British and American Productivity Re-examined

March, 2007

British and American

Productivity Re-examined

A New Industry of Origin Benchmark of

Comparative Manufacturing Productivity for 1935

By

Pieter Jacob Woltjer1

Supervisors Dr. Herman de Jong

Dr. Marcel Timmer

Abstract

This thesis provides a new estimate of British and American comparative productivity for the manufacturing sector in the year 1935. Exploiting existing census data, I review the comparative levels of value added per worker and per man-hour employing a variant of the unit value approach akin to the double deflation methodology. The new estimates for the Anglo-American comparative manufacturing productivity illustrate that even though on an aggregate level the results corroborate existing estimates, they also reveal important differences at industry level and highlight the relative productivity lead of the United States. On average the American manufacturing productivity level was 218 percent on a per worker basis, and 274 percent on a per man-hour basis, when compared to the United Kingdom.

JEL Code: N64, O47

Introduction

Studies of comparative economic performance of nations have come to form an integral part of economics. Research based on the catch-up and convergence literature - popularized by Abramowitz, Romer and Baumol - relies heavily on these studies to provide empirical proof and backing for their theories of economic growth and long-term development.2 _{Extensive studies on comparative economic performance were already} performed in the 1940s and 1950s by Kuznets and Mitchell in the U.S. and Colin Clark in the U.K., who looked primarily at national income and product statistics. However, perhaps the best known comparison of long-run productivity performance is the work of Angus Maddison.3 Part of the appeal of his approach is the transparent methodology and sole reliance on national time-series published by statistical offices. Furthermore, the wide coverage in terms of countries and time-span makes it exceptionally well suited for research on economic growth. A common problem with the Maddison benchmark (or any long-term study on economic trends) however, is the matter of deflation. It is well known that problems of interpretation arise, when time-series of different origin are projected from a benchmark-year into distant periods. Particularly, changes in the relative price structures can cause large distortions in these time series, as the debate in the Journal of Economic History between Ward, Devereux and Broadberry has shown.4 Historically, as growth occurs, the composition of production, consumption and relative prices all vary, and the economic meaning of comparing real product per head based upon remote Purchasing Power Parities becomes entirely questionable; so it could occur that comparisons based upon PPP projections might generate larger errors than comparisons using conventional exchange rates.5 In addition, economists have increasingly argued that our ability to improve upon our understanding of economic growth is constrained by a severe lack of data.6 _{The long-term studies on economic}

Abramowitz, M. (1986), “Catching Up, Forging Ahead and Falling Behind”; Romer, P. (1986),

“Increasing Returns and Long Run Growth”; Baumol, W. (1986), “Productivity Growth, Convergence and Welfare: What the Long Run Data Show”

3

Maddison, A. (1995), Monitoring the World Economy 1820-1992; Maddison, A. (2001), The World

Economy: A Millennial Perspective 4

Ward, M. and Devereux, J. (2003), “Measuring British Decline: Direct Versus Long-Span Income Measures.”; Broadberry, S. (2003), “Relative Per Capita Income Levels in the United Kingdom and the United States since 1870: Reconciling Time-Series Projections and Direct-Benchmark Estimates.”; Ward, M. and Devereux, J. (2004), “Relative U.K./U.S. Output Reconsidered: A Reply to Professor Broadberry”

5

Eichengreen, B. (1986), “What Have We Learned from Historical Comparisons of Income and productivity?”

6

trends simply do not provide the level of detail necessary to systematically extend our knowledge of the comparative economic performance of nations.

A promising alternative to the long-term comparative economic performance studies is the so called industry of origin approach. Industry of origin studies focus on comparisons of output and productivity on a level of industries, thus furnishing valuable insights into the comparative economic structure and relative productivity. This type of research was popularized by Rostas who employed a quantitative approach for his interwar study for the United Kingdom, Germany and the United States, where he compared output per worker figures based on quantitative data on the production of comparable commodities and the number of employees producing these respective quantities.7 _{Paige and Bombach took this type of research one step further in their} comparison of national output and productivity of the United Kingdom and the Unites States, and adopted a so called unit value approach.8 _{Here unit value ratios (UVRs) - the} ratio of the average value of a single commodity for the countries under comparison - are used to construct industry, branch and sector PPPs; which in turn can be used to convert gross or net output in a single monetary unit. As noted by van Ark, this particular methodology proves to be more reliable and effectively provides more information regarding the productivity differences than the available alternatives will.9 _Furthermore the unit value approach can be extended to take account of differences in input per output of individual commodities, differences in the technical input-output coefficient of industries and variations in the price levels of inputs and outputs. This particular methodology is often referred to as double deflation, as the value added (or net output) is weighted using a combination of two PPPs, one for gross output and one for intermediate inputs.

Even though, as the discussion above has shown, the need for such coherent accounts of economic history are becoming increasingly pressing, there are currently only a handful of industry of origin studies that closely examine the pre-1950 period. For the interwar years, the only available study dealing with the British and American comparative productivity in the manufacturing sector is the work by Rostas, which dates

Rostas, L. (1943), “Industrial Production, Productivity and Distribution in Britain, Germany and the United States”

8

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

Kingdom and the United States 9

back from 1948.10 _{Unfortunately, this study lacks the detail and the systematic} comparative approach which is required in modern economic historical comparative productivity research, as I will show in section 2. In this thesis I intend to fill this hiatus with a new, more robust, real value added based comparisons of welfare and productivity. I will employ a variant of the unit value approach akin to the double deflation methodology; the results of which are presented in the table below. The new estimates for the Anglo-American comparative manufacturing productivity illustrate that even though on an aggregate level the results corroborate the existing estimate by Rostas, they also reveal important differences at industry level and highlight the relative productivity lead of the United States.

In the next section I will discuss both the U.K. and the U.S. censuses for the year 1935, which form the bulk of my data sources. Here I will identify and quantify possible inconsistencies between the censuses, particularly the issue of annual hours worked, which has a major impact on the British and American comparative level of labor productivity; as the table above clearly shows. In section 2 I will discuss the alternative methods by which output per worker can be compared in the same industry for different countries. The unit value approach will be discussed in detail in this section. Section 3 will illuminate the results of the comparative productivity study. In addition I will briefly elaborate on the exact procedure of productivity measurement which I applied, as lack of data on intermediate inputs ruled out the application of the double deflation technique. In Section 4 I will compare my findings to the estimate by Rostas and present an estimation of German and American comparative labor productivity levels based on a paper by Fremdling, de Jong and Timmer.11

1.

