Earlier this week, I posted on the classic question of whether certainty in punishment is more important than severity. By some coincidence, when I then received the latest issue of Criminology, I found a fascinating new article that explores another dimension of the certainty problem.
The terminology is a little awkward here, but the authors are interested in what they call the “ambiguity” surrounding the risk of punishment. The basic idea is this: two different people contemplating a crime might both reach the same estimate of the risk of getting caught, but one might have much more confidence in his estimate than the other. The less confident criminal perceives the same level of risk, but a higher level of ambiguity. What effect, if any, does such ambiguity have on deterrence? That is the question that Thomas A. Loughran and his coauthors consider in ”On Ambiguity in Perceptions of Risk: Implications for Criminal Decision Making and Deterrence,” 49 Criminology 1029 (2011).
Some criminologists have theorized that ambiguity might have an independent deterrent effect apart from the actual risk of getting caught. (1031) If so, then one potential implication is that police can maximize their effectiveness by being unpredictable.
For instance, imagine that police resources are such that arrests can only be obtained in five percent of robbery cases. It doesn’t seem that many robbers would be deterred by that level of risk. But now let’s say that the police department concentrates its robbery response resources in particular neighborhoods, shifting each day on a random basis. The overall risk of arrest is still only five percent, but an unlucky robber in the wrong neighborhood on the wrong day now faces a risk of, say, 80 percent. If robbers are indeed “ambiguity averse,” the uncertainty in risk in the random system should result in more deterrence (less crime) than in the nonrandom system.
Among other things, this theory suggests that compstat-based policing may be misguided: rather than allocating resources based on analysis of crime trends, the police would do better just to flip a coin.
Why would prospective criminals be averse to ambiguity? There is much empirical research to support the existence of ambiguity aversion in a variety of different contexts. Loughran et al. note possible psychological mechanisms:
Frisch and Baron (1988) theorized about why ambiguity might affect decisions involving uncertainty and risk. One theme from their research is that when a probability judgment involves ambiguous information, an opponent or adversary may possess superior information to the decision maker, giving the adversary an inherent advantage. A would-be offender’s adversary is law enforcement. Thus, ambiguity might make an individual reticent to offend for fear that law enforcement has an inherent advantage. Second, ambiguity might make an offender hesitant to act because by waiting the offender may obtain more and better information about the apprehension risks, and then he or she may make a better informed decision at a later time. (1036)
Whatever the underlying psychological mechanisms, Loughran and his coauthors sought in the new article to test whether real offenders actually do behave in ambiguity-averse ways.
Their subjects were 1,354 adolescent offenders who were interviewed at six-month intervals over a three-year period. The subjects were asked about the perceived risks of arrest for seven different crimes. They were also asked about whether they had committed any offenses in the past six months.
Consistent with classic deterrence theory, the researchers found a negative relationship between perceived risk and commission of crimes: the more certain punishment was perceived to be, the less likely the subjects were to offend.
As to ambiguity, the researchers found a much more complicated pattern. First, there seemed to be a distinction between what they called “no one around” crimes (breaking in, stealing, theft, vandalism) and “face-to-face” crimes (fighting, stabbing, robbery with a gun). With the former, when perceived risk was low, there was less offending as ambiguity grew higher. This result is consistent with the ambiguity-aversion theory. On the other hand, when risks were perceived to be high, ambiguity had the opposite effect: it was associated with more offending. (1054)
And as to face-to-face crimes, the effects were also reversed. When perceived risks were low, increased ambiguity was again associated with more offending. This result seems to support the common intuition that face-to-face crimes are more emotion-driven and less deliberate than no-one-around crimes, with quite different psychological dynamics driving the offender’s weighing of costs and benefits.
Although ambiguity-aversion was not consistently observed, the research may supply some support for the strategic use of ambiguity to fight some types of crime in some circumstances. For instance, robbery would probably not be a good candidate, but burglary (a no-one-around crime with a low clearance rate and hence low risk) might be.
Loughran and his coauthors focus just on ambiguity as to the likelihood of getting caught. I wonder, though, about whether ambiguity in the severity of punishment might display similar effects. Imagine a system of randomized punishment: rather than a judge selecting a sentence within a statutory range, a wheel would be spun to get a number with the range. Would this be more efficacious in reducing crime than existing approaches?
Bernard Harcourt once proposed randomized criminal justice in a provocative paper. I responded critically in print. (Harcourt’s argument and my response, along with those of other critics, are available here.) I may be warming up to the randomization idea, though, in part because of potential ambiguity-aversion benefits.