Tax-reporting implications of asymmetric treatment: direct subsidies vs. tax deductions.

ABSTRACT

The Treasury Department has criticized use of the federal income tax system to deliver indirect subsidies to taxpayers in the form of tax deductions, and recommended that all such deductions be eliminated. It recognized, however, that it would be necessary to replace some tax deductions

with direct subsidies. Using an experiment, we examined whether the form of the subsidy (tax deduction or direct subsidy) affects tax reporting. Despite the economic equivalence of the two forms of subsidy, we found that taxpayers who are disadvantaged by not receiving a direct subsidy reported more income than taxpayers disadvantaged by not receiving a tax deduction.

Keywords: subsidies; tax expenditures; tax compliance; horizontal equity.

Data Availability: Data used in this study are available from the authors upon request.

INTRODUCTION

It is generally accepted that a tax system characterized by asymmetric deductions is horizontally inequitable, and horizontal inequity has a negative effect on tax-reporting decisions (see, for example, Spicer and Lundstedt 1976; Spicer and Becker 1980; Hite 1988; Moser et al. 1995). In this study we use an experiment to examine whether tax-reporting decisions would change if tax deductions were replaced with economically equivalent direct subsidies. The negative effect of asymmetric deductions is significant because over the last 40 years, the federal income tax system has increasingly been used to achieve various economic and social objectives.

Tax expenditures include preferential tax treatment in the form of deductions, exclusions, credits, preferential rates, and tax deferrals (U. S. Congressional Budget Office 1981, 1). Tax expenditures result in reduced tax liabilities for the taxpayers receiving preferential treatment, providing them with indirect subsidies. An important characteristic of the tax expenditure concept is its equivalence to direct subsidies. In fact, the term "tax expenditure" is derived from the assumption that the goals of these favorable tax provisions could be accomplished by replacing them with direct expenditures (U.S. Congress 1985, 2). Although one form of subsidy might be preferred to the other for a number of reasons, including ease of enactment, ease of administration, and visibility to the public, it is generally recognized that tax expenditures (indirect subsidies delivered through the tax system) and direct subsidies constitute alternative means of achieving the same policy objectives. (1)

Using the federal income tax system to administer indirect subsidies through tax deductions was sharply criticized by the U.S. Treasury Department in its 1984 Report to the President that eventually resulted in the Tax Reform Act of 1986. Noting that indirect subsidies delivered through the tax system create complexity for taxpayers and undermine taxpayer morale, the Treasury Department recommended specifically that all such provisions be eliminated or curtailed. The Report was also careful to state, however, that the recommendation should not be construed to mean that none of the tax-favored activities were worthy of government support. It specifically recognized that were such a proposal enacted, many tax deductions would be replaced with direct government subsidies (U.S. Department of the Treasury 1984, 20).

The Treasury Department's concern with using the tax system to deliver indirect subsidies is based on the idea that the "growing use of the income tax to subsidize various forms of economic activity is a major source of the increase in the perceived lack of fairness of the tax system" (U.S. Department of the Treasury 1984, 16). Further, the perception that the tax system is unfair is often cited as a reason why taxpayers fail to voluntarily comply with the tax law. According to the Treasury Department, "If the basis for this belief [that taxes are fair] comes under suspicion, voluntary compliance...is jeopardized."

This study uses an experiment to examine the effect on tax-reporting decisions of replacing tax deductions with direct subsidies. The experimental design compares economically equivalent systems of asymmetric tax deductions (some subjects receive tax deductions and others do not), asymmetric direct subsidies (some subjects receive direct subsidies and others do not), and an equitable system (no deductions or subsidies).

Prior research has identified a number of dimensions of tax equity, including but not limited to the fairness of the taxpayer's burden compared to other similarly situated taxpayers (horizontal equity), the fairness of the taxpayer's burden compared to other taxpayers who are not similarly situated (vertical equity), and the fairness of the benefit received from the government for taxes paid (exchange equity) (for discussions of types of equity, see, for example, Leventhal 1976; Gerbing 1988). The suggestion that eliminating tax deductions will have a favorable effect on tax reporting is based on the idea that their elimination will increase horizontal equity.

Our results support the Treasury Department's suggestion that replacing tax deductions with direct subsidies would increase compliance. Despite the economic equivalence of the two forms of subsidy, subjects who are inequitably disadvantaged by not receiving direct subsidies reported more income than subjects who were inequitably disadvantaged by not receiving tax deductions.

The remainder of this paper is organized as follows. The second section provides background and hypothesis development. The third section describes the experimental method, and the fourth section presents the results. The last section provides a general discussion and concluding comments.

