Pay/productivity gap graphs are nonsense

Matt Rognlie left three excellent comments after my recent pay/productivity post, which confirmed some of my suspicions, but also cleared up a few points I was slightly confused about. In doing so, he made me even more skeptical of the graph that supposedly shows a huge gap between growth in productivity and growth in pay. Basically the graph is bogus, and has no useful information on productivity and pay, which have risen in tandem since 1965.

Matt points out that a small part of the supposed gap reflects the fact that the diagram starts at 1973, a very good year for labor.  As I pointed out (and Matt confirmed), the labor share of income has been stable since 1965, with a few ups and downs.  And 1973 was an up year.  One can debate which year is an appropriate starting point, but either way the labor share gap is no big deal.

It’s the terms of trade gap where Matt cleared up a misconception on my part. Although I didn’t mention this in the post (fortunately), I had assumed the gap reflected the fact that consumer goods prices rose faster than investment good prices. Rognlie shows that’s not the issue:

The other major component of the EPI graph is the “terms of trade” wedge, which is meant to be the gap between the price index for workers’ net output and the price index for their consumption. It’s important to emphasize that this is almost all spurious. This is clear enough from a quick sanity check – if you directly compare the behavior of the price indices for net domestic product (NDP) and personal consumption expenditures (PCE) since 1973, you’ll see that they have moved roughly in tandem:

[Some seriously lengthy and obscure price index discussion ensues…]

The EPI graph appears to show otherwise only because it opts to use the CPI-U-RS, rather than the PCE price index, to deflate workers’ compensation – and the CPI-U-RS has risen by 15% more than the PCE index since 1973. On its face, this seems pretty odd: for one thing, the NDP price index (like PCE) is chained while the CPI-U-RS isn’t. If you deflate net production by the former and compensation by the latter, the result is a “productivity-pay” gap that’s partly due to formula bias!

Matt provides much more detail, and is entirely convincing on this point.  And I just want to make sure readers are not getting lost in the weeds here.  This is not one of those “he said, she said” where reasonable people can disagree on whether the PCE or CPI is a better price index.  This is a pay/productivity gap being invented by using the slowly moving price index (NDP, which is similar to the PCE) to make worker productivity look better, and the faster moving price index (CPI) to make real wages look lower.  That’s not kosher.  You need to use the same type of index for both lines on the graph.

So now we have demolished one part of the pay gap, and shown the other goes away if you start the clock in 1965 (still a good year for American workers, according to the left) and not 1973.  What’s left is the biggest category, wage inequality.  But of course it’s bizarre to put this on a pay productivity graph, as it has nothing to do with productivity and pay.  It’s all about inequality of pay among workers.  I’ll conclude with another Rognlie comment that points to the subtle sleight of hand involved in putting this on a pay/productivity graph; it makes readers think that low-skilled workers are not being compensated for their productivity, without actually providing any evidence that their productivity rose as fast as high-skilled workers.  Here’s Matt:

So… as the lengthy digressions above demonstrate, neither the “loss in labor’s share” nor the “terms-of-trade” wedges in the widely circulated EPI graph are all that meaningful. The one remaining component that is unambiguously meaningful and important is the “inequality of compensation” gap, i.e. within-labor inequality.

Now, it’s not clear that this really belongs in a chart of the “productivity-pay” gap: the implicit assumption here is that there has been no change in mean vs. median labor productivity underlying the change in mean vs. median compensation. Needless to say, that’s a rather sketchy hypothesis – most economists would regard it as unlikely, and I don’t see much evidence for it.

It’s probably best to interpret the chart, therefore, as a clever way to rebrand plain old labor inequality: the idea that pay has failed to keep up with productivity sounds like more of a scandal than does “inequality” in the abstract. The scandal looks even bigger when you add the (dubious) terms-of-trade wedge on top.

Like Scott, I’m not moved much by this kind of rebranding, but I am a utilitarian who cares about distribution and individual welfare, so I am greatly concerned about the rise in within-labor inequality. That’s why I want to keep focused on that issue – and, with any luck, stop the endless tide of confusing productivity-pay charts.

