Economics 201

Case of the Day: Information Technology and Productivity

You can see the computer age everywhere but in the productivity statistics. (Robert Solow, New York Times Book Review, July 12, 1987)

When I was a college student 40+ years ago, telephones had cords and minutes of long-distance connection were expensive, the campus computer had its own building at which we assembled decks of punch-cards onto which we had punched our programs, theses and course papers were written on typewriters that didn't identify spelling errors or allow us to fix mistakes easily, and VCRs (let alone DVDs or DVRs) didn't exist--we watched what was being broadcast on television at the moment or went to a movie. Since then, we have undergone a technological revolution built on an ever-growing array of electronic wizardry that has made our lives better and made us more productive.

Few would argue with the "made our lives better" part of this statement but as the Solow quotation above suggests, the "made us more productive" part was a puzzle for a long time. Although it seemed that modern electronics were being applied to the production process everywhere, the data on productivity didn't indicate for a long time that we were becoming any more productive.

This case examines the empirical evidence about productivity growth since World War II with special emphasis being placed on the role of information technology. The broad facts are these:

• During the "golden age of economic growth" after World War II, the U.S. economy grew steadily with advancements in technology apparently raising productivity rapidly and steadily.
• After about 1973, growth in productivity slowed down considerably in the United States and other rich countries. This slow growth continued for about two decades, despite the increasing penetration of marvelous new electronic communication and data-processing equipment.
• A period of strong productivity growth began in the mid-1990s and continued into the 2000s, before the Great Recession hijacked the macroeconomy in 2008.
• During and since the economic recovery after the Great Recession, productivity growth has been tepid at best, even in places such as the United States that have recovered robustly.

Before examining the data, we consider the method of "growth accounting" by which economists attempt to estimate growth in productivity and its causes.

Measuring productivity: Growth accounting

The theory of economic growth tells us that improvements in technology are the driving force behind long-run improvements in per-capita GDP and material standards of living. However, there is no good way to measure "technology" at an aggregate level. It consists of millions of individual improvements in the methods of production at firms ranging from tiny to large. We can collect data on some variables that are probably correlated with "aggregate technological progress" such as R&D expenditures and patents issued, but we cannot measure progress directly.

Because we cannot measure the technology's effects on output directly, we have to measure them through the back door: whatever gains in output we cannot attribute to other causes must (or at least may) have been caused by technological progress. Thus, we measure the effect of technology as the "Solow residual"—the part of economic growth that is left unexplained after we remove all of the other measurable causes.

The most obvious "other causes" that we need to remove are the effects of increased physical quantities of labor and capital. If we think of an aggregate production function given output (Y) as a function of aggregate inputs (K and L) and the state of technology (A), then we can write Y = F(K, L, A). Our goal in growth accounting is to measure the contribution of A to the growth of Y, so we need to extract out the effects of growth in K and L. Eliminating the effects of K and L on the growth in Y gives us a measure of "total-factor productivity" (TFP), which is the basis for measuring technological progress. (Note that TFP is a better potential measure of technological effects than labor productivity Y/L, though the latter is sometimes used because it is much easier to measure. If there is no change in technology, then an increase in the capital stock would increase Y/L, but not TFP.)

But while the task of growth accounting seems relatively simple, the actual application is very complex. For example, we shouldn't just count up worker-hours to measure labor input, we should also take account of increases in the amount of human capital that those workers have. (Human capital is traditionally counted as part of labor input rather than capital.) Similarly, many new technologies are embodied in new kinds of capital, making it difficult to distinguish the increase in capital from the effects of advancing technology.

Of course, just taking out the effects of growth in labor and capital inputs does not mean that the remaining growth in "output per unit of input" is necessarily due to technological progress. If there are economies of scale in important industries then growth in production would lead to higher TFP even with unchanged technology. Similarly, if the economy eliminates inefficiencies by improving its resource allocation (perhaps moving redundant farm workers to more useful jobs in the cities) then TFP will increase as the economy moves out toward the production-possibilities frontier. The effects of these factors on TFP growth are very difficult to estimate, as are other non-technology effects such as changes in environmental regulations and effects of the weather.

Among the pioneers in the field of empirical growth accounting are Edward Denison, John Kendrick, and Angus Maddison. Their methods vary somewhat, but all of them follow the basic strategy of identifying TFP growth as a residual between the growth of output and the weighted growth rates of the inputs. Dale Jorgenson has recently produced remarkable detailed accounts of the effects of information and communications technology on growth. We shall examine a few numbers from these authors below.

