Content uploaded by Richard J. Cebula
Author content
All content in this area was uploaded by Richard J. Cebula on Dec 22, 2014
Content may be subject to copyright.
PSL Quarterly Review, vol. 67 n. 271 (2014), 451-481
© Economia civile
The underground economy in the U.S.A.: preliminary
new evidence on the impact of income tax rates (and
other factors) on aggregate tax evasion 1975-2008
RICHARD J. CEBULA*
1. Introduction
Studies of income tax evasion behaviour essentially fall into three
categories. First, there are the principally theoretical models of tax evasion
behaviour, such as Allingham and Sandmo (1972), Falkinger (1988),
Klepper et al. (1991), Dasgupta (1994), Pestieau et al. (1994) and Caballé
and Panadés (1997). As a rule, most such studies are not empirical in nature,
but they potentially can provide insights into new variables that might
influence tax evasion behaviour in the aggregate. Second, there are a number
of studies that either (a) use questionnaires or (b) undertake experiments,
such as Spicer and Lundstedt (1976), Spicer and Thomas (1982), Baldry
(1987), Alm et al. (1992), Thurman (1991) and Alm et al. (1999). These
studies are empirical in nature, deriving the data largely (if not entirely) from
the experiments, and also potentially can provide insights into new variables
that might influence tax evasion behaviour in the aggregate. Certain of these
studies indicate an aversion to the prospect of being audited while others
reveal a lack of such risk-averse behaviour; still others imply that taxpayers
may be averse to tax evasion on moral grounds. Third, there are those
studies that largely or in some cases exclusively adopt what is referred to as
“official data”, i.e. data obtained from the IRS (the Internal Revenue
Service, or its counterparts outside of the U.S.) and/or some other
“official source”, i.e. a credible government agency. Among the types of
information thusly obtained and analysed are data on income tax evasion,
income tax rates, penalties assessed on detected unpaid income taxes and
audit rates. Such studies endeavour typically either seek to estimate the
relative extent of tax evasion or to identify determinants thereof (Tanzi,
* Jacksonville University; email: rcebula@ju.edu.
452 PSL Quarterly Review
1982; 1983; Clotfelter, 1983; Carson, 1984; Crane and Nourzad, 1987;
Poterba, 1987; Pyle, 1989; Erard and Feinstein, 1994; Feige, 1994; 1996;
Joulfaian and Rider, 1996; Cebula, 1997; 2001; 2004; 2011; 2013; Ali et al.,
2001; Alm and Yunus, 2009; Cebula et al., 2009).
In the empirical literature, it has often been found that the degree of
federal personal income tax evasion is positively affected by income tax
rates (Tanzi, 1982; Clotfelter, 1983; Crane and Nourzad, 1987; Poterba,
1987; Feige, 1994; Joulfaian and Rider, 1996; Cebula, 1997; 2001; 2004;
2011; 2013). Interestingly, Yaniv (1994) characterises Clotfelter (1983) as
“the most relevant study” with respect to the impact of income tax rates on
tax evasion, whereas Cox (1984) questions his findings. In any event, in a
purely static framework, this perspective appears simple and
straightforward, namely, the higher the income tax rate, the greater the
benefit (in terms of a reduced tax liability) from not reporting taxable
income, ceteris paribus. However, as observed by a number of scholarly
papers, including Crane and Nourzad (1986), Caballé and Panadés (2007),
Gahramanov (2009), and Freire-Serén and Panadés (2013), there is an
apparent contradiction between the empirical evidence on the response of
taxpayers to increased or decreased income tax rates and the predictions of
theoretical models of income tax evasion. In other words, on theoretical
grounds it can be argued that the effect of, say, an increase in the income tax
rate on income tax evasion is ambiguous. For example, it follows from
Crane and Nourzad (1986) that the net response of tax evasion to a change in
the tax rate depends upon the relative strength of the substitution and income
effects of a tax rate change.
The purpose of this exploratory study is twofold. First, it seeks to add
to the rich literature on income tax evasion by identifying key
determinants of aggregate federal personal income tax evasion in the
U.S. using data up to and including the year 2008. For the most part,
earlier studies of aggregate tax evasion in the U.S. do not go beyond the
year 1997, although the recent study by Cebula (2013) uses official IRS
data based on the AGI gap approach and runs through to the year 2005.
Thus, by investigating tax evasion through 2008, the study period is more
current than the nearly all of the existing published literature. Second, it
provides a number of empirical estimates of aggregate income tax
The underground economy in the U.S.A. 453
evasion that adopt two alternative income tax rate measures in the effort
to provide information that may be pertinent to the controversy and
debate concerning the actual net response of aggregate income tax evasion
to, say, an increase in the tax rate, depending upon the relative strengths of
the substitution and income effects of a tax rate change.
It is noteworthy that focusing on aggregate income tax evasion and its
determinants permits the analysis of actual (official) as opposed to
hypothetical or experimental tax evasion data and also permits the analysis
of a variety of actual real-world explanatory variables; furthermore, the
aggregate time-series approach adopted in this study permits the analysis of
aggregate federal personal income tax evasion over time. Finally, the use of
aggregate data can provide researchers and policymakers with a convenient
tool for estimating the lost tax revenues resulting from tax evasion and
potential tax receipts increases that various public policies can potentially
generate; indeed, policymakers arguably might well be much more
interested in the results from aggregative analysis using official data because
such information may be easier to present to elected officials seeking to
enact new legislation to limit income tax evasion.
The framework/model is presented in section 2. Section 3 provides the
formal empirical analysis. These estimates first consist of OLS (ordinary
least squares) results and, as a test of robustness, subsequently of 2SLS (two
stage least squares) results. Section 4 provides a summary of the study
findings and certain policy observations.
2. A benefit-cost model of under-reporting taxable income
In this study, the relative probability that the representative economic
agent will under-report his/her taxable income to the IRS (pur) is treated as
an increasing function of the expected gross benefits to the agent of under-
reporting taxable income, egb, and as a decreasing function of the expected
gross costs to the agent of under-reporting taxable income, egc. Thus, the
ratio of the probability of under-reporting taxable income to the IRS, pur, to
454 PSL Quarterly Review
the probability of fully reporting taxable income1 to the IRS, (1-pur), is
described for the representative economic agent by:
pur/(1-pur) = f(egb, egc), such that fegb > 0, fegc < 0 (1)
Expressing the probabilities described in equation (1) in relative terms
possesses the virtue that it thereby reflects the form of the tax evasion data,
i.e., data where (as described in section 3) the aggregate degree of federal
personal income tax evasion is expressed in such relative terms. Needless to
say, these probabilities, pur and (1- pur), can differ from one taxpayer to
another. The aggregative approach adopted here, like that of its predecessors,
is effectively a de facto average of those probabilities and is not perfect.
2.1. Expected direct benefits of under-reporting taxable income
Arguably, on a superficial static level, an example of potential expected
benefits from income tax evasion could simply be the value of the taxes not
paid to the IRS (Internal Revenue Service). To the extent that an individual
engages in income tax evasion and underpays income taxes by, say, $X,
those $X are a direct benefit to the individual, who can (in theory) spend
and/or save the $X. It follows that the higher the marginal federal income tax
rate that the individual is subject to, the greater the pecuniary benefit from
underreporting his or her income. For instance, if the maximum marginal
income tax rate for the taxpayer in question is 10%, evading tax on the
amount of $5,000 would potentially be expected to yield the tax evader
$500; however, if the taxpayer in question faced an increased or higher
income tax rate, say of 40%, the expected direct benefit of this same degree
of tax evasion would appear to be $2,000. Thus, it might be logical to
deduce that, ceteris paribus, the higher the maximum marginal income tax
rate faced by a taxpayer, the greater the incentive to engage in income tax
evasion because the higher that income tax rate, the greater the expected
benefits of the tax evasion behaviour. Clearly, the probability of
underreporting income may differ by income level and from one person to
another, even with the same income. The way in which the tax evasion data
1 Fully reporting all taxable income is complete income tax compliance.
The underground economy in the U.S.A. 455
are gathered reflects a de facto average response of taxpayers to tax-evasion
incentives and influences.