BACKGROUND AND HYPOTHESIS DEVELOPMENT

Congress has long used tax legislation to provide indirect subsidies to constituents in the form of income tax deductions or some other type of tax preference. As an alternative, the same policy objectives could be achieved by direct subsidies. Although tax deductions might be preferred to direct subsidies for reasons not related to the tax system, use of the tax system to deliver indirect subsidies could have unintended effects. This possibility was recognized by the Treasury Department (1984, 16):

The U.S. Government has long spent public funds in ways that many taxpayers question. While this may cause many to believe that their tax dollars are being wasted, it does not raise doubts about the equity of the tax system itself. The situation is very different when the tax system, rather than direct spending, is used to provide subsidies. Similarly situated taxpayers can pay considerably different amounts of tax, depending on how they earn and spend their income, and high-income families may pay tax on a smaller portion of their income than do poorer families. The result is a perception that the income tax itself is unfair.

This statement implies that replacing asymmetric tax deductions with asymmetric direct government subsidies will have a favorable effect on the perceived equity of the tax system, and consequently on tax reporting. This suggestion is interesting in light of the government's acknowledgment of the economic equivalence of tax deductions and direct government subsidies (U.S. Congress 1985, 2), and is inconsistent with the invariance assumption of expected utility theory that predicts two versions of a choice problem that are economically equivalent will elicit the same response.

The Treasury Department's position suggests that taxpayers will not understand the economic equivalence of the two alternatives. Consistent with this suggestion, prospect theory (Kahneman and Tversky 1979, 1984; Kahneman et al. 1986; Tversky and Kahneman 1981, 1986) provides a theoretical basis for predicting that taxpayers will respond differently when the subsidy is in the form of a direct payment vs. a tax deduction. Prospect theory maintains that the first phase of the choice process is framing the decision. A frame is the reference point from which a problem is solved, and a robust finding of this literature is that risk preferences vary depending on whether an outcome is framed as a gain or a loss. The framing effects of prospect theory have been studied by researchers from many disciplines and applied to many different types of decisions.

More recently, Levin et al. (1998), after reviewing the extensive literature on framing effects, proposed that many of the prior studies had utilized different operational definitions of framing and had actually been studying different underlying processes. They proposed a typology of three different types of framing effects: risky choice framing, attribute framing, and goal framing. The three types of framing differ in terms of what is framed, what the frame affects, and how the effect is measured. The deduction vs. subsidy issue is an example of goal framing.

Goal framing involves situations where attention may be focused either on the potential to provide a benefit or on the potential to avoid a loss, but in both frames the behavior has the same beneficial or negative consequences. Studies of goal framing have generally found that the loss frame has a stronger impact on responses than the gain frame (Levin et al. 1998). This effect is the result of a negativity bias in processing information, where negative information has a stronger impact on judgment than objectively equivalent positive information (Meyerowitz and Chaiken 1987; Levin et al. 1998).

Goal framing provides a basis for predicting that individuals would be more aggressive when a subsidy is provided in the form of a tax deduction than when it is provided in the form of a direct subsidy. A direct subsidy can be thought of as a potential gain, while the objectively equivalent tax deduction can be thought of as avoiding a potential loss. If a taxpayer receives a direct subsidy, then it increases his or her income. It is reasonable that the taxpayer would frame the possibility of receiving additional dollars as a potential gain. On the other hand, a tax deduction reduces the amount of tax that a taxpayer would otherwise pay. Again, it is reasonable that the taxpayer would frame the possibility of receiving a tax deduction as the potential to avoid paying out dollars or avoiding a potential loss.

Thus, while expected utility theory predicts that individuals will respond the same regardless of whether a subsidy is provided in the form of a tax deduction or direct payment because the two choices are economically equivalent, goal framing provides an alternative prediction. Goal framing does not dispute the economic equivalence of the choice, but rather bases its prediction on how individuals process the alternatives. Goal framing would suggest that taxpayers would process the information differently, with the deduction possibility having a stronger impact on judgment. Consistent with the Treasury Department's suggestion and the alternative prediction provided by goal framing, we hypothesize: Hypothesis: Subjects who are disadvantaged by not receiving direct subsidies will report a larger percentage of their incomes than subjects who are disadvantaged by not receiving tax deductions.

EXPERIMENTAL METHOD

A laboratory setting was used to test the research hypothesis. The subjects earned money based on their decisions during the experimental sessions, and were paid in cash at the end of the sessions. No deception was employed in the experiment, and the subjects had full information concerning tax rate, audit rate, penalty, and deductions or subsidies.

Experimental Design Overview

The subjects received income in the form of an endowment at the beginning of each period, and decided how much of that income to report. After making the reporting decision, each subject rolled a die to determine whether the tax return would be audited. Subjects paid a tax on reported income, and a tax plus penalty on unreported income if audited. Taxes collected each period were approximately doubled (2) and distributed back to the subjects as public goods. Asymmetric treatment was achieved in one session by allowing a tax deduction for only half of the subjects. Another session replaced the tax deduction with a direct subsidy for half the subjects, holding wealth constant.

