Skip to content

What happened to productivity after the Great Recession?

The Great Recession produced a sudden and permanent reduction in the growth rate of labor productivity:

Productivity today is 20% lower than it would have been if it had kept growing at its rate of the previous few decades.

But what was the cause of the slowdown? The recession itself? Our tepid response to it? Or was the old growth rate a bubble that we were simply never going to be able to keep up forever?

Whatever the answer, it left a vast amount of wealth on the table. If the old growth rate had continued, median household income would probably be cracking six figures this year instead of just the $80,00 we actually ended up with.

55 thoughts on “What happened to productivity after the Great Recession?

  1. FrankM

    This is what happens when you create an extrapolation line based on some quadratic fit of a portion of the data. The line is meaningless, so the "gap" is likewise meaningless. I can just as easily create a half-dozen empirical lines going all over the graph. Some would undoubtedly show current productivity above the extrapolated line. On what basis would you choose the one you've drawn? If I'd presented a graph like this when I was working I'd probably have been summarily fired.

    There's an ironclad rule for extrapolating data: never extrapolate based on an empirical model. If you don't know what factors go into the model, you can't know whether they will hold in the future.

    1. illilillili

      Wowza. What a BS comment. It's perfectly legit to point out that over extended time period X something was happening and over exended time period Y something different was happening, and to then ask for opinions as to what factors might explain the difference.

      1. FrankM

        Drawing lines is nothing but artwork. If it's an empirical line there's no underlying meaning behind it. Just because you draw a couple of different lines doesn't mean anything different is happening.

        I could just as easily draw one line connecting the 1980-2000 portion with the 2015 and later portion, which would imply the period between 2000 and 2015 was the anomaly. Or I could draw dozens different lines leading to dozens of different conclusions. Or you could first decide on the conclusion and draw a line to support that.

        Modelling data is an art that involves more than just drawing lines.

        1. Jasper_in_Boston

          Drawing lines is nothing but artwork. If it's an empirical line there's no underlying meaning behind it.

          Seems likely Kevin's line is based on actual statistics. You do realize productivity numbers can be looked up. Right?

          1. bouncing_b

            Seems likely Kevin's line is based on actual statistics.

            Would have been more convincing if the method or rationale behind those statistics had been stated.

            It looks to me like a short section of slope pre-2006 has been extrapolated to suggest an expected growth rate larger than any other period in this record.

            Sorry, I’m with FrankM here. You need a good reason why that slope should be expected, otherwise it contains no real information. Artwork indeed.

        2. ScentOfViolets

          I think that's exactly what FrankM is saying; that it's all down to data points you choose to include for your curve fitting.

        1. jahoosafat

          Sorry, folks. FrankM's analysis isn't out there at all. The statistics behind his point are basic and uncontroversial. Kevin is great on a lot of things/topics, but his tendency to draw simple regression lines on scatterplot data - especially time series data - is often problematic. In this case, the quadratic fit line is a very subjective choice with no theory behind it. If he has an explicit defense/justification for this, then he should elaborate. Otherwise, there's no there there.

          1. FrankM

            Kevin's model quiver has but two arrows: linear and quadratic. If it looks straight it's linear. If it looks curved it's quadratic. Lots of data that appears straight is not linear, and there are many, many models for curved data besides quadratic.

            Empirical models that fit well can be useful for interpolation. I've done that lots of times. But never, ever, extrapolate an empirical model.

            1. ScentOfViolets

              Yeppers, you can do whatever degree of polynomial you like (well, there is an upper bound imposed by the amount of data you have.) It can be exponential, trigonometric, whatever function you care to model with; in the end, it's _all_ linear regression. For those of you who say waitaminute, a cubic certainly isn't linear, well you're not alone in making that objection. But you see, it's not the function itself, but the _regressors_ that are linear. For example, a cubic can be thought of as a linear combinations of the form a + bx + cx^2 + dx^3 where a, b, c and d are the coefficients. Sorry for that bit of um digression there.

        2. glipsnort

          I thought it was spot on. I do lots of statistical analysis, and I would never just pick an arbitrary stretch of time series data, fit an arbitrary function to it, and assume extrapolations from that curve meant anything.

      2. DonRolph

        So let's use the eyeball curve running up to 2006. If you extrapolate ahead it lines up about where we are now.

        Unless a sensitivity analysis is performed it is impossible with the data given to which extrapolation is more reasonable.

    2. lower-case

      i agree; looks to me like there was a temporary increase in the early 2000's then returning to trend

      businesses taking advantage of the internet in the aughts would be a reasonable guess

    3. skeptonomist

      What, Kevin is not allowed artistic license? He takes it anyway with his segments (sometimes non-linear) arbitrarily breaking at various points.

