Comparative productivity in British and American manufacturing.

Ark, Bart van

1. Introduction

Studies on comparative economic performance can greatly benefit from comparisons of relative productivity levels by industry of origin. In the past decades various comparisons of productivity in the manufacturing sector between the United Kingdom and the United States have been made by individual scholars.(1) Recently the publication of the 1987 US Census of Manufactures (US Department of Commerce) has been completed, which allows a reassessment of the comparative productivity performance between the two countries. The present comparison for 1987 is particularly interesting in the light of the accelerated growth in manufacturing productivity in both countries during the 1980s compared to the previous decade.

This study has been carried out within the framework of the International Comparisons of Output and Productivity (ICOP) project at the University of Groningen (Netherlands). The methodology is also directly related to recent NIESR studies of this type between Britain on the one hand and the Netherlands (van Ark, 1990a), France (van Ark 1990b) and Germany (O'Mahony, 1992) on the other.

In this article 'productivity' is measured according to various concepts, including 'output per employee', 'output per hour worked' and in terms of 'output per joint unit of labour and capital input'. Section 2 reports on the results of the 1987 benchmark comparison. I will briefly summarise the methodology but I refer to other studies for a more extensive discussion.(2) In section 3 the productivity results for 1987 are extrapolated backwards to 1968 and forwards to 1990.

2. Comparative price and productivity levels in 1987

Unit value ratio-method

Cross-country comparisons of real output and productivity are either obtained by comparisons of output in physical quantities weighted by some indicator (employment or net output) or they require nominal output to be expressed in a common currency. In the present study value added in British manufacturing has been converted into US dollars on the basis of unit value ratios (or, 'industry of origin' purchasing power parities). These unit value ratios (UVRs) represent comparisons of prices for similar products in the two countries. The 'prices' are obtained by dividing the sales value (at ex-factory prices i.e. excluding excise taxes and duties) for the items compared by their corresponding quantities from the production censuses, and are therefore better called 'unit values'. To arrive at the average unit value ratios, detailed comparisons for individual product items were made between the countries. In practice it was not possible to 'match' all products. The census information is not always complete with regard to product information, and in other cases quality differences between the products in the two countries made comparisons infeasible.

For the present study a total of 171 UVRs were compiled, which cover approximately 15 per cent of total manufacturing sales, with slightly higher coverage percentages in industries with relatively many products of a homogeneous nature, such as food products, textiles and wearing apparel. The unit value ratios for products which were not covered were assumed to be the same as those for the products included in the same manufacturing industry or branch.

The first column of Table 1 summarises the UVRs for six major groups of manufacturing branches. This table only shows the geometric average of the UVRs weighted British and American weights, but in appendix Table 1 the results are shown separately for by each of the weighting systems and for all sixteen manufacturing branches. In all cases the UVRs show the dollar value of the pound sterling at or below the exchange rate. For total manufacturing the UVR is 1.41 US dollars to the pound compared to an exchange rate of 1.63 US dollars to the pound. This implies that in 1987 the price level of TABULAR DATA OMITTED manufacturing output in Britain (the unit value ratio compared to the exchange rate) was some 16 per cent above that of the United States in 1987.(3)

The unit value ratio-method used here is preferable to the use of expenditure-based purchasing power parities (PPPs) such as those provided on a regular basis by Eurostat and the OECD. These PPPs refer to price measures which take account of imports and of trade and distribution margins, which should not be included for domestic output in manufacturing. Moreover, expenditure PPPs leave out a good deal of price information on intermediate goods, which make up a significant share of manufacturing output. In fact, the purchasing power parity between the UK and the USA for the total economy was very close to the official exchange rate in 1987, namely 1.65 US dollars to the pound. After an adjustment of the GDP PPP by excluding PPPs for non-manufactured items, the 'proxy' PPP for manufacturing came at 1.51 US dollars to the pound, which is in between the exchange rate and the unit value ratio used for this study.(4)

