Coercive and legitimate authority impact tax honesty: evidence from behavioral and ERP experiments

Abstract

Cooperation in social systems such as tax honesty is of central importance in our modern societies. However, we know little about cognitive and neural processes driving decisions to evade or pay taxes. This study focuses on the impact of perceived tax authority and examines the mental chronometry mirrored in ERP data allowing a deeper understanding about why humans cooperate in tax systems. We experimentally manipulated coercive and legitimate authority and studied its impact on cooperation and underlying cognitive (experiment 1, 2) and neuronal (experiment 2) processes. Experiment 1 showed that in a condition of coercive authority, tax payments are lower, decisions are faster and participants report more rational reasoning and enforced compliance, however, less voluntary cooperation than in a condition of legitimate authority. Experiment 2 confirmed most results, but did not find a difference in payments or self-reported rational reasoning. Moreover, legitimate authority led to heightened cognitive control (expressed by increased MFN amplitudes) and disrupted attention processing (expressed by decreased P300 amplitudes) compared to coercive authority. To conclude, the neuronal data surprisingly revealed that legitimate authority may led to higher decision conflict and thus to higher cognitive demands in tax decisions than coercive authority.

Introduction

Social systems are characterized by individuals who cooperate in the provision of non-excludable public goods, or defect and free-ride (Dawes, 1980). Tax honesty and tax evasion represent highly relevant examples of such cooperation. They do not only impact the possibilities of modern societies to provide schools or roads, but can also fuel social turbulences based on perceived injustice, if some taxpayers are perceived to systematically avoid paying their fair share of taxes. To uphold cooperation in these fragile systems, modern societies employ centralized institutions who wield harsh coercive or soft legitimate authority to ensure individual cooperation (Raven et al., 1998; Turner, 2005; Güth et al., 2007; Baldassarri and Grossman, 2011). Empirical evidence shows that the two qualities of authority lead to different motivations to cooperate (Kastlunger et al., 2012; Hofmann et al., 2014; Hofmann et al., 2017), but rarely differ in their positive effect on cooperation (Alm et al., 1992; Tyler and Fagan, 2008; Isakov and Rand, 2012; Hofmann et al., 2014; Hartl et al., 2015).

Coercive and legitimate authority promote cooperation originating from different motivational states (Koslowsky et al., 2001; Kirchler et al., 2008; Gangl et al., 2015). Coercive authority based on control and punishment is assumed to fuel mistrust (Mulder et al., 2006), to be perceived as ferocious and unfair (Mossholder et al., 2009), and to induce calculative cost-benefit analyses (Kirchler, 2007), which frame the decision to cooperate as a business decision rather than as an ethical decision (Tenbrunsel and Messick, 1999). Thus, the decision to cooperate might become a deliberate rational decision about one’s own egoistic utility and societal goals become less salient. Consequently, coercive authority is seen to cause resistance, reactance (Brehm, 1966) and negative emotional arousal (Coricelli et al., 2010), and it enforces motivation to comply out of feared penalization (Frey, 1997; Kirchler et al., 2008). In contrast, legitimate authority is based on acceptance, appreciation, perceived expertise and information provision (Raven et al., 1998; Gangl et al., 2015). It is assumed to increase trust and fairness (Tyler and Fagan, 2008; Hechter, 2009) and might be perceived as the communication of a widely accepted rule by community members (Bendor and Mookherjee, 1990), or as an instrument to establish fairness and a social norm of cooperation (Fehr and Fischbacher, 2002). Legitimate authority might induce moral pressure and psychological stress (e.g. a conflict, Erard and Feinstein, 1994; Dulleck et al., 2016) in response to the felt urge to reciprocate to legitimate authority (Fehr et al., 1997). Legitimate authority might provoke a deliberate conflict between self-interest and community interests which, in turn, fosters a voluntary motivation to cooperate (Tyler, 1997; Kirchler et al., 2008; Hofmann et al., 2014). Thus, decisions about cooperation under coercive or legitimate authority do reflect rational economic choices (Camerer, 2003) and social motives of reciprocity or inequity aversion (Fehr and Fischbacher, 2002). Although coercive and legitimate authority are perceived so differently, their positive effects on cooperative behavior do not seem to vary dramatically (Hartl et al., 2015).

