Inflation inequality meets wage compression
Did unequal inflation undo the post-pandemic wage compression?
A historically tight labor market after the pandemic led to sharp real wage gains at the bottom. Based on data from the EPI Data Library (which in turn uses Current Population Survey data), the 10th percentile real wages were up by around 15% between 2019 and 2024 — exceeding the cumulative wage growth for the 10th percentile over the prior 40 years. In contrast, the median real wage rose by around 6%, and the 90th percentile by 7%. So in a striking reversal of the past trend, real wages grew much more at the bottom than the top, as I documented in my paper with David Autor and Annie McGrew, “The Unexpected Compression: Competition at Work in the Low Wage Labor Market”.
But when I say “real” wage, I’m using a single, national price index to deflate nominal wages. We may worry that lower-income households face higher inflation. Most importantly, food, rent, and energy weigh more in their basket — so the national CPI under-deflates their nominal gains.1
This post addresses this concern using the distributional chained CPI that accounts for between-category differences in inflation by household income deciles, as constructed by Xavier Jaravel.
So, how much of the measured wage compression survives once we deflate each wage decile by its own between-category chained CPI?
Let’s dig in.
First things first. Jaravel provides chained CPI inflation rates by household income decile. As Jaravel’s distributional DCPI shows, post-pandemic inflation was regressive: between 2019 and 2024, cumulative inflation of 23.8% at the bottom decile exceeded the 20.4% at the top decile. The key drivers? Food, shelter, and energy — all more heavily weighted for low-income families. So the same nominal raise buys less at the bottom than at the top.
Figure 1: Cumulative inflation since December 2019, by household income decile.
But household incomes are not the same as wages. The deciles based on the two concepts only imperfectly map onto each other. Using the 2019 Current Population Survey Annual Social and Economic Supplement (ASEC), I estimate the joint distribution: while earners in most top-income families have high wages, low-wage earners are scattered across low- and middle-income families. As a result, a naive application of low-income family inflation to low-wage earners would overstate the inflation-inequality correction on compression. So I estimate a rank-to-rank mapping between wage and household-income percentile from the 2019 CPS ASEC, and use that to deflate wages by percentile. Using this mapping to re-weight, we can construct wage decile specific inflation rates. While it’s still the case that low wage deciles experienced greater inflation than high wage deciles, the differences are less pronounced than for household income.
Figure 2: Cumulative inflation since December 2019, by wage decile.
To see how this affects compression, I use nominal wages by percentile from EPI’s Data Library, which is based on the Current Population Survey. Then, I deflate these using the wage-decile-specific inflation measures I constructed. So, what’s the punchline?
Figure 3: Real wage growth by wage decile between 2019 and 2024, based on decile-specific inflation rates.
For the 90–10 ratio of wages, the 2019–2024 compression using national prices is 8.4 percentage points. Correcting for wage-decile-based inflation differences brings the compression down to 7.1 pp — a 15% reduction. Not nothing, but fairly modest. For what it’s worth, naively applying household-income-decile inflation rates to wage deciles would have implied a 5.5 pp compression instead, or roughly twice the correction.
Figure 4: Real wage growth by wage decile between 2019 and 2025, based on decile-specific inflation rates.
We can also look at wage growth through 2025. The baseline compression (using a common chained CPI) fell somewhat from 8.4 to 6.2 percentage points, as wage growth was weaker at the bottom in 2025; adding inflation inequality further reduces the compression to 4.8 pp … a proportionally larger correction (22%) than over 2019–2024. This highlights the bigger risk: a somewhat weakened labor market (and the ongoing supply-shock inflation) poses a serious risk to broad-based prosperity and the post-pandemic wage compression.
Wasner (2026) sharpens this concern with NielsenIQ scanner data showing household-specific food inflation outran the food CPI for most bottom-decile households 2019–23 — though limited to food-at-home (~7.5% of spending), where within-category dispersion may be unusually high. Unfortunately, data limitations prevents the ability to extend this approach to create broader inflation measures with within-category variation.
I am interested to hear of the pushback you get and how much of it comes with specifics on your model versus those who employ their anecdotal experience.
The methodological contribution here is worth naming clearly: using a single national price index to measure real wage gains at the bottom isn't just imprecise — it's systematically biased in the wrong direction. The households that experienced the strongest nominal wage compression also experienced the highest inflation on their actual consumption basket. So the instrument we use to validate the good news story was measuring something slightly different than the good news story itself. What strikes me about this from a systems perspective is that the bias isn't random — it consistently runs in the direction of understating how much of the nominal gain gets eroded at the bottom. That's not a coincidence. Food, shelter, and energy are more heavily weighted in low-income baskets because those households have less discretionary room to substitute away from essentials when prices rise. The inflation hits harder precisely because the system has less give at that level. The compression story survives your correction — which matters — but the correction itself reveals something structural: the standard measurement framework doesn't see the mechanism that makes low-wage households more exposed to supply-shock inflation in the first place.