OP here. These are not two estimates with error bars. They are two definitions applied to the same BLS series over the same 250 quarters, 1964 Q1 to 2026 Q3.
Two choices do the work. Counting employer benefits as pay moves the pay index by 71.6 points. Deflating pay by the price of what the sector produces, instead of consumer prices, moves it by 54.6. Doing both adds 29.8 points more than the sum, so neither choice explains the gap on its own.
The convention EPI uses (production and nonsupervisory wages, CPI-U-RS) gives 2.58x here. The one Feldstein argued for (total compensation, output deflator) gives 1.18x. The BLS publishes neither: its own real hourly compensation series is consumer-deflated and sits at 1.67x.
In real time: this morning's BLS release has output per hour up 3.6% annualized in Q3 and real hourly compensation up 1.1%. Since 2025 Q4, output per hour is up 1.75% and real hourly compensation up 0.17% (2017 = 100: 119.8 to 121.9, and 110.6 to 110.8).
Two limits I would rather state than have found. Productivity here is nonfarm business with a 1964 base, not EPI's total economy from 1979, so 2.58x is my computation of their convention, not their published number. And AHETPI is an average over production and nonsupervisory workers, not a median.
Data: BLS series OPHNFB, COMPNFB, COMPRNFB, IPDNBS, AHETPI and CPIAUCSL (FRED IDs), with this morning's Q3 values checked against the BLS API; the monthly wage and CPI series are averaged to quarters. The CPI-U-RS deflator is derived as COMPNFB / COMPRNFB and reproduces the published BLS series to 0.000 index points. Quarterly CSV and the convention registry are on the page. Happy to discuss methodology.
I appreciate the effort here, but can you re-run these figures easily while using a more uniform inflation figure?
The driver behind the oft misunderstood productivity/pay gap is a difference in inflation measures. Productivity is a figure derived from GDP - GDP is deflated via the GDP deflator. Real incomes are adjusted per CPI, and in general CPI runs about a half percent per year hotter than the GDP deflator.
See here: https://fred.stlouisfed.org/blog/2023/03/when-comparing-wages-and-worker-productivity-the-price-measure-matters
> The results are quite different, and the decoupling isn’t as stark. The original real labor compensation line describes what purchasing power workers have, as it is deflated by a price index that tracks a typical basket of goods a household would buy. The two new lines look at this from the employer side: How much are workers paid compared with (i) the value of all things produced (GDP deflator) and (ii) the prices that producers are getting for their wares (PPI)? These two lines are much closer to the productivity line, indicating that the choice of prices matters.
> Why would you take one price series over another? It depends on what question you’re asking of the data. If you want to understand how businesses decide to allocate resources to factors of production (labor, capital, intermediates), then the second graph is relevant. If the topic is about worker purchasing power, then the first graph is relevant. But more fundamentally, why would these price indexes differ so much? That’s the topic of our next blog post…
Basically think about it this way: the formula for labor productivity is “Productivity= GDP/Labor inputs”. GDP is a real figure to compare to wages you also need a real wage figure. If you use two different measures of inflation then a decoupling isn’t possible it’s inevitable.
Also the next post referenced here is really good, it’s a summary of why these price indexes diverge and why CPI tends to have higher prints than any other measure of inflation: https://fred.stlouisfed.org/blog/2023/03/the-differences-among-price-indexes/
Yes, that's already one of the four. Same pay (total compensation), only the deflator changes: 1.67x with consumer prices, 1.18x with the nonfarm business output deflator, which is what the productivity series itself uses. The GDP deflator is a bit different since this is nonfarm business, not GDP. With it I get about 1.40x. Agreed that it depends on the question: consumer prices for what pay buys, output prices for what labor costs the employer.
Yeah, the important thing is comparing a CPI driven real wages here doesn’t make sense because it’s not the comparison being made. If the question is “have wages kept up with productivity” then you need to use the same deflator across both measures.
If the question is “have wages increased in real terms” then CPI is fine, but that’s unrelated to productivity.
Did you purposefully use Q1 as the index quarter, or was that random? if not, what was the reason, does it have something to do with Q1 generally having a lower productivity than the other quarters of the year(often as much as half as much LINK: https://www.stlouisfed.org/on-the-economy/2017/march/residual-seasonality-return-old-first-quarter-friend )? I have generally learned that unless you seasonally/quarterly-adjust the data it is best to use the same index quarter to compare. (just want to know if it was something you thought about or if there is a reason why you decided to do so, not attacking your post, the graph looks nice : ) )
Thanks! Not really a choice, the production wage series only starts in Jan 1964 so Q1 is the first quarter where all four work. Ran it Q3 to Q3 to check and it barely moves: 2.58 / 1.68 / 1.81 / 1.18 vs 2.58 / 1.67 / 1.82 / 1.18.
