Thursday, 10 March 2011

Computers and productivity

As noted in a previous post, Chris Trotter wrongly attributes New Zealand's post-reform increases in MFP growth to the concurrent adoption of information technologies by business in New Zealand:
Have you given any thought to the fact that the period of rapid MFP growth depicted in the graph coincides with the widespread adoption of the personal computer in New Zealand workplaces; the opening up of the Internet from 1992 onwards; and the rapid take-up of the mobile phone as an essential tool of business?
While Chris's timing is wrong he does inadvertently raise the interesting issue of what is the relationship between information technologies and productivity growth. Here the U.S. is an interesting case study.

Alcaly (2003: 20) says,
Whatever else we might wish it where, a new economy is one that has changed significantly through the adoption of innovative new technologies and business practices, leading to a meaningful and sustainable increase in the rate of productivity growth.
Robert Solow famously quipped in a 1987 review of the book “Manufacturing Matters: The Myth of the Post-Industrial Economy” that: “[y]ou can see the computer everywhere but in the productivity statistics.” A remark that has given rise to what is often called the “Solow productivity paradox”. We have computers but where are the productivity gins? Post-1995 the effects of computers finally showed up in the U.S. productivity statistics. In the post mid-90s period paradox seemed resolved. Computers did finally show up in the productivity stats.
It [productivity growth in the United States] finally began to pick up after 1995, rising over the next five years at a rate of more than 2.5 percent a year, almost twice as fast as its pace between 1973 and 1995 and within striking distance of the rates achieved during the golden age of 1948-1973. The surge during the last half of the 1990s raised the average growth rates of productivity and living standards for the entire decade to roughly 2 percent a year, about the same as for the century as a whole. (Alcaly 2003: 37-8)
The average rates of productivity growth in the U.S. for the period 1948-73 was 2.9 percent, for 1974-1995 it was 1.4 percent and 1996 to the third quarter of 2002 it was 2.6 percent. Coyle (2001: 27) explains that “[ . . . ] the improvement [in U.S. productivity growth in the late 1990s] came mainly from greater use of information technology and greater efficiency in its production. Average U.S. growth climbed from 2.75 percent in 1991-95 to 4.82 percent in 1996-99. Of this two-point improvement, 0.5 point come from growth in the input of information-technology capital, 0.9 from other capital and labor input, and 0.6 from increased growth in total factor productivity. The contribution to growth from this measure of technical progress shot up from 0.48 percent a year in the early 1990s to 1.16 percent in the second half of the decade.” So around two-thirds of the mid-to-late-90s acceleration in productivity growth was due to investment in computers, software, networks infrastructure etc along with efficiency gains in the production of computer equipment and semiconductors. By 1996 the new economy had finally arrived. For the U.S. at least.

The productivity surge was not worldwide. As Robert Gordon notes Europe, for example, did not follow the U.S. in having a post 1995 productivity increase, “[ . . . ] since 1995 Europe has experienced a productivity growth slowdown while the United States has experienced a marked acceleration. As a result, just in the past eight years, Europe has already lost about one-fifth of its previous 1950-95 gain in output per hour relative to the United States. Starting from 71 percent of the U. S. level of productivity in 1870, Europe fell back to 44 percent in 1950, caught up to 94 percent in 1995, and has now fallen back to 85 percent.” (Gordon 2007: 176).

But wherever the acceptance of the new economy went scepticism about the causes of the productivity increases was soon to follow. Robert Gordon is one who argues that by themselves computers could not match the effects of the innovations of the past which involved a cluster of new technologies being developed contemporaneously. As an example he points to the combination of innovations which occurred over the period 1860-1900 and resulted in developments such as electricity, air and motor transport, radio and movies and indoor plumbing.

As to the reasons for the apparently small effects and slow appearance of the new economy, in the aggregate data, Coyle (2007: 60-1) offers three observations,
There are several responses to the argument that computers have not been very important for growth. One is that measuring the impact of steam or electricity in exactly the same way as the impact of computers is measured (using the growth accounting described above), you find that steam and electricity look pretty small too: a “small” percentage point difference in growth rates is the statistical footprint of a large economic and social change (Crafts 2004). [ . . . ] A second is that any radical innovation takes a long time to have measurable aggregate impact because people take many years to adjust: perhaps new infrastructure must be built, new skills learned, workplaces reorganized (David 1991). Indeed, many people have an incentive to resist innovations. As Niccol`o Machiavelli put it in The Prince, “Innovation makes enemies of all those who prospered under the old regime, and only lukewarm support is forthcoming from those who would prosper under the new.” And, lastly, although popular attention has focused on computers, there is a cluster of new technologies today, including biotechnology, new materials, and nanotechnology. Their combined impact on our well-being is likely to be just as profound as the cluster of technologies introduced around the start of the twentieth century.
Interestingly, unlike the macro-level data we have just been looking at, micro-level data provides little evidence in support of Solow’s productivity paradox. Pilat (2004a: 11) explains “[s]tudies with firm-level data often find the strongest evidence for economic impacts of ICT.” Recent research on the productivity paradox based on firm-level data suggests that ICT use is beneficial to firm performance and productivity, even for industries and countries where there is no evidence at the more aggregate levels. This result holds for all countries in which micro-level studies have been carried out. For example, Hempell, Van Leeuwen and Van Der Wiel (2004) found that ICT capital deepening increased labour productivity in services firms in Germany and the Netherlands. A close correlation between labour productivity and ICT use was found for Swiss firms by Arvanitis (2004). Maliranta and Rouvinen (2004) looked at ICT use in Finland and concluded there are productivity-enhancing effects associated with ICTs. Baldwin, Sabourin and Smith (2004) found that greater use of ICTs was associated with higher labour productivity growth in the nineties for Canada. Clayton et al (2004) analysed U.K. data and found a positive effect on labour productivity and multi-factor productiviy associated with the exploration of computer networks for trading. U.S. data was used by Atrostic and Nguyen (2002) to demonstrate that average labour productivity was higher in plants with computer networks with labour productivity being around 5 percent higher for such plants.