However, from a more rigorous theoretical perspective, it can be
hypothesised that the effect of an income tax rate change on income tax
evasion is actually ambiguous (Crane and Nourzad, 1986; Caballé and
Panadés, 2007; Gahramanov, 2009; Freire-Serén and Panadés, 2013). The
sign on the partial derivative of tax evasion with respect to the income tax
rate theoretically depends upon the relative strengths of the substitution
effect associated with the tax rate change on the one hand and the income
effect associated with the tax rate change on the other hand. Assume an
income tax rate increase is implemented. On the one hand, the substitution
effect, which is always positive, implies an increased incentive to engage in
greater income tax evasion. As Freire-Serén and Panadés (2013, p. 810)
observe, “This effect generates incentives to substitute tax evasion for
honesty”. On the other hand, under the assumed income tax rate increase, the
income effect could either be positive or negative, depending upon the
taxpayer’s attitude towards, i.e. tolerance of, risk of detection of tax evasion
by the IRS. If the income effect is positive, then the taxpayer will likely
engage in increased income tax evasion as a reaction to the tax rate increase.
However, if the income effect is negative and if it outweighs/dominates the
positive substitution effect, then the partial derivative of tax evasion with
respect to the income tax rate is negative. Consequently, the theory implies
that, ceteris paribus, the net effect of an income tax increase on tax evasion
may well be ambiguous.
This study seeks provide potential insight into the actual net impact of
higher income tax rates on income tax evasion in the economy as a whole.
To reflect the federal personal income tax rate, this study adopts two
different measures of the federal personal income tax rate: the maximum
marginal federal personal income tax rate (MAXMARGTX)2 and the average
effective personal income tax rate (AVETXRATE). Arguably some very
recent related research on aggregate personal income tax evasion has used a
maximum marginal tax rate and found it to exercise a positive and
2 This particular measure of the income tax rate is adopted because it can be argued that it is
not only an actual income tax rate but also potentially reflects to some degree the extent to
which the income tax rate schedule at any point in time is progressive.
456 PSL Quarterly Review
statistically significant impact on personal income tax evasion (Alm and
Yunus, 2009; Cebula, 2011; 2013); interestingly, unlike the other studies of
aggregate income tax evasion, the study by Cebula (2011) provides
autoregressive (AR) estimates. In any event, it is observed that these two tax
rate measures are considered in separate estimates because they are rather
highly correlated, i.e. r (AVETXRATE, MAXMARGTX) = 0.631.
Aside from empirically investigating the substitution effect/income
effect controversy at the aggregate level, this exploratory study seeks to
provide contemporary insights into other potential income tax evasion
determinants. For example, in the U.S., Form 1040 (Schedule A) provides a
variety of itemised deduction types that ultimately enable the taxpayer to
reduce his or her taxable income. These include allowances for medical
expenses, state and local government taxes, mortgage interest payments,
charitable contributions and other such outlays. The larger the pecuniary
value of these deductions, the lower the taxable income of the taxpayer.
Furthermore, given the rather varied, numerous and sometimes complex
forms of these deductions, and given the limited ability for the IRS to
directly verify many of these itemised deductions, taxpayers filing their tax
returns and claiming itemised deductions have an opportunity to derive
direct tax benefits by overstating their itemised deductions. Thus, it is
hypothesised that the greater the proportion of taxpayers who itemise their
tax deductions (PCTITEM),3 the greater the degree to which itemised
deductions are exaggerated and hence the greater the direct expected benefits
of federal personal income tax evasion. Interestingly, this variable, which
reduces taxable income, has been effectively ignored in the official tax
evasion literature to date.
Whereas higher income tax rates and the opportunity to claim
exaggerated itemised deductions can yield expected direct tangible benefits
from income tax evasion, there is at least one course of action that can
tangibly reduce the expected direct benefits of income tax evasion.
Moreover, it is a legal course of action, namely, it is legal tax avoidance.
One avenue through which this course of action is made possible is the
existence of the municipal bond market, where qualified bonds issued by
3 Instead of claiming the “standard deduction”.
The underground economy in the U.S.A. 457
cities, counties and states in the U.S. pay interest that is free from federal
income taxation for the owners of record of such bonds.4 In particular,
following Cebula (2004), it can be argued that the greater the ratio of tax
free interest rate yield on high grade municipals relative to taxable
interest rate yield such as that on 10-year U.S. Treasury notes, TFTEN,
the greater the benefits of tax avoidance, which is legal, and hence the
lower the expected direct benefits of tax evasion, which of course is
illegal.
Thus, the expected gross direct benefits from income tax evasion are
expressed as:
egb = egb (MAXMARGTX or AVETXRATE, PCTITEM, TFTEN) (2a)
2.2. Expected secondary benefits of under-reporting taxable income
Aside from the issues of tax rates and itemisation of deductions,
persons who have taxable income may be able in certain cases to derive
secondary benefits from income tax evasion behaviour, i.e. from under-
reporting taxable income. For example, if people disapprove of the way in
which the federal government is spending their tax dollars, they may be
angered or frustrated. To relieve this anger and/or frustration, they may
consider income tax evasion as a means of expressing this anger and/or
frustration (Feige, 1994).
Consider the case of the U.S. military being engaged in an unpopular
war. For example, there is empirical evidence (Feige, 1994; Cebula, 2001)
that the unpopularity of the Vietnam War so angered the public that many
chose to underreport their income and/or exaggerate their itemised tax
deductions. This form of income tax evasion behaviour was undertaken
because of the disapproval by much of the U.S. public of the U.S.
involvement in and expenditures to finance the Vietnam War using their tax
dollars. As the Vietnam War raged on, at least some portion of the public
received secondary benefits/gains from the experience that they were
4 In states that impose a state income tax, there is an exemption from that tax for state
residents purchasing qualified tax-exempt bonds issued within that state (the so-called
“dual exemption”).
458 PSL Quarterly Review
withholding financing of that military action. Similarly, it is hypothesised in
the present study that the U.S. military involvement in Iraq, i.e. the war in
Iraq (IRAQWAR), that began in 2003 and was still raging on through and
beyond the end of the study period for the present paper, quickly became
very unpopular with the U.S. populace as a whole. Indeed, Cebula et al.
(2007) found strong empirical evidence that, unlike the (brief) war involving
not only the U.S. and Iraq but also numerous others nations that took place
in 1991 and invoked immense patriotism (and approval) within the U.S., the
war in Iraq that began in 2003 quickly became very unpopular with a large
percentage of the U.S. population. Hence, the latter war in Iraq is
hypothesised here to have created a circumstance for people to express their
dissatisfaction by engaging in a greater degree of income tax evasion that
yielded them a “secondary gain”.
Thus, the expected gross direct and secondary benefits from income tax
evasion are expressed as the following:
egb = egb (MAXMARGTX or AVETXRATE, PCTITEM, TFTEN,
IRAQWAR) (2b)
2.3. A control variable: the Tax Reform Act of 1986
In 1987, Musgrave observed (1987, p. 59), “The Tax Reform Act of
1986 is the most sweeping reform since the early 1940s[…].” Indeed, the
TRA (Tax Reform Act) did introduce a number of reforms, many of which
are outlined in broad terms in Barth (1991), Barth and Brumbaugh (1992),
Ott and Vegari (2003) and Sanger et al. (1990). For example, as observed in
Ott and Vegari (2003, p. 279),
“[t]he Act introduced major cuts in the personal tax rate. When fully effective
(1988), only two tax brackets set at 15 and 28 percent, were to replace the 14
bracket tax schedule with rates in the range of 11 to 50 percent [while it]
broadened the tax base by reducing the itemized deduction.”