You wouldn't expect it at one factory. Across a whole sector it's closer to accounting: output gets split between workers and capital, so if labor's share holds steady, pay and productivity grow together (with pay deflated by output prices). That's what the 1.18x measures: labor's share is about 15% below 1964.
That suggests they should be correlated, but not necessarily 1:1. Capital is also involved in the process of increasing output (indeed it’s more involved in this example).
Yes but if there is a linear relation as you agree because scale should not affect the wage share of productivity if the degree of competition remains the same then the proportion increased is 1 to 1. Why would the wage share decrease with productivity holding everything else constant
What do you mean lol at least in developed countries the economy is mostly service based not exactly the most capital intensive of sectors and we have offshored pretty much all industry except that with extremely high added value like luxury products or cars. The only plausible reason for wage share decreasing is a decrease in market competition caused by consolidation due to the information revolution and a larger open global market. Also the destruction of medium sized businesses
The proportion of growth of both productivity and average wage should stay the same for any constant wage share. That happened for at least 40 years after the second world war then stopped. It's not because capital was not as involved as it is now as the private investment share of GDP was higher back then
Why should we assume constant wage share? Assuming constant wage share is the same thing as assuming wages should be 1:1 correlated with wages. That’s just circular reasoning.
Well we shouldn't you're right. But we know historically it was the same for decades while there was a very high level of private investment. So how can you explain that
that hasn't been how it works since the beginning of time. it makes sense on paper but the people or businesses that take the risk to start a business will always get a much larger percentage. workers will get replaced by another worker willing to do the work at a lower cost well before they get a wage increase that matches the productivity increase
this is just the economics behind how the world runs. yes public companies do rely on investors....so they have a board that determines whether or not they keep their job as CEO or CFO etc etc and whether than keep their job is based on performance and returns to the investors. raising labor costs is a risk they lose their own job if then don't reward the shareholders. companies will always find someone else to work unless you have a special set of skills that is not replaceable.
I own small businesses and there are a lot of small business owners out there. everyone only thinks of companies in the s&p but there are a lot of other small businesses out there where labor cost is by far the most significant expense and where the failure rate is extremely high.
either way you look at it whether it's a small business or a large corporation, the common worker is not going to ever get close to matching output due to higher efficiency. it should definitely correlate higher but giant companies are not exactly giant companies because they pay everyone what they output
Because companies compete for workers. If all factories become more productive then they will have to pay workers more or they will work at a different factory.
If there isn't an overabundance in demand for labor this isn't the case entirely. All that matters is the marginal effects. If the market doesn't demand a higher wage, a higher wage doesn't become the equilibrium.
But productivity might decrease need for workers. If you need 50% less workers because your machine is 100% more efficient, you’d expect demand for workers to fall.
Wages used to keep pace with GDP growth in the postwar period.
As the percent of workers in unions went down the percent of surplus going to workers went down.
I dont expect a 1 to 1 ratio to happen naturally, by some spooky invisible hand. People have to organize and fight for a bigger share of the surplus.
If workers work to make a factory successful enough that it can invest in better technology that makes each worker more productive, workers should be able to fight for better compensation. And we should have laws in place to allow this so that we have a society that benefits the many and not the few.
Workers have been getting better compensation. Real median wages have increased by 56.2% since 1974. But as productivity is mainly driven by technology, it wouldn't be fair for labor to keep the same percentage.
You also have to causation backwards. Unions failing didn't cause the percentage of surplus going to labor to decrease. That's a common misunderstanding. Unions drove American manufacturing to be uncompetitive. Thus reducing their own opportunities in the long run.
There's no reason we should us force to take from one group and give to the other.
Because I care more about people than I do treating economics like a thought experiment, and my expectations are aligned as such. You got a faster production line? Great so that means the workers you had up to that point played a key role in delivering success that allowed you to afford to buy those upgrades and they deserve to benefit from the investment as well. You would not be in the position to upgrade without them.
[OP] Low_Ability4450 | an hour ago
OP here. These are not two estimates with error bars. They are two definitions applied to the same BLS series over the same 250 quarters, 1964 Q1 to 2026 Q3.