Taxpaying can be considered as a highly relevant real life example for cooperation in a social context. Tax behavior is frequently investigated in laboratories. Referring to experimental games, income, tax rates, audit probabilities and fine rates are manipulated and tax payments over several tax filing trials are registered as dependent variable (Mittone, 2006; Alm et al., 2010; Hartl et al., 2015). Although a bulk of empirical studies on tax behavior has been published, studies on cognitive processes underlying tax behavior are rare and neurophysiological insights are missing.

We performed an event-related potentials (ERPs) study, investigating how tax payment decisions are affected by different types of centralized authorities. ERPs have been repeatedly used to gain insights into social decision-making, with their high temporal resolution enabling detailed insights into the mental chronometry of decision-making. In particular, amplitude variation of the Medial Frontal Negativity component (MFN; Gehring and Willoughby, 2002) has been investigated in this regard. It is a negative scalp potential within 200–300 ms after (feedback) stimulus onset at fronto-central electrodes, with the anterior midcingulate cortex (aMCC) and ventral striatum as potential neuronal generators (Debener et al., 2005; Becker et al., 2014). Generally, MFN amplitudes are more negative after unfavorable compared to favorable (Miltner et al., 1997), unexpected compared to expected (Alexander and Brown, 2011; Pfabigan et al., 2011), and salient compared to insignificant outcomes (Talmi et al., 2013). Moreover, MFN enhancement was interpreted as a signal of enhanced cognitive control (Botvinick et al., 2001; Yeung et al., 2004), which describes, among others, the monitoring and regulation of response strategies, feedback processing yielding strategy regulation and response conflict (i.e. concurrent activation of incompatible response options; Folstein and Van Petten, 2008). Taken together, MFN amplitude variation reflects early and coarse stimulus evaluation processes (Yeung et al., 2004; Hajcak et al., 2006). Regarding social decision-making, the MFN is usually observed in economic games focusing on asset distribution, such as the Ultimatum Game (UG; Güth et al., 1982). There, MFN amplitude enhancement indicates unfair compared to fair outcomes (Boksem and De Cremer, 2010; Alexander and Brown, 2011; Alexopoulos et al., 2012, 2013), concurrently also reflecting cognitive control.

Another ERP component frequently investigated during social decision-making is the P300. This positive-going component peaks within 300–500 ms after (feedback) stimulus onset at posterior electrodes (Polich, 2007). More pronounced P300 amplitudes have been reported in response to a greater change in evaluative stimulus categorization (Cacioppo et al., 1993; Ito et al., 1998) and greater attention allocation (Polich, 2007). Larger P300 amplitudes were found after positive compared to negative outcomes (Bellebaum and Daum, 2008; Pfabigan et al., 2015). In UG studies, P300 amplitudes were also sensitive to outcome fairness–usually larger after fair compared to unfair offers (Wu et al., 2012; Qu et al., 2013).

We conducted two experiments introducing taxpaying in two centralized institutional contexts. In the first experiment, we collected behavioral data and reaction times while participants performed repeated fast tax decisions under coercive and legitimate authority. In the second experiment, the tax paradigm was slightly adapted to an individual assessment and such that the speed of participants’ proceeding the tax decision was reduced, to allow a more valid assessment of ERP data.