Furthermore, as Barth (1991, pp. 45-46 observes, among other things,
that under the Tax Reform Act (TRA) the 10 percent investment tax credit
for the purchase of equipment was repealed, and the “life” of the investment
The underground economy in the U.S.A. 459
was increased for depreciation purposes.5 Thus, it is hypothesised here that
at the time the TRA was being enacted and fully implemented (1986-1987),
there were many complex and new provisions added to the U.S. Internal
Revenue Code. Consequently, taxpayers in general, including would-be as
well as “repeat” tax evaders, were unfamiliar with all of the sweeping
changes in IRS policies. It logically follows that not only honest taxpayers
but also those contemplating or planning income tax evasion required time
to climb the “learning curve” associated with the TRA, resulting in at least
some temporary diminution of the aggregate degree of federal personal
income tax evasion. Indeed, the idea that this reaction to the TRA might be
only temporary was originally revealed in the words of Slemrod (1992, p.
45), who some years ago argued that it would take at least some time for
taxpayers “to learn about and adjust to the new law [the TRA]”. In any case,
it is hypothesised here that, for the period when the TRA was initially
implemented, 1986, through the year the TRA became “de facto fully
effective”, 1987 (Barth, 1991; Barth and Brumbaugh, 1992; Cebula et al.,
2009), the value of egb was reduced. Accordingly, (2b) above is replaced by
(3):
egb = egb (MAXMARGTX or AVETXRATE, PCTITEM, TFTEN,
IRAQWAR, TRA) (3)
2.4. Expected gross costs of under-reporting taxable income
The expected gross costs of not reporting or under-reporting taxable
income to the IRS and/or of reporting exaggerated itemised deductions to the
IRS are hypothesised to be an increasing function of the expected risks/costs
thereof (Pestieau et al., 1994; Erard and Feinstein, 1994; Caballé and
Panadés, 1997). In this study, to the representative economic agent, the
expected risks/costs from underreporting taxable income to the IRS are
enhanced by an increase in AUDIT, the percentage of filed federal personal
income tax returns that is formally audited by IRS examiners/personnel,
ceteris paribus. Indeed, the experience of an IRS tax audit could imply non-
5 The Tax Reform Act of 1986 was actually signed into law by President Reagan in October,
1986.
460 PSL Quarterly Review
pecuniary (“psychic”) costs (such as psychological stress) as well as direct
pecuniary costs (including outlays for attorneys and/or other representation
such as accountants or financial advisors, along with the value of one’s own
time) above and beyond any potential added taxes, penalties, and interest
assessed by the IRS. Hence, this study adopts the probability of a formal
audit as a measure of risk to the would-be tax evader. In addition, IRS
penalty assessments on detected unreported income are also adopted as a
measure of the risks/costs associated with tax evasion. In particular, the
greater the average penalty assessed by the IRS per audited tax return (PEN),
the greater the expected costs of tax evasion, ceteris paribus.
Hence, the expected gross costs of engaging in income tax evasion at
the aggregate level is represented in this study by:
egc = egc (AUDIT, PEN) (4)
2.5. The synthesised model
To express the full model simply requires substituting from equations
(3) and (4) into equation (1), yielding:
pur/(1-pur) = f(MAXMARGTX or AVETXRATE, PCTITEM, TFTEN,
IRAQWAR, TRA, AUDIT, PEN) (5)
Let AGI represent the actual total value of the aggregate federal
adjusted gross income in the economy, i.e. AGI = UAGI+RAGI, where
UAGI is the dollar size of the unreported aggregate federal adjusted gross
income in the economy,6 and RAGI is the dollar size of the reported
aggregate federal adjusted gross income in the economy. It logically follows
that the relative degree of aggregate income tax evasion can be expressed as:
(pur)/(1-pur) = UAGI/RAGI. Thus, it follows that equation (5) can be
replaced by equation (6):
UAGI/RAGI = f (MAXMARGTX or AVETXRATE, PCTITEM, TFTEN,
6 For an explanation of how unreported adjusted gross income is calculated using the
General Currency Ratio model, see Feige (2009) and Cebula and Feige (2012).
The underground economy in the U.S.A. 461
IRAQWAR, TRA, AUDIT, PEN) (6)
Equation (6) constitutes the foundation for the empirical model
considered in the next section of this study.7
3. Empirical analysis
3.1. The empirical model
Based on the framework provided in equation (6) above, the following
reduced-form equations are to be estimated initially:8
(UAGI/RAGI)t = a0 + a1 MAXMARGTXt-1 + a2 PCTITEMt-1 + a3 TFTENt-1 +
a4 TRAt + a5 IRAQWARt + a6 AUDITt-1 + a7 PENt-1 + u’ (7)
(UAGI/RAGI)t = b0 + b1 AVETXRATEt-1 + b2 PCTITEMt-1 + b3 TFTENt-1 + b4
TRAt + b5 IRAQWARt + b6 AUDITt-1 + b7 PENt-1 + u” (8)
where:
(UAGI/RAGI)t = the ratio of the aggregate unreported federal adjusted
gross income in year t to the aggregate reported federal adjusted gross
income in year t, expressed as a percentage; a0, b0 = constant terms;
MAXMARGTXt-1 = the maximum marginal federal personal income tax rate
in year t-1, expressed as a percentage;
AVETXRATEt-1 = the average effective marginal federal personal income tax
rate in year t-1, expressed as a percentage;
PCTITEMt-1 = the percentage of federal personal income tax returns that
included Schedule A, itemising deductions, in year t-1;
TFTENt-1 = the ratio of the average nominal interest rate yield on high grade
7 Clearly, UAGI = (pur)*AGI and RAGI = (1-pur)*AGI. It then follows that:
UAGI/RAGI = (pur)*AGI/(1-pur)*AGI = (pur)/(1-pur). Substitution of UAGI/RAGI for
pur/(1-pur) in (5) yields: UAGI/RAGI = f(MAXMARGTX or AVETXRATE, PCTITEM,
TFTEN, TRA, IRAQWAR, AUDIT, PEN), which is equation (6).
8 Cebula (2011) includes variables reflecting public dissatisfaction with government and
the opportunity cost of tax compliance but overlooks the variables PCTITEM and
AVETXRATE and measures PEN differently from this study, i.e. in terms of the average
penalty per audited return.
462 PSL Quarterly Review
tax free municipal bonds in year t-1 to the average nominal interest rate yield
on 10-year Treasury notes in year t-1, expressed as a percentage;
IRAQWARt = a binary (dummy) variable for the years 2003-2008 of the
study period during which the U.S. was involved in an unpopular war, i.e.,
the war in Iraq (“Operation Iraqi Freedom”): IRAQWARt = 1 for the years
2003-2008 and IRAQWARt = 0 otherwise;
TRAt = a binary (dummy) variable for the years 1986 and 1987: TRAt = 1 for
the years 1986 and 1987, and TRAt = 0 otherwise;
AUDITt-1 = the percentage of filed federal personal income tax returns in year
t-1 that was subjected to a formal IRS audit involving IRS examiners;
PENt-1 = IRS imposed penalties plus interest on detected unreported income
in year t-1, expressed as a percent of per capita real GDP in year t-1; and
u’, u” = stochastic error terms.
The study period runs from 1975 through 2008, reflecting availability
of the data used, in particular, for the variables PCTITEM and PEN, in the
analysis. The data are annual. For the interested reader, descriptive statistics
for the study period for each of the variables are found in table 1 of this
study. Note that the number of observations is only 33 due to data
limitations; clearly a larger N would be preferable, but more recent data on
the dependent variable is currently unavailable. A group unit root test (which
assumes a common unit root process) reveals that the variables in the model
represented in equation (7) are stationary in levels over the study period;
similarly, a group unit root test reveals that the variables in the model
represented in equation (8) are stationary in levels over the study period as
well.9
The series adopted to measure income tax evasion, in this case
represented by the variable (UAGI/RAGI), were obtained from Cebula and
Feige (2012).10 Based on the General Currency Ratio (GCR) model, Cebula
and Feige (2012, table B-2) provide estimates of the ratio of aggregate
unreported adjusted gross income to aggregate reported adjusted gross
income. These data are provided in table 2 of the present study. The data for
MAXMARGTX and AVETXRATE were obtained from the Internal Revenue
Service (2010, table 6). The AUDIT, PCTITEM and PEN data were obtained
from the Government Accounting Office (1996, table I.1) and the U.S.
9 These results will be supplied upon e-mail request.
10 See also Feige (2009); the GCR model actually refers to underreported taxable income.
The underground economy in the U.S.A. 463
Census Bureau (1994, table 519; 1998, table 550; 1999, table 556; 2001,
table 546; 2010, table 469). The data for the variable TFTEN were obtained
from the Council of Economic Advisers (2013, table B-73). The IRAQWAR
and TRA variables are binary (dummy) variables.
Table 1 – Descriptive Statistics
Variable
Mean Period:
1975-2008 Standard Deviation
(UAGI/RAGI) 21.013 2.027
MAXMARGTX 48.157 15.73
AVETXRATE 13.99 1.087
TRA 0.0513 0.224
PCTITEM 31.863 3.692
TFTEN 89.9 10.8
AUDIT 1.362 0.4
PEN 2.648 1.832
IRAQWAR 0.181 0.381
N = 33
3.2. The initial estimation results: the linear specification
The OLS estimation of equations (7) and (8) in linear form are provided
in columns (a) and (b), respectively, in table 3; in both cases, the Newey and
West (1987) heteroskedasticity correction was adopted.11 In column (a), all
seven of the estimated coefficients exhibit the expected signs. Four of these
estimated coefficients are statistically significant at the 1% level, and two are
statistically significant at the 5% level. The coefficient of determination (R2)
is 0.72, so that the model explains approximately seven-tenths of the
variation in the independent variable. Based on the DW and Rho statistics,
there is no concern regarding autocorrelation. Finally, the F-statistic is
statistically significant at the 1% level, attesting to the overall strength of the
11 Testing for heteroskedasticity revealed a need to make such a correction in all of the
estimations in this study.
464 PSL Quarterly Review
model.
Table 2 – Data for Dependent Variable UAGI/RAGI, by Year, 1960-2008
Year UAGI/RAGI Year UAGI/RAGI
1960 16.10 1985 21.11
1961 15.47 1986 18.89
1962 15.86 1987 17.42
1963 16.44 1988 18.74
1964 15.88 1989 21.06
1965 14.62 1990 21.06
1966 14.86 1991 21.39
1967 15.36 1992 19.04
1968 15.21 1993 17.70
1969 15.32 1994 17.98
1970 16.30 1995 20.01
1971 16.04 1996 18.64
1972 16.16 1997 18.66
1973 16.27 1998 18.30
1974 17.47 1999 20.55
1975 18.81 2000 22.29
1976 20.17 2001 22.73
1977 20.37 2002 23.94
1978 20.63 2003 23.17
1979 21.14 2004 21.57
1980 22.84 2005 21.98
1981 22.25 2006 23.85
1982 22.93 2007 24.90
1983 21.46 2008 23.94
1984 21.86
Note: UAGI/RAGI is expressed as a percentage.
Source: Cebula and Feige (2012, table B-2, p. 282).
According to the results provided in column (a) of table 3, the
coefficient on the maximum marginal federal personal income tax rate
variable (MAXMARGTX) is positive and statistically significant at the 1%
level. Thus, the higher the maximum marginal federal personal income tax
Table 3 – Empirical Estimates 1975-2008, OLS
The underground economy in the U.S.A. 465
Dependent Variable:
Specification: Linear
(UAGI/RAGI)
(a) (b)
MAXMARGTX 0.07***
(3.69) ------
AVETXRATE -------- 0.616*
(1.81)
PCTITEM 0.27**
(2.07) 0.292*
(1.72)
TFTEN -12.4***
(-3.81) -11.88***
(-4.14)
TRA -4.313***
(-4.22) -4.292***
(-4.62)
AUDIT -3.223***
(-4.29) -1.872**
(-2.37)
PEN -0.218
(-1.22) -0.302
(-1.30)
IRAQWAR 2.328**
(2.27) 3.39*
(1.85)
Constant 24.81 16.43
R2 0.72 0.70
AdjR2 0.64 0.61
F 8.78*** 8.02***
DW 1.76 1.72
Rho 0.12 0.14
Note: Terms in parentheses are t-values; ***indicates statistical significance at the 1% level; **indicates
statistical significance at the 5% level; *indicates statistical significance at the 10% level.
rate, the greater the extent of that income tax evasion. This finding is
consistent with several previous empirical studies of income tax evasion
(Tanzi, 1982; Clotfelter, 1983; Crane and Nourzad, 1987; Poterba, 1987;
Feige, 1994; Joulfaian and Rider, 1996; Ali et al., 2001; Cebula, 1997; 2001;
2004; 2011; 2013). As observed above, the sign on the partial derivative of
income tax evasion with respect to (in this case) the maximum marginal
federal personal income tax rate theoretically depends upon the relative
strengths of the substitution effect on the one hand and the income effect on
the other hand, associated with the tax rate change. The finding in column
(a), of and in itself, would seem to imply that, for the representative
466 PSL Quarterly Review
taxpayer, the substitution effect, which is always positive, outweighs the
income effect (be it positive or negative). And as Freire-Serén and Panadés
(2013, p. 810) observe, “This effect [the substitution effect] generates
incentives to substitute tax evasion for honesty.” But of course this is but one
finding in one estimation.
The estimated coefficient on the PCTITEM variable is positive and
statistically significant at the 5% level. This finding implies that the greater
the percentage of taxpayers that itemise their personal deductions (on
Schedule A of Form 1040), the greater the degree to which taxable income is
underreported and hence the greater the degree of aggregate federal personal
income tax evasion. This finding has effectively not been formally
researched in the tax evasion literature to date.
The estimated coefficient on the tax free/taxable interest rate variable,
TFTEN, is negative, as hypothesised, and statistically significant at the 1%
level, providing compelling empirical evidence that the greater the rewards
for legal tax avoidance (as measured here), the less the aggregate degree of
illegal personal income tax evasion. This finding is consistent with Cebula
(2004), who first proposed and tested this hypothesis, albeit with data
running only through the year 1997.
Consistent with the arguments in Musgrave (1987) and findings in
Cebula et al. (2009) and Cebula (2011), the results for TRA variable are
compelling. In particular, the estimated coefficient on variable TRA is
negative and statistically significant at the 1% level. Thus, there is evidence
that the Tax Reform Act of 1986 is shown to have reduced federal personal
income tax evasion in the U.S., albeit only briefly. Given the specification of
TRA as applying to the short-term period of just 1986 and 1987, these results
would seem to confirm the arguments in Slemrod (1992) and the findings in
Cebula et al. (2009), who also argue that it would take at least some time for
taxpayers to understand the revisions in the Internal Revenue Code and to
adjust to those revisions.
The estimated coefficient on the IRAQWAR dummy is positive and
statistically significant at the 4% level, which is consistent with the
hypothesis proffered above that an unpopular war elicits a reduction in
income tax compliance. This behaviour arguably reflects a “secondary gain”
from income tax evasion.
The underground economy in the U.S.A. 467
Next, there is the audit rate (by IRS examiners) variable. As shown in
the first column of table 3, the estimated coefficient on this variable is
negative and statistically significant at the 1% level. Thus, it appears that the
audit rate variable (AUDIT), of and in itself, may be viewed as a deterrent to
federal personal income taxation evasion. This finding is consistent with
previous studies such as Pestieau et al. (1994), Erard and Feinstein (1994)
and Caballé and Panadés (1997), who suggest that IRS policies such as a
higher audit rate by IRS personnel impose a variety of costs, both pecuniary
and psychic, that act to dissuade income tax evasion behaviour.
Finally, the estimated coefficient for the IRS penalty variable, PENt-1,
is negative, as expected; however, it is not statistically significant at even the
10% level. Thus, in this linear estimate, the evidence implies that higher
levels of IRS imposed penalties and interest do not materially act to
discourage federal personal income tax evasion.
In column (b) of table 3, the OLS estimate of equation (8) is
provided. This estimate differs in construct from that in equation (7)
only insofar as it adopts an alternative measure of the aggregate
income tax rate structure, namely, AVETXRATEt-1. In column (b), all
seven of the estimated coefficients exhibit the hypothesised signs, with
two statistically significant at the 1% level, one statistically significant
at the 5% level and three statistically significant at the 10% level;
hence, these results appear to be somewhat less robust than those in
column (a). The coefficient of determination is 0.70, so that the model
in equation (7) explains approximately seven-tenths of the variation in
the dependent variable. Once again autocorrelation is not a problem.
