There’s no one correct measure of the labor share. There are a lot of ways to consider what should go in the numerator and the denominator (Karabarbounis 2024). But no matter how you approach it, I believe three things are generally agreed upon for the United States:
After being in a steady range from the 1950s to 1990s, it has fallen notably since 2000.
It fell again during the pandemic and reopening, and after stabilizing, appears to be falling again since early 2025. While there are many explanations for the fall in the 2000s (indeed, it can be explained “many times over” (Grossman and Oberfield, 2022)), there are fewer ones for the more recent fall.
This is true if you include depreciation; depending on what you are measuring, gross measures tend to be lowest on record and net measures among the lowest on record, but even then the recent falls since 2025 may make them the lowest.
Throw whatever you want at it, this is generally true. This has gotten more attention lately, especially with recent BLS gross measures (the only thing the government releases with the title “labor share”) showing the lowest values on record.
Richard DiSalvo and Erica York at the Tax Foundation have a new post pushing back on this, arguing that we should understand the labor share as making a “round trip” over the postwar era rather than falling. Their conclusion:
The labor share is back to a historically precedented, not “never-before seen,” level. And the “downward trend throughout” needs to be replaced with “the labor share rose, then fell, over the postwar era.” It has made a round trip, rather than declining consistently from its starting level.
So nothing to especially worry about, just a “round trip” back to older, historically precedented values.
Take a second to read their piece. Where to start? I think five things. (Github code for replication of these exercises here.)
First, their data concedes my first two points above. Here’s their “unambiguous labor share” in Figure 1:
After being stable in the 1950s through 1990s, it fell, and fell again over the past six years. So we’re in agreement there. Are we done? Well we’re already here, let’s keep going. The open question is number 3 above, what to make of the idea about this “round trip.”
Second, if you include proprietary (prop) income as labor income for their measure, then it becomes the lowest on record.1 Let’s stick with their measure, and add in some proprietary income as labor income. Their post is in a bind, because it wants to argue, following the latest evidence, that we should think of prop income as being mostly labor income, on the order of three-quarters (Smith et al 2019), to try and talk up the idea that there’s a lot of mis-measured and hidden labor income out there. Fair enough. But if you do that, given the higher share of prop income in the 1940s, it raises the labor share then. Here is their measure with 0%, 50%, and 100% of prop income as labor income (Figure 2):
Adding any prop income to labor share raises the level of income today but it also raises it even more in the 1940s, since prop income was a much bigger share of the economy then. At any sufficient prop share to income, and especially at 50% or at 100%, the current quarter isn’t just low, it’s the single lowest reading in the entire 1947-2026 series. I tend to use 50%, a value I should probably update by increasing it, but either way it annihilates any “round trip” discussion.
That’s really the entire thing, but let’s move on to the fragility of their result as they present it.
Third, as you can see from Figure 1, their idea of a “round trip” is really dependent on a handful of data points in the late 1940s when the quarterly data starts. But that’s a brief window, and a very weird economy reopening from World War II. Sticking with their unambiguous labor share measure, the average in the 21st century (2000Q1-2026Q2) is 70.8%, which is below the 73.9% average from 1950-1999. While the average is 69.2% in 1947-1949, by 1950 to 1952 it’s 71.3% and it increases fast from there.
Their round trip language invokes a staple of the labor share literature debates of the 2010s, which is that net share is back to earlier historical ranges. As Grossman and Oberfield (2022) put it: “The gross labor share was relatively stable through 2000, but then declined to a level below its historical range. In contrast, as noted by Bridgman (2018) and Rognlie (2015), the net labor share had been rising between 1940 and 1980, so the recently lower level is not without precedent.”
That was definitely true in 2015 and 2018. But is that true now? I want to use a labor share more consistent with the literature than what the Tax Foundation uses.2 The measure I watch most closely here is the net labor share of the non-financial corporate sector (so there’s no prop income to worry about), benchmarked against its own 1947-49 average (Figure 3):
As you can see from the dotted line here, during the 2000s and 2010s, this hung out in roughly that late 1940s and 1950s level. So a round trip? No. Sadly, since COVID, this has collapsed further, and fallen even more since Trump took office at the beginning of 2025. Something is different.
Fourth, they imply that the labor share was stable at a “starting level” before this higher midcentury range. Since they use quarterly data they can only go to 1948 in the NIPAs. But the real heads know if you really dig deep into those NIPA flat files, which you can conveniently do with the tidyusmacro::getNIPAFiles() function from my R library, you can get annual data from 1929 onward.
As you’ll see below, it’s very unstable during that pre-1948 period. Here’s their measure dating back to 1929 (Figure 4):
It itself is making many round trips! I’m not sure exactly what to make of it, given how, to put it gently, complicated the economy was 1929 to 1948. But the topsy-turvy nature of this is why Robert Solow argued, in the 1950s and using this data from 1929 to 1954, that he was “skeptical” about any consistency in the labor share which “may be an optical illusion” instead (Solow 1958).3 More, there’s a whole literature around sectoral shifts creating an illusion of stability here. So I’m not convinced that there’s a “starting level” at all.
Fifth, separately and speculatively from above, labor share actually mirrors 1% inequality this way. Using the annual data again, take the capital share of net income and put it up against top 1% share (from the World Inequality Database through 2024), and take a look (Figure 5):
It’s the same story with different decades? They both have certainly comes back over the full 1929-2024 overlap. I think we have a story that 1% inequality has largely been a choice we’ve made, though obviously many people argue it’s strictly a matter of supply and demand. We certainly don’t, in the back of our heads, think that the 1% share is fixed deep in the economy. Perhaps it’s time we think the same thing with the capital share.
But whether or not you buy this argument, you shouldn’t buy a labor share “round trip” one. It collapses like a weak Jenga block structure the moment you pull any piece.
BEA defines proprietors’ “prop” income as “Current-production income of sole proprietorships, partnerships, and tax-exempt cooperatives. Excludes dividends, monetary interest received by nonfinancial business, and rental income received by persons not primarily engaged in the real estate business.”
A funny thing about their definition is that the non-profit sector, because it generates no corporate profits, rental income, or proprietors’ income in the NIPAs, gets treated as having a 100% net labor share. Truly, we non-profit workers are the vanguard of the working class!
The title of the piece is “A Skeptical Note on the Constancy of Relative Shares” and it is fun to read, starting with the line: “Ever since the investigations of Bowley and Douglas it has been widely believed that the share of the national income accruing to labor is one of the great constants of nature, like the velocity of light or the incest taboo.”
Am I misreading something? It seems like they’re saying including proprietors matters, but none of their arguments are supported at all by including proprietors. Round trip as well as the late 1940’s being all time lows both seem to require 0% prop to me.
It looks to me like labor force participation is falling as boomers retire, recently exacerbated by immigrants exiting. We can measure this by count or income, but the answer looks similar-- definitional variations be damned. Choosing a five-year period from 80 years ago as the historical benchmark is risible. Is it this simple, or am I missing something?
Since the US population will continue to age, labor income as a proportion of the economy will continue to fall. Meanwhile, government outlays for retirement and healthcare of an aging society will continue to rise. The policy implication is that we can't continue to rely on taxing labor income to fund the government to the same extent without crushing the shrinking labor pool. So where do we find more tax money? Through process of elimination, I think we end up raising tax rates on capital, one way or another (dividend, capital gain, profit, estate, wealth tax), to levels seen in the middle to late 20th century. Also, import more labor.