The Impact of Mental Accounting on Tax Evasion (An Experimental Study in Accounting Students)

* Corresponding author, email address: 1aprilia.kardianzah.putri95@gmail.com, achdiar.redy17@gmail.com ABSTRACT The purpose of this study is to fi nd out the impact of mental accounting on tax evasion decisions. This research is a quantitative research using experimental research methods. Analysis was conducted using 2x3 between subjects experimental design. The sample used in this experiment was accounting students at Trunojoyo University, Madura, with concentration on taxation. The results for all treatments in this study indicate that there are no differences between in gain scenario and loss scenario. These fi ndings can be explained by mental accounting and prospect theory. This study shows that individuals evaluate tax payments and tax refunds asymmetrically. In addition, the individuals also underlie mental accounting when making tax evasion decisions.


INTRODUCTION
For most developed and developing countries, taxes are one of the important elements in managing the national income. For Indonesia, as a developing country, taxes are also an important element to support the state revenue budget. Almost 80% of the state revenue in the state budget comes from the tax sector which is used to support the Indonesian economy (Suparmanto, 2013). The importance of state revenue from the tax sector makes the government increase tax revenues every year.
The low awareness of taxpayers and the high level of tax manipulation are the factors that cause a reduction in the target of tax revenues, where taxpayers do not pay the tax burden owed according to what has been charged. In addition, the less optimal tax revenue, one of which is infl uenced by the practice of tax planning which aims to reduce the amount of tax burden that must be paid. Tax planning is divided into two: tax avoidance and tax evasion.
Taxpayers will be more likely to choose to commit tax evasion than tax avoidance (Sari, 2015). This is because committing tax avoidance requires insight and detailed understanding of tax regulations so that taxpayers can fi nd gaps that can be penetrated to reduce the tax burden paid without violating the tax regulations. Meanwhile, tax evasion is easier for taxpayers to do, even though they have to violate regulations. In addition, tax evasion is an act that violates tax regulations by reporting understatement of income or overstatement of the deductions in the Annual Tax Return. Tax evasion is a phenomenon that is very diffi cult to observe and examine.
The diffi culty of observation cannot be separated from the diffi culty of controlling and verifying the behavior of taxpayers. Although there have been many studies on tax evasion, few have focused on tax evasion in terms of profi ts and losses (Fochmann and Wolf, 2015). The fact is that taxpayers can commit tax evasion by reporting understatement of income or overstatement of the deductions. On the other hand, according to the research conducted by Fochmann and Wolf (2015), in an effort to avoid tax, people prefer underreporting income to over-deduction of expenses. This can be explained by mental accounting in paying taxes and tax returns. The results show that mental accounting plays an important role in tax evasion decisions. Mental accounting is a fi nancial behavior or economic behavior when a person classifi es inputs and outputs based on posts, such as accounting models.
This research is a replication of the research conducted by Fochmann and Wolf (2015). This is because there are still only few experimental studies on tax evasion. This research is used as the basis for knowing the behavior of tax evasion when it is faced with profi t scenario and loss scenario. The difference between this research and the previous research lies in the sample used, that is, Accounting Students at Trunojoyo University, Madura, with concentration on taxation. This research is important and useful in preparing strategies for increasing taxpayer compliance, and as important source of information for political and economic observers.

Prospect Theory
Daniel Kahneman and Amos Tversky were the initiators of Prospect Theory in the early 1980s that covered two scientifi c disciplines: psychology and economics or often referred to as psycho-economics (Mahastanti and Wiharjo, 2012). Prospect Theory is an alternative account of individual decision making under risk (Kahneman and Tversky, 1979). This theory was developed to explain a person's reasons for making certain decisions from his psychological side.
In Prospect Theory, the picture adopted by a person can infl uence his decision, and in conditions of uncertainty the person will choose the choice that will produce the biggest expected utility. Experimental research conducted by Kahneman and Tversky (1979) shows that attitudes about the risk of facing gain would be very different from the attitude about the risk of facing loss. The value function is defi ned in the form of gains and losses. The function of the value explains that someone, in making decisions, tends to risk averse when in a state of gain and risk seeking when in a state of loss.
Prospect Theory shows that people will have irrational thinking tendencies to risk more profi ts than losses. In conditions of loss, people will tend to be more determined to bear the risk than in a successful condition. Prospect Theory can be used to see human behavior in a decision-making process that is sometimes absurd (Mahastanti and Wiharjo, 2012).