      Economists actually claim to have semi-theoretical models in order to make their extrapolations (such as the CBO projections), but the predictions from them are almost always wrong, so just ignore Kevin's lines if they are not to your taste.

      The main problem in the US economy is that most of the productivity increase has gone to the uppermost incomes. Real production worker wages (by the CPI) have not yet risen above the 1972 level. There is nothing about the current economy that suggests this will change soon. Finance and banking keep taking a larger share of the economy and profits have increased over the last few years (that's where a lot of the productivity has gone).

  2. illilillili

    I always vote for "our tepid response". But peering at the bumps in the graph, it looks like we recovered to trend in 2010 and then failed to keep up.

    1. Joseph Harbin

      Yep. I think 2010 was the pivot year (more than 2006).

      What happened? The GOP took Congress and we were destined for a decade of austerity. Fiscal stimulus was nil. (Even Obama/Dem stimulus in '09 was too small.) Emergency Fed policy (zero rates, rounds of QE) was the single instrument for assisting economic growth. Productivity growth (along with other things) lagged.

      George Bush left the White House just in time for Republicans. If he had stayed on through 2010, the election would have come after the recession bottomed. Dems would have have won in '10 with an historic mandate. Instead, Obama won in '08 and paid the political price for being prez while the economy crashed after the 2008 fiscal crisis. If Obama won in '10, after the real pain was felt, he may have become the next FDR. No such luck. The tilt toward fiscal policy had to wait for the pandemic, and Biden has pushed the pendulum toward a long overdue bigger role for government than many expected.

      ETA:
      The timing of elections matter. I imagine if FDR had been elected in 1930, he'd have presided over the economy as it slid into the Great Depression. Democrats would have paid. The New Deal may never have happened. GOP might have followed with a landslide, and if so, we may slid into fascism then with other natioins.

  3. Jasper_in_Boston

    Virtually everything I've seen on this topic suggests US productivity is doing very well indeed, post-pandemic, compared to other high income countries. Which means they're doing even worse. Which suggests the slowdown is a global phenomenon.

  4. Jasper_in_Boston

    Just for kicks, I ran Kevin's idea's past Claude (Anthropic). Here's what this LLM had to say:

    The observation of slower productivity growth post-pandemic was a topic of concern among economists and policymakers. Several factors were proposed as potential causes for this trend:

    1. Disruptions to supply chains: The pandemic caused significant disruptions to global supply chains, which may have led to inefficiencies in production processes.

    2. Labor market shifts: The "Great Resignation" and changes in work patterns (like the shift to remote work) may have temporarily reduced productivity as businesses and workers adapted to new arrangements.

    3. Skill mismatches: Rapid changes in the job market may have led to mismatches between available skills and job requirements, potentially reducing overall productivity.

    4. Reduced business investment: Uncertainty during and after the pandemic may have led to decreased business investment in productivity-enhancing technologies and processes.

    5. Sectoral shifts: Changes in the relative sizes of different economic sectors (e.g., services vs. manufacturing) could have impacted overall productivity measures.

    6. Measurement issues: Some economists argued that traditional productivity measures might not adequately capture the value created in a increasingly digital and service-oriented economy.

    7. Long-term effects of the pandemic: Factors such as long COVID, mental health issues, and changes in work-life balance priorities could have had lasting effects on workforce productivity.

    Claude was a bit more skeptical of Kevin's math with respect to household income, as am I.

    1. Justin

      Quite the load of nonsense word salad there. Everything causes everything! If that's the best AI can do, then it's no risk to anyone but the talking heads on CNBC.

    2. Joseph Harbin

      Looks like Claude has the wrong decade. Maybe LLMs are good at guessing the next word but still can't tell you what year it is.

      "Slower productivity growth post-pandemic" is not really the issue. Kevin's chart shows the slowdown coming post-GFC (really, post-2010).

      The pandemic created a burst of volatility in productivity; first a spike up in 2020, then down into 2022. But since mid-'22 (i.e., "post-pandemic"), productivity has grown rapidly. Growth has now been 2.7% or better for 3 straight quarters, better than almost the entire last decade.

      Details here:
      https://fred.stlouisfed.org/series/PRS84006091

      Best recipe for continued growth is a Dem sweep in the election, or we could see hostage-taking and austerity, or even much worse.

  5. Ken Rhodes

    Ignore all the hoo-rah about drawing lines—what the lines mean, and why they mean anything at all. Kevin draws lines because he thinks they’re easier to visualize than a page full of numbers. Meanwhile, though, the numbers are the *facts* which are the subject, and the question is valid:

    “Look at these facts. There seems to have been something about a certain time when the facts suggest something changed in the environment that produced these facts. What do you suppose changed? Can we identify it? Can we quantify it?