Labour and joint factor productivity

The relative level of manufacturing productivity between Britain and the United States is estimated on the basis of a comparison of output and factor inputs for 16 branches which together constitute the manufacturing sector as a whole. Value added and employment is obtained from each country's census of production, i.e. the 1987 US Census of Manufactures (US Department of Commerce, Washington D.C.) and the Report on the Census of Production 1987 (HMSO, London). Value added (in the British context, 'net output') is defined as gross value of output minus the cost of materials, energy inputs and industrial services. This is a somewhat broader concept than the nowadays more widely accepted 'gross value added'-concept, as is used in for example the national accounts, which also excludes non-industrial services (see also below under 'measurement issues'). Labour input is expressed in terms of the number of employees (excluding working proprietors and self-employed persons), and is also presented as the number of person-hours worked. The latter estimate refers to the number of hours actually worked, i.e. adjusted for the loss of working time due to sickness, holidays, industrial disputes, etc..

Column (2) of Table 1 shows the ratio of value added by group in the USA as a percentage of the corresponding level in Britain, after it was converted to a common currency. On average value added in British manufacturing is 13.6 per cent of the American level. This is a relatively low ratio compared to the ratio of gross domestic product between the UK and the United States, which was 19.0 per cent of the United States in 1987.(5) On the other hand, by 1987 manufacturing employment in the UK was still at 24 per cent of total employment (based on the UK Census of Employment), whereas the US Bureau of Labor Statistics (BLS) estimated the manufacturing employment share in the United States at only 16 per cent in 1987.

An important measure of the relative strength of the manufacturing sector in an economy is the comparative productivity performance of the manufacturing labour force. Column (3) in Table 1 shows that UK value added per employee in manufacturing was 53.6 per cent of the American level in 1987. However, employees in US manufacturing work more hours than their British colleagues, so that the labour productivity gap is somewhat smaller on the basis of a comparison of output per employee-hour, as is shown in column (4) of Table 1. The average annual working time in US manufacturing was estimated at 1,909 hours per year compared to only 1,763 hours by British manufacturing employees in 1987. Although the number of paid hours in manufacturing is similar for the two countries at approximately 2,100 hours per year, British workers take more holidays and show slightly higher sickness rates.(6)

In Table 2 the information on unit value ratios, real output and labour productivity is rearranged into three main types of manufacturing activity, i.e. (1) the production of durable and non-durable consumer goods; (2) the manufacturing of basic goods, including intermediate products; and (3) the making of investment goods largely by engineering industries. The productivity gap is largest in the consumer goods industries, about average in the investment goods industries and relatively narrow in the basic goods industries. The relative position of the USA which emerges from this comparison is confirmed by Pilat and van Ark (1991) in a study of productivity levels in German, Japanese and US manufacturing. In these comparisons the US lead is also shown to be relatively strong in most light industries, whereas its productivity level was below that of, in particular, Japan in many investment industries. However, the relatively good productivity performance in basic goods industries including textiles, chemicals and basic metals, in Britain can be seen as a unique feature among these four countries.

In order to take account of differences in the capital intensiveness of the manufacturing production process in the two countries, productivity was also calculated as value added per joint unit of labour input (as defined above) and capital input.

Capital input is defined as the gross stock of machinery, vehicles and structures, which was obtained for both countries on the basis of the perpetual inventory method.(7) With this method investments in manufacturing are cumulated on the basis of assumptions concerning the service lives and the scrapping pattern of the assets. According to the official capital stock estimates, the average service lives in Britain are considerably higher than the assumptions for American asset lives. For non-residential structures, service lives in Britain are 60 years compared to 32 years in the United States. The difference for machinery is less but still substantial at 26 years for British machinery and 17 years for US machinery. To my knowledge there is no work of a truly comparative nature which confirms these very large differences in service lives, which incidentally are also much higher for Britain in comparison to other European countries. Furthermore, the evidence on declining asset lives over time (as is assumed in the official UK estimates, but not for those of the USA) is also slim.(8) I provisionally standardised the service lives for both countries on the basis of an average for 14 OECD countries, which is 45 years for structures and 17 years for equipment and vehicles (Blades 1989, 1991). Instead of assuming that all assets are retired at once at the end of the average life time, I used a 'delayed linear' retirement pattern, which assumes that structures are scrapped proportionally after between 36 and 54 years, whereas machinery and vehicles (taken together) are scrapped after between 14 and 20 years of service (see also O'Mahony 1992). The value of the capital stock was converted to a common currency on the basis of the purchasing power parity for investment.

