Income - Averages vs. Medians

August 21, 2007
Posted by Jay Livingston

David Cay Johnston reports on taxes and income for the New York Times, and he does a great job. Not too many other publications have reporters on this beat. But why, in today’s article, is he using averages?
While incomes have been on the rise since 2002, the average income in 2005 was $55,238, still nearly 1 percent less than the $55,714 in 2000, after adjusting for inflation, analysis of new tax statistics show[s].
Here’s the graph that accompanies the article.

Much of the fluctuation in income is accounted for by those at the top. When average income rises, it’s probably because those at the top are getting a lot richer. But the decline in the average income in the early years of this decade may have been driven by the bursting of the Internet bubble, a bubble which affected the top 10% or less.

When economists want a single figure that shows income trends, they use not the average but the median. In this case, the news is about the same. Here are some data from the Regional Economic Information System (REIS) of the U.S. Department of Commerce, Bureau of Economic Analysis. It uses 3-year averages.

In constant dollars, median income is about $1000 (2%) lower in 2005 than it had been in 2000.

There was one bit of good news in the Times article: “The I.R.S. data showed that the number of Americans making less than $25,000 a year shrank, down by 3.2 million, or 5.5 percent.” Unfortunately, we don't know how may of those 3.2 million people increased their income and how many just dropped out of the labor market entirely


Anonymous said...

This is great! I think I'm going to use it in my Intro Soc class to demonstrate why we need to critically analyze what we read.

Brad Wright said...

You're right about median being perhaps more informative than the mean here... though it's a big hard to tell with the different groupings of years.

Jay Livingston said...

Brad, You're right that the figures are not exactly comparable. I was surprised how difficult it was to find 2000-2005 median income data on the Internet. (FWIW, I also think the 3-year averaging is a good idea.)

But the larger point stands: when you look at income, you should use median, not mean. As the currently popular example has it, if Bill Gates walks into the room, the mean income rises to the point where everybody except one person, Mr. Gates, is below average.

Anonymous said...

Agreed - this drives me crazy.

But so do graphs that don't have a zero on the Y axis.

Jay Livingston said...

Matt, you're absolutely right. I think it's the ancient text "How to Lie With Statistics" that refers to this as a "gee-whiz graph." It exaggerates the differences so that you look at it at say, "Gee whiz."

But I couldn't figure out how to change the scale range on the Y-axis in Excel.