A true and fair view of productivity

Politicians, business people and financial commentators talk about economic growth as if they were discussing objective data, like population growth, or temperature. But national income accounting is every bit as much a subjective enterprise as the private sector accounting on which it ultimately depends.

The collapse of Enron has shaken markets because it has reminded everyone that corporate accounts are interpretations, not facts. even the most conservative of accountants has recently been under pressure to confirm stories of heroic leadership, organisation transformation and technological revolution. Everyone knew these things were true, even if figures were sometimes slow to reveal it.

Now reality is breaking through. The hubris of the last decade did not only distort perceptions of the performance of American companies. It distorted perceptions of the performance of the American economy itself. Politicians, business people and financial commentators talk about economic growth as if they were discussing objective data, like population growth, or temperature. But national income accounting is every bit as much a subjective enterprise as the private sector accounting on which it ultimately depends.

Until 1999, the American economy obstinately confirmed the observation of Bob Solow, Nobel Laureate for his work on economic growth. “You can see computers everywhere except in the productivity statistics”. In that year, the Bureau of Economic Analysis discovered that American growth had been understated. The understatement was mainly the result, not of the discovery of new facts about the economy, but of reinterpretation of existing data.

When politicians and pundits talk about economic growth they talk about movements in the level of gross domestic product (GDP) at constant prices. Only when things get rough are people interested in how accounts are compiled, which is why, when things go well, it is possible to get away with so much. There are books written on how these concepts are defined and measured, and they are not much fun to read. Those who can’t be bothered should notice two features. GDP is a gross measure – no allowance is made for depreciation of assets. And it is measured at constant prices. National income statisticians measure the value of output but growth requires an attempt to measure the change in the volume of output. This was easy when output was mostly steel but is much harder in the knowledge economy.

A bar of steel is – more or less – a bar of steel but the volume of computers is an elusive concept. Computer prices have been falling rapidly. Still an entry level PC has cost £700 or so for several years. You just get more computer for your money. But how much more? Economic statisticians use two main estimation techniques. One is to try to track the falling price of the same computer. Another is a statistical process known as hedonic price measurement. Depending on the details of the method, the answers obtained by different national statistics offices for the fall in computer prices over the last five years range from 30% to 75%. Britain is relatively conservative and the United States very aggressive.

This makes a big difference. Actual expenditure on computers in 2000 in Britain was about £10 bn. The Office for National Statistics estimates that the computers which cost £10 bn in 2000 would have cost £18 bn in 1995. If American price indices were used, the figure would be £37 bn. The difference amounts to 2% of British GDP. Over the five years, Britain’s reported growth rate would have been almost ½ % per year higher if our statisticians had used American price indices.

So what is the right answer? We are not trying to measure the benefits of computers. That may well be much larger than £18 bn, or even £37 bn. You could only calculate it by doing extensive surveys of the experience of individual firms, which no one in any statistical office anywhere has attempted. And they don’t need to make the attempt. Whatever the effect of computer investment on the productivity of other sectors, it has already been counted in the official figures. Most computers are bought by businesses. The contribution of this investment to their productivity is included in the output of financial services companies, retailers, telecoms firms, and all the other activities that make use of computers.

The £18 bn figure is an attempt to answer a different question. What part of business spending on computers should be capitalised rather than treated as a cost against current output, because it is intended to enhance future rather than current output? General accounting practice does allow you to capitalise such expenditures. But it insists on two qualifications. One is that you capitalise it at its actual cost, not at some hypothetical measure of what it might have cost in the past or be worth in the future. And the other is that you write off the capitalised expenditure over the lifetime of the asset. The rules for measuring GDP do not impose either of these conditions. You can engage in extravagant revaluation. You don’t have to depreciate the capitalised expenditure. And you don’t even have to persuade a compliant auditor to go along with what you’re doing.

Under standard accounting principles, the maximum expenditure which you could capitalise would be the whole of actual spending on computers in 2000 – £10 bn or so. And you could only justify this if you could argue that there was no need for any write down of previous expenditure on computers – as a result of scrapping or technological obsolescence.

My own estimates are that the stock of computers in Britain in 2000 was probably worth about £20 bn. That is what it would have cost to replace. In that year, available computing power probably rose by 20% or so as a result of net new investment, minus depreciation and scrapping. Because of the falling price of computers, this larger stock of computing power was probably not worth any more at the end of the year than the smaller stock was worth at the beginning. A kindly auditor, like the Andersen people down in Houston, might allow you to capitalise £5bn or so of this expenditure. If a commercial company seriously proposed to credit £37 bn to its P & L – on the grounds that this is what it might have had to pay if it hadn’t bought them so cheaply – its directors would probably end up in jail.

