3.2.2Long Learning Lags

This argument is mainly due to David (1990). It states that it takes time for technological revolutions to produce significant effects as reported by traditional macroeconomic indicators. For instance, David showed that productivity growth did not accelerate until forty years after the introduction of electric power in the early 1880s. Part of the reason is that it took until 1920 for at least half of American industrial machinery to be powered by electricity. This was also the time needed to re-organize businesses around electric power. David argued a technology starts to have significant effects only when it has reached 50% penetration rate. The Economist (2000) suggested that this level of diffusion was reached only recently for IT capital in the United States. In fact, labor productivity has increased to an annual average 2.9% since 1996, from 1.4% on average between 1975 and 1995. In the second quarter of the year 2000 this rate was estimated at 5.2%. These facts tend to support this long learning lags hypothesis. Thus, because it takes time for new technologies to produce visible aggregate effects, comparing current costs with current benefits might not show high returns to IT investment. Brynjolfsson and Hitt (1993) gave a numerical example, starting by assuming it would take thirty years for IT capital stock to represent 100% of the current level of gross national product (GNP). Then, if returns to investment in IT are 20%, then GNP should increase by 20% over thirty years, which means only 0.06% a year.

Furthermore, rapid technical progress in the computer-producing industry makes information technologies change rapidly. IT-users have to upgrade constantly not only their equipment but also their skills at the same time. Not only computers do things faster they also do it in different ways that are constantly evolving. It takes time to learn new techniques. Indeed, a commercial for a famous computer brand states that the average user exploits only 30% of computers’ capabilities. One reason for this “under utilization” may be the difficulties for users to understand the language and communicate with their computers, and the time needed to receive benefits from the use of information technologies fully.

On the other hand, using computers is becoming easier. For instance, the Windows-type operating system certainly facilitated the use of computers compared to the previous DOS-type system. Similarly, Internet editing software avoids the use of the complicated HTML language. Hence, over time, it becomes easier to exploit new technologies. However, computers might remain relatively complicated to use for a large part of the labor force. Long learning lags are then necessary for improvements made by computers and information technology to appear in national productivity statistics. Depicting a brighter picture for the future, Powell (2000) argued:

‘Today’s college students were born at the same time as PCs, and they’ll enter the workforce having grown up with them as part of their landscape. For them, there’s no transition to computer technology. It’s always been there and they’ve always used it. In the hands of a generation for whom computer technology is less of a novelty and more of a given, and who have no outdated work habits to break, the promise of computerized productivity can finally be realized.’

Economists have observed that the adoption of new technologies is usually slow. According to Greenwood (1999) it is regulated by two interrelated factors that form a feedback loop: the speed of learning and the speed of diffusion. On one hand, the harder it is for users to learn about a new technology, the slower its diffusion. On the other hand, the faster its diffusion, the easier it may be to learn about it. This may be a reason for some of today’s new marketing techniques. A few years ago, to prevent espionage, manufactures would keep an innovation secret until the date it was first publicly sold. New technologies are now introduced to the public with commercials and samples before they even are officially sold (DVD copiers for instance). Apart from marketing strategies, manufactures may want to test the market and get early feedback from potential consumers in order to refine the new product, which will then be able to diffuse more rapidly, and therefore produce benefits faster.

Price decline is also the engine of diffusion. At the beginning, a new technology needs enormous investments before it can be sold to the public. Its price is therefore very high. But, over time, manufactures learn how to produce more efficiently and begin to achieve economies of scale. In addition, new competitors start entering the market. All these factors drive the price of the new technology down, accelerating its diffusion at the same time. Still, Roach (1998), a productivity paradox advocate, emphasized the differences between previous revolutions and the breakthrough in IT capital. Previous innovations were made for tangible goods whereas computers are dealing with intangible output. The comparison with past technological revolutions should therefore be made cautiously.

Thus, the long learning lags hypothesis seems to be a valid explanation for the productivity paradox, supported by the fact that most authors now seem to agree that IT capital has proven its productive capacity since the late 1990s. Still, another argument explaining the productivity paradox refers to the mismanagement of information technologies.