Let’s end on a positive note.  Krugman, Rognlie and I all believe that wage inequality is the real issue.  Let’s stop with all this pay/productivity nonsense and focus on the real issue.  Here’s a great proposal to reduce pay inequality, without raising unemployment.



33 Responses to “Pay/productivity gap graphs are nonsense”

  1. Gravatar of benjamin cole benjamin cole
    11. September 2015 at 05:23

    The latest issue of The Economist magazine reports that two Fed researchers report US labor receiving a declining share of GDP….

  2. Gravatar of ssumner ssumner
    11. September 2015 at 05:55

    Be, That’s true in the last 10 years, but not the last 50 years (if you exclude depreciation, which it’s a no-brainer to exclude).

  3. Gravatar of Patrick R. Sullivan Patrick R. Sullivan
    11. September 2015 at 06:35

    From the NY Times piece by Oren Cass, to which Scott linked;

    ‘For example, if the target wage is $12 an hour, then a worker earning $8 an hour receives a $2 an hour subsidy, added by the government to the worker’s paycheck.’

    AKA, Friedman’s negative income tax.

  4. Gravatar of Ray Lopez Ray Lopez
    11. September 2015 at 07:34

    Another brilliant Sumner post, with an assist from Rognlie! It’s just getting better and better here… I’ve not much to add, except that inequality seems to increase, as Piketty says, during peaceful times, which seems to support the theory that prosperity causes economic growth causes inequality (standard Gaussian statistics as N-> infinity). Put another way: if everybody has to sacrifice for the common cause, said common cause being war, then everybody will be on a level playing field. But the tradeoff is everybody will be dressed in rags living off rations…yes, wartime economics beats depression, but at a cost.

  5. Gravatar of Mike Sax Mike Sax
    11. September 2015 at 08:59

    Wow Scott. I got a lot of posts to catch up on. You now may be the most prolific blogger on the Internet-other than me. LOL

  6. Gravatar of collin collin
    11. September 2015 at 09:09

    The more I read about the economy from 1948 – 1973 and the high inflation 1970s, the more I understand that the high wages of the 1950s – 1960s was simply a function of the low developed world labor supply of the era. We had a shortage of workers due to WW2, the depression baby bust and a workplace with defined sex and racial discrimination. By 1974, the economy could not afford high wages for the massively growing Baby Boom workforce and the huge increase of labor supply of declining discrimination. (Just think the annual number job increases of the Carter administration was higher than it was during the Reagan administration. And yes foreign policy against the Middle East did not help either.)

    That said now that we are the other end of the Boomer’s careers, I do believe job shortages will continues.

  7. Gravatar of ChrisA ChrisA
    11. September 2015 at 09:20

    The criticism that such a wage subsidy scheme would receive is that the Government would be subsidising low wage payers. This is the referenced reason that the UK government is raising the minimum wage. I am not convinced by this argument but personally I prefer an alternative solution – higher state pensions – which entirely avoids this issue. To me old age is the highest utilitarian time to provide additional cash to poor people, especially when most of them have difficulty in actually saving because of hyperbolic discounting.

  8. Gravatar of Patrick R. Sullivan Patrick R. Sullivan
    11. September 2015 at 10:11

    It didn’t take Larry Summers long to go from the sublime, to the ridiculous;

    ‘Few people believe that private unions in the United States had excessive power in the 1970s. Yet the share of private sector workers in unions has fallen from over 24 percent in 1973 to under 7 percent today. Unions are surely too weak today.’

    Speaking as someone who was a union member (for a few minutes) in the 1970s; Ha, ha, ha, ha, ha….

    ‘Their weakness most clearly shows up in workers who have difficulty maintaining a middle class lifestyle.’

    With their own computers and smart phones?

    ‘In a more profound way, the weakness of unions leaves a broad swath of the middle class largely unrepresented in the political process, paving the way for the kind of disillusionment that has driven the Trump candidacy.’

    Then how can you claim they are ‘unrepresented in the political process’?

    ‘After all, no society is going to remain stable and confident in its central institutions if parents do not believe their kids can lead better lives than they did.’