    My guess is that there was a very real, very important, and somewhat temporary change in the environment during the twenty year period from 1980-2000. That was the revolution in computer usage. In the seventies, computers were the tools of a few—the scientists and engineers. They were big, expensive, and required a support staff. By the turn of the century, it was a vastly different world. And in my recollection, the turn of the century was (approximately) when the revolution ended and the environment (in that particular regard) returned to evolution.

    If my hypothesis is correct, then it would be important to examine the data (or view the line, if you like) including the prior decades. Start the sample at 1960 and see if the specific time from 1980-2000 is the exception to the historical trend.

    1. cmayo

      Yeah, this seems right to me. Once the workplace was sufficiently computerized/saturated with workstations/PCs, the boost to productivity provided by these new tools tapered off.

    2. FrankM

      The difficulty is in determining what is "normal". Worker productivity is extremely complex. Lots of things change all the time. Trying to identify which factors are responsible for what seems to me to be a rather small effect in an uncontrolled experiment is well-nigh impossible. Which means people can debate it forever.

  6. Austin

    Nothing can grow at the same rate forever. Perhaps humans are reaching a plateau in productivity until the next big technological leap forward?

    And piggybacking off of what Ken Rhodes said: it’s quite possible the slope of the line changed right at about 1980 too with the previous tech leap of widespread computer usage. We’d have to see more of the slope pre-1980 to know if the extrapolated line is reasonable.

  7. geordie

    When charting growth, the line will eventually approach horizontal because no more efficiency gains are possible. As an example better routing allows Amazon to increase the number of deliveries per driver but at some point the speed limits on the roads are the limiting factor.

    1. aldoushickman

      "When charting growth, the line will eventually approach horizontal because no more efficiency gains are possible"

      Well, sure, someday we will asymptotically approach perfect efficiency. And someday the sun will run out of hydrogen and go red giant on us. But since it's pretty plain that we are nowhere near perfect efficiency, it's not bumping up against the base-level laws of physics that caused whatever discrepancy Kevin is identifying around 2008.

        1. aldoushickman

          yes, yes, but that's in the same category as, in response to somebody wondering if something is up because the oven is taking longer than usual to preheat, sagely noting that in the long run the oven cannot preheat because everything will reach thermal equilibrium with its environment b/c thermodynamics. It's a pedantic argument that has nothing to do with the phenomenon being observed.

    2. IncorrigibleTroll

      I disagree that the speed limits will be a limiting factor. I watched an Amazon truck on the little map travel over a mile in about 45 seconds on a 35mph road.

      The horsepower of the trucks will be the limiting factor.

  8. golack

    This is for non-financial services, yet over this time, financial services, as percent of GDP, has grown a lot. Maybe too much for a healthy economy, but still, that factors into household income.

    That said, the apparent growth under Bush (the son) was a bit of a bubble. Wars and deficit spending fed it, as did the "irrational exuberance" of the stock market. That burst with the Great Recession, and McConnell wanted to keep Obama a "one-term president", so blocked needed stimulus when he could--depressing growth. Trump got into office--and never really helped non-financial part of economy, but did feed into a financial bubble--burst by the pandemic. And Biden, well, had to clean up that mess.

  9. cmayo

    Why are we assuming that households would capture that productivity growth in terms of income increases? Why would we assume that household income growth would be any higher than it is right now, and not assume that the ownership class would keep all of that productivity growth for itself? Because the evidence certainly suggests the latter rather than the former.

    Also, the answer is probably computers and it's not the post-Recession period that's special/different, but the higher trendline prior to that.

  10. azumbrunn

    The median income's rise would depend on what percentage of the gains would trickle down to lower classes. This is the "disadvantage" of using the median: It reflects the situation of the average Joe which the Wall Street Journal types would rather people didn't know.

  11. Adam Strange

    My guess as to why the measured productivity is slowing down would be that pollution is finally catching up with us.

    It takes an increasingly larger part of profits to extract the next barrel of oil.
    It takes an increasingly larger amount of resources to educate people to a higher level, and then they have shorter working lives.

    Eventually, when you have to invest a full dollar to get 50 cents worth of benefits, you are in negative productivity land and you need to simplify.

    Joseph Tainter said that this accounts for the collapse of complex societies, which overrun their resources and ignore their pollution.

    Now, if humans had access to a small, local black hole, we could convert matter to energy at 50% efficiency, and we could keep the band playing until we run out of matter to dump into it.

    1. emjayay

      And we are increasingly requiring that the externalities like pollution are paid for by the polluters or the government which has to collect more taxes to pay for it, and limited by the government which is not free for the polluters who have to install new equipment, do things in more expensive ways etc.

      OTOH government incentives for things like new lighting sources may end up saving money for those companies.

      Pollution used to be free to the polluters and only paid for in destroyed environments and people's health, including their workers.

    2. aldoushickman

      "It takes an increasingly larger part of profits to extract the next barrel of oil."

