Bullshit Alert: Real Median Income Increase Makes No Sense



  • US Census says real median income increased by 3.2% in 2016.
  • There is data set after data set which casts doubt this conclusion is way off.
  • Should this conclusion been tested before being released?

The U.S. Census says that “real median household income increased by 3.2 percent between 2015 and 2016“. That is significantly higher than real GDP which grew at 1.5%. This increase does not make sense.

Historically, GDP grows at a faster rate than income – except for periods after the Great Recession. And when one visits FRED (graph above red line) – it shows year-over-year gains of 4.5% using the same U.S Census data. Interesting.

To add to concerns about the accuracy of the median household data, inflation adjusted wage / salary growth is growing at half the rate of median household income (graph below).

And if one looks at the year-over-year growth rate of real disposable personal income per household – the AVERAGE growth was 0.4%.

Could it be that there are more people in a household working? There was an increase of 1.5% employment year-over-year- and the prime working age population did not grow.

But on the other hand – there were 1.0% more households in 2016. This means the gain in employment per household would be 0.5% (1.5% employment growth less 1.0% household growth).

It does not seem employment growth was much of a contributor to household income.

I could produce chart after chart which demonstrates how significantly this data seems out of whack with income data from other data sets from Census, BLS or BEA. This real median income data was derived from a sample of around 50,000 households (well less than 0.5% of all USA households). The measurement error uncertainty for such a small sample must be at least +/-3% and possibly more. All of the income related numbers above are either approximately the same as the measurement uncertainty or significantly smaller. Thus no significant changes in income can be claimed from the survey data alone.

Whenever one extrapolates analysis, one tests data against other data sets. This data fails this test and is likely that nothing can be defined about the growth in real median income reported for the last year.