As in previous studies, we expected that in contrast to legitimate authority, coercive authority will lead to self-reported enforced motivation to comply, low voluntary motivation to cooperate (Kastlunger et al., 2012; Hofmann et al., 2014) and enhanced rational decision-making and reactance than legitimate authority (Brehm, 1966, Kirchler, 2007). Overall tax payments should not differ between the authority conditions (Hartl et al., 2015). For reaction times (experiment 1) and ERP data (experiment 2) we tested two opposing hypotheses. On the one hand, coercive authority might induce a more complex decision process than legitimate authority. It provokes a comparatively effortful calculative cost-benefit analysis and deliberation about possible ways to reduce the tax burden. Legitimate authority might be perceived as a positive signal and thus produces a fast response to spontaneously cooperate (Tenbrunsel and Messick, 1999; Kirchler, 2007). On the other hand, legitimate authority might induce more complex cognitive processes than coercive authority, as personal utility optimization motives might be in conflict with the moral pressure to pursue societal goals. Reciprocity towards the legitimate authority might elicit a cognitive response conflict whereas coercive authority produces a fast response to cooperate (Fehr et al., 1997). Reaction times serve as indicator of cognitive effort and indicate whether cost-benefit analysis induced by coercive authority or the heightened response conflict induced by legitimate authority lead to more complex, deliberate and, thus, slower decisions (Rubinstein, 2007). For the ERP data, we expected MFN enhancement to indicate heightened cognitive control demands due to response conflict either in the coercive or the legitimate context (Botvinick et al., 2001). Enhanced P300 amplitudes should reflect enhanced categorization demands and attentional processing (Cacioppo et al., 1993; Ito et al., 1998; Polich, 2007) in case either coercive or legitimate authority induce a more complex decision process. In addition, we explored the P2 component prior to the MFN, which is indicative of arousal levels (Carretié et al., 2001) and attention capture (Potts, 2004).

Experiment 1

Method

The following section describes the central methodological aspects of our study; additional methodological details are provided in Supplementary Material.

Sample

The sample consisted of 80 volunteers (38 men, 1 did not indicate sex; M_age = 24.89, SD_age = 6.48). Participants were randomly assigned to one of two taxpaying conditions (coercive authority of tax administration followed by legitimate authority [n = 39], or legitimate authority followed by coercive authority, [n = 41]. The study was conducted in accordance with the Declaration of Helsinki (7th revision, 2013) and local ethical guidelines for experimentation with human participants (including approval by an institutional review board) at the Faculty of Psychology, University of Vienna.

Procedure

Participants were asked to imagine being self-employed, earning money in Experimental Currency Units (ECU) and paying taxes (40% of their income) over several trials in a fictitious country. They were informed that the tax administration would conduct tax audits with a chance of 15%. In case of detected tax evasion they would have to pay back the evaded amount plus a fine of the same amount. Participants were informed that a randomly chosen trial determinates their final remuneration for participation. After the introduction to the rules of the tax game, in counterbalanced order, 40 coercive and 40 legitimate authority trials followed, presented on a PC. After introducing the first authority, each trial started with the endowment of one out of five incomes (each presented eight times per authority in exactly the same sequence for all participants; the randomized sequence was determined prior to the experiment) and the presentation of 40% tax rate. Then participants decided how much tax they wanted to pay. After 40 trials in the first country, participants filled out the first self-report questionnaire. Subsequently, they were told to move to another country with a changed authority where they filed taxes during another 40 trials, ending by filling out the second self-report questionnaire. Participants were reminded of the respective authority manipulation every 10th trial. Finally, one prior to the experiment randomly selected trial determined the payment of all participants based on their behavior in this trial. On average, individuals earned €10.78.

Material

To manipulate authority, two scenarios were developed which described tax administrations that basically work with harsh controls and punishments (coercive authority) or with professional experts who support taxpayers filing their taxes (legitimate authority). To allow a direct comparison of the effect of both scenarios, a within-subjects design was used in which all participants were presented with both scenarios, in counterbalanced order. A manipulation check revealed that the centralized institutions were perceived as manipulated (see Supplementary Material [1.1.3, 3] for material and details on the manipulation check).

Tax behavior was assessed by averaging taxpaying decisions over 40 tax filing trials. For each tax filing trial one out of five incomes (60.000 ECU, 70.000 ECU, 75.000 ECU, 85.000 ECU, or 90.000 ECU) was assigned to the participants. Participants read the information about coercive or legitimate authority, their income, and the tax rate until they choose to terminate the slide by pressing key 1. Afterwards, participants were presented with a slide with five possible tax amounts to pay corresponding to 0% (fully dishonest), 25%, 50%, 75% and 100% (fully honest) of the tax due. This slide was presented until participants choose one of the options via button press (keys 1–5, pre-assigned to the respective percent value). The different income levels ensured that participants varied the responses over the 40 trials.

