Comparative levels of labour productivity in Dutch and British manufacturing.

Van Ark, Bart

COMPARATIVE LEVELS OF LABOUR PRODUCTIVITY IN DUTCH AND BRITISH MANUFACTURING

1. Introduction

In recent years there has been a renewed interest in comparisons of labour productivity levels between industrialised economies. Figures on comparative growth performance, as well as on comparative levels of productivity, are widely quoted in academic works, official reports and in the press. The increased interest in comparative productivity performance can be traced to the re-structuring of industrialised economies during the last two decades. The shift of resources from manufacturing to services was already evident before the oil crises of the 1970s, and the subsequent recession of the early 1980s reinforced the reduction of the manufacturing labour force to around a quarter of total employment in most industrialised nations.

Nevertheless, manufacturing production remains of crucial importance in creating new opportunities for technological change and economic growth. Important structural changes within the sector have taken place as well. Manufacturing branches in which technical innovations have been important - for example chemicals, electrical engineering, office and data processing machinery - increased their relative market share, whereas in traditional industries - for example metals and textiles - growth has slowed down, and in some cases falls in output and employment have occured.

This article is primarily concerned with the measurement of productivity levels. Several studies in recent years have shown that British productivity levels have stayed far behind those of Germany and the United States. The present study was undertaken to compare the productivity performance of British industry with a much smaller, but important, economic counterpart. It presents comparisons of output per person-hour worked in manufacturing branches of the Netherlands and the United Kingdom, which are derived from information collected in each country's census of production. A comparison between the two countries for the late 1950s (Mensink 1966) implied that the productivity gap in manufacturing was negligible at that time. However it will be shown that at present the productivity level in Dutch manufacturing industry is more like that in Germany than that in Britain.

The Netherlands has a population of about a quarter and a geographical area of less than 20 per of the United Kingdom. Its total gross domestic product is slightly over a quarter of British GDP. However, it would be wrong to think of the Netherlands as a small isolated economy, since it recognised an early need to join in a larger economic grouping by helping to form the Benelux. Today this has been overtaken by its involvement in the European Community. The close geographical and economic ties of the Netherlands with the main EC countries give it a greater effective market than its own size might imply.

A comparison between the Netherlands and the United Kingdom is even of specific interest. After 1973 the Netherlands began to suffer from similar problems as Britain in re-structuring its manufacturing sector. The development of a substantial natural gas field in the north of the country contributed to an appreciation of the guilder, which affected the competitive position of export-oriented and import-competing industries. This was by many economists seen as an extreme case of de-industrialisation, which even acquired its own label, the `Dutch disease.'

Both countries applied active policies to restore manufacturing growth after 1980, but the emphasis was rather different. In Britain contractionary fiscal and monetary policies were supplemented by the liberalisation of labour markets and financial markets and the privatisation of nationalised industries. In the Netherlands, there has been a continuous drive to cut government spending. At the same time the government has moved away from a direct involvement in central wage negotiations with employers' for negotiations at the level of industrial branches and firms. Among other things an employment-creating policy of redistribution of labour was implemented by cutting weekly working hours from 40 to 36 hours in most parts of the economy.

Governments in both countries have stressed the need to concentrate on supply-side factors as well, in particular to increase the level of skills of the labour force. The effect of these policies on the actual productivity performance remains a question of concern.

The following section presents details of the method and procedures used in calculating the productivity ratio between the two countries. It would go beyond the scope of this study to provide a full account of the complex reasons why the Netherlands has a relatively high level of productivity in manufacturing compared to Britain. Section 3 attempts to provide some perspective by measuring the effect of the different composition of the Dutch and British manufacturing sectors; it also looks at the variation in relative and absolute levels of productivity among manufacturing branches. Section 4 takes the analysis a step further by examining the relation between the average size and the degree of vertical integration of manufacturing units.

2. Dutch-UK comparison of productivity in

manufacturing

The main results of this study are set out in table 1. It shows the calculations of output per person-hour worked in the Netherlands and the United Kingdom in 1984 for total manufacturing and for 16 constituent groups of industries (`branches'). Output is defined as gross value added, that is, the total sales of goods and services minus the purchases of materials and other inputs. Output per person-hour in national currencies is converted to prices of the other country on the basis of unit value ratios (see below). The comparisons are made both in Dutch and UK prices; these yield slightly different ratios of productivity, because of differences in the mix of products within each industry, which is the familiar `index number problem'. The estimate for total manufacturing, taking a geometric average of the two sets of prices, shows Dutch manufacturing industry as having a 65 per cent greater output per person-hour worked than British manufacturing. The gap would be 52 per cent if calculated on the basis of output per person employed instead of output per person-hour worked, due to the larger number of hours worked per person in Britain compared to the Netherlands (see below). [Tabular Data Omitted]

Before commenting further on the findings, the main sources for the calculations and the most important features of the method of calculation need to be briefly outlined. The censuses of production record the total value of output of all manufacturing establishments (above a minimum size) classified to each industry, together with their main costs of production and the number of persons employed. It provides the most comprehensive and detailed account of the activities of manufacturing industries. The production census is well-suited for productivity calculations because the figures on output and labour input are based on returns from the same establishments. In annex A the output and employment figures used for this study are compared with corresponding information from other sources, that is, the national accounts and employment statistics.

Both countries also publish the quantities (in terms of tonnages or other physical units) and values of the major products sold by manufacturing industries; unit values for these products can be obtained by dividing values by quantities. The ratios of unit values for matched pairs of products in the two countries provided the basis for converting each branch's total gross value added, measured in the currency of its own country, into the currency of the other. The product-detail for both countries did not always match, but it was possible to match 107 major products for 1984. These matched products accounted for 15 per cent of the total sales of manufactured product items in the Netherlands and for 12.5 per cent of total manufacturing sales in the United Kingdom. It should perhaps be emphasised that these coverage percentages concern only the unit value ratios used to convert the value of output to a common currency; the final estimates of relative productivity are based on total value added and employment for all manufacturing activities.

