Northwestern University Economics professor Robert Gordon shares his thoughts on the Productivity Paradox. Using the U.S. economy as an example, he explains why the U.S. has seen and continues to be plagued by decreasing growth even with technological breakthroughs.
This Dialogue intends to provide a forum for prominent scholars, policymakers and industry experts to exchange views and collaborate on possible new and better measurement approaches.
Luohan Academy has been in close collaboration with some of the world's foremost social scientists, policy experts, and business practitioners to document and analyze society's and the economy's ongoing digital transformation worldwide. Our very first Frontier Dialogue focuses on the measurement of the new economy, a crucial step in the direction of better understanding the impact of digital transformation.
Robert Gordon: Great. Thank you. Just like magic, the digital economy. I want to put the previous comments about measuring the digital economy in the context of a slowdown of the productivity growth that has been pervasive in the advanced economies. In the United States since 2005 and other economies even before that. Ultimately, we want to determine whether improvements in measurement that Erik has certainly led the research effort to measure the welfare effects of the digital economy. How big are these impacts in contrast with the magnitude of the slowdown? In other words, how much of the slowdown could be explained by understating the growth of, if you want to call it, GDP-B, a welfare enhanced measure of GDP. I'm going to give you some a rough idea of what has been happening in the US and Western Europe. But there have also been slowdowns in productivity growth in Japan and what I call developed East Asia, Singapore, Hong Kong, Taiwan, and South Korea. So the same story would apply to them. But my own work applies to the United States and Western Europe.
So here to the left of the vertical line, we have four time periods. The first one is 1950 to 1972. The next one is 1972 to 1995. The third one is 1995 to 2005. And the fourth one is 2005 to 2015. And I might add that the numbers for the final period would be just about the same if we were to extend that to 2019. Red is the United States and blue is what I call the EU 10, which are 10 of the countries of Western Europe that were members of the common market before 2004. They include all of the big countries, the UK, France, Germany, Italy, and Spain. And here's the rate of productivity growth in these four periods. You'll notice for the blue bars, the EU, we've had a steady and continuous slowdown in productivity growth, very rapid productivity growth at the beginning, much more rapid than the US because initially after World War II, Western Europe had fallen way by due to wartime destruction and also a long lag in the adoption of electricity and the internal combustion engine compared to the United States.
Then the catch-up period in Europe gradually ended through the 1970s and eighties followed by continuous slowdown until we reached a barely half a percent growth per year in the final decade. The United States has had a zigzag pattern with modest growth in the first period. A much-discussed productivity slowdown in the second period. A much-discussed revival in the third period. And then finally, another slowdown. Now, one of the things we have emphasized in our research, and I've done several papers now with a young graduate student. What we've emphasized is if you look at the first red bar and the second blue bar, you see, they're almost the same, that Europe in the second period grew just about the same speed as the United States in the first period. And so what we call those are the two early periods, discounting the initial big blue bar as simply a war related catch-up and not relevant for more fundamental issues.
Now, if you compare the initial United States experience of two and a half percent with the second blue bar of two and a half percent, and compare them with the final period, you see the United States and Europe had almost exactly the same slowdown, down to the second decimal point. So the slowdown that we're really concerned with is about one and two thirds percent for the United States and Europe. The fact that it's so close and so identical suggests that there is something fundamental going on, a great magnitude, because essentially two thirds of the productivity growth that had become typical in the early postwar years has been wiped out. And so when I referred to the early to late slowdown, I'm starting from two different periods, one for the United States, and one for Europe.
In our research, we’ve broken this down by industry and we found that if you divide the economy up into commodities like manufacturing and utilities and mining versus the large service sectors, the slowdown in Europe versus the United States is just about the same in each of those two parts of the economy. The big puzzle on which we've written separately is the fact that the United States had a revival in the late 1990s associated with the digital revolution. Whereas Europe did not. And I can speak briefly to that in a moment.
Here again are the same four time periods. The US with its zigzag pattern is on the top with total productivity growth shown by the width of the several colored bars. And the same is shown for Europe in the bottom half of the chart. And here we just divide up the responsibility for this productivity change into capital deepening in purple and total factor productivity, the effect of innovation, in green. And we see, for the United States that there was very little impact of capital deepening investment in the first three periods. All the action was in total factor productivity. And then in the final period, there was a decline in both TFP and investment. For Europe, going from the first to the final period, we had the slowdown equally shared between TFP and declining capital investment.
Now, one of the main points that I've made in talking about the slowdown is that investment is not strictly exogenous. We don't go off and look at interest rates or other kinds of monetary policy to understand what's happening to investment. There's strong feedback to investment, not only from population growth, which has been declining, but also from innovation itself. If innovation declines, then there is less incentive to invest. And that I think is one of the main reasons why investments slowed down so much, particularly after 2005. So again, in the United States, it's changing multifactor productivity growth that does most of the explaining with capital deepening for the US contributing mainly in the final period. For the EU 10, it's a steady slowdown in both MFP and capital deepening.
