RIETI-KEIO Conference on Japanese Economy

Leading East Asia in the 21st Century?-21世紀の日本経済:東アジア諸国との競争と協調-

イベント概要

  • 日時:2003年5月30日(金)9:00-16:45
  • 会場:慶應義塾大学(北館ホール)
  • 開催言語:英語
  • Speech Summary

    PART ONE:Assumptions for productivity projections

    There still is a definite slowdown in Japan, and there is a non-sustainable element in US growth during the 1990s. In the case of the United States, the number of hours worked grew at about 2% a year, but the working age population grew at about 1% a year. And we can say from that the growth of hours worked was not sustainable, and therefore in projecting growth going forward, we would want to assume that the growth rate is more like the growth rate of the working age population.

    The reverse is true in Japan, that during the 1990s, the working age population in relationship to hours worked increased, or the hours worked in relation to the working age population shrank, but there were very important reasons for that in addition to slack aggregate demand. The most important is one that was analyzed in the Review of Economic Dynamics, namely institutional change, the institution of the 40-hour week. That is something that we assume is not going to be reversed in looking at the outlook for the Japanese economy going forward.

    So what I would like to do is to look at sources of US economic growth analysis similar to the one you have just seen for Japan, and then I would like to project potential growth for average labor productivity, potential growth being one in which the employment rate remains in a constant relationship to the working age population, and then look at a similar set of projections for Japan.

    Let us look at the historical record. The growth of GDP is the sum of hours worked and the growth of labor productivity. My comparison is going to be before and after 1995. And the reason for that is the change in the rate of productivity growth that took place in 1995. If you look at the figures before 1995 - I am going back here to 1973 for the United States - hours worked grew at about 1.4% per year, and afterward at about 2% a year. Labor productivity went from about 1.3% to 2.07%.

    What are the sources of labor productivity growth? One is capital deepening. Capital deepening is defined as capital input per hour worked. The intuition is that investment provides more and better capital to workers, so it is a classic substitution of a less expensive factor, namely capital, for a more expensive one, namely labor. Labor quality growth is the increase in the proportion of more productive workers. And then finally we come to total factor productivity growth where TFP is defined as output per unit of capital and labor inputs.

    What changed after 1995? Between this earlier period of slow growth, 1973-1995, and the period of rapid growth from 1995 to 2000, there was an increase in labor productivity growth of about three quarters of a percent. Half a percent of that was due to capital deepening in the form of IT inputs. That was the substitution of IT inputs for labor, but also for non-IT inputs, because capital deepening for non-IT, the other category here, was actually negative. In other words, there was substitution away, not only from labor, but also from non-IT. And the reason for that of course is because of the price changes in IT.

    Labor quality is this concept of the growth of more productivity per worker, the average marginal productivity of labor. And that turned out to be slightly negative. You would think ordinarily that the labor force would be upgraded as more highly educated and more experienced people came into the labor force. But just the reverse occurred from 1995 to 2000 with the huge increase in hours worked. The unemployment rate declined, and the labor force employment rate increased. The result is that more people with lower qualifications, lower productivity, were drawn in, and that had a negative effect on the growth of labor quality.

    Finally we come to total factor productivity, output per unit of input which I have allocated in the same between IT and non-IT, and for IT production that was about a quarter of a percent per year, and for all other, about half that. And so the bottom line is that all of the growth in labor productivity can be attributed to IT. About two-thirds of that is capital deepening, and one-third is output per unit of input in IT production in the very narrow set of industries that produce software communications equipment and computers. The other elements here cancel out. So it is basically an IT story.

    In making projections, I am going to make two key assumptions to remove those transitory effects. One is that I am going to assume going forward that output and reproducible capital grow at the same rate. Reproducible capital is the kind that is produced within the economy; and non-reproducible capital is the kind that is produced by nature, namely land. The land surface of the United States has not changed in over 100 years, and it turns out that the reproducible capital is about 80% of the total, and that is the part that is going to grow at the same rate as output. The second thing we need assume is that hours growth matches labor force growth, and that is important, because it means that instead of having a reduction in the unemployment rate and an increase in the employment rate of the population, the proportion of people who work, we are going to assume that those factors stabilize.

    Those are two assumptions that I am also going to make about Japan, and that is going to differentiate this lost decade from the resurgence of the US economy where the business cycle factors were a strong positive. I am going to make three different scenarios to illustrate the different possibilities, the pessimistic, base case, and optimistic assumptions. I am going to allow assumptions about total factor productivity growth in IT and elsewhere in the economy to vary across the scenarios. And I am also going to allow for differences in capital quality growth, which is the rate at which IT is substituted for non-IT within the capital stock. And then I am going to assume in common across all the scenarios that hours and labor quality growth will be the same and can be derived from demographic projections.

    It makes sense that that would be the case because we know almost all the people who are going to be in the labor force ten years from now: They are either in the labor force now, or they are in the school system and are going to join the labor force in the next 10 years. Furthermore, the rate at which people achieve various levels of education has not changed in about three decades in the US. The shares of capital labor and the output of IT as a proportion of the total economy have also been the same for about a decade. In fact the capital labor shares had been constant going back a good bit before that, so I am going to assume they are constant at historical averages.

    Now we are ready to do our calibration. The base case is going to use 1990-2000 averages, because we have had a shift in the pace of information technology progress reflecting a more rapid product cycle, and I am going to project an eventual reversion to the old product cycle, and it turns out that 1990-2000 is just about evenly divided between the two. That means that I am going to assume there is going to be a slowdown in the rate at which technology progresses in IT. I am not going to assume that that is going to continue. I do not have a model of endogenous productivity growth. I do have, however, an alternative case, and that is an optimistic case. Let us suppose, for example, that the very progressive information technology industry continues at the pace that characterized the late 1990s. That is my optimistic scenario. My pessimistic scenario is that we revert to the period before 1975 and say that the period of the growth resurgence in the US was a historical quirk and was not sustainable.