    Larry has a Harvard Phd.

    ‘In 2009, at the Brookings Institution, I spoke about the important role of unions in generating broadly shared growth.’

    I guess Larry hasn’t talked to Sonia Sotomayor lately. I.e., ‘Unions EXCLUDE people. It’s not nice, but it’s what they do.’

    ‘I said, “If we want to propel this economy forward and we want to have a sound expansion, it has to be an expansion whose benefits are more broadly shared. And that goes to the question of tax policy and progressivity, it goes to the question of education over the longer term, and it goes to the question of having a healthy and well-functioning trade union movement.”’

    Like in France?

  9. Gravatar of collin collin
    11. September 2015 at 10:45

    One aspect of the the growth and fall of Unions is why did they gain such power in the 1950s even after Taft-Hartley? I assumed that companies were willing to deal with unions with the labor shortages of the 1950s that they had get employees to accept lower wages.

  10. Gravatar of ssumner ssumner
    11. September 2015 at 11:46

    Patrick, Actually even better than the NIT. You have to work to get the benefits, and the more you work the more benefits you get.

    Chris, I’m familiar with that argument but I don’t buy it. This proposal is far superior to a high minimum wage, which will lead to higher unemployment. Then you need welfare for the unemployed. I believe this is similar to what the Germans did, and it was a huge success in Germany.

    I can’t even imagine why someone would object to the government subsidizing firms that hire low skilled workers, unless they were a libertarian purist. But in that case why advocate a minimum wage? It makes no sense.

    Patrick, That’s pretty funny that Summers didn’t recall that people thought unions were too powerful in the 1970s. Wage controls were installed, precisely because even Democratic economists thought unions were too powerful.

  11. Gravatar of Justin D Justin D
    11. September 2015 at 11:57

    I second ChrisA’s point – the political left is going to see such a program as a reward to those bad companies which have been mistreating workers all along.

    Additionally, how much does the wage subsidy really do to remedy inequality?

    In the article, the target wage is $12, so a worker earning $8/hr will be brought up to $10/hr. A good percentage increase for the worker, to be sure, but pretty de minimis from an inequality perspective. It’s like climbing a hill and saying that you’re closer to the moon. Technically true, though also trivially true.

    Can someone explain why adding adding a wage subsidy is better than just expanding the EITC?

    And why concede that income inequality is an issue worth thinking about at all? Shouldn’t the real concern be whether people have basic needs? People are unequal in an enormous variety of ways – attractiveness, health, intelligence, confidence, etc., and some of these are more important than income. I’d gladly trade places with a person earning 20% of my income but in perfect health.

  12. Gravatar of benjamin cole benjamin cole
    11. September 2015 at 15:29

    This FRED chart indicates a decline in labor share of nonfarm income…at least I think…goes back to 1947

  13. Gravatar of E. Harding E. Harding
    11. September 2015 at 15:31

    But in that case why advocate a minimum wage?

    -Because wage subsidies use incentives and help the expropriators. As all income inequality is unjust, it is only fair to expropriate the expropriators, and wholly unfair to subsidize them. Have you looked in the NYT comments? They’re all along that vein.

  14. Gravatar of Ray Lopez Ray Lopez
    11. September 2015 at 15:49

    @Patrick R. Sullivan – nice spot on Summers’ ridiculousness advocating for unions. Next he’ll be asking for a EU-style Fortress USA. I think Summers is angling for a spot on the B. Sanders team–I’m serious. These guys use national media as a sort of broadcast email to potential employers. At least our own Scott Sumner has no hidden agenda and tells it like it is. He’s not beholden to some vested interest (Duda aside, but Duda is for the common good apparently).

  15. Gravatar of ChrisA ChrisA
    11. September 2015 at 20:32

    Scott – as I said I don’t agree that wage subsidies are a reward for low paying employers, just that a lot of other people see it that way. Osborne’s solution of raising the minimum wage is popular for that reason. Ideas have to be politically sellable as well economically, which is why my alternative solution, higher old age state pensions, is to me a better solution to inequality in earnings. Old age pensions are very popular especially among those who vote. And there is a grain of truth in the idea that if you get the same wage regardless of what you do, why bother to seek to upgrade your skills or look for more rewarding work? Whereas an old age pension recognises by the time you get to be old all essential interventions whether personal or by society are moot. You truly are a creature of fate and thus there is no moral hazard of society giving you unearned compensation.

