“Optimal Honesty” in the Context of Fiscal Crimes

Abstract

This paper begins by contrasting the caricatures ‘homo and femina economicus’ with ‘homo and femina realitus’. Against this backdrop, the paper considers three ‘apparently falsified’ empirical predictions of the standard expected utility model of individual decision-making concerning participation in fiscal crimes: that tax evasion and benefit fraud can be treated identically; fiscal crimes should be endemic; and that all individuals, depending on parameter values, should be either honest or dishonest. A utility function relating to decisions with a moral dimension is used to offer insight into the rationalization of the predictions and involves defining an individual’s ‘optimal honesty’ in the context of fiscal crimes. The policy implications of the approach are briefly explored.

1. Introduction

Economists apply their ‘cultural’ perspective (Rubinstein 2008, p. 11) to almost all individual choices and actions. Such choices and actions could be influenced by a myriad of social, psychological and biological factors. Economists cut the Gordian knot of multiple potential competing analyses by adopting the Paretian value judgments and ascribing particular attributes to the unit of analysis—individuals. The preferences and capabilities of homo and femina economicus are the focus. This ‘representative’ individual is described as: (i) ‘rational’; (ii) egoistic; (iii) with egoism predicated on self-interest narrowly defined in terms of income or wealth (Brennan and Lomasky 1993). Armed with this caricature, the world is the economists’ oyster. The world of fiscal crimes is captured in the ‘workhorse model’ developed by Allingham and Sandmo (1972) in particular. Doubtlessly, it offers great insights by modeling tax evasion as amoral expected utility maximizing decisions. In this context, policy prescriptions also become clear—manipulate the detection probabilities and the penalties contingent on detection.

What is wrong with this? For some economists absolutely nothing—job done! For others, there are a number of misgivings about the theory. The question as to “what do you want economic theory to do?” highlights aspects of the misgivings. As a bold generalization, you might want theory to be pre-(post-) dictive, prescriptive, descriptive, and elegant in the sense of being sparse (conforming to Occam’s razor). Famously, Friedman (1953) has claimed the only relevant criterion for a ‘good’ economic model is its pre-(post-) dictive accuracy—the ‘gold standard’ test. In the ‘workhorse model’, as taxpayers, individuals consider the net expected utility gain of under-reporting income. As benefit claimants (of cash and in-kind benefits), individuals consider the net expected utility gain of under-reporting income. This ‘economic’ analysis of the decision to commit criminal activity focuses on expected outcomes alone (Becker 1968). However, for all the merits of this approach, it seems to present three puzzles. More specifically, three predictions of this theoretical approach seem falsified.

First, Halla and Schneider (2014, p. 412) note, “… tax evasion and benefit fraud are almost identical in the standard neoclassical model of compliance”. If individuals experience the same financial loss from benefit fraud as from tax evasion, there is no reason in the neoclassical world to assume that individuals’ perceptions and attitudes toward these fiscal crimes will differ. But they do. Empirical studies indicate that citizens are far more condemnatory of benefit fraud than of tax evasion, even though the estimated financial loss from tax evasion is greater than the estimated financial loss from benefit fraud (e.g., see Cullis et al. 2015). By comparison, the ‘economic’ approach (based on amoral, instrumental motivations) predicts that citizens will be more condemnatory of the activity (tax evasion) that produces the greatest financial loss (ceteris paribus).

Second, Alm et al. (1992) are early authors who complain that the ‘economic’ analysis of an ‘evasion gamble’ predicts that most individuals will evade, when, in practice, most declare income honestly. In Andreoni et al.’s (1998) seminal literature review, the authors note that, in 1995, the audit rate in the USA for individual tax returns was 1.7 percent, and the civil penalty for underpayment of taxes was 20 percent of the underpayment. They emphasize the implicit prediction that many more individuals were likely to evade tax than the number estimated for 1995.