Finally, the F-statistic is statistically significant at the 1% level, as its
counterpart in column (a) was.
According to the results in column (b) of table 3, the coefficient on the
average effective federal personal income tax rate variable (AVETXRATE) is
positive but statistically significant at only the 8% level. Thus, there is
evidence, but not terribly compelling evidence, that the higher the average
effective federal personal income tax rate, the greater the expected benefits
of tax evasion and hence the greater the extent of that income tax evasion.
This finding is arguably inconsistent with several previous empirical
studies (Tanzi, 1982; Clotfelter, 1983; Crane and Nourzad, 1987;
468 PSL Quarterly Review
Poterba, 1987; Feige, 1994; Joulfaian and Rider, 1996; Ali et al., 2001;
Cebula, 1997; 2001; 2004; 2011; 2013), which found more compelling
evidence that higher income tax rates elicit higher income tax evasion. As
observed above, in the aggregate, for the representative taxpayer, the sign
on the partial derivative of income tax evasion with respect to, in this
case, the average effective federal personal income tax rate, theoretically
depends upon the relative strengths of the substitution effect associated
with the tax rate change on the one hand and the income effect associated
with the tax rate change on the other hand. The finding in column (b), in
and of itself, does not convincingly imply that the substitution effect,
which is always positive, outweighs the income effect (be it positive or
negative). Indeed, the failure of this estimated coefficient to be
statistically significant at the 5% level, could potentially be interpreted as
implying that the impact of a higher income tax rate (as measured by
variable AVETXRATE) on income tax evasion is ambiguous, depending
upon whether statistical significance at the 8% level is considered robust
enough to deduce something resembling a clear positive impact of tax
rates on tax evasion. Nevertheless, based on this result in table 3, one
must be cautious about accepting or rejecting this inference. Results
shown in table 4 are relevant to this issue.
In column (b), the estimated coefficient on the PCTITEM variable is
positive and statistically significant at beyond the 10% level. This finding
implies, although un-compellingly, that the greater the percentage of
taxpayers that itemise their personal deductions (on Schedule A of Form
1040), the greater the degree to which taxable income is underreported
and hence the greater the degree of aggregate federal personal income tax
evasion.
The estimated coefficient on the tax free/taxable interest rate variable,
TFTEN, is negative, as hypothesised, and statistically significant at the 1%
level, providing additional empirical evidence (along with the counterpart
finding shown in column (a) that the greater the rewards for legal tax
avoidance (as measured here), the less the aggregate degree of illegal
personal income tax evasion.
The underground economy in the U.S.A. 469
Table 4 – Empirical Estimates 1975-2008, OLS
Dependent Variable:
Specification: (Log-Log)
log (UAGI/RAGI)
(a)
(b)
logMAXMARGTX 0.059
(1.03) --------
logAVETXRATE -------- 0.303
(1.07)
logPCTITEM 0.491**
(2.40) 0.48*
(1.95)
logTFTEN -0.633***
(-3.65) -0.604***
(-3.85)
TRA -0.201***
(-4.52) -0.204***
(-5.21)
logAUDIT -0.157***
(-5.59) -0.1604***
(-3.19)
logPEN -0.028
(-1.05) -0.033
(-1.20)
IRAQWAR 0.103**
(2.37) 0.145*
(1.77)
Constant 1.085 0.539
R2 0.73 0.74
AdjR2 0.65 0.66
F 9.16*** 9.51***
DW 1.82 1.93
Rho 0.08 0.03
Note: Terms in parentheses are t-values; ***indicates statistical significance at the 1% level; **indicates
statistical significance at the 5% level; *indicates statistical significance at the 10% level.
Consistent with the arguments in Musgrave (1987) and findings in
Cebula et al. (2009), the results for the TRA variable are compelling.
In particular, in column (b), the estimated coefficient on the variable
TRA is negative and statistically significant at the 1% level. Thus,
there is further evidence that the Tax Reform Act of 1986 reduced
federal personal income tax evasion in the U.S., albeit only briefly. As
470 PSL Quarterly Review
observed above, given the specification of TRA as applying to just
1986 and 1987, these results would seem to confirm the argument in
Slemrod (1992) and the later findings by Cebula et al. (2009), who
also argued that it would take at least some time for taxpayers to
understand the many revisions in the Internal Revenue Code and to
adjust to those revisions.
Unlike the finding for this variable in column (a), the estimated
coefficient on the IRAQWAR dummy is positive and statistically significant
at only the 8% level, which is “weakly” consistent with the hypothesis
proffered above that an unpopular war elicits a reduction in income tax
compliance. This behaviour implies, but not strongly, that the unpopularity
of the War in Iraq provided a “secondary gain” for taxpayers from income
tax evasion.
Next, there is the audit rate (by IRS examiners) variable. As shown
in column (b), the estimated coefficient on this variable is negative and
statistically significant at the 3% level. Thus, it once again appears that
the audit rate variable (AUDIT) may be viewed as a deterrent to federal
personal income taxation. This finding is consistent with the result in
column (a) and with previous studies such as Pestieau et al. (1994), Erard
and Feinstein (1994) and Caballé and Panadés (1997), who suggest that
IRS policies such as a higher audit rate by IRS personnel impose a
variety of costs, both pecuniary and psychic, that act to dissuade income
tax evasion behaviour.
Finally, the estimated coefficient for the IRS penalty variable is
negative, as expected; however, as found to be the case in column (a) of
table 3, it is not statistically significant at the 10% level. Thus, in this linear
estimate, the evidence implies that higher levels of IRS imposed penalties
and interest do not measurably act to discourage federal personal income tax
evasion.
3.3. Alternative estimation results: the log-log specification
In this sub-section of the study, the results of estimating the models in
equations (7) and (8) in log-log form are provided in columns (a) and (b) of
table 4. Once again, it is noted that the Newey and West (1987)
The underground economy in the U.S.A. 471
heteroskedasticity correction was adopted in both estimates. In column (a) of
table 4, all seven of the estimated elasticity/coefficient values exhibit the
expected signs, with three statistically significant at the 1% level and two
statistically significant at the 5% level. The F-statistic is statistically
significant at the 1% level, and the R2 value is 0.73, so that the model
explains in excess of seven-tenths of the variation in the dependent variable.
According to the results provided in column (a) of table 4, the
coefficient/elasticity on the maximum marginal federal personal income tax
rate variable (MAXMARGTX) is positive but it fails to be statistically
significant at even the 10% level. This finding is inconsistent with several
previous empirical studies of income tax evasion (Tanzi, 1982; Clotfelter,
1983; Crane and Nourzad, 1987; Poterba, 1987; Feige, 1994; Joulfaian and
Rider, 1996; Ali et al., 2001; Cebula, 1997; 2001; 2004; 2011; 2013), as
well as the results in column (a) of table 3 (and, arguably column (b) of table
3) in this study, which found higher tax rates to elicit increased tax evasion.
As observed above, for the representative taxpayer, the sign on the partial
derivative of tax evasion with respect to, in this case, an increase in the
maximum marginal federal income tax rate, theoretically depends upon the
relative strengths of the substitution effect on the one hand and the income
effect on the other hand, associated with the tax rate change. The finding in
column (a), of and in itself, would seem to imply that for the representative
taxpayer the substitution effect, which is always positive, merely offsets a
negative income effect, so that in this estimate the impact of the
MAXMARGTX measure of the income tax rate on tax evasion is on balance
ambiguous.
Also in column (a) of table 4, the estimated coefficient/elasticity on the
PCTITEM variable is positive and statistically significant at the 2.5% level.
This finding implies that the greater the percentage of taxpayers itemising
their personal deductions (on Schedule A of Form 1040), the greater the
degree to which taxable income is underreported and hence the greater the
degree of aggregate federal personal income tax evasion. For example, a 1%
increase in the value of PCTITEM would lead to a 0.491% increase in
(UAGI/RAGI).
