Sunday, April 11, 2010

How to Lie Without Statistics

Today's Parade Magazine (with a circulation of 32 million) includes its "Annual Salary Survey." What this means is that the magazine presents about a hundred photos of various people with their names, occupations, and annual earnings.

At first, you might think this is a good way to give readers a sense of the distribution of income in society. And it would be, if the sample were at all representative. But it isn't. Parade decides to oversample celebrities. That is understandable--after all, readers are more interested in hearing about famous actors and sport stars than about a plumber in Dubuque. But one result of this choice is that the sample is far from representative, making the whole affair misleading as a piece of journalism.

By my count, about 14 percent of the people in Parade's sample earn more than $1 million a year. In the real world, the actual percentage is about 0.2 percent. So, in a truly representative sample of a hundred people, you would most likely have zero, or perhaps one, person with a million dollar income. Finding two would be highly unlikely. 14 would be nearly impossible.

Does this matter? I think it might. There is a common perception in some circles that we can solve all our fiscal problems if only we were willing to tax the rich some more. Yet, in reality, there are not enough rich for this to work. By presenting such a skewed cross-section of incomes, Parade inadvertently feeds an all-too-common misperception.