American productivity has run ahead of American pay since 1964 by a factor of 2.55, or by a factor of 1.18. Both figures come from the same BLS series over the same 249 quarters. The distance between them is made of two methodological choices that are almost never stated. Sixty-two years of data cut across the era-by-era account of macro-financial regimes, which is one reason the conventions matter.
Eco3min Research · 8 September 2026 · 249 quarterly observations, 1964 Q1 to 2026 Q2 · CSV under CC BY 4.0
US productivity and pay: the four conventions in use
Index, 1964 Q1 = 100. Each pay line pairs one pay measure with one deflator.
Sources: U.S. Bureau of Labor Statistics via FRED (OPHNFB, COMPNFB, COMPRNFB, AHETPI, IPDNBS). Chart: Eco3min Research.
In short
Between 1964 Q1 and 2026 Q2, US labour productivity per hour rose to an index of 333.4 (1.95% a year). Pay over the same window ends at 130.9 under the convention the Economic Policy Institute uses, and at 281.9 under the one Martin Feldstein argued for. That is a decoupling ratio of 2.55 against 1.18. The BLS publishes neither: its own real hourly compensation series is consumer-deflated and sits between the two, at 1.67. Two choices separate them: whether employer benefits count as pay, worth 68.9 index points, and whether pay is deflated by consumer prices or by the price of what the sector produces, worth 53.7 points. Their joint effect adds 28.3 more. This page publishes all four combinations rather than the one that flatters a prior. Whichever convention is used, productivity remains the slow variable underneath the economic cycle and its successive phases.
The spread, at 2026 Q2
2.55xwidest reading, production wages deflated by CPI-U-RS
1.18xnarrowest reading, all-worker compensation deflated by output prices
1.67xthe BLS own published real compensation series, between the two
151index points of pay between the two, out of 333.4 for productivity
249quarters, 1964 Q1 to 2026 Q2, one deflator switch documented
The four conventions, side by side
The productivity and pay debate is not a disagreement about data. Every figure below comes from the same five BLS series. It is a disagreement about two definitions, and the table is the whole argument.
| Convention | Pay measure | Deflator | Pay index 2026 Q2 | Ratio | Used by |
|---|---|---|---|---|---|
| C1 | Production & nonsupervisory hourly wages | BLS research deflator (CPI-U-RS) | 130.9 | 2.55x | Economic Policy Institute |
| C2 | Hourly compensation, all workers | BLS research deflator (CPI-U-RS) | 199.8 | 1.67x | BLS published real hourly compensation |
| C3 | Production & nonsupervisory hourly wages | Nonfarm business output deflator | 184.6 | 1.81x | used by neither camp |
| C4 | Hourly compensation, all workers | Nonfarm business output deflator | 281.9 | 1.18x | Feldstein (2008), labour share framing |
Productivity is the same line in all four rows: an index of 333.4 at 2026 Q2, growing at 1.95% a year. Only the pay line moves.
The two choices that move the answer
The first choice is what counts as pay. Hourly wages exclude employer contributions to health insurance, retirement plans and payroll taxes. Total compensation includes them. Over 249 quarters that gap is worth 68.9 index points, and almost all of the growth in it is employer health premiums, which reach the worker as coverage rather than as money to spend.
The second choice is which price index deflates pay. Consumer price indices measure what workers buy. The output deflator measures the price of what the sector produces. Productivity is already expressed in output prices by construction, so comparing it to consumer-deflated pay compares two different price bases. That choice is worth 53.7 points.
Neither choice on its own accounts for the distance. Applied together they add 28.3 points more than the sum of their separate effects, because the two adjustments compound over six decades.
What separates the two answers, step by step
Pay index at 2026 Q2, 1964 Q1 = 100, decomposed from the widest reading to the narrowest.
Sources: U.S. Bureau of Labor Statistics via FRED. Chart: Eco3min Research.
Which camp uses which convention
The Economic Policy Institute compares production and nonsupervisory workers’ hourly wages, deflated by the CPI-U-RS, to productivity for the total economy. That is row C1, and it is the source of the familiar chart showing a wide and widening gap.
Martin Feldstein’s 2008 objection uses total hourly compensation deflated by the output price index of the same sector. That is row C4, and it shows pay tracking productivity far more closely.
The BLS itself publishes neither row. Its headline real hourly compensation series is deflated by consumer prices, not by output prices, which makes it row C2 at 1.67. That is the check on the whole pipeline: reconstructing the published series from the derived deflator returns 0.000 index points of error. The statistical agency’s own number sits between the two advocacy positions rather than endorsing either.
The narrow reading carries a detail neither camp tends to quote. Under C4, pay was not merely close to productivity, it was ahead of it in 74 of the 249 quarters, and stayed ahead as late as 2001 Q3. On that convention the familiar story of a gap opening in the early 1970s does not appear: the divergence starts in the 2000s. Under C1 the same crossing last happened in 1965 Q2, and never again.
Row C3 is used by nobody. It exists here because it isolates the deflator choice on its own, which is what makes the decomposition possible. It also makes a point neither camp emphasises: C2 and C3 land close together, at 199.8 and 184.6, which means the two choices carry comparable weight rather than one dominating.
What this does not settle
Publishing four numbers is not the same as saying the question has no answer. Three arguments survive the exercise, and they pull in different directions.
The case for the consumer deflator is that workers experience their pay as purchasing power over consumption goods, not over the output of the nonfarm business sector. If the question is whether living standards tracked productivity, the consumer basket is the relevant basis, and the wider reading is the honest one.
The case for the output deflator is arithmetic. Productivity is output per hour valued in output prices. Comparing it to pay valued in consumer prices embeds the relative price of consumption against production into a statistic that claims to be about the division of output. On that reading the narrow figure is the only internally consistent one.