Behavioral Finance Theory
Behavioral fi nance studies how psychological phenomena affect fi nancial behavior (Shefrin and Statman, 2000). In another study, Nofsinger (2001) defi nes that behavioral fi nance studies how humans actually behave in a fi nancial setting. Thus, behavioral fi nance studies the process on how individual makes decisions in fi nancial terms. Behavioral fi nance infl uences fundamental decisions and decision-making that seems good, and turns bad, and vice versa (Tjandrasa, 2014).

Mental Accounting
Thaler (2008), developed a mental accounting theory based on the concept of psychological accounts introduced by Kahneman and Tversky (1979) through Prospect Theory. Mental accounting is a description of the way a person performs an accounting process that can only be learned by making observations about one's behavior or concluding rules that apply in society (Mahastanti and Wiharjo, 2012). According to Thaler (2008), mental accounting is a cognitive operation series used by individuals in coding, making criteria, and evaluating fi nancial activities. Mental accounting focuses on how individuals should respond to and evaluate a situation when there are two or more possible outcomes, especially on how to combine these results. Thaler and Shefrin (1981), defi ne mental accounting as economic behavior when one classifi es input and output based on posts such as account code according to behavioral life cycle theory. Mental accounting refers to the behavior or way of thinking of someone who has a tendency to group and apply money differently depending on where the money comes from (Thaler, 1999). Mental accounting plays an important role in decision making, how they should respond to and evaluate a condition when there are two or more possible outcomes.

Tax Evasion
Tax evasion is an effort carried out by taxpayers to alleviate the tax burden by violating laws (Mardiasmo, 2009). Mughal and Akram (2012) state that tax evasion can be described as an activity of taxpayers where they do not comply with and intentionally violate the law or violate tax laws with the aim of escaping from the payment of taxes that have become their obligation. So it can be concluded that tax evasion is carried out by taxpayers intentionally to reduce their economic burden to pay taxes.
According to Reskino et al (2012), citing the results of research conducted by the Head of Wonosari KPP for his doctoral dissertation (released at www.ugm.ac.id), the background of tax evasion actions is usually because they view taxes as a burden that will reduce one's economic capacity. Tax evasion will have an impact on people's lives, in terms of fi nance, economics, and psychology (Sumarsan, 2013).

