Our first FRED graph traces the evolution of the median weekly wage in the United States. See the spike in 2020? Does this huge increase mean everyone got a huge raise? No, it does not. As we’ll try to show in this post, the so-called composition effect is misleading us here.

We’ve discussed the composition effect before, which is basically that group averages can mask true individual experiences.

Our second graph shows employment for various education levels, excluding non-salaried workers: We see that more education made it less likely to lose a job in 2020.

The third graph shows the number of employed people by education, including non-salaried workers: All categories decreased, but the decreases were disproportionally larger for the less educated.

The second and third graphs show us where that composition effect kicks in: After losing more low-wage jobs than high-wage jobs, the median wage had to go up. Remember: With a median, in this case median wage, half the jobs are above and half are below. If you remove more jobs from the bottom half than the top half, the median wage will rise.

How these graphs were created: For all graphs, start at the Weekly and Hourly Earnings from the Current Population Survey release tables and navigate to the table of interest. Check the series you want to display and click “Add to Graph.”

Suggested by Christian Zimmermann.