Perspective Chapter: Rationality, Social Norms, and Dishonesty in Everyday Life

Abstract

Many forms of unethical behavior are social in character. We are often heavily influenced by what others do. The social characteristic of dishonesty and crime nevertheless does not contradict that there is a rational element in people’s behavior. But people usually act within social conditions that are not determined by themselves. Inspired by recent findings in experimental economics, this chapter combines the standard economic model of crime, which emphasizes individual rationality, with more criminologically oriented approaches, which emphasize norms, social context, and peer influences. The aim of the chapter is to develop a framework for analyzing social interaction and how situational factors influence the spread of crime and other types of unethical behavior. It is demonstrated that various crime-preventing interventions can have effects that deviate significantly from what the standard model of rational crime predicts. Finally, it is argued that the distance between the theory of rational crime and more criminological approaches to crime is not as great as many make it out to be.

1. Introduction

Fifty-five years have passed since Gary Becker published the article Crime and Punishment: An Economic Approach [1]. The paper has been highly influential and has inspired much theoretical and empirical work on the problem of crime. The theoretical framework developed by Becker was soon given the name ‘the standard economic model of crime’ and has since dominated the field of law and economics. At the time when Becker published his work, the theory of rational choice was the most consistent and systematic framework for describing and predicting human behavior. For Becker, the idea of rational choice was also a natural starting point for understanding criminal behavior. He argued that there is nothing special about the decision to involve oneself into something criminal. It is a question of costs and benefits, as it is in all areas in life that involve choosing between alternatives according to rational choice theory. Hence, the economic model of crime holds that a potential criminal rationally weighs expected benefits of crime against expected costs.

While Becker’s work quickly gained recognition and popularity among economists, the economic model of crime has been hotly debated and criticized in other academic fields such as psychology, sociology, and criminology. The model’s lack of empirical relevance is one type of criticism that has been raised [2]. The main policy implication of the economic model of crime is that the government can reduce crime by increasing the cost of committing crime, where the probability of detection and the penalty if detected are the most important factors. Increasing the cost of crime will lead to a decrease in the number of crimes, an effect known as the ‘deterrence hypothesis’. However, despite a huge empirical research effort, there is no consensus on the validity of this hypothesis [3]. Some have also pointed out that in many settings, the probability of detection is extremely low, while the payoff from an offense is high. This should lead to a very high crime rate, but this is often not the case. The economic model of crime therefore seems to predict much more crime than what we actually observe [4]. Others hold that the theory of rational crime oversimplifies the decision-making process by not considering the many complex factors that can influence an individual’s behavior, for instance, different social and environmental factors. There are also those who simply reject the model because of its underlying behavioral assumptions, in particular the idea of rationality. Various cognitive biases strongly influence people’s behavior, and these biases are difficult to reconcile with the idea of rationality.

During the past few decades, the theory of rational choice has also been questioned by the economics profession itself. Experimental economics is a field of economics that uses controlled laboratory experiments, field experiments, or online experiments to study economic behavior. An important goal of experimental economics is to test the theory of rational choice to better understand how people make decisions [5, 6]. The methodology in experimental economics allows researchers to test economic theories by manipulating certain variables and observe how these changes affect the decision making of participants in a controlled environment. Experimental methods have particularly been used to test different hypotheses on how people make decisions under uncertainty and risk, how they respond to incentives, how they bargain, and how they interact and cooperate with others [7].

Experimental economics has also made significant contributions to the field of crime and dishonesty by providing insights into how people make decisions related to these issues in experimental settings. Dan Ariely is a leading researcher in this field. By providing insights into how people make decisions related to stealing, cheating, and lying in controlled environments, Ariely and colleagues have made several significant contributions to the study of dishonesty and unethical behavior [4, 8, 9, 10, 11, 12]. All in all, Ariely’s research provides little support for the economic model of rational crime. People’s dishonesty is only weakly influenced by expected (material) costs and benefits. Social context and psychological factors seem to be more important. Still, Ariely’s results do not provide sufficient evidence to completely reject the economic model of crime. People’s behavior is not unaffected by incentives.

