Income tax evasion revisited: the impact of interest rate yields on tax-free municipal bonds.
Income tax evasion remains a timely issue, as evidenced by recently published studies (e.g., see Feige 1989, 1994; Alm, Jackson, and McKee 1992; Pestieau, Possen, and Slutsky 1994; Cebula 1997, 2001; Alm, McClelland, and Schulze 1999; Atkins 1999; Panteghini 2000; Ali, Cecil, and Knoblett 2001; Saltz 2001). In an effort to provide further insight into the determinants of tax evasion, this study hypothesizes that the tax-evasion decision may include an assessment of the tax-free interest rate yield on, say, high-grade municipal bonds relative to the taxable interest rate yield on alternative high-quality bond issues, such as 10-year Treasury notes. Presumably, the higher the interest rate yield on high-grade municipals (whose interest is legally exempt from federal income taxation) relative to the taxable interest rate yield on 10-year Treasury notes, the more attractive those municipals become, ceteris paribus. In effect, the interest paid on tax-free bonds, especially for households in higher tax brackets, may offer an attractive legal alternative (income tax avoidance) to illegal income tax evasion. (1) Thus, it is hypothesized that the higher the tax-free interest rate yield on high-grade municipals relative to the taxable yield on 10-year Treasury notes (or equivalent taxable issues), the less the incentive for and, hence, the less the degree of income tax evasion.
This study investigates empirically whether higher yields on tax-free municipals (relative to the taxable yields on 10-year Treasury notes) reduce the degree of aggregate federal personal income tax evasion. Section 2 provides the model, which includes a variety of variables already established in the literature to affect tax evasion. Section 3 describes the data. Sections 4 and 5 provide the empirical findings and concluding remarks.
2. The Model
The economy consists of agents who generate economic value that is reflected in the form of income and choose whether to report none, some, or all of their income to the Internal Revenue Service (IRS). The probability that the representative economic agent will not report taxable income, pnr, is treated as an increasing function of the expected gross benefits to the agent of not reporting income, eb, and a decreasing function of the expected gross costs to the agent of not reporting income, ec:
(1) pnr = f(eb, ec), [f.sub.eb] > 0, [f.sub.ec] < 0.
The expected gross benefits from not reporting income to the IRS are hypothesized to be an increasing function of income tax rates. Unlike previous studies, this study focuses on both the average effective federal personal income tax rate (AEPT) and the maximum marginal federal personal income tax rate (MAX), such that
(2) eb = g(AEPT, MAX), [g.sub.AEPT] > 0, [g.sub.MAX] > 0.
Furthermore, as suggested by Feige (1994) and Cebula (2001), the higher the level of public dissatisfaction (DIS) with the perceived performance of government employees and elected government officials, the greater the subjective benefit taxpayers derive from tax evasion. Accordingly, Equation 2 becomes
(2') eb = g(AEPT, MAX, DIS), [g.sub.AEPT] > 0, [g.sub.MAX] > 0, [g.sub.DIS] > 0,
Finally, as explained in section 1, it is hypothesized in this study that the higher the tax-free interest rate yield on high-grade municipal bonds (TF) relative to the taxable interest rate yield on 10-year Treasury notes (TEN), the greater the incentive to engage in legal tax avoidance and the lower the incentive to engage in tax evasion. Thus, Equation 2' becomes
(2") eb = h(AEPT, MAX, DIS, TF/TEN), [h.sub.AEPT] > 0, [h.sub.MAX] > 0, [h.sub.DIS] > 0, [h.sub.TF/TEN] < 0.