Sources and Data

Production censuses are the most suitable sources for cross-country comparisons of labor productivity levels; consequently census data forms the backbone of this thesis. For the United Kingdom, I consulted the Fifth Census of Production of 1935, published by the Business Statistics Office (BSO) of the Board of Trade.12 _{For the United States, I}

Rostas, L. (1948), Comparative Productivity in British and American Industry

11

Fremdling, R., de Jong, H. and Timmer, M. (2006), “British and German Manufacturing Productivity Compared”

12

used the Biennial Census of Manufactures of 1935 published by the Bureau of the Census of the U.S. Department of Commerce.13 _{These production censuses are} particularly well suited for productivity comparisons, since they contain detailed figures on physical quantities, prices, gross output, intermediate input and employment. As the information for outputs and inputs is based on one and the same questionnaire - for which the information is supplied by the same firms - internal consistency is guaranteed. A common drawback of census data however is that the production censuses do not always provide a complete picture of industrial activity; smaller firms in particular are not always required to report their production data, which might generate inconsistencies in cross-country comparisons. Countries might also differ in their definitions and concepts of gross output, intermediate input and employment. Before starting the comparison, I will briefly discuss the respective censuses and the possible differences in coverage and concepts between British and American censuses.

The aim of any cross-country comparison of relative industrial productivity is to get a better view of the realized and potential differences in labor productivity. Consequently the year under comparison should be carefully chosen, as business cycle and capacity utilization effects can have a significant influence on the outcome of the study. Laszlo Rostas addressed this very question in his study of British and American manufacturing productivity for the interwar period and came to the conclusion that 1937 is the best year for comparison.14 _{He argued that this was a representative year for both} the U.K. as well as the U.S., where the use of available capacity was roughly comparable. Unfortunately, statistical material is not fully available for the U.K. in this year; the Import Duties Act Inquiry of 1937 does cover some of the manufacturing branches, but the majority of the manufacturing industries were not yet tabulated or published by the time the Second World War broke out.15 _{To overcome this problem} Rostas chose to compare the British year of 1935 with the American year of 1939 and work towards 1937. The primary reason Rostas relied so heavily on the American 1939 Biennial Census of Manufactures and not the 1935 or 1937 censuses, is that it met the sizable data requirements of the quantitative study he employed.16 _{For the purpose of} this thesis, both the 1935 censuses for the U.K. and the U.S. are perfectly suited for the

U.S. Department of Commerce, Biennial Census of Manufactures, 1935

14

Rostas, L. (1948), Comparative Productivity in British and American Industry¸ p 24

15

Board of Trade, Preliminary Reports of the Import Duties Act Inquiry, 1937

16

This particular census is part of the 1940 decennial census and contains detailed figures on the size of plants, horse power of machinery installed, etc. See: U.S. Department of Commerce, Sixteenth Decennial

task at hand. The 1935 comparison provides a good view of the realized labor productivity levels and, if necessary, can be adjusted for the effects of the business cycle, capacity underutilization and long term productivity movements. One way to realize this adjustment is to make use of existing productivity time series to calculate the average movement in productivity levels in the manufacturing sector for both countries between 1935 and 1937. This estimation shows that productivity would not change in the U.S. compared to the U.K. when the more representative year 1937 is used for the comparison.17

The production censuses embrace the entire manufacturing sector, but also contain data on mining, construction works, public utilities and government industries. For the purpose of this thesis I will exclude the latter industries and focus primarily on the manufacturing sector. The remainder of the British census distinguishes between 108 manufacturing industries or trades, whereas the American census encompasses no less than 327 industries and sub-industries. I have reclassified the industries from both countries into 12 branches and 95 common industries based primarily on the U.K. Census (see Appendix A for further details). Admittedly, this arrangement is the easiest to implement - as the more detailed American census can be readily fitted into the British framework - but it also adheres to the classification of the British-German comparison by Fremdling, de Jong and Timmer; making it feasible to compare the two as I will show in the fourth section.18

The British canvass covers the whole of Great Britain and Northern Ireland, and includes all productive operations in the United Kingdom. Following the method adopted at the 1930 census, detailed returns were not obtained from firms employing not more than 10 persons as a yearly average; the only information required from these firms was a statement of the nature of their business and the average number of their male and female employees in the year. From this data Rostas estimated that about 55,200 firms, employing 536,600 persons, were excluded from the U.K. census.19 _{The average} number of operatives employed during the year, covered by the census, was 4,441,300, whereas the number of administrative, clerical and technical staff employed amounted to approximately 672,500. The U.S. census covers the 48 contiguous states of the mainland of America and the District of Columbia (thus excluding Alaska and Hawaii),

Broadberry, S. (1997), The Productivity Race, p 44

18

Fremdling, R., de Jong, H. and Timmer, M. (2006), “British and German Manufacturing Productivity Compared”

19

and its returns fully represent the calendar year’s 1935 operations. All establishments that had been engaged in manufacturing, printing or publishing during any part of 1935, and whose production during the year had been valued at $5,000 or more, were required to participate in the inquiry.20 _{For the U.S. census, the reported average number of} operatives and staff employed were 7,182,700 and 1,059,100 respectively. The less stringent exemption levels of the U.K. census allowed for a greater number of small firms to be excluded from the figures for Great Britain, which will undoubtedly influence the levels of comparative productivity. In line with the estimate by Fremdling, de Jong and Timmer, this would result in a downward bias of average productivity for U.S. manufacturing as a whole vis-à-vis the U.K. of 2 percent, but certainly not more.21

Even though the terminology employed by the British and the Americans differs slightly, the concepts of gross output, intermediate input and value added are equivalent for both censuses. The gross output is the ex-factory value of products (the selling value at the factory or plants), whereas intermediate input represents the cost of materials, fuel and contract work. The net output, or value added, is the figure which results from the deduction of the intermediate input from the total value of gross output. This figure constitutes the sum of wages, salaries, rent, royalties, rates and taxes, depreciation of plant and machinery, advertisement and selling expenses and all other similar charges, as well as profits.

In some cases, excise duties and consumer taxes were included in the ex-factory value of products. To obtain a comparable measure of productivity at the industry level, I deducted these excises from the value of gross output. For the U.K., I subtracted excises on Silk and Artificial Silk, Sugar and Glucose, Beer, Aerated Waters, Tobacco, Chemicals, Dyestuffs and Drugs, Matches and Printing. In addition I adjusted the gross output value of Tobacco for the U.S. All values of the excise duties have been taken from the respective censuses, with the sole exception of the U.K. excise on Tobacco, which has been based on the estimate in the study of British and German manufacturing productivity by Fremdling, de Jong and Timmer.22

As the value of production per employee was $5,450 in 1935, the scope of the U.S. census is considerably wider than the British census. The American census did not require the exempted firms to produce any information regarding their total workforce or the nature of their business however, making it difficult to determine the percentage of the workforce that was excluded from the survey.