The aim of this chapter is to review and discuss some of the main research findings of Ariely and colleagues, in particular their contribution to understanding how social context influences dishonesty and in what way people’s behavior deviates from the theory of rational crime. It is demonstrated that by incorporating key findings from Ariely’s research into the traditional economic model of crime, we can gain deeper insights into some of the shortcomings of the traditional model while at the same time putting Ariely’s experimental findings into a more general theoretical framework. By applying this theoretical framework, the aim of this chapter is also to show that the distance between the theory of rational crime and more psychological and criminological approaches to crime is not as great as many make it out to be. Finally, a wider version of the theory of rational crime can also contribute to a better understanding of why recent empirical research on the relationship between stricter deterrence and the rate of crime achieves such divergent results.

2. The problem of dishonesty

Cheating, stealing, and other forms of unethical behavior represent major challenges for society. While the standard economic model of crime seems to predict more crime than what we witness, dishonesty of one form or another is not difficult to spot [4]. While the media usually highlight the most extraordinary cases, these extraordinary cases often join the ranks of cheating and stealing in society more broadly. Citizens avoid taxes, bureaucrats and politicians accept bribes and supply corrupt services, students cheat on exams, people call in sick when they are not, people trick insurance companies into paying them more than they are entitled to, customers steal from shops, and so on. These are examples of “unethical behaviors committed by “ordinary” people who value their morality highly but cut corners when faced with an opportunity to gain from dishonest behavior.” ([12], p. 125). Stated differently, a lot of crime and dishonesty are not limited to examples of one person causing harm to many. Rather, the general picture is that many cheat a little, in sum causing harm to many.

Examples of the same pattern are also easy to find in the corporate world [8]. Over the past few decades, we have witnessed several corporate scandals that have had major consequences for society, for the economy, and for individuals. Just to name a few: The Enron Scandal in 2001 is often mentioned as the most infamous corporate scandals in history. Enron, an energy company based in Houston, USA, was found to have largely manipulated its financial statements to inflate its stock price and deceive investors. This led to the bankruptcy of the company and the loss of thousands of jobs and billions of dollars for investors. Several executives of the company were convicted of fraud and other crimes related to the scandal. The year after, the telecommunications company WorldCom was found to have falsified financial statements in order to hide billions of dollars in losses. The scandal led to the company’s bankruptcy. The solid and well-known German car manufacturer Volkswagen (VW) was also close to bankruptcy after what is often referred to as “dieselgate”. The VW emissions scandal began to roll when the United States Environmental Protection Agency announced that the company had installed software in certain diesel vehicles that could detect when the vehicle was being tested for emissions. The software would activate controls that reduced emissions of nitrogen oxides (NOx) to pass the test, but during normal driving, the controls would be turned off, leading to higher emissions of NOx. The company was caught and subsequently admitted to the use of the software. Another case that involved both money and politics is the so-called Cambridge Analytica scandal in 2018. Cambridge Analytica, a political consulting firm, was found to have harvested the personal data of millions of Facebook users without their consent and used it to influence their opinions and behavior during the 2016 US presidential election. Finally, the financial crisis that struck the world in 2008 was also a result of unethical behavior and shady dealings, particularly within the financial sector. Creative banking, lax lending standards, and development of complex financial instruments, which proved to destabilize the entire financial sector, forced governments around the world to intervene and to stabilize the financial system, including bailouts to many financial institutions. Despite governments’ effort, the financial crisis led to a strong global economic downturn, hurting many families and communities.

All these examples illustrate that the problem of dishonesty is usually not that it is one person or one corporation that causes harm to many. It is rather as Gino et al. ([10], p. 4) emphasize: “an increasing amount of empirical evidence in the social psychology and management literatures demonstrates that dishonesty often results not from the actions of a few people who cheat a lot, but from the actions of a lot of people who cheat a little.” This observation has been confirmed by Dan Ariely through his own research [13]. Ariely has carried out a large number of experiments to study various aspects of cheating and dishonesty, where the participants have been incentivized with money. The experiments have involved more than 30,000 participants around the world. Ariely reports that 12 of these acted maximally dishonestly and as a result have run away with approximately $300. In contrast, around 20,000 participants behaved “a little” dishonestly and as a result have run away with around $50,000. This illustrates that the problem of cheating and dishonesty is not a question of picking out a couple of rotten apples. It is the group of “small cheaters” that represents the largest cost to organizations, firms, and society. But how should such “everyday crime” be dealt with? Designing effective measures against unethical behavior requires a deeper understanding of the factors that drive such behavior, and this is where Ariely’s research comes in.

