Where’s the Automation in the Productivity Accounts?

June 5th, 2014 at 7:57 am

Yesterday’s productivity report for 2014q1 was predictably negative—we already knew that real GDP fell in the quarter while employment grew apace—but I don’t read much into the noisy quarterly changes.

But then there’s this: year-over-year, productivity growth was up 1% last year and has averaged 0.8% since 2011.  The figure below plots the yearly changes, which are themselves pretty noisy.  What’s more instructive is the smooth trend through the numbers.

Year-over-Year Productivity Growth and Smooth Trend


Source: BLS, my analysis

The trend suggests that the pace of productivity growth has decelerated since the first half of the 2000s and this begs an important question.  There’s considerable speculation that the pace at which machines are displacing workers has accelerated.  I keep hearing about “the end of work” based on the assumption that the pace of labor-saving technology—robots, AI—has accelerated.

Maybe it has—there’s lots of good anecdotes to that effect, most recently that geeky-looking Google self-driving car.  And data being the plural of anecdote, I’m certainly open to the possibility.  But the robots-are-coming advocates need to explain why a phenomenon that should be associated with accelerating productivity is allegedly occurring over a fairly protracted period where the trend in output per hour is going the other way.

A shave with Occam’s razor would lead one to conclude that over this weak expansion characterized by large output gaps, a simpler explanation for decelerating productivity would be weak demand and its corollary, weak capital investment.  Perhaps once we close those gaps, we’ll be better positioned to get a read on the pace at which labor-saving technology is entering the workforce.

Finally, until someone can convince me what’s wrong with the above argument, I don’t want to hear that automation-induced productivity gains are precluding full employment.  The problem isn’t productivity; it’s negligent policy.

Print Friendly, PDF & Email

14 comments in reply to "Where’s the Automation in the Productivity Accounts?"

  1. Larry Signor says:

    “…negligent policy.” certainly gets my vote. Labor saving technology need not also be job destroying technology as seems to be assumed by the high unemployment apologists. Just another smoke screen.

    • Nick Batzdorf says:

      It gets my vote 100% too, and for sure it’s being used a smokescreen by unemployment apologists. Negligent policy is about the best one can call what we have. “Cynical” seems like it fits too.

      But that’s today. 15 years on, how could, say, 3D printing possibly not lower employment? I desperately want to believe that there’s always going something else for people to work on – that’s certainly been the case historically – but I just don’t see how it could happen with the next stages of the digital revolution.

      Please explain how I’m wrong. I mean that sincerely, not rhetorically!

      Again, the follow article scares me. If 45 percent of America’s occupations are going to be automated within the next 20 years, a large number of people would have to be SOL.


      • Smith says:

        You are confusing work with employment. Computers can do all the work for everyone and that does not mean everyone can not be employed. Certainly the owner proprietor will pay himself.
        “Over 70 percent of U.S. businesses are owned and operated by sole proprietors or sole traders.”
        That’s 22 million (keep in mind entire full time workforce is 120 million)
        Partnerships and Corporations will always pay themselves (another 9 million).
        What about the other 3/4 of the workforce?
        Assuming the 1/4 business owners and corporate executives aren’t all computer and robotics experts, they might need someone to assist. Now we’re half way done.
        Being that 1/4 of the workforce now has the phoney job of looking after the robots and computers (because actually the computers and robots are self monitoring but only the computer and robotics workers know that), the owner or executive might decide to offload some of his onerous burden running the company, perhaps hiring a secretary, a relative, or a manager to manage the computer and robotics employee. Lastly and most importantly, if he wants to stay in business more than a few years, he needs to invest in R & D, because otherwise his competitors will eat his lunch. Most of the R & D is done by computers, but he needs an actual person and luck to realize any gains.

        In truth, it matters not if all work is automated, all that matters is that a mechanism exists to divide output equitably. I would argue that we’ve already mastered the art of creating non-essential jobs that produce nothing in order to distribute output of the economy through antiquated structures that resemble remuneration for work. What is alarming is that people with the most are hoarding instead of sharing, and will continue to do so until a political movement directs otherwise.

        • Nick Batzdorf says:

          “all that matters is that a mechanism exists to divide output equitably”

          “What is alarming is that people with the most are hoarding instead of sharing, and will continue to do so until a political movement directs otherwise”

          Isn’t that the same conclusion I’m reaching, Smith, only from a different angle?

          I’m all for higher marginal tax rates, but even if that were possible politically, simple logic and everything I’ve read says it alone isn’t enough. That’s why I’m in favor of FHA-style loans and anything else we can do to expand ownership.

      • Nathan J. Kerr says:

        There are hedges to this of course. One is to reduce the cost of human capital by removing income taxes and medical costs in favor of consumption and sales taxes and singe payer catastrophic/ chronic illness insurance with health savings accounts for actual primary care opening it up to more competitive pricing pressures with Nurse Practitioners replacing most MD’s/DO’s in primary care.

        Currently we give tax breaks to automate (AI and robotic business equipment as business expenses with depreciation tax shields) while adding costs (taxes, workers comp insurance, Obamacare, FICA, Social Security, unemployment insurance premiums etc.) to hiring human capital. This is “bass ackwards”! We incentivize automation and dis-incentivise hiring people!

        If we remove the costs of hiring humans, most small businesses will hire humans over the initial costs of buying expensive equipment. Additionally, this equipment would start being taxed making it a little more expensive and businesses will be pickier in what the automate.

  2. Robert Buttons says:

    Productivity is declining at my business because we are devoting many more man hours than ever before to govt compliance.

    • Peter K. says:

      “And data being the plural of anecdote, I’m certainly open to the possibility.”

      Most anecdotes I’ve come across have to do with a lack of demand rather than onerous government regulations.

      DeLong described Pikettyworld and Summersworld (secstags) as being orthogonal. Secular stagnation reduces productivity. (Does it reduce profits or do they just take it out of labor’s share?) They attempt to boost profits (or productivity) by increased automation.

      And the apologists declare there is nothing that can be done about labor’s reduced income share because automation is a necessary response to secular stagnation. (Likewise government’s role needs to be reduced so that the job creators can create. Meanwhile, cutting government’s role in stablizing the economy reduces growth and productivity levels.)

      • Robert Buttons says:

        Check my comment, I never tried to extrapolate my personal business experience to the economy as a whole.

        The Luddites will disagree, but automation makes us all more productive. BTW, you are free to bypass the ‘net and save USPS jobs by sending your posts via snail mail.

    • Larry Signor says:

      If your business is “devoting any more man hours than ever before to govt compliance” an economist might posit several reasons. Perhaps the business generates unfunded negative externalities such as pollution, traffic congestion, monopsony or monopoly power, the list could go on. Perhaps the business has certain labor disputes that interest government regulators and demand a legal reply, employing company resources. Perhaps the business is expanding into areas of non-expertise where government oversight is more stringent but the business mis-priced the costs of expansion. An anecdote is a story until the data piles up. Still waiting for the data.

  3. Smith says:

    One would need more data to actually create a model and check for suspected causes of lower measured productivity growth. In the absence of data one strongly suspects weak gdp growth and slack labor markets are overwhelming actual productivity improvements. The spike and slow descents following recessions of 2001 and 2008 are telling. In a recession, payroll is cut creating immediate efficiencies. The modern service economy is more sensitive to this effect. One worker creating useless reports, emails, and attending meetings easily supplants the productivity that two workers formerly accomplished. When no one is left to downsize, productivity begins it’s slow descent as a) the effect of layoffs takes hold furthering weakening demand, b) fixed costs can not be cut c) the partly illusionary boost to layoff induced productivity gains make subsequent years appear worse than they actually are. In this case 3 year moving average exaggerates the ups and downs in our slow motion recession.
    There is also an effect that low inflation has on productivity measurements, since below a threshold, prices and hence GDP may be held historically lower than would be otherwise.
    Finally, I’ve made the case previously that GDP is often inflated by useless trading of expensive services by middle and upper classes and businesses. Other factors also affect measurement of productivity that have nothing to do with actual technological progress. Productivity is only inferred by looking at hours reportedly worked and GDP. Cutting down on unneeded medical tests will by itself lower GDP and productivity growth. Again data and modeling is needed to quantify all the effects.

  4. Rima Regas says:

    Right on, Jared! The automation is erasing the jobs zombie myth is just that. While we might be seeing stories about artificial intelligence and self-driving cars, those things are not in mass-production and in use by the general public. What’s more, 100% of the jobs that have been eliminated since 2007 fall into two categories: the largest, outsourced to much cheaper Asian labor, and the smaller eliminated altogether not because the jobs were redundant, but because the labor force is so fearful that you can spread extra work around with impunity, in the knowledge that it will be done, even at the cost of not reporting overtime.

    All this talk by Friedman and Brooks about new “efficiencies” is a bunch of hooey. Millions have been abandoned and are suffering in silence because no one in the press bothers to report on what the public now sees as “losers.”

  5. jhaskell says:

    “Finally, until someone can convince me what’s wrong with the above argument, I don’t want to hear that automation-induced productivity gains are precluding full employment. The problem isn’t productivity; it’s negligent policy.”

    It’s partially lazy analysis. Much like the “skills gap” argument, skills biased technological change has a ‘cool’ sounding name that people accept as gospel. State and local level policymakers in my state have been able to avoid any critical analysis by peddling these types of “analyses”.

  6. Larry Signor says:

    Just another anecdote; I have to repair a roof leak today. Hard to get a robot to do that. It is also very difficult to quantify the productivity of service workers. Just thinkin’.

  7. Gerald Huff says:

    I think most of the noise about automation destroying jobs is about the near future, not today. As you say it’s the “robots are coming” crowd, not the “robots are here” crowd (though to be fair, there have been some robots-are-coming folks in the past). Google’s self-driving car is a great example. Google has made tremendous progress in the just last few years, but we are probably at least 10 years away (perhaps more depending on regulatory friction) from widespread adoption. But if you cast your vision that far down the road, you have to ask what the millions of truck and taxi drivers are going to do. The technologies emerging from labs today have the potential in the coming decades to disrupt occupations and industries that employ a significant % of people. And most of the people who raise this issue are not anti-tech Luddites who say we should “stop progress”. They are just asking – how will we accommodate this coming disruption and ensure that the bounty of this incredible tech innovation is shared more broadly? And if the answer is “we will just retrain people for the new jobs that will be created in new industries”, then we need some evidence that (a) the new industries will not start automation heavy, employing a small number of highly skilled people (cf WhatsApp) and (b) that we actually know how to retrain 30-50 year old cashiers and truck drivers for those jobs. And if the answer is “many will move to the entrepreneur/sharing/maker economy” then we should figure out how to retrain/educate them for that very different world and ensure they can make a living when millions more people are competing to share/make. There may be other answers too, all perfectly sound. The point is to do a risk assessment and make sure we’re attempting to build alternative models for a future with amazing AI, robots, 3D printing, personalized medicine, etc. Some of the systems that *might* need to adapt – e.g. education, health care, safety net, taxation – have very long lead times and are historically difficult to change, and today’s gridlocked political environment doesn’t instill a lot of confidence in our ability to change that.