The expected gross costs of not reporting income to the IRS are hypothesized to be an increasing function of the expected risks thereof. In this study, to the representative economic agent, the expected penalty from not reporting taxable income to the IRS, if said activity is detected by the IRS, is measured in part by the total pecuniary penalties (including both penalties per se and interest) previously assessed by the IRS on audited tax returns (PEN). Furthermore, these expected risks are presumably enhanced by an increase in AUDIT, the percentage of filed federal personal income tax returns that is formally audited by agents of the IRS. Finally, it is also arguable that, to the extent that is perceived as successful, the IRS's IMP (income-matching program) technology (MATCH) may act to dissuade tax evasion. As argued in Cebula (2001, p. 405), "the ability to successfully underreport income, especially taxable non-wage income, is influenced by the ability of the IRS to detect such income ... improving IMP technology ... should act to reduce [tax evasion] ... by ... increasing the risks [of detection] associated with income-underreporting activities." Thus,
(3) ec = j(AUDIT, PEN, MATCH), [j.sub.AUDIT] > 0, [j.sub.PEN] > 0, [j.sub.MATCH] > 0.
Substituting from Equations 2" and 3 into Equation 1 yields
(4) pnr = f(AEPT, MAX, DIS, TF/TEN, AUDIT, PEN, MATCH), [f.sub.AEPT] > 0, [f.sub.MAX] > 0, [f.sub.DIS] > 0, [f.sub.TF/TEN] < 0, [f.sub.AUDIT] < 0, [f.sub.PEN] < 0, [f.sub.MATCH] < 0.
3. Empirical Model
Before providing the empirical evidence based on Equation 4, the data adopted for measuring the aggregate degree of federal personal income tax evasion (DTE) are addressed. A number of studies have estimated the magnitude of the DTE, including Bawley (1982), Tanzi (1982, 1983), Feige (1989, 1994), and Pyle (1989). Based on these studies, there appear to be at least three basic approaches to estimating the size of the DTE: (i) the aggregate adjusted gross income (AGI) gap approach, (ii) the Taxpayer Compliance Measurement Program (TCMP), and (iii) Currency Ratio (CR) Models, including the General Currency Ratio (GCR) Model.
The aggregate AGI gap data are compiled by the Bureau of Economic Analysis (BEA), which computes the discrepancy between the aggregate AGI reported to the IRS and an independent estimate of the aggregate AGI derived from the National Income and Product Accounts estimate of aggregate personal income.
A second approach is that prepared by the IRS based on its TCMP. In each year when the TCMP is prepared, a sample of roughly 55,000 taxpayers is subjected to a detailed examination by IRS auditors. In 1973, the IRS estimated that tax evasion, expressed as the ratio of aggregate unreported adjusted gross income (UAGI) to aggregate reported adjusted gross income (RAGI), was 0.1484, that is, 14.84%. (2)
A third approach relies on some variant of the CR model, including the GCR model, which is perhaps best described in Feige (1989). This model can take a number of forms. In its simplest and most restrictive form, the CR model assumes that (i) currency is the exclusive medium of exchange for unreported transactions, (ii) the ratio of currency to checkable deposits is affected only by the growth of unreported domestic transactions, (iii) the income velocities of reported and unreported transactions are equal to one another, and (iv) in some base year, unreported income (expressed as the ratio UAGI/RAGI) was known, so that the observed currency/checkable-deposit ratio (as a percentage) in that base year serves as a surrogate for the currency ratio associated with that degree of tax evasion. As the observed currency/checkable-deposit ratio rises and falls over time, so accordingly does the degree of income tax evasion; that is, so does the estimated ratio (UAGI/RAGI).
In this study, the data set used to measure the DTE is the series by Edgar Feige; this series is available for the period 1973-1997. Feige (1989, 1994) has generated revised and updated estimates of UAGI as a percentage of RAGI, employing the estimated IRS ratio for UAGI/RAGI of 14.84% for 1973 as the base year. Feige actually uses a variant of the CR model, namely, the GCR model, which somewhat relaxes the assumptions previously listed for the CR model. Interestingly, the Feige data and the well-known series on the size of the "underground economy" as a percentage of GNP developed by Tanzi (1982, 1983) are very highly correlated (+0.93) for the years the two series have in common.
