A visual explainer of a working paper

Your raise was 3 percent.
Prices rose 9.
The difference came out of your paycheck —
and for millions of workers, it never went back.

In 2021–2023, America ran the fastest inflation in four decades. Most firms answered with the same modest annual raise they had always given. This page explains the machinery — the firm wage-growth norm — that turned a temporary spike in prices into a permanent step down in pay, and the evidence that it is why the country stayed furious long after inflation cooled.

The study underneath this page — Sticky Wage Norms and the Real Wage Cost of Unexpected Inflation, by Erik Hurst, Christina Patterson, Nela Richardson and Ye Liv Wang (Becker Friedman Institute Working Paper 2026-108, August 2026) — reads the payroll records of roughly 16 million American workers a month, captured in ADP administrative data and benchmarked against the Current Population Survey.

It finds a quiet piece of machinery inside almost every firm: a single customary annual raise, applied to nearly everyone in a single firm-specific month. That machinery ran smoothly for decades of low, stable inflation. Then prices jumped — and the machine did not.

  • What it is: the modal raise a firm hands out on autopilot.
  • What it did: left 43% of long-tenured workers poorer in real terms after four years.
  • Why it matters: the missing pay became profit, and the missing purchasing power became a grudge that outlived the inflation itself.
Open the ledger ↓

Item 01 · filed under wage-setting

The machine: one raise for almost everyone

Most firms do not negotiate each worker’s raise. They pick one number — a modal annual nominal increase — and apply it to nearly everybody in a single firm-specific “on-cycle” month.

Inside the ADP records, the pattern is almost mechanical. Among workers who receive exactly one wage adjustment in a year, raises bunch tightly around a firm-specific mode: a majority of all annual wage changes land within half a percentage point of it, and more than 90% within one and a half points. Before the pandemic, the modal firm’s number was 3%.

Fig. 1 Piled up against the mode Annual nominal wage changes relative to the firm’s modal increase · workers receiving a raise
Set this firm’s norm:

The histogram’s shape is stylized to the paper’s Figure 7; the band shares — a majority within ±0.5 pp, >90% within ±1.5 pp — are as reported. On-cycle month: the calendar month in which the largest share of a firm’s workers received a wage change.

Source: Hurst, Patterson, Richardson & Wang (2026), Fig. 7; ADP payroll data.

The result is a workforce sorted less by individual bargaining than by which number their employer happened to adopt. Before the pandemic, 89% of workers were at firms whose norm was 2, 3, or 4%. Then consumer prices rose sharply and unexpectedly from mid-2021 through late 2023 — peaking at 9.1% inflation in June 2022 — and the norms barely moved.

Fig. 2 The norm map barely shifts Share of workers by their firm’s wage-growth norm · employment-weighted
Period:

The 2–4% band is the paper’s reported total (89% → 76%); the split of that band across 2, 3 and 4 is illustrative. Post-inflation: the ~3% modal norm was restored, with more firms at ~4% and fewer at 2%.

Source: Hurst et al. (2026), Fig. 8 and text; band totals as reported.

Fig. 3 The collision: prices soar, the dial clicks once CPI-U inflation, year over year · median firm wage-growth norm · 2019–2025

Rigid on the way down, rigid on the way up. Economists have long documented that wages resist cuts in recessions. This paper documents the mirror image: firm norms resisted rises during an inflation surge. The norms were built for a low-and-stable-inflation regime — the surprise was never priced into the norm or into workers’ original wage bargains.

Source: Hurst et al. (2026), Fig. 9. Inflation: CPI-U, BLS. Norm path: median firm modal base-wage change (stylized to the paper’s stated range).

Item 02 · filed under your paycheck

Watch a paycheck come apart

Nominal pay kept climbing on schedule. Prices climbed faster. What a worker could actually buy — the real wage — stepped down, then resumed rising on a permanently lower track.

Fig. 4 The stub, month by month A representative stayer’s pay deflated by consumer prices · Dec 2020 = $20.00/hr

Nominal path is stylized to the paper’s reported milestones — a median real decline of ~4% at the trough, a return to the Dec-2020 level only in late 2024. Prices: CPI-U level, BLS.

Source: Hurst et al. (2026), Figs. 1–2; CPI-U, BLS.

Fig. 5 Recovered is not made whole Real wage index vs. the trend it left · Dec 2020 = 100
Benchmark trend:

“Back to 2020” sounds like healing; it is not. A given worker’s real wage normally grows — the 2017–2019 trend ran +2.2%/yr, the longer 2000–2019 trend +1.5%/yr. By Dec 2025 the index stood ~8.5 points (~7%) below the extrapolated 2017–19 trend, ~4% below the more forgiving 2000–19 trend. The gap is the red hatch: a loss that compounds, not a pause that ends.