Procedure

The subjects were 30 university student volunteers. Ten subjects participated in each experimental session. (3) Upon entering the lab, each subject selected a station. Experimental packets (identical in appearance) were randomly distributed before the subjects entered the lab. The packets consisted of instructions, assignment to treatment, report sheets, and an earnings report.

The lab supervisor read the instructions along with the subjects. Participants were informed that the purpose of the research was to learn more about economic decision making, and if they followed the instructions carefully and made good decisions, they would earn money that would be distributed in cash at the end of the experiment. They were informed that they could refer back to the instructions at any time, and ask for clarification if the instructions were not clear. The subjects were asked to not talk to any other participants or make any general comments that might disclose information about their decisions or earnings during the experiment. The subjects were informed that the experiment would consist of a number of independent periods, but they were not informed about the length of the experiment or the number of periods. They were told that they would earn money during the experiment based on their decisions. Throughout the experiment, neutral terms were used. Although the use of contextually rich lan guage in experiments is sometimes desirable because of its potential for information, salience, motivation, or memory-coding effects (Haynes and Kachelmeier 1998), we avoided explicit reference to the tax setting. Our experimental design captured the distinguishing features of tax deductions vs. direct subsidies, and the additional information or motivation provided by a tax context might have introduced into the experiment additional variables (e.g., individual ethics and respect for or fear of authority). Wartick et al. (1999) provide evidence that the use of tax terminology affects subjects' economic decisions. The use of neutral terms allowed us to focus on the deduction vs. subsidy issue with respect to the economic incentives and variables controlled in the experiment.

The Basic Experiment

At the beginning of each period each subject was endowed with 1,200 Francs (experimental currency). The subjects had to decide how to allocate their 1,200 Franc income between two accounts (decide how much income to report on their tax returns and how much income they would not report). One account was a risk-free account on which the subject was required to pay a 30 percent commission (report income and pay a 30 percent tax). The other account was a risky account on which the subject would pay either no commission or a 60 percent commission (choose not to report income and keep all income if not audited, or pay the 30 percent tax plus a 100 percent penalty if audited). After all subjects recorded their decisions, assistants watched each subject as he/she individually rolled a ten-sided die and recorded the result of the die toss. If a zero was rolled, the subject's allocation was examined (tax return was audited), and the amount allocated to the risky account was subject to the 60 percent commission (the amo unt not reported on the tax return was subject to tax plus penalty at an effective rate of 60 percent). The subject's payoff before redistribution was the initial endowment less commissions (income less taxes and penalties).

The subjects were told that they would also receive dividends (public goods) each period. At the end of each period, the supervisor determined the total commissions (taxes) paid during the period. The total dividends (public good) available for distribution were determined by multiplying the total commissions (taxes) collected during the period by a factor. (4) The total commissions (taxes) collected, the total dividend (public good) available for distribution, and the individual dividend (public good) were announced at the end of each period. Thus, the subject's total payoff each period was the initial endowment after commissions (income after taxes and penalties), plus the subject's share of the total dividend (public good).

After all periods were completed, the subjects' earnings were converted from Francs to dollars at the rate of .06 cents, and the subjects were paid in cash. Finally, the subjects completed a debriefing questionnaire.

One session followed this basic experimental design and is referred to as the equitable treatment because all subjects were treated the same. The other two sessions manipulated an asymmetric deduction or subsidy.

Tax Deduction

The neutral instructions for this session stated that for the purpose of calculating the commission on the risk-free account (calculating the tax), some of the subjects would receive a discount (deduction) of 400 Francs and some of the subjects would receive no discount (deduction). The subjects were not told how many of them would receive the discount (deduction). In fact, only half of the subjects received the discount (deduction). After the instructions were read, subjects learned privately whether they would receive the discount (deduction). Subjects remained in the same experimental condition for all periods.

Direct Subsidy

The subjects in this session were told that after all commissions (taxes) for the period had been paid, some of the subjects would receive an additional allotment (direct subsidy) of 120 Francs and some of the subjects would not receive the additional allotment (direct subsidy). The subjects were not told how many of them would receive the additional allotment (direct subsidy). In fact, only half of the subjects received the additional allotment (direct subsidy). Again, after the instructions were read, subjects learned privately whether they would receive the additional allotment (direct subsidy), and subjects remained in the same experimental condition for all periods.

The amount of the deduction and direct subsidy were such that wealth was held constant between the deduction and subsidy treatments. Maximum theoretical earnings for all experimental conditions are presented in Table 1.

RESULTS

Results from the post-experimental questionnaire indicated that no subjects thought the experiment was about tax reporting. The subjects confinned that they made their decisions to try to maximize earnings, and they considered themselves adequately compensated for the time spent in the experiment. Overall, 17 periods were run during the equitable treatment session, and 21 periods were run for each of the other sessions. The average payout for the 90-minute session was $15.05. The standard deviation of the payments was $1.93.