Two choices do the work. Counting employer benefits as pay moves the pay index by 71.6 points. Deflating pay by the price of what the sector produces, instead of consumer prices, moves it by 54.6. Doing both adds 29.8 points more than the sum, so neither choice explains the gap on its own.
The convention EPI uses (production and nonsupervisory wages, CPI-U-RS) gives 2.58x here. The one Feldstein argued for (total compensation, output deflator) gives 1.18x. The BLS publishes neither: its own real hourly compensation series is consumer-deflated and sits at 1.67x.
In real time: this morning's BLS release has output per hour up 3.6% annualized in Q3 and real hourly compensation up 1.1%. Since 2025 Q4, output per hour is up 1.75% and real hourly compensation up 0.17% (2017 = 100: 119.8 to 121.9, and 110.6 to 110.8).
Two limits I would rather state than have found. Productivity here is nonfarm business with a 1964 base, not EPI's total economy from 1979, so 2.58x is my computation of their convention, not their published number. And AHETPI is an average over production and nonsupervisory workers, not a median.
Data: BLS series OPHNFB, COMPNFB, COMPRNFB, IPDNBS, AHETPI and CPIAUCSL (FRED IDs), with this morning's Q3 values checked against the BLS API; the monthly wage and CPI series are averaged to quarters. The CPI-U-RS deflator is derived as COMPNFB / COMPRNFB and reproduces the published BLS series to 0.000 index points. Quarterly CSV and the convention registry are on the page. Happy to discuss methodology.
RIP_Soulja_Slim | an hour ago
I appreciate the effort here, but can you re-run these figures easily while using a more uniform inflation figure?
The driver behind the oft misunderstood productivity/pay gap is a difference in inflation measures. Productivity is a figure derived from GDP - GDP is deflated via the GDP deflator. Real incomes are adjusted per CPI, and in general CPI runs about a half percent per year hotter than the GDP deflator.
See here: https://fred.stlouisfed.org/blog/2023/03/when-comparing-wages-and-worker-productivity-the-price-measure-matters
> The results are quite different, and the decoupling isn’t as stark. The original real labor compensation line describes what purchasing power workers have, as it is deflated by a price index that tracks a typical basket of goods a household would buy. The two new lines look at this from the employer side: How much are workers paid compared with (i) the value of all things produced (GDP deflator) and (ii) the prices that producers are getting for their wares (PPI)? These two lines are much closer to the productivity line, indicating that the choice of prices matters.
> Why would you take one price series over another? It depends on what question you’re asking of the data. If you want to understand how businesses decide to allocate resources to factors of production (labor, capital, intermediates), then the second graph is relevant. If the topic is about worker purchasing power, then the first graph is relevant. But more fundamentally, why would these price indexes differ so much? That’s the topic of our next blog post…
Basically think about it this way: the formula for labor productivity is “Productivity= GDP/Labor inputs”. GDP is a real figure to compare to wages you also need a real wage figure. If you use two different measures of inflation then a decoupling isn’t possible it’s inevitable.
Also the next post referenced here is really good, it’s a summary of why these price indexes diverge and why CPI tends to have higher prints than any other measure of inflation: https://fred.stlouisfed.org/blog/2023/03/the-differences-among-price-indexes/
[OP] Low_Ability4450 | 58 minutes ago
Yes, that's already one of the four. Same pay (total compensation), only the deflator changes: 1.67x with consumer prices, 1.18x with the nonfarm business output deflator, which is what the productivity series itself uses. The GDP deflator is a bit different since this is nonfarm business, not GDP. With it I get about 1.40x. Agreed that it depends on the question: consumer prices for what pay buys, output prices for what labor costs the employer.
RIP_Soulja_Slim | 55 minutes ago
Yeah, the important thing is comparing a CPI driven real wages here doesn’t make sense because it’s not the comparison being made. If the question is “have wages kept up with productivity” then you need to use the same deflator across both measures.
If the question is “have wages increased in real terms” then CPI is fine, but that’s unrelated to productivity.
