Demographic changes are the most important factors in explaining where productivity growth went. In essence, the United States used some of its growth dividend in the form of more people living alone.

Read the whole paper. The puzzle is the fact that median household income growth over the past 30 years has been much less than growth in *per capita* GDP. Rose points out that much of this is due to more households *per capita*, with poor people particularly likely to be unmarried or divorced–the Marriage and Caste story.

Rose does not mention the other demographic factor that has added households in the lower part of the income distribution–immigration.Suppose that twenty years ago you started with 4 households, with the following incomes:

$24,000

$28,000

$36,000

$48,000

Average income was $34,000 and median income was $32,000. Next, suppose that today every household’s income is 20 percent higher (adjusted for inflation), so that we now have:

$28,800

$33,600

$43,200

$57,600

As a result, average income is now $40,800 and median income is $38,400. Average income and median income have grown at the same 20 percent rate.

But suppose that today we have also added an immigrant household, with an income of $30,000. Including the immigrant household lowers the average income from $40,800 to $38,640. It lowers the median income from $38,400 to $33,600. Including the immigrant household, average income for current households is 13.6 percent higher than it was 20 years ago, but median income is up only 5 percent.

The purpose of this point is not to bash immigration. On the contrary, the immigrant family’s income may have gone up the most in percentage terms. The point is that one has to be careful about interpreting slow growth in median income as a failure of prosperity to be shared across income groups. In the example, prosperity is perfectly widespread, but the demographic change skews the relationship between growth in the median and growth in the mean.

## READER COMMENTS

## Michael Sullivan

## Jun 14 2007 at 10:34am

You also have to be careful with your examples. The fact that the median here got hit by adding the immigrant family much more than the mean did is mostly a property of your toy model and the numbers you chose.

With a larger sample and very small differences between individual families in the sample around the median (IOW, a model that looks more like the real world), there might be a difference between the affect on the mean vs. the median, but it’s unlikely to be anywhere near that large. I’d adjust this toy model’s numbers to work more like real ones would in that regard.

Comments are closed.