21

Fremdling, R., de Jong, H. and Timmer, M. (2006), “British and German Manufacturing Productivity Compared”

22

In cross-country productivity comparisons the concept of value added per worker is more commonly applied, even though in a purely technical sense the concept of value added per man-hour is the more relevant one. For the present inquiry this distinction is of particular importance since the average American working week for the manufacturing sector in the interwar period was substantially shorter than the average British working week, which will result in a sizable downward bias of U.S. comparative productivity when figures for value added per worker are used. Colin Clark estimated that the average 1935 working week in Great Britain was 47.8 hours, whereas this was only 37.2 hours in the United States;23 _{likewise, Rostas presumed the average length of the U.K. and U.S.} working week to be 47.8 and 36.6 hours respectively.24 _{Unfortunately, the values} reported in the 1995 study by Maddison are at variance with these figures, it appears that Maddison overstated the average length of the American working week.25 The differences in working hours, as described by Colin Clark and Rostas, are confirmed by Jones and the Year-book of Labour Statistics of 1939.26 _{In addition, the values given by} Maddison for the total annual hours worked in the German manufacturing sector are at odds with the data published in the Statistisches Handbuch vor Deutschland.27 _The actual hours of work per week for the U.K., U.S. and Germany from 1929 up to 1938, as published by the latter sources, are given in figure 1.1. In light of this overwhelming evidence, I can only conclude that the data by Maddison must be rejected.

Figure 1.1 below also shows that 1935 was not the only year in the post-depression period where the American working week in the manufacturing sector was shorter than the British working week. From 1929 onwards the British and German working weeks became slightly shorter due to the introduction of short-time, but by 1934 the average working week had basically returned to its pre-depression level. The American average working week on the other hand dropped by nearly 30 percent in the period 1929 to 1934 and it never attained its pre-depression level again, even after the outbreak of the war. As made obvious by Huberman and Minns, the reasons for this

Colin Clark, M. (1951), The Conditions of Economic Progress, p 68

24

Rostas, L. (1948), Comparative Productivity in British and American Industry¸ p 29

25

Maddison, A. (1995), Monitoring the World Economy 1820-1992

26

Jones, E. (1963), "New Estimates of Hours of Work per Week and Hourly Earnings, 1900-1957"; International Labour Office, Year-book of Labour Statistics, 1939

27

considerable deviation are diverse - ranging from cuts in work time and the introduction of work sharing for unemployed - but lay beyond the scope of this thesis.28

Figure 1.1 Weekly Hours Worked Manufacturing, U.K., U.S. and Germany (1929-1938)†

†) Sources:

- Germany from Länderrat des Amerikanische Besatzungsgebiet, Statistisches Handbuch vor Deutschland

1928-1944, 1949 and Huberman, M. and Minns, C. (2005), “Hours of Work in Old and New Worlds”, p 27

- U.K. from Hart, R.A. (2000), “Hours and Wages in the Depression: British Engineering 1929-1938” and Colin Clark, M.A. (1951), The Conditions of Economic Progress

- U.S. from Jones, E.B. (1963), "New Estimates of Hours of Work per Week and Hourly Earnings, 1900-1957"

Regardless of the causes of the considerable deviations in working hours, it is clear that adjustments must be made, not only for the substantial differences in hours worked per week, but also for differences in the number of holidays and vacations and variations in working hours per branch. The Year-book of Labour Statistics of 1939 contains detailed statistics on the average hours of work per worker per week for several industries and industry-groups for both the U.K. and the U.S., which I weighed by employment to obtain a branch classification that adheres to the British census of production.29 Data by Huberman and Minns on the number of vacation and holidays for the U.K. and U.S. in

Huberman, M. and Minns, C. (2005), “Hours of Work in Old and New Worlds”

29

1938 allowed me to construct the figures for the annual hours worked. The data on weekly and annual average hours worked in 1935 are presented in table 1.1 below.

Table 1.1 Weekly and Annual Average Hours Worked, U.K. and U.S. (1935)

U.K. U.S.

Branch / Sector Weekly

hours worked† Annual hours worked† Weekly hours worked Annual hours worked Textile Trades 48.5 2,312 35.8 1,774 Leather Trades 48.5 2,312 38.6 1,915 Clothing Trades 47.0 2,242 32.2 1,597

Iron and Steel Trades 47.5 2,264 36.6 1,813 Engineering, Shipbuilding and Vehicle Trades 47.5 2,264 36.5 1,809 Non-ferrous Metals Trades 47.5 2,264 37.1‡ 1,838‡ Food, Drink and Tobacco Trades 48.4 2,306 39.5 1,962 Chemical and Allied Trades 48.3 2,302 38.1 1,892 Clay and Building Materials Trades 48.1 2,293 36.5 1,812

Timber Trades 47.2 2,250 39.5 1,958

Paper Trades 48.5 2,312 38.2 1,896

Miscellaneous Trades 47.5‡ _2,264‡ _{33.9 1,682}

Manufacturing Sector 47.9 2,283 36.6 1,817

†) The U.K. working hours were adjusted upwards by approximately 1 percent to compensate for the difference between normal hours and hours actually worked.

‡) The sector averages - provided by the 1939 Year-book of Labour Statistics - were used when detailed industry level data for hours worked was unavailable.

2.

Methods of Productivity Comparison

Broadly speaking there are three alternative methods by which output per worker can be compared in the same industry for different countries; (i) the sample approach; (ii) the quantitative approach; and (iii) the valuation approach.30 _{Before presenting a brief} overview of the strengths and weaknesses of the fore mentioned methods, I would like to note that although in principal these three methods serve the same purpose, each of them provides specific information indispensable to explaining the variations in productivity between countries; meaning that they are in no way mutually exclusive, but are in effect complementary to one another.

The first (sample) approach is based on the comparison of the performance of a small number, or sample, of mills or factories producing identical products under broadly identical conditions. An example, relevant to this thesis, is the Platt Report; an inquiry into the relative productivity differences in the cotton industry, between the United Kingdom and the United States during the Second World War and the preceding period.31 _{The main advantage of this approach is that it ascertains data on the} productivity of labor employed in the industry, but it also provides valuable information regarding the relative productivity in the separate phases of the productive process. This makes this type of research ideally suited to identify the main factors owing to the productivity differences, as was the case for the Platt report.32 _{A major drawback of this} method is that by trying to equate a small number of typical factories one basically eliminates the effects of some of the structural factors behind the productivity differences, for instance location and size of firms. Furthermore, determining what a representative factory is and what representative products are will always remain a somewhat arbitrary process, as it is often not possible to find two identical factories producing identical products under broadly the same conditions. Finally, due to its time-consuming nature, this type of research usually focuses on a single or small group of industries and thus provides only a limited view of the differences in labor productivity between nations.