In this experiment, the experimental sessions were designed such that there was common knowledge of group outcomes, creating cross-sectional dependence and making statistical analysis of the results problematic. The most conservative interpretation of this design is that since there is a potential for each session to develop its own idiosyncratic "personality," the results should be viewed as one observation per treatment. In this particular study, however, we would argue that while it is common for interactive sessions to develop their own personalities, it is not likely that this tendency would lead to predictable individual differences within a session. In other words, while the cross-sectional dependence might lead to either high or low reporting across sessions, there is no reason to predict that it would cause differences between the advantaged and disadvantaged subjects within sessions. Thus, our interest in the relative reporting rates of the disadvantaged vs. advantaged subjects within sessions miti gates somewhat the cross-sectional dependence problem.

The mean percentage of income reported within each experimental treatment by the advantaged and disadvantaged subjects is shown in Panel A of Table 2. Panel A reports means separately for the entire experiment, the last ten rounds, and the last five rounds. However, because the results are robust, discussion is in terms of the entire session. Panel B provides the mean percentages of income reported by each subject. (5)

Implicit in the suggestion that tax reporting would increase if tax deductions were replaced with direct subsidies is the assumption that underreporting occurs in a tax system with asymmetric deductions. This is based on the idea that taxpayers who believe that they are being treated inequitably (by not receiving an available tax deduction) will report less income. Our results support this assumption. The disadvantaged/deduction treatment group reported a mean percentage of 4.41, and the equitable treatment group reported 7.98 percent. The difference in means is statistically significant (p < .05 using a t-test and p < .10 using a Wilcoxon Z-test). This result is consistent with the idea that individuals who perceive that they are being treated inequitably will act to restore equity, supporting the Treasury Department view that tax deductions are perceived to be inequitable and lead to noncompliance.

Our research question, however, addresses whether reported income would increase if tax deductions were replaced with direct subsidies. We predict that a larger percentage of income will be reported when taxpayers are inequitably disadvantaged by not receiving an available direct subsidy than when taxpayers are inequitably disadvantaged by not receiving an available tax deduction. The results support this hypothesis. In Table 3, we report the average compliance rate of the disadvantaged participants across periods for the deduction and subsidy treatments.

The mean reported income of those in the disadvantaged/deduction treatment was 4.41 percent. The mean for the disadvantaged/subsidy treatment was 33.26 percent. Subjects inequitably disadvantaged by not receiving a direct subsidy reported more income than subjects inequitably disadvantaged by not receiving a tax deduction (p < .10 using a t-test and p < .05 using the Wilcoxon Z-test). This result is consistent with the Treasury Department's suggestion that replacing tax deductions with equivalent direct subsidies would have a positive effect on the tax system.

It is interesting to note that the participants in the equitable treatment group, the disadvantaged/deduction group, and the disadvantaged/subsidy group were all treated the same with regard to receiving preferential treatment. None of the participants received a deduction or a subsidy. The difference in how subjects were treated is that in the deduction and subsidy sessions, there were other participants in the session who received tax deductions or direct subsidies.

Although we did not predict how much income the disadvantaged/subsidy subjects would report relative to the equitably treated subjects, the disadvantaged/subsidy subjects reported more income (p < .01 using either a t-test or Wilcoxon Z-test). The disadvantaged/subsidy subjects reported 33.26 percent, and the equitably treated subjects reported 7.98 percent. One explanation for this result is the way subjects view direct subsidies. If subsidies are viewed as potential gains, then it is reasonable that even subjects who do not receive the direct subsidies perceive a larger collective "gain" as part of the terms of trade with the government. If so, a system of direct subsidies may encourage taxpayers to report more income because such a system is perceived to provide more collective benefits in exchange for taxes paid. (6)

To confirm the results reported above, we ran a repeated measures ANOVA with treatment, whether the subject was disadvantaged, and an interaction term as independent variables. Gender and age were also included to control for their potential effect. Age was found to be insignificant and dropped from the analysis. Table 4 shows the distribution of subjects by age and gender across treatments.

The repeated measures ANOVA results are reported in Panel A of Table 5, and are consistent with the means tests. We predicted that subjects disadvantaged by not receiving a direct subsidy would report a greater percentage of their income than subjects disadvantaged by not receiving a tax deduction. The repeated measures ANOVA results are consistent with this prediction. The results show that while the percentage of income reported does not differ depending on either treatment or whether the subject was disadvantaged, the interaction of treatment with disadvantaged is significant (F = 4.29, p < .05). We followed up the ANOVA results with additional analysis, and least square means and t-tests of differences are reported in Panel B of Table 5.

The results presented in Table 5, Panel B are also graphically shown in Figure 1. Figure 1 depicts the relative treatments. For the deduction treatment, the disadvantaged subjects reported significantly less than the advantaged subjects (3.29 and 18.68 percent, respectively; t = 4.9659, p < .01). For the subsidy treatment, the disadvantaged subjects reported significantly more than the advantaged subjects (32.15 and 14.86 percent, respectively; t = 5.5264, p < .01). differ (14.86 and 18.68 percent, respectively, t = 1.2219). it would cause the pattern of results observed here.