Mean reaction times were assessed from the onset of the tax amount screen until button press. For each trial, outliers (2.17% of all data points) were discarded in case reaction times were longer than the mean reaction time plus three times the standard deviation.

Self-reports were assessed with two identical questionnaires. Perception of coercive authority and legitimate authority was assessed with the Interpersonal Authority Inventory (Raven et al., 1998; Hofmann et al., 2014) adapted to our tax context. Reactance was measured by the adapted Hong Psychological Reactance Scale (Shen and Dillard, 2005). Deliberate rational decision-making was measured by the adapted Rational-Experiential Inventory (Pacini and Epstein, 1999). Enforced compliance was measured by the TAX-I (Kirchler and Wahl, 2010), voluntary cooperation by items adapted from the motivational posture scale capitulation (Braithwaite, 2003). Response scales were 7-point Likert scales (1–disagree, 7–agree). Items and their reliability scores are reported in Supplementary Material (4).

Statistical analyses

To analyze the differential impact of the two centralized institutions, 2 x 2 univariate analysis of variance with authority manipulation (coercive vs legitimate) as within-subject factors, and order of manipulation (coercive followed by legitimate authority or vice versa) as between-subjects factor was performed, with self-report, behavioral and reaction time data as dependent variables. Effect sizes of significant results are specified with partial eta squared (η_p²) and Cohen’s d; the alpha-level was set at P < 0.05. In the following, we present only the significant results of main effects. Detailed results and robustness checks are presented in Supplementary Material (1.1.4, 1.1.5).

Results

Self-report data

Self-reported enforced compliance was significantly affected by the type of authority (F(1,77) = 54.55, P < 0.001, η_p² = 0.42). As Figure 1 shows, enforced compliance was higher in the coercive (M = 5.39, SE = 0.18) than in the legitimate authority condition (M = 3.88, SE = 0.20). This is especially true if coercive authority follows legitimate (for details on an interaction effect see Supplementary Material). On the other hand, voluntary cooperation was only affected by authority (F(1,77) = 59.51, P < 0.001, η_p² = 0.44), it was higher under legitimate (M = 4.65, SE = 0.15) than coercive authority (M = 2.90, SE = 0.15). Also, rational decision-making was affected by authority (F(1,78)=11.55, P < 0.001, η_p² = 0.13) and was higher under coercive (M = 5.00, SE = 0.17) than legitimate authority (M = 4.35, SE= 0.18). Reactance was also higher under coercive (M = 5.25, SE = 0.15) than legitimate authority (M = 3.80, SE = 0.18; F(1,78) = 81.77, P < 0.001, η_p² = 0.51; for details see Supplementary Material).

Fig. 1.

Differences in self-reports in experiment 1 and 2. Note: The bars represent means with the standard error. T1 and T2 indicate whether the authority was presented first (T1, i.e. first 40 trials) or second (T2, i.e. second 40 trials). 1 = disagree, 7 = agree.

Behavioral data

Tax compliance was significantly higher under legitimate (F(1,78) = 5.96, P = 0.017, η_p² = 0.07; M = 4.07; SE = 0.12) than coercive authority (M = 3.77, SE = 0.14). Reaction times were significantly affected by the contrast between coercive and legitimate authority (F(1,78) = 177.63, P < 0.001, η_p² = 0.70). Compared to time 1 (first 40 trials) (t(78) = 1.05, P = 0.295), at time 2 (second 40 trials) (t(78) = –1.97, P = 0.052, d = 0.45) coercive authority (M = 1784.67 ms, SE = 138.61) by trend yielded faster responses than legitimate authority (M = 2230.02 ms, SE = 138.98).