Once again, the estimated coefficient/elasticity on the tax free/taxable
interest rate variable, TFTEN, is negative, and statistically significant at the
472 PSL Quarterly Review
1% level, providing additional compelling empirical evidence that the
greater the rewards for legal tax avoidance (as measured), the less the
aggregate degree of illegal personal income tax evasion (Cebula, 2004).
Indeed, a 1% increase in the value of TFTEN would lead to a 0.633%
decrease in (UAGI/RAGI).
Consistent with arguments in Musgrave (1987), Slemrod (1992) and
Cebula et al. (2009), the results for the TRA variable are again robust. In
particular, the estimated coefficient on variable TRA is again negative and
statistically significant at the 1% level. Thus, there is further evidence that
the Tax Reform Act of 1986 is shown to have reduced federal personal
income tax evasion in the U.S., albeit only briefly.
The estimated coefficient on the IRAQWAR dummy is positive and
statistically significant at the 3% level, which is consistent with the
hypothesis that an unpopular war elicits a reduction in income tax
compliance. This behaviour arguably implies the presence of secondary
gains from income tax evasion. Next, there is the audit rate (by IRS
examiners) variable. As shown in the first column of table 4, the estimated
coefficient/elasticity on this variable is negative and statistically significant
at the 1% level. Thus, there appears to be further evidence that the audit rate
variable (AUDIT) may be viewed as a deterrent to federal personal income
taxation. For instance, a 1% increase in the value of AUDIT would lead to a
0.157% decrease in (UAGI/RAGI). Finally, the estimated
coefficient/elasticity for the IRS penalty variable, PENt-1, is once again
negative; however, it is not statistically significant at even the 10% level.
Thus, in this log-log estimate, the evidence implies that higher levels of IRS
imposed penalties and interest do not act to discourage federal personal
income tax evasion.
In column (b) of table 4, all seven of the estimated elasticity/coefficient
values exhibit the expected signs, with three statistically significant at the
1% level and two statistically significant at the 10% level. The F-statistic is
statistically significant at the 1% level, and the R2 value is 0.74, so that the
model explains nearly three-fourths of the variation in the dependent
variable.
According to the results provided in column (a) of table 4, the
coefficient on the average effective federal personal income tax rate variable
The underground economy in the U.S.A. 473
(AVETXRATE) is positive but it fails to be statistically significant at even the
10% level. This finding also is inconsistent with several previous empirical
studies of income tax evasion (Tanzi, 1982; Clotfelter, 1983; Crane and
Nourzad, 1987; Poterba, 1987; Feige, 1994; Joulfaian and Rider, 1996; Ali
et al., 2001; Cebula, 1997; 2001; 2004; 2011; 2013), as well as the results in
column (a) of table 3 (and, arguably column (b) of table 3) in this study,
which found higher tax rates to elicit increased tax evasion. As observed
above, for the representative taxpayer, the sign on the partial derivative of
tax evasion with respect to, in this case, the average effective federal income
tax rate, theoretically depends upon the relative strengths of the substitution
effect on the one hand and the income effect on the other hand, associated
with the tax rate change. The finding in column (a), of and in itself, would
seem to imply that for the representative taxpayer the substitution effect,
which is always positive, merely offsets a negative income effect, so that the
impact of the AVETXRATE measure of the income tax rate on tax evasion is,
on balance, ambiguous.
Also in column (b) of table 4, the estimated elasticity on the PCTITEM
variable is positive and statistically significant at the 6% level. Arguably,
this finding implies that the greater the percentage of taxpayers that itemise
their personal deductions (on Schedule A of Form 1040), the greater the
degree to which taxable income is underreported and hence the greater the
degree of aggregate federal personal income tax evasion. This outcome is
effectively consistent with our three previous results for this variable. For
example, a 1% increase in the value of PCTITEM would presumably lead to
a 0.48% increase in (UAGI/RAGI).
Once again, the estimated elasticity on the tax free/taxable interest rate
variable, TFTEN, is negative, and statistically significant at the 1% level,
providing additional compelling empirical evidence that the greater the
rewards for legal tax avoidance (as measured), the less the aggregate degree
of illegal personal income tax evasion. Indeed, a 1% increase in the value of
TFTEN would apparently lead to a 0.604% decrease in (UAGI/RAGI).
The results for the TRA variable are once again robust. In particular, the
estimated coefficient on variable TRA is again negative and statistically
significant at the 1% level. Thus, there is further evidence that the Tax
Reform Act of 1986 is shown to have reduced federal personal income tax
474 PSL Quarterly Review
evasion in the U.S., albeit only briefly.
The estimated coefficient on the IRAQWAR dummy is positive and
statistically significant at the 9% level, which is modestly supportive of the
hypothesis that an unpopular war elicits a reduction in income tax
compliance through a secondary gain from income tax evasion.
Next, there is the audit rate (by IRS examiners) variable. As shown in
the first column of table 3, the estimated coefficient/elasticity on this
variable is yet again negative and statistically significant at the 1% level.
Thus, there appears to be further evidence that the audit rate variable
(AUDIT) may be viewed as a deterrent to federal personal income taxation
evasion. For instance, a 1% increase in the value of AUDIT would lead to a
0.134% decrease in (UAGI/RAGI).
Finally, although the estimated coefficient/elasticity for the IRS
penalty variable, PENt-1, is negative, it once again is not statistically
significant at the 10% level. Thus, in this log-log estimate, the evidence
implies that higher levels of IRS imposed penalties and interest do not
dissuade federal personal income tax evasion.
3.4. 2SLS estimation results: a robustness test
As a test of the robustness of the results of the basic model, this sub-
section of the study provides 2SLS estimates of the following re-specified
versions of equations (7) and (8):
(UAGI/RAGI)t = a0 + a1 MAXMARGTXt-1 + a2 PCTITEMt-1 + a3 TFTENt-1 +
a4 TRAt + a5 IRAQWARt + a6 AUDITt + a7 PENt-1 + u’ (9)
(UAGI/RAGI)t = b0 + b1 AVETXRATEt-1 + b2 PCTITEMt-1 + b3 TFTENt-1 + b4
TRAt + b5 IRAQWARt + b6 AUDITt + b7 PENt-1 + u” (10)
These specifications differ from those in equations (7) and (8)
insofar as the audit variable is now shown as un-lagged. This
specification implies that the greater the current IRS audit rate, which
might well be the case if the relevant IRS audit has been publicly
announced or otherwise made known to the public during the tax-filing
The underground economy in the U.S.A. 475
season, the greater the risk of tax evasion and hence the lower the current
aggregate degree of tax evasion. As such, the dependent variable,
(UAGI/RAGI)t, and the explanatory variable AUDITt are
contemporaneous. As a result of this circumstance, the possibility of
simultaneity bias arises. To address this issue, the model in both
equations is estimated by 2SLS. The instrumental variable adopted is the
two-year lag of the federal/central government budget deficit expressed
as a percentage of GDP, DEFYt-2. The choice of this instrument is based
on the fact that DEFYt-2 and AUDITt are highly correlated, whereas
DEFYt-2 is uncorrelated with the error terms in the system.
The 2SLS estimations of equations (9) and (10) are provided in
columns (a) and (b) of table 5. For the most part, the non-tax rate results in
this table parallel those in table 3, where OLS results for the models were
reported. What is most interesting is that in column (a), the estimated
coefficient on the tax variable MAXMARGTXt-1 is positive and statistically
significant at the 1% level, whereas in column (b) the estimated coefficient
on the tax variable AVETXRATEt-1 is not statistically significant at even the
10% level. Thus, the higher the maximum marginal federal personal income
tax rate, the greater the extent of aggregate income tax evasion; however, a
higher average income tax rate elicits no statistically significant change in
the aggregate degree of tax evasion. These results are consistent with the
OLS estimates reported in table 3 and imply that the impact of higher federal
personal income tax rates is “ambiguous”, i.e., it can be fairly stated that this
exploratory study once again, this time using 2SLS, finds evidence most
compatible with “ambiguity”. Indeed, the same conclusion is reached if
equations (9) and (10) are estimated by 2SLS in log-log form.