    Interestingly the hike in the minimum wage in the UK is also seen as a way to attack immigration. The theory being that low wage jobs are only attractive to immigrants (legal or otherwise) as the locals have higher reservation wages or better options. So for instance you see largely Eastern European immigrants working in the hospitality industry in the UK. Make those jobs go away with the new minimum wage and of course you will see less immigrants. This to me is the most cynical politics ever – essentially destroy high visible jobs so voters don’t get upset by immigration.

  16. Gravatar of James Alexander James Alexander
    11. September 2015 at 22:02

    Scott, Benjamin

    It is possible the share of income to “labor”, shown in the labor share, has fallen. But the profits share also get shared with “labor” if they have 401k mutual fund or other equity investments.

  17. Gravatar of benjamin cole benjamin cole
    11. September 2015 at 23:55

    James: maybe so. It is interesting to ponder why the Fed has a hard numerical target for inflation but not a hard numerical target for unemployment.

    Imagine if the Fed instead targeted a 3%
    unemployment ceiling, but had a squishy zone for an inflation target.

    Imagine if at Jackson Hole there were six panel discussions all on unemployment and not one on inflation.

  18. Gravatar of Anand Anand
    12. September 2015 at 04:33

    This is all fine. But this was already recognized many years ago, that a direct comparison is problematic. I wonder what you and your readers think of this short analysis by Dean Baker, which Krugman cites elsewhere. It adjusts for various things:

  19. Gravatar of Orn Gudmundsson, Jr. Orn Gudmundsson, Jr.
    12. September 2015 at 04:34


    Making the assumption that we are talking about both salary and hourly wages as components, aren’t wages over a certain amount are simply renting intellectual assets and talent rather than time?

    Since you previously stated you aren’t concerned with inequality in income, just wage inequality, can you please elaborate? From my understanding of a utilitarian perspective, I can’t see why you would be concerned with wage inequality (a poor proxy for standard of living in a country as diverse, geographically and demographically, as the U.S.) but more, I don’t see why renting someone’s talent (which often requires a heavy cash investment for them to obtain) is any different than renting their apartment.

  20. Gravatar of ssumner ssumner
    12. September 2015 at 04:48

    Justin, I agree that those other forms of inequality are much bigger problems, but that doesn’t mean wage inequality is not a problem.

    As far as the left hating the program, there’s nothing I can do about that. But the EITC subsidizes low pay companies just as much as this proposal, and lots of Democrats support the policy. The opposition from the left didn’t stop the EITC from being enacted.

    You said:

    “In the article, the target wage is $12, so a worker earning $8/hr will be brought up to $10/hr. A good percentage increase for the worker, to be sure, but pretty de minimis from an inequality perspective. It’s like climbing a hill and saying that you’re closer to the moon. Technically true, though also trivially true.

    Can someone explain why adding adding a wage subsidy is better than just expanding the EITC?”

    You are presumably not a low wage worker. If you are trying to live on $8/hour, then $10 would be a really nice change. And of course the numbers are just illustrative. You could make $14/hour the target, but the program would be much more expensive. Also, low wage workers qualify for programs like food stamps and Medicaid, so their actual income is higher than this suggests.

    I see two advantages over the EITC. Less fraud and it gives more incentive to work hard. Under EITC workers who decide to work overtime often lose benefits. Under this plan they would always gain benefits by working more hours.

    Chris, I’m told the new German minimum wage was also aimed at stopping immigration. Of course it will only do that if it “fails” i.e. if it causes lots of unemployment (which its supporters insist it won’t do.) If it doesn’t create more unemployment, it will be an even greater magnet for immigration.

    In America, the over 65 age group are already better off than the under 65.

    On subsidizing companies, see my response to Justin.