ConsiderationOne4022 | an hour ago
Did you purposefully use Q1 as the index quarter, or was that random? if not, what was the reason, does it have something to do with Q1 generally having a lower productivity than the other quarters of the year(often as much as half as much LINK: https://www.stlouisfed.org/on-the-economy/2017/march/residual-seasonality-return-old-first-quarter-friend )? I have generally learned that unless you seasonally/quarterly-adjust the data it is best to use the same index quarter to compare. (just want to know if it was something you thought about or if there is a reason why you decided to do so, not attacking your post, the graph looks nice : ) )
[OP] Low_Ability4450 | 43 minutes ago
Thanks! Not really a choice, the production wage series only starts in Jan 1964 so Q1 is the first quarter where all four work. Ran it Q3 to Q3 to check and it barely moves: 2.58 / 1.68 / 1.81 / 1.18 vs 2.58 / 1.67 / 1.82 / 1.18.
thinkB4WeSpeak | 42 minutes ago
Only beneficial if you're a profit sharing or co op business. People need to stop giving so much to companies not giving to them or unionize.
Key-Organization3158 | 30 minutes ago
So the gap grew by about 1.2 when you adjust properly. Considering how much more capital intensive work is these days, that's not bad at all.
rhino369 | an hour ago
Why would you expect productivity to 1:1 correlated with wages?
If a factory gets a faster production line why would the works expect to get more pay.
[OP] Low_Ability4450 | an hour ago
You wouldn't expect it at one factory. Across a whole sector it's closer to accounting: output gets split between workers and capital, so if labor's share holds steady, pay and productivity grow together (with pay deflated by output prices). That's what the 1.18x measures: labor's share is about 15% below 1964.
Simeh | an hour ago
Because workers are involved in the process of increasing output
runningraider13 | an hour ago
That suggests they should be correlated, but not necessarily 1:1. Capital is also involved in the process of increasing output (indeed it’s more involved in this example).
lilychouchou-guy | an hour ago
Yes but if there is a linear relation as you agree because scale should not affect the wage share of productivity if the degree of competition remains the same then the proportion increased is 1 to 1. Why would the wage share decrease with productivity holding everything else constant
Appropriate-Bid8671 | an hour ago
Because why should Capital take less profit?
lilychouchou-guy | an hour ago
It can take more or less the thing is why is the share changing
DragonBank | an hour ago
Because 2026 production is very capital intensive and not as labor intensive.
lilychouchou-guy | 54 minutes ago
What do you mean lol at least in developed countries the economy is mostly service based not exactly the most capital intensive of sectors and we have offshored pretty much all industry except that with extremely high added value like luxury products or cars. The only plausible reason for wage share decreasing is a decrease in market competition caused by consolidation due to the information revolution and a larger open global market. Also the destruction of medium sized businesses
Adventurous-Roof488 | 26 minutes ago
Not all services are labor intensive. Healthcare certainly is but internet search definitely isn’t.
lilychouchou-guy | 12 minutes ago
They aren't very capital intensive either
runningraider13 | 26 minutes ago
Can you add some punctuation? Struggling to understand what you mean here
lilychouchou-guy | 16 minutes ago
The proportion of growth of both productivity and average wage should stay the same for any constant wage share. That happened for at least 40 years after the second world war then stopped. It's not because capital was not as involved as it is now as the private investment share of GDP was higher back then
runningraider13 | 9 minutes ago
Why should we assume constant wage share? Assuming constant wage share is the same thing as assuming wages should be 1:1 correlated with wages. That’s just circular reasoning.
lilychouchou-guy | 8 minutes ago
Well we shouldn't you're right. But we know historically it was the same for decades while there was a very high level of private investment. So how can you explain that
kloakndaggers | an hour ago
that hasn't been how it works since the beginning of time. it makes sense on paper but the people or businesses that take the risk to start a business will always get a much larger percentage. workers will get replaced by another worker willing to do the work at a lower cost well before they get a wage increase that matches the productivity increase
13Krytical | an hour ago
How long does “taking a risk” mean something?
It’s only “taking a risk” in very specific situations.
Most big businesses rely on investors and others so face NO risk, but still take a huge share of the pie.
Appropriate-Bid8671 | an hour ago
In perpetuity, obviously.
kloakndaggers | 59 minutes ago
this is just the economics behind how the world runs. yes public companies do rely on investors....so they have a board that determines whether or not they keep their job as CEO or CFO etc etc and whether than keep their job is based on performance and returns to the investors. raising labor costs is a risk they lose their own job if then don't reward the shareholders. companies will always find someone else to work unless you have a special set of skills that is not replaceable.
I own small businesses and there are a lot of small business owners out there. everyone only thinks of companies in the s&p but there are a lot of other small businesses out there where labor cost is by far the most significant expense and where the failure rate is extremely high.
either way you look at it whether it's a small business or a large corporation, the common worker is not going to ever get close to matching output due to higher efficiency. it should definitely correlate higher but giant companies are not exactly giant companies because they pay everyone what they output
13Krytical | 57 minutes ago
Right.
Owner mentality.
You think because you own, nobody else is as smart as you unless they also own.
But no, you are spouting ideas that exploiting people over some idea that your original risk is worth more than their daily efforts, forever, is fine?
I hope your businesses tank.
rhino369 | an hour ago
But they aren’t always
Adventurous-Roof488 | an hour ago
Capital is involved in securing a faster production line, not workers.
Simeh | an hour ago
Capital doesn't increase productivity on its own. You need skilled workers to roll out, run and manage the upgraded processes
Adventurous-Roof488 | an hour ago
You may need fewer workers to do so and they may not need to be as skilled as before.
SghettiAndButter | 57 minutes ago
So we are in a race to the bottom?
Adventurous-Roof488 | 29 minutes ago
Real incomes are higher than in the past. Is that how you define a race to the bottom? I don’t.
SghettiAndButter | 28 minutes ago
Needing fewer workers is a race to the bottom. What do we do when there’s more people than good jobs?
a_library_socialist | an hour ago
And capital comes from the work that workers did previously
jayantsr | an hour ago
No it comes from how much the top executives fooled the market to invest in them
a_library_socialist | 49 minutes ago
where do you think "the market" got resources? Who do you think builds the machines that constitute "capital"?
jayantsr | 49 minutes ago
Some poor child in communist china most prob
schfiftyfiveshades | an hour ago
Maybe not necessarily 1:1, but compensation up at 1/3 of the productivity is theft.
rhino369 | an hour ago
How much went to driving costs down?
Glupscher | an hour ago
Because companies compete for workers. If all factories become more productive then they will have to pay workers more or they will work at a different factory.
DragonBank | an hour ago
If there isn't an overabundance in demand for labor this isn't the case entirely. All that matters is the marginal effects. If the market doesn't demand a higher wage, a higher wage doesn't become the equilibrium.
rhino369 | an hour ago
But productivity might decrease need for workers. If you need 50% less workers because your machine is 100% more efficient, you’d expect demand for workers to fall.
Glupscher | 52 minutes ago
Sure, that's one of the many reasons why the assumption only works in perfect markets.
SghettiAndButter | an hour ago
Demand for workers decreasing is generally bad for society
rhino369 | 48 minutes ago
Ok and?
SghettiAndButter | 45 minutes ago
Ok and I don’t want to live in a society where the average worker is destitute
rhino369 | 43 minutes ago
You should try to move to America then.
SghettiAndButter | 41 minutes ago
What? I live here. If you think there is no workers in America that are destitute then you just live in a bubble
Key-Organization3158 | 33 minutes ago
Worker pay is up significantly since 1974. 56.2% increase in the real median wage.
BroughtBagLunchSmart | an hour ago
Reagan I thought you died decades ago.
rhino369 | an hour ago
No serious economist thinks productivity and wages should have a 1:1 correlation.
pluralofjackinthebox | 46 minutes ago
Wages used to keep pace with GDP growth in the postwar period.
As the percent of workers in unions went down the percent of surplus going to workers went down.
I dont expect a 1 to 1 ratio to happen naturally, by some spooky invisible hand. People have to organize and fight for a bigger share of the surplus.
If workers work to make a factory successful enough that it can invest in better technology that makes each worker more productive, workers should be able to fight for better compensation. And we should have laws in place to allow this so that we have a society that benefits the many and not the few.
Key-Organization3158 | 33 minutes ago
Workers have been getting better compensation. Real median wages have increased by 56.2% since 1974. But as productivity is mainly driven by technology, it wouldn't be fair for labor to keep the same percentage.
You also have to causation backwards. Unions failing didn't cause the percentage of surplus going to labor to decrease. That's a common misunderstanding. Unions drove American manufacturing to be uncompetitive. Thus reducing their own opportunities in the long run.
There's no reason we should us force to take from one group and give to the other.
mynameisrockhard | an hour ago
Because I care more about people than I do treating economics like a thought experiment, and my expectations are aligned as such. You got a faster production line? Great so that means the workers you had up to that point played a key role in delivering success that allowed you to afford to buy those upgrades and they deserve to benefit from the investment as well. You would not be in the position to upgrade without them.