    Here is how it comes out. The TPF contribution from IT for the period of 1995 to 2001 was about 0.4% per year, and in the base case I have assumed that the projection for productivity growth for IT is lower, the optimistic case is higher, and the pessimistic case a good bit lower. But these are important sources of growth. I am going to take the average from 1995 to 2000 because 2001 was a recession period, and that gives me the base case assumption of 0.11 for the whole decade of the 1990s, 0.17 for the optimistic assumption, and the pessimistic assumption would be that we have slow growth of the period before 1995.

    Capital quality growth, again, is the rate at which IT is substituted for non-IT. That is an extremely important part of this story. And it turns out that the average for 1995-2001 was a whopping 2.5% per year. To translate that into the impact on growth, you have to multiply by the capital share, and that gives you about a 1% effect on growth. The base case assumption is that that is going to slow down to something more like an average of the 1990s, whereas the optimistic case is going to be an acceleration like the last half of the 1990s, and the pessimistic case is the period before 1995.

    We can now put it altogether. The demographic projections that I am going to draw on are from the Bureau of the Census. The standard projection is that hours growth reflecting the growth of the working age population is going to be about 1% a year, and you can see here that that is the same across all three of these scenarios and is far below what it was from 1995 to 2001.

    My second assumption is that labor quality growth is going to continue to slow. The proportion of people with higher levels of education is going to stabilize at the levels that have prevailed for new cohorts entering the labor force over the last 20 years. And as the older parts of the labor force retire, we are going to have a slowdown because everybody is going to have basically the same qualifications. From 1995 to 2001 there was a fairly rapid growth in labor quality. If you multiply this by the labor share 0.6, you get the number 0.24, and that is going to be considerably lower going forward. Demographic assumptions give us that hours growth is going to slow and that labor quality growth is going to slow, and now we factor in our assumptions about productivity and capital quality and we obtain the result.

    PART TWO:Projections for the US and Japan

    If you look at labor productivity growth defined as GDP per hour worked, from 1995 to 2001, the number was about 2%, a little higher than that if you just average from 1995 to 2000, again, because 2001 was a period of recession. The appropriate number for the base case is 1.78, so what we would see would be a slightly faster growth of TPF, a slightly slower growth in capital deepening in 1.2, and considerably slower labor quality growth reflecting the slowdown in that factor. How did we get from 1.4 down to 1.2? Why would the base case capital deepening be lower than it was during our historical period?

    The hours worked are growing more slowly, and output and reproducible capital have to grow at the same rate, and that being the case, capital deepening has to slow. If you put this together with our estimate of hours worked at 1%, you see that the base case growth projection for potential GDP in the US is 2.78. That turns out to be precisely what output growth was from 1973 to 1995. In other words, the US economy, despite the vitality of the information technology industry and the rapid pace at which we are going to be substituting IT for other kinds of capital, the growth rate, because of the slowdown in the demographic factors, is going to be precisely what it was before 1995, in other words during the period of relatively slow growth.

    I would like to derive lessons for Japan, and let me summarize the assumptions that I am going to make. First, I am going to use the standard for the working age. I am going to use the official number of -0.55% per year. Remember the corresponding figure for the US? One percent. That means the growth rate of the US economy is 1.5% faster simply because of the difference in hours worked growth. That is extremely important. Demography is not destiny, as someone said, but it is important in the story. Labor quality growth in Japan has been fairly high for quite a period of time, however, and I am going to project that into the future because I do not have any better set of assumptions.

    Here are the base case and pessimistic and optimistic projections. By comparison with 1995-2000, I foresee a sizeable slowdown in the growth of labor productivity. TFP is higher in Japan because TPF in IT grow at the rate that is comparable to that in the US (since both countries have access to the same technology) and there is the ongoing factor of catching up. There will be considerable catch-up if we simply project the TFP growth for non-IT in the past. Because of the slowdown in hours worked, there will be a big reduction in capital deepening. This estimate of capital deepening is far higher than you would get from the official statistics, because of our internationally harmonized prices. Going forward, the capital deepening is going to be considerably less, 0.82. And 0.28 is the labor quality projection. Finally, what is the output growth going to be? About 1.72. About a percentage point below that in the US.

    Is that a good-news or bad-news story? Let us go back to the labor productivity story. US labor productivity growth was going to be about 1.78. That is probably overly precise, but the corresponding projection for Japan is half a percentage point higher. If you are going to focus on per capita output, given the fact that the Japanese population is shrinking in the next decade, this says that there is going to be quite a bit of catching up in terms of labor productivity. Labor productivity is going to converge toward US levels at a pretty hefty rate, about half a percent a year. That is a good-news story.

    But you can say that the bad news is that economic growth itself is going to be one percent below that in the US. I prefer this story, but if you believe in the lost decade, you will probably want to emphasize this one. At any rate, we now have our picture. We have taken the revised version of the lost decade. We have found a decade, at least, in the internationally harmonized prices, where the Japanese economy has already begun to recover.

    What looms over the next decade is the forthcoming demographic downturn, the reduction in the size of working age population at about half a percent per year. And, as the 40-hour work is established and is going to continue and the employment rates are not going to change, you are going to end up with a picture in which Japanese growth has to be considerably less. If you look at it from a more optimistic point of view and turn it into a story about per capita growth, which is of course what is relevant to the Japanese standard of living, then this turns out to be a very optimistic picture indeed in which the Japanese per capita standard of living is going to be rising relative to other industrialized countries at about a half a percentage per year.