Also in view of the cross-sectional dependence issue, we reanalyzed our data including a variable for dividend distributions in the prior period. Because this variable is the same for every member of the experiment for that period, it is collinear with period, and we were unable to estimate a repeated measures ANOVA. Including the variable in an ANOVA, however, provided results that were identical to those without the variable, indicating that the lagged dividend distribution was not significant.

A research design issue that experimentalists have had to deal with in tax-reporting studies and the one that led to the cross-sectional dependence problem in this experiment, is the public good issue. The standard model of taxpayer compliance treats reported income as a function of the taxpayer's income, tax rate, audit rate, and penalty (Allingham and Sandmo 1972). Other researchers have recognized, however, that the standard model is incomplete in that it ignores the fact that tax revenues are used to provide public goods. Based on the idea that a tax system with no public good is an inequitable exchange, a number of studies have shown that the presence of a public good affects tax-reporting behavior (see, for example, Aim et al. 1992; Becker et al. 1987; Bordignon 1993; Cowell and Gordon 1988). The research design of this study incorporated a public good into the experiment through the redistribution of taxes collected.

While inclusion of a public good represents a more complete model of taxpayer reporting, it complicates the analysis because the reporting decision is not simply a trade-off between the potential costs of evasion and the gain from underreporting, but is now also made within the context of the fundamental public good dilemma. Ledyard (1995) describes the typical public good experiment as follows: Four individuals each are given an endowment of $5 and told that they can invest some or all of their endowment in a group project. Without discussion, each individual places an amount in an envelope. The experimenter collects the contributions, doubles the amount collected, and divides the money among the group. No subject knows others' contributions, but all know the total contributions. There are alternative theories about what to expect in a public good experiment. The economic/game-theoretic prediction is that no one will ever contribute anything to the public good because each subject will attempt to free ride (a Nash equilibrium). This is the classic public good problem because the group would be better off if all contributed the entire $5 (each subject would take home $10). The sociologic/psychologic prediction is that each subject will contribute something. The basis for this prediction is that altruism, social norms, or group identification will lead to the optimal group outcome.

As stated before, the problem that arises in interpreting the results of experiments with a public good feature is with cross-sectional dependence. The reporting decisions of the individuals making up each treatment group are not independent because of their common knowledge of group outcomes. This means that there is the potential for each session to develop its own personality such that each session should be treated as a single observation. Although it would generally be necessary to have multiple sessions for each treatment, for this particular experiment our interest in the relative reporting rates of the disadvantaged vs. advantaged subjects within sessions mitigates the problem. This continues to be an issue, however, and one that future researchers should carefully consider in tax-reporting studies.

CONCLUSION AND DISCUSSION

The report issued by the Treasury Department prior to enactment of the Tax Reform Act of 1986 made recommendations based on the idea that more income would be reported if tax deductions were replaced with direct subsidies. The effect on tax reporting of the form of the subsidy was tested in an experiment.

Our results support the Treasury Department's prediction. The subjects who were inequitably disadvantaged because they did not receive a direct subsidy reported significantly more income than the subjects who were inequitably disadvantaged because they did not receive a tax deduction. The results are consistent with goal framing, which is based on the idea that taxpayers view a possible tax deduction as avoiding a potential loss and view a possible direct subsidy as a potential gain. This result is especially strong given the artificial environment of the laboratory where the economic equivalence of tax deductions and direct subsidies is likely apparent to the subjects. The connection that taxpayers make between tax deductions and direct subsidies in the real world may not be as direct.

Although laboratory experiments cannot provide measures of absolute levels of compliance, a disappointing result was the very low level of reported income. (7) We believe that the low percentage of reported income was due to the relatively low audit rate used in this study, and the fact that subjects viewed the decision as an investment decision with no criminal penalties or social sanctions for not reporting.

The methods used in this study offer some advantages with regard to generalizability. A neutral context allows the theory to be generalized to other applications with the same economic characteristics. For example, the differences found in the deduction/subsidy treatment may be applicable to the context of salaries vs. fringe benefits, and may explain the low participation in flexible spending plans. On the other hand, recent findings by Wartick et al. (1999) indicate that although the neutral context allows for strong internal control, contextual variables that are not present in an experimental economics setting potentially affect tax-reporting decisions. A limitation of this study is that it is unknown if variables associated with the tax context (e.g., political orientation, respect for or fear of authority, or personal ethics), as well as economic variables controlled in the study such as wealth, might also affect the deduction vs. subsidy issue. A final issue related to the experimental context was the use of an investment context rather than a purely abstract setting. Although the level of abstractness necessary to inhibit role-playing is unknown, the investment context may have invoked investment schemas in some of the subjects.

