We should probably give up on the hopeless task of figuring out the “correct” measure of productivity. Real output is an abstraction with an increasingly tenuous connection to reality. And it follows directly from this that the productivity data have little practical meaning.
Trying to raise what we call productivity from its recently dismal pace of advance would still be wise, I guess. The point here is not to be luddite. But when fussing over the measurement of or outlook for productivity, it is probably best to ask first: why do we care? How “productivity” is doing will be a function of that.
US productivity growth is looking pretty crummy.
Friday’s Productivity and Cost report for 2016Q2 showed that output per hour in the nonfarm business sector was roughly unchanged from the end of 2014. Outside recessions or major supply-side shocks, that is unusual.
Just looking at the chart, which is admittedly amateur, the underlying trend of productivity growth looks to have settled at just 1/2% a year this decade. That would be quite the step down from the 2 ¾% average annual advance achieved in the fifteen years to 2010. And it makes the early 1970s slowdown look like a boom by comparison. Alas, the 1970s economy is often unfairly disparaged. Certainly, nobody wants to go back to those horrible days, people say.
There are two debates over this weak showing, which are running concurrently. The first is whether the data are even correct. Economists at Goldman Sachs, among others, claim that real output growth is particularly likely to be understated in the current environment, because the output data miss both the speed of tech price decline and the consumer surplus associated with new product releases, some of which are immediately transformative.
Others, such as these economists at the Fed and IMF, claim that measurement problems have always been with us and that, if anything, the recent slowdown of output and related productivity growth may be even sharper than the official data imply.
The second debate is over whether we should extrapolate the current weakness. Robert Gordon’s tome on the The Rise and Fall of American Growth, published earlier this year, says that we should. Cell phones and web surfing are cool, but they pale in comparison with rural electrification, the assembly line or even indoor plumbing. For Gordon, the glory days ended a few decades ago.
Larry Summer’s secular stagnation thesis is based in part on the idea that the information technology revolution has run its course and that productivity growth for the next while will be weak. But he does not seem to be on board with Gordon’s claim that all the low-hanging fruit have all been picked and that the boom of 1870 to 1970 cannot be repeated. Summers could imagine a revival, if the right policies were put in place. These policies almost invariably increase the speed at which Summers can get home from work or to his speaking engagements in London!
And of course the guys driving the IT revolution have no idea what the academics are even talking about. For them, the information economy is just getting started.
I do not have much to offer on the fundamental question of whether the best days of productivity advance, as understood by the academics, are behind us. I would offer only that the consensus view among the experts has never been well formed on this issue and has never made both strong and accurate predictions.
On the measurement issue, I also don’t have anything technical to offer. But I would like to address a glaring omission in this whole discussion. It is no brilliant insight to identify this omission. Nor is it even controversial. What I see is something that just gets swept under the rug, because people don’t really know how to deal with it.
The notion of real output has always been an abstraction and its connection to what we ultimately care about is becoming increasingly tenuous. As a direct result, the productivity data are losing relevance for most people, excepting perhaps those who like to argue about them. It is not that the data are biased up or down. It is that they are hopeless abstractions.
To see this, contrast Goldman’s insistence on the social value added of the social network with how Robert Gordon begins his empirical review of American Growth.
Goldman points out that real economic activity associated with Americans’ use of Facebook is systematically undercounted by the output data, because those data capture only the improvements to Facebook after it has entered into the government’s accounting framework. That is a major issue because the very idea of Facebook was transformational.
The weakness in the government bean counting highlighted by Goldman is an important issue, although not one on which I can add anything technical. But here is a more philosophical issue, if you will pardon the big word. Is your use of Facebook services a rivalrous or non-rivalrous form of consumption?
I think one could argue that Facebook time is rivalrous and thus closer to defense activity than to, say, enjoying a nice lonely walk in the woods. “My life is more fabulous than yours, as evidenced by these photos of my great vacation with my good-looking family in these good-looking places!” “Just to let you know, I got a promotion!” Or more darkly, possibly from a teenager being cyberbullied: “No those mean things you said about me are not true and here’s why?”
Or what about when you post something you regret? I do that all the time, here. Because the subject is economics and because I have very few readers, no biggie? But Facebook is different. Facebook can be forever, for a lot more people, and about a lot more personal stuff.
This is just one example, cuing off the seemingly-valid technical point that Goldman economists raise. But reasonable people might disagree about whether Facebook has even been a net source of satisfaction to Americans.
I am not able to judge that debate. But I can tell you that there is nothing in national income accounts that takes a view on any of the nuances there. Nor could there ever be. And if the dispute over how quickly productivity is advancing boils down basically to how we treat things analogous to Facebook, then forget about it. There is no way the productivity data can measure what they purport to.
Compare this how Robert Gordon opens his historical review of American productivity advance. After documenting that that the glory days for US productivity growth were from 1920 to 1970 and how that relates to the standard growth accounting framework, he turns to consumer spending patterns in the US, with a particular focus on caloric intake, something we can easily get our brains around.
The first chart getting down into the weeds, found on page 64 of American Growth, tracks per capital caloric intake from 1800 to 2010. For the first 180 years of that period, there was no net advance. And then, this measure of economic welfare – and the benefits of productivity advance – took off.