Based on the model in Equation 4, the following reduced-form equation is to be estimated:
(5) [(UAGI/RAGI).sub.t] = [a.sub.0] + [a.sub.1] [MAXFPTR.sub.t-1] + [a.sub.2] [AEPT.sub.t-1]+ [a.sub.3] [AUDIT.sub.t-1]+ [a.sub.4] [PEN.sub.t-1] + [a.sub.5] [DIS.sub.t] + [a.sub.6] [MATCH.sub.t] + [a.sub.7] [(TF/TEN).sub.t-1] + u,
[a.sub.o] = constant term;
[(UAGI/RAGI).sub.t] = aggregate unreported AGI as a percentage of aggregate reported AGI in year t as estimated by Feige (1989, 1994);
(2) For a detailed discussion of the TCMP data. see the two studies by Feige (1989, 1994).
[MAXFPTR.sub.t-1] = the maximum marginal federal personal income tax rate in year t - 1 as a percentage;
[AEPT.sub.t-1] = the average effective federal personal income tax rate in year t - 1 as a percentage;
[AUDIT.sub.t-1] = the percentage in year t - 1 of filed federal personal income tax returns that was subjected to a formal IRS agent audit;
[PEN.sub.t-1] = a measure of the average penalty from underreporting income to the IRS in year t - 1, computed as the aggregate value of pecuniary penalties on detected unreported income in year t - 1, including net interest charges, scaled by the aggregate RAGI in year t - 1;
[DIS.sub.t] = the dissatisfaction index in year t: [DIS.sub.t] lies between the values of 1.5 for least dissatisfied to +1.5 for most dissatisfied;
[MATCH.sub.t] = the estimated percentage of the aggregate taxable nonwage income detected by the IRS's IMP technology in year t;
[(TF/TEN).sub.t-1] = the ratio of the average annual nominal interest rate yield on Standard and Poor's high-grade municipal bonds in year t - 1 to the average annual nominal interest rate yield on 10-year Treasury notes in year t - 1 expressed as a percentage; u = stochastic error term.
The time series examined in this study are annual but, because of data limitations on the variable [MATCH.sub.t], can cover only the 1975-1997 period. To measure the personal income tax rate, this study adopts two tax rate measures: (i) the maximum marginal federal personal income tax rate, MAXFPTR, because tax-free municipals are likely to be more appealing to higher-income households, and (ii), following Feige (1994), the average effective federal personal income tax rate, AEPT. To measure the public's dissatisfaction with government, the variable DIS is represented by the "dissatisfaction index." This index is constructed as an equally weighted average of three normalized indices reflecting answers to surveys by the University of Michigan's Institute for Social Research (ISR) concerning whether government employees and elected government officials can be trusted to complete their assigned job duties, whether they are dishonest, and whether they waste tax dollars. Values for the dissatisfaction index range from a low of -1.5, which corresponds to least dissatisfied, to a high of +1.5, which corresponds to most dissatisfied. To obtain this series, see Feige (1994) and the University of Michigan (1998). The data for the remaining variables were obtained from Feige (1989, 1994), the IRS (1997, 2004), the Council of Economic Advisors (2003, table B-73), and the Tax Foundation (2004).
The Phillips-Perron (P-P) unit root test statistics indicate that six of the explanatory variables (MAXFPTR, AEPT, AUDIT, PEN, DIS, and MATCH) are stationary only in first differences over the study period. In addition, the P-P test indicates that the Feige data (UAGI/RAGI), as well as the relative interest rate yield data (TF/TEN), are stationary in levels with a trend variable (TREND). Consequently, in the estimation provided in this study, variables MAXFPTR, AEPT. AUDIT, PEN, DIS, and MATCH are all expressed in first differences, and a trend variable is included.
Given that the tax-evasion variable is contemporaneous with the variables MATCH, and [DIS.sub.t], the possibility of simultaneity bias exists. Accordingly, an instrumental variables (IV) approach is adopted, with the instruments for [MATCH.sub.t] and [DIS.sub.t], being the two-year lag of the actual percentage year-to-year CPI inflation rate (P2) and the two-year lag of the average annual unemployment rate of the civilian labor force (U2). The choice of instruments is based on the findings that P2 and U2 are highly correlated with variables [MATCH.sub.t] and [DIS.sub.t], respectively, while not being significantly correlated with the error terms in the system. Variables P2 and U2 were obtained from the Council of Economic Advisors (2003, tables B-42 and B-64).
