What’s (not) up with productivity growth? A quick overview

August 22nd, 2017 at 11:35 am

For a long while now, I’ve been thinking and reading about the great productivity growth slowdown, so it seemed like a good time to give the lay of the land as I see it:

–Facts of the case, 1: As measured (as you’ll see, this caveat is important), since 2010, output per hour has been growing about 1% per year, half the growth rate of the long-term average. Slowdowns of similar magnitudes have occurred across most advanced economies.

Source: Furman, https://www.vox.com/the-big-idea/2017/3/21/14938698/growth-trump-economic-us-slowdown-demographics-stagnation

–Facts of the case, 2: The bulk of the slowdown is attributable to a decline in total factor productivity (TFP), or the growth in output when you take out all the measurable inputs. TFP is reasonably considered a proxy for innovation.

–The dreaded “empty hole problem:” Outside of accounting exercises that raise as many questions as they answer, economists do not understand the underlying forces that make productivity speed up and slow down. This creates the “empty hole problem:” since no one knows the answer, partisans fill the hole with their favorite candidate. E.G., here in DC “tax cuts and deregulation!” become the solution du jour.

–Optimists and (sort of) pessimists: When it comes to how lasting our plodding productivity growth rates will be, commentators fit roughly into pessimistic and optimistic camps. The pessimists are the larger group and, at least in my judgement, have better evidence. Their focus is on the slowdown in TFP, and for all the talk about it, no one really knows what drives innovation cycles. In that sense, “who knows?” is a subgroup among the pessimists wherein I place myself. The real pessimist caucus is chaired by the productivity expert Robert Gordon, who argues that the big-ticket productivity movers—e.g., electricity diffusion, air conditioning, indoor plumbing, air travel—are long behind us. Candy Crush is a fun, free diversion, but it ain’t a big efficiency play.

–What about mismeasurement? The optimists largely depend on mismeasurement and they bring some evidence to the table. Since we’re talking about growth rates, showing evidence of mismeasurement alone is not proof of anything. It must be shown that mismeasurement is getting worse, i.e., that we’re increasingly leaving out value added in our measures of real output. Some mismeasurement claims stem from the observation that sectors wherein it is harder for national accountants to pick up true declines in quality-adjusted prices—health care, software, the “app” economy—are the very sectors that are growing as a share of value added, meaning even constant mismeasurement in those sectors could lead to downward bias in measured output and thus productivity.

The biggest mismeasurement advocates are the typically hard-nosed economics team at Goldman Sachs. The figure below shows their portentous adjustments to output from significantly goosing the quality adjustments to IT hardware, software, and “free digital content.” Based on this work, they conclude that both GDP and productivity growth are understated by 1/4-1/2 percentage points, which is big in this biz.

Source: GS Research

However, I don’t find all their adjustments fully convincing. Careful research points out that we’re doing a better job than we used to measuring hardware and software, thus the productivity slowdown may be understated (in the US, we’re also producing less IT hardware). Other work finds that, yes, our price indices are missing tech improvements, such that TFP in that sector has hardly slowed at all. But this just implies that TFP outside of tech has decelerated even faster than we thought. Then there’s research showing that productivity is falling across many countries, and its decline is uncorrelated with their production of IT.

Also, a bunch of what’s allegedly being increasingly mismeasured – e.g., the value of software – are intermediate goods, meaning you’ve got to show the links in the chain such that final demand is increasingly biased down.

Wherein I fill the empty hole: Here are three explanations that make sense to me. First, some of the most interesting research in this space shows an historically unique divergence between the productivity growth of so-called “frontier” and “laggard” firms. Why has the latter failed to adopt the technologies of the former, and why hasn’t that failure led to their demise? This may be an important market failure.

Second, though the productivity slowdown predates the Great Recession, “secular stagnation” has been upon the land for quite a while now, and thus it might be a mistake to reject the hypothesis that weak demand is a factor. I can think of a simple, intuitive model wherein strong demand boosts unit labor costs, squeezing unit profits, such that maintaining profit margins means finding ways to produce more efficiently (this is the “full employment productivity multiplier” about which I’ve theorized). Third, the most accurate forecasts of productivity growth over the next few years require the use of very long—as in 40 years—autoregressive lags, so perhaps we will eventually mean-revert back to healthier productivity growth rates.

