Statistics, Generalizations, and Stereotypes

Statistics can create generalizations, generalizations can create stereotypes. All three — statistics, generalizations, stereotypes — can be good or bad, depending on how they are used.

My educational background consists of mathematics, psychology, and computer science. I learned early on that applying statistics to an individual person can sometimes be a bad idea.

It’s well known that making a generalization is usually beneficial. For example, early humans probably learned a generalization that “some animals are dangerous” after a small statistical sampling of interactions. Even though some animals likely won’t be dangerous, making a generalization that all snakes or apes or saber tooth tigers are dangerous must have been advantageous from an evolutionary point of view.

Generalization reduces cognitive load, and allows decisions to be made more quickly. If an early human was walking in the jungle and an ape appeared on the trail up ahead, it’d make sense for the human to go to a different trail, rather than mentally calculate a probability of danger.

Generalizations are typically based on some form of explicit or implied statistics. And so generalizations don’t necessarily apply to a particular individual. This can be a problem when statistical information is applied to people on the basis of group membership, such as (Australian, Brazilian, Canadian), or (Anglican, Baptist, Catholic), or (Anaheim, Baltimore, Cambridge), or (Alpha, Beta, Chi), or anything else.

Suppose you see these actual statistics, which were pulled from various State and Federal government agency Web sites, where X represents a particular group membership:

Percent of X children born illegitimately: 81.1%
Percent of X males arrested and convicted: 31.5%
Percent of X residents on some form welfare: 59.1%
Percent of X males drop out of high school: 46.6%
Percent of all murders committed by X males: 52.5%

The exact numbers vary depending on year, reporting agency, specific geographical region, and so on. And the numbers are implicitly reinforced by media reports. But the exact values aren’t the relevant issue with regards to generalization and stereotyping. The significance of the data is when values for one group are much different than the corresponding values for other reference groups. (Note: I’m a bit skeptical that statistics like these can be collected by government agencies with complete accuracy. For one thing, by the time the data is collected it’s already out of date.)

---

Here is a result of a Google image search for “teens arrested”. Media results and depictions are a type of implicit statistics that can create generalizations.

---

There is no clear distinction between a generalization and a stereotype, but most research papers seem to consider stereotypes as strong or extreme generalizations.

An interesting research paper by P. Bordalo out of Harvard suggests that people construct stereotypes that are, “selective, however, in that they are localized around group features that are the most distinctive, that provide the greatest differentiation between groups, and that show the least within-group variation.” This makes sense because the point of a stereotype is to be able to quickly make a decision of some sort.

None of this is rocket science. But statistics apply to a group, not directly to an individual person or item.

Social media is full of people telling other people how and what to think. My advice to myself is what my parents told me: Think for yourself. Treat others as you wish to be treated. Doing the right thing is always the right thing to do.