In his Trends in American Economic Growth, 1929-1982 (Brookings Institution Press, 1985), Denison decomposes real output growth into components due to inputs and productivity factors. Over the 1948-73 period, which has been characterized as the "Golden Age of Economic Growth," U.S. national income grew at an average rate of 3.89 percent per year. Of this, Denison estimates that 1.46 percentage points were due to growth in labor input (including education and other effects), 0.77 percentage points were due to capital growth, and total-factor productivity grew at 1.66 percent per year.

The Golden Age ended some time around 1973, the year in which the Arab oil embargo occurred. Denison's estimates for 1973-82 show a 2.61 percent per year growth in real output, of which 1.86 percentage points were due to labor input, 0.67 points due to capital input, and a measly 0.08 percent per year increase in TFP. The decline in TFP growth from 1.66 percent to 0.08 percent per year is dramatic; it has been called the Great Productivity-Growth Slowdown.

The slowdown in productivity growth was not limited to the United States. Maddison in his book Dynamic Forces in Capitalist Development (Oxford University Press, 1991) presents estimates for the U.S. and five other advanced countries for the periods 1950-73 and 1973-87. Maddison's estimates of total-factor productivity growth in these two periods are shown in Table 1 below. All countries except the United Kingdom experience a large slowdown in productivity growth after 1973. (Remember that the UK became an oil producer with the emergence of the North Sea fields during this period. It was also the period in which Margaret Thatcher's government liberalized the economy on many fronts.)

Table 1. Maddison's cross-country estimates of TFP growth

This apparent change in the behavior of TFP stimulated economists' interest in the determinants of productivity growth and has led to a resurgence of interest both in growth theory and (together with the release of a wonderful new data set) in empirical investigation of the proximate correlates of growth.

Information technology and productivity growth

Productivity growth continued to be sluggish throughout the 1980s, which surprised many economists and others because this was a period of explosive growth in computer technology. The IBM personal computer was first released in 1981, followed three years later by Apple's Macintosh. Fax machines were spreading rapidly to facilitate the quick transmission of printed information, and electronic mail began to emerge from obscurity into common use in the late 1980s.

Despite all these amazing new technologies that appeared to enhance business productivity, the TFP numbers didn't show any significant acceleration in growth. Solow's quip quoted at the top of this case reflects the puzzlement among economists at the apparent contradiction between observation of business practice and the hard, cold numbers.

However, the delay in the productivity effects of revolutionary innovation would not surprise many scholars of the economic history of technological innovation. Over the course of technological history, new innovations have always required lengthy periods—often several decades or more—before their contributions to production achieved significant reductions in resource costs. This was particularly true of other revolutionary innovations such as the steam engine and the electric motor. It simply takes a long time for firms to recognize how these technologies can best be used and how to lower their costs with them. Moreover, the technologies themselves undergo massive refinement during their first decades to make them more useful to firms. Predictably, perhaps, just as economists began to puzzle over the absence of an effect of computers on productivity, the effect began to show up.

The most authoritative studies of the effects of information technology (IT) on productivity growth have been done under the direction of Dale Jorgenson (Reed, '56). In his 2001 presidential address to the American Economic Association, Jorgenson presented detailed estimates of TFP growth and of the contribution of IT to TFP growth for various periods. By Jorgenson's estimates, shown in Table 2 below, overall TFP growth before 1973 was driven almost completely by not IT sources, but since 1973 and especially during the re-emergence of rapid growth since 1995, IT has been leading the way.

Table 2. Jorgenson's estimates of effects of IT on TFP growth

Evidence for the period since 1999 suggests that labor productivity growth continued to be strong up until about 2004, then plummeted in 2006-08 and has been very low since 2010 after a brief recovery. The diagram below shows the pattern in labor productivity in the non-farm business sector since 1968:

Economists' understanding of the determinants of productivity growth leaves much to be desired. Why did information technology seem to add so much to productivity in the 1990s, but so little 20 years later?

Questions for analysis

1. Why might the oil embargo and the resulting large oil-price increases lead to a reduction in the growth of total-factor productivity? Is it possible for TFP growth to be negative in such a situation (as some estimates suggest)? How might this happen? (Hint: Think about the necessary re-allocation of resources after an oil-price increase and the utilization of oil-intensive capital goods.)

2. Do you expect rapid productivity growth associated with IT technology to continue in the coming decade? Why or why not?

3. The IT revolution has been compared to a new industrial revolution in that it has led not only to rapid productivity advances, but to new ways of organizing the workplace and society in general. To what extent do you think that modern electronic technology can be reasonably called a revolution?

Answer the questions in Moodle