Third, the model cannot predict a separating equilibrium in which some individuals tax evade alongside others who do not. It generates pooled equilibria in which, depending on parameter values, all individuals will evade or all individuals will comply with the law (this result is independent of different degrees of risk aversion individuals might display). Unfortunately, for the theory, it is the former separating equilibrium that is widely observed. The ‘workhorse’ theory seems to fail the ‘gold standard’ test of the ‘non-falsified predictions’ purpose of theory. As regards the other criteria noted above, the theory would not seek descriptive accuracy but, as long as you accept ‘homo and femina economicus’, would meet the prescriptive purpose for that caricature. It can also claim elegance. Given this judgment, what is to be done? The old joke goes: Q: “What do you do if the empirical data does not support your theory?” A: “Amend the data”. Sadly, this is not as far-fetched as it may seem, as empirical work is not that often replicated in economics, and Duvendack et al. (2017) note that one reason why replication is important is to avoid data error and ‘outright fraud’.

But what of the current context? In the face of these apparently empirically falsified predictions, it is not surprising that many have sought to amend the theory in various ways. In particular, this paper explores the relevance and possible determinants of moral costs in the decision to commit or not to commit the fiscal crimes of tax evasion and benefit fraud. When is ‘honesty the best policy’ and what ‘policy is best for honesty’?

The paper is organized as follows: Section 2 provides a brief literature review on various factors influencing ‘homo and femina realitus’ in their decision to engage in fiscal crimes, which helps extend Levitt and List’s utility function model (Levitt and List 2007) by incorporating the moral hinterland variables discussed in Section 3. A discussion of the policy implications of this revised model, followed by concluding remarks, is presented in Section 4 and Section 5, respectively.

2. Responding to Evidence

Gordon (1989) has individuals with an honesty characteristic who suffer a private stigma cost when they tax evade. Further, they incur public reputation costs depending on the strength of the evolving evasion norms that surround them. Privately they feel shame, and publicly they are shamed. These themes are explored elsewhere. Spicer (1986) and Kirchler (2007) argued that individuals experience a ‘psychic cost’ when they engage in tax evasion. Cullis et al. (2012) have citizens experiencing discomfort when they feel they are not complying with a social norm. Another strand in the literature draws attention to the intrinsic value that individuals derive when acting honestly. Deci (1971, p. 105) argues that an individual is “…intrinsically motivated to perform an activity when… (the individual) … receives no apparent reward except the activity” (see also Deci and Ryan 1980).

If this broader analysis is to shed insight (and be more than a tautology), it is important to focus on the determinants of perceptions of the intrinsic value of action. It has been argued that moral beliefs are a determinant of individuals’ perceptions of the intrinsic value of action. Brekke et al. (2003, p. 1969) argue: “People want to think of themselves as socially responsible (‘What kind of person am I?’)”. They argue that “… individuals first determine their morally ideal effort by asking themselves the following question: ‘What would the consequences for social welfare be if everybody acted like me’”. Essentially a Kantian perspective. Individuals choose to trade-off their wish to be socially responsible against the cost they will incur (in terms of consumption and leisure).

Frey (1997) emphasizes that individuals’ perceptions of the intrinsic value of action also depend on recognition—i.e., on an awareness that others acknowledge their action. To date, analysis has focused largely on the impact of acknowledgement of individuals’ action. When paying tax, individuals are far more likely to act honestly the more they see that others act honestly (Kirchler 2007). By comparison, less attention has been paid to the relevance of morality and to the way in which moral beliefs are likely to influence the decision to evade tax and the decision to commit benefit fraud. An important contribution here is Rabin (1995), who follows the logic of viewing acting morally as an ‘internal constraint’ on what otherwise would be narrowly self-interested behavior.

This paper questions the relevance of moral cost and the different ways in which moral costs are likely to influence the decision to act honestly. To facilitate this task, a different caricature to that of ‘homo and femina economicus’ is invoked.