Rostas refers to these three methods as “(i) the sample method, (ii) the global method, and (iii) the net output value method”. Although the general meaning of these methods is essentially the same, I chose not to use these designations since the latter does not make an adequate distinction between the comparison of gross output values and value added figures, which will prove to be a key element in this thesis. See: Rostas, L. (1948), Comparative Productivity in British and American Industry, pp 6-10

31

Ministry of Production (1944), Report of the Cotton Textile Mission to the Unites States of America

32

The second (quantitative) approach compares output per worker figures based on quantitative data on the production of comparable commodities and the number of employees producing these respective quantities. The rather straightforward computations and readily interpretable results are the main appeal of this approach. Still, it must be noted that in practice various factors will complicate the application of this method. The main disadvantage lies in the fact that it is very difficult to find comparable commodities and maintain a sufficiently high coverage ratio for the entire industry or branch that is being investigated. Usually allowances will have to be made for differences in quality; but a great many non-measurable differences in quality will be neglected, which inevitably will influence the final outcome of the study. Other issues concerning this approach will be discussed in a later paragraph dealing with one of the first major attempts to compare real output and productivity by industry; a study by Laszlo Rostas which utilizes this method for a comparison of British and American manufacturing productivity for the years 1935 and 1937.33

The third (valuation) approach is based on a comparison of the value of output per head in the countries under scrutiny, converted into the same monetary unit at the purchasing parity or exchange rate. As with the quantitative approach, a single study utilizing this method is able to deal with a great number of industries at the same time, providing comparable data on labor productivity for, for instance, the entire manufacturing or agricultural sector. The difficulties that arise with this approach are primarily related to the ascertainment of a suitable conversion factor. The official exchange rate or existing expenditure Purchasing Power Parities (PPPs) can be used for economy-wide international comparisons of productivity; an example of this approach is the well-known comparison of long-run productivity performance by Angus Maddison.34 As Paige and Bombach demonstrated however, more appealing conversion factors can be obtained by constructing new, so called industry of origin, PPPs from either output price data alone (single deflation) or from price data for both outputs as well as intermediate inputs (double deflation).35 _{A drawback of this approach, and the double} deflation procedure in particular, has been the paucity of comparative price data. Consequently, despite the general consensus that this is the preferred approach for sectoral comparisons, studies utilizing this approach have been rather scarce;

Rostas, L. (1948), Comparative Productivity in British and American Industry

34

Maddison, A. (1995), Monitoring the World Economy 1820-1992

35

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

particularly for international comparisons in the first half of the 20th _century.36 Nonetheless a recent study by Fremdling, de Jong and Timmer proofs that double-deflation is feasible for a cross-country productivity comparison of the manufacturing sector using census data predating the Second World War.37 _{The industry of origin} approach will form the basis of this thesis as well, details of which will be discussed below.

Quantities versus Prices

Both industry of origin approaches (the quantitative approach and the single and double deflated valuation approaches) are closely related. Not only are they very similar in their methodology, they also make use of the same census data and they are both able to make predictions regarding sectoral as well as economy-wide labor productivity. Van Ark argues that when output is fully covered in both countries, the ‘physical quantity’ (or quantitative) method and the ‘unit value’ (or single deflation) method lead to exactly the same result; these two methods are in fact each other’s mirror-image.38 _{This begs the} question, is there a need for another industry of origin comparison of the British and American manufacturing sector for the interbellum, when this has already be done in great detail by Rostas?39 _{In my opinion this question should be answered with a} resounding “yes”. There are two primary reasons to support this claim; the single- and double deflated approaches are in many ways more reliable than the quantitative approach; and, the single- and double deflation approaches offer additional information that allows us to identify the main factors owing to the productivity differences.

As previously mentioned, it is very difficult to find comparable commodities and maintain a sufficiently high coverage ratio using the quantitative approach. In a comparison of physical quantities one necessarily has to confine oneself to industries that have a rather simple product structure (e.g. industries that produce only one or a few products or industries that produce approximately homogeneous products). Unfortunately these specific types of industries are in practice scarce at best. To make matters worse, finding industries that adhere to these requirements in a cross-country

A detailed overview of studies in international comparisons in manufacturing has been published in: van Ark, B. (1993), International Comparison of Output and Productivity

37

Fremdling, R., de Jong, H. and Timmer, M. (2006), “British and German Manufacturing Productivity Compared”

38

van Ark, B. (1993), International Comparison of Output and Productivity, p 14

39

comparison is even more difficult, mainly due to the fact that one has to deal with differing market requirements, climates, tastes, etc. Regardless, in order to study the productivity of labor in terms of physical output per worker, roughly comparable principle products have to be selected for which the quantity ratios can be computed. Products of the given product class can be included in the total quantity of the principle product by either taking the aggregate directly or, when the product structure is more complex, introducing a weighting factor to aid the aggregation. The resulting quantity ratio of the aggregate product for a given industry neglects conceivable variations in quality entirely, for which allowances will have to be made in order to get credible results. The weighting of products and the compensation of quality differences complicates the application of the quantitative industry of origin comparison considerably.

As it is generally easier to match products by using the unit value approach than by using the quantitative approach, a comparison using the former method should always yield higher coverage ratios. Regardless Rostas, using the latter approach, reports an overall coverage ratio of about one half of the value of net output in Britain and two-fifths of the value of net output in the U.S., which is generally higher than in unit value industry of origin studies.40 _{Unfortunately, it is not possible to retrace the steps} taken by Rostas - as he has published very little on the non-aggregated quantity figures or on the compensation of quality differences he applied - making it exceedingly hard to either confirm or negate his claim. Still, from the published data it is possible to deduce that the 31 product groups covered by Rostas indeed embrace approximately 48 percent of manufacturing employment of Great Britain.41 _{The ‘machinery’ section however, which} encompasses 10 percent of employment in the manufacturing sector, should be deducted from this figure. Owing to the extremely complex structure of the mechanical and electrical engineering industries, Rostas chose to compare output prices (converted at the official exchange rate) instead of quantities for this section. Unfortunately there is no guarantee that the internal price level and structure of a country coincides with world market conditions, a required condition for the official exchange rate to be a valid deflator. Consequently this part of Rostas’ research will not result in meaningful figures, and in my opinion should thus be excluded entirely. The remaining product groups consist firstly of either single products or small groups of comparable commodities and

Rostas, L. (1948), Comparative Productivity in British and American Industry, pp 27-28

41

I solely examined the British coverage ratios, as Rostas’ data for the U.S. has been taken from various censuses (1935, 1937 and 1939) making it considerably more difficult to come up with a single,

secondly of groups that represent a large number of commodities (in some cases well over 50 products) with widely differing product characteristics. The former groups, which cover approximately 13 percent of manufacturing employment, are perfectly suited for the quantitative comparison of Rostas. The latter groups however are not, but Rostas has treated them as single groups nonetheless; mainly due to the lack of detailed employment data. The resulting productivity figures for these groups heavily depend on the weighting factors introduced to overcome differences in quality for these considerably more complex product structures. In my opinion Rostas is unable to substantiate his claim that the latter group of products, that represents approximately 24 percent of manufacturing employment, can be reduced to a sufficient degree of homogeneity to successfully apply a productivity comparison based on physical quantities. Also, in light of other similar studies (e.g. Frankel, who reports coverage ratios of 16 and 18 percent of manufacturing employment for the U.K. and U.S. respectively) the coverage ratios presented by Rostas appear to be considerably overstated.42

Another factor to consider is the representativity of comparison of matched output for non-matched output. Van Ark notes that a general consensus has emerged that the representativity of measured prices for unmeasured prices is better than that of measured quantities and unmeasured quantities.43 _{Apart from the reliability issues, the} unit value approach effectively provides more information regarding productivity differences than the quantitative approach. Most importantly, a study based on the unit value approach provides figures for labor productivity using data on gross output as well as value added.44 _{In summary, the unit value approach will provide a higher coverage} ratio, takes quality differences into account, provides a better measure of the non-matched products, and the value added comparison adds considerably to the available knowledge on labor productivity. Therefore, in my opinion, this thesis will prove to be a valuable addition to the available literature on comparative productivity in the British and American manufacturing industry.