The bottom line of all this is that the published data on GDP probably overstates output growth in the UK over the last five years, but by less than 1%. Economists have known for years that constant price GDP is a flawed measure of output. (The National Institute for Economic and Social Research has produced corrected series for some time). But the errors were not large, and offset by the advantages of data series that were comparable over time and between countries.

The ICT revolution has changed all that. US statisticians were quick to realise that the accelerated pace of price decline might create problems, and introduced a new mechanism called chain linking. Britain’s ONS plans to follow, but has not done so yet. As a result, the distortions in US data are not as large as would be introduced in Europe if American assumptions were adopted here.

But the effects are large nevertheless. Over the period 1996 – 2000, ICT investment contributed almost 1% to reported US growth. Simply substituting net investment at cost for gross investment at revalued prices reduces this contribution by about half. Reported US GDP growth overstates the actual growth of American output by around ½ % per annum over the period.

Table 1: US Contribution of ICT to growth, BEA Growth Accounting framework (% of GDP, Total 1996 – 2000)

This difference is equivalent to the major part of the productivity miracle which so enthused believers in the new economy.

It is not only American companies whose figures are now in question. USA plc capitalised much of its software expenditure, revalued it at the highest price it might ever have paid, and calculated its profits without any depreciation of its revalued assets. Who were its officers at the time? The former CEO, Bill Clinton, may be spending more time with his family. But Alan Greenspan, who has repeatedly argued that American economic statistics should be more consistent with the optimistic reports of US business people, is still the company’s Chief Financial Officer.

An extended exposition of the problems, with references to relevant literature.

John’s paper on National Accounting for the New Economy provides a more thorough discussion of the issues raised in this FT article (of 15 February 2002). It also presents further details of his estimates of the capital stock of computers and revised US/UK growth figures.

The following text highlights some of the key points and suggests some good resources.

History of GDP and National Accounting

The practice of national income accounting was developed during the Second World War, in particular by Richard Stone and James Meade under the tutelage of Keynes (Stone, 1986, Weale, 1993). Some aspects of national accounts conventions – particularly the emphasis on gross rather than net measures and the emphasis on the real flows corresponding to the circular flow of income rather than the financial flows important to private sector accounting – seem to be the product of a perspective based on war conditions and Keynesian economics.

Suggested reading:

BEA, 2000, GDP: One of the Great Inventions of the 20th Century’, Survey of Current Business, (January)

Stone, R., 1986, ‘Nobel Memorial Lecture 1984: The Accounts of Society’, Journal of Applied Econometrics, Vol. 1, Issue 1 (January), pp. 5-28

Weale, M., 1993, ‘Fifty Years of National Income Accounting’, Economic Notes, Vol. 22, No. 2, pp. 178-199.

Defining GDP

A basic glossary item usually describes GDP as “the total market value of goods and services produced in a given year within the borders of a country (whether or not they are sold)”.

An economist would not deviate too much for this definition by defining GDP as the sum of firms’ gross value added. (Gross value added is the extra a vale a firm adds to raw materials in order to make output.) And because gross value added is used to pay employees’ wages and shareholders’ dividends national accountants can estimate GDP by looking at peoples’ incomes, as well as what they spend on goods and services.

GDP is strictly defined as the sum of consumption (C), investment (I), government expenditure (G) and net exports (Z). However, defining which fall under the term ‘investment’ can be a subjective matter. In particular, whilst investment goods are included in GDP because they enhance future output, any intermediate goods (or current inputs) are ignored as they form part of the vale of final output. Recently this has caused many to argue that a greater proportion of computer software should be included as investment because it may enhance production beyond the current year. Indeed this caused the US national accounts to be retrospectively altered.

There are two important things to remember about this definition:

(i) GDP is a gross measure. Investment that firms undertake in order to replace worn out equipment is added to GDP. Even if all investment is used to replace obsolete machines GDP still rises.

(ii) GDP is measured in constant prices. The total market value of a particular good or service can increase (a) if the price rises, and/or (b) if more units are produced. To calculate economic growth in terms of the volume of output national accounts try to eliminate (a) by valuing all output in the prices that existed in some arbitrary base year (1995 in the UK at present).