Recent decades have witnessed the development of a different caricature for theory to work with, which has been dubbed in ‘cod’ Latin ‘homo and femina realitus’ (see Barile et al. 2018). This caricature draws on the type of arguments made above and numerous empirical findings from social and cognitive psychology and experimental economics. Relevant economics Nobel Laureates who made fundamental contributions in this field are 1978 winner Herbert Simon (see, for example, Simon 1955), 2002 winner Daniel Kahneman (see, for example, Kahneman 2011), and 2017 winner Richard Thaler (see, for example, Thaler 1994). This newer actor has a number of characteristics, exhibiting:

1. Bounded abilities. Individuals are not all powerful, and the sources of their limitations can be subdivided into (a) bounded rationality in choices and (b) bounded self-will over actions.

2. Bounded self-interest being concerned with more than pure self-interest narrowly defined—a ‘bigger and richer person’ than ‘homo and femina economicus’. Again, a sub-division may be helpful: (a) individuals have an internal moral or ethical dimension that shows up in concepts like intrinsic motivation—a desire to do ‘the right thing’ for its own sake and (b) an external dimension where they are wary that if they follow their narrow self-interest, they will unjustifiably impose costs on others and/or disappoint by failing to act in line with an accepted social norm. The flipside is where others act in a narrowly self-interested way, imposing cost on them and/or offending by showing disregard for an accepted social norm. Such norms are likely to be culturally specific. The estimated varying sizes of ‘black’ or shadow economies in different countries seems to bear witness to this (see, for example, Schneider and Williams 2013). The Corruption Perceptions Index is published annually by the non-governmental organization Transparency International on a scale of 0 (“highly corrupt”) to 100 (“very clean”). Denmark has topped the index for the last six years and had a score of 90 for 2023. Somalia was the most corrupt country in the same year with a score of 11. An implication of this is that in any policy discussions it is very unlikely that “one size will fit all”.

3. A preference map that is endogenous and malleable (as opposed to the traditionally assumed exogenous and fixed preferences that facilitate much of Neo-classical economic analysis). Here a distinction can be made by (a) looking at what might be termed transient endogeneity (e.g., by emotion priming in experiments to influence results or ‘micro’ framing effects); (b) focusing on more permanent endogeneity where the actor is responsive to public policy and other signals that affect preferences (‘macro’ framing effects). This raises much debated issues as to the appropriate role of government in a market economy, especially related to the use of “nudges”.

In what follows, insights relating to characteristic (2) above are employed in Section 3, whereas characteristics (1) and (3) are germane to Section 4.

3. ‘Optimal’ Honesty: The Relevance of Moral Costs

In this section, a utility function canvassed by Levitt and List (2007) is adopted to illustrate the impact of both instrumental and intrinsic motivation in determining optimal honesty. The objective is to apply and adapt the Levitt and List formulation to ‘solve’ the three falsified predictions noted above. What part (if any) are moral convictions likely to play in individuals’ assessment of the ‘wrongness’ of action: (i) to evade tax and (ii) to engage in benefit fraud? To what extent (if at all) are perceptions of the ‘wrongness’ of action relevant in individuals’ decisions: (i) to evade tax and (ii) to engage in benefit fraud?

Following Levitt and List (2007), an individual, i, has a choice over action a (here to be tax evader or benefit fraudster) that generates two sources of utility. First, there is the utility derived from income (Y). Second, there is a non-pecuniary source of (dis)utility derived from the moral costs (intrinsic value) of acting dishonestly or honestly. If an individual chooses to act dishonestly, the disutility (M) is the ‘psychic cost’ incurred as the intrinsic value of seeing yourself as honest atrophies—the moral cost of acting dishonesty. To begin, both Y and M depend on the value of the financial sum (v) that might be achieved if the individual were to act dishonestly.