4. Summary and closing observations
This exploratory empirical study of the underground economy in the
U.S. examines a new and updated series based on the General Currency
476 PSL Quarterly Review
Table 5 – Empirical Estimates 1975-2008, 2SLS
Dependent Variable:
Specification: Linear
(UAGI/RAGI)
(a)
(b)
MAXMARGTX 0.156***
(3.30) --------
AVETXRATE -------- 0.448
(0.63)
PCTITEM 0.005
(0.04) 0.244*
(1.72)
TFTEN -19.7***
(-2.76) -27.8***
(-3.09)
TRA -3.69***
(-3.96) -3.37***
(-4.01)
AUDIT -7.67***
(-2.92) -6.14**
(-2.57)
PEN -0.02
(-0.10) -0.47*
(-2.00)
IRAQWAR 4.49***
(3.47) 4.50**
(2.13)
Constant 40.42 39.93
F 5.00*** 3.97***
DW 1.74 1.80
Rho 0.13 0.10
Note: Terms in parentheses are t-values; ***indicates statistical significance at the 1% level; **indicates
statistical significance at the 5% level; *indicates statistical significance at the 10% level.
Ratio (GCR) model on aggregate personal income tax evasion for the period
1975 through 2008, with 2008 being the most recent year for which income
tax evasion data are available for the U.S. Focusing on aggregate income tax
evasion and its determinants permits the analysis of actual (official) as
opposed to hypothetical or experimental tax evasion figures and also permits
the analysis of a variety of actual real-world explanatory variables;
furthermore, the aggregate time-series approach adopted in this study allows
analysis of tax evasion over time. Finally, the use of aggregate data can
provide researchers and policymakers with a convenient tool for estimating
the lost tax revenues resulting from tax evasion and potential tax receipts
The underground economy in the U.S.A. 477
increases that certain public policies may potentially generate. In any event,
the IRS has data based on the AGI gap approach; however, the most recent
year for which this data is available is 2005 (Internal Revenue Service, 2010,
table 6). Thus, the present study is more “contemporary” than previous
studies using “official data”.
The purpose of this exploratory study is two-fold. First, adopting a cost-
benefit model of income tax evasion decision-making, this study seeks to
identify key determinants of aggregate federal personal income tax evasion
in the U.S. using data available up to and including the year 2008.
Second, it provides a number of empirical estimates of aggregate income
tax evasion that adopt two alternative income tax rate measures in the
effort to provide information that may be pertinent to the controversy and
debate concerning the actual net response of aggregate income tax evasion
to, say, an income tax rate increase and the relative strengths of the
substitution and income effects of a tax rate change. The estimates initially
take the form of OLS estimates; however, as a test of robustness, 2SLS
estimates are also provided. Both the OLS and 2SLS estimates yield the
same conclusions regarding the impact of income tax rates on income tax
evasion: ambiguity.
More specifically, over the study period, the principal conclusions are
the following: federal personal income tax evasion is an increasing function
of the percentage of filed federal personal income tax returns that itemises
deductions, and U.S. involvement in an unpopular war, in this case, the war
in Iraq. The study also finds persistent evidence that (a) the Tax Reform Act
of 1986 acted to (briefly) discourage tax evasion; (b) a higher IRS audit rate
by IRS personnel acted to discourage tax evasion; and (c) the greater the
benefits of legal tax avoidance, as measured by the ratio of the tax free
interest rate yield on high grade municipals to the taxable interest rate yield
on 10-year Treasury notes, the less the degree of illegal tax evasion. By
contrast, the study finds consistent evidence that IRS imposed penalties and
interest on detected unreported income to have no discernible impact on
aggregate personal income taxation evasion.
Six different estimates addressed the impact of higher income tax rates
on aggregate federal income tax evasion. In the linear OLS model estimates,
the coefficient on the maximum marginal personal income tax rate was
478 PSL Quarterly Review
positive and statistically significant at the 1% level, whereas the average
effective personal income tax rate while positive was statistically significant
at only the 8% level. In the two log-log OLS estimates, both income tax rate
measures were found to be statistically insignificant at even the 10% level.
Finally, in the 2SLS estimates, the coefficient on the maximum marginal
personal income tax rate was positive and statistically significant at the 1%
level, whereas the average effective personal income tax rate, while positive,
was not statistically significant at even the 10% level. On balance, it is
unclear what the net impact of a higher income tax rate is.
Given the tax rate measures considered, it would appear that there is a
reasonable argument that can be made that the tax-evasion impact of higher
income tax rates may well be ambiguous. However, this is but one study.
The decision to engage in tax evasion is complex and difficult to model,
especially across an entire economy with so many millions of taxpayers,
each with her/his own utility function; indeed, behavioural patterns will also
change over time both as a reflection of taxpayer “turnover” and changing
cultural, political and economic conditions, among other things. Moreover,
for any given time frame, alternative specifications may yield different
results than these, as might studies of longer time periods or studies based on
experimentation. Nevertheless, it can be fairly stated that this exploratory
study finds preliminary evidence most compatible with “ambiguity”.
The findings imply, among other things, that in the pursuit of greater
tax revenues, limiting the ability to itemise personal tax deductions may be a
fruitful path to consider; indeed, it was a significant component of the Tax
Reform Act of 1986. However, it is a path requiring careful planning so as to
avoid political fall-out with the taxpaying public. Obviously, avoidance of
what is likely to be an unpopular war would be wise not only on moral,
ethical and political grounds, but also on the practical grounds of avoiding
tax revenue losses. Finally, although increasing the IRS examiner audit rate
may yield additional tax revenues, political pragmatism would seem to
require a circumspect implementation of such a policy change: draconian
IRS empowerment would likely not be well received by the populace.
In closing, it is observed that subsequent related research might seek to
identify additional factors potentially influencing income tax evasion in the
U.S., including the presence of undocumented immigrants within its borders.
The underground economy in the U.S.A. 479
In addition, although 2SLS estimates have been provided in this study,
future research might undertake a more in-depth investigation of the
possibility of simultaneity issues between tax evasion and other variables
(Cebula, 2001; Alm and Yunus, 2009). Finally, as more current data become
available over time, revisiting the issues at hand may yield further insights.
REFERENCES
ALI M.M., CECIL H.W. and KNOBLETT J.A. (2001), “The Effects of Tax Rates and
Enforcement Policies on Taxpayer Compliance: A Study of Self-Employed
Taxpayers”, Atlantic Economic Journal, vol. 29 n. 2, pp. 186-202.
ALLINGHAM M.G. and SANDMO A. (1972), “Income Tax Evasion: A Theoretical
Analysis”, Journal of Public Economics, vol. 1 nn. 3-4, pp. 323-338.
ALM J. and YUNUS M. (2009), “Spatiality and Persistence in U.S. Individual Income Tax
Compliance”, National Tax Journal, vol. 62 n. 1, pp. 101-124.
ALM J., JACKSON B. and MCKEE M. (1992), “Institutional Uncertainty and Taxpayer
Compliance”, The American Economic Review, vol. 82 n. 4, pp. 1018-1026.
BALDRY J.C. (1987), “Income Tax Evasion and the Tax Schedule: Some Experimental
Results”, Public Finance/Finances Publiques, vol. 42 n. 3, pp. 357-383.
BARTH J.R. (1991), The Great Savings and Loan Debacle, Washington (DC): American
Enterprise Institute Press.
BARTH J.R. and BRUMBAUGH R.D. (1992), The Reform of Federal Deposit Insurance,
New York: Harper Business.
BAWLEY D. (1982), The Subterranean Economy, New York: McGraw-Hill.
CABALLÉ J. and PANADÉS J. (1997), “Tax Evasion and Economic Growth”, Public
Finance/Finances publiques, vol. 52 nn. 3-4, pp. 318-340.
––––– (2007), “Tax Rates, Tax Evasion, and Growth in a Multi-period Economy”,
Hacienda Pública Española/Revista de Economía Pública, vol. 183 n. 4, pp. 67-80.
CARSON C. (1984), “The Underground Economy: An Introduction”, Survey of Current
Business, vol. 64 n. 5, pp. 21-37.