Source: Hurst et al. (2026), Figs. 1 & 19. Index path stylized to reported milestones; trends as reported.

Item 03 · filed under who lost what

The toll: 43 out of every 100

Among workers who stayed at the same firm from December 2020 through December 2024 — the cleanest measure of what the norm did — 43% ended with lower real wages than they started with.

Fig. 6 One hundred stayers, four years Share of Dec-2020 job-stayers ending the period with a real wage decline
Horizon:
Find yourself:

Among those who fell behind over four years, the mean loss was ~9% and the median ~7%. More broadly, 55% of stayers averaged under 1% real wage growth per year across the whole window.

Source: Hurst et al. (2026), Table 1; all-workers and age-50+ figures from text.

Fig. 7 Fewer workers behind — a deeper hole for those who are Share of stayers with a real decline · conditional mean loss · by horizon from Dec 2020

The erosion arrives immediately — two-thirds of stayers lost ground in year one — and then refuses to heal. The equivalent pre-pandemic cohort (Dec 2015 → 2019) closed with 21.4% behind and a −6.9% mean loss: roughly half the incidence.

Source: Hurst et al. (2026), Table 1.

Item 04 · filed under ways out

Two escape hatches — both leak

Workers were not helpless. They could change employers, or win a large off-cycle raise. Both routes worked — for the minority who took them.

Fig. 8 Hatch one: switch employers Median nominal wage growth vs. CPI inflation · job-changers vs. job-stayers

Job-changers’ nominal growth tracked inflation nearly one-for-one — the market repriced them on the way out the door. But moves are infrequent, and switching rose only modestly. Adding changers to the ledger moves the four-year real-decline share only from 43% to 37%; 58% of all workers still ended below the pre-pandemic trend.

Source: Hurst et al. (2026), Figs. 13 & 15. Paths stylized to the reported ~1:1 pass-through for changers.

Fig. 9 Hatch two: the off-cycle raise Distribution of raise sizes, on-cycle vs. off-cycle · share of stayers with >1 adjustment per year

Firms did not reset norms — but they did hand out bigger individualized increases outside the annual review. On-cycle raises cluster between 2–4%; two-thirds of off-cycle increases exceed 4% and one-third exceed 8%. The share of stayers getting more than one base-wage adjustment a year jumped from ~16–18% to ~27% in 2021–22. The within-firm distribution grew a fatter right tail while the middle stayed anchored.

Source: Hurst et al. (2026), Figs. 10–12. Distributions stylized to the reported shares.

Both hatches reached a minority, so the average concealed a split screen: a growing group sprinting ahead on repriced offers and outsize raises, while most people quietly slid backward on the same old number.

Item 05 · filed under who absorbed it · where it went

The accounting

The losses were not evenly shared, the forgone pay did not vanish, and a simple accounting exercise shows how much of the shortfall the norm itself explains.

Fig. 10 Older workers absorbed more Share with a cumulative real wage decline, Dec 2020 → Dec 2024

Roughly 55% of workers aged 50+ saw a real decline. They had less access to both escape hatches: they switch employers less often, gain less when they do, are less likely to receive a large within-firm raise, and already had flatter age–earnings profiles.

Source: Hurst et al. (2026), Fig. 17 and text.

Fig. 11 The lowest-paid were shielded — briefly Real wage growth in 2021 by position in the wage distribution

Higher job-switching rates protected the bottom two deciles in 2021 — their real growth stayed near its pre-period pace while every other decile fell about 2% (~4 points below their pre-period pace). Over the full 2021–2024 window, wage compression looked much like the pre-period.

Source: Hurst et al. (2026), Fig. 16 and text. Bars approximate the reported pattern.

Fig. 12 Where the money went U.S. corporate profits, share of nominal GDP · quarterly, 2016–2026

The wage growth workers did not receive reappeared as margin. The profit share jumped from a steady 11.4% (2016–2019) to 13.1% (2021–2025) — up 1.7 points, within the 1.2–2.4-point range the wage shortfall implies, and the highest sustained level in half a century. The paper treats this as a consistency check, not a full decomposition.

Source: Hurst et al. (2026), Fig. 20; BEA corporate profits via FRED (A445RC1Q027SBEA). Path stylized.