Levitt and List (2007) also draw attention to other considerations that are likely to affect the costs of acting dishonestly. The first of these are the costs of non-compliance with social norms, n. These costs increase, the more that your action offends against a norm (or a legal rule) in society. Such norms vary from society to society. In a series of articles, Wenzel (2004, 2005a, 2005b) theoretically and empirically, using Australian data, finds a complex relationship between personal and social norms in relation to tax compliance. For example, social norms “behavior and shared ethics attributed to others” (Wenzel 2004, p. 214) become relevant when attributed to a group with which an individual identifies.

The second is the extent of scrutiny (s). Scrutiny increases the likelihood that individuals will experience stigma or shame costs the more that others become aware of an individual’s immoral action1. The individual’s additively separable utility function is described in Equation (1):

U_i = Y_i (a, v) + M_i (a, v, n, s)

(1)

In the absence of a ‘moral dimension’ the utility function collapses to an ‘income’ (or ‘wealth’) maximizing utility function. When moral considerations are relevant, individual choices are likely to deviate from a pure income-maximizing choice. The moral costs increase with the size of the deception v, the strength of the (honesty) norm n and the extent of scrutiny/stigma costs s. Pecuniary gains naturally increase as v increases.

Apart from considering n and s (are they different for tax evasion and benefit fraud?) and holding v constant, what other variables might impact whether tax evasion and benefit fraud are viewed differently? Given this question, it is reasonable to extend the list of variables affecting moral costs with an eye to a possible empirical investigation of decisions discussed here that involve, for most, moral costs. In particular, political affiliation, p, religion, r and moral rectitude (mr) suggest themselves as relevant factors. Questions on religion and political affiliation are common in social surveys on fiscal crime. Moral rectitude is a variable that is likely to require ‘proxying’. A priori, what might be the impact of these variables on the willingness to tax evade or indulge in benefit fraud.

p refers to political affiliation. Your political leaning provides a lens through which you view tax evasion and benefit fraud. A right-wing leaning is associated with a “low tax, small spending government”, “standing on your own two feet”, “carrying your own pack”, etc., predicting a weaker disapproval of tax evasion compared with benefit fraud. The equivalent caricature on the left-leaning perspective is a “high tax, big spending government”, “we are our brother’s keeper”, “take from the rich to give to the poor”, etc., predicting a weaker disapproval of benefit fraud compared to tax evasion. In short, other things equal, left-leaning individuals will be less inclined to view fraud benefits as ‘wrong’ and more likely to perceive tax evasion as ‘wrong’ than right-leaning individuals.

Further, r is for religion. Where moral costs arise, religion ought to matter. After all, the eighth commandment says, “thou shalt not steal”. It is the case that most religions see stealing as morally wrong. Given this, it can generally be predicted that religious individuals are less likely to be tax evaders or benefits fraudsters and be more disapproving of the activities. A priori, it is not clear whether tax evasion and benefit fraud would be seen as equally wrong or one more wrong than the other.

Finally, there is the moral stance an individual adopts. Much of economics has a largely amoral actor that seems at odds with the notion of people as moral beings. Individuals differ. The saying “being more Catholic than the pope” captures the notion that individuals display different degrees of moral rectitude (mr) in making choices. Zamir (2012) discussed below provides an argument that the moral costs of benefit fraud will exceed the moral costs of tax evasion. Adopting these amendments, Equation (1) can be modified to:

U_i = Y_i (a, v) + M_i (a, v, n, s, p, r, mr)

(2)

The variables n, s, p, r, and mr are seen here as defining an individual’s ‘moral hinterland’2 which feeds into the moral cost of acting dishonestly.