4. Empirical Findings
Estimating Equation 5 by IV, using the Newey-West correction for heteroskedasticity, yields
(6) [(UAGI/RAGI).sub.t] = 48.5 + 0.16 [zMAXFPTR.sub.t-1] - 1.12 [zAEPT.sub.t-1] - 7.34 [zAUDIT.sub.t-1] (+2.46) (-1.07) (-2.98)
0.0008 [zPEN.sub.t-1] + 5.78 [zDIS.sub.t] 0.87 [zMATCH.sub.t] - 0.26[(TF/TEN).sub.t-1] (-2.22) (+2.25) (-3.29) (-4.70)
-0.001 TREND, DW = 2.05, F = 10.29 (-2.02),
where z is the first-differences operator and terms in parentheses are t-values.
In Equation 6, the estimated coefficients on six of the seven explanatory variables are statistically significant at the 5% level or beyond with the expected signs. The coefficient for the variable TREND is negative and significant at nearly the 6% level. This marginally significant variable may reflect a variety of subtle changes in social attitudes whose explicit inclusion in the model would be very difficult. Finally, the F-ratio is significant at the 1% level.
The estimated coefficient on the MAXFPTR variable is positive and statistically significant at the 3% level. Thus, there is evidence that the maximum marginal federal personal income tax rate, per se, increased the DTE. This finding is effectively consistent with certain other studies, including the study of audits of individual tax returns by Clotfelter (1983) and more recent studies by Cebula (1997) and Saltz (2001). By contrast, the coefficient for the AEPT variable is both negative and statistically insignificant, implying that the AEPT did not influence the DTE.
The estimated coefficient on the AUDIT variable is negative and significant at the 1% level. Thus, it appears that higher IRS audit rates discourage tax evasion. In addition, the estimated coefficient on the PEN variable is negative and significant at the 5% level. Hence, there is evidence that higher IRS penalties on detected unreported income may have acted to reduce the DTE. Next, the estimated coefficient on the IMP technology variable, MATCH, is negative and significant at the 1% level. This result, which is consistent with Ali, Cecil, and Knoblett (2001) and Cebula (2001), implies that the IRS's improving IMP technology over time has reduced the DTE. The estimated coefficient on the DIS variable is positive, as hypothesized, and statistically significant at the 5% level. This result implies that as the public's level of dissatisfaction with government increases, the DTE increases.
Finally, the estimated coefficient on the (TF/TEN) variable is negative and statistically significant at well beyond the 1% level. This finding implies that as the tax-free interest rate yield on high-grade municipals increases relative to the taxable interest rate yield on 10-year Treasury notes, the DTE declines. This constitutes strong empirical support for the hypothesis introduced in this study that the higher the (TF/TEN) ratio, the less the extent of federal personal income tax evasion as households find tax-free municipals an increasingly attractive alternative to tax evasion, that is, one that enables them to reduce tax liabilities by legal tax avoidance.
Based on this study of the period 1975-1997, it appears that federal personal income tax evasion in the aggregate is a decreasing function of IRS audit rates, IRS penalties on detected unreported income, and IRS income-matching technology. In addition, there is evidence that the higher the maximum marginal federal personal income tax rate, the greater the degree of tax evasion. These results for AUDIT, MATCH, PEN, and MAXFPTR are consistent with tax evasion theory. Furthermore, it appears that the greater the public's dissatisfaction with government, the greater the extent of income tax evasion. Finally, there is strong empirical support for the central hypothesis investigated by this study, namely, that higher tax-free interest rate yields on high-grade municipal bonds relative to, say, taxable yields on 10-year Treasury notes reduce income tax evasion. Indeed, the option of tax-free bonds would seem to have created an environment in which legal tax avoidance can potentially replace illegal tax evasion.