Frankel, M. (1957), British and American Productivity, p 16

43

van Ark, B. (1993), International Comparison of Output and Productivity, p 14

44

Single- and Double Deflation

Although the single- and double deflation procedures are quite similar in their implementation, double deflation is the preferred method as it takes variations in input per output of individual commodities into account. Single deflation will inevitably give too much weight to those of the industry’s products with relatively high inputs per unit of output (and vice versa), since these commodities are weighted by ex-factory output prices instead of value added figures. Furthermore, in cross-country comparisons, differences in the technical input-output coefficient of an industry can result in notable variations in the results between single- and double deflation. For instance, the labor productivity of a country relying heavily on imported or otherwise expensive materials may be considerably overstated when the single deflation method is used. Another factor that needs to be taken into account is the difference in price level between inputs and outputs; the single deflation approach implicitly assumes that price movements of inputs are similar to those of outputs, while this is not necessarily the case.

In practice however, apart from its considerable data requirement, the volatility of the deflated value added measure poses a problem. As Paige and Bombach noted, the accuracy of the individual industry value added comparisons depends on consistent measurement of the inputs and outputs of each industry.45 _{Inconsistencies appear to be} unavoidable, and their effect may be magnified by the fact that value added is obtained as a residual of gross output and intermediate input. The errors in inputs and outputs are likely to be correlated however, so that the cumulative effect might be negligible. Studies applying the double deflation technique confirm that the method itself is feasible and generates reliable results in line with expectations.46 _{Nonetheless, for the sake of} completeness, both the results from the single- as well as the double deflation (when available) will be presented in this thesis. Below I will provide a very brief description of the single and double deflation methodology.47

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

Kingdom and the United States, pp 80-81 46

Fremdling, R., de Jong, H. and Timmer, M. (2006), “British and German Manufacturing Productivity Compared”

47

The cross-country comparison of unit values (uv) forms the basis of the industry of origin PPPs, which are used to compare the value of output per head in the countries under scrutiny. A unit value is essentially the average value of a single unit of a commodity or group of similar commodities. A unit value of good ‘i’ which is part of industry ‘j’ of country ‘A’ can thus be derived by dividing the ex-factory output value (o) by the produced quantity (q) for that good.

A ij A ij A ij

q

o

uv

=

In bilateral comparisons (between pairs of countries), the ratio of two unit values, the so-called unit value ratio (UVR), indicates the relative producer price of the matched product. A UVR with country ‘A’ taken as the base country would look like equation (2).

A ij B ij A AB i

uv

uv

UVR

=

The UVRs are aggregated to obtain the Gross Output Purchasing Power Parity (GOPPP). In this study, the UVR’s are weighted several times; first according to their share in the total value of matched products for the industry, then according to the industry share in the branch and finally according to the branch share in the manufacturing sector. In all cases two sets of indices will be constructed; one using the weights of the base country and one using the non-base country’s weights. In a bilateral comparison, when one assumes country ‘A’ to be the base country, the first GOPPP presented below (3) is an index of the Laspeyres type, whereas the weights of the (non-base) country ‘B’ would have to be used for constructing a Paasche index (4). It may be worth noting that generally Laspeyres PPPs will be higher than Paasche PPPs. Kravis refers to this phenomenon as the “Gerschenkron effect”, which arises from the fact that each country’s gross output structure adapts itself to the country’s own price structure; where gross output tends to be large when prices are low and vice versa.48 _Hence, valuation of gross output by a set of foreign quantities tends to inflate its aggregate

value. The Laspeyres and Paasche GOPPP for industry ‘j’ with a total of ‘s’ matched products and country ‘A’ as base country, are thus given by equation (3) and (4).

∑

∑

∑

∑

The Fisher index, which is the geometric mean of the Paasche and Laspeyres indices, will be predominantly used in the remainder of this study. The Intermediate Input Purchasing Power Parity (IIPPP) is constructed in a similar fashion; however, instead of output values, input values (v) are used. An example of a Laspeyres IIPPP is presented below. B ij A ij A ij B ij s i s i A ij A ij A AB j q v q v v v IIPPP ⋅ ⋅ ⋅ =

∑

∑

When the GOPPP - which is the single deflated gross output PPP - and the IIPPP are constructed, the double deflated value added Purchasing Power Parity (VAPPP) can be derived. The Laspeyres and Paasche VAPPP for industry ‘j’ are given by (6) and (7) respectively; where GO and II denote the value of Gross Output and Intermediate Inputs.

3.

Results

This section will illuminate the results of the comparative productivity study, summarized in the introduction. Before starting a discussion of the final results however, I will briefly elaborate on the exact procedure of productivity measurement which I applied. This procedure deviates slightly from the methodology presented in the second section due to some data restrictions. Although data with regard to gross output is quite abundant in both the U.K. and the U.S. censuses, specific pricing and quantity data for intermediate inputs is unavailable in the American census for the majority of branches. Consequently, Intermediate Input and Value Added PPPs could only be drawn up for the Textile and Iron and Steel Trades. The lack of data on intermediate inputs thus rules out the application of the double deflation technique, even though industry specific data on value added is readily available. Nonetheless, I chose to present not only the results of the single deflation (the output per worker/man-hour figures) but also the figures on value added per worker/man-hour, deflated by so called ‘adjusted single indicator’ Value Added PPPs. Below I will address the representativeness of these PPPs and in section 4, I will discuss the option of performing a full double deflation study and framing ‘double deflated’ Value Added PPPs by using data from input-output tables. The remainder of this section contains a discussion of the Purchasing Power Parities and the relevant coverage ratios, as well as an in-depth discourse of the final results.

Purchasing Power Parities

the number of matches differ across branches; this can be explained by the availability and heterogeneity of products, by differences in quantity specifications, the unique national character of some products and by differences in quality across countries.49 Coverage ratios were highest for the Food, Drink and Tobacco Trades and the Textile Trades, together accountable for approximately one-third of the total value and employment of production of the entire manufacturing sector (see table 3.4). The Paper Trades yielded the lowest coverage ratios, mainly because printing products could not be matched.

Table 3.1 Coverage Ratios and Matched Products, U.K. and U.S. (1935)

Gross Output PPP Intermediate Input PPP Branch / Sector Cov.

ratio (U.K.) Cov. ratio (U.S.) Nr. of products matched Cov. ratio (U.K.) Cov. ratio (U.S.) Nr. of products matched Textile Trades 65.7 45.2 42 63.2 40.7 44 Leather Trades 40.3 47.4 6 Clothing Trades 37.6 36.3 20

Iron and Steel Trades 42.9 37.9 26 49.9 41.3 23 Engineering, Shipbuilding and Vehicle 28.0 36.0 44

Non-ferrous Metals Trades 37.8 20.5 26 Food, Drink and Tobacco Trades 64.4 52.5 41 Chemical and Allied Trades 41.6 55.0 83 Clay and Building Materials Trades 41.6 33.3 13

Timber Trades 15.8 45.0 13

Paper Trades 19.2 14.0 17

Miscellaneous Trades 46.3 31.2 34

Manufacturing Sector 45.1 40.4 365 58.5† 41.0† 67† †) Due to the lack of data, valid Intermediate Input PPPs could only be calculated for the Textile Trades and the Iron and Steel Trades; the manufacturing sectors total for the Intermediate Input PPPs thus represent only the latter branches.