That last point is in the spirit of the most honest answer: who knows?

Print Friendly, PDF & Email

13 comments in reply to "What’s (not) up with productivity growth? A quick overview"

  1. spencer says:

    I make profits growth a function of the spread between unit labor cost and the price deflator for non-farm business.

    Over the years this has had a very tight fit with actual profits growth as measured by the BEA or by Standard & Poor.

    This implies that the productivity data that goes into the calculation of unit labor is fairly accurate and has not become more or less significant in recent years.

    • Smith says:

      I think there is a misinterpretation in your comment. GS studies are saying free services aren’t factored in, meaning they wouldn’t add to profits. Your tight fit is not inconsistent with GS claim productivity is underestimated (which I don’t buy).

  2. Smith says:

    I am aghast to think that growth or lack thereof is not considered, discussed, or assessed as the primary and perhaps obvious reason for slower productivity growth.
    Try FRED Gross Domestic Product (GDP)
    Units: Percent Change, Seasonally Adjusted Annual RateFrequency: Annual,Average
    Not only does the graph show 2/3 to 1/2 the average annual growth in GDP 2006 – 2016 compared to 1996 – 2006, but the latter period also includes the much longer and deeper 2008 – 2009 Great Recession, vs abbreviated (though jobless recovery) 2001 – 2002 recession.
    Those growth figures alone would tend to explain completely slow productivity growth. You don’t even need productivity gains when demand is weak.
    Not a new idea, and formally expressed in Verdoorn’s law of 1949.

    We could add a few other factors to boot, that would enhance the effect.
    1) Inequality grew dramatically in the Clinton boom years. This meant new growth didn’t fuel new demand as it did in the past, no virtuous cycle. You’ve shown many times that productivity growth didn’t accrue to most citizens. Money was saved instead of spent. Obama years also saw recovery of inequality levels.
    2) The 10 percent reduction in manufacturing employment was never recovered after the Great Recession. That means this sector had no reason to seek productivity improvements due to excess capacity and reserve army of labor.
    3) Executives, large corporations increase profits overseas, not here. If you relocate to cheaper labor environs, however small the gains ( labor accounts for small portion, often 10 to 20% of production costs), even when laid off workers find employment elsewhere, that doesn’t foster productivity gains. Just the opposite.

    With all these obvious reasons, “who knows” is not an acceptable answer. Who knows? Everyone who bothers to look.

  3. Bob Palmer says:

    The productivity slowdown is a puzzle. Retailing is becoming more efficient at the hands of innovators like Amazon and other online retailers. As a result goods are delivered to end users at lower overall cost than previously. While little value can be added in retail generally, retailing is such a large part of GDP that the positive effect of Amazon et al on productivity should be significant. But is it?

    Service and retail are low value added activities, and potential gains in productivity in them are correspondingly small. We are rapidly becoming a service-and-retail economy with a government devoted to health care and war making. Greater potential gains are, or were, in industrial activities that make up a decreasing portion of our activities. Perhaps industrial activities were already optimized to the point of diminishing returns, or were exported to low-wage nations. That leaves service and retail to carry the day. But it still doesn’t add up.

  4. John S. says:

    Perhaps an aging population has a negative impact on productivity growth.

    What is the effect on productivity of inadequate funding for public education and that impact on the labor force? What about the effect of the “opioid crisis” on productivity?

    It seems this issue of accurately measuring productivity and productivity growth is an opportunity for a dedicated and determine stellar team of economists to resolve. Let’s get with the program boys and girls!

  5. Serene says:

    I don’t know much about productivity, but I do hope somebody begins charting aggregate anxiety vs. labor productivity. We have a natural experiment in progress.

  6. Flex says:

    Maybe I shouldn’t comment, but since I said effectively the same thing to one of out VP’s in the company today, and I’ve finished most of a bottle of champagne (what can I say, I like the taste and I can afford it), I’ll throw this out there.

