One source of data that people often use is median household income. It’s a good idea to use the median rather than the mean -– the mean can be very misleading. For example, the mean income of Harvard graduates who studied economics is going to be very high in the year that Jeremy Lin graduated. John Elway, another econ grad, pulled up the mean dramatically for Stanford grads that year.

But there is a problem with median household income and those who use it relentlessly to grind their policy axes never mention it. The problem is that when household structure is unstable, comparing medians over time is a very poor way of assessing the progress of the typical person.

As I have written many times, rising divorce rates in the 1970?s for example, meant that the number of households in the US grew 26.7%. Population grew only 11.5%. There was an increase in the number of households as one household became two. If both people were working, that alone would likely decrease median household income. If only one of the spouses was working, it was usually the man. The former wife found herself in the labor force unexpectedly. Her income is likely to be below the median. Both of these effects create new households with incomes below the median, dragging down the median over time.

via Inequality and Stagnation.