A limitation of this study concerns the size of the economy. It was necessary to hold the number of subjects constant across all experimental sessions in order to control marginal per capita return. From an external validity perspective this is analogous to constraining growth in the economy. An extension might manipulate the size of the economy. An additional limitation, also arising from holding wealth constant, was that output was constrained such that it was constant across all subjects. Collins and Plumlee (1991) found that the subjects who underreported were also willing to exert more effort to earn more income.

In conclusion, our results provide evidence that replacing tax deductions with direct subsidies could increase compliance. In addition, our results contribute to the tax-reporting literature by providing additional support of a causal relation between horizontal inequity and noncompliance.

[FIGURE 1 OMITTED]

TABLE 1

THEORETICAL MAXIMUM PAYOUTS

Panel A: (Total tax collected * 2) MPCR (a) = .21

                                  Disadvantaged          Equitable
                                 No           No      No Deduction
                              Deduction     Subsidy    of Subsidy

Income                          F1200        F1200
Deduction                           0            -
Income subject to tax            1200         1200
Tax rate                          .30          .30
Tax                               360          360
Income after tax                  840          840
Direct subsidy                      -            0
Income before public good         840          840
Public good                       600          600
Final payout per round           1440         1440

                                  Advantaged
                             Received    Received
                            Deduction    Subsidy

Income                        F1200       F1200
Deduction                       400           -
Income subject to tax           800        1200
Tax rate                        .30         .30
Tax                             240         360
Income after tax                960         840
Direct subsidy                    -         120
Income before public good       960         960
Public good                     600         600
Final payout per round         1560        1560

Panel B: (Total tax collected * 1.8333) MPCR = .195

Income                     F1200
Deduction                      0
Income subject to tax       1200
Tax rate                     .30
Tax                          360
Income after tax             840
Direct subsidy                 0
Income before public good    840
Public good                  660
Final payout per round      1500

(a) Marginal per capita return (MPCR) is the return an individual would
receive if he or she invested in the public good and no one else
invested in the public good, relative to the private good. In this
experiment the private good is equivalent to the probability of keeping
all income. The details of how MPCR was calculated are included in
footnote 4.

TABLE 2

MEAN PERCENTAGES OF REPORTED INCOME

Panel A: Mean Percentages of Income Reported by Disadvantaged and
Advantaged Subjects within Treatments

                   Deduction Treatment                      Subsidy
                                               Equitable    Treatment
           Disadvantage  Advantage   Total     Treatment  Disadvangage

All            4.41        19.80     12.10       7.98        33.26
  periods     n = 5        n = 5     n = 10     n = 10       n = 5
Last 10        2.50        20.04     11.27       6.96        35.50
  periods     n = 5        n = 5     n = 10     n = 10       n = 5
Last 5         1.33        20.01     10.67       6.58        35.33
  periods     n = 5        n = 5     n = 5      n = 10       n = 10

             Subsidy Treatment
           Advantage   Total     Total

All          20.35     26.81     16.15
  periods    n = 5     n = 10    n = 30
Last 10      20.31     27.90     15.38
  periods    n = 5     n = 10    n = 30
Last 5       24.77     30.05     15.77
  periods    n = 5     n = 5     n = 10

Panel B: Mean Percentages of Income Reported by Each Subject within
Treatments

 Deduction       Equitable Treatment        Subsidy
 Treatment                                 Treatment
  Subject        Mean       Subject       Mean       Subject

Disadvantage                                       Disadvantage
     1            0            1          0             1
     2           22.2          2          0             2
     3            0            3          0.10          3
     4            0            4          8.30          4
     5            0            5         22.55          5
 Advantage                                          Advantage
     6            0            6          0             6
     7           33.33         7         16.42          7
     8            0            8         24.61          8
     9           31.75         9          5.90          9
    10           33.92        10          1.00         10

                Subsidy
               Treatment
  Subject        Mean

Disadvantage
     1           0
     2          31.94
     3          35.32
     4           1.43
     5          97.62
 Advantage
     6          51.25
     7          15.48
     8           0.01
     9          35.02
    10           0

TABLE 3

MEANS TESTS TAX DEDUCTION VS. DIRECT SUBSIDY AVERAGE COMPLIANCE RATES OF
DISADVANTAGED SUBJECTS COMPARISON OF DISADVANTAGED/DEDUCTION TREATMENT
TO DISADVANTAGED/SUBSIDY TREATMENT

                 Disadvantaged/  Disadvantaged/  t-test of
                   Deduction        Subsidy      difference

All periods           4.41           33.26        -1.58 *
                     n = 5           n = 5
Last 10 periods       2.50           35.50        -1.79 *
                     n = 5           n = 5
Last 5 periods        1.33           35.33        -1.85 **
                     n = 5           n = 5

                 Wilcoxon Z-test
                  of difference

All periods        -1.7827 **

Last 10 periods    -1.4120 *

Last 5 periods     -1.4100 *

*, **p < .10 and p < .05, respectively, one-tailed tests.

