AB, BS)

Master's, doctoral, or professional school degree (e.g., MA, PhD, JD)

Q18: What is (or was) your major in college?

Value

Value Label

I never attended college

Arts and Humanities

Business, Accounting, and Economics

Health and Medicine

Multi-/Interdisciplinary studies

Public and Social Services

Science, Math, and Technology

Social Sciences

Q19: How many business, accounting, and economics college classes have you completed?

Value

Value Label

None

One or two

Three to five

More than five

Q20. What is your marital status?

Value

Value Label

Married

Separated

Divorced

Widowed

Never married

Q21. How many children do you have?

Value

Value Label

Zero

One

Two

Three

Four

Five

Six or more

Q22. In 2016, what was your household's total combined income? This includes money from

jobs, net income from business, farm or rent, pensions, dividends, interest, social security

payments, and any other money received by family members.

Value

Value Label

Less than $5,000

$5,000 to $7,499

$7,500 to $9,999

$10,000 to $12,499

$12,500 to $14,999

$15,000 to $19,999

$20,000 to $24,999

$25,000 to $29,999

$30,000 to $34,999

$35,000 to $39,999

$40,000 to $49,999

$50,000 to $59,999

$60,000 to $74,999

$75,000 to $99,999

$100,000 to $149,999

$150,000 or more

Q23. Which of the following best describes your 2016 employment status?

Value Value Label

An employee of a private company or business, or of an individual for wages,

salary, or commissions

Government employee (local, state, or federal)

Self employed

Not employed

Q24. [If Q23==4] Which of the following best describes why you were not employed in

2016?

Value

Value Label

Retired

Student

Disabled or unable to work

Homemaker

Not looking for work

IV. Debriefing

Thank you for completing this survey. This concludes the study. Please wait and you will be

given further instructions for receiving your payment for participating in this study.

This is a copy of a public record, reproduced as it was published. It is not legal advice, and it may not be the version a court would rely on. Check the official source before you cite it.