Data story · the 1973 break
Productivity has left wages in the dust
For a generation after the war, the typical worker’s pay rose right alongside how much they produced. Around 1973 the two lines split — output kept climbing, pay flattened — and the wedge between them has only widened since. That wedge is the income workers generated but didn’t take home.
Since 1948, net productivity has grown +298% but the typical worker’s pay only +143%. They rose together until 1973 (+110% vs +87% that year) — then pay stalled while output kept climbing.
Net productivity of the total economy vs. the inflation-adjusted hourly compensation of production/non-supervisory workers — about 80% of the private-sector workforce — as cumulative % change since 1948.
And the little growth there was wasn’t shared
Real median hourly wage growth since 1973, by group, against the same productivity line. Every group trailed productivity badly — women gained most (+52% since 1973) but from the lowest base, while men's typical wage barely moved (+12%).
Median wages (not total compensation), inflation-adjusted to 2025 dollars. Asian/AAPI is omitted from this since-1973 view — its series begins 1989, so it shares no 1973 baseline.
The next surge · will it be different?
AI may explode productivity. History says pay won’t follow.
Productivity is about to get another shove, and forecasters disagree wildly on its size. Goldman Sachs estimates generative AI could add roughly 1.5 percentage points a year to U.S. labor productivity over a decade with broad adoption; MIT’s Daron Acemoglu puts the total nearer 0.7% over the same ten years — a rounding error. Call it a range from transformative to negligible.
But size was never the worker’s problem. Since 1973 productivity grew +90% and the typical worker’s pay captured only about a third of it (+30%). The economy never stopped producing more per hour — the question was always who kept the gains.
Illustrative — if history’s pass-through holds
So the AI question for wages isn’t how much will productivity rise — it’s whether this surge passes through to pay or follows the post-1973 pattern and accrues to owners and the top. Faster productivity is necessary for broad raises. The last fifty years say it is nowhere near sufficient.
Data: EPI State of Working America Data Library — productivity and pay indexes, and median hourly wages by gender and race/ethnicity — the authoritative source behind EPI’s published productivity–pay gap. Feeds: public/data/pay-gap-productivity.csv + pay-gap-demographics.csv, built by scripts/build-pay-gap.mjs. AI productivity forecasts: Goldman Sachs & Acemoglu (NBER 32487). View the original on Tableau Public →