    James, That’s true, and labor also gains from that fact that a growing part of “profits” is the implicit rent they earn on their own owner-occupied housing. In fact, labor’s share is unchanged over the past 50 years, right at 68% of national income.

  21. Gravatar of ssumner ssumner
    12. September 2015 at 05:02

    Anand, What aspect of that do you want me to comment on?


    Income is a meaningless indicator of economic wellbeing for all the standard public finance reasons. If you haven’t studied public finance, then check out this post:

    And yes, human capital has some similarities to physical capital. But since the production of human capital is already highly subsidized, I don’t have a big problem with taxing the returns on human capital. Also, a big part of college is consumption, not investment. It’s mostly about signaling. Studies show that if Harvard students are offered the option of going there for free for 4 years, but only getting the education, no diploma, they are not interested.

  22. Gravatar of Patrick R. Sullivan Patrick R. Sullivan
    12. September 2015 at 07:07

    ‘Wage controls were installed, precisely because even Democratic economists thought unions were too powerful.’

    And we have no less an authority than Fed Chairman Arthur Burns, in congressional testimony, back in that day, making the claim.

  23. Gravatar of Patrick R. Sullivan Patrick R. Sullivan
    12. September 2015 at 07:20

    Here’s Robert Hetzel in 1998 quoting Arthur Burns from the time period (BOG meeting in November 1970) Larry Summers is talking about in his blog post;

    ‘…prospects were dim for any easing of the cost-push inflation [**]generated by union demands[**]. However, the Federal Reserve could not do anything about those influences except to impose monetary restraint, and he did not believe the country was willing to accept for any long period an unemployment rate in the area of 6 percent. …. he did not believe that the Federal Reserve should be expected to cope with inflation single-handedly. The only effective answer, in his opinion, lay in some form of incomes policy.’

    Which is pretty much what the consensus of economic polite society was then.

  24. Gravatar of James Alexander James Alexander
    12. September 2015 at 07:58

    “Imagine” is that dreamy, utopian, song of John Lennon’s. It always seems a more appropriate anthem for socialists than the Red Flag.

    Short of our MM dream coming true it is still excellent to see so many comments out there these days about the minuscule costs of inflation versus risking another bout of 2008-09 medicine. Hopefully, it will have some impact on the central bankers.

  25. Gravatar of collin collin
    12. September 2015 at 08:04


    The fall of unions is well known and deserved. I think somebody who is against private and public unions has to explain why unions were so popular and gaining membership and power during the 1950s. Even after a lot of post-WW2 anti-union movement and passage of Taft-Hartley in 1947. I am guessing Unions gained strength from the short labor supply of the early 1950s and companies gave union and benefits to win gain employment while controlling wages. (Companies liked giving pension benefits instead of wage negotiations.)

    Considering, the US unemployment rate is 5.1% and labor supply does not seem to growing, we are starting to see a skilled labor shortage. (Also, jobless benefits are at exceptionally low.) So how are companies going to gain good skilled employees without wage increases?

  26. Gravatar of Anand Anand
    12. September 2015 at 08:58

    The point of the study was simply that if one adjusts the data for various things (like using median instead of average, to incorporate intra-labour inequalities), and several other adjustments mentioned in the analysis, there still exists a smaller but significant gap. I think the gap may still arise from the choice of deflator: I am not sure if this will still be true if one uses a different one. Also the comparison is from 1973 onwards: if one uses 1965 perhaps the gap will disappear.

    I agree that intra-labour inequality is a large part of the story.

  27. Gravatar of Steve F Steve F
    12. September 2015 at 09:44

    Scott, I can’t piece together what you mean by wage inequality being a problem, especially since you claim income inequality is meaningless. To the lay (I’m lay), the two don’t sound like they should play by different rules. Furthermore, I would think that low wages are the problem as opposed to equality, just like how poverty is the problem as opposed to income inequality. But even then, wages are supposedly set by the market, so when we see a greater rate of divergence in incomes of Tom Cruise and GM workers, I don’t see how this suggests inequality is the culprit.

    Please explain to me what is keeping me from understanding why you say wage inequality is a problem.