It is nevertheless worth considering whether the unit value ratios (UVRs) for the matched items can be taken as unbiased estimates for the uncovered part of manufacturing. In this study the calculations were staged at different levels of aggregation. First, UVRs were calculated for a group of 20 industries, where at at least 30 per cent - but in most cases over 50 per cent - of the output value could be matched. Although other assumptions are possible, it seems most plausible to assume that the price relationship for non-covered items in each industry is similar to that of covered items.(3) After that UVRs were aggregated to the level of manufacturing branches and finally for the sector as a whole. Annex B describes the method of conversion in more detail for each branch, as the method varied slightly according to the availability and representativity of matched items in each branch.

The method described here is called the industry-of-origin approach, because it concentrates on sectoral output, by using manufacturers' ex-factory prices to `deflate' value added. It has important advantages over possible alternative ways of converting value added to the currency of another country. For example, market exchange rates are subject to fluctuation and are likely to be influenced by capital movements. Alternatively prices to the final purchaser, which are used for the estimation of purchasing power parities (PPPs) by the EC and the OECD, incorporate transport and distributive margins; they also include prices of imported goods and exclude prices of semi-manufactured goods.

One cannot directly obtain from the statistics a PPP based on prices to final purchasers for domestic output in manufacturing, but it is of interest to estimate an average PPP based on retail prices for predominantly manufacturing products. On this basis the `proxy PPP' for manufacturing output in 1984 was 4.28 guilders to the pound sterling, which was almost identical to the official exchange rate.(4) For comparison, the (geometric) average unit value ratio for total manufacturing calculated for this study was 3.76 guilders to the pound sterling in 1984. It seems that the `proxy PPP' overestimates the price relationship for manufacturing output, because it only reflects prices of final goods, and not those of semi-manufactured products. Annex B shows that the average UVR of the latter group, in which the Netherlands is more strongly represented than Britain (see also section 3), was about 15 per cent below the average UVR for final goods. Consequently the manufacturing productivity gap between the Netherlands and the UK is 65 per cent instead of 45 per cent as assumed on the basis of proxy PPPs.

The average price level of manufacturing products, which is derived by dividing the UVR by the market exchange rate, was about 12.5 per cent lower in the Netherlands compared to Britain in 1984.

The estimates of the productivity gap in table 1 are adjusted for differences in the number of hours worked per employee in the Netherlands and the United Kingdom. On average a British employee worked about 8.5 per cent more hours per annum compared to his or her Dutch counterpart, that is, 1,749 hours compared to 1,611 hours. Annex C provides details of the method of calculation of the hours worked per person. The difference in numbers between the two countries is reversed in Maddison's estimates, that is, 1,520 hours per annum in Britain compared to 1,640 hours in the Netherlands in 1984 (Maddison, 1987); but his estimates are for the total economy whereas those in the present study are for manufacturing only. It appears that the policy of redistribution of labour by cutting weekly working hours per person had more impact on Dutch manufacturing than on other parts of the Dutch economy; on the other hand overtime was more substantial in British manufacturing than in the rest of the British economy.

Table 2 shows the results of an extrapolation of the `benchmark year' estimates for 1984 backwards to 1975 and forwards to 1988. For this purpose indexes of production and employment in the Netherlands and the United Kingdom were used for combinations of branches. Britain's relative level of output per person-hour in manufacturing was at its lowest compared to the Netherlands at the beginning of the 1980s. Between 1984 and 1988 the gap narrowed by one-third. This was not so much due to a further acceleration of manufacturing productivity growth in Britain as to a near standstill of productivity growth in the Netherlands between 1984 and 1987. However, by 1988 output per person-hour worked in Britain was still less than 70 per cent of the Dutch level.

Table : Table 2. Output per hour worked in combined

 manufacturing branches: Netherlands compared
 to the United Kingdom, 1975-1988
 Netherlands/United Kingdom (UK=100)

Food, beverages and

tobacco 149.0 161.9 155.9 144.9

Textiles, wearing

apparel, leather and

footwear 134.3 148.4 163.6 146.3

Wood products, stone,

clay and glass

products 134.4 158.2 165.6 157.2

Paper products, printing

and publishing 153.5 164.7 175.3 145.9

Chemicals, petroleum

refining, rubber and

plastic products 145.4 178.0 177.8 154.6

Basic metals, metal

products, engineering

 and other 124.8 149.1 142.3 120.1
Total manufacturing 143.3 169.1 165.5 144.2

This is not the first study to find that Britain lags behind many other industrialised nations in its manufacturing productivity performance. A number of detailed comparisons of productivity levels were made between Britain and the United States(5); a study at the National Institute by Smith, Hitchens and Davies (1982) showed that the German output per person employed in manufacturing was 135 per cent of the UK level in 1968. The authors used indexes of output and employment to update the 1968 German/ British productivity ratios to 1977. If this productivity ratio is extrapolated further, one would find that the output per person-hour worked in German manufacturing was over 150 per cent of the British level in 1984 and close to 145 per cent in 1988(6).

The important conclusions from this section are firstly that the gap between Dutch and British output per person-hour in manufacturing is substantial and roughly of the same magnitude as that between Britain and Germany. Secondly, the productivity gap between Britain and the Netherlands was reduced by one-third between 1984 and 1988. The compound growth rates of manufacturing productivity between 1979 and 1988 were 4.3 per cent in the United Kingdom and 3.0 per cent in the Netherlands. On the assumption that Britain could sustain this faster growth over the coming decades, it would still take up to the year 2010 before output per person-hour in British manufacturing has reached the same level as in the Netherlands.

3. Variations amongst industrial branches

Table 1 in the previous section shows both the comparative and absolute levels of productivity for manufacturing branches. Comparative productivity measures the ratio of Dutch output per person-hour worked to its British equivalent. It appeared that the productivity advantage is small in some branches, for example leather and footwear and transport equipment. On the other hand, output per person-hour in Dutch textiles and basic metals is twice the corresponding level in Britain, and for paper products it is as much as three times as high.