CEBULA R.J. (1997), “An Empirical Analysis of the Impact of Government Tax and
Auditing Policies on the Size of the Underground Economy: The Case of the United
States, 1973-94”, The American Journal of Economics and Sociology, vol. 56 n. 2,
pp. 173-185.
––––– (2001), “Impact of Income-detection Technology and Other Factors on Aggregate
Income Tax Evasion: The Case of the United States”, Banca Nazionale del Lavoro
Quarterly Review, vol. 54 n. 219, pp. 401-415.
––––– (2004), “Income Tax Evasion Revisited: The Impact of Interest Rate Yields on
Tax-Free Municipal Bonds”, Southern Economic Journal, vol. 71 n. 2, pp. 418-423.
––––– (2011), “An Empirical Analysis of Determinants of Recent Federal Personal
Income Tax Evasion in the U.S.”, Journal of Public Finance and Public Choice,
480 PSL Quarterly Review
vol. 29 nn. 1-3, pp. 33-54.
––––– (2013), “New and Current Evidence on Determinants of Aggregate Federal
Personal Income Tax Evasion in the United States”, The American Journal of
Economics and Sociology, vol. 72 n. 3, pp. 701-731.
CEBULA R.J. and FEIGE E. (2012), “America’s Unreported Economy: Measuring the Size,
Growth and Determinants of Income Tax Evasion in the U.S.”, Crime, Law and
Social Change, vol. 57 n. 3, pp. 265-286.
CEBULA R.J., COOK J. and ISSA T.A. (2007), “Economics, Wars and Scandals: Their
Impacts on the US Public Approval Ratings of its President over the Long Run”,
The ICFAI Journal of Applied Economics, vol. 6 n. 3, pp. 31-39.
CEBULA R.J., COOMBS C. and YANG B.Z. (2009), “The Tax Reform Act of 1986: An
Assessment in Terms of Tax Compliance Behavior”, International Economics, vol.
62 n. 2, pp. 145-161.
CLOTFELTER C.T. (1983), “Tax Evasion and Tax Rates: An Analysis of Individual
Returns”, The Review of Economics and Statistics, vol. 65 n. 3, pp. 363-373.
COUNCIL OF ECONOMIC ADVISERS (2013), Economic Report of the President, Washington
(DC): U.S. Government Printing Office.
COX D. (1984), “Raising Revenue in the Underground Economy”, National Tax Journal,
vol. 37 n. 3, pp. 283-288.
CRANE S.E. and NOURZAD F. (1986), “Inflation and Tax Evasion: An Empirical
Analysis”, The Review of Economics and Statistics, vol. 68 n. 2, pp. 217-223.
––––– (1987), “On the Treatment of Income Tax Rates in Empirical Analysis of Tax
Evasion”, Kyklos, vol. 40 n. 3, pp. 338-348.
DASGUPTA A. (1994), “A Theory of Hard-to-Get Groups”, Public Finance/Finances
publiques, vol. 49 supplement, pp. 28-39.
ERARD B. and FEINSTEIN J.S. (1994), “The Role of Moral Sentiments and Audit
Perceptions in Tax Compliance”, Public Finance/Finances Publiques, vol. 49
supplement, pp. 70-89.
FALKINGER J. (1988), “Tax Evasion and Equity: A Theoretical Analysis”, Public
Finance/Finances Publiques, vol. 43 n. 3, pp. 388-395.
FEIGE E.L. (2009), “New Estimates of Overseas U.S. Currency Holdings, the
Underground Economy and the ‘Tax Gap’”, MPRA Paper, n. 30353, Munich
University Library.
––––– (1994), “The Underground Economy and the Currency Enigma”, Public
Finance/Finances Publiques, vol. 49 supplement, pp. 119-136.
FREIRE-SERÉN M.J. and PANADÉS J. (2013), “Do Higher Tax Rates Encourage/Discourage
Tax Compliance?”, Modern Economy, vol. 4 n. 12, pp. 809-817.
FRIEDLAND N. (1982), “A Note on Tax Evasion as a Function of the Quality of Information
about the Magnitude and Credibility of Threatened Fines: Some Preliminary
Research”, Journal of Applied Social Psychology, vol. 12 n. 1, pp. 54-59.
GAHRAMANOV E. (2009), “The Theoretical Analysis of Income Tax Evasion Revisited”,
Economic Issues, vol. 14 n. 1, pp. 35-41.
GOVERNMENT ACCOUNTING OFFICE (1996), Individual Audit Rates, Washington (DC):
U.S. Government Printing Office.
INTERNAL REVENUE SERVICE (2010), Aggregate Unreported Income Statistics,
The underground economy in the U.S.A. 481
Washington (DC): U.S. Government Printing Office.
JOULFAIAN D. and RIDER M. (1996), “Tax Evasion in the Presence of Negative Income
Tax Rates”, National Tax Journal, vol. 49 n. 4, pp. 553-570.
KLEPPER S., NAGIN D. and SPURR S. (1991), “Tax Rates, Tax Compliance, and the
Reporting of Long-Term Capital Gains”, Public Finance/Finances publiques, vol.
46 n. 2, pp. 236-251.
MUSGRAVE R.A. (1987), “Short of Euphoria”, The Journal of Economic Perspectives, vol.
1 n. 1, pp. 59-71.
NEWEY W.K. and WEST K.D. (1987), “A Simple, Positive Semi-Definite,
Heteroskedasticity and Autocorrelation Consistent Covariance Matrix”,
Econometrica, vol. 55 n. 3, pp. 703-708.
OTT A.F. and VEGARI S.B. (2003), “Tax Reform: Chasing the Elusive Dream”, Atlantic
Economic Journal, vol. 31 n. 3, pp. 266-282.
PESTIEAU P., POSSEN U.M. and SLUTSKY S.M. (1994), “Optimal Differential Taxes and
Penalties”, Public Finance/Finances Publiques, vol. 49, supplement, pp. 15-27.
POTERBA J.M. (1987), “Tax Evasion and Capital Gains Taxation”, The American
Economic Review, vol. 77 n. 2, pp. 234-239.
SANGER G.C., SIRMANS C.F. and TURNBULL G.K. (1990), “The Effects of Tax Reform on
Real Estate: Some Empirical Results”, Land Economics, vol. 66 n. 4, pp. 409-424.
SLEMROD J.B. (1992), “Did the Tax Reform Act of 1986 Simplify Tax Matters?”, The
Journal of Economic Perspectives, vol. 6 n. 1, pp. 45-57.
SPICER M.W. and LUNDSTEDT S.B. (1976), “Understanding Tax Evasion”, Public
Finance/Finances Publiques, vol. 31 n. 2, pp. 295-305.
SPICER M.W. and THOMAS J.E. (1982), “Audit Probabilities and the Tax Evasion
Decision: An Experimental Approach”, Journal of Economic Psychology, vol. 2 n.
3, pp. 241-245.
TANZI V. (1982), The Underground Economy in the United States and Abroad, Lexington
(MA): Lexington Books.
––––– (1983), “The Underground Economy in the United States: Annual Estimates, 1930-
80”, IMF Staff Papers, vol. 30 n. 2, pp. 283-305.
THURMAN Q.C. (1991), “Taxpayer Noncompliance and General Prevention: An
Expansion of the Deterrence Model”, Public Finance/Finances Publiques, vol. 46
n. 2, pp. 289-298.
U.S. CENSUS BUREAU (1994), Statistical Abstract of the United States 1994, Washington
(DC): U.S. Government Printing Office.
––––– (1998), Statistical Abstract of the United States 1998, Washington (DC): U.S.
Government Printing Office.
––––– (1999), Statistical Abstract of the United States 1999, Washington (DC): U.S.
Government Printing Office.
––––– (2001), Statistical Abstract of the United States 2001, Washington (DC): U.S.
Government Printing Office.
––––– (2010), Statistical Abstract of the United States 2010, Washington (DC): U.S.
Government Printing Office.
YANIV G. (1994), “Tax Evasion and the Income Tax Rate: A Theoretical Reexamination”,
Public Finance/Finances Publiques, vol. 49 n. 1, pp. 107-112.
