The Other McCain

"One should either write ruthlessly what one believes to be the truth, or else shut up." — Arthur Koestler

The Error of Comparing Group Averages

Posted on | October 7, 2016 | Comments Off on The Error of Comparing Group Averages

One of the basic tricks of Lying With Statistics is to use certain characteristics — age, race, sex, etc. — to define a group, and then report some data about the group as an average. CNN does this:

Working class white men saw their income drop 9% between 1996 and 2014, according to a new report from Sentier Research. This group, who Sentier defines as having only a high school diploma, earned only $36,787, on average, in 2014, down from $40,362 in 1996.
Meanwhile, college educated white men saw their income soar nearly 23% over the same period, from $77,209 to $94,601.
Published by two former Census Bureau officials, the Sentier report shines yet another light on the fortunes of the white working class. This group has become a force in the 2016 presidential election, serving as the backbone of Donald Trump’s support. And the Republican candidate’s campaign has tailored much of his campaign to the working class, with promises that he will bring back the manufacturing jobs that once allowed them to support their families.
The study first looked at the 1996 incomes earned by 10 groups of men in two-year cohorts ranging in age from 25 to 26 to 43 to 44. Sentier then looked at what men earned 18 years later, when the youngest cohort were 43 to 44 and the oldest were 61 to 62.
The results varied greatly by age. The youngest group of working class white men, who were 25 to 26 in 1996, saw their incomes rise by 19%, from $32,677 to $38,803, over the 18-year period. However, their college educated peers enjoyed a 133% explosion in their incomes, from $40,487 to $94,252. . . .
In addition to showing the tough times the working class has faced, the report also shows the big income boost that comes with a college degree, [study co-author Gordon] Green said.
“People say it may not be worth it to go to college. These numbers show that isn’t true,” he said.

You can read the rest. Whatever these statisticians have discovered, it does not explain much about any individual person’s life, because the groups — men with or without a college diploma — are so large that their “average” incomes are essentially meaningless.

If you are college-educated and your income has not “exploded” since 1996, why not? That’s because this group is so large and diverse, including both computer engineers and English majors, that to make generalizations based on their “average” income tells us very little.

This is something you would understand if you have read, for example, Thomas Sowell’s The Vision of the Anointed: Self-Congratulation as a Basis for Social Policy, or Charles Murray and Richard Herrnstein’s The Bell Curve: Intelligence and Class Structure in American Life.

To generalize as broadly as does the CNN story — to describe college-educated white males as if they constituted a uniform class — implies that they necessarily share the same fate on the basis of those particular characteristics. But is it really true that having a college degree (any degree) automatically confers a “big income boost”? And what about that category “white”? This includes 247 million Americans — 77% of the U.S. population — and attempting to make economic generalizations about “white males” (i.e., more than 120 million people) doesn’t tell us much of anything about “the backbone of Donald Trump’s support.”

(Via Memeorandum.)



Comments are closed.