Absolute productivity refers to the nominal value of output per person-hour in each branch. Absolute productivity levels are highest in chemicals and petroleum refining, because these are very capital-intensive industries. It is also high in the food manufacturing branch and in basic metals.

The comparative productivity for total manufacturing is not only determined by the productivity ratios for the individual branches, but also by the absolute levels of productivity and the composition of the manufacturing sector in the two countries(7). Table 3 shows that output and labour input in Dutch manufacturing are more concentrated in relatively few branches than is the case in Britain. This is a characteristic feature of a relatively small open economy. The food manufacturing industry, the chemicals branch and the electric engineering branch each account for approximately 15 per cent of manufacturing employment in the Netherlands. The high comparative productivity in Dutch chemicals has a strong impact on the productivity ratio for overall manufacturing because of its high absolute productivity and the high employment (or output) share in manufacturing. Similarly, the relatively low comparative productivity ratios for wearing apparel, footwear and transport equipment do not do much to reduce the overall productivity ratio because these are branches with low absolute productivity levels and relatively small shares in total employment and output.

Table : Table 3. Share of gross value added and total

 hours worked by branch: Netherlands and the
 United Kingdom, 1984
 (percentage terms)
 Netherlands United Kingdom
 Gross Total Gross Total
 value hours value hours
 added worked added worked

Food products, beverages,

 and tobacco 17.59 16.77 13.43 11.79
Textiles 2.31 2.88 3.33 5.14

Wearing apparel, leather

 and footwear 1.12 2.00 3.15 5.52
Wood products 2.08 3.09 2.81 3.97
Paper products 3.04 2.96 2.89 3.09

Printing and publishing 7.25 7.34 6.80 5.45

Chemicals, rubber, plastic

and petroleum

refining 22.52 15.21 15.90 10.31

Stone, clay and glass

 products 3.40 3.70 4.74 4.35
Basic metals 5.48 3.84 3.60 3.39
Metal products 6.85 8.85 6.46 8.02
Machinery 8.53 9.90 11.24 12.13
Electric engineering 13.40 14.23 12.48 11.59
Transport equipment 5.30 7.90 10.59 12.13

Instruments and other

 manufacturing 1.12 1.46 2.58 3.03
Total manufacturing 100.00 100.00 100.00 100.00

This compositional component can be estimated - at least in part - by weighting the productivity ratios for each branch by constant weights, using the labour input of either the Netherlands or the United Kingdom. The estimated gap in output per person-hour would then be 52-55 per cent.

One should be aware that the estimate above is based solely on differences between branches and does not take account of compositional differences within branches, that is the product mix. At this stage, it is not possible to provide exact figures, but a closer look at the product mix within some branches provides some perspective on the significance of this matter.

In food manufacturing, important differences in the product mix can be observed in the dairy industry 40 per cent of British dairy output consists of liquid milk compared to only about 10 per cent in the Netherlands; butter and cheese alone account for almost 50 per cent of Dutch dairy production. Another example concerns the textile industry. During the 1970s the traditional Dutch textile industry declined as a result of severe competition from the newly industrialising countries(8); it then moved into a higher quality segment of the market, in particular by increasing the share of artificial fibres in output; similar changes took place in Britain, but still about a quarter of value added in textiles represents hosiery and knitwear products, which are characterised by relatively low absolute levels of productivity.

In the chemicals branch one can distinguish between basic chemicals (for example, fertilizers, resins and synthetic fibres) and finished chemical products (for example, paints, pharmaceutical products, soap, rubber and plastic products) The Dutch productivity advantage for basic chemicals was about 60 per cent compared to 40 per cent for finished chemical products. The first category is relatively predominant in the Netherlands - accounting for about 50 per cent of employment in chemicals - whereas the latter category accounted for 80 per cent of employment in British chemicals. In combination with the higher value added per person-hour for basic chemicals in nominal terms compared to finished products, it appears that this difference in product-mix accounts for about one-third of the 70 per cent productivity advantage in the Dutch chemicals branch.

Part of the productivity gap between the Dutch and British basic metals industry is also accounted for by compositional differences. Apart from producing basic hot rolled products and sections, the British steel industry manufacturers a relatively wide range of more labour-intensive end-products, such as tubes, cold rolled sheets, tinplates and galvanised sheets, which are less important in the Dutch steel industry. On the other hand, the Dutch industry is relatively strongly involved in the first phase of non-ferous metal production, for example aluminum, which is a highly capital-intensive production process. However, it appears that in this branch there is also another reason for Britain's relatively low productivity. This concerns the relatively slow introduction of new technology, such as basic oxygen steelmaking and the modernisation of casting and rolling facilities in British steel plants. An intensive round of new investments at the end of the 1970s coincide with a collapse in the steel market.(9) Between 1980 and 1984, employment in the production of basic metals was almost halved, but ouput growth was restored again. The British steel industry has caught up rapidly with its continental competitors, though in 1984 output per person-hour was still only about half of the Dutch level.(10) The productivity gap between the Dutch and British steel industry was reduced to only 20 per cent in 1988.

The effect of the delayed introduction of new technologies is also an important cause of the most extreme result of this study: in paper production, according to table 1 Dutch plants in 1984 produced over three times as much output per person-hour worked as in the UK. This result is consistent with the outcome from a detailed international comparison of paper-producing plants at the beginning of the 1980s, which showed that the plants in West Germany and France were producing 4-5 times as many tons per person as in the average UK mill.(11) The paper industry in Western Europe has undergone a fundamental transformation in the past decades, involving a greater use of recycled paper, the use of pulp delivered to the paper mill by tanker in `semi-processed form, and in particular the introduction of highly mechanised plants. The Dutch paper industry successfully made these changes during the mid-1970s, whereas the British paper industry delayed modernisation until after 1982.(12) In recent years, British paper output has steadily increased, and the productivity gap between the countries has been slightly reduced.