TABULAR DATA OMITTED

The joint factor productivity ratio between the two countries is estimated on the basis of Solow's traditional specification of the Cobb-Douglas production function, with weights for labour and capital which represent the factor shares in total GDP. For the sake of simplicity I applied fixed weights which are based on an average of the labour share in the UK and the USA derived from each country's national accounts.(9) The joint factor productivity ratio is obtained by deducting the logarithmic index of the relative capital-labour ratio of both countries (|K.sup.UK~/|L.sup.UK~ over |K.sup.US~/|L.sup.US~) from that of the corresponding ratio of labour productivity (|Y.sup.UK~/|L.sup.UK~ over |Y.sup.US~/|L.sup.US~), with |Alpha~ representing the share of labour input in gross value added:

in |A.sup.UK~/|A.sup.US~ = in (|Y.sup.UK~/|L.sup.UK~/|Y.sup.US~/|L.sup.US~) - (1 - |Alpha~) in (|K.sup.UK~/|L.sup.UK~/|K.sup.US~/|L.sup.US~) (1)

The last column of Table 1 shows that the joint factor productivity gap is somewhat smaller than the labour productivity gap between the two countries, i.e. UK joint factor productivity is 62.9 per cent of the US level. Compared to the labour productivity gap this implies that capital intensity accounts for less than one eighth of the productivity gap for total manufacturing.

It goes beyond the scope of this article to go in more detail into the reasons for the productivity gap observed here. Differences in the employment mix among the 16 manufacturing branches did not account for any part of the productivity gap. Neither did differences in firm size play a substantial role. The median size (i.e. the average size of a local unit with half of the employees working in smaller units) is only slightly smaller in UK manufacturing compared to the US, i.e. 240 employees in Britain compared to 263 in the United States. The overall effect of differences in firm size distribution is that the manufacturing productivity gap between the two countries in fact marginally increases.(10)

Some studies have pointed at other factors of importance in explaining the relatively large productivity gap between British and American manufacturing, but some of these may be of less importance now compared to earlier decades. On the basis of a cross section analysis by industry, Davies and Caves (1987) have argued that low management skills and poor labour relations, particularly, in large firms largely accounted for Britain's relatively low productivity level during the 1970s. Similar results emerged from a study by Prais (1981) comparing American, British and German industry.

Measurement issues

In the final part of this section the results of the comparisons presented above are compared with 'unadjusted' estimates based on information which is more readily available from official statistical sources. The unadjusted estimates are based on manufacturing GDP at factor cost, employment and capital stock which are directly derived from each country's national accounts, with the value estimates being converted at the currency exchange rate.(11) Table 3 shows that the unadjusted estimates show a somewhat smaller labour productivity gap between the two countries, i.e. about three-quarters of the gap which appears after the adjustments made in this article.

This smaller productivity gap to some extent is explained by the relatively important role of non-industrial service inputs in US manufacturing. These inputs, which include outsourcing of cleaning services, accounting, cafeteria services etc., should be excluded from value added, but the production censuses are weak on the reporting of these services inputs. Although the UK Report on the Census of Production shows value added inclusive ('net output') and exclusive ('gross value added') of non-industrial services inputs, the US Census of Manufactures in fact does not even provide an estimate for it. Estimates on the basis of each country's input-output tables suggest that the share of non-industrial services in total intermediate inputs in manufacturing is indeed somewhat larger in the United States, namely 23.6 per cent in 1986, than in the United Kingdom where it was 19.2 per cent in 1984.

Despite the slight overstatement of value added derived from the production census, this source is to be preferred over national accounts for comparisons of productivity levels. Firstly, the level of manufacturing GDP in the national accounts of both the United States and the United Kingdom is not directly derived from output sources (such as the production census) but rather from income estimates with income allocated to industries on the basis of output shares. Secondly, the output and employment estimates in the national accounts are based on different primary statistical sources. This may not be essential for trend comparisons of productivity, but it can seriously affect level estimates. The census information used here is based on output and employment information based on returns from the same establishments.

As mentioned above the estimates of the capital stock have been adjusted to standardised asset lives and scrapping patterns, and were converted to a common currency on the basis of a PPP for investment. Because of the much higher asset lives according to the official British estimates of the capital stock, the 'adjusted' productivity gap between the two countries is reduced by 7.5 percentage points, but this effect is more than offset by the other effects in Table 3.

The most important reason for the deviation between the unadjusted and the adjusted productivity level TABULAR DATA OMITTED estimates in this article is the use of unit value ratios instead of the exchange rate for converting manufacturing output to a common currency. Exchange rates are increasingly affected by capital movements and speculation and therefore do not reflect the actual price relationships of manufacturing products.

3. Relative UK/US productivity levels in manufacturing since 1968

Comparative productivity trends, 1968-90

It is now half a century since Rostas provided the first meticulous analysis of comparative productivity levels in Britain and the United States (Rostas, 1943 and 1948). The Rostas study, which can still be regarded as the pioneering work for cross-country productivity comparisons, set the stage for a range of studies confirming Britain's relatively low productivity level in manufacturing compared to the United States.(12)

Instead of comparing individual benchmark studies of cross country productivity levels, which are not available on an annual basis and often based on slightly different samples and methods, the benchmark result for 1987 can be extrapolated backwards to 1968 on the basis of time series on output and labour input from the national accounts and employment censuses. Between 1968 and 1990 manufacturing GDP per hour increased at a rate of 3.5 per cent per year in the United Kingdom compared to 2.7 per cent in the USA. However, in both countries the growth rates for the 1970s were significantly lower than for the 1980s. From 1980 up to 1989 UK productivity growth in manufacturing accelerated to an average growth rate of 4.7 per cent per year. In the USA the major break in the trend occurred in 1982, showing an average rate of 4.4 per cent over the period 1982 to 1989.

Table 4 shows the comparative labour productivity ratios for the six major groups of manufacturing branches for selected years of the period under consideration. It appears that throughout the period the manufacturing productivity gap between Britain and the United States has continued to narrow. However, the trends are different the major branches. The improvement in comparative productivity in the UK was fastest in basic metals and metal products, for which the productivity level was the lowest in 1968. The productivity gap in chemicals, which was already relatively narrow in 1973, further reduced during the 1970s and 1980s and made chemicals the most productive branch in relative terms to the USA in 1989. Changes were more modest in other branches, such as textiles and wearing apparel. Particularly striking is the movement of the UK/US productivity ratio in the engineering sector. Up to 1982 the productivity gap in engineering between the two countries narrowed significantly but it widened again during the 1980s.

Chart 1 compares the labour productivity series with the series showing joint factor productivity from 1968 to 1990. The variables are defined and calculated as explained in the previous section. The chart clearly shows TABULAR DATA OMITTED the narrowing of the UK/US labour productivity gap over the past two decades. As this increase in relative UK productivity went in conjunction with a rise in relative capital intensity, the joint factor productivity gap narrowed at a slightly slower pace.