Discussion experiment 1

Corroborating previous studies, experiment 1 shows that a perceived coercive compared to a legitimate authority led to less voluntary cooperation (Hofmann et al., 2014). For enforced compliance and reactance, we observed sensitivity to the change from coercive to legitimate authority. More specifically, legitimate authority’s negative effect on enforced compliance and reactance was particularly strong when a legitimate authority was perceived as the direct change following coercive authority. Overall and in line with previous studies (e.g. Hofmann et al., 2014), the effect sizes indicated that the difference between coercive and legitimate authority was the most important factor affecting enforced compliance and reactance. Results also showed that legitimate authority led to slightly higher tax payments than coercive authority. Coercive authority induced more self-reported rational decision-making than legitimate authority. However, reaction time results indicate that the decision under coercive authority was quicker to reach compared to legitimate authority. At time 2 (second 40 trials), coercive authority led to faster reactions than legitimate authority. Experiment 1 is one of the rare attempts showing that the different psychological meanings of coercive and legitimate authority can also be detected by indirect measures of cognitive processes such as reaction times. To clarify the exact nature of these processes and how coercive and legitimate authority affect them, data providing access to the temporal sequence and different cognitive processes underlying decision-making and behavior are needed. Therefore, our second experiment assessed ERPs while participants made tax decisions under different tax authorities.

Experiment 2

Method

 

Sample

The sample consisted of 81 volunteers of which three were excluded because of limited task comprehension or excessive alpha-band EEG activity (final sample: n = 78, 40 men, M_age = 24.51, SD_age = 5.05). All participants were right-handed (Oldfield, 1971), had normal or corrected-to-normal vision, and reported no past or present neurological or psychiatric disorder. All participants gave written informed consent prior to the experiment. Recruitment and ethical guidelines were comparable to experiment 1. They were again randomly assigned to one of two taxpaying conditions (coercive authority of tax administration followed by legitimate authority [n = 38], or legitimate authority followed by coercive authority [n = 40]).

Procedure

Participants were tested individually in a sound-attenuated, shielded chamber. As in experiment 1, they were asked to imagine being self-employed, earning income, and paying taxes over several years. To ensure that participants understood the instructions, they were led through a taxpaying example and performed 10 training trials. Then, identical to experiment 1 (order counterbalanced), 40 coercive and 40 legitimate authority trials followed. However, different to experiment 1, participants were reminded every 5th trial to make the manipulation more salient. Each trial started with the presentation of a white fixation cross on black background (1000ms). Afterwards, the flag of the respective country, the fictitious income, and the 40% tax rate in total numbers were presented centrally on black background (3000 ms). Prior to the actual tax decision, another screen was blended in asking ‘How much tax would you pay in [current country]?’ (2000 ms). This slide was added to the experimental design to avoid fast responses to the tax decision as in Experiment 1. Afterwards, participants were presented with five options for taxpaying, either representing 0%, 25%, 50%, 75% or 100% of the 40% tax rate (absolute numbers). Participants had to choose between the keys 1–5 on a standard keyboard to indicate their tax decision via button press. No time limit was given for their decision. A variable inter-stimulus-interval (1400–1600 ms) was presented afterwards depicting the fixation cross. After 40 trials in the first context, participants filled out the first self-report questionnaire and were subsequently informed that they would move to a different country with a different tax administration. After introducing the second context, another 40 trials were presented applying the same experimental design as before, ending with filling out the second self-report questionnaire. At the end, participants were remunerated based on a €10 show-up fee and their tax decision in one randomly chosen tax trial (on average €22.12).

Material

The same material as in experiment 1 was used (Supplementary Material 1.2.3).

EEG was recorded from 57 equidistantly arranged electrodes in a cap. Signal preprocessing and artefact correction was conducted using EEGLAB (Delorme and Makeig, 2004). Please refer to Supplementary Material for further details on data collection and preprocessing (1.2.2).

To assess ERP amplitudes, data were epoched time-locked to the onset of the income separately for coercive and legitimate trials. MFN amplitudes were extracted at FCz as peak-to-peak-to-peak values (Yeung and Sanfey, 2004); i.e. difference between the MFN component and the mean of the preceding P2 and the subsequent P300 component) in the time window 150–400 ms post income. P300 amplitudes were extracted at Pz as peak-to-peak values (Pfabigan et al., 2011); i.e. difference between the P300 and the preceding N2 component in the time window 200–600 ms post income.

Statistical analyses

Several 2 x 2 univariate analyses of variance with manipulation of authority (coercive vs legitimate authority) as within-subjects factors, and order of manipulation (coercive followed by legitimate authority and vice versa) as between-subjects factor were performed with self-report data, behavioral data, and P2, MFN and P300 amplitudes as dependent variables. Again, we present only significant results. Detailed results and robustness checks are presented in Supplementary Material (1.2.4–6).

Results

Self-report data

As shown in Figure 1, self-reported enforced compliance was again significantly higher under coercive (M= 5.65, SE = 0.17) than legitimate authority (M= 4.13, SE= 0.20; F(1,76) = 50.78, P < 0.001, η_p² = 0.40). Again, voluntary cooperation was significantly higher under legitimate (M = 4.92, SE = 0.14) than coercive authority (M = 3.03, SE = 0.15; F(1,76) = 103.11, P < 0.001, η_p² = 0.58). Rational decision-making did not differ between coercive and legitimate authority. Again, reactance was significantly higher under coercive (M = 5.18, SE = 0.17) than legitimate authority (M= 3.79, SE = 0.18; F(1,75) = 50.17, P < 0.001, η_p² = 0.40).

Behavioral data

Tax payments were not affected by our experimental manipulation (all P-values > 0.195). See Supplementary Material 1.2.5 for descriptive statistics.

ERP data

P2 amplitudes were significantly more pronounced under legitimate (M = 5.20, SE = 0.52) than coercive authority (M = 4.13, SE = 0.58; F(1,76) = 5.30, P = 0.024, η_p² = 0.07). As Figure 2a indicates, MFN amplitudes were significantly more pronounced under legitimate (M = –9.35, SE = 0.54) than coercive authority (M =  –8.27, SE = 0.56; F(1,76) = 9.43, P = 0.003, η_p² = 0.11). P300 amplitudes were significantly more pronounced under coercive (M = 9.67, SE = 0.62) than legitimate authority (M = 8.64, SE = 0.52; F(1,76) = 6.81, P = 0.010, η_p² = 0.08). See Supplementary Material 1.2.6 for descriptive statistics.

Fig. 2.

(A) Time-courses of the P2 and MFN component at FCz (upper panel; also showing P300 peak used for peak quantification [Yeung and Sanfey, 2004]) and the P300 component at Pz (lower panel; also showing N2 peak used for peak quantification [Pfabigan et al., 2011]). Time point zero indicates onset of the flag, tax income, and the 40% tax rate. Negative is drawn upwards per convention. (B) Scalp topographies of the mean difference between coercive and legitimate trials in the P2 time interval (170–210 ms; left side), MFN time interval (260–300 ms; in the middle) and the P300 time interval (300–400 ms, right side). (C) Bar graph depicting mean and SE of P2, MFN and P300 peak values per condition and time point. Please note that absolute values of the three ERPs (in µV) were plotted for a uniform display. Note: b2p = base-to-peak; p2p = peak-to-peak; p2p2p = peak-to-peak-to-peak approaches for ERP quantification.

Discussion experiment 2

Experiment 2 confirmed most of the results of experiment 1, apart from the absent differences in tax payments and self-reported rational decision making. However, these absent differences replicate existing findings (Hartl et al., 2015) and might originate from the different settings in both experiments. Concerning ERP data, results show that coercive authority differentially affected attentional processing (P2, P300) and reduced cognitive control demands (MFN) compared to legitimate authority.

General discussion

We investigated whether or not coercive and legitimate centralized institutions elicit comparable cognitive and neuronal processes to reach the final tax payment decision. In experiment 1, coercive authority led to less tax payments, faster decisions and more self-reported rational decision-making, and to less voluntary cooperation than legitimate authority. In experiment 2, no difference in tax payments and self-reported rational decision-making was found. This can be due to differences in experimental procedures, fostering less spontaneous decisions in experiment 2. On the neuronal level, tax authority influenced all ERPs. P2 and MFN amplitudes were more pronounced in legitimate compared to coercive conditions, while P300 amplitudes were more pronounced in coercive than legitimate ones.