It can be concluded that in general the Dutch manufacturing industry tends to concentrate more on producing `capital-intensive' and semi-manufactured products compared to Britain, Although exact estimates are not available, the differences in the relative shares of branches and in the product-mix within branches appear to account for a quarter to one-third of the observed productivity gap between the two countries. However, other `branch-specific' factors, such as the absorption of new technology and the efficiency of raw materials, also play an important role in explaining the variation in the productivity differential among branches.

It must be noticed that fourteen of the sixteen branches in the Netherlands had a productivity advantage of over 20 per cent compared to Britain. This implies that common factors, which are valid for all manufacturing branches, are at least as important in explaining the productivity differential between the Netherlands and Britain as branch-specific factors.

Matched plant' studies are required to analyse the impact of factors such as the quality of the labour force, labour relations and the utilisation of capital on the observed productivity gap more accurately. A comparison of `matched plants' in the food manufacturing industry of the Netherlands and Britain is underway at the National Institute. Food manufacturing accounts for 15 per cent of Dutch manufacturing employment and for 10 per cent in British manufacturing. Its productivity advantage is relatively close to the average observed for total manufacturing. New developments in food technology have become apparent, which in conjunction with a better skilled labour force have the potential to further improve the productivity performance in this branch and increase the quality of the products it produces.

4. Average size of the manufacturing unit and

vertical integration

There is a long-standing discussion on the possible relationship between the average size of plants and their productive performance. It is known from earlier studies that British plants on average operate at a similar or even slightly larger scale - in terms of numbers of employees - compared to countries such as Germany and the United States, but that there is no productivity advantage in this for Britain (Prais, 1981). On the contrary, in Britain larger plants appeared to be more hampered by unfavourable labour relations than in other countries.(13)

Table 4 shows that the difference in the typical size of manufacturing units between the Netherlands and the United Kingdom is very pronounced. The median size (that is, the number of employees of which half are employed in plants with more than that number) of the manufacturing enterprise in the Netherlands is 254 persons compared to 413 persons for the average establishment in Britain(14). Dutch production units are only significantly bigger than in Britain in branches which are dominated by one or two big companies, such as chemicals and petroleum refining, basic metal, and electric engineering.

It is remarkable that Dutch manufacturing units, which appear on average to be about 40 per cent smaller than in Britain, are able to produce 65 per cent more output per person-hour worked. This effectively implies that one median production unit in the Netherlands has about the same output as its British counterpart; but, of course, produces that output with over 40 per cent less labour input.

The substantial difference in size to some extent reflects a higher degree of vertical integration in British manufacturing units. The defining characteristic of integrated plants is that they include a succession of stages in the production process. On the other side, specialised units tend to concentrate on a limited number of operations, by purchasing more semi-manufactures from other units which are then further processed. A high degree of specialisation (or low degree of vertical integration) should be reflected in the census by a relatively low ratio of value added to gross value of output: each transfer of output from one production unit to another unit appears in the census as output, whereas this transfer does not lead to any increase in value added. Table 4 (right hand side) shows that the ratio of net output(15) to gross value of output is about 25 per cent lower in the Netherlands than in the United Kingdom. [Tabular Data Omitted]

A simple regression of the two variables in table 4 shows a fairly strong positive relationship - significant at the 1 per cent level - between the differential in net output/gross output-ratios and the of the median size of production units in the two countries. Naturally, other factors may contribute to a lower net output content per unit of gross output as well: the higher cost of intermediate inputs per unit of output can be caused by quality differences; or by a bigger share of imported raw materials, which tend to be more expensive because of import duties.

5. Discussion of the results

This article has shown that average output per person-hour in British manufacturing was only some 60 per cent of that achieved in the Netherlands in 1984. Some of the variation in productivity performance among branches is explained by differences in industrial structure and product mix in the two countries. The relatively strong concentration of Dutch manufacturing in capital-intensive industries and in the production of semi-manufactured goods accounts for a quarter to one-third of the observed productivity gap. The delay in the introduction of new technologies seems to account for part of the productivity gap in some branches as well. However, as Dutch output per person-hour was more than 20 per cent above the British level in fourteen of the sixteen manufacturing branches, common factors such as the intensity of skills and the utilisation of capital, play an important role as well.

Between 1979 and 1988, British manufacturing productivity has increased at 4.3 per cent per year on average, compared to 3.0 per cent a year in the Netherlands. A report by the British Treasury(16) suggested that the high growth rates in British manufacturing productivity are not only the result of overcoming the recession of the early 1980s, but also indicate a more fundamental change in performance, that is, a better use of technology, changes in industrial relations and the increase in international capital mobility. The present study suggests that even if the Treasury view is correct and Britain could sustain its relatively high growth rate of manufacturing productivity of the last decade, it would be well into the next century before the Dutch level of output per person-hour is achieved. In 1988, output per person-hour in British manufacturing was still less than three-quarters of that in the Netherlands and this reduction in the gap was primarily due to a near standstill of productivity growth in the Netherlands between 1984 and 1987, which picked up again in 1988.

It is obviously of considerable interest to know whether the performance of British manufacturing relative to the Netherlands is representive of its performance in relation to other European countries. This study provided some indication that, on the basis of earlier comparisons, the present German/British productivity gap seems to be of a similar size as the gap observed between the Netherlands and the United Kingdom. A comparison of the British production census for 1984 with the census in France is now underway, which, in combination with a renewed comparison between Britain and Germany, will provide a more complete picture of Britain's relative productivity performance.

Finally, comparative labour productivity is not quite the same thing as competitiveness. Obviously, a relatively high output per person-hour might improve a country's competitiveness, but the latter also depends, for example, on the cost of a unit of labour input.