Focusing (in the first instance) on the decision to evade tax (ignoring for the moment the curves MCb in panel (a) and TCb in panel (b) of Figure 1), it is possible to illustrate the relevance of moral/intrinsic and instrumental/money motivations in Figure 1. In this skeletal analysis, the utility costs are illustrated on the y-axis, and the extent of dishonesty is illustrated on the x-axis (running from ‘total honesty’, at the origin, to ‘total dishonesty’). If the individual is ‘totally honest’, (‘psychic’) moral costs (MCt) are zero. The more that the individual is dishonest, the greater the moral costs, with ‘total dishonesty’ generating very high costs. Income costs (YC) are essentially tY, where t is the tax rate and Y represents income (for the benefit fraud case, YC would be the value of potential benefits not fraudulently claimed). When the individual becomes more dishonest, YC = tY falls. This cost is zero when the individual is ‘totally dishonest’3.

For any individual, ‘optimal honesty’ occurs where total costs (YC + MCt), captured as TCt in Figure 1b, are minimized. This will occur where the absolute values of the slopes of MCt (the moral costs) and YC (the income costs) are equalized. As drawn, optimal honesty is at Ht* in Figure 1b, which represents a high degree of honesty. Empirical evidence, using different research methods, accords with this depiction. In an ‘indirect’ applied econometrics approach, Engström and Hagen (2017) analyze the consumption patterns of the ‘tax evader tarred’ self-employed. They try to establish whether their consumption is consistent with their reported income. Employing the permanent income concept, they find, for their Swedish data, that correcting for transitory income fluctuations in current income, leads to an overestimate of income underreporting by as much as 40%. Clearly, this is a substantial correction in the ‘honesty’ direction.

In a ‘direct’ behavioral approach, Cohn et al. (2019) had researchers drop 17,000 apparently “lost wallets” in 355 cities in 40 countries. Surprisingly, the majority of wallets containing money were reported, with reporting rising with the size of the ‘find’. Some 72% of wallets containing a 100 USD were reported, compared to 61% of wallets with just over 13 USD. Interestingly, only 46% of wallets containing no money were brought forward. It appeared that individual honesty was not dependent on economic gain (the instrumental-pecuniary motivation) but rather how bad the act of dishonesty made them feel (the intrinsic-psychic motivation). One explanation was altruism, so that individuals show empathy for the (other) person who lost the wallet. More consistent with the framework outlined here, and favored by the authors themselves, is that individuals want to maintain a positive image of themselves as honest moral beings, so that not reporting wallets containing more money engenders greater shame or moral costs to themselves. The economics of tipping raises similar issues of contrasting pecuniary and non-pecuniary motivations. Azar (2020) reports the results of a survey he conducted in the USA and Israel on the main reasons for tipping. ‘Tipping being a social norm’ and ‘Show gratitude’ attracted strong support (‘Avoiding feeling guilty’ was also prominent in the USA responses). Psychological and social motivations for choices seem commonplace for the majority of individuals. However, what about the minority. Not all individuals are the same, and some individuals are much more extreme than others in their choices. The ‘least cost’ optimal honesty approach allows ‘corner’ solutions.

For individuals displaying ‘real honesty’ there is an infinitely steep MCt curve. These ‘saints’, whose moral rectitude is beyond reproach, always incur YC and enjoy MCt = 0. Real ‘sinners’, the morally bankrupt, are at the other end of the scale; MCt is the x-axis, so that cost minimization always predicts complete dishonesty. In practice, it is likely that an individual’s ‘optimal honesty’ will lie somewhere in the interior. Corner solutions can also be obtained if the MCt (YC) curve is everywhere steeper in absolute value than the YC (MCt) curve producing ‘apparent’ saints and sinners. ‘Apparent’ because these individuals give credence to illegal pecuniary gains and moral costs, respectively, thereby keeping the door to future dishonesty and honesty ajar as slopes might change.

The picture is of a minority of saints and sinners alongside a majority of individuals who dabble with a socially acceptable level of tax evasion. Or as psychologist Abigail Marsh put it, commenting on the lost wallets study noted above: “…what I like about this study is that it supports so much of the data out there… that most people are trying to do the right thing most of the time.” (Quoted in Fieseler 2019, emphasis our). The optimal honesty approach, so far, sheds light on the second and third implications above—fiscal crimes will not be endemic and not all will act in the same way. But what about the first implication? How is an asymmetrical attitude to tax evasion and benefit fraud to be explained?

Cullis et al. (2015) provide a hedonic coding explanation of this implication. Tax evasion is about illegally not putting into the community pot, whereas benefit fraud is about illegally taking out of the community pot. With the former coded as a foregone gain and the latter as a loss, the particular shape of the Kahneman and Tversky (1979) value function does the rest. Illegal monetary gains of equal size are labeled by their different source/mental account. However, can this coding be given a more fundamental moral foundation? Zamir (2012) has considered why tort law is more developed than unjust enrichment law and why there is more constitutional protection of civil and political rights compared to social and economic rights. He argues that the explanation is premised on moral intuition. He draws attention to the concept (in moral philosophy) of ‘common sense morality’. If morality is relevant, there is every likelihood that individuals will be more condemnatory of benefit fraud than of tax evasion ceteris paribus.

By definition, ‘common sense morality’ is deontological. It focuses on action rather than outcome. Zamir (2012, pp. 876–77) explains that: “Deontology does not primarily judge the morality of an action (or anything else) according to its outcomes but rather focuses on the morality of the action itself”. With ‘common sense morality’, individuals weight the possibility of “…intentionally or actively harming other people…” far more heavily than the benefits that might be achieved. Zamir (2012) provides many examples. The first of these is the “…instinctive immorality of killing one person to harvest organs that will save the lives of three others”. He notes that such acts are “…inherently wrong, they are impermissible even as a means to furthering the overall good.” (our emphasis). Common sense morality “…distinguishes between harming a person and not benefiting her”.

When focusing on individuals’ preferences, the “… law conforms to prevailing moral intuitions… (that)…are closely linked to notions of reference points and loss aversions…” (ibid, p. 876). The implication is that individuals are likely to perceive tax evasion as a foregone or unrealized gain to the community (via ‘not benefiting’ the public purse), while benefit fraud is likely to be perceived as a loss to the community (via ‘harming’ the public purse). The moral costs of taking from the public purse are instinctively greater than the moral costs of not contributing to the community. As the intrinsic value of honesty, when claiming government benefits, is likely to be greater than the intrinsic value of paying tax honestly, the moral costs of dishonesty when fraudulently claiming government benefits are likely to be greater than the moral costs of dishonesty when making false income declarations.

The implications are that, in Figure 1, MCb (the costs of acting dishonestly) when committing benefit fraud are: (i) greater than MCt (the costs of acting dishonestly) when evading tax, and (ii) that MCb will increase at a faster rate when committing benefit fraud than MCt when committing equivalent tax evasion (see Figure 1a). As a consequence, in Figure 1b in the cost-minimizing model, the ‘optimal’ level of honesty will be greater when individuals consider benefit fraud than when they consider tax evasion (referring to TCb, the ‘optimal level of honesty’ will be higher at Hb* compared to TCt and Ht* in Figure 1b).

The impact of perceptions of the ‘wrongness’ of benefit fraud as a brake on criminal activity is likely to be greater than the impact of perceptions of the ‘wrongness’ of tax evasion as a brake on tax evasion. Is there evidence consistent with this? United Kingdom government statistics for the financial year ending in 2024 estimate the ‘tax gap’ between what is collected and what should be collected at 4.8% of total theoretical tax liabilities, or £39.8 billion in absolute terms (see HM Revenue 2024). For the financial year ending 2024, the level of fraud and error in the Great Britain benefit system amounted to £9.7 billion in absolute terms and was 3.7% of total benefit expenditure (see Department of Work and Pensions 2024). If the direct translation of these data to the theoretical depiction in Figure 1 were permissible and total honesty is 100%, then Ht* = 95.2% and Hb* = 96.3%. While this is casual empiricism, it is clearly not inconsistent with the depiction.

4. Policy Implications and Discussion

The policy relevance of this paper can be captured by reference to Figure 1. In the standard Allingham and Sandmo (1972) model, the YC income costs line would, in effect, become the expected utility (EU) maximizing income declaration (D) out of actual income (Y) determined by:

EU = (1 − p) U(Y − tD) + pU(Y − tD − F[Y − D])

(3)

where t is a proportional tax rate, p is the probability of detection, and F is the fine rate that exceeds t. This narrowly instrumental approach naturally directs policy towards setting t, p, and F to influence D. It might be noted that raising p will have a limit in terms of its real resource implications, and raising F will have a limit in terms of making the punishment commensurate with the crime. That said, ideal policy would result in D = Y, no false declarations. It relies on one blade of a scissors. This formulation ignores the MC curves in Figure 1, highlighting the limitation of ignoring intrinsic/moral costs.

The difference between the set of variables that influence the behavior of ‘homo and femina economicus’ and the set of variables that influence ‘homo and femina realitus’ is evident when comparing Equations (2) and (3). If policies are to be made actionable, information is required about the impact of each variable. Different empirical methodologies have been employed. Questionnaire surveys and interview surveys report individuals’ attitudes toward changes. Experiments and field trials indicate individuals’ actual responses. This literature has identified the importance of the way a decision is framed. Tax authorities now frame the decision to comply by informing taxpayers of the prevalence of honest compliance by their peer group. The literature has also identified behavioral anomalies (when ‘homo and femina realitus’ responses are deemed ‘irrational’ by comparison with ‘homo and femina economicus’ responses). With evidence that individuals might regret ‘irrational’ decisions, there is a case for government intervention (by way of a ‘nudge’) that encourages individuals to avoid mistakes. Such intervention is premised on ‘liberal paternalism’ (Thaler and Sunstein 2009). Of course, policy might also be designed simply to minimize criminal activity.

Once ‘homo and femina realitus’ supplants ‘homo and femina economicus’ as the relevant caricature actor, the tax evasion/benefit fraud policy palette is transformed from monochrome to Van Gogh vibrant. Alm et al. (2023) illustrate the many interventions that are possible when government relies on a behavioral approach to improve tax compliance.