The orange colored little small segments in that graph where the effective composition of the labor force by age and sex. And that has very little effect. So what are some of the measurement issues? And here, I'm going to focus on the United States because more work has been done to quantify the issue of measurement. It's widely agreed that price indexes for digital goods, particularly for computer hardware, have an upward bias, that is, official measures of the prices of computers don't decline fast enough. And that, it turns out, goes in the wrong direction if we want to explain why productivity growth has slowed down so much, because the bias from this mis-measurement of computer hardware has gotten smaller for several reasons. And for a bias to get smaller, it goes in the opposite direction of what we're trying to explain. Now, one of the reasons why the bias has gotten smaller is that IT investment is simply a smaller share of GDP. It's the nominal share of any given component of GDP that measures its importance.
So the share of total information technology investment in GDP in the two peak years of the late nineties was 2.8%. In the two most recent years, 2018 and 2019, it was only 1.9%. So it had fallen by a third. If there had been a regular constant upward bias in price indexes for computers, the importance of that for GDP measurement would have declined. Now in the measurement paper that discussed all this written four years ago by David Burns and co-authors, they showed that the measurement bias for labor productivity coming from IT investment had fallen by half from the late 1990s to the mid 2010s. And most of the remaining bias and price indexes for computers is in the semiconductor chips themselves, which are an intermediate good in the manufacturer of final computer equipment where the bias has become smaller.
So this is one aspect of the measurement issue. We're looking for bigger biases to explain the productivity slow down, but so far we've found one source of a smaller bias. Now here's another really important thing, and this applies perhaps more for the United States than for some other countries, and certainly would not apply to China. If you look at the price index for computer equipment for imports shown by blue, and all of these are negative growth rates with zero at the top of the graph and the vertical axis. If you can't see it, it goes down to about minus 25% per year. So very little decline in the import price index, rapid decline in the red line for domestic price of computers. And the average, which weights domestic and imported computer equipment production, is the black line in between.
And you see that that black line gets a lot closer to the blue line at the end, reflecting the increased importance of imports. Well, we know that that import price index is nonsense. There's no way that import prices of computers could be declining so much less than domestic prices when at the same time purchasing has been switching rapidly from domestic U.S. production to imported computer equipment. So, what is the importance or the significance of what is clearly a very large upward bias in the import price index? It's obviously gotta be wrong because import computers, if anything, are getting cheaper than domestic computers. So let's say that the import price index is wrong, and we replace it by the domestic price index for computers. There's no impact on GDP or labor productivity. We simply get more investment with faster declining prices, but also more imports, which are subtracted out of GDP.
But what it implies is that because investment has gone up due to the faster decline in prices, we get more capital deepening given labor productivity growth that implies less total factor productivity growth, thus deepening the mystery of why total factor productivity growth is growing so slowly in the wake of all this investment, all these patents. Here's something that goes along with the previous graph. I mentioned that the average gets closer to the imports. Here is the share of computer equipment investment in the United States from 2002 to 2013. And there was a massive shift right after the great financial crisis in reliance on imported computers. So again, this doesn't affect GDP or productivity. What it does is it raises investment and reduces measured total factor productivity. And if the imported component of computer output is a hundred percent, that means that computer output is no longer part of GDP. So any further discussion of price index bias is irrelevant for GDP to the extent that we're talking about computer hardware. This does not affect anything that Erik talked about, free internet services.
So let's just say a couple of things to conclude about free internet services. First of all, they're not free. Both smartphone subscriptions and wifi subscriptions do cost money. They're not part of marginal costs. They are fixed cost that has to be taken into account. When we look at our own expenditure, it depends how long you keep your smartphone and how rapidly you depreciate it. But you're talking 100 or 150 dollars a month for the cost of these internet services. What Erik is doing with this research goes beyond some of his earlier work that valued hours of internet use by using the employed wage and instead the value of internet time should be compared with the previous uses of that time.
I mean, previously people weren't just staring at the ceiling, they were listening to the radio, they were listening to music, they were watching television, they were talking on the phone, they were playing board games. They were getting consumer surplus from what they were doing, clearly less than they are now but not zero. Now my expense focused on this and it's something that is at the heart of our discussion today. The consumer surplus produced by smartphones and the internet is much greater than the increase in business productivity. That's obvious from Erik's
computer consumer surplus graphs. Facebook use is outside of the realm of the production of business goods and services. And in addition to that, we know that some portion of the value of the hardware of the phone is imported because the actual construction of these phones is not part of GDP.
So let me just conclude by appreciating Erik's discovery that people still value indoor plumbing. Throughout the history of the industrial revolution, we've had inventions that add the consumer welfare without being included in GDP or affecting business productivity. If we think about running water, indoor plumbing and the associated reduction in infant mortality and infectious diseases, they make the rise in consumer surplus during the period of 1870s to 1930, much faster than the rise of GDP or productivity, which if we look back at the data, looks mysteriously slow during that period that was over overwhelmingly influenced by the production of electricity and the internal combustion engine, the end, of course, droppings. Much of the infrastructure for running water and indoor plumbing was the investment of local government. It was not part of the business sector at all. So these are things that we need to consider. It's not just all those great things we're getting now, but how valuable are there compared to all those great things we got at various points in the past. Thank you.
For more information, please visit Luohan Academy's youtube channel: Luohan Academy