Some studies therefore presented `unit labour costs', which is the ratio of hourly compensation to the output per person-hour, as a measure of competitiveness. Compensation per person-hour in British manufacturing is about two-thirds of the corresponding level in Netherlands (17), which seems to suggest that the backlog in Britain's comparative productivity level is balanced by it lower labour costs. However, this cost advantage only improves a country's competitive position if it either lowers the total production costs per unit of product or leads to a more effective mix of costs. As mentioned above, the relative price level of manufactured products in Britain was 12.5 per cent above that of the Netherlands in 1984. This indicates that either non-labour costs in manufacturing production were relatively higher than in the Netherlands, or that the British pound was overvalued against the guilder which would have eroded Britain's cost advantage.(18)

Usually countries with higher labour costs tend to have higher productivity levels as well. This is particularly so if low labour costs are associated with a larger proportion of unskilled or semi-skilled labour in the labour force. Earlier comparisons of British manufacturing with Germany and the United States (Prais, 1981; Daly, Hitchens and Wagner, 1985; Davies and Caves, 1987; Steedman and Wagner, 1989), showed that these factors account for a substantial part of Britain's manufacturing productivity gap. In addition the higher quality of products in Germany was a significant factor in comparisons between Germany and Britain. Comparisons of `matched plants' in the food manufacturing industry of the Netherlands and Britain are underway at the National Institute to analyse the impact of all these factors on the productivity gap observed in this study.

ANNEX A. STATISTICAL SOURCES OF MANUFACTURING OUTPUT, EMPLOYMENT AND PRODUCTIVITY

Table A1 shows the output and labour input figures used for the calculations of output per person-hour worked in this study. The value of output and the number of persons employed are taken from the national production censuses, that is, the Produktiestatistieken and the Report on the Census of Production (estimates of person-hours worked are described in detail in annex C). The Dutch production census excludes establishments with less than 10 persons employed. [Tabular Data Omitted]

In this study, output corresponds to `value added', which is defined as the value of production minus the costs of raw materials and other industrial and services inputs. It is expressed at factor cost, that is, excluding indirect taxes and including subsidies. The number of persons employed includes employees and working proprietors, but excludes self-employed persons.

Production censuses are sometimes criticised for their anachronistic concepts, and their lack of comparability in an international framework. For example, in former days value added in production censuses (often called `net output') was a slightly broader concept compared to the definition given above: it included purchases of services inputs. Other problems in production censuses concern the diffuse treatment of indirect taxes and subsidies; the registration of stocks at book value instead of at constant prices; and the limited coverage of establishments in the lowest (employment) size bands.

Nevertheless, such problems, the production census has some major advantages over national accounts and employment statistics for comparisons of labour productivity:

(1) It is the most comprehensive and detailed account of the

activities of manufacturing industries: one can make adjustments

even for minor differences between the Dutch and

British classification of industrial activities; it makes it possible

to remove net indirect taxes from output; and the purchase

and sales of merchanted goods can be dealt with in

the same way for both countries.

(2) The information on output and labour input is based on

returns from the same establishment in the manufacturing

sector; in Britain, the national accounts statistics do not

provide an estimate of employment which is fully consistent

with the output estimates of GDP; in the Dutch national

accounts, information on employment is provided but it

contains a substantial margin of error.

At present the value added concept in both the Produktiestatistieken and the Report on the Census of Production is the same as that used in the national accounts. Nevertheless, table A2 shows that differences in levels of output, employment and productivity are sometimes large. Value added in manufacturing as a whole is 3 per cent lower according to the Dutch production census (estimates for establishments with less than 10 employees are included here) compared to the national accounts. In Britain, value added from the production census (adjusted for price changes in the book values of stocks) was 3 per cent higher compared to the national accounts. A comparison of employment information shows a 2 per cent lower estimate in the Dutch production census, and an 8 per cent lower estimate in the British production census compared to alternative employment statistics. As a result, the Dutch/UK productivity gap would have been about 12 per cent higher on the basis of information from the national accounts and employment statistics, compared to what was found in this study. [Tabular Data Omitted] ANNEX B. THE USE OF UNIT VALUE RATIOS

Some of the earlier cross-country comparisons of output and productivity (Rostas, 1948; Paige and Bombach, 1959; Mensink, 1966) were based on direct measures of physical quantities of selected major products (such as, tonnages). Nowadays, developed countries have such a complex output structure that it becomes very difficult to find enough representative quantity indicators for total real output.

Instead, price relatives for matched pairs of products in the two countries can be used to convert their total values of output to a common currency. The prices in this study represent unit values, which are obtained by dividing the ex-factory sales value by corresponding quantities. The sources are the Produktiestatistieken (CBS) and the Quarterly Sales Inquiries (BSO). The unit value ratios (UVRs) are expressed as guilders per pound sterling. [Some studies use the term `purchasing power parties' not only for expenditure comparisons but for output comparisons as well; however, this term only has a clear meaning in the expenditure approach, which is why the more neutral term `unit value ratios' is preferred.]

Of course it was not possible to match all products in the manufacturing industry of both countries. Products cannot be matched if the information on their sales value or quantity is not given, for example for reasons of confidentiality. Neither can unit value ratios be obtained if the products are too heterogenous in their characteristics; in other words quality problems reduce the number of `matcheable' items. Table B1 shows the percentage share of matched items in the ex-factory sales value of branches.

Table : Table 1. Ex-factory sales value of matched

 product items as a percentage of total sales for
 manufacturing branches Netherlands and the
 United Kingdom, 1984
 Sales of matched items as a %
 of total branch sales
 Netherlands United Kingdom
Food products 26.1 37.2
Beverages 77.3 60.8
Tobacco products 94.9 100.0
Textiles 32.6 24.1
Wearing apparel 50.3 35.6
Leather and footwear 47.4 35.9
Wood products 0.0 0.0
Paper products 49.2 32.2
Printing and publishing 0.0 0.0

Chemicals, rubber and plastic

products 17.4 14.4
Petroleum refining 0.0 0.0
Stone, clay and glass products 13.3 5.1

Basic metals, metal products, machinery and electric engineering and transport equipment 3.3 2.6