  28. Gravatar of Matt Rognlie Matt Rognlie
    12. September 2015 at 23:00

    benjamin cole,

    Yes, the graph you’re linking from fred is the BLS LPC program’s gross labor share series for the “nonfarm business sector”. This is the most commonly cited labor share measure in most popular discussion – but it’s important to recognize that it’s only one such measure, and provides a rather incomplete picture.

    Most important, it’s a gross labor share measure, meaning that the denominator includes depreciation. This has led to a downward drift over time in large part due to the rising role of software and other very short-lived capital, for which depreciation comprises most of gross value added. If you transform this measure by replacing the denominator with net value added rather than gross value added, you get a series that looks surprisingly stationary over time – with the most notable exception being the recent (possibly cyclical) dip:

    Even then, there are serious difficulties having to do with the treatment of proprietors’ income, which has both a labor component (which is not directly measured) and a capital component. When computing the labor share, the BLS has to make a very special imputation: namely, that proprietors’ labor implicitly earns the same average compensation per hour as the rest of the workforce. This imputation is nice in some respects – for instance, it means that proprietors leave the compensation per hour figures, which are part of the same data release, unaffected. But it obviously isn’t a great way to measure the actual labor contribution from proprietors, and there a number of papers showing how this approach has interacted with changes in the composition of proprietors to produce a downward bias in the labor share.

    To give some sense of the potential magnitudes at stake here, if we ignore the implicit labor component of proprietors’ income altogether and just count explicit compensation as part of the business sector’s labor share, we get the purple, green, and yellow lines in the following graph:

    [Some needlessly technical detail on why I am including three lines: the purple line most straightforwardly replaces the numerator of the labor share with only labor compensation. The green line changes the denominator from the one that the BLS uses – which is “business sector value added” as determined from the product side of NIPA – and replaces it with the equivalent measure obtained from the income side. This is to make the numerator and denominator more consistent: horrifyingly enough, as the labor share is currently calculated, the negative NIPA “statistical discrepancy” pushes up the labor share by a nontrivial amount. The yellow line further subtracts product taxes and transfer payments from the denominator, leaving us with only “factor income”; this is probably the best measure.

    My point here is not so much to revel in these abstruse distinctions as to emphasize just how much tedious methodological details can matter: the headline labor share that everyone cites is not the only choice, and indeed not a very good one. (As I mentioned in comments on the previous post, I prefer avoiding the impossible proprietors’ income issue altogether by looking at the labor share in the non-proprietor sectors of the economy…)]

  29. Gravatar of ssumner ssumner
    13. September 2015 at 04:46

    Patrick, Yes, I remember all that.

    Anand, I’m not sure the point of your comment. Simply talking about “the point of the study” does no good unless you address Matt’s critique. He demolished the study. The study is wrong, so it doesn’t matter what the point is, there is no pay productivity gap over the past 50 years. If they are interested in inequality, they should have talked about inequality, not pay/productivity gaps.

    Steve, Income data is meaningless because it adds labor and capital income, which means you are basically counting the same resources twice. If you don’t understand why, this post might help:

    And yes, inequality is mostly a problem at the low end. My only point about the rich is that we can tax a dollar from them at less cost than from the middle class, dollars we will use to help the poor.

  30. Gravatar of ssumner ssumner
    13. September 2015 at 04:55

    Matt, Thanks for that info. I’m interested in the huge gap between the red line, and the three adjusted lines (purple, green, yellow.) What’s the intuition here? Is it mostly just the removal of imputed labor income in proprietor’s income? If so, what’s going on there? For instance, there’s been a big decline in the number of self-employed farmers since 1948—does that factor into these trends? (I’m not even sure whether they are part of the business sector.)

  31. Gravatar of Matt Rognlie Matt Rognlie
    13. September 2015 at 10:59

    Scott, yes, it’s just the removal of imputed labor income – a surprisingly big deal.

    Some of it, but not all, is indeed due to farms. (The “business sector” stats shown in those graphs do include farms; I would have liked to do the more popular “nonfarm business sector”, but the NIPA table needed to calculate that isn’t on fred, so it couldn’t be a nice, clickable graph.) In 1947, farms’ net operating surplus was a bit less than 8% of business sector net value added; today, it is a bit less than 1%.