In tax experiments which applied a similar paradigm, both coercive and legitimate authority are shown to increase tax payments (Wahl et al., 2010; Hofmann et al., 2014; Hartl et al., 2015). This, however, is based on different motivations. Coercive authority was shown to lead to less trust in institutions, more enforced compliance and less voluntary cooperation than legitimate authority (Hofmann et al., 2014). The different psychological processes underlying both types of authority might explain why some previous studies (Hofmann et al., 2014) as well as experiment 1 show that legitimate authority generates higher tax payments than coercive authority. Enforced compliance elicited by coercive authority reduces tax payments; voluntary cooperation based on legitimate authority increases tax payments (Kastlunger et al., 2012; Gangl et al., 2015). Thus, particularly when individuals decide spontaneously these different motivational processes related to voluntary cooperation and enforced compliance might generate higher tax payments for legitimate than for coercive authority.

In ultimatum game studies, enhanced MFN amplitudes are usually observed after unfair compared to fair offers, and most likely reflect fairness-related norm enforcement in these situations (Feng et al., 2015). Along these lines, Fehr and Camerer (2007) reported that unfair offers in economic games induce motivational response conflict between economic self-interest and norm enforcement. MFN amplitude variation in the current study might therefore reflect enhanced response conflict induced by legitimate compared to coercive tax authority since evading tax might be perceived as more conflicting in the legitimate than the coercive context because of moral pressure to pursue community goals on costs of egoistic motives. Legitimate authority is likely perceived as positive and as community-serving, which in turn calls for reciprocity (Fehr et al., 1997; Tyler and Fagan, 2008), mirrored by enhanced voluntary cooperation (Hofmann et al., 2014). Thus, MFN and voluntary cooperation data indicate that legitimate authority might make it more difficult to defect than coercive authority. Moreover, we observed significantly larger P2 amplitudes during legitimate than coercive trials. This further corroborates the MFN results. Thus, even before indicating response conflict by MFN variation, legitimate tax authority stimuli induced heightened arousal levels (Carretié et al., 2001) or higher attention capture (Potts, 2004) than coercive tax stimuli.

In contrast to the early evaluation indicated by P2 and MFN components, enhanced P300 amplitudes were observed in coercive compared to legitimate trials. This finding suggests at first glance that coercive authority might yield greater evaluative changes and attention (Cacioppo et al., 1993; Ito et al., 1998; Polich, 2007), which disagrees with a simple calculative cost-benefit heuristic. However, Fabre et al. (2015) proposed that larger P300 amplitudes during asset distribution might reflect more automatic processing in line with simple heuristics in working memory (Khader et al., 2011), while decreased amplitudes might reflect a disruption of heuristic-driven processing. For example, P300 amplitudes are larger for fair compared to conflicting unfair and mid-fair proposals (Hewig et al., 2011; Wu et al., 2012; Qu et al., 2013). Also, in dual-task settings, P300 amplitudes are larger in the main task if the second task is an easy compared to a difficult cognitive task (Kramer et al., 1985). Therefore, we belief that legitimate compared to coercive authority is related to a rather difficult decision conflict which disrupts also stimulus categorization (mirrored in P300 decrease).

Faster decisions in experiment 1 in the coercive condition also indicate that a simple calculative cost-benefit analysis might represent an easy-to-apply cognitive heuristic, which is processed faster than the response conflict induced by reciprocity during legitimate authority. Thus, our findings indicate that tax decisions under coercive authority were processed more easily than under legitimate authority.

Self-reported rational decision-making was enhanced under coercive compared to legitimate authority in experiment 1. However, no such difference was observed in experiment 2. Although reaction time and ERP data indicate that the legitimate condition is the more complex one, this self-report finding suggests that the decision complexity under legitimate authority was not consciously perceived by the participants. The different experimental settings might be partly responsible for this. The rather spontaneous tax decision in experiment 1 might have led to more conscious cognitive workload in the coercive than in the legitimate condition. In contrast, the experimentally prolonged tax decision in experiment 2 might have reduced the conscious workload for coercive compared to legitimate authority. Alternatively, the rational decision-making scale might not be able to differentiate between simple calculative cost-benefit analysis and deliberate decisions about the social consequences of one’s decision. The present results suggest that both, coercive and legitimate authority, lead to deliberate considerations, but with varying cognitive demands.