Instruments and other

manufacturing 0.0 0.0
Total manufacturing 14.9 12.5

The method of applying the UVRs for matched products to the output value in the manufacturing sector as a whole is largely derived from a study by Maddison and Van Ark (1988 and 1989), but it has been further refined in some aspects. The procedure was carried out in various steps: (1) Seventy-six of the 107 (UVRs) were located in one of 20 industries, for which more than 30 per cent of the sales value could be matched. Most of the industries were in the food, beverages-and tobacco branch (poultry, dairy prodtcts, vegetable, conserves, grain milling, bakery products; other food products; beer; soft drinks; and tobacco) or in the chemicals branch (fertilizers; synthetic resins, paints; soap and detergents); other branches with one or more industries with acceptable matching percentages were textiles, wearing apparel, paper products, and stone, clay and glass products. The UVRs for matched products were aggregated to the industry level at either Dutch or UK quantities, so that important products get more weight than less important products, that is: [Mathematical Expressions Omitted] where: [Q.sup.D] and [Q.sup.U] are Dutch and UK quantities

[P.sup.D] and [P.sup.U] are Dutch and UK unit values The value added in these industries was converted on the basis of the weighted average UVR for matched products. (2) In those industries within the branches mentioned under (1) for which less than 30 per cent of the total sales value or no products at all could be matched, value added was converted on the basis of the (quantity) weighted UVR for all matched products in the branch as a whole. (3) For metals, engineering and transport equipment some products could be matched, but their share in the total sales value of the branch was less than five per cent; here value added was converted by the arithmetic average of the UVR for the total of branches with sufficiently matched product items (mentioned under (1), weighted at their value added) and the (quantity) weighted UVR for the few matched products within these branches. (4) In the three remaining branches (wood products, printing and publishing, and other manufacturing) not a single product could be matched. Here value added was converted on the basis of the average UVR for branches with sufficiently matched product items (the branches mentioned under (1), weighted at their value added).

The question arises how representative the unit value ratios for matched products are for total manufacturing, even after the detailed aggregation procedure described above. For this purpose the 107 UVRs were tested on the basis of some hypotheses. For example, a strong negative relationship (significant at the 1 per cent level) was found between the importance of the matched items (in terms of their ex-factory sales value) and the corresponding unit value ratios. On the assumption that the `non-matched' items are to be found among the smaller items, the average UVR in this study would be understated. However, this assumption is questionable. Firstly, as mentioned above, quite a number of the more important items were not matched, because value or quantity information was not disclosed in the census for reasons of confidentiality, or because there were important quality differences between pairs of products. Secondly, many of the large items which were matched consist of a range of smaller products. These smaller products were lumped together because they could not be matched separately if product specifications were not exactly the same.

By dividing the 107 matched product items into final and semi-manufactured goods, it appeared that the arithmetic average of unit value ratios for final goods was about 15 per cent higher than for semi-manufactured goods. This gave rise to the hypothesis that some of the Dutch matched items might have been of a more `semi-manufactured' nature than their British counterparts, and therefore charge a lower absolute price. In general, the likelihood of such `mis-matches' appeared to be insignificant, because the specifications of the matched products were sufficiently detailed. For example the UVRs for the paper industry, which were exceptionally low at only 2.3 guilders to the pound on average - which therefore resulted in a relatively high productivity ratio - were carefully examined on the assumption of matching semi-manufactured Dutch paper products with final British paper products. The assumption was rejected: firstly, the UVRs for final goods in the paper industry (such as envelopes, toilet paper, kitchen towels and wallpaper) were also very low; secondly, a similar comparison between British and French paper products resulted in low unit value ratios as well, suggesting that British paper products have relatively high unit values.

As mentioned above, the unit value ratios relate to the ex-factory sales value of products (adjusted for the impact of indirect taxes). The method assumes that the price relationship between the two countries at the sales level is also representative for value added.

This so-called single indicator method is preferred over the alternative of calculating separate unit value ratios for inputs, because these are often subject to important measurement errors (see for example Paige and Bombach, p. 80; Smith, Hitchens and Davies, p. 162). However, for this study a price comparison was made for major energy inputs (natural gas, electricity and oil-based fuels) in Dutch and British manufacturing branches, which led to a downward adjustment of the original UVR by 1.5 per cent on average.

ANNEX C. CALCULATION OF HOURS WORKED

The calculation of the number of hours worked per person has become an important aspect of labour market accounting. Earlier studies showed that there are substantial differences in the average number of hours worked, even among industrialised countries. However, most estimates are for the economy as a whole (see for example Maddison 1987), and less attention has been given to the manufacturing sector separately.

Annex table C1 shows the percentage shares of the various components of labour time - leave, sickness, overtime - in the manufacturing sectors of the Netherlands and the United Kingdom. Potential labour time can be defined as normal weekly paid hours of work per employee, which is taken from national wage cost surveys, and multiplied by 52 weeks. Contractual labour time is derived by deducting time for annual leave, holidays and work-sharing agreements (arbeidstijdverkorting) from the potential labour time. Finally `actual hours worked per employee' are obtained by further deducting time for sickness, maternity leave and industrial disputes and by adding paid and unpaid overtime hours.

Table : Table C1. Components of potential labour time

 in manufacturing in the Netherlands and the
 United Kingdom, 1984
 (in percentage terms)
 Nether- United
 lands Kingdom
Potential labour time 100.0 100.0
 Annual leave -9.6 -8.9
 Holidays (bank-holidays, etc.) -2.4 -3.1
 Worksharing agreements[(sup.a)] -1.7 -
Contractual labour time 86.3 88.0
 Sickness -7.7 -3.6
 Strikes -0.0 -0.1
 Overtime +1.6 +7.5
Hours worked per employee 80.2 91.8

Number of hours worked per employee

per annum 1611 1749

Note:(a) Only the reduction of labour-time that was treated as extra leave-days. Reduction of weekly hours due to worksharing agreements is already accounted in the potential labour time.

The reduction of labour-time due to worksharing agreements is already accounted for in the potential labour time as far as it affects weekly hours (for example working 2 or 4 hours less each week). If reduced labour-time takes the form of additional enforced leave days (which was the case for 80 per cent of the agreements in the Netherlands in 1985, see Hesemans 1988), it is accounted for separately.