Measurement issues

One major problem in linking national time series to cross country comparisons, is that this extrapolation method will not necessarily provide exactly the same result compared to individual benchmark comparisons. For example, a comparison of the extrapolated results with benchmark estimates for 1968, 1975 and 1977 (Smith, Hitchens and Davies, 1982; van Ark 1990c; Smith 1985) reveals that the latter show somewhat larger productivity gaps between the UK and the USA than the extrapolated results.(13) One reason for such differences is that the benchmark comparisons are based on country weights for a particular year, whereas the time series are based on base year weights which either are fixed throughout the whole period (as in the USA) or shift every five years or so (as in the UK).(14)

In addition to the inconsistency of index numbers over time and across countries, there have also been critics arguing that the national accounts series themselves are inadequate indicators of the trends in real output by industry. In recent years the US national accounts series on GPO (Gross Product Originating) have amongst other things been criticised for their use of a fixed base year for long periods, and for an inadequate double deflation procedure. Furthermore, the US national accounts make for the use of an hedonic price index for computers which allows for a substantially larger quality element in computer prices than the conventional approach of comparing the price change in computers on the basis of matched models, comparing new models directly to older ones.(15)

However, the evidence of biases in the growth rates of US GPO in manufacturing is by no means conclusive. Recent revisions by the US Department of Commerce include an estimate of real output growth based on shifting base years (i.e. 1977 for the period 1977-82, 1982 for the period 1982-87 and 1987 for the period from 1987 onwards). This index shows no change compared to the original series for manufacturing GPO using 1982 weights for the period 1977 to 1987 as a whole. However, the revisions suggest that the slowdown in manufacturing output growth from 1977 to 1982 was less substantial than the earlier estimates suggested, whereas the rise since 1982 was more moderate. As can be seen from the dotted line in Chart 1, these revisions suggest that the narrowing of the productivity gap between the UK and the USA between 1977 and 1982 has been more moderate, whereas there has been no significant widening of the gap during the early 1980s.

There can be little doubt that the hedonic price deflator used for data processing equipment in the United States has some effect on the relatively high growth rates of US real output in manufacturing. By way of sensitivity testing I therefore dropped the hedonic price index from the US series to make them more comparable to the UK series. I replaced the price deflator used in the US machinery-branch (which includes computers) by the price deflator for electrical engineering. The dashed line in Chart 1 shows that, adopting the 1987 benchmark results, the trend in comparative productivity is particularly affected for the early 1980s. However, the difference with the original comparative productivity index is in fact not more than between 2 and 4 percentage points.(16)

It could be argued that the best approach to provide a dynamic perspective of cross-country comparisons is to make benchmark comparisons for each individual year. Apart from the practical problem that complete production censuses are not available for the USA annually, there is also a methodological objection to such an approach. Annual benchmark comparisons imply that the weighting system changes each year, which gives it the nature of a kind of 'chain index'. Chain indices affect the comparibility of the figures over longer periods, as they may make the series 'path dependent'.

A compromise needs to be sought between a regular updating of benchmark comparisons and the use of time series for extrapolation, but one cannot determine an unambiguous time span for updating benchmarks. It partly depends on the different speeds at which structural changes in the two countries occur. If output growth by industry is growing or declining much faster in one country than in another there is more need to reconcile extrapolated series with benchmarks than when growth rates have been fairly similar. The major breaks in the manufacturing productivity growth rates in the UK and the USA in 1980 and 1982 respectively justify the 'rebasing' of benchmark comparisons between these two countries from the 1970s to 1987.

4. Summary and conclusions

In this article the results from the most recent Anglo-American productivity comparison for manufacturing show that census value added per hour worked in the UK was at 58 per cent of the US level in 1987. Between 1968 and 1990 the labour productivity gap in manufacturing narrowed by about ten percentage points. The lower capital intensity in British manufacturing accounts for only a few percentage points of the UK/US productivity gap.

In absolute terms the overall productivity leadership of the United States is apparent, particularly in light industries (e.g. food products, wood and paper products, etc.). In these industries, which still account for 45 to 50 per cent of manufacturing employment, the United States achieved productivity leadership at an early stage of its industrialisation process, supported by the mass production-nature of the products. However, in general consumer goods industries are less open to international trade and show less potential for raising productivity quickly by introducing innovations in the production process. This raises concern about the long term vitality of US manufacturing. On the other hand, in recent years American manufacturing has improved its relatively productivity performance in engineering industries compared to the UK, although it is strongly challenged in these branches by Japanese manufacturing (see Pilat and van Ark, 1991).

