The Five Dumbest Things on Wall Street This Week

Well, Kozlowski couldn't have done it. The Census would have recorded his salary at a mere million bucks.

"The fact that we're not recording the full dollar value is going to understate the share of income controlled by households at the highest levels," says Welniak.

But, says Welniak, there's a good reason for capping monetary entries at six digits: It limits the potential for error. One extra digit at the high end, and you're talking about, say, a $9 million paycheck instead of a $900,000 payout. Errors at the high end of the income scale have a much larger impact than errors at the bottom. The increased accuracy introduced by more possible digits, says Welniak, would be more than offset by the decreased accuracy from newly enabled errors.

Welniak has even investigated the exact effect of rounding all multimillion-dollar income sources down to a megabuck. According to his analysis of numbers from 1999 -- a year for which 26 respondents reported employment compensation of at least $1 million in at least one category -- data-entry limitations effectively understated income inequality by 1%, using a standard measure of income distribution known as the Gini Index.

But, given that the error appears to be constant year after year, says Welniak, "Measuring changes in income inequality from one year to the next is not going to be affected." In other words, ignore the absolute number and look at the trend.

Mindful of that, we point out that over the past decade, the Census Bureau's Gini Index has been creeping upward -- implying increased income inequality. Starting at 45.4 in 1993, it peaked at 46.6 in 2001 but retreated to 46.2 last year. (For purposes of comparison, the United Nations Development Program -- which puts the U.S. at 40.8 -- says Japan is a 24.9 and Brazil is a 60.7.)