3. Dishonesty explained

Dan Ariely and colleagues have conducted numerous experiments to explore various factors that may influence dishonesty, including the factors that are central to the economic model of crime. The experiments are quite simple yet cleverly designed. Here is an example (see [4, 9, 10] for further details). The participants are invited to take part in an experiment where they are briefed to believe that it is their mathematical skills that will be tested. They are given a sheet of 20 simple math problems to be answered within 5 minutes, where they are offered money per correct answer (the amount varies over different versions of the experiment). When 5 minutes have passed, the participants hand in the answer sheet, which is checked, and the participants are paid money according to the number of correct answers. This is the control group.

In another version of the experiment, called the shredder condition, participants are asked to do the same but with a twist. Participants correct their answers themselves, then run the answer sheet through a shredding machine, and self-report how many correct answers they have. Since the answer sheet is shredded, participants can cheat without the possibility of being detected. What happened in this version of the experiment? The participants in the shredder condition claimed to have solved on average two more questions compared to the control group. That is, they seem to cheat somewhat but much less than what the economic model of crime suggests [4].

What happened when the participants were offered more money for each correct answer, and the probability of detection remained zero? Little happened. In fact, the extent of cheating decreased when participants were promised the maximum amount of $10 for each correct answer. This is clearly contrary to what the economic model of crime predicts.

The upshot of all of this and similar types of experiments is that the participants’ behavior is quite insensitive to changes in the variables that drive behavior in the economic model of crime. How can this be explained? Ariely argues that most people have a strong self-concept of being honest and that this self-concept is important to them. This still does not prevent people from cheating or engaging in dishonest behavior. People seem to be equipped with a kind of “moral wiggle room” that gives quite a lot of flexibility when it comes to bending or breaking rules. That is, people are not completely honest or dishonest, they rather seem to operate within a range of acceptable behavior. People cheat or engage in dishonest behavior but only up to a point where the amount of cheating would not damage their self-concept of being an honest person. In the words of Ariely ([4], p. 27):

our behavior is driven by two opposing motivations. On one hand, we want to be able to view ourselves as honest, honorable people. … On the other hand, we want to benefit from cheating and get as much money as possible (this is the standard financial motivation). Clearly these two motivations are in conflict. How can we secure the benefits of cheating and at the same time still view ourselves as honest, wonderful people? This is where our amazing cognitive flexibility comes into play.

Another interesting finding from the experiments described above is that people’s behavior is sensitive to small changes in situational factors (other than material costs and benefits). That is, people do not seem to have a fixed set of moral standards. Their moral wiggle room can expand or contract depending on the social context. Ariely [4] and Mazar et al. [9] show that people’s degree of dishonesty can be particularly influenced by:

• Distance: The greater the physical, psychological, or emotional distance between the person and the target of the dishonest act, the more likely the person is to engage in dishonesty.

• Diffuseness: The more vague or abstract the description of the target of the dishonest act, the more likely people are to engage in dishonesty.

• Doping: The more people can tie their actions to a higher purpose, such as helping a good cause, the more likely they are to engage in dishonesty.

• Moral reminders: When people are reminded of moral values or ethical principles, they are less likely to cheat.

4. Dishonesty, norms, and social dynamics

While most people have a general inclination to be honest, their moral standards are not fixed. As just noted, people’s tendency to stretch ethical standards is affected by situation and context. Ariely and colleagues have also found that peer influence is a strong factor in unethical behavior [10]. Dishonesty can spread through social contagion, a process by which people’s behavior is influenced by the behavior of those around them. In one experiment, participants were given the opportunity to cheat on a task and were told that their performance would be compared to that of others in the group. The results showed that participants were significantly more likely to cheat when they were told that others in the group had cheated, compared to when they were told that others in the group had not cheated. In another experiment, participants were exposed to a person (an actor) who exhibited dishonesty, a manipulation that significantly increased dishonesty in the rest of the group. This effect was particularly strong when the dishonest person was clearly identified as an in-group member [10].

4.1 Dishonesty and social norms

Since people are more likely to cheat when they perceive that others around them are also cheating, unethical behavior can escalate. Mazar and Ariely [8] argue that social norms play an important role here. Social norms shape the extent to which people consider certain forms of dishonesty to be acceptable and therefore also people’s self-concept of what it means to be an honest person. Social norms serve as a benchmark against which an individual compares his behavior.

It is well documented more generally that people care about the norms and values of their society [14, 15, 16]. People want to conform to social norms to avoid disapproval from others, which creates an inner feeling of discomfort. As part of socialization, people often also internalize social norms and self-enforce them. Internalized norms thus motivate us through our commitment to them, created by internally generated rewards or punishments. Feeling of guilt, shame, and remorse may arise if a social norm is violated, even if no one else knows about it. Hence, social norms influence our behavior through our inner reward system. Social norms have a ‘grip on the mind’, as Elster puts it [17].

4.2 Integrating social norms into the economic model of crime

As noted, the economic model of crime is based on the idea that people engage in crime (dishonesty) when they believe the potential benefits outweigh the potential costs. The decision is based on a cost–benefit analysis, where the following elements are essential: (1) the payoff from crime, which we denote as $a$; (2) the payoff from the legal alternative, which we denote as $b$ (where $a>b$); (3) the probability of being caught, which we denote as $s$; and (4) the magnitude of punishment if caught, which we denote as$c$. On the basis of these variables, the individual chooses the action that maximizes his payoff, which means that the individual follows the criminal strategy if the expected payoff from crime, ${\pi }^C=\left(1-s\right)a-\mathit{sc}$, is greater than the alternative, $b$. If we set $b=0$ (to economize on notation), we can clarify the decision problem as illustrated in Figure 1, where the black solid line represents the expected payoff from crime (${\pi }^C$).

Figure 1.

The decision problem.

The individual will follow the criminal strategy if ${\pi }^C>0$, which is met up to the point where $s=s^{*}$ in Figure 1. That is, the individual will follow the criminal strategy if the probability of being caught (s) is small enough (less than $s^{*}$). If $s>s^{*}$ the individual will choose the legal strategy.

An increase in the cost of crime due to harsher punishment ($c$ goes up) shifts the payoff from crime schedule (${\pi }^C$) down. For $s=s^{*}$, the individual will now follow the legal strategy, and this will be the case down to the point where $s=s^{**}$. These effects capture in a nutshell the basic message of the standard economic model of crime. Increasing the probability of detection or the magnitude of punishment will reduce or even eliminate crime.

The basic idea behind introducing a norm of honesty into this framework is that norms also influence behavior. As Mazar et al. ([9], p. 634) note: “if a person fails to comply with his or her internal standards for honesty, he or she will need to negatively update his or her self-concept, which is aversive. … Notably, this perspective suggests that to maintain their positive self-concepts, people will comply with their internal standards even when doing so involves investments of effort or sacrificing financial gains.” Applied to the context of rational crime, this suggests that people who consider doing something illegal will be influenced not only by external costs and benefits but also by the way the act might make them perceive themselves.

Assume a large group or a society where there exists a norm saying that “you should not commit crime”. Assume further that noncompliance with the norm leads to negative rewards (i.e., punishments) or what I will denote a moral cost. Since individuals differ in their psychological constitutions, they may have different moral costs. We capture this by assuming that individuals are heterogeneous with respect to the moral cost, where z_i denotes the moral cost for individual i.

Integrating this idea into the framework illustrated in Figure 1 produces a small but potential important adjustment. The expected payoff from the criminal strategy is now ${\pi }^C=\left(1-s\right)a-\mathit{sc}-z_i$, and the individual follows this strategy if this is greater than the legal alternative, which gives

$$z_i<\left(1-s\right)a-\mathit{sc}$$E1

Eq. (1) says that only those with sufficient low moral costs follow the criminal strategy. The existence of a moral cost $z_i$ leads to a shift down in the payoff from crime schedule (${\pi }^C$). For a sufficiently high moral cost, an individual will not commit crime even if the probability of being caught ($s$) is zero. The last situation is illustrated in Figure 1 by the green payoff from the crime schedule.