Whilst this system has been fairly robust up until the 1990s, the introduction of new goods and technology has posed a few problems for national accountants (and users of the data they produce).

Suggested reading:

Office for National Statistics website: ‘Gross Domestic Product: A brief guide’ explains the circular flow and the different methods of calculating GDP. Links to more detailed information provided.

THE PROBLEMS:

(i) Estimating price changes: new goods and rapidly changing technology

Measuring the price fall of standardized goods is easy. A barrel of oil is a barrel of oil in whatever two time periods you look at it. Comparing prices is child’s play. But what happens when we compare the price of a computer or a car, whose characteristics and quality may vary dramatically even over short periods of time? A base-level PC costing £700 this year offers far greater power than a machine costing £700 in 1995.

Economic statisticians have two main estimation techniques at their disposal:

(i) Matched-model approach. This simply tracks the falling price of the same computer over time.

(ii) The hedonic pricing approach. This involves carefully estimating the value each characteristics of goods. For computers this would involve estimating the market prices of the amount of memory, the processor speed the hard disk size, and so on. By comparing old and new models a price differential is calculated using regression analysis. This differential can be positive or negative and, for computers, this is normally negative – reflecting some fairly dramatic falls in prices.

Suggested reading:

BEA, 2000, A note on the impact of hedonics and computers on real GDP, Survey of Current Business, (December)

Hausman, J.A, 1994, Valuation of New Goods under Perfect and Imperfect Competition, NBER Working Paper No. w4970. Also published in Bresnahan, T. and Gordon, T. (eds.) The ‘Economics of New Goods’, Studies in Income and Wealth 58: 209-237.

University of Chicago Press for the National Bureau of Economic Research.

Lequiller, F., 2001, The new economy and the measurement of GDP growth, INSEE working paper

Schreyer, P., 2000, “Computer Price Indices and International Growth and Productivity Comparisons” for an international perspective.

(ii) The distortion of base prices.

As mentioned above, real GDP attempts to disentangle the volume growth of output from the increase in prices and, to do this, all output is valued in the prices that exist in some base year. The problem is the results can be very sensitive to the base year we pick. This is because relative prices (and therefore consumer tastes) change over time. A good that was relatively new and expensive in 1995 may become quite cheap and have gained serious sales momentum in 2000. The good’s contribution to real GDP growth will be upwardly biased if it’s increased production is valued at its high introductory price.

This is precisely what has been happening with computers and communications technology because the price of these investment goods has fallen so fast.

As a result of these concerns the US real GDP is now constructed under a system called “chain linking” which uses the prices of adjacent years rather than one fixed base year. [So, for the year 2000, the prices of 1999 and 2000 will be used and for 2001 the price of 2000 and 2001.] See BEA (1997) for more details. The UK’s Office for National Statistics is planning to follow suit.

Suggested reading:

BEA, 1997, BEA’s Chain Indexes, Time Series, and Measures of LongTerm Economic Growth, Survey of Current Business, (May)

See the ONS Chain-linking project web page for more details of the ONS plans to introduce chain linking and the effect on GDP.

(iii) Depreciation

GDP is measured gross and is therefore insensitive to counting investment that simply replaces old and obsolescent equipment and investment that enlarges the capital stock (bringing fresh productive capabilities). National accountants do produce a version of GDP – Net Domestic Product – that does take account of such depreciation. But, up until recently, the growth in GDP and NDP has been sufficiently similar to warrant a focus GDP growth as the main indicator of economic growth.

Computers, software and communications technology have upset the balance somewhat. Whilst land and industrial equipment may depreciate at around 2% and 15% per cent per annum respectively, computers depreciate far faster than rest of capital stock, perhaps at a rate of over 30% (Lequiller, 2000). So, whilst a greater proportion of firms’ investment spending is on computers much of this spending has merely replaced computers of a slightly older vintage. The diverged between NDP and GDP growth has become even more pronounced.

The impact on growth figures

John’s paper looks at the impact of these issues on rates of economic growth.

Suggested reading:

Jorgenson, D.W. and Stiroh, K.J., 2000, ‘Raising the Speed Limit: U.S. Economic Growth in the Information Age’, Brookings Papers on Economic Activity 1, pp. 125-211.

Oliner, S.D., and Sichel, D.E., 2000, ‘The Resurgence of Growth in the1990s: Is Information Technology the Story?’, Journal of Economic Perspectives, Vol.14, pp. 2-22.