      Sorta, although the price of a barrel of oil is a more-or-less random walk and not out of line with historical prices. Now, price in dollars of course is not the same thing as the cost of extraction (in money or in effort), but it's not unrelated.

      The bigger thing, though, is the decoupling of economic growth--particularly in the US--from oil consumption. Example: US oil consumption has been flat for over two decades, despite US population increasing by 15% and GDP increasing by 122% over the same period.

      So, it's doubtful that any slowdown in productivity growth is because of increasing difficulty in getting oil.

  12. Boronx

    Seems like you're expecting a "Moore's Law" of productivity, but just like with Moore's law, the low hanging fruit get picked out. Improvement slows until the next big breakthrough.

    Edit: As I remember it, the great recession started in 2008, so maybe you got the causes reversed. Drooping productivity growth caused the failure of too optimistic, unbacked futures.

    1. aldoushickman

      "the great recession started in 2008, so maybe you got the causes reversed. Drooping productivity growth caused the failure of too optimistic, unbacked futures."

      The great recession was not caused by overly rosy assessments of productivity growth--it was a credit crunch brought on by financial institutions being overleveraged on risky mortage backed securities and credit default swaps.

  13. Jim B 55

    Could the reason be demographic? There a fewer young people. Young people tend to get up-to-date training when they enter the work force. More young people means more skilled workforce entrants, fewer less.

  14. Gilgit

    Not sure why two recent commenters mentioned income. That has nothing to do with the chart. I suppose they were better than the first people to comment who seem to have no idea how to read a chart, but talked at length about how you can cherry pick data even though Kevin showed 40+ years of data.

    Contrary to what some people have posted, Kevin's observation has been widely discussed for years among economists. I assume that China would be a likely culprit. With another likely choice being that more of the economy started going into the financial sector which was really a drain on everyone everywhere. I did recently read how manufacturing productivity is not growing, but only in the US. In the rest of the world it is still increasing. No one knows why this is either, but I remember one suggestion was that America was already productive and there wasn't much room to increase. I think another possibility was we use robots less than China and a few other countries.

    I do also remember reading a few years ago about how hard it was to retain lower wage tech workers in chip manufacturing plants. The problem wasn't even the pay. It was that they kept changing the work schedules - like they do with retail workers. Basically, the corporate higher ups couldn't be bothered to value the lower paid techs on the factory floor, so the company ran less efficiently. Genius!

    I guess I'll make up a scenario. With too much of our manufacturing moving overseas, America doesn't have enough critical mass to keep manufacturing productivity increasing. Less overall manufacturing means less investment, including in automation/robots. And diverting money and many of our most talented people into the financial sector means we can't catch up. Sounds as good as any other idea.

    I recently found out that the inflation reduction act really is causing a lot of new factories to be built. Maybe that will reverse the made up scenario mentioned above.

  15. jdubs

    Is this productivity stat/estimate actually a good measurement of inputs and outputs?

    Has it always been an extremely good measurement of this?

    The economy of the last 20+ years has very different mix of inputs and outputs than it did in the 80s and 90s. Is there any reason to think that productivity is a good (accurate) way to measure every single type of input and output?

    Often when empowered with a stat or measurement or metric, people forget what the point of this metric was and simply assume that the metric is the meaningful thing instead of merely a point in time guesstimate of some important idea or concept.

  16. emjayay

    Not doing any numbers but just stupidly sticking a ruler from the left end of that productivity line up to 2006, where Kevin thinks it deviated downward. The right end of the ruler is way closer to real productivity than Kevin's projection. It looks like he took an increasing change over a relatively short period and projected it into the future. As others have pointed out that period of more productivity increase is when computers/internet entered the picture.

    The real story about productivity is that at around the Reagan presidency the growth in wages that had been tracking productivity increases forever deviated from productivity growth and the money started to be sent only to higher incomes, and the higher the more.

  17. ProgressOne

    To see how productivity is doing, all that matters is the rate of change in recent years compared to the past rate of change. Some drops and humps now and then in the overall curve are expected as economic forces take effect (computerization, recessions, etc.).

    If you draw a straight line from 1980 to 2023, the productivity trend looks surprisingly linear. And looking specifically at the rate of change, 1980-2000 is about the same as for 2010-2023. This suggests the US is doing just fine for productivity growth.

  18. duncanmark

    I spent decades successfully improving productivity
    Most improvements were not because of automation or increase capital but just by "working smarter"
    My team of engineers would grab any good ideas we could find (or think of) and then implement them
    The engineers would contribute about half of the good ideas - the rest from the shop floor
    In all that time I can't think of a single good idea from the executives
    But almost all of the actual BENEFIT from those good ideas would flow to the executives
    I suspect the slowdown is because more and more of the people who actually did the work have realised that
    Which has had a bad effect on their motivation

Comments are closed.