People care about norms, but they also care about what others do. To bring this last element into the analysis, we follow Funk [18] and Weibul and Villa [19] and assume that the strength of the honesty norm decreases in the level of crime. This can be formalized by writing up the moral cost of breaking the norm in the following way:

$$z_i={\alpha }_i-\mathit{\sigma p}$$E2

where ${\alpha }_i$ is a fixed moral cost component, which can vary across individuals, and $p\in \left[0,1\right]$ denotes the share of dishonest people in the group or in society or simply the crime rate. The term $\mathit{\sigma p}$ thus takes care of the effect of decreasing moral costs when crime becomes more widespread.

It is instructive for the analysis that follows to illustrate the decision problem of an individual using a slightly modified version of Figure 1, where we measure the crime rate in society along the horizontal axis. As before, the ${\pi }^C$-schedule represents the expected payoff to an individual from committing a crime, where the moral cost specified in (2) is included. The expected payoff from crime is ${\pi }^C=\left(1-s\right)a-\mathit{sc}-{\alpha }_i+\mathit{\sigma p}$. An individual will follow the criminal strategy if ${\pi }^C>0$ as illustrated in Figure 2.

Figure 2.

The decision problem with a norm of honesty.

Figure 2 shows that committing crime is preferable to not committing crime if the crime rate is greater than $p^{*}$. Hence, $p=p^{*}$ is a critical threshold or cross-over point, defined as the proportion of people that must be criminal before an individual joins in. That is, what an individual chooses to do depends on what others do.

Figure 2 also illustrates that when people are heterogeneous with respect to the size of the moral cost (${\alpha }_i$), people have different cross-over points. An increase in the moral cost shifts the payoff from the crime schedule downward and moves the critical threshold $p^{*}$ to the right.More people need to be criminal before the individual joins in. A decrease in the moral cost has the opposite effect. For individuals with high moral costs, it is possible that ${\pi }^C<0$ for all $p\in \left[0,1\right]$, which means that these individuals will not commit a crime regardless of the crime rate in the group or in society.

4.3 Thresholds, tipping, and social dynamics

The main insight from Figure 2 is that an individual’s decision whether to commit a crime is conditional on the behavior of those around him. It follows from this that individuals facing the same external costs and benefits related to crime can make very different decisions. When individuals have different moral costs, they also have different thresholds for committing crime, where these thresholds act as tipping points in decision making by marking the point where an individual comes to believe that committing a crime is preferable to not committing a crime. As McGloin and Rowan ([20], p. 485) note: “Despite the prominence of threshold models in the sociological literature, criminological research leveraging this concept has been sparse”. As we will show below, there is something to be gained from closing this research gap. By incorporating the idea that individuals are influenced by social norms and have different thresholds for committing crime into a traditional model of rational crime, we reach results that differ from the traditional model in various respects - results that are difficult to derive without a more formal analysis. However, a formal analysis requires a more precise formulation of how moral costs are distributed over the population.

4.3.1 A stylized numerical example

The following simple numerical example, inspired by Kuran [21], can be illustrative to demonstrate how a small change in the distribution of moral costs may generate a large difference in the crime rate at the societal level. Consider again the moral cost $z_i$ specified in Eq. (2). Suppose that people are equally sensitive in their moral costs when it comes to changes in the level of crime in society (they do not differ with respect to the second term in Eq. (2)), but they differ when it comes to the size of ${\alpha }_i$. Remember also from Figure 2 that the size of ${\alpha }_i$ determines the size of the critical threshold $\left(p\right)$ for an individual (for a given level of $a$, $c$, and $s$), where this threshold specifies the proportion of people who must be criminal before the individual joins in. An individual with a large ${\alpha }_i$ will have a high threshold for committing a crime, while an individual with a small ${\alpha }_i$ will have a lower threshold.

To fix ideas, imagine a population consisting of 10 individuals – call them 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 – where individual thresholds are distributed as follows: Individual 1 has a threshold of 0, individual 2 and 3 have a threshold of 20, while the remaining individuals have higher thresholds as summarized in the upper panel of Table 1. Let us denote this group of individuals by society A.

