Racial Profiling and a Misunderstanding of Probability

Racial Profiling and A Misunderstanding of Probability

Posted Sep 13, 2009

Racial profiling can be broadly defined as a practice in law enforcement based on the belief that a person's category membership (e.g., ethnicity, national origin) functions as an indicator of criminal propensity. In my previous posts on the topic, "why moral education has not reduced racial profiling (1) & (2)," I examined some cognitive root of the biased practice. The current discussion explores another type of cognitive distortions that supports racial profiling-a misunderstanding and misapplication of probability/statistics.

One of the most typical justifications that officers use for racial profiling involves the statement that "it is true that certain ‘groups' are more like to commit offenses or infractions than the others." I think that this statement suggests the confusion about basic probability and statistics concerning criminal behavior. This confusion consists of two cognitive errors:
1) unawareness of the distinction between the relative frequency and the absolute frequency regarding a particular crime-prone attribute, and
2) ignorance about the multiplication rule of probability for criminal behavior. Namely, failure to recognize that criminal behavior is determined by the interaction of multiple variables or factors.

Regarding the first error, research in criminology and psychology has identified numerous variables that may increase an individual tendency to commit crime. These variables range from social disorganization, dysfunctional environments, crime learning conditions, developmental problems, lack of social control, lack of education, stigmatization, cognitive deficiencies, impulsivity, childhood trauma, as well as some associated demographic variables such as gender, social class and race. The distinction between the relative frequency and the absolute frequency can be explained with the following examples:

Research has shown that people with only high school education are more likely to commit crime than people with graduate degrees. The same can be said about high impulsive persons, who are more inclined to commit crime than low impulsive persons. Let's assume that in both cases, the former are 90% and the latter are 10% in criminal propensity. The 90% statistics looks impressive but it is quite misleading. The number only represents the relative frequency of the tendency in comparison. It does not suggest that 90 % of the people who have high school diploma or are impulsive will commit crime, because the absolute frequency (the number of offenders with the attribute out of the total population with the same attribute in the United States) is extremely low, it is certainly below 5% of the total population. Another example may make the distinction more clear. Statistically, more than 80% serial killers are white males, but it does not suggest that being a white male makes the person 80% more likely to become a serial killer, because the absolute frequency is so low.

Additionally, the official crime statistics may be problematic. According to FBI's Uniform crime report (2007) on arrest by race, of all adults and juveniles arrested nationwide in 2007, 69.7 percent were whites, 28.2 percent were blacks, 1.3% were American Indians or Alaska Natives, and 0.8 were Asians or Pacific Islanders. Although the sum is 100%, there is a problem: The arrest data on Hispanic Americans, which are 15% of the U.S. population, are conspicuously missing. Someone may argue that Hispanic populations are not really a different race and their arrest data were distributed among the four groups above (are they really four "different" races)? If it is the case, it only shows the lack of precision in the crime statistics.

The second error involves the ignorance about the multiplication rule of probability for criminal behavior. To understand the rule, we can look at how it is applied to the offender profiling for serial killers. On the basis of analyzing the characteristics of known serial killers in America (e.g., Ted Bundy, Jeffrey Dahmer, Aileen Wuornos, Robert Lee Yates, and many others), researchers have developed a profile for the offenders. The offender characteristics include: White males (80%), above average intelligence, committing the first murder at the age between 24 and 40, 50% married with a stable family life and stable jobs, and a sign of cruelty to animals in childhood, among other factors.

This offender profiling has certain validity in helping law enforcement personnel to detect and investigate serial offenders, because it is based on the understanding of the multiplication rule of probability for criminal behavior. That is, the probability of a suspect as a serial killer is a product of race, gender, intelligence, marital status, childhood deviant behavior, and other factors. In other words, because the variables are independent from one another, an individual attribute alone on the list (e.g., race, intelligence) has no predictive value about the criminal likelihood of the target person. In fact, as shown in criminological research, most criminal behaviors are the product of the interaction of multiple causes and variables, rather than determined by one or two variables.

The application of the multiplication rule to predicting and detecting serial killers has limitations, because not all identified attributes on the profiling list can be generalized to all serial killers. For example, it took about 20 years to catch Gary Ridgway - aka the "Green River Killer," because his name came out early but police initially thought the killer was unmarried while he was married. In addition, important psychological variables about serials killers are not included in the profile.

In short, racial profiling is biased, because:

1) It is based on misunderstanding the scientific theory of probability.

2) It contradicts the mission of the justice system, which has two purposes: First, to punish the guilty, and second, to protect the innocent. Racial profiling has violated the principle of protecting innocent individuals by turning thousands of innocent people across the country into the victims.

3) It creates a self fulfilling prophecy, because when officers only focus on certain categories of individuals (yes, they will catch some criminals among them), they will also bypass many criminals about whom they do not do profiling.

Please read my related post "How the use of 'white vs. nonwhite' contaminates criminology."