Finally, as Gilman and Madura (2001, p. 39) observe, the tax-free feature of municipal bonds (issued to help finance various public projects for state and local governments) "enables municipalities to obtain funds at a lower cost ... investors are willing to accept a lower pre-tax return on municipals because they tend to be more concerned with the after-tax return." It is not clear that tax-free bonds are necessarily an efficient way for the federal government to subsidize state and local governments. However, despite the finding presented in Equation 6, it is beyond the scope of this brief study to determine whether the diminished tax evasion resulting from tax-free municipal bonds would yield sufficient tax revenues to offset the revenue loss to the Treasury from this form of subsidy.
(1) Thus, the choice between tax evasion and tax avoidance is expressible in terms of opportunity costs.
Ali, M. M., H. W. Cecil, and J. A. Knoblett. 2001. The effects of tax rates and enforcement policies on taxpayer compliance: A study of self-employed taxpayers. Atlantic Economic Journal 29:186-202.
Alm, James, Brian Jackson, and Michael McKee. 1992, Institutional uncertainty and taxpayer compliance. American Economic Review, 82:1018-26.
Alm, James, George H. McClelland, and William D. Schulze. 1999. Changing the social norm of tax compliance by voting. Kyklos 52:141-71.
Atkins, Frank J. 1999. Macroeconomic time series and the monetary aggregates approach to estimating the underground economy. Applied Economics Letters 6:609-11.
Bawley, Dan. 1982. The subterranean economy. New York: McGraw-Hill.
Cebula, Richard 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. American Journal of Economics and Sociology 56:173-86.
Cebula, Richard J. 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 53:401-15.
Clotfelter, Charles T. 1983. Tax evasion and tax rates: An analysis of individual returns. Review of Economics and Statistics 65:363-73.
Council of Economic Advisors. 2003. Economic report of the president, 2003. Washington, DC: U.S. Government Printing Office.
Feige, Edgar L. 1989. The underground economies: Tax evasion and information distortion. Cambridge: Cambridge University Press.
Feige, Edgar L, 1994. The underground economy and the currency enigma. Public Finance/Finances Publiques 49:119-36.
Gitman, Larry J., and Jeff Madura. 2001. Introduction to finance. Boston, MA: Addison-Wesley.
Internal Revenue Service (IRS). 1997. Data book. Washington, DC: U.S. Government Printing Office.
Internal Revenue Service (IRS). 2004. Tax statistics. Accessed 1 May 2003. Available http://www.irs.gov.
Panteghini, Paolo M. 2000. Tax evasion and entrepreneurial flexibility. Public Finance Review 28:199-209.
Pestieau, Pierre, Uri Possen. and Steve Slutsky. 1994. Optimal differential taxes and penalties. Public Finances/Finances Publiques 49:15-27.
Pyle, David. 1989. Tax evasion and the black economy. New York: St. Martin's Press.
Saltz, Ira S. 2001. An empirical note on tax auditing and the size of the underground economy in the United States, 1962-1980. International Review of Economics and Business 48:119-24.
Tanzi, Vito. 1982. The underground economy in the United States and abroad. Lexington, MA: Lexington Books.
Tanzi, Vito. 1983. The underground economy in the United States: Annual estimates. IMF Staff Papers 30:283-305.
Tax Foundation. 2004. Tar bites. Accessed 13 February 2004. Available http://www.taxfoundation.org/prtopincometable.html.
University of Michigan. 1998. Political attitudes and behavior. Accessed 10 December 1998. Available http:// www.isr.umich.edu.
Richard J. Cebula , Economics Department, Hawes Hall 208J, Armstrong Atlantic State University, Savannah. GA 31419, USA; E-mail email@example.com.
I am grateful to Ed Feige for data and to Richard Connelly for data collection and to Mike Toma, Rick McGrath. two anonymous referees, and Laura Razzolini for helpful comments and suggestions.
Received June, 2003; accepted April, 2004.
|Printer friendly Cite/link Email Feedback|
|Author:||Cebula, Richard J.|
|Publication:||Southern Economic Journal|
|Date:||Oct 1, 2004|
|Previous Article:||Employment, wages, and alcohol consumption in Russia.|
|Next Article:||Determinants of cross-border intellectual property rights enforcement: the role of trade sanctions.|