On the British side, during the early 1980s a considerable shake-out of inefficient activities in manufacturing took place, which had been built up during the period of Britain's slowdown in earlier decades. The comparative productivity level in British manufacturing is now at its highest in some of the industries where it was at its worst in the early 1970s, in particular in basic goods industries such as basic metals. This once-for-all catch up effect has by now lost most of its force. Recently productivity growth rates in manufacturing have significantly slowed down from 5.5 per cent in 1988 to 3.9 per cent in 1989 to almost zero in 1990 and 1991. As the estimates for 1990 in this article show, the recent recession in Britain has deferred any further catching-up on the US manufacturing productivity level. Furthermore, although the productivity gap is significantly smaller than a decade ago, British manufacturing productivity levels are still some 20 to 30 per cent behind those of its main European counterparts (see van Ark, 1990c; O'Mahony, 1992).

This article also dealt with some of the methodological problems of estimating trends in comparative productivity with the use of national time series linked to a benchmark. There is a clear case for frequent updates of benchmark comparisons, in particular when productivity growth in one or both countries is exceptionally fast or slow.

TABULAR DATA OMITTED

TABULAR DATA OMITTED

TABULAR DATA OMITTED

TABULAR DATA OMITTED

Appendix Table 5. Census value added per hour worked, capital
intensity and joint factor productivity, United Kingdom/United
States, 1987

 UK value UK capital UK joint
 added stock factor
 (net output) per hour productivity
 per hour worked
 worked

1968 49.2 58.2 57.1
1969 50.2 58.9 58.0
1970 51.2 56.7 59.9
1971 51.1 58.3 59.3
1972 52.0 63.2 58.9
1973 52.4 64.8 59.0
1974 54.4 64.5 61.4
1975 53.6 61.8 61.2
1976 53.2 65.2 59.8
1977 51.6 65.7 57.9
1978 52.6 68.0 58.5
1979 53.7 69.6 59.3
1980 55.2 72.3 60.3
1981 57.7 78.3 61.7
1982 60.0 73.5 65.3
1983 60.0 74.1 65.1
1984 59.1 76.7 63.5
1985 57.7 74.0 62.7
1986 57.8 74.3 62.8
1987 58.0 74.4 62.9
1988 58.6 74.7 63.2
1989 61.0 74.8 65.5
1990 58.8

Sources:

1987 benchmarks appendix Tables 3 and 4. Extrapolation with
time series as follows: US manufacturing GPO (up to 1977) and
employment series from US Dept. of Commerce (1986), National
Income and Product Accounts, 1929-82; for recent years from
Survey of Current Business, various issues (including January
and April 1991); 1989-1990 from production index. US hours per
person from Pilat and van Ark (1991). UK manufacturing GDP at
constant prices from CSO, United Kingdom National Accounts,
various issues; em-ployees kindly provided by UK Dept. of
Employment, also published in Dept. of Employment Gazette,
various issues; hours from O'Mahony (1992). For capital stock
estimates see sources appendix table 4.

REFERENCES

Ark, B. van (1990a), 'Comparative levels of labour productivity in Dutch and British manufacturing', National Institute Economic Review, No. 131, February.

Ark, B. van (1990b), 'Manufacturing productivity levels in France and the United Kingdom', National Institute Economic Review, No. 133, August.

Ark, B. van (1990c), 'Comparative levels of manufacturing productivity in postwar Europe: measurement and comparisons', Oxford Bulletin of Economics and Statistics, November.

Ark, B. van (1993), 'The ICOP approach: its implications and applicability', in: A. Szirmai, B. van Ark and D. Pilat (eds.), Explaining Economic Growth, North Holland, forthcoming.