TABLE 4

DEMOGRAPHIC CHARACTERISTICS OF SUBJECTS BY TREATMENT

            Deduction  Equity  Subsidy

Gender
  Male          6        9        5
  Female        4        1        5
  Total        10       10       10
Age
  Under 21      5        3        3
  21-25         3        3        7
  26-29         1        3        0
  30-35         0        1        0
  Over 35       1        0        0
  Total        10       10       10

TABLE 5

ANOVA RESULTS

Panel A: Repeated Measures ANOVA

                           Degrees
                             of     Sum of    Mean
Source                     Freedom  Squares  Square  F-Val  Significance

Between-Subjects Effects
  Treatment                   2      1.4483  0.7242  1.25      0.3039
  Disadvantaged               1      0.0122  0.0122  0.02      0.8855
  Treatment*Disadvantaged     1      2.4818  2.4818  4.29      0.0492
  Gender                      1      5.2192  5.2193  8.73      0.0071
  Error                      24     13.8803  0.5783
Within-Subjects Effects
  Period                     16      0.3388  0.0243  1.08      0.3729
  Period*Treatment           32      0.6388  0.1996  0.89      0.5384
  Period*Disadvantaged       16      0.3407  0.0213  0.95      0.5164
  Period*Treatment*          16      0.3570  0.0223  0.99      0.4656
  Disadvantaged
  Period*Gender              16      0.5207  0.0325  1.45      0.1174
  Error                     384      8.6451  0.0225

Panel B: Least Squares Means


                Subsidy      Subsidy       Deduction    Deduction
                Advantage    Disadvantage  Advantage    Disadvantage

Subsidy
  Advantage     14.86
Subsidy
  Disadvantage   5.5264 ***  32.15
Deduction
  Advantage      1.2219      -4.3433 ***   18.68
Deduction
  Disadvantage  -3.6997 ***  -9.3091 ***   -4.9659 ***  3.29
Equity          -0.4840      -6.5541 ***   -1.8419 *    3.5458 ***


                Equity

Subsidy
  Advantage
Subsidy
  Disadvantage
Deduction
  Advantage
Deduction
  Disadvantage
Equity          13.42

*, **, ***Significant at p < . 10, p < .05., and p < .01, respectively.

Least squares means reported on the diagnol. T-test of difference in
least squares means reported off the diagonal.

We thank Andy Cuccia, Mark Isaac, Ron King, Silvia Madeo, and Jeff Schatzberg for helpful comments. We also thank Fran Ayres and two anonymous reviewers for their contributions. We are grateful to The University of Arizona Economic Science Laboratory and the University of Missouri--St. Louis School of Business Administration for their financial support of this project.

Submitted: June 1999

Accepted: October 2002

(1.) As noted previously, indirect subsidies can be in the form of any type of preferential tax treatment. Although we specifically investigate tax deductions and use that term throughout the remainder of this paper, the discussion is applicable to other forms of preferential treatment.

(2.) In the deduction and subsidy treatments, taxes collected were multiplied by 2.0. In the equitable treatment, taxes were multiplied by 1.8333. The different multipliers were to control for wealth.

(3.) In accordance with the findings of Isaac et al. (1984), Isaac and Walker (1988), and Isaac et al. (1994) the number of subjects was kept constant across experimental sessions. If the number of subjects in the experiment varied across sessions, then the payoff would need to change to keep the marginal per capita return constant. However, if both number of participants and payoffs varied, then there would be a net wealth effect to the subjects. Thus, marginal per capita return and wealth are controlled by keeping the number of participants constant across all experimental sessions.

(4.) Applying a multiple to the total amount contributed is standard in public good experiments and reflects the consumer surplus of the public good. Taxes collected are doubled in the deduction and subsidy treatments, resulting in a marginal per capita return of 0.21. In the equitable treatment, taxes are multiplied by 1.8333 to control for wealth in the experimental design, resulting in a marginal per capita return of 0.195. Marginal per capita return is calculated for subjects in the deduction and subsidy treatments as follows: the numerator is ($1 x 2)/10 or .20, reflecting that these subjects would get back 20 cents for each dollar paid in taxes if no one else paid any taxes. Because the subjects face a 10 percent chance of a 60 percent penalty on income not reported, the denominator is .94, calculated as ((.90 x $1) + (.10 X $.40)). For subjects in the equitable treatment, the numerator would be ($1 x 1.8333)/10, and the denominator would be the same as for the subjects in the deduction and subsidy trea tments.

(5.) As can be seen in Panel B of Table 2, one subject in the disadvantaged/subsidy group reported all of his or her income every period except one. If this subject is removed from the analysis, the mean for the disadvantaged/subsidy treatment is 17.17 percent. The inclusion or exclusion of this subject does not change the results with regard to reporting significant differences.

(6.) It should be noted that this is a perception and not reality. Lowering the recipient's tax liability by an equal amount could confer the same benefit. We believe that reducing the tax liability would not be viewed as a gain, but as reducing a loss.