    I live and work down in the trenches. For 20+ years I’ve worked as a technician and then engineer in the automotive sector and I recently managed to get into front-line management. I enjoy studying economics as a sideline, but I recognize the limitations of the dismal science.

    The difficulty we’ve been having with productivity in our company is two-fold.

    1. Changes are coming too rapidly. There are companies who’s business plan is based on change. For example, we need to design circuit boards. There are many CAD tools to help us do that. But these tools are changing every 2-3 years. By the time the designers are familiar with one tool, it is no longer supported by the company who developed it, In order to get support we need to purchase an ‘updated’ tool which takes time to learn. I understand why software companies who build these tools are doing this, if the tool is good enough to use perpetually then these companies will be out of business. But practically it means that there is never enough time to learn to use a tool to it’s fullest capability before it is replaced. In the 1990’s the increase in functionality of these tools was enough to improve productivity. But these days the changes are severe enough to limit productivity gains.

    2. Support. Many companies in the automotive sector (which means probably other sectors are also affected), have decided that the creative teams, i.e. the engineers, can also manage all the support tasks. When I entered the field, more than 20 years ago, it was expected that an engineer would be able to keep 2-3 technicians busy all the time and there were a number of administrative assistants who would track requests, bird-dog drawing changes, manage schedules, and even do things like follow-up on patent submissions. Today I’m managing a team of 20 engineers and we have 3 technicians and no administrative support. My team spends far too much time on work that a technician could perform because there is no one else to do the work.

    The first hurdle to our productivity has come about clearly because the business model of too many companies who make development tools rely on software updates to continue to bring in business. I understand where they are coming from, but it hurts overall productivity gains. Rolling out a new tool takes time, and mistakes will be made. If a new, or updated, tool is required every year there will be a reduction in productivity simply because of the mistakes made while learning the tool. Currently several tools are being rolled out every year, all promising productivity improvements. Which never materialize because it will take 2-3 years to see those improvements, and there is never enough time to get those improvements before the next update is introduced.

    The second hurdle seems to be a combination of many things; including a reduction in “non-essential” employees (as if support is not essential) and the feeling that the tools provided make these employees unnecessary. There is a problem with an employee who is supposed to be spending their time performing engineering work but spends most of their time on time-sheets or fighting to get their travel expenses approved. We are informed that the new, improved software tool will solve that problem (once they learn how to use it), but it rarely is that easy. (Note: I spent 3 hours at work today activating a new Travel Credit Card, eventually calling both the corporate HQ and the issuing bank, and finding ways to speak to a person rather than a phone-tree, in order to sort out what should have been a 3 minute activation.)

    Don’t get me wrong. I’m a big supporter of tools which will reduce the workload of my team, and I’m looking at tools which will help me do so. But what I see is that changes are made to these tools in order to keep the tool-maker in business, I can’t introduce the tool fast enough before it becomes unsupported.

    I also see that executives are willing to hire very bright STEM graduates but think that hiring the support needed to help them be productive is unnecessary. Why this is the case, I don’t know. If I was to guess, it would be because there are few people in management positions who started by performing the support tasks, or even were in the positions where support personnel were important.

    Does this explain the falling productivity numbers? Probably not entirely. But that’s my $0.02.

    • Smith says:

      Yes, this rings true. I would judge your observations shrewd and insightful, but I would offer some alternative or additional explanations. I have firsthand experience in tech, which has included everything from evaluating tools, creating tools, using tools, watching others figure out how to use new tools There are just a few inconsistencies in your logic, if one is only skeptical enough to envision a less ordered less meaningful purposeful universe.
      Although it’s possible that constant training for new tools erases any advantage the new tool provides, that would imply any pause in tool introduction would bring about a quantum leap in productivity as all the accumulated advantages countless improvements are given a moment to surface.
      The alternative and more plausible explanation is that the new tools offer little to no advantage. Look at Microsoft office for example. Very little useful functionality has been added in over 20 years. Yet every time they change the interface, it sets back basic office productivity. If you are fortunate enough to use a Macintosh computer, office retains it’s menu system. On a PC, you’re stuck with the ribbon of icons.
      Likewise in operating systems, Microsoft stumbled so badly they had restore Classic Windows interface to their latest release. It’s a New Coke Classic Coke marketing gaffe.
      Apple IOS upgrades got so bad a few years ago, customers threatened to sue, bloated software seemed designed more to make their phones obsolete and encourage new sales vs offering new functionality.
      New software that sports fins and chrome and 12 cylinder cpu’s with more horsepower is nothing new to American marketing. The new information highway invites more traffic than your daily commute on the interstate, emails in your in and outbox do not enhance productivity.
      Although you seem to find a lack of support personnel is a problem, several issues interact with each other. Eliminating support gets rid of hourly employees, and their work can be picked up by higher paid salaried employees who are exempt. This is tied to the erosion and elimination of middle income jobs so we are left with only high and low income work, more inequality especially for lower income workers. Ironically also software tools, just as you say, are supposed to make the elimination of support possible, but they don’t.
      But of course software vendors are in business to make money not to sell you the best software or easiest to use software. Same thing for software engineers.
      This is nothing new, as the automotive business is beset by planned obsolescence. Dealers make all their money on repairs. Cars that lasted 20 years would cut sales in half. The average of US vehicles is currently a record 11.5 years

    • Karl says:

      I’m not an expert but I think the productivity showdown is going beyond the problems people face in the trench. Not that those problems you have outlined are not very real problems you face. But that in theory if there were much better ways of doing those tasks, (In your example, hiring more administrator assistants an using tools for longer periods of time being the potential better way) that at least one company would try to adopt those methods and if those methods are better they would out compete their competitors, which results in the eventual adaption of those methods in the broader market and the failure of companies that can’t adapt.

      This means that if its truly management processes like that limiting productivity then we are seeing a market failure were companies with inefficient business practices are somehow not eliminated by competition on a broad macro scale. AKA its not just 1 subsection of the economy that is failing but a broad failure of some sort in economic regulation.

      • Smith says:

        When was the last time you thought about starting your own bank, auto company, pharmaceutical firm? Maybe you thought it would be profitable to enter the oil business, open a factory producing clothes or electronics? Open a national chain of stores to compete with Walmart? Maybe could capture the search market where Yahoo failed and compete with google? Try opening a book warehouse to compete with Amazon or Barnes & Noble. Or maybe you can find a niche in computers, somewhere between Microsoft and Linux (which is free).
        Large companies that drive the economy are oligopolies and monopolies. Competition is something avoided. However, competition is also slow. World War I produced extreme competition and yet with more than a few players, countries, generals, there was little innovation, as mostly death and exhaustion, decided the winner. The American Civil War was like that as arms outpaced tactics and generals ignored productivity gains, or the gains were mostly in death.
        How does productivity improve in the arms race of advertising? How is financial wizardry producing anything?
        Companies are in business to make money, not to improve productivity. Cost cutting by hiring cheaper workers and lowering payroll is the very definition of reducing productivity.

  7. Serene says:

    Our political system doesn’t handle productivity gains well, which is why we needed Piketty to point out that war and its associated destruction of capital increases GDP and demand. Because almost all productivity gains flow to capital rather than labor, capital destruction stimulates the economy.

    If our political system could handle this better then productivity gains would be good, and everyone would benefit. But because we don’t benefit more equally, productivity gains have produce political instability rather than an increase in quality of life for most.

    Measurement is a tough problem and the Solow method is helpful but very imperfect, and modern technology gains are not measured properly at all. We don’t know what the real productivity gains are due to computers.

    Machine learning is already amplifying the instabilities because ownership of information is concentrated even more than capital.

  8. ezra abrams says:

    respectfully, as a scientist, I think you economists are missing the huge huge advances in biomedical research cause you are not measuring the right thing “happyness”

    we are about to allow paralyzed people the ability to move stuff with their brains
    we are gonna give blind people sight
    if you can’t measure this, then there is something really really wrong with what you are measuring

    • Jared Bernstein says:

      I think the problem is that we measure “output” and you’re measuring “welfare” or happiness. So, if blind people get sight and thereby contribute more to output, we got it. If they get sight and enjoy the majesty of the sunset for the first time…we don’t.

      In Bhutan, btw, they measure gross nat’l happiness!