But the evidence also suggests that turning investment in ICT into higher productivity is not a forgone conclusion - something policymakers in New Zealand should keep in mind - and that to do so requires complementary investments and changes in areas such as human capital, organisational change and innovation. Countries which better support a process of creative destruction, with successful firm growing and failing firm disappearing, are better able to seize the advantages of ICTs.

Pilat (2004b: 56-8) argues there are six reasons why we find a productivity paradox in the aggregate data but do not see it in the micro-level data:
“[f]irst, aggregation across firms and industries, as well as the effects of other economic changes, may disguise the impacts of ICT in sectoral and aggregate analysis. This is also because the impacts of ICT depend on other factors and policy changes, which may differ across industries. The size of the aggregate effects over time depends on the rate of development of ICT, their diffusion, lags, complementary changes, adjustment costs and the productivity-enhancing potential of ICT in different industries (Gretton et al., 2004). Disentangling such factors at the aggregate or industry level is not straightforward.

Second, the firm-level benefits of ICT in many OECD countries may not yet be large enough to translate into better outcomes at the aggregate level. The firm-level benefits may be larger in the United States (and possible also in Australia) than in other OECD countries, and thus show up more clearly in aggregate and sectoral evidence. For example, Haltiwanger et al. (2003) suggest that the impacts of ICT are smaller in Germany than in the United States. Given the more extensive diffusion of ICT in the United States, and its early start, this interpretation should not be surprising. This is particularly the case if it takes time before the benefits from ICT become apparent, e.g. because of high costs of adjustment to the new technology. Moreover, the conditions under which ICT is beneficial to firm performance, such as having sufficient scope for organisational change or process innovation, might be more firmly established in the United States than in many other OECD countries. Small firm-level benefits in most OECD countries might thus lead to relatively small productivity benefits at the aggregate level.

Third, firms that are successful in implementing ICT may be better able to gain market share and grow in a competitive market such as the United States than in less competitive markets. This would contribute to greater overall impacts of ICT in the United States. For example, some of pick-up in US productivity growth over the second half of the 1990s can be attributed to the growth in market share of Wal-Mart, a company that replaced many less efficient retailers, partly owing to its effective use of ICT throughout the value chain. If the most efficient firms in Europe find it difficult to expand and gain market share, even if they do benefit from ICT, the overall impacts on productivity might be more limited than in the United States.

Fourth, measurement may play a role. The impacts of ICT may be insufficiently picked up in macroeconomic and sectoral data outside the United States, due to differences in the measurement of output. For example, the United States is one of the few countries that have changed the measurement of banking output to reflect the convenience of automated teller machines. Since services sectors are the main users of ICT, inadequate measurement of service output might be a considerable problem.

Fifth, countries outside the United States may not yet have benefited from spill-over effects that could create a wedge between the impacts observed for individual firms and those at the macroeconomic level. The discussion above has already suggested that the impacts of ICT may be larger than the direct returns flowing to firms using ICT. For example, ICT may lower transaction costs, that can improve the functioning of markets (by improving the matching process), and make new markets possible. Another effect that can create a gap between firm-level returns and aggregate returns is ICTs impact on knowledge creation and innovation. ICT enables more data and information to be processed at a higher speed and can thus increase the productivity of the process of knowledge creation. A greater use of ICT may thus gradually improve the functioning of the economy. Such spill-over effects may already have shown up in the aggregate statistics in the United States, but not yet in other countries.

Finally, the state of competition may also play a role in the size of spill-over effects. In a large and highly competitive market, such as the United States, firms using ICT may not be the largest beneficiaries of investment in ICT. Consumers may extract a large part of the benefits, in the form of lower prices, better quality, improved convenience, and so on. In other cases, firms that are upstream or downstream in the value chain from the firms using ICT might benefit from greater efficiency in other parts of the value chain. In countries with a low level of competition, firms might be able to extract a greater part of the returns, and spill-over effects might thus be more limited.
So Chris Trotter raises an important question about the relationship between information technologies and productivity growth, even if he doesn't have an answer. The answer largely depends on what data you are looking at. We see Solow's productivity paradox in the macro-level data but the micro-level data provides little evidence in support of the paradox.

No comments: