Download Those elusive productivity gains

Productivity: lost in cyberspace
Office equipment|capital spending and productivity; Technology|return on
investment;Computers|return on investment
13-Sep-97 798
Productivity: lost in cyberspace
IF THIS is an era in which economies can enjoy rapid growth and low unemployment with no risk of inflation,
information technology gets much of the credit. Much of the case for a new economic paradigm rests on the belief
that computers and telecoms have boosted the growth of productivity and thereby removed the speed limit on
economic growth. In the words of George Gilder, a noted techno-optimist, information technology is “the biggest
technological juggernaut that ever rolled”.
Big it certainly is. Last year spending on computers and telecommunications accounted for two-fifths of American
firms’ total investment in capital equipment. Yet if official figures are to be believed, the computer revolution has
not made the economy more efficient. The annual average rate of productivity growth in America’s business sector
has slowed from 2.6% in 1960-73 to around 1% in more recent times. This apparent contradiction—speedier
computing and communications along with slower productivity growth—is known to economists as the
“productivity paradox_ 8;. Its existence leaves a gaping hole in the “new economy” debate.
One gloomy view is that much spending on technology has been a waste. Paul Strassmann, the author of a recent
book called “The Squandered Computer”, claims that computer systems are often not subject to proper investment
appraisal before they are purchased, and that they are badly used. Mr Strassmann finds no correlation between
spending on information technology and profitability in any industry.
There is clearly some merit in this argument. Many firms have made unwise investments in computers or use them
in ways that bring little obvious benefit to their bottom lines. In addition, as computers have sprouted on every
desktop, they can easily be used for less productive tasks, such as playing Minesweeper.
But the productivity paradox may be explained in less troubling ways. Economic history suggests that there are
often long time lags before technological breakthroughs deliver productivity gains. For instance, the electric dynamo
was invented in 1881, but it took 40 years before companies learned how to reorganise their factories efficiently
around electric power to take advantage of its flexibility. Likewise, 26 years after the microprocessor was introduced
firms are still learning how to employ it. The economic benefits of computers may lie in the future.
Yet there is already some evidence that computers boost productivity at the individual company level. Two
American economists, Erik Brynjolfsson and Lorin Hitt, have calculated an annual return on investment in
computers of around 50%, well above the average return on capital. Likewise, Alan Greenspan, the chairman of the
Federal Reserve Board, has recently pointed to the increase in corporate profit margins as evidence of faster
productivity growth.
Why does this not show up at the economy-wide level? One possible answer is that computers matter less than we
think. In another new book, “The Computer Revolution”, Daniel Sichel, a Federal Reserve economist, claims that
computers account for only 2% of America’s total capital stock. Even if the return on computers is higher than on
other investments, he argues, their contribution to overall economic growth will be modest. The high level of
corporate spending on computers exaggerates their importance: much of the money goes to replace equipment which
rapidly becomes obsolete, so the net amount of new investment in computers is relatively small.
Mr Sichel’s measure, however, understates the full impact of information technology because he focuses only on
computers themselves. Add in all the equipment used for gathering, processing and transmitting information, and the
total accounts for 12% of America’s capital stock, exactly the same as the railways at the peak of their development
in the late 19th century, when they were providing a significant boost to America’s economy.
The second and most persuasive argument for why productivity gains do not show up in the economy as a whole is
mismeasurement. Three-quarters of all computers are used in the service sector, such as finance and health, where
output is notoriously hard to measure. And many of the benefits arrive not as cost savings or increased output, but in
the form of improvements in quality and convenience. Replacing some bank tellers with automated teller machines
gives bank customers 24-hour-a-day access to their money, but this does not show up in national accounts. Indeed, if
the machines cause customers to write fewer cheques, measured productivity will fall, not rise.
Thus there are strong grounds for believing that computers have boosted productivity growth. For the moment,
however, this has to be taken on trust. Faster productivity growth is certainly to be welcomed, as it would make it
possible for the economy to expand faster before inflation takes off. But this does not mean that the economy has
shaken off the fetters that limit its growth, only that those fetters may be a tad looser than before.
A Survey of the world economy:
The hitchhiker's guide to cybernomics - Paradox lost (part 3 of 11)
Information technology|innovation and productivity; Computers|productivity, elusive gains;
Productivity|computers, elusive gains; Labour markets|productivity and information technology;
Statistics|productivity measurement, innovation effect; Inflation|measurement difficulties with
new products;
28-Sep-96 2609
Those elusive productivity gains
OVER the past two decades America's investment in computers has risen by 20-30% a year in real terms. The share
of IT in firms' total investment in equipment has jumped from 7% in 1970 to over 40% this year (see chart 5). Add
in the billions of dollars that companies spend on software and spending on IT now exceeds investment in traditional
machinery. This investment should in theory have lifted productivity and boosted growth. Yet productivity gains in
the big industrial economies have slowed sharply since the early 1970s.
Labour-productivity growth in the big seven economies which in 1960-73 averaged 4.5% a year has been a mere
1.5% since then (see chart 6). Growth in total factor productivity (output per unit of labour and capital combined)
over the same period slowed from an average of 3.3% to 0.8% a year. In other words most growth since the mid1970s can be explained simply by increased inputs of labour and capital; the contribution from technological
progress appears to have been tiny. Robert Solow a Nobel-prize-winning economist once remarked that 'you can see
computers everywhere but in the productivity statistics'. The apparent absence of a productivity boost from the new
technology has come to be known among economists as the productivity paradox.
Despite hundreds of studies the dismal scientists remain deeply divided on why the computer revolution has failed to
spur productivity. One possible if depressing explanation is that there has been no revolution and that computers are
simply not particularly productive. Paul Krugman an economist at MIT and never one to dodge controversy argues
that recent technological advances are not in the same league as those achieved earlier this century. Looking back to
the 1950s and 1960s when productivity surged he points out that changes in technology then affected every aspect of
life. In 1945 crossing America by train could take three days and groceries were bought in mom-and-pop stores; by
1970 the journey from one end of the country to the other took five hours by plane and groceries came from big
efficient supermarkets. By comparison he claims IT has less effect on the average person's life. 'Computerised
ticketing is a great thing but a cross-country flight still takes five hours; bar codes and laser scanners are nifty but a
shopper still has to queue at the checkout.'
Mr Krugman has a point. Just because today's computer is 50 times more powerful than it would have been ten years
ago it will not necessarily be 50 times more productive. And as computers become cheaper people sometimes use
them for less-than-vital tasks such as playing solitaire. One reason why spending on computers is high is that
equipment is replaced ever more frequently by more powerful newer models long before it is worn out.
Also investment in new technology may not always be intended to boost productivity. In a recent study 'Technology
Productivity and Jobs' the OECD points out that a growing slice of spending on R&D and IT is devoted to product
differentiation and marketing in a battle for market share not to making existing production more efficient.
21 st century dynamo
While there may be an element of truth in the argument that some IT spending is 'wasteful' it cannot be the whole
story. A second more hopeful explanation for the paradox draws upon history which shows that there is often a
delay of several decades before technological breakthroughs deliver economy-wide productivity gains. Firms take
time to identify the most efficient way to use new technology and to make organisational changes. It is the wide
diffusion of a technology rather than its invention that brings the biggest benefits.
Paul David an economic historian at Stanford University explained in a classic study* how the introduction in the
early 1880s of the electric dynamo (which opened the way for the commercial use of electricity) took 40 years to
yield significant productivity gains. Growth in productivity in industrial economies actually slowed down after 1890
and did not revive until the 1920s. This partly reflected the slow rate of adoption of electricity. In 1899 electricity
accounted for less than 5% of power used in American manufacturing; only in 1919 did its share reach 50%. And
even when firms had installed electricity it still took them a long time to learn how to organise their factories around
electric power and to take advantage of its flexibility. Previously machines had had to be put next to water wheels or
steam engines; electricity allowed them to be placed along a production line to maximise the efficiency of the work
flow.
In the same way spending lots of money on IT is not enough; businesses also have to learn how to use it efficiently.
When computers first appeared in offices in the 1970s they were used to automate existing tasks such as typing. It
took time for managers to understand that computers not only allowed them to do the same things differently but
also to do completely different things. Moreover about three-quarters of all spending on IT has gone into service
industries such as telecommunications and financial services. Until recently these were often sheltered from
competition so had little incentive to use IT efficiently.
Although the computer was invented 50 years ago the computer revolution did not begin in earnest until Intel
introduced the microprocessor in 1971. The IBM PC was not introduced until 1982. If it took 40 years before
electric power started to show productivity gains perhaps it is not surprising that the benefits of computers have yet
to materialise. On the other hand thanks to plunging prices computers are being diffused through the economy more
rapidly than electricity was. About half of all American workers now use some form of computer (up from a quarter
in 1984). Since Mr David's research seems to suggest that productivity benefits start to be reaped once the diffusion
rate of a technology passes 50% the productivity gains may be just around the corner.
Indeed at the company level rather than the level of the economy as a whole there is already evidence that computers
boost productivity. Erik Brynjolfsson and Lorin Hitt at MIT in a study*** which examined 367 big American
manufacturing and services firms between 1988 and 1992 found that investment in computers yielded an average
return of more than 50% a year.
Lost in a statistical black hole
If the return is so good why does it not show up in the economy as a whole? The explanation say some is that
computers account for only 2% of America's total capital stock. By contrast at the peak of the railway age in the late
19th century railways accounted for 12% of its total capital stock. On the surface this might provide a third
explanation for the productivity paradox: that the capital stock of computers at present is too small to make much
difference to overall productivity growth. However the 2% figure covers only computers. Add in
telecommunications equipment and computer software and the total approaches 12% the same as for railways.
Moreover if as many believe the IT revolution is still in its early stages with some sectors still barely touched by it
then as it spreads through the economy so will its potential to raise productivity.
According to a fourth explanation of the productivity paradox (and perhaps the most persuasive of all) the benefits
are already coming through but standard economic statistics are failing to capture them. The tools used for
measuring productivity are more suited to the output of 19th-century dark satanic mills than 21st-century electronic
wizardry.
Zvi Griliches an economist at Harvard University points to a clue to the productivity paradox: the slowdown in
productivity growth has largely been in the service sector where productivity is notoriously hard to measure. In
manufacturing where the statistics are more reliable productivity growth has held up rather better. This is significant
because many service industries have invested proportionately more in IT than manufacturing. In air transport
telecommunications retailing health care banking and insurance investment in IT has been equivalent to an average
of almost 6% of output compared with under 3% in manufacturing. According to Mr Griliches the measurement
problem is twofold: first economies are shifting towards services which have always been tricky to measure; and
second the nature of the gains from IT often makes them hard to quantify.
In many services it is hard even to define the unit of output partly because higher output often comes in the form of
quality improvements. In areas such as finance health care and education government statisticians typically assume
that output rises in line with the number of hours worked. The bizarre effect is that measured productivity growth is
zero by definition. Likewise telecommunications output is measured in minutes of calls leaving out the huge
increase in information transmitted via faxes or faster modems. Or suppose a road-haulage firm introduces a
computer system which helps drivers to pick the quickest route and thus provide a better service for its customers. If
mileage drops as a consequence the official statistics will show a fall in real output.
The share of the economy that can be measured with any degree of accuracy is shrinking. In 1947 calculates Mr
Griliches about half of America's output was in agriculture mining and manufacturing and so readily measurable. By
1990 that share had declined to only 30%*. And yet the other 70% of output contains some of the fastest-growing
sectors.
One of the biggest problems is calculating price changes which are used to deflate increases in nominal output
between two periods to obtain changes in the volume of output. This is relatively easy for basic commodities such as
wheat or steel but much harder for goods and services where new and better products are continuously emerging. A
tonne of steel is much the same as 20 years ago but the quality of a television set say will have improved vastly so
the increase in price needs to be adjusted for the higher quality.
Official price indices overstate the rate of inflation and hence understate growth in a number of ways. For instance
government statisticians tend to delay the inclusion of new goods or services in the consumer-price index until they
have been on the market for several years thereby leaving out the rapid fall which usually takes place in the early
years of a new product. When video-cassette recorders first came on the market in the late 1970s they sold for
around $ 1000. By the mid-1980s the price had fallen to $ 200 and many new features such as electronic controls
and programming capabilities had been added. Yet it was not until 1987 that they were included in America's
consumer-price index so the huge drop in price in their first decade went unrecorded. In the same way personal
computers were not included until 13 years after the first home computer was launched.
Another example is that the price of medical services is based on the cost of inputs such as doctors' fees and hospital
charges. But this fails to reflect improvements in treatment which reduce the consumption of these inputs. For
example in the early 1960s a typical cataract patient would spend seven days in hospital whereas now the operation
can be done in a day as an outpatient. None of this shows up in the official statistics.
Riches beyond measure?
A study*** by Leonard Nakamura an economist at the Federal Reserve Bank of Philadelphia estimates that
America's inflation rate has been overstated by an average of two to three percentage points a year since 1974
mainly because new products or product improvements have been neglected. If this is correct then productivity
growth in the rich industrial economies has been much higher than the official figures suggest.
This would matter less if new products and quality changes were introduced at a constant pace. The overstatement of
inflation (and hence the understatement of growth) would be constant over time so it would not distort relative
productivity growth between two periods and hence could not explain the productivity paradox. However there is
good reason to believe that the degree of mismeasurement has worsened as product cycles have shortened. Mr
Nakamura estimates that the mismeasurement of inflation has been 1.7 percentage points a year higher since 1974
than in the previous 15 years. The greater understatement of growth that this implies would be almost enough to
explain all of the apparent slowdown in America's productivity growth in the past two decades. And if
mismeasurement has been getting worse over the period then true productivity growth may now be surging not
sagging.
The mismeasurement problem becomes more acute as output shifts to goods and services where quality is an
important factor. Henry Ford offered his customers a car of any colour as long as it was black. But IT makes it
economical to provide greater choice even to tailor goods to the needs of individual customers. A large share of the
benefits of IT comes through not as cost savings but in the shape of quality improvements wider consumer choice
better customer service time savings and convenience. These are important to consumers but difficult to measure.
For example 24-hour automatic teller machines have clearly made life more convenient for banks' customers but this
added convenience does not show up in the national accounts. On the contrary as fewer cheques are processed
measured productivity appears to fall.
Many American economists believe that revisions made earlier this year to the methodology used to calculate
America's GDP (a shift from fixed to annually adjusted weights) although defensible in themselves have had the
effect of understating growth even more than before. Even if the number-crunchers are trying getting the statistics
right is clearly difficult. One reason is that government statistical agencies in many countries have been starved of
funds. And even where the money is available it is often badly allocated. The budget for America's national accounts
statistics for example is less than one-third that for agricultural statistics.
However an even bigger reason for the inadequacy of official statistics may be that the economy is simply becoming
unmeasurable. Conventional methods of measuring the economy are no longer up to the task and economists cannot
agree on how to improve them. It does not help that within the profession economic statistics is considered a less
exciting subject than economic theory so it fails to attract as much talent as it needs.
But in any case whatever the statisticians may eventually come up with it is misleading to judge the value of IT by
comparing productivity growth in the past two decades with that in the 1950s and 1960s. That was an exceptional
period when growth in many countries was making up for time lost in the 1930s depression and the second world
war. Taking a longer view recent growth has been quite respectable (see chart 7). Allow for the fact that
mismeasurement is now much worse than earlier this century and the rich economies' output per head may well be
growing more swiftly than in any previous technological revolution. But what does that mean for jobs?
* 'The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox'. By Paul David.
American Economic Review May 1990
*** 'Information Technology as a Factor of Production: The Role of Differences among Firms'. By Erik
Brynjolfsson and Lorin Hitt. Economic Innovation and New Technology 1995 vol. 3
* 'Productivity R&D and the Data Constraint'. By Zvi Griliches. American Economic Review March 1994.
*** 'Is US Economic Performance Really that Bad?'. By Leonard Nakamura. Fed. Res. Bank of Philadelphia
working paper No 95-21
© 1996 The Economist Newspaper Limited. All rights reserved
A Survey of the world economy: The hitchhiker's guide to cybernomics - A game of monopoly?
(part 8 of 11)
Information technology|monopolistic concentration, theory of increasing returns; Economic
theory|increasing returns and the technology age; Companies|monopolistic concentration, theory
of increasing returns
28-Sep-96 1790
The hitchhiker's guide to cybernomics - A game of monopoly?
Perhaps; but a lot more competition as well
IN 1958 the Harvard Business Review predicted that computers would lead to a greater concentration of power in
American business because they would allow bosses to keep better track of information within large firms.
Eventually it predicted the economy would be dominated by a few giants. In 1967 the economist JK Galbraith
argued that new technology would inevitably lead to increasing dominance by big corporations immune from market
forces. These predictions have so far turned out to be wrong: the average size of firms has shrunk and competition
has increased since the 1960s. But now the Cassandras are at it again. According to a recent article in Wired
magazine 'the economics of monopolistic competition is the economics of the technology age'.
Brian Arthur an economist at Stanford University is the leading proponent of the so-called theory of increasing
returns*. He argues that in a growing number of industries there is a natural tendency for the market leader to get
further ahead causing a monopolistic concentration of business.
By contrast says Mr Arthur traditional economics works on the assumption of diminishing returns. As a firm
expands the theory suggests it eventually hits a limit where costs per unit of output start to rise and unit profits fall.
For example a farmer will use his best land first; if he wants to increase his output he will have to use less and less
productive land so he runs into diminishing returns. In the same way as a manufacturing firm expands it will hit
some limit such as the capability of its management or the size of its regional market beyond which unit costs rise.
Each producer will expand his output until he hits increasing unit costs. No one firm can corner the market so
competition thrives. Mr Arthur agrees that traditional industries such as wheat and steel are subject to diminishing
returns but insists that the new knowledge-based industries are different.
According to Mr Arthur these businesses tend to have three things in common. First they have high fixed costs such
as R&D but low variable costs. For example the cost of writing a computer program is the same regardless of the
number of copies sold so the higher the sales the greater the profit margin. The same is true for many other
industries that are heavy on know-how and light on material resources from pharmaceuticals to defence.
A second common characteristic is what Mr Arthur calls 'network externalities'. In the software industry for example
this means that the more widely an operating system is used the more likely it is to become a standard for the
industry and the more people will want to use it to ensure that their software is compatible with that of other network
users. This makes it harder for rivals to compete.
The third factor which strengthens a leader's grip on a market is the customer lock-in effect. Many high-tech
products are difficult to use so once a customer has learnt how to use a computer program say he is loth to switch. If
all three factors are present Mr Arthur argues increasing returns will magnify the market leader's advantage. By
cutting its price a leader can grab a bigger share of the market earn bigger profits and spend more on R&D than its
rivals sharpening its edge even further.
Economists have been aware of the possibility of increasing returns for a long time but believed them to be rare in
practice. However IT and the general shift of economies from processing tangible goods to processing information
and ideas is now introducing increasing returns into a growing number of industries according to Mr Arthur. In hightech businesses heavy R&D costs and network and lock-in effects are common. IT and globalisation also allow
firms to sell cheaply into a bigger global marketplace and so exploit bigger economies of scale. Mr Arthur reckons
that banking and insurance are becoming subject to mild increasing returns. As banks move to low-cost computers
and on-line networks to process their customers' business he explains variable costs will plunge. The bank with the
largest customer base will be able to spread its fixed costs more widely than its competitors and therefore offer the
best interest rates which will enable it to attract yet more customers.
Big Brother is (not) watching
In a world of increasing returns firms could in theory grow without limit amassing enormous monopoly power. The
lack of competition could push up prices and stifle technological development. Luckily the reality does not match
the model. True a firm such as Microsoft can conquer a market at least for a time by exploiting increasing returns
but its ability to abuse its power is limited.
Although the size of firm at which diminishing returns kick in may have increased in some industries there are good
reasons to question the idea that increasing returns are becoming widespread. For one thing the advantage of the
lock-in effect is tempered by the rapid pace of innovation which favours the emergence of new products and thereby
opens the door to new entrants. If a market leader becomes complacent and jacks up its prices or neglects R&D it
risks being shoved aside by rivals with better products. Moreover concentration of output need not be harmful so
long as barriers to entry are low: the mere threat of competition can make a firm behave competitively. Microsoft
has a near-monopoly in PC software but it remains innovative because its markets are contestable. Falling
communications costs and the Internet are helping to lower barriers to entry in many markets. In the past only big
companies could afford big data networks and a global marketing presence; now tiny firms can have the same
access.
Another factor favouring small firms is that new technologies tend to be less lumpy than old ones. Unlike car plants
or steel mills new businesses in knowledge-based industries can often be set up at minimal capital cost: just a PC
and a telephone. IT also brings down the minimum size a firm has to be to operate profitably by making overheads
more divisible. Legal and accounting expertise for example is now available relatively cheaply on-line.
A study by MIT economists Erik Brynjolfsson and Thomas Malone*** found that the size of the average American
firm whether measured by number of employees sales or value added had shrunk since the 1970s and that firms that
invested most in IT tended to be smaller than others. Over the past two decades the average number of employees
per firm in America has fallen by a fifth. An analysis by the OECD shows that in most rich industrial countries the
average firm size has declined in most sectors except for computer services drugs and supermarkets. This is not
because workers have been displaced by computers but because firms have narrowed their focus. Computers make it
cheaper to farm out work to specialists than to do everything under one roof.
In theory computers could wipe out the need for firms in the traditional sense altogether. In 1937 Ronald Coase a
Nobel-prize-winning economist asked why workers were organised in firms instead of acting as independent buyers
and sellers of goods and services at each stage of production. He concluded that firms were needed because of the
lack of information and the need to minimise transaction costs. A world without firms in which production was
organised entirely through markets would require full information and no transaction costs; but in the real world it
takes time and money to find out about the product being bought or sold. A firm resolves these problems. Mr Coase
argued that the size of firms is determined by the relative costs of buying in services from outside and the overhead
cost of providing them in-house.
For instance a car firm can either make tyres itself or buy them from a supplier. The tyres will probably cost less if
bought in the competitive marketplace but some of this saving may be offset by higher transaction and co-ordination
costs. The higher these costs are the greater the likelihood that firms will find it more profitable to provide services
internally which will increase their size. However IT in the form of e-mail the Internet the fax machine and
computerised billing reduces these costs and so increases the attraction of buying goods and services from outside.
As these costs fall argues Peter Huber an American telecoms specialist so the traditional logic of the firm becomes
less persuasive. All kinds of goods and services can be outsourced and many employees replaced by outsiders linked
by electronic networks. In this way IT encourages vertically integrated corporate giants (such as AT&T) to break
themselves up into smaller more efficient firms loosely connected by networks.
Two opposing forces are therefore at work. In industries such as software and entertainment where 'network
externalities' are powerful IT will favour a greater concentration of business to exploit larger economies of scale so
firms will tend to increase in size. Elsewhere falling communications costs will favour decentralisation.
The end of science friction?
A related prediction about new technology is that it will make 'frictionless capitalism' possible: middlemen the firms
or individuals who intermediate between producers and consumers will become redundant. This view is strongly
argued by Microsoft's Bill Gates in his book 'The Road Ahead' (published by Viking). Through the Internet
consumers will have direct access to information about goods and services without having to pay middlemen. Travel
agents estate agents insurance brokers retailers and even dating agencies the argument goes are doomed. The
Internet will make it easier for people to make their own travel bookings and find their own properties. Insurance
companies will be able to sell direct rather than through an agent. Shops will be eliminated because producers will
be able to sell through electronic stores on the Net with customers making their choice on-line from home.
Certainly many middlemen will be squeezed out by the Internet but the species is unlikely to disappear altogether.
Lower communication and transaction costs on the Internet will also make it easier for new middlemen to set up
business providing information services or customised newspapers for example. To succeed in this more competitive
world however middlemen will have to become more efficient and find new ways to provide extra value. It will no
longer be enough for a travel agent say just to sell tickets and package holidays. As services improve and transaction
costs plunge consumers can only gain.
* 'Increasing Returns and the Two Worlds of Business'. By Brian Arthur. Harvard Business Review July 1996
*** 'Does Information Technology Lead to Smaller Firms?'. By Erik Brynjolfsson Thomas Malone Vijay Gurbaxani
and Ajit Kambil. Management Science Vol. 40 No 12 December 1994
© 1996 The Economist Newspaper Limited. All rights reserved