A	Individual	1	2	3	4	5	6	7	8	9	10
	Threshold	0%	20%	20%	30%	40%	50%	60%	80%	100%	100%
B	Individual	1	2	3	4	5	6	7	8	9	10
	Threshold	0%	10%	20%	30%	40%	50%	60%	80%	100%	100%

Table 1.

Distribution of threshold in society A and B.

The outcome in society A is clear: Individual 1 will commit crime irrespective of what others do. But since individual 1 constitutes only 10 percent of the population, and that is below the thresholds of each of the others, all of them therefore opt for the legal alternative. The stable outcome is that only 1 in 10 individuals commits crime.

Now perturb the distribution of thresholds slightly, as illustrated in the lower panel of Table 1, which we denote society B. Remove individual 2 with threshold 20 and replace him by one with threshold 10. Society A and B are essentially identical, but the outcome in society B is quite different. The outcome in society B can be described as a bandwagon effect: As above, individual 1 will commit crime irrespective of what others do, but this will now activate individual 2, again activating individual 3, and so on, until individual 7 has committed crime. Since individuals 1 to 7 constitute 70 percent of the population, and that is below the thresholds of the remaining three individuals, the process will stop. The stable outcome is that 7 in 10 individuals commit crime.

This skeletal numerical example illustrates two main points:

1. Two societies that are identical when it comes to the payoff from crime ($a$), the probability of being caught ($s$), and how severely crime is punished ($c$) may end up in very different situations due to minor differences in the process of aggregation, produced by a small change in the threshold for one individual (in this case individual 2).

2. It may be misleading to infer individual dispositions from aggregate outcomes [22]. It is tempting, for example, to conclude that society A mostly consists of individuals with high moral standards and who keep within existing laws and regulations, while society B is characterized by the majority being unscrupulous criminals. This is clearly not correct. As we know, since we have constructed the example, society A and B are almost identical in individual dispositions.

4.3.2 A more general model

A more general model can be developed by assuming a society with many individuals where each individual is associated with a level of moral costs (see Andvig and Moene [23], Naylor [24], Lindbeck et al. [25], and Funk [18] for a related analysis in the context of corruption, collective action, work incentives in the welfare state, and crime, respectively). Remember from Figure 2 that an individual will follow the criminal strategy if the expected payoff from committing a crime ($\left(1-s\right)a-\mathit{sc}-z_i$) is greater than the payoff from a legal alternative, which is identical to the inequality

$${\alpha }_i<\left(1-s\right)a-\mathit{sc}-\mathit{\sigma p}$$E3

Eq. (3) is essentially identical to (1), but here we have used $z_i={\alpha }_i-\mathit{\sigma p}$. Equation (3) states that only those with sufficient low moral costs follow the criminal strategy, constituting the crime rate ($p$) in society. What the crime rate will be depends on how ${\alpha }_i$ is distributed in the population.

When the distribution of ${\alpha }_i$ is bell shaped, we may have the case illustrated in Figure 3. The $f\left(·\right)$ curve represents the cumulative distribution function, illustrating how many individuals in society follow the criminal strategy for a perceived level of crime ($p^e$) and for given levels of $a$, $c$, and $s$. Individual decisions are guided by the following process: Each individual knows his own moral cost ${\alpha }_i$, observes the level of crime in society, and adjusts his strategy in the next period given that information. The horizontal axis in Figure 3 represents the perceived crime rate, while the vertical axis records the realized crime rate. The 45-degree line is for reference. It shows all points where the perceived crime rate is equal to the actual crime rate.

Figure 3.

The equilibrium crime rate.

Point A in Figure 3 is a stable social crime equilibrium. To see why, imagine that the perceived crime rate somehow starts out at $p_2^e$. The $f$-curve shows that this produces a crime rate that is higher than the perceived. Having turned out to be an underestimation, the initial perceived crime rate will be revised upward, inducing more people to switch to the criminal strategy because of a weakening social norm. This will now activate other individuals with slightly higher moral costs. This process will continue until the perceived crime rate is $p_1^e$. If the perceived crime rate somehow starts out to the right of $p_1^e$, the process will go the other way around. To the right of $p_1^e$, the actual crime rate is lower than the perceived rate. Having turned out to be an overestimation, the initial perceived crime rate will be revised downward, inducing more people to switch to the legal strategy because of a strengthening of the social norm. This will now activate other individuals with slightly lower moral costs, and this process will continue until the perceived crime rate is $p_1^e$.