Previous Studies and Formulation of Hypotheses
Based on research conducted by Kahneman and Tversky (1979), a person's attitude about risk in the face of gain is different from the attitude about risk in the face of loss. These results prove that different perceptions infl uence the decision process. The value function explains that someone, in making decisions, tends to risk averse when he is in a state of profi t and tends to risk seeking when he is in a state of loss.
Tax evasion is a phenomenon that is very diffi cult to observe and examine. Although there have been many studies on tax evasion, very few have focused on tax evasion on the side of profi ts and losses. According to Fochmann and Wolf (2015), people like avoiding taxes by underreporting income rather than by over-deduction of expenses. This can be explained by mental accounting in tax payment and tax returns. The results show that mental accounting plays an important role in tax evasion decisions. Therefore, researchers would like to reveal tax evasion behavior by replicating the research conducted by Fochmann and Wolf (2015), that is, by conducting experiments in which participants are faced with gain scenario and loss scenario. It is assumed that participants conduct positive income evaluations in gain scenario, and vice versa. Participants who face positive income in the gain scenario must pay taxes, but participants who face negative income in the loss scenario will receive a tax return. Although there have been no clear predictions based on theory, there is some empirical evidence that individuals avoid taxes more than profi ts. According to Chang et al. (1987), tax evasion behavior depends on whether the tax is framed as a loss or reduced profi t. In addition, according to Kirchler and Maciejovsky (2001), self-reported tax compliance depends on the gain and loss situation, whether faced with tax payments or tax returns. Torgler and Schaffner (2007) state that a higher tax spirit reduces tax evasion from profi t scenario rather than loss scenario. According to Torgler, (2012), taxpayers like avoiding taxes by reducing tax credit rather than by reporting income. So the hypothesis can be formulated as follows: H1: At the baseline treatment, someone would rather commit tax evasion in gain scenario than in loss scenario.
In observing tax evasion behavior, it does not need to separate between gain scenario and loss scenario. According to Thaler (2008), various studies show that someone underlies mental accounting, and the results of one's decisions depend on the decisions made. In the design that the researchers made, mental accounting refers to two different mental accounts as tax evasion decisions, namely gain scenario and loss scenario. In aggregation treatment, it still separates between gain scenario and loss scenario. This means that profi ts and losses continue to be isolated until the accumulation, payment of taxes, and the possibility of penalties that will be calculated by individuals for each scenario. In other words, even though payments are accumulated at the end, each individual decides to commit tax evasion in an overall independent gain scenario with a loss scenario decision, because the individual is not likely to realize the results of the accumulation when making the decision. From the above perspective, the following hypothesis can be formulated as follows: H2: At the aggregation treatment, someone would rather commit tax evasion in gain scenario than in loss scenarios.
In one tax declaration treatment, taxes and possible penalties are only calculated as taxable basis for profi t and loss and as accumulation after the tax evasion decision. Therefore, there is no signifi cant difference in tax evasion either reducing reporting profi ts or overestimating losses. According to Chang et al. (1987), taxpayers, in doing tax evasion, depends on how individuals view taxes as a deduction of certain gains or losses. In contrast to aggregation treatment, tax evasion will be more visible in the one tax declaration treatment. Therefore, participants will only use one mental account for this treatment. Thus, the mental accounting phenomenon no longer plays a role. As a result, in this study there is no difference in tax evasion behavior between in profi t scenario and in loss scenario. So the hypothesis can be formulated as follows: H3: At the one tax declaration treatment, there is no difference in tax evasion behavior between in gain scenario and in loss scenarios.

RESEARCH METHOD
This research is a quantitative research using experimental research methods. The experimental design used is a 2x3 betweensubject design. Participants are divided into three groups: baseline treatment, aggregation treatment and one tax declaration treatment. Grouping is done to make it easier for researchers to compare and analyze experimental results. Aggregation treatment is almost the same as baseline treatment. Aggregation treatment distinguishes only the amount of payment in the profi t scenario and the loss scenario that is accumulated in each period, while in the baseline treatment, the amount of payment in the gain scenario and the loss scenario is not accumulated. One tax declaration treatment is the same as the aggregation treatment, but the difference is in the tax reporting. In one tax declaration treatment, there is only 1 tax reporting, while at aggregation treatment, there are 2 tax reporting. The researchers used randomization to each treatment. With randomization, each participant received one of the three treatments. It was expected that each treatment would produce the same number of participants.
Participants in this experiment were accounting students who took concentration on taxation because researchers assumed that they had already understood tax issues. The number of participants in this experiment was 38 participants. This experiment was carried out in the laboratory room. The procedures in this experiment were: 1) the participants worked on the given command; 2) the researchers gave an explanation of the purpose of the experiment and guided the steps to work on the experiment and fi lled in the demographic data; 3) the researchers also provided experimental exercises for 2 periods.
In the Introduction section, the participants were told that they would get capital of IDR 5,000, -as initial capital, and the capital would increase and decrease depending on decisions and opportunities. Then participants were asked to fi ll in demographic data and read an overview of the gain scenario and loss scenario. Furthermore, participants were asked to determine the actual profi t before tax and the actual loss before tax for 5 periods using the free random number generator application.
After participants knew the actual profi t before tax and actual loss before tax, they were asked to do tax reporting for 5 periods in accordance with the tax reporting limits available on the instrument provided. At each period, 30% of their tax reporting would be audited and 70% would not be audited. If an audit occurred, a penalty would be imposed. The amount of penalty in the profi t scenario was twice as much as tax avoidance. Whereas in the loss scenario, the amount of penalty was twice as much as overpayment obtained from tax returns.
The result of period for audits on profi t scenario was after-tax profi ts minus penalties, and the result of loss scenarios was after-tax profi ts plus penalties. Whereas in the case of no audit, the result of period on profi t scenario was after-tax profi t, and on the loss scenario was after-tax loss.
At baseline treatment, after participants made a decision for 5 periods, 1 period would be drawn randomly. To determine whether there were gains or losses in the scenario, the participants were asked to throw the dice. If the dice stopped at numbers 1, 2 and 3, the participants were in a profi t scenario. But if the dice stopped at numbers 4, 5 and 6, the participants were in a loss scenario. Then the results were converted into money. So, if it was in a profi t scenario, the initial capital would increase. But on the contrary, if it was in a loss scenario, the initial capital would decrease. Whereas at aggregation treatment, the total yield period no longer used dice, but the total payment result from the experiment was done by subtracting the period results in the profi t scenario with the period results in the loss scenario. It was the same as at aggregation treatment, although in a one tax declaration treatment each participant faced profi ts and losses and reports gains and losses in each period, reporting gains and losses would be calculated to get the amount to be used as the basis for taxation. This calculation was carried out every period before the decision took place. So that only 1 tax reporting would be used.
The variables used in this study were mental accounting and tax evasion. Tax evasion behavior was analyzed by adopting research conducted by Fochmann and Wolf (2015). Data analysis technique used in this study was nonparametric statistical tests using different tests, such as the Wilcoxon Signed Rank test and the Mann-Whitney test. The Wilcoxon test was used to test the difference in two paired samples from two data whether there was a difference or not. While the Mann-Whitney test was conducted to determine the differences in two samples that were not related or paired with each other.
Manipulation checks in this study were seen from tax reporting made by participants for 5 periods whether in accordance with the reporting limits set or not. The fi rst manipulation check was reporting on profi t scenarios, while the second manipulation The results of the manipulation check showed that of 38 participants, 13 participants did not pass the manipulation check. So, the participant data that could be processed was only 25 people with a total of 125 data. The average number of male respondents was 32% and female was 72%.