Gordon’s point is not that the productivity surge began in 1980. The quality of calories and presumably the satisfaction associated with getting them in the belly rose over much of the nineteenth century. But in the early history of the United States a big part of economic welfare — and its advance — was about food.
This comparison of Facebook with Food is not dispositive. Rather, it draws your attention to something I assume is obvious and does not really need arguing. Our conception of real economic output is now far too abstract to allow productivity advance to be measured with any precision. And there does not seem to be a technical fix imaginable, short of metal probes inserted into the pleasure centers of our brains.
This brings me, as many things do, to the main insight of philosophical pragmatism. If you find yourself puzzling over the meaning of the productivity data, your first instinct should not be to attempt to find the truth. That would be impossible. Rather, your first question should be, as almost always, why do I care? When I have an answer, what do I plan to do with it? **
I will get to that below. But first, let’s consider a glaringly obvious fact, which highlights that people mostly already intuit that the productivity don’t do what they purport to do.
The LEVEL of productivity in the United States has never been higher.
If you believe that there is such a thing as material wellbeing and that productivity strongly influences the measure of it, then things have never been better. Moreover, things can only get better from here, because even the productivity pessimists concede that productivity will rise further, just less quickly than in the past. When could be a better time to be alive, that is, besides tomorrow?
I think there is a reason that nobody looks at it that way, and that people instead obsess over the RATE OF GROWHT of productivity. The reason is that people instinctively realize that real output is an abstraction and view productivity growth as an influence on, and therefore proxy of, the things that actually matter.
Strikingly, the idea that productivity growth is a proxy for what really matters is at odds with the standard welfare economics model that is behind the idea that productivity can even be measured. I find that a fun thought, although not one I am competent to pursue very far.
Returning to theme, here are some examples of how measured productivity growth might be indirectly relevant, to the extent it relates to what we actually care about.
Productivity advance might distract people from their concern about income inequality. If the rising tide is quickly lifting my boat, then I might be less concerned with the other boats. The left-leaning economist, Brad Delong, makes basically this point in his recent blog post explaining why he prefers the neoliberal “growth” agenda over the harder-left redistributionist agenda offered by Bernie Sanders and his allies. Delong might also be alluding to the importance of habit formation, which inclines the level of satisfaction to vary with the rate of change of consumption. That would be an analogous argument. In any case the real issue here is containing resentment, not really achieving productivity growth per se.
Faster productivity growth might reduce the importance of frictions in the labor market or the effect of the zero lower bound on nominal interest rates. It might be easier to achieve full employment, in a world where productivity growth – as conventionally understood – is high. This might be the case so long as the underlying technology advance is not extremely biased against the labor share. That makes an interesting debate, but the thing we are really trying to achieve here is full employment.
Faster productivity advance in the US might correlate with America outperforming other countries, which would be relevant from a defense perspective. For example, we probably want to stay well ahead of China on productivity, and military tech advance in particular. From this perspective, productivity advance is a zero sum game, globally. That does not fit into the standard welfare model underlying the conventional measure of productivity, to say the least.
Narrowing our focus to mercenary concerns, cyclical spikes and declines of productivity growth, which are sometimes called productivity “surprise” are the main determinant of profit margin changes. If you own equities, this is a pretty important short-term consideration. Over the longer haul, this consideration is dominated by the economic fundamentals that determine the capital income share in equilibrium, when the level of productivity is presumably as expected. Those forces are invisible to we mere mortals, which is why we focus on productivity surprise instead. But even in this context, productivity growth is a proxy for something else, basically operating leverage.
Finally, have you noticed that on Wall Street the doves have tended to be productivity optimists while the hawks have tended to be pessimists? This occasionally reflects the assumption that slowdowns of productivity advance push up unit labor costs (true), which feeds directly into higher inflation and the need for Fed tightening (apparently not true). People arguing about productivity are actually arguing about what the Fed should do, and often quite poorly.
The point is that much of the popular and even advanced discussion of productivity is ultimately really about something else. Rather than get bogged down in the fruitless search for the “true” measure of productivity, then, we should look instead at how economic developments are influencing the things we actually care about. In many cases, following Ocham, and under the advice of the pragmatists, we can just leave the productivity middle man out. The questions that then arise can occasionally be easier to answer, as well as more relevant.
It is a bit like the joke about the hapless bear hunter whose presumably-last experience hunting is hearing the hungry bear say: “you are not really here for the hunting, are you?” What he thinks he is focused on and what he is really focused on are different things. For the bear hunter it is too late. But it might save you some trouble to figure out what you are really up to when thinking about productivity.
* The first charts, on pages 14 and 16, show that labor productivity growth was by far the quickest during 1920-70, primarily because total factor productivity, a proxy for the impact of tech advance, was quickest then.
** Here is a nice definition of philosophical pragmatism. Pragmatism may be presented as a way of clarifying (and in some cases dissolving) intractable metaphysical and epistemological disputes. According to the down-to-earth pragmatist, bickering metaphysicians should get in the habit of posing the following question: “What concrete practical difference would it make if my theory were true and its rival(s) false?” Where there is no such difference, there is no genuine (that is, non-verbal) disagreement, and hence no genuine problem. Those of you actually schooled in philosophy itself might turn up your noses at such low ambition. But I am pragmatic in my taste for pragmatism. I guess that might circular, but full disclosure anyway.