Concerning our opposing hypotheses, we found that coercive authority induces simpler calculative cost-benefit analysis instead of more complex social evaluation processes, which in turn make it less conflicting to defect under coercive compared to legitimate authority (Tenbrunsel and Messick, 1999). Surprisingly, the hypothesis that the cost-benefit analysis induced by coercive authority is more complex than the tax decision under legitimate authority is not supported by our data. Relating our findings to previous literature (Coricelli et al., 2010; Dulleck et al., 2016) suggests that coercive authority might induce less deliberate and less emotional demands compared to legitimate authority, which might provoke a straining conflict on reciprocity and fairness accompanied by psychological stress. Thus, coercive and not legitimate authority might provoke more instinctive economic choices (Rubinstein, 2007) and allows a fast calculative response whereas legitimate authority introduces a complex conflict between self- and social interests. Our finding that legitimate authority leads to more fairness-related response conflict might also serve as explanation for why legitimate authority sometimes leads to higher tax payments than coercive authority (e.g. when spontaneous decisions are made such as in experiment 1 and Hofmann et al., 2014). Legitimate more than coercive authority might trigger and explain individuals’ tax morale (Frey and Torgler, 2007).

Taxpaying is an excellent example for social decision-making and implies financial decisions which affect both self-interest and other-orientation (Fehr and Camerer, 2007). Several authors suggested that a dual-process system is at work during decision-making reflecting the interplay of self-interest vs other-orientation (see Sanfey and Chang, 2008). System 1 is assumed to represent an automatic, fast, effortless, unconscious and slow-learning system implementing automatic and heuristic-based judgements. Contrarily, system 2 is assumed to represent a controlled, slow, effortful and fast-learning system implementing more deliberate reasoning and potentially reappraising input from system 1, and balancing competing interests. Thus, the question arises whether tax decisions under coercive and legitimate authorities are purely based on heuristic-based judgments (system 1) or involve more deliberate reasoning (system 2). The calculative cost-benefit analysis as a heuristic might be used to easily assess self-interests, while social norm enforcement might be reflective of other-orientation. Enhanced MFN amplitudes during legitimate compared to coercive trials might reflect system 1 activity, since MFN variation usually reflects automatic, coarse stimulus evaluation (Hajcak et al., 2006)–indicating enhanced response conflict between self- and other-interests. Subsequent P300 variation indicates that participants might have countered system 1 and tried to costly implement more deliberative reasoning during legitimate trials resulting in diminished P300 amplitudes compared to coercive trials (Fabre et al., 2015).

Limitations of the current study concern the consequences of the necessary compromise between research methods in tax psychology and neuroscience. The experimental setting (within-subject design), number of assessed trials (40) and the assessment of tax decisions (5-point Likert-type scale) deviates from the standard tax paradigms but was necessary to meet the requirements of EEG measurement. However, replication of the effects of coercive and legitimate authority on enforced compliance and voluntary cooperation (Hofmann et al., 2014) show that the present experiments can be compared with previous ones, providing evidence for the internal validity of the present results.

Future studies on cognitive and neuronal processes involved in tax decisions should examine more realistic settings by analyzing the impact of a combination of high vs low coercive with high vs low legitimate authority on cooperation (Hartl et al., 2015). Moreover, trust is argued to be as important for cooperation as coercive and legitimate authority (Kirchler et al., 2008).

Cooperation in social systems is organized by centralized institutions such as tax administrations which aim to motivate tax honesty. The present experiments are the first attempt of generating insights into psychological, cognitive and neuronal processes involved in taxpaying under coercive or legitimate authority. Our results also have policy implications: legitimate compared to coercive authority makes tax compliance to a complex social instead of a simple economic problem, thus should be a preferred strategy of tax administrations to foster general tax morale. A better understanding of tax compliance might help to ensure sufficient funds for public goods and might increase public confidence if all taxpayers are perceived to contribute their fair share instead of exploiting the system.