Bacon, R.W., and W.A. Eltis, (1974), The Age of US and UK Machinery, NEDO.

Blades, D.W., (1989, 1991), 'Capital measurement in OECD countries: an overview', paper presented at International Seminar on Science, Technology and Economic Growth, June 1989. In revised form published in: Technology and Productivity: The Challenge for Economic Policy, OECD, Paris.

Davies, S. and R.E. Caves (1987), Britain's Productivity Gap, NIESR, Cambridge University Press.

Denison, E.F. (1989), Estimates of Productivity Change by Industry--An Evaluation and an Alternative, Brookings Washington D.C..

Frankel, M. (1957), British and American Manufacturing Productivity, Urbana, University of Illinois.

Gordon, R.J. and M.N. Baily, 'Measurement issues and the productivity slowdown in five major industrial countries', in: OECD, Technology and Productivity: the Challenge for Economic Policy, Paris, pp. 187-206.

Krijnse Locker, H. and H.D. Faerber (1984), 'On the estimation of Purchasing Power Parities on the basic heading level', The Review of Income and Wealth, Series 30, No. 2, June.

Maddison, A. (1991), Dynamic Forces in Capitalist Development, Oxford University Press.

Maddison, A. and B. van Ark (1988), 'Comparisons of real output in manufacturing', Policy, Planning and Research Working Papers, WPS5, World Bank, Washington D.C..

Mishel, L. (1988), Manufacturing Numbers: How Inaccurate Statistics Conceal US Industrial Decline, Economic Policy Institute, Washington D.C..

O'Mahony, M. (1992), 'Productivity levels in British and German manufacturing industry', National Institute Economic Review, February.

Paige, D. and G. Bombach (1959), A Comparison of National Output and Productivity, OECD, Paris.

Pilat, D. and B. van Ark (1991), 'Productivity Leadership in Manufacturing, Germany, Japan and the United States, 1973-89', Research Memorandum, no. 456, Institute of Economic Research, Groningen.

Prais, S.J. (1981), Productivity and Industrial Structure, Cambridge University Press.

Rostas, V. (1943), 'Industrial production, productivity and distribution in Britain, Germany and the United States', Economic Journal, April.

Rostas, V. (1948), Comparative Productivity in British and American Industry, National Institute of Economic and Social Research, Cambridge University Press.

Sinclair, J. and B. Catron (1990), 'An experimental price index for the computer industry', Monthly Labor Review, October, pp. 16-24.

Smith, A.D., D.W.W.N. Hitchens and S. Davies (1982), International Industrial Productivity, National Institute of Economic and Social Research, Cambridge University Press.

Smith, A.D. (1985), 'Changes in comparative Anglo-American productivity in manufacturing industries', National Institute Discussion Paper no. 101, National Institute of Economic and Social Research, London.

Summers, R. and A. Heston (1988), 'A new set of international comparisons of real product and price levels, estimates for 130 countries, 1950-85', The Review of Income and Wealth, Series 34, No. 1, March.

Summers, R. and A. Heston (1991), 'The Penn World Table (Mark 5): An expanded set of international comparisons, 1950-88', Quarterly Journal of Economics, Vol. CVI, No. 2, May.

Szilagyi, G. (1984), 'Updating procedures of international comparison results', The Review of Income and Wealth, Series 30, No. 2, June.

NOTES

(1) See Rostas (1948), Frankel (1957), Paige and Bombach (1959), Smith, Hitchens and Davies (1982), van Ark (1990c) and Smith (1985).

(2) See for example, Maddison and van Ark (1988), van Ark (1990c, 1993).

(3) Relative price levels between the two countries have been very different over the 1980s due to the volatile movements of the US$/|pounds~ exchange rate. During the years 1983 and 1985, when the dollar was relatively expensive, relative price levels in British manufacturing were significantly below that of the United States.

(4) These PPPs are based on Fisher-indexes (kindly provided by Eurostat) and are therefore more comparable to our UVRs than the multilaterally weighted PPPs.

