NCN Summit 2001: Professor Robert Solow, MIT
Information Technology and the Recent Productivity Boom in the US

view video -- return to video home

1. My general subject is the fantastic U.S boom of 1995-2000, and especially its relation to ICT. The unexpected combination of low and falling unemployment and low, fairly steady inflation is not by itself the main concern here (though it is of great importance). If the U.S had experienced an acceleration of productivity and had screwed it up macroeconomically, we would still be interested in the productivity surge. In fact there is an important connection between the productivity surge and the combination of low unemployment and low inflation. Without the first, the second would probably not have been attainable. But that is not our focus now.

2. The 'new economy' hype in the popular and business press - the notion that the laws that govern the behaviour of the economy have changed in the IT age, that the business cycle is a thing of the old-fashioned past, that there are no basic constraints on economic growth - all that was always foolishness and should have been seen to be foolishness from the beginning. In any case, the slowdown had already begun by the middle of 2000, and a standard recession was probably already in the cards well before the attack on the WTC. Its cause was classical: businesses perceived that they had overinvested in ICT equipment, and began to pull back, despite continued high consumer spending. The situation was pretty clearly worsened by the extra uncertainty imposed on both consumers and firms by the events of September 11 and after.

3. Our real topic can be introduced this way: between 1947 and 1972, real output per hour worked in the private non-farm economy rose at an average annual rate of 2.9 percent; between 1972 and 1995, that figure was cut in half, to 1.4 percent; and then between 1995 and 2000 it went back up to 2.5 percent, not quite reaching the early post-war figure, but a big gain. Differences of that magnitude have important implications for the path of the standard of living. What made the difference at the end of the 1990s? In particular, was it mainly the - long delayed - fruition of the computer age? Is it likely to continue?

4. I am going to go at this by way of an unfamiliar breakdown of the 'contribution' of different industries to the acceleration of productivity after 1995. If a very small industry experiences a big jump in its rate of productivity growth, that may be intrinsically interesting, and one would want to know why, but it will contribute little to the acceleration of productivity on a national scale. The importance of a particular industry depends on both the size of its jump in productivity growth on its share of total employment. Looked at this way, a mere six out of 59 industries contributed essentially ALL of the acceleration of productivity from 1987-95 to 1995-2000. The other 53 industries included some positive and some negative (i.e. decelerating) industries, but netted out to zero. What distinguished 1995-2000 from other periods was not the mere existence of a small group of industries whose productivity growth rates jumped - that is common - but the fact that the industries in question were big employers. I do not think you would guess the identities of the three largest contributors to the productivity surge: they were, in order, wholesale trade, retail trade, and security and commodity brokers. The three next-largest contributors were more expectable: electronic and electric equipment (i.e. semiconductors), industrial machinery and equipment (i.e. computers), and telecommunications. The contribution of the first three was almost three times the size of the contribution of the last three.

5. Remarkably, then, if we want to understand the productivity surge, we have to understand what happened in wholesale and retail trade and financial services. If this was mainly something to do with ICT, then it is something to do with ICT in those three sectors. That is not far-fetched, of course. There has been a lot of computerization in just those sectors. We know that sales of computers and peripherals to business firms have been heavily concentrated in sales to firms in the service sector of the economy. But buying IT is clearly not enough, in the service sector. Retail banking, for instance, has been a very big investor in IT, and so far has little or nothing to show for it. It turns out, on close inspection, that computers have something to do with the productivity surge, but not nearly everything.

6. For example, in wholesaling, warehouse centralization and automation was the most important factor. There was a large technological component in relation to picking and shipping, but the technology used - including IT - was quite old when the productivity acceleration occurred. The key factors were really managerial innovations in the organization of functions, some of them made possible by increases in scale. It appears that the whole process was forced by pressure from large retailers. In retailing, the story is similar, and even more closely traceable. It can be summed up in the Wal-Mart phenomenon. Wal-Mart opened up an enormous productivity gap over the rest of the industry, which was then forced to imitate in order to survive. (It has not caught up yet, which suggests that there may be rapid productivity growth still to come.) The Wal-Mart advantage owed something to IT, but again not to anything new or exotic. However most of Wal-Mart's innovation was managerial, especially the large-store format and improved logistics.

7. There is an interesting semi-technical point of interpretation here. It is natural to measure labor productivity in terms of real value added per hour worked. One big advantage is that value added by industry can be added across industries to give national income. A consequence of this choice is that if a Wal-Mart, for example, shifts to selling higher-value goods, its measured productivity will increase even if, so to speak, the sales people perform the same physical actions at the same pace. This seems odd, but it is probably appropriate. One would not hesitate in the case of manufacturing: if an innovation allows a machine-tender to produce a better widget by doing the same motions in the same time, we have no trouble in labelling that an increase in productivity. So we do the same in trading activities. Perhaps the increase in measured productivity should be imputed to whatever was the source of the higher-value throughput. Don't think of the routine calculation as imputing any 'deep' productivity to individual participants in the chain that brings real value to final consumers. It is just a calculation, and all the value-added in the economy has to be accounted for somewhere. (But remember that there remains the 'deep' productivity gain that does come from improved retailing formats and better organization of the functions and tasks that they perform.)

8. Next I want to tell you about some evidence that goes mildly the other way, and then I will suggest that this evidence is incomplete. My reason for this giving-and-taking-away is that it helps me to make an important analytical point about the whole new-economy issue.

If you look at a lot of different industries, and compare their productivity growth rates with the extent of their investment in IT, you find a positive correlation across industries. The correlation is not terribly close or sharp but it is there. You could conclude that the productivity surge of 1995-2000 was indeed the fruit of the Information Age.

But be careful. The question is not whether IT investment increases productivity. Of course it does. Any intelligent investment increases productivity; that is what it is for. The New-Economy question is whether IT investment has a substantially larger productivity-increasing effect than other investment in equipment. Equivalently, is the return to IT investment larger than to other forms of investment? The correlation I just mentioned throws no light at all on that issue.

Some of the detailed studies I was citing earlier suggest that there was no such super-profitability effect. (One or two such studies do show it, but in such an extreme form as to create instant doubt.) There is also a casual but telling reason to doubt this possibility. If it were generally true that the return on IT investment had exceeded the return on other investment - which, to repeat, is what this argument amounts to - businesses should find themselves wishing that they had invested more in computers than they did. (There were no supply constraints holding them back.) No such groundswell has appeared. On the contrary, there was probably over-investment, though that reaction may be more a business-cycle effect than a statement about the longer run.

9. So New-Economy romanticism needs to be cut back to size. But then what is it reasonable to expect for the next five years, after the current recession has run its course? (I am not offering an opinion on the nature of the current recession.) Between looking at some important industries individually, and doing some iffy extrapolation of aggregates, it is reasonable to suppose that about half of the gain from 1.4 percent to 2.5 percent annual productivity growth is likely to endure for the next five years. So one might look forward to productivity gains of 2 percent a year, after cyclical effects are washed out. That is significant acceleration, if not the stuff that hype is made of. Keep in mind that five-year extrapolation makes some sense precisely because it is too short a time-span for brand-new as yet undiscovered technologies to play a role. I have no idea about that, but then neither does anyone else.