Only in point A does the actual crime match the perceived crime rate that generated it. A perceived crime rate of $p_1^e$ and a corresponding actual crime rate of $p_A$ are self-fulfilling and thus self-reproducing. Hence, point A is a unique stable equilibrium. For later analysis, it is helpful to note that we have a stable equilibrium if the $f$-curve crosses the 45-degree line from above (and an unstable equilibrium if the $f$-curve crosses the 45-degree line from below).

Exogenous changes in costs and benefits: Assume now a society that is in equilibrium A and that the government allocates more resources to fight crime. Within the model, we can interpret that the probability of being caught and convicted increases ($s$ increases to $s^h$). An increase in $s$ shifts the $f$-curve downward. For a given perceived crime rate, fewer individuals follow the criminal strategy since the expected cost of crime increases. The crime rate goes down. We can think of this as a movement from point A to B in Figure 3. This effect is in line with the traditional model of rational crime. A higher expected cost of a crime induces more individuals to choose the legal alternative. But when people also care about social norms, we get an additional effect [18]. When the crime rate falls, the social crime norm is strengthened. Individuals with high moral cost switch to the legal alternative, which in turn leads to a further reduction in the crime rate. This in turn leads to even more individuals switching to the legal alternative and so on. In sum, this bandwagon effect leads society to the new equilibrium in point C in Figure 3. Stated differently: an increase in the probability of being caught does not cause the crime rate to fall from a high-crime equilibrium (point A) to another slightly lower high-crime equilibrium (point B) but to a completely new type of low-crime equilibrium (point C). This effect is produced by a gradual strengthening of the social norm.

The same effect will occur with a decrease in $a$ (the payoff from crime) or an increase in $c$ (the punishment of being caught and convicted). Note also that the same mechanisms can work the other way around, moving the society from the low-crime equilibrium (point C) to the high-crime equilibrium (point A) in Figure 3. The central point to notice is that when a social norm and associated moral costs are included in a standard model of rational crime, small changes in external costs and benefits can lead to large changes in the crime rate at the societal level (much larger than the standard model predicts).

Multiple equilibria: There is no reason why the social dynamics should produce only one equilibrium level of crime. Figure 4 illustrates a case with three possible crime equilibria A, B, and C but where only the first and last are stable. The middle equilibrium B is unstable since nearby perceived crime rates generate movements away from it. If the perceived level of crime happens to be $p_B^e$, it will be confirmed (we have an equilibrium) but a slight movement away from $p_B^e$ will produce adjustments toward one of the stable equilibria. If the perceived level of crime is to the right of $p_B^e$, the actual crime rate converges step by step to $p_A$. If the perceived level of crime is to the left of $p_B^e$, the actual crime rate converges step by step to $p_C$.

Figure 4.

Multiple equilibria crime rates.

What does Figure 4 tells us? An important observation is that two societies that are identical in terms of external costs and benefits related to crime, which also have the same social norms and an equal distribution of moral costs, can end up with very different levels of crime (which echoes the result established in the somewhat simpler context of Section 4.1). The process is path dependent in the sense that initial expectations determine where society ends up. A society that starts somewhere to the right of B ends up in a high crime equilibrium (point A), while a society that starts somewhere to the left of B ends up in a low crime equilibrium (point C).

In the case of multiple equilibria, it is also possible to move from a high crime equilibrium to a low crime equilibrium by temporary changes in the underlying parameters. Assume a society that is in the high-crime equilibrium A and where the probability of being caught and convicted ($s$) increases (in the same way as analyzed above). An increase in $s$ shifts the $f$-curve downward. A higher expected cost of crime reduces the proportion who follow the criminal strategy. The crime rate in society goes down, which in turn strengthens the social norm. This produces a bandwagon that drives the society all the way down to the new low-crime equilibrium D in Figure 4. A crime-reducing shock is reinforced by the strengthening of the social norm. An important result of this is that even if $s$ is reversed back to its initial level, the crime rate will not move back to where it was (point A) but rather end up in the low-crime equilibrium (point C). A temporary increase in $s$ creates long-lasting changes in the level of crime in society.