In some cases it was possible to use output UVRs for the framing of input PPPs, particularly for branches which use outputs from other branches as inputs. The Clothing Trades for example gets the majority of its inputs from the Textile Trades, whereas iron and steel outputs serve as intermediate inputs for the Engineering, Shipbuilding and Vehicle Trades. Available U.K. data on quantities and prices for intermediate inputs allowed for the proper allocation of these output UVRs to the input PPPs. Furthermore, limited data in the American census on inputs (for the Engineering, Shipbuilding and

Vehicle and Food, Drink and Tobacco Trades) made it possible to construct some additional input UVRs. The coverage ratios for these Intermediate Input PPPs, which are generally rather low, are not presented in table 3.1 as these ratios, and the Intermediate Input PPPs they represent are not suited for a full-scale double deflation analysis, but merely serve as a rough estimate. Overall the resulting double deflated Value Added PPPs are comparable to the adjusted single indicator Value Added PPPs, which will serve as the main deflator for this thesis and which will be presented below.

Table 3.2 provides an overview of the Gross Output and Intermediate Input PPPs per branch that could be framed with the available data. The table shows the Laspeyres and Paasche PPPs as well as their geometric mean, the Fisher index. The PPPs are presented in dollars per pound, but can very easily be converted into pounds per dollar without any loss of generality.50 Gross Output PPPs are generally high in Textile, Leather, Food, Drink and Tobacco and Clay and Building Materials Trades. Ex-factory output prices for these branches were especially high in the U.S. compared to the U.K. relative to other branches. The output PPPs of the Timber Trades were particularly low, which might be explained by low U.S. prices for tree trunks, the main intermediate input of this branch. In all but one of the cases the Laspeyres PPP is higher than the Paasche PPP, implying that the Gerschenkron effect does indeed occur. The non-existence of a Gerschenkron effect for the Leather Trades implies that consumer preferences are not fully reflected in price setting. Still there is no reason to assume that the formation of prices and the allocation of production were being distorted in the U.S., as the Leather Trades makes up only an inconsiderable portion of the manufacturing sector. The overall Fisher Gross Output PPP is 4.74 dollar per pound, which is remarkably close to the official exchange rate of 4.94 dollar. Still, the large cross-industry variation of the output PPPs - not uncommon for industry of origin studies - show that the exchange rate would function poorly as a PPP. The two available Intermediate Input PPPs show little variation; both across industries and compared to their output equivalents.

Table 3.2 Gross Output and Intermediate Input PPPs, U.K. and U.S. (1935) Gross Output PPP

($/£)

Intermediate Input PPP ($/£) Branch / Sector

Laspey-res Paa-sche Fisher Laspey-res Paa-sche Fisher Textile Trades 6.4 5.5 5.9 6.0 5.0 5.5

Leather Trades 5.6 5.9 5.8

Clothing Trades 5.2 4.6 4.9

Iron and Steel Trades 5.6 5.4 5.5 5.6 5.2 5.4 Engineering, Shipbuilding and Vehicle 4.1 3.6 3.8

Non-ferrous Metals Trades 5.4 5.2 5.3 Food, Drink and Tobacco Trades 6.2 5.4 5.8 Chemical and Allied Trades 5.1 3.2 4.0 Clay and Building Materials Trades 5.4 5.4 5.4

Timber Trades 2.2 2.2 2.2

Paper Trades 4.0 3.7 3.9

Miscellaneous Trades 6.2 4.4 5.2

Manufacturing Sector 5.3 4.2 4.7 5.9† 5.1† 5.5† †) Due to the lack of data, valid Intermediate Input PPPs could only be calculated for the Textile Trades and the Iron and Steel Trades; the manufacturing sectors total for the Intermediate Input PPPs thus represent only the latter branches.

Value Added PPPs can be derived for the Textile and Iron and Steel Trades using the available Gross Output and Intermediate Input PPPs; Section 2 illustrates how the Laspeyres and Paasche VAPPPs can be constructed. The resulting double deflated (DD) VAPPPs are presented in the second set of columns in Table 3.3. Alternatively, the available gross output UVRs can be used to construct so called adjusted single indicator (ASI) VAPPPs. This method has been used in various earlier industry of origin studies and tends to provide more robust results than an incomplete double deflation study can.51 _{In his 1993 study, van Ark shows that the ASI VAPPPs can easily be derived by} weighting the output UVRs by the value added of the corresponding industries instead of by gross output, as was the case for the GOPPPs.52 _{This can be done by replacing} equations 3 and 4 of section 2 by equation 8 and 9 below.

∑

∑

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

Kingdom and the United States, p 82 52

∑

∑

Table 3.3 Value Added PPPs, U.K. and U.S. (1935) Adjusted Single Indicator Value Added PPP ($/£) Double Deflated Value Added PPP ($/£) Branch / Sector

Laspey-res Paa-sche Fisher Laspey-res Paa-sche Fisher Textile Trades 6.3 5.3 5.8 7.1 6.3 6.7

Leather Trades 5.6 5.9 5.8

Clothing Trades 5.2 4.8 5.0

Iron and Steel Trades 5.6 5.4 5.5 5.6 5.6 5.6 Engineering, Shipbuilding and Vehicle 4.2 3.6 3.9

Non-ferrous Metals Trades 5.3 4.9 5.1 Food, Drink and Tobacco Trades 6.3 5.6 5.9 Chemical and Allied Trades 4.8 3.2 3.9 Clay and Building Materials Trades 5.4 5.5 5.4

Timber Trades 2.2 2.2 2.2

Paper Trades 3.8 3.4 3.6

Miscellaneous Trades 6.5 4.4 5.4

Manufacturing Sector 5.2 4.1 4.6 6.5† 5.9† 6.2† †) Due to the lack of data, double deflated Value Added PPPs could only be calculated for the Textile Trades and the Iron and Steel Trades; the manufacturing sectors total for the double deflated Value Added PPP thus represent only the latter branches.

apply for all branches. The double deflated VAPPP for the Textile Trade for instance is raised because of relatively inexpensive U.S. intermediate inputs for this industry. Unfortunately, with the currently available data on prices and quantities for the intermediate inputs, it is not possible to construct more reliable estimates for the double deflated Value Added PPPs. Consequently the (Fisher) adjusted single indicator Value Added PPPs will be used for the remainder of this thesis. In Section 4 I will discuss the option of performing a full double deflation study and framing double deflated Value Added PPPs by using data from British and American input-output tables.