The discussion above focused particularly on the second characteristic of ‘homo and femina realitus’—bounded self-interest. A subdivision was made between an internal and external stimulus to the individual stimulus to ‘honesty’. Most studies identify two important determinants of the perceptions of the intrinsic value of action (e.g., Frey 1997; Luttmer and Singhal 2014). The first is the value individuals attach to the moral (or social) norm itself-i.e., to acting honestly. The second is the importance individuals attach to evidence that others acknowledge honest behavior (and reciprocate). Government policies might trade on one (or both) of these two determinants of citizens’ perceptions of the moral costs of dishonesty. As noted, it may not just be about how others view you but also about how you view others’ actions. ‘Hot lines’ that invite you to report suspected tax evaders and benefit fraudsters allow you to take some responsibility for the fiscal environment. In this vein, the UK His Majesty’s Revenue and Customs (HMRC) publishes a ’shame list’ of the names of individuals who have incurred a penalty for deliberately providing inaccurate documents or failing to comply with the tax rules where there is more than £25,000 of tax at risk. Characteristic 1—bounded abilities were subdivided into: (a) bounded rationality in choices and (b) bounded self-will over actions. Bounded rationality recognizes the cognitive limitations of individuals and their information stock. With respect to tax/benefit systems, the level of confusion and misunderstandings among the vast majority of individuals seems legion. Less complicated tax codes and benefit regulations recommend themselves. Policies providing information and illumination could be intensified and the ‘fiscal connection’ between taxes and public sector benefits highlighted. Information has to come from ‘trusted’ sources if it is to be acted on. Who are the community leaders that are listened to rather than using the organs of government who may be mistrusted. Policy in Practice estimated in 2024 that some £23 billion in benefits go unclaimed each year due to stigma (how others perceive you) and the complexities of the welfare system (Ghelani and Walker 2024; emphasis ours). Bounded self-will is consistent with individuals deliberately overpaying tax in a withholding system using the tax system effectively as a compulsory savings scheme—a self-control mechanism. Individuals who succumb to the temptation of tax evasion and/or benefit fraud may regret their actions and feel guilt. In this respect, tax and fraud amnesties would have appeal. Characteristic 3—preferences that are endogenous and malleable suggests any communications with individuals should take advantage of what is known about ‘framing’ effects to encourage honesty—the micro perspective. As regards the macro perspective, Jones et al. (1998) use a shifting (endogenous to government policy signals) preference map to indicate how government redistribution (that acknowledges the involvement and competence of private voluntary redistribution) will mitigate the ‘one for one crowding out’ predicted by neoclassical theory. Proxying the efficiency and effectiveness of government spending programs is often part of econometric work in this area. While the optimal allocation of resources between different fiscal enforcement policies on tax evasion and benefit fraud remains to be determined, it is evident from this brief discussion that there are many diverse fiscal enforcement policies that can be explored if ‘homo and femina realitus’ are front and center. The use of policy prescriptions based on findings in behavioral economics is not without controversy. Many of the policies are seen as paternalistic and have been criticized on the grounds “they arbitrarily privilege one set of preferences over another” (Le Grand 2018, p. 281). Here the objection seems to have less force because the behavior under discussion is illegal. Fiscal laws and regulations are about imposing a ‘state’ set of preferences. The issue may then become whether the tax code and benefit system rules are set by a legitimate authority/government. In this respect, it might be expected that legitimacy might be attributed to more democratic regimes than to autocratic ones.

As regards methodological considerations, compared to the standard model, the analysis presented here lacks elegance and prescriptive precision. However, at the core of behavioral economics “…is the belief that increasing the realism with which individual behavior is seen will improve the ability to predict behavior and to devise policies” (Alm and Sheffrin 2017, p. 4). If the approach appears ‘messier’ (because it does not always assume that individuals are “…rational, outcome-oriented, self-controlled, selfish, and egoistic…”), it is now “… an essential part of the public sector economics dialogue” (Alm and Sheffrin 2017, pp. 4–9). In this context, Olsen et al. (2018, p. 408) argue that: “a model should be evaluated in terms of the reasonableness of its assumptions, its predictive power, and its potential usefulness to policymakers”.

5. Conclusions

Decisions with a moral dimension raise the question of the moral costs associated with the choices that individuals make. Here individuals display optimal honesty if they weigh illegal pecuniary gain alongside non-pecuniary intrinsic/moral costs associated with crimes in a cost-minimizing way. More specifically, this paper explores the role of moral costs in defining ‘optimal’ honesty in the context of tax evasion and benefit fraud. The framework is consistent with shedding light on three, at minimum, uncomfortable implications of the ‘standard’ expected utility approach to fiscal crimes by suggesting: a high level of individual honesty, allowing for honest, mostly honest, and dishonest individuals to be observed alongside one another, and rationalizing benefit fraud as being viewed as more morally wrong than pecuniary equivalent tax evasion. Extant research is used to support the theoretical depiction in Figure 1. In addition, variables relevant to defining an individual’s ‘moral hinterland’ as a precursor to possible econometric work were explored.

Considering the question posed at the end of Section 1, “honesty is the best policy” when the moral costs of dishonesty are high relative to the pecuniary payoff from dishonesty. While this may be somewhat of a tautology, what is not a tautology is the answer to the second part of the question, which asks what is “the best policy for honesty”. Here the answer is an eclectic set of policies that include detection rates and penalties but also a carefully chosen and culturally tailored selection of policies, often associated with behavioral economics and the ‘nudge’ agenda, that serve to raise the moral costs of dishonesty.