<< McMiranda | Main | Officer's killer is put to death >>


A funny way of saying it

| 1 Comment
A mostly off-topic note on the oddities of language about people and statistics.
Today I received this release from the Gallup Poll:

The Republican Party in 2011 remains demographically and ideologically similar to the way it looked in 2008, although slightly more Americans call themselves Republicans today. However, Republicans are now slightly less likely than they were in 2008 to be male and to be highly religious.
Really?  All Republicans are?  I'm a Republican, and I am no less likely to be male now than I was in 2008.  My wife is also a Republican, and her probability of being female has not increased from its prior value of 100%.

What Gallup really means by "likely," of course, is that if you select a person at random from today's population of Republicans, you are a shade more likely to select a female than if you had made a random selection from the 2008 population of Republicans.  Sampling is their business, so that is how they think and how they express themselves.  But it sure strikes me as a strange way to say it.

The important point is to remember that means, correlations, and probabilities are just mathematical abstractions.  The "average family with children" has 1.86 children, but no real families have 1.86 children.  Numbers describe the abstract average person in a group, however that group is defined.  But they don't describe everyone in the group, and they may not describe anyone at all.

In public policy debates over crime, we often discuss statistics about crime, criminals, and victims.  It is too easy in such debates to forget about the real people and real crimes behind the numbers.  We need to guard against that tendency.

1 Comment

"In public policy debates over crime, we often discuss statistics about crime, criminals, and victims. It is too easy in such debates to forget about the real people and real crimes behind the numbers."

And that tendency is magnified by sob stories about individual criminals. But the macro effect of policies that arguably impose unfair hardships on certain criminals prevent untolled suffering to nameless hypothetical victims. It's easy to look at one criminal and think that it may be acceptable to give him a break--but when that happens again and again (as will always be the case) there are real costs to real people. So for example, perhaps a state's policy of providing LWOP for juveniles has some results which are wrong. Before that issue gets corrected, the flipside has to be analyzed as well, whether the LWOP policy would prevent some victimization. Then it becomes a balancing issue. And where I come out is that there needs to be a thumb on the scale against those who choose to commit serious crimes.

Leave a comment

Monthly Archives