The case against reading either as a measure of fairness is that the compensation series covers all workers including the highest paid, so a widening distribution of pay can leave average compensation tracking productivity while the median worker sees neither. Row C4 is consistent with both a healthy and an unequal labour market, and does not distinguish between them.
Methodology
Productivity is real output per hour for the nonfarm business sector (FRED series OPHNFB), quarterly, seasonally adjusted. Pay is either average hourly earnings of production and nonsupervisory employees (AHETPI, monthly, averaged to quarters) or hourly compensation for the nonfarm business sector (COMPNFB). Deflators are the nonfarm business output price index (IPDNBS) and the BLS research deflator.
The BLS research deflator is not published as a standalone FRED series. It is derived here as nominal hourly compensation divided by the BLS real hourly compensation series (COMPNFB divided by COMPRNFB). The BLS documents that series as deflated by the CPI-U-RS through the end of 2005 and by the CPI-U thereafter, and the derived series reproduces exactly that break: it diverges from the published CPI-U between 1974 and 1994, then stops diverging after 2005. Reconstructing real compensation from the derived deflator returns the published BLS series to 0.000 index points, which is the cross-check for the whole pipeline.
All series are indexed to 100 at 1964 Q1, the first quarter in which all four conventions can be computed, because AHETPI begins in 1964. The two compensation-based conventions can be extended back to 1947, and are not, so that the four rows share one window.
decoupling_index = productivity_index / pay_index
pay_index = (nominal pay / deflator) rebased to 100 at 1964 Q1
Limits
- Using the published CPI-U instead of the BLS research deflator widens C1 from 2.55x to 2.78x and C2 from 1.67x to 1.82x. Both variants are in the CSV. Neither camp uses the published CPI-U for this comparison, which is why it is a sensitivity rather than a row.
- Productivity here is nonfarm business, while the EPI comparison uses total economy productivity. The two differ by roughly the government and household sectors, and the difference is smaller than any of the choices decomposed above.
- AHETPI covers production and nonsupervisory workers, about four fifths of private employment. It is an average within that group, not a median, and it is not adjusted for changes in the composition of that group.
- Every figure is a level comparison between two quarters sixty-two years apart. Nothing here identifies a cause, a mechanism or a turning point, and no line on this page is a forecast.
Questions
Has US pay kept up with productivity?
It depends on two definitions. Measured as production and nonsupervisory wages deflated by consumer prices, productivity outran pay by 2.55 times between 1964 and 2026. Measured as total compensation deflated by output prices, the factor is 1.18. The BLS own published real compensation series gives 1.67. All three come from the same BLS data over the same 249 quarters.
Why do different sources give different productivity-pay gaps?
Because they make different choices on what counts as pay and which price index deflates it. Including employer benefits moves the pay index by 68.9 points, switching from a consumer deflator to an output deflator moves it by 53.7 points, and doing both adds 28.3 points beyond the sum of the two.
Which convention does the Economic Policy Institute use?
Production and nonsupervisory hourly wages deflated by the CPI-U-RS, compared to total economy productivity. That is row C1 in the table above, giving a ratio of 2.55 at 2026 Q2. It is not the convention the BLS uses for its own published real compensation series, which is row C2 at 1.67.
What is the output deflator and why does it change the answer?
It is the price index of what the nonfarm business sector produces, rather than of what workers consume. Productivity is measured in output prices by construction, so deflating pay the same way removes the relative price of consumption from the comparison. It narrows the ratio from 2.55 to 1.81 when applied to production wages alone.
Do employer benefits belong in pay?
They are part of what an employer pays for an hour of labour, and most of their growth is health insurance premiums. Counting them moves the pay index up by 68.9 points over the period. Whether that is income a worker can use is the substantive disagreement, and this page does not resolve it.
Does a narrow gap mean workers are doing well?
No. The compensation series is an average across all workers, so it is consistent with a widening distribution in which the median worker gains far less than the average. The narrow reading describes the aggregate division of output, not its distribution.
Where does the data come from and can it be reused?
Five BLS series retrieved from FRED, none of which carries a third-party copyright notice. The quarterly series and the convention registry are published as CSV under CC BY 4.0, covering 249 quarters from 1964 Q1 to 2026 Q2.
Sources
- DataU.S. Bureau of Labor Statistics, Productivity and Costs, series OPHNFB, COMPNFB, COMPRNFB, IPDNBS, retrieved from FRED, St. Louis Fed.
- DataU.S. Bureau of Labor Statistics, Current Employment Statistics, average hourly earnings of production and nonsupervisory employees (AHETPI), retrieved from FRED.
- MethodU.S. Bureau of Labor Statistics, Handbook of Methods, Major Sector Productivity, data sources for compensation and deflation.
- PaperMartin Feldstein, “Did Wages Reflect Growth in Productivity?”, NBER Working Paper 13953, 2008.
- PaperAnna Stansbury and Lawrence Summers, “Productivity and Pay: Is the Link Broken?”, NBER Working Paper 24165, 2017.
- PaperRobert Z. Lawrence, “Recent Declines in Labor’s Share in US Income: A Preliminary Neoclassical Account”, Peterson Institute Working Paper 15-10, 2015.
- SeriesEconomic Policy Institute, The Productivity-Pay Gap, methodological appendix.
Eco3min Research, “The US productivity-pay gap under four conventions”, 8 September 2026. Quarterly dataset, 249 observations, 1964 Q1 to 2026 Q2, CC BY 4.0.
Disclaimer – Financial Information: The analyses, commentary, and content published on eco3min.fr are provided for informational and educational purposes only. They do not constitute investment advice or a solicitation to buy or sell financial instruments. Past performance is not indicative of future results. All investment decisions involve risk and are the sole responsibility of the reader.