For the Netherlands all information on person-hours, including absence and overtime, was taken from the Wage Cost Survey for 1984. However, in this source the numbers relate to full-time employees only. The hours of full-time employees were therefore multiplied by the number of full-time-equivalent employees (see below for method of calculation) to obtain the total number of hours worked in each branch. This could then be divided by the number of employees (that is full-time and part-time) to arrive at the average number of hours per employee.

Figures for Dutch employees in establishments with 10 employees or more were taken from the Produktiestatistieken, and the number in establishments with less than 10 employees were taken from the Employment Census for September 1984. These figures were adjusted to an annual basis by using employment data from a research project on contractual hours at the Dutch Central Bureau of Statistics (see Bos, 1987, and Bos c.s., 1988). It provides information on the contractual hours per job (that is the average for full-time and part-time jobs) and per full-time job. One can obtain the number of full-time-equivalent employees by multiplying the number of employees by the hours per job and dividing it by the hours per full-time job.

A comprehensive estimate of annual hours worked per person was not directly available for the United Kingdom, not even for full-time employees. Normal weekly labour time was obtained from the New Earnings Survey 1984 (NES). However, NES excludes employees who are exempted from paying tax or National Insurance contributions. These are mostly part-timers working very few hours. Average hours for employees working less than 30 hours a week were therefore calculated on the basis of unpublished data obtained from the Labour Force Survey for 1984. Weekly hours were adjusted for short-time (taken from the Employment Department's `L2 Survey') and overtime (also derived from NES). Overtime in manufacturing amounted to 3 hours a week on average.

The average number of working weeks in Britain was then derived by first calculating the annual leave, holidays (8 days a year), sickness and industrial disputes. This information was taken from the following sources: * annual leave from Employment Department, Time-Rates of Wages and Hours of Work, Appendix III, `Holidays with pay'. * sickness and maternity leave from unpublished information on `days lost due to sickness' collected for the Labour Force Survey 1984. * industrial disputes from information collected by the Employment Department, as published in the Employment Gazette. It appeared that after deducting these various types of absence, the number of weeks actually worked amounted to 43.8. This was multiplied by the weekly hours to obtain the number of annual hours per employee.

Table C1 shows that the average number of hours worked per annum was about 8.5 per cent higher in Britain compared to the Netherlands. It appears that the loss of labour-time attributed to sickness (including maternity leave) is about twice as high in the Netherlands as in Britain. The other main source of the higher number of hours worked in the United Kingdom related to the higher amount of overtime hours worked in Britain, which is almost 3 hours a week on average compared to only 40 minutes in the Netherlands. Annual hours per employee for manufacturing branches are presented in table A1.

REFERENCES W Bos (1987), `Overeengekomen jaarlijkse arbeidsduur, jaargemiddelde aantal banen en arbeidsvolume van werknemers, 1977-1986; methode en resultaten van een proefonderzoek' (Contractual Annual Working Time, Average Number of Jobs and Volume of Labour of Employees, 1977-1986; Methodology and Results of a Pilot Study), Supplement bij de Sociaal-Economische Maandstatistiek, No. 5. W Bos, WP Leunis, AH Sprangers, CJ Veenstra and CG Verhage (1988), `Towards a System of Labour Accounts', Netherlands Official Statistics, No. 4. Business Statistics Office, Business Monitor, Report on the Census of Production, various issues, HMSO, London. Business Statistics Office Business Monitor, Quarterly Sales Inquiry various issues, HMSO, London. Business Statistics Office (1987), Business Monitor, Purchases Inquiry 1984, HMSO, London. Central Bureau of Statistics (CBS), Produktiestatistieken, various issues, 's Gravenhage. Central Bureau of Statistics, Nationale Rekeningen, various issues, Staatsuitgeverij, 's Gravenhage. Central Bureau of Statistics, Statistiek Werkzame Personen, various issues, 's Gravenhage. Central Bureau of Statistics, (1986a), Statistiek van het Ondernemingen-en Vestigingenbestand, 's Gravenhage. Central Bureau of Statistics, (1986b), `Lonen en Loonkosten', Sociaal-Economische Maandstatistiek, July. Central Statistical Office (CSO), United Kingdom National Accounts, various issues, HMSO, London. A Daly, DMWN Hitchens and K Wagner (1985), `Productivity, Machinery and Skills in a Sample of British and German Manufacturing Plants: Results of a Pilot Inquiry', National Institute Economic Review, February. SW Davies and RE Caves (1987), Britain's Productivity Gap, NIESR, Cambridge University Press. Department of Employment, Employment Gazette, various issues, London. Department of Employment (1984), New Earnings Survey 1984, HMSO, London. CH Feinstein (1988), `Economic Growth Since 1870: Britain's Performance in International Perspective', Oxford Review of Economic Policy, Spring. JCM Hesemans (1988), `Ontwikkeling van de arbeidsduur 1982-1987' (Trends in Hours of Work) Supplement bij de Sociaal-Economische Maandstatistiek, No. 3. A Maddison (1987), `Growth and Slowdown in Advanced Capitalist Countries', Journal of Economic Literature, June. A Maddison and B van Ark (1988), Comparisons of Real Output in Manufacturing,Working Papers, WPSS, World Bank, Washington D.C. A Maddison and B van Ark (1989), `International Comparison of Purchasing Power, Real Output and Labour Productivity: A Case Study of Brazilian, Mexican and US Manufacturing', Review of Income and Wealth, March.

GJA Mensink (1966), Comparisons of Labour Productivity in the United Kingdom and the Netherlands, 1958, CBS, Statistical Studies, No. 18, The Hague. OECD (1987), Purchasing Power Parities and Real Expenditures, Paris. Office of Population Censuses and Surveys (1986), Labour Force Survey 1983 and 1984, HMSO, London. D Paige and G Bombach (1959), A Comparison of National Output and Productivity, OEEC, Paris. SJ Prais (1981), Productivity and Industrial Structure, NIESR, Cambridge University Press. L Rostas (1948), Comparative Productivity in British and American Industry, Cambridge University Press. AD Smith, DMWN Hitchens and SW Davies (1982), International Industrial Productivity, NIESR, Cambridge University Press. H Steedman and K Wagner (1989), `Productivity, Machinery and Skills: Clothing Manufacture in Britain and Germany' National Institute Economic Review, May.