Labor Productivity

productivity lead of the U.S. over the U.K., which is confirmed in table 3.5 below. Table 3.4 also shows that the U.K. and U.S. Textile, Leather and Clothing Trades were rather labor intensive, whereas the food and chemicals sectors enjoyed relatively high labor productivity levels. The difference in labor productivity can be explained by the high capital intensity of the latter branches.

Table 3.4 The Structure of the Manufacturing Sector, U.K. and U.S. (1935) Value Added† Employment Branch / Sector U.K. U.S. U.S./

U.K. U.K. U.S. U.S. U.K. / Textile Trades 13% 9% 187% 21% 15% 118%

Leather Trades 1% 1% 282% 1% 1% 176%

Clothing Trades 7% 7% 341% 10% 11% 163% Iron and Steel Trades 9% 10% 319% 10% 11% 174% Engineering, Shipbuilding… 21% 20% 397% 22% 20% 150% Non-ferrous Metals Trades 3% 3% 364% 2% 3% 197% Food, Drink and Tobacco Trades 17% 16% 257% 10% 11% 176% Chemical and Allied Trades 8% 9% 481% 4% 5% 205% Clay and Building Materials Trades 5% 3% 185% 5% 3% 95%

Timber Trades 3% 4% 988% 4% 7% 293%

Paper Trades 9% 12% 558% 8% 9% 185%

Miscellaneous Trades 4% 6% 458% 3% 4% 203% Manufacturing Sector 100% 100% 350% 100% 100% 161% †) The values in national currencies were converted with the (Fisher) adjusted single indicator Value Added PPPs from table 3.3.

The main results of this thesis are presented in table 3.5 below. As previously noted, the labor productivity can be represented as output per worker and value added per worker, both of which are rendered below in terms of American productivity relative to British productivity. In total four deflators were used; the official exchange rate and the Gross Output PPPs were used for deflating output per worker, whereas the double deflated Value Added PPPs (where available) and the adjusted single indicator Value Added PPPs were used for deflating value added per worker. These four methods of deflation all agree on the fact that the comparative level of American labor productivity for the entire manufacturing sector was at least twice that of Britain, as has been commonly assumed.53 _{On a disaggregate level however, there are widespread differences in labor}

productivity between the methodologies; on this intermediate level of aggregation, it matters a lot which type of deflation is applied.

Table 3.5 Labor Productivity per Worker, U.K. and U.S. (1935) Output per Worker

(% U.S. / U.K.)

Value Added per Worker (% U.S. / U.K.) Branch / Sector ($/£=4.94) GOPPP DD VAPPP ASI VAPPP

Textile Trades 150% 125% 137% 158%

Leather Trades 148% 127% 160%

Clothing Trades 223% 226% 209%

Iron and Steel Trades 191% 173% 180% 184% Engineering, Shipbuilding and Vehicle 235% 303% 265% Non-ferrous Metals Trades 156% 146% 185% Food, Drink and Tobacco Trades 222% 189% 145% Chemical and Allied Trades 235% 288% 235% Clay and Building Materials Trades 222% 202% 195%

Timber Trades 129% 291% 338%

Paper Trades 223% 285% 302%

Miscellaneous Trades 226% 213% 225%

Manufacturing Sector 211% 220% 218%

As was to be expected, the use of the official exchange rate smoothes out the differences in labor productivity across branches. This inevitably leads to the underestimation of the comparative levels of labor productivity in U.S. branches that perform well relative to the U.K. and other branches (and vice versa); this effect is particularly obvious for the Timber Trades. As previously noted, another problem with the exchange rate as deflator is that the internal price level and structure of a country may depart significantly from world market conditions; if so, the official exchange rate may differ substantially compared to actually observed PPPs, as was the case for the study of comparative labor productivity by Fremdling, de Jong and Timmer.54 _{The differences} between single deflation (GOPPP), adjusted single deflation (ASI VAPPP) and double deflation (DD VAPPP) are slightly less pronounced and can, for the most part, be explained by the fundamental differences in methodology. As illustrated in section 2, the single deflation will inevitably give too much weight to those of the industry’s products with relatively high inputs per unit of output (and vice versa), since these commodities are weighted by ex-factory output prices instead of value added figures. This effect is

particularly obvious for the Textile Trades, where the relatively large share of British inputs in the total value of products for this branch results in the understatement of American single deflated output per worker (the reversal is true for Food, Drink and Tobacco Trades). Furthermore, the single deflation and adjusted single deflation approaches implicitly assume that price movements of inputs are similar to that of outputs, while this is not necessarily the case. In theory, double deflation is thus the preferred methodology; but unfortunately, the results of the double deflation are incomplete. Regardless, the application of the different deflation techniques did not considerably change the rank order of comparative productivity levels among the twelve branches. Consequently, I have given preference to the adjusted single deflation methodology.

Even though in cross-country productivity comparisons the concept of output or value added per worker (as presented in table 3.5) is more commonly applied, the concept of output or value added per man-hour is the more relevant one; as explained in section 1. Table 3.6 presents the labor productivity per man-hour statistics. The branch specific employment data was multiplied by the data on annual hours worked from table 1.1; both the single deflated as well as the adjusted single deflated labor productivity per man-hour figures were computed. The last column of table 3.6 shows the difference between the comparative labor productivity in terms of workers and in terms of hours actually worked. As previously noted the difference is substantial, mainly due to the fact that the average American working week for the manufacturing sector in the interwar period was substantially shorter than the average British working week. The labor productivity figures for the total manufacturing sector for output and value added are raised by 26 percent, to 276 percent and 273 percent respectively. Likewise, the actual annual hours worked is higher in all the British branches. On this intermediate level of aggregation the rise in labor productivity ranges from anywhere between 15 percent (Timber Trades) up to 40 percent (Clothing Trades).

A final distinction regarding labor productivity which can be made is between the types of labor supplied. Both censuses provide data on the number of supervisory staff and operatives.55 _{The composition of the labor force is however roughly identical for the} two countries, both for the entire manufacturing sector and the individual branches.

Consequently the comparative labor productivity is left unaffected when it is computed on a per operative basis, instead of a per worker basis.

Table 3.6 Labor Productivity per Man-hour, U.K. and U.S. (1935) Labor Productivity per

Man-hour (% U.S. / U.K.)

Difference per worker (W) and per man-hour (H)

Branch / Sector Output

(GOPPP) Value Added (ASI VAPPP) (% W / M) Textile Trades 160% 204% 130% Leather Trades 153% 194% 121% Clothing Trades 312% 291% 140%

Iron and Steel Trades 216% 229% 125%

Engineering, Shipbuilding and Vehicle 379% 332% 125% Non-ferrous Metals Trades 180% 228% 123% Food, Drink and Tobacco Trades 223% 171% 118% Chemical and Allied Trades 351% 286% 122% Clay and Building Materials Trades 256% 247% 127%

Timber Trades 335% 388% 115%

Paper Trades 348% 368% 122%

Miscellaneous Trades 288% 303% 135%

Manufacturing Sector 276% 273% 126%

4.