    Furthermore, backing out the BLS’s imputed labor compensation in proprietor’s income from a mix of BLS and BEA data, it looks like in 1947, about two-thirds of farm net operating surplus was imputed as labor income. Today, only about one-sixth of the much smaller total is imputed as labor income; hence the farm imputation added something like 5 percentage points more to the labor sector back then than it does today. This is a pretty huge deal, and it is responsible for a decent chunk of the red vs. (purple,green,yellow) discrepancy in my graph.

    But changes in the imputation of nonfarm proprietors’ income have also played an important role. In 1948 (the necessary table doesn’t go back to 1947), nonfarm noncorporate NOS was over 14% of total business sector factor income, and it looks like fully 69% of it was imputed to be labor income by the BLS. Today, nonfarm noncorporate NOS is even a slightly higher share, 15%, of total business sector factor income, but only 38% of it is imputed to be labor income by the BLS. Obviously, the fall from imputing 69% to imputing 38% has led to some downward movement in the labor share: indeed, (69%-38%)*15% = 4.65 percentage points, which is more than enough to account for the long-term decline in the net labor share.

    The natural question is then: is this fall in the fraction of proprietors’ income that is imputed as labor income justified? Probably not – as I mentioned above, it isn’t the result of a deliberate attempt by the BLS to come up with the best possible estimate of the true labor component, but instead the result of an assumption that proprietors have the same implicit compensation per hour as the rest of the workforce. Back in the 40s and 50s, when many proprietors ran small, low-key farms or businesses (often making far less than unionized workers in the formal sector), this estimate may have even erred on the high side in estimating their effective compensation. Today, when many proprietors are actually professionals of some stripe – and farmers, although still mostly families, run much larger and more sophisticated enterprises – this assumption almost certainly lowballs their labor contribution.

    With this bias in both directions, the fall is surely exaggerated.

    Of course, the rise of wealthy, professional sole proprietorships and partnerships providing legal, health, financial, etc. services is itself an instance of inequality at the high end. So this definitely doesn’t mean that inequality itself is overstated; just that the (already small) component coming from labor vs. capital is commonly exaggerated, and that it’s much more of a labor vs. labor affair.

  32. Gravatar of Matt Rognlie Matt Rognlie
    13. September 2015 at 11:16

    Anand, I’ve seen the 2007 Baker study before, and it is best interpreted as a precursor to the recent (and, with many years of incremental development, somewhat more polished) Bivens and Mishel EPI study.

    Once Baker makes all his adjustments, in Figure 5, he’s left with something very similar to the current EPI exercise: he’s comparing net productivity for the entire economy to a measure of median hourly compensation. This figure, by using the same deflator for both productivity and compensation, excludes the “terms of trade” gap that I lengthily ranted about in the previous post. As such, it’s left with the two other components of the gap between net productivity and median compensation growth: (1) the gap between median and mean compensation, and (2) the gap between mean compensation and productivity (i.e. changes in the labor share).

    I’ve also argued, along with Scott, that the role of (2) – already quite small – is exaggerated for several reasons: it’s made to look bigger by choosing 1973 as the start date, and also problems with the classification of proprietor’s income have led to some bias in the trend. To some extent, this bias actually means that compensation inequality (1) is being mislabeled as a falling labor share (2).

    Hence, in Baker’s study just like the recent EPI one, the “gap” that remains after appropriate analysis and adjustments is overwhelmingly driven by compensation inequality within labor. Scott’s and my point (which has also been made by other critics) is that this isn’t really a “productivity-pay” gap, since we have no idea whether productivity inequality has mirrored compensation inequality or not.

    Really, these graphs are just a circuitous restatement of the fact that within-labor inequality is a serious and large-scale problem, and that it’s driven a sizable wedge between the mean and median. It’s a branding exercise that makes sense politically, but isn’t helpful analytically.

  33. Gravatar of ssumner ssumner
    14. September 2015 at 05:37

    Thanks Matt, I find that to be extremely helpful.

Leave a Reply