The widespread perception that US middle-class incomes have stagnated for decades contributed to the force of populism in the 2016 presidential election. If the recent Democratic party presidential debates are anything to go by, the same thing will happen again next year. Yet a clear-eyed look at the statistics shows that it is a mistake to judge that American capitalism is broken.
Leading candidate Bernie Sanders cites declines of $5,000 in inflation-adjusted median household income since 1999-2000, based on the 2014 census data available during the 2016 election.
However, the picture is far better when a different price index is used, the trend towards smaller households is taken into account, and we look at long-term trends rather than the unusually high starting point of 1999 and use more recent 2017 data.
On this basis, I calculate that real median household incomes rose 50 per cent over the past 50 years, rather than the 21 per cent suggested by the census data. They have even increased about 8 per cent between 1999 and 2017 (rather than only 2 per cent suggested by the census data) despite the worst recession since the 1930s.
Now let’s look at the individual components. Household incomes have to be adjusted for price increases to make meaningful comparisons over time; otherwise the data would seriously exaggerate gains during the inflationary 1970s, for example. The Census Bureau uses a research version of the consumer price index. Both the Federal Reserve and the Congressional Budget Office prefer the personal consumption expenditure index.
The PCE gives much more scope for substitution among products in response to the changes in relative prices. Over the past 50 years, the research CPI has risen 12 percentage points more than the PCE, so using the former lops off that much-perceived progress in real incomes over this period.
Meanwhile, average household size declined from 3.3 persons in 1967 to 2.6 by 2000 and 2.5 by 2017. That larger average household size in the late 1960s means there were more potential workers per household. However those extra household members are unlikely to earn as much as the main breadwinner.
The census makes no adjustment for household size, but the CBO adjusts for household size by assuming that a household of four would be expected to earn only twice as much as a single-person household. Using this size adjustment, household incomes in the 1960s need to be shrunk by about 12 per cent to be meaningfully compared to household incomes by 2017.
Cherry-picking years for comparisons is the third data distortion that needs to be avoided. At the end of the dotcom boom, in 1999-2000 incomes were well above their trendline; in 2014, still in the wake of the financial crisis, they were below their trendline. A statistical trendline giving each year equal weight rather than just comparing endpoints shows real growth in median household income was 1 per cent annually from 1967-70 to 2000 (compared with 0.6 per cent in the census data), and 0.5 per cent annually from 2000 to 2017 (versus zero in the census data).
Finally, with the same adjustments the data show that incomes for the “lower-middle class” (those from the 20th to the 40th percentiles) are not lagging as far behind the median as one might fear, given concerns about rising inequality. This group’s trendline income rose by 40 per cent over the past half-century, less than the 50 per cent increase for the median but not radically so.
Moreover, income growth in the post-2000 trendline can be expected to rebound somewhat as the 2008 crisis recedes further into the past and its weight in the statistical trend diminishes. The middle- and lower-middle classes are doing better than some politicians think.
The writer, who heads Economics International Inc, is a senior fellow emeritus at the Peterson Institute