(7.) The means reported in this study are much lower than means reported in other experimental studies of taxpayer compliance. For example, Aim et al. (1992) report compliance levels of between 20 and 35 percent. An important difference in their study was that audits were conducted for the current and four prior periods. We believe our low compliance rate was due to the low audit and penalty rates.

REFERENCES

Allingham, M., and A. Sandmo. 1972. Income tax evasion: A theoretical analysis. Journal of Public Economics 1 (November): 323-338.

Alm, J., B. Jackson, and M. McKee. 1992. Estimating the determinants of taxpayer compliance with experimental data. National Tax Journal 45 (March): 107-114.

Becker, W., H. Buchner, and S. Sleeking. 1987. The impact of public transfer expenditures. Journal of Public Economics 34: 243-252.

Bordignon, M. 1993. A fairness approach to income tax evasion. Journal of Public Economics 52: 345-362.

Collins, J. H., and R. D. Plumlee. 1991. The taxpayer labor and reporting decisions: The effect of audit schemes. The Accounting Review 66 (July): 559-576.

Cowell, F. A., and J. P. F. Gordon. 1988. Unwillingness to pay: Tax evasion and public good provision. Journal of Public Economics 36: 305-321.

Glerbing, M. 1988. An empirical study of taxpayer perceptions of fairness. Unpublished dissertation, The University of Texas at Austin.

Haynes, C. M., and S. J. Kachelmeier. 1998. The effects of accounting context on accounting decisions: A synthesis of cognitive and economic perspectives in accounting experimentation. Journal of Accounting Literature 17: 97-136.

Hite, P. 1988. The effect of peer reporting on taxpayer compliance. The Journal of the American Taxation Association (Spring): 47-64.

Isaac, R. M., J. Walker, and S. Thomas. 1984. Divergent evidence on free riding: An experimental examination of possible explanations. Public Choice 43 (1): 113-149.

-----, and -----. 1988. Group size effects in public goods provision: The voluntary contributions mechanism. Quarterly Journal of Economics 103: 179-201.

-----, -----, and A. W. Williams. 1994. Group size and voluntary provision of public goods: Experimental evidence utilizing large groups. Journal of Public Economics 54: 1-36.

Kahneman, D., and A. Tversky. 1979. Prospect theory: An analysis of decision under risk. Econometrica (March): 263-291.

-----,and -----. 1984. Choices, values and frames. American Psychologist (April): 341-350.

-----, J, Knetsch, and R. Thaler. 1986. Fairness and the assumption of economics. Journal of Business (October): 285-300.

Ledyard, J. 1995. Public goods: A survey of experimental research. In The Handbook of Experimental Economics, edited by J. Kagel, and A. Roth. Princeton, NJ: Princeton University Press.

Leventhal, G. 1976. Fairness in social relationships. In Contemporary Topics in Social Psychology, edited by J. Thibaut, and R. Carson. Morristown, NJ: General Learning Press.

Levin, I., S. Schneider, and G. Gaeth. 1998. All frames are not created equal: A typology and critical analysis of framing effects. Organizational Behavior and Human Decision Processes 76 (2): 149-188.

Meyerowitz, B., and S. Chaiken. 1987. The effect of message framing on breast self-examination attitudes, intentions, and behavior. Journal of Personality and Social Psychology 52: 500-510.

Moser, D. V., J. H. Evans, and C. K. Kim. 1995. The effects of horizontal and exchange inequity on tax-reporting decisions. The Accounting Review 70 (October): 619-634.

Spicer, M., and S. Lundstedt. 1976. Understanding tax evasion. Public Finance 31 (2): 295-305.

-----, and L. Becker. 1980. Fiscal inequity and tax evasion: An experimental approach. National Tax Journal 33 (June): 171-175.

Tversky, A., and D. Kahneman. 1981. The framing of decisions and the psychology of choice. Science 211: 453-458.

-----,and -----. 1986. Rational choice and framing of decisions. Journal of Business 59: 251-278.

U. S. Congress. 1985. Estimates of Federal Tax: Expenditures for Fiscal Years 1986-1990. Washington, D.C.: U.S. Government Printing Office.

U. S. Congressional Budget Office. 1981. Tax Expenditures: Current Issues and Five-Year Budget Projections for Fiscal Years 1982-1986. Washington, D.C.: U.S. Government Printing Office.

U. S. Department of the Treasury. 1984. Tax Reform for Fairness, Simplicity and Economic Growth. Washington, D.C.: U. S. Government Printing Office.

Wartick, M., S. Madeo, and C. Vines. 1999. Reward dominance in tax-reporting experiments: The role of context. The Journal of the American Taxation Association (Spring): 20-31.

Cynthia C. Vines is an Associate Professor at the University of Kentucky and Martha L. Wartick is an Associate Professor at the University of Northern Iowa.