NOTES

(1) The author is grateful to statisticians at the Business Statistics Office (Newport), the Central Statistics Office (London), the Department of Employment (London), and the Central Bureau of Statistics (The Hague) for their advice and for providing additional information. Representatives from various industry branches provided useful comments for interpreting the results in this study. I am indebted to colleagues at the Training Agency, the University of Groningen, the Central Planning Bureau (The Hague) and the National Institute for their comments. I am most grateful to SJ Prais (NIESR) for help and advice at all the research. Financial support was provided by the Training Agency. (2) With many thanks to Stephen J. Davies for tracing this highly appropriate quotation. (3) For example, as an alternative it was assumed in some very early studies that the quantity ratios for matched items was representative for non-matched items, but this alternative was largely rejected in later studies; for an extensive discussion of this issue and related matters see Maddison and van Ark (1988 and 1989). (4) The `proxy PPP' for manufacturing products was estimated on the basis of the purchasing power parities for final expenditure on food, beverages and tobacco; clothing and footwear; furniture, household textiles and household appliances; personal transport equipment, and machinery and equipment in 1985 (from OECD, Purchasing Power Parities and Real Expenditures, 1987 Paris, part two, table 9). The result was close to a calculation by Donald Roy on the basis of more detailed categories, which he kindly provided. The `proxy PPP' for 1984 was then obtained on the basis of 1984-1985 changes in the national producer price indexes for manufacturing of the two countries. (5) Rostas (1948) showed that the average American output per person employed for 31 manufacturing industries was 215 per cent of the level in Britain at the end of the 1930s; according to Paige and Bombach (1959) the US/UK ratio for manufacturing industry as a whole was as high as 273 per cent in 1950. Smith, Hitchens and Davies (1982) showed that the average US output per person employed for 117 manufacturing industries was 289 per cent of the UK level in 1968. (6) Some words of caution must be added to prevent the reader from interpreting the `extrapolated' German/British productivity ratios for the 1980s too rigorously in this context. Firstly, although the method applied by Smith, Hitchens and Davies is basically the same as the one in this study, some of the concepts are slightly different. For example, in their study, output is defined as `net output', for which the cost of service inputs is not deducted from gross output. Moreover, the unit value ratios in their study are in a few cases derived from prices to the final purchaser instead of ex-factory sales prices. Secondly, indexes of production and employment within a country are most suitable for short-term comparisons, but become increasingly liable to errors as the years pass. For this reason these time series are usually re-based about every 5 to 10 years on the basis of a more recent census of production. The same need for revision applies to cross-country comparisons. The considerable decline in total employment in British manufacturing during the 1980s and the accompanying changes in the balance of manufacturing re-inforce the desirability for an up-to-date benchmark comparison of manufacturing productivity between Germany and the United Kingdom. However, as a proxy estimate the extrapolated German/British productivity ratio satisfies the present purpose. (7) This point is extensively discussed by Smith, Hitchens and Davies (1982, p. 25 and pp. 163-5). (8) The number of employees in Dutch textiles was cut by more than half in a a decade, from around 50,000 to less than 25,000 in 1985 (CBS, Produktiestatistieken). (9) See A Cockerill (1988), `Steel', in P Johnson (ed), The Structure of the British Industry, 2nd edition, London, p. 85. (10) The measure of value added per man-hour in the steel industry in this study shows a higher productivity gap than that which can be derived from a simpler measure of the total tonnage of crude steel per man-hour. The later measure, which would show tons of crude steel per man-hour in Britain at three-quarters of the corresponding Dutch level, des not take account of the relatively stronger involvement of the British steel industry in the production of more labour-intensive products. (11) NEDO, International Productivity Comparisons, Visits to German and French Paper Mills by Teams from the UK Paper and Board industry, Paper and Board EDC, 1983. (12) See for example a survey on the pulp and paper industry in the Financial Times, 13 December 1989, which argues that the British paper industry had difficulties to compete on a cost effective basis due to lack of investment in new machinery in the early 1980s. This is consistent with the high `ex-factory' price level of British paper products in this study, which was almost double the corresponding price level in the Netherlands (see also discussion in annex B). (13) The statistics on plant size for the United Kingdom appear to show that at present the average plant size in terms of numbers of employees is about 75 per cent of the average size in 1979. This is partly due to a statistical fallacy because of the introduction of a new business register in 1985 which increased the coverage of plants with less than 20 employees (see Business Monitor, Size Analyses of United Kingdom Businesses, 1985 and 1988). For the other part the decrease in average plant size is largely due to a substantial fall in the number of plants with over 500 employees. (14) A Dutch enterprise can consist of more than one establishment, but this hardly affects the comparison: the Dutch manufacturing sector comprised 8,903 enterprises (see sources table 4) employing more than 10 persons on 1 January 1985 compared to 9,073 establishments of the same size for 1984 (Produktiestatistieken.) (15) `Net output' is gross value added plus purchases of non-industrial service inputs; this is more suitable here because non-industrial services cannot be purchased from other manufacturing units. (16) HM Treasury (1989), Economic Progress Report, No. 201, April. (17) See for example, Swedish Employers' Confederation (1989), Wages and Total Labour Costs for Workers, International Survey 1978-1988, Stockholm, November; or U.S. Dept. of Labor, Bureau of Labor Statistics (1988), International Comparisons of Hourly Compensation Costs for Production Workers in Manufacturing, Washington D.C., February. In both studies labour costs are converted to a common currency on the basis of market exchange rates. (18) The Dutch-UK ratio of unit labour costs for 1987, which was calculated on the basis of the relative labour costs and the extrapolated productivity differential in that year, was also close to unity, whereas the price level of manufacturing products in Britain in that year was almost 15 per cent above that in Holland.