(5) GDP for 1987 derived from Maddison (1991, table A.2) with an adjustment by substituting Fisher purchasing power parities (PPPs) for Paasche PPPs (kindly provided by Eurostat) to improve the comparability with the estimates for manufacturing in this study.

(6) The ratio of hours worked to hours paid in manufacturing is estimated at 0.909 for the United States (US Bureau of Labor Statistics, Monthly Labor Review, February 1990) compared to 0.838 in the United Kingdom. The latter estimate is based on a detailed accounting of weekly hours paid (including overtime hours) and the number of weeks actually worked for 1984 by van Ark (1990a). The 1984 hour-estimates are updated to 1987 on the basis of an index of total weekly hours paid divided by the trend in employment.

(7) Depreciation has not been taken out of the capital stock estimates, because it usually shows little if no relation to the actual decline in the 'productivity capacity' of items. As productivity is calculated in terms of value added I only took into account fixed assets and excluded inventories.

(8) See for example Blades (1989, 1991) for a comparison of service life assumptions for assets in different OECD countries. Bacon and Eltis (1974), casts doubt on whether large service life differences do exist for machine tools. Clearly, changes in the composition of the capital stock may cause a change in asset lives, which has been taken into account for the standardised estimate in this article as far as the increase in the share of equipment compared to structures is concerned. Maddison (1991, appendix D) has also argued that studies of second-hand markets do not suggest declining asset lives, as assumed for example in Britain. New work on international comparisons of service lives of assets is necessary to settle this issue in a more satisfactory way.

(9) The share of labour income for manufacturing was estimated at 0.729 for the United States and at 0.722 for the United Kingdom in 1987. These labour share excludes the income of self-employed persons.

(10) See van Ark (1993) for more details. It should be emphasised that at this stage I made adjustments only at the level of sixteen manufacturing branches. At a more disaggregated level factors such as industry composition and firm size distribution may well account for a greater part of the productivity gap.

(11) The number of hours per person employed is assumed to be the same as for the comparison based on production census information.

(12) Some of these studies are cited in note 1. For a more extensive overview of these studies and their methods see van Ark (1990c, 1993).

(13) For 1968 the UK/US ratio of census value added per person employed is 0.370 on the basis of the benchmark estimate (Smith, Hitchens and Davies, 1982, adjusted to exclude working proprietors in the UK) compared to 0.493 as extrapolated backwards from 1987. The differences are less for the benchmark estimates of the 1970s, which show 1975 at 0.443 (van Ark 1990c, adjusted from national accounts to census value added) compared to 0.508 on the basis of extrapolation. For 1977 Smith (1985) estimated UK census value added per person employed at 40.2 per cent of the USA (adjusted to exclude working proprietors in the UK), compared to 49.2 per cent based on the present study.

(14) See for example Krijnse Locker and Faerber (1984) and Szilagyi (1984) for a discussion of index number problems when linking time series and cross-country comparisons. Similar problems of weights inconsistency were also faced in the construction of the Penn World Tables of a System of Real National Accounts (Summers and Heston, 1988; 1991).

(15) For the most explicit criticism of the US national accounts output series in recent times, see Mishel (1988), Denison (1989) and Gordon and Baily (1991). The hedonic price index for computers in the US shows a very substantial price fall of as much as 40 per cent between October 1988 and January 1992 (see Sinclair and Catron, 1990, and subsequent issues of BLS, Producer Price Indexes). There is no separate computer price index available for the United Kingdom. For comparison, in Germany the price decline between 1987 and May 1992 was estimated at about 10 per cent (Statistisches Bundesamt, Preise und Preisindizes fur gewerbliches Produkte, various issues).

(16) It should be emphasised that this is only a hypothetical experiment making the series more comparable, on the basis of the unrealistic assumption that there was almost no price fall for computers in the two countries. I would not want to suggest that one should go back to the 'matched model' approach in devising price deflators, which is still common practice in most European countries.