The status of the deterrence hypothesis: The deterrence hypothesis suggests that increasing the severity or certainty of punishment for a crime will decrease the likelihood of that crime being committed. The hypothesis follows directly from the economic model of crime. When individuals weigh the potential costs and benefits of committing a crime, and the cost increases (in the form of harsher or more certain punishment), the crime becomes less appealing. On aggregate, this will decrease the number of crimes.

As noted in the introduction, the empirical status of the deterrence hypothesis is mixed. Some studies find clear support for the hypothesis, while others are not able to find any relationship between harsher punishment and the crime rate. There are also some studies that find that harsher punishment leads to higher crime rates (see, e.g., van der Weele [3] and Opp [26] for further discussions and references to the empirical literature).

The framework developed in this chapter may help in explaining why the empirical status of the deterrence hypothesis is so mixed. We have seen that an increase in the probability of detection (or the severity of punishment) does not cause the crime rate to fall from a high-crime equilibrium to another slightly lower high-crime equilibrium (as the standard model predicts) but to a completely new type of low-crime equilibrium. A crime-reducing shock is reinforced by the strengthening of the social crime norm. And as just demonstrated, if the governments’ deterrence is reversed back to its initial level, the crime rate will not move back to where it was but rather end up in a low-crime equilibrium. A temporary increase in deterrence creates a long-lasting change in the level of crime not captured by the standard model.

It is important to note, however, that these results are sensitive to the assumptions made about the underlying social dynamics. A slight modification of how the moral cost is distributed in the population may produce a completely different result. This observation supports van der Weele ([3], p. 406), who notes that: “It is somewhat puzzling that decades of intensive empirical research on the deterrence hypothesis have not yielded consensus. One reason may be that we still have not looked hard enough, or nature has not given us clear enough evidence. … More fundamentally, however, there may be something wrong with the question people are trying to answer. … there is no such thing as the effect of deterrent policies. Rather, they seem to work in some times and places, while not in others.”

5. Conclusion

Most people want to follow the rules and to be kind. This assertion is supported by recent findings in experimental economics demonstrating that people generally have a strong self-concept of being honest. This still does not prevent people from engaging in crime and other forms of dishonesty. We seem to be equipped with a cognitive flexibility that gives us quite a lot of maneuvering room when it comes to bending or breaking rules. Most people are not completely honest or dishonest. We are often tempted and willing to engage in cheating, stealing, and lying but only up to a point where the amount of dishonesty would not damage our self-concept of being an honest person [4].

Available evidence also demonstrates that where the threshold goes for stretching ethical standards without damaging our self-concept is not fixed. Our cognitive flexibility is indeed flexible. Social norms seem to play an important role here. Social norms shape people’s self-concept of what it means to be an honest person since they serve as benchmarks against which we compare our behavior. But what others do also affects our perception of what is acceptable. A social norm loses its grip on the mind if many people do not follow the norm.

By incorporating social norms and peer influences into the standard economic model of rational crime, we are able to analyze in more depth the rather complex relationship between government deterrence, individual behavior, and aggregate outcome. We have for instance demonstrated that societies that are identical in terms of social norms and external costs related to crime may end up with very different levels of crime due to minor differences in initial beliefs and the process of aggregation.

Another key finding from the framework presented in this chapter is that when expected external costs and benefits of crime influence behavior directly as well as indirectly through their impacts from social norms, the outcome may differ significantly from the standard economic model. When the moral cost of breaking a norm of honesty increases in the number of norm followers, there is a ‘double dividend’ of deterrence [3]. A very ‘mild law’ may be sufficient to coordinate beliefs and behavior in society into a low crime equilibrium.

The standard economic model of rational crime has been heavily criticized by criminologist and other social scientists. Guided by recent findings in experimental economics, the main aim of this chapter has been to enrich the economic model of crime by embedding it in a social context. The hope is that this can contribute to an understanding that the distance between psychological approaches to criminal behavior and the idea of rational choice is not as great as many make it out to be.

Acknowledgments

I thank Bjørn Hansen and Sevgi Güney for helpful comments and suggestions.

Conflict of interest

The author declares no conflict of interest.