Comparative Productivity in Perspective

Even though the publication of Rostas’ first comparison of real output and productivity in 1943 raised fierce discussions among British economists, it undeniably left its mark on all subsequent industry of origin studies.56 The various problems and methods of comparison, as described in detail by Rostas in his second, full scale, comparative productivity study, are to a large extend still applicable today.57 _{Still, as the previous} sections have shown, the industry of origin approach has evolved over time and the quantitative comparison employed by Rostas has become obsolete. Nonetheless, this thesis is part of only a select few industry of origin studies dealing with the pre-1950 period, and also the first attempt to reassess the comparative productivity in American and British industries for the interwar years. One might be tempted to assume that so few new benchmark studies have been carried out because, in practice, the improved industry of origin approach is unable to substantiate its theoretical superiority. This

Rostas, L. (1943), “Industrial Production, Productivity and Distribution in Britain, Germany and the United States”

57

section will address this question and present a direct comparison of the results by Rostas and the results presented in section 3. In addition, I will combine this study with the comparison of British and German manufacturing productivity by Fremdling, de Jong and Timmer to provide an estimate of the level of German and American comparative productivity.58 _{Lastly I will elaborate on the options of extending this study, which might} aid in gaining new insights into the determinants of growth and stagnation during the interwar period.

The comparison of the outcomes of the study by Rostas and my own findings are presented in Appendix D.1 on a per worker basis.59 _{This table features a selection of} industries that broadly match the products and product-groups on which Rostas based his estimate. In addition, the figures for several essential industries - which were lacking in the 1948 study - were added to present a more balanced picture of comparative productivity for the manufacturing sector. The value added per worker figures found upon my own estimates have been based primarily on industry specific Adjusted Single Indicator Value Added PPPs. In addition value added per worker figures deflated by branch PPPs have been added for all industries since it was not always possible to calculate separate industry PPPs, even though these figures generally provide less detailed information regarding the comparative productivity of the respective industry. The second set of columns of Appendix D.1 present the value added per worker figures for several base years as estimated by Rostas. Unfortunately however, the findings of the two studies are not directly comparable, as the comparison is slightly complicated due to the fact that different methodologies were used. Therefore, prior to discussing the discrepancies in Rostas’ study, I will clarify the steps taken to make the two studies analogous.

The first issue hampering the direct comparison is the fact that Rostas used various base years. Strictly speaking, the two benchmarks can only be compared when identical years are used. Existing productivity time series demonstrate that the relative level of U.S. productivity in the manufacturing sector is nearly 10 percent higher when the U.S. base year of 1939, instead of the year 1935, is used (when compared to the

Fremdling, R., de Jong, H. and Timmer, M. (2006), “British and German Manufacturing Productivity Compared”

59

British base year of 1935).60 _{Still, Rostas provided productivity figures for most industries} based on the years 1935 and 1937, for which the average movement in productivity levels was only very small. Secondly, the aggregation of Rostas’ comparative productivity figures presents a number of issues that need to be addressed in order to get a clear view of the differences between the results from Rostas and my own. The correct procedure for the aggregation of Rostas’ comparative productivity figures is discussed in detail in Appendix D.2.

In table 4.1 the results by Rostas and the outcome of my own study are presented as the level of comparative productivity for the 12 branches and the manufacturing sector as a whole. To overcome the above mentioned issues, I compared, as far as possible, the same years (mostly 1935) and used the implicit weighting method (as described in Appendix D.2) to aggregate Rostas’ comparative productivity figures. The value added per worker figures for Rostas are based on the uncompensated figures - for which the undervaluation by Rostas of American productivity was not taken into account - as these are the most representative figures for the quantitative study that solely rely on Rostas’ study. Value added per worker figures for the Leather, Non-ferrous Metals and Timber Trades could not be estimated, since no industries belonging to these branches were included in Rostas’ study. Overall, Rostas could only cover about half of the manufacturing sector; the relative value of production per branch for the industries that were covered is listed in table 4.1 as well.

The last column of table 4.1 illustrates that on an aggregate level the outcome of the industry of origin studies is quite resistant against different research strategies. The index of Rostas’ adjusted value added per worker figure for the manufacturing sector is 205, which is fairly close to my own estimate.61 _{On the branch level however, greater}

The level of British and American comparative productivity in the manufacturing sector, relative to the base years of 1935. See Broadberry, S. (1997), The Productivity Race, p 44

(1935/1935 = 100) U.S. base year 1935 1937 1939 U.K. base year 1935 100 105 110 1936 96 101 106 1937 96 100 105 1938 97 102 107 61

differences can be found. The value added per worker figures for Rostas are higher in the Food, Drink and Tobacco Trades, but substantially lower in the Clothing and Chemical and Allied Trades. These differences can be traced back to the industry level, were the estimates by Rostas deviate even further from my own; as shown in Appendix D.1.

Table 4.1 Value Added per Worker, Woltjer and Rostas (1935) Value Added per Worker (% U.S. / U.K.) Percentage of Industries Coveredζ Relative Comp-arison Branch / Sector Woltjer† _Rostas‡ _Rostas

(U.K.) Rostas (U.S.) Rostas / Woltjer Textile Trades 158% 151% 74% 82% 0.95 Leather Trades 160% Clothing Trades 209% 165% 25% 25% 0.79

Iron and Steel Trades 184% 180% 54% 64% 0.98

Engineering, Shipbuilding and Vehicle 265% 268% 88% 96% 1.01

Non-ferrous Metals Trades 185%

Food, Drink and Tobacco Trades 145% 175% 53% 46% 1.20

Chemical and Allied Trades 235% 198% 22% 14% 0.84

Clay and Building Materials Trades 195% 197% 67% 79% 1.01

Timber Trades 338%

Paper Trades 302% 252% 21% 22% 0.83

Miscellaneous Trades 225% 203% 54% 63% 0.90

Manufacturing Sector 218% 205% 53% 52% 0.94

†) The values of value added per worker for Woltjer are identical to the Adjusted Single Indicator Value Added deflated figures in table 3.5.

‡) The values of value added per worker for Rostas are identical to the uncompensated implicit PPP weighted figures in table D.2.1.

ζ) The coverage ratios presented in the second set of columns of this table represents the value of net output of the industries covered by Rostas in terms of percentages of the total value added of the branches for the U.K. and the U.S. As noted in section 2, these ratios cannot be compared to the coverage ratios of table 3.1.

The only industry covered by Rostas in the clothing branch was the Boots and Shoe Trade. The relatively low estimate of 141 percent for this industry can most likely be explained by the non-homogeneity of the end-products for this category. Rostas had to rely on a rather rudimentary conversion factor and in addition had to integrate the sizable American ‘boot and shoe cut stock and findings’ trades into the industry. The difference between the estimate by Rostas and my own findings became even more pronounced for the Clothing Trades due to the low coverage ratio of only 25 percent for this branch.

possible the same years (mostly 1935) in both countries. See Rostas, L. (1948), Comparative Productivity