The transfer

Worker → firm

The real wage decline borne by workers who did not act is a transfer: pay that stayed in the employer’s ledger as lower labor costs. It moves money; it does not burn it.

The deadweight loss

Effort → smoke

The searching, switching, renegotiating and conflict that workers spent defending a real wage is a deadweight loss — time and strain burned by the economy, recovered by no one.

Fig. 13 The switch that was never flipped Observed real wage index vs. counterfactuals that index raises to inflation · Dec 2025 gap to trend
Counterfactual:
Benchmark:

The exercise holds everything else — off-cycle raises, job-changer wage growth, group shares — at observed values, and indexes only the written rule. Indexing the modal raise alone lifts the Dec-2025 index ~3.2 points, closing ~40% of the gap to the 2017–19 trend and ~73% of the gap to the 2000–19 trend. Indexing all job-stayer growth adds ~4.8 points — more than half of the first gap, essentially all of the second. These are accounting counterfactuals, not general-equilibrium policy experiments; the paper stresses the qualitative conclusion, not the decimals.

Source: Hurst et al. (2026), Fig. 19. Counterfactual paths stylized to reported index-point effects.

Item 06 · filed under the grudge

Why it still stings

Inflation ended; the anger did not. The paper’s cleanest explanation is not the price level itself — it is the unprotected real wage. Wherever income was indexed to prices, sentiment healed.

Fig. 14 Fury outlived the spike U. Michigan consumer sentiment · share naming inflation the top family financial problem

Sentiment fell to 56.1 in Q3 2022 — below its Great-Recession trough of 57.4 (Q4 2008), and far below the 74.1 it held in Q2 2020 while jobs vanished. Over 40% of Americans still named inflation the family’s top financial problem in early 2024 (vs. an ~8% average in 2000–2021, which never exceeded 18% even at the recession’s peak); ~30% still did in 2025.

Source: Hurst et al. (2026) and notes; U. Michigan Surveys of Consumers; Gallup; Pew Research Center (2022). Sentiment path stylized to reported points.

Fig. 15 The Belgium control group Same inflation, same shocks — one indexed wage system · 2020Q1 → 2024Q4
Compare Belgium with:

Belgian wages are automatically indexed to inflation through collective bargaining. Belgium, Germany, the Netherlands and the Euro Zone all absorbed ~22–24% price increases over 2020–2024 (Denmark ~16%), with nearly identical unemployment and job-vacancy rates. Yet Belgian real wages regained pre-inflation levels by 2023 and confidence recovered; among the peers, neither had recovered by end-2024. Confidence is the paper’s Table 3 verbatim (percentage-point deviations from each country’s 2016–19 mean); real-wage paths are stylized to the stated outcomes.

Source: Hurst et al. (2026), Fig. 21 & Table 3; Eurostat/FRED confidence series.

Fig. 16 Indexation at home U.S. consumer sentiment by age group · deviation from the group’s 2016–19 baseline
Year:

The same pattern runs through the U.S. itself: the closer a group’s income is to being indexed — Social Security by law, plus asset income — the smaller its sentiment collapse. Retired-age groups fell least and healed most; working-age groups, exposed to sticky norms, fell furthest and stayed down. The share of income from retirement sources is printed inside each bar.

Source: Hurst et al. (2026), Table 4 (verbatim); U. Michigan Survey of Consumers micro data; retirement-income shares from CPS.

Sentiment recovered where real income was protected — not simply where inflation stopped.

Reference · filed under the record

The ledger, 2020–2025

The whole episode on one strip — scroll or use the buttons.

  1. Dec 2020

    100

    Baseline. The four-year stayer cohort starts here; real wage index = 100.

  2. Mid 2021

    ↑

    Prices ignite. CPI climbs from ~2% to ~6% in months; firm norms stay pinned near 3%.

  3. Jun 2022

    +9.1%

    Peak inflation. The norm has clicked up barely a notch; real wages trough ~4% below Dec 2020.

  4. Q3 2022

    56.1

    Consumer sentiment undercuts its Great-Recession trough (57.4, Q4 2008).

  5. Mid 2023

    3.0%

    Inflation returns to normal and so does real wage growth — but the lost ground is not made up.

  6. Late 2024

    ≈100

    Median real wage regains its Dec-2020 level — four years to get back to the starting line.

  7. Dec 2024

    43%

    Cohort closes: 43 of 100 stayers end below where they began; mean loss ~9% among them.

  8. Dec 2025

    −7%

    Real wage index still ~7% below the extrapolated 2017–19 trend (~4% below 2000–19). The lower track persists.

More demos
Get started with Devin CLI