Description of Variable
In this experiment, at baseline treatment group, the mean value of tax evasion was 46.5%. However, the level of tax evasion differs between in the gain scenario and in the loss scenario. In the profi t scenario, the mean value of tax evasion was 22.1% with a standard deviation of 0.180, while in the loss scenario, the mean value of tax evasion was only 24.4% with a standard deviation of 0.267.
At aggregation treatment group, the mean value of tax evasion was 27.2%. In the gain scenario, the mean value of tax evasion was 12.8% with a standard deviation of 0.126, while in the loss scenario the mean value of tax evasion was only 14.4% with a standard deviation of 0.147.
At tax declaration treatment group, the mean value of tax evasion was 37.8%. However the level of tax evasion differs between in the gain scenario and in the loss scenario. In the gain scenario, the mean value of tax evasion was 16.4% with a standard deviation of 0.22, while in the loss scenario, the mean value of tax evasion was 21.4% with a standard deviation of 0.25.

Results of Wilcoxon Signed-Rank Test
The Wilcoxon signed-rank test was used to analyze the differences between in the gain scenario and in the loss scenario (Egain and Eloss) at each treatment.

Source: Data Processed
The results of Wilcoxon signed-rank test showed that at all treatments there was no difference tax evasion decision both in gain scenario and in loss scenario.

Results of Mann-Whitney Test
The Mann-Whitney test was conducted to determine the differences in two samples that were not related or paired with each other.  Table 4, it can be seen that the value of 0.193 is obtained from Etotal baseline treatment (0.465) minus Etotal aggregation treatment (0.272), then for the calculation of Egain, Eloss and Edifference is also the same as Etotal. The results of the Mann-Whitney test above show that the value of sig. > 0.05 so that it can be said that there is no signifi cant difference between baseline treatment and aggregation treatment on the tax evasion ratio on Egain, Eloss, and Edifference. Meanwhile, for Etotal, the value of sig < 0.05. Thus Etotal has a difference in the amount of tax evasion in baseline treatment and aggregation treatment.
The results of the Mann-Whitney test on baseline treatment with one tax declaration treatment show that the value of sig > 0.05 so that it can be said that there is no signifi cant difference between the baseline treatment and the aggregation treatment on each measure of tax evasion, except on the tax evasion ratio in the gain scenario that has a signifi cance value of 0.016. The results of the Mann-Whitney test at baseline treatment with one tax declaration treatment can be seen on Table 5.
In addition, the results of Mann-Whitney test at aggregation treatment with one tax declaration treatment can be seen on Table 5.

Source: Data Processed
Based on the results of Mann-Whitney test (aggregation treatment with one tax declaration treatment) in the above table, there is no signifi cant difference in each measure of tax evasion. This can be seen in table 6 which shows that the sig value > 0.05 in Etotal, Egain, Eloss, and Edifference.

Results and Discussion
Based on the results of data analysis, the descriptive statistical test shows that the average participant commits more tax evasion in the loss scenario than in the gain scenario. This can be seen in table 2 where at baseline treatment in gain scenario, the average person commits tax evasion is 24.4% compared to in the loss scenario of 22.1%. In other words, at baseline treatment, someone is more likely to commit tax evasion in loss scenario than in gain scenario. But this is different from the results of Wilcoxon Signed-Rank test. In table 3, it can be seen that the p value is 0.881 which means that the signifi cance value is more than 0.05 or (0.881 > 0.05), which means that hypothesis 1 is rejected. The researchers consider that there is no difference in tax evasion in both gain scenario and loss scenario because they assume that taxes are a deduction from their income, so that the tax burden they pay will be reduced to a minimum. They can carry out tax evasion with overstatement of income and underdeclaration of income. Based on these results it can be concluded that at baseline treatment there is no difference in tax evasion decisions, both in gain scenario and in loss scenario, depending on the situation.
The results of this study are in accordance with behavioral fi nance theory that psychological factor affects someone in fi nancial decision making. This factor is infl uenced by prospect theory, where in making decisions, people tend to risk averse when they are in a state of profi t, and they tend to risk seeking when they are in a state of loss.
The researchers also argue that the model design in this study is based on mental accounting, in which in making tax evasion decisions, individuals are faced with two account mental decisions, that is, in the gain scenario and in the loss scenario. This is in accordance with the mental accounting theory that focuses on how one should respond to and evaluate if there are two or more possible outcomes (Thaler, 2008). This research is in line with the research conducted by Kirchler and Maciejovsky (2001) which states that self-reported tax compliance depends on the situation of gains and losses, whether faced with tax payments or tax returns. This study also supports the research conducted by Chang et al. (1987) that tax evasion behavior depends on whether taxes are framed as gains or as losses. However, this study is not in line with the research conducted by Fochmann and Wolf (2015) and Torgler and Schaffner (2007) which state that a person tends to commit tax evasion by underreporting income rather than by overdeducting expenses. The results of the research conducted by Torgler (2012) also show that there are differences in under-declaration and over-deduction. People prefer to embezzle taxes on over-deduction rather than underdeclaration.
At aggregation treatment, based on the results of data analysis in descriptive statistical test, it can be seen that the average participant commits more tax evasion in the loss scenario than in the gain scenario. This can be seen in table 2 where at aggregation treatment in the gain scenario, the average participant commits tax evasion is 14.4% compared to in the loss scenario of 12.8%. But this is different from the Wilcoxon Signed-Rank test results. In table 3, it can be seen that the p value is 0.647, which means that the signifi cance value is more than 0.05 or (0.647 > 0.05), which means that hypothesis 2 is rejected. The researchers consider that there is no difference in tax evasion, both in the gain scenario and in the loss scenario, because they assume that taxes are a deduction from their income, so that the tax burden they pay will be reduced to a minimum. They can carry out tax evasion with overstatement of income and under-declaration of income. Based on these results it can be concluded that at baseline treatment there is no difference in someone committing tax evasion, both in gain scenario and in loss scenario, depending on how the situation he wants.
Based on these results, it can be concluded that at aggregation treatment there is no difference in someone committing tax evasion, both in gain scenario and loss scenario. This means that the modifi cation of the payment in gain scenario and in loss scenario will be accumulated in both scenarios for payment in each period, or it can be said that a person is not different in terms of making tax evasion in both gain and loss situations. This is also in line with the behavioral fi nance theory that psychological factors affect someone in fi nancial decision making.
Although in modifying the provision of payments is accumulated at the end, participants keep making tax evasion decision in two scenarios: gain and loss. The researchers also argue that the model design in this study is based on mental accounting. In making tax evasion decisions, individuals are faced with two mental accounting decisions, namely in the gain scenario and loss scenario.
The results of this study support the research conducted by Kirchler and Maciejovsky (2001) which states that selfreported tax compliance depends on the situation of profi ts and losses, whether faced with tax payments or tax returns. Chang et al. (1987) also argues that taxpayers, in conducting tax evasion, depend on how individuals view the tax, as a deduction of certain profi ts or losses. This research is not in line with the research conducted by Fochmann and Wolf (2015) and Torgler and Schaffner (2007) which state that a person tends to commit tax evasion by underreporting income rather than overdeducting expenses. Torgler (2012), in his research shows that there are differences in under-declaration and over-deduction. People prefer to commit tax evasion on over-deduction rather than on under-declaration.
However, in the one tax declaration treatment there is no difference in tax evasion behavior, between in gain scenario and in loss scenario. This is indicated by the p value of the Wilcoxon test of 0.062, which means that the p value > 0.05. From these results it can be said that one will commit tax evasion based on the desired situation. Although the calculation on this treatment is not much different from aggregation treatment, at the one tax declaration treatment it is only one mental account which is used. In hypothesis 3 shows that mental accounting plays an important role in a person's decision to commit tax evasion.
This research is in line with the research conducted by Fochmann dan Wolf (2015) which states that there is no difference in the tax evasion behavior, both in gain scenario and in loss scenario. In addition, Kirchler and Maciejovsky (2001) state that self-reported tax compliance depends on the situation of gains and losses, whether faced with tax payments or tax returns. This research is also in line with the research conducted by Chang et al. (1987) that tax evasion behavior depends on whether taxes are framed as gains or as losses. However, this research is not in line with the research conducted by Torgler dan Schaffner (2007) which states that people tend to commit tax evasion by underreporting income rather than by over-deducting expenses. Torgler (2012), in his research, also shows that there are differences in under-declaration and overdeduction.

CONCLUSION AND SUGGESTION Conclusion
In this study, the researchers draw conclusions as the following: First, for participants who are at the baseline treatment, there is no difference in tax evasion, either in gain scenario or in loss scenario. This means that taxpayers will tend to commit tax evasion whether they are in a gain situation or in a loss situation. Second, for participants who are at aggregation treatment, there is no difference in tax evasion, either in gain scenario or in loss scenario. This means that taxpayers will tend to commit tax evasion whether they are in a gain situation or in a loss situation. Third, for participants who are at the one tax declaration treatment, there is no difference in tax evasion, either in gain scenario or in loss scenario. This shows that taxpayers will tend to commit tax evasion whether they are in a gain situation or in a loss situation. So, it can be concluded that mental accounting plays an important role in tax evasion decisions.

Research Limitation
This study has several limitations, such as: First, the number of samples used in his study was reduced due to a reduction in manipulation checks, so that only 25 sample participants that could be analyzed; Second, the experiment was scheduled for only one day, but it turned out that on the scheduled day, the number of participants who came was so few that the researchers had to conduct the experiment again the following week; Third, the participants were only limited to accounting students at Trunojoyo University, Madura, with concentration on taxation, so that there might be different results if the population was expanded to the actual taxpayers; Fourth, the lack of learning process in experimental experiments made participants less aware of experimental instruments so that many participants failed during manipulation checks; Fifth, the treatment, which was presented verbally caused differences in participants' perceptions and understanding compared to using a recorder system.

Suggestion
Based on the analysis results and discussion of the effect of mental accounting on tax evasion decisions, the suggestions for further research are as follows: First, it is suggested that further research use actual taxpayers as experimental samples; Second, it is suggested that further research change the research variables and fi nd other variables that have strong infl uence and can be benefi cial to all parties; Third, it is suggested that further research not only use experiments but can also conduct direct interviews; Fourth, it is suggested that further research use case instruments so that participants in the experiment understand the desired treatment; Fifth, it is suggested that further research use a system record so that the delivery of intent and purposes in treatment does not cause a difference in understanding; Sixth, it is suggested that further research take into account the learning process so that the participants understand the task of experiments given and to avoid the number of participants who fall during manipulation check.