Comparative Productivity in British and German Manufacturing Before World War II: Reconciling Direct Benchmark Estimates and Time Series Projections

Stephen Broadberry; Carsten Burhop

doi:10.1017/S0022050707000125

Comparative Productivity in British and German Manufacturing Before World War II: Reconciling Direct Benchmark Estimates and Time Series Projections

Published online by Cambridge University Press: 29 May 2007

Stephen Broadberry and

Carsten Burhop

Show author details

Stephen Broadberry: Affiliation:
Professor, Department of Economics, University of Warwick, Coventry CV4 7AL, United KingdomCo-ordinator of the Economic History Initiative at the Centre for Economic Policy Research, London. E-mail: S.N.Broadberry@warwick.ac.uk.
Carsten Burhop: Affiliation:
Heisenberg-Fellow, Max-Planck-Institute for Research on Collective Goods, Kurt-Schumacher-Strasse 10, 53113 Bonn, Germany. E-mail: burhop@coll.mpg.de.

Article contents

Abstract
A BENCHMARK FOR CIRCA 1907
TIME SERIES PROJECTION, 1871–1938
INCOME BASED ESTIMATES
CONCLUSION
Appendix 1: Underlying Data for 1907 Benchmark
Appendix 2: Data For Time Series Projections
Appendix 3: Comparative Productivity in Manufacturing
References

Rights & Permissions

Abstract

This article provides a new benchmark estimate of comparative Germany/U.K. labor productivity in manufacturing for circa 1907, and experiments with alternative German manufacturing production indices for time series projection from a circa 1935 benchmark. A consistent picture of broadly similar levels of manufacturing labor productivity in Britain and Germany throughout the period 1871–1938 is established. We also show that a substantial German productivity lead had already emerged in heavy industry by 1907, but was offset by a substantial British productivity lead in light industry. For the pre-1914 period, an additional check is provided using nominal income-based estimates.

Information

Type: ARTICLES
Information: The Journal of Economic History , Volume 67 , Issue 2 , June 2007, pp. 315 - 349
DOI: https://doi.org/10.1017/S0022050707000125 [Opens in a new window]
Copyright: © 2007 The Economic History Association

Controversy continues to surround the issue of Germany's level of labor productivity in manufacturing compared with that of Britain before World War II. A long tradition going back at least as far as Ernest Williams has suggested that Britain was already falling behind Germany in manufacturing in the late nineteenth century, and this view was bolstered by the views of Alexander Gerschenkron and David Landes, who saw Germany as modernizing before World War I on the basis of heavy industry.¹

Williams, Made in Germany; Gerschenkron, Economic Backwardness; and Landes, Unbound Prometheus.

Worries about competition from Germany resurfaced during the interwar period, which saw Germany reclaim its position in export markets. Furthermore, this view of Germany leapfrogging ahead of Britain also underpins Alfred D. Chandler Jr.'s characterization of German manufacturing as closer to the successful U.S. model of “managerial capitalism” than the supposedly failing British model of “personal capitalism.”²

Chandler, Scale and Scope.

However, this traditional view has proved difficult to square with work in the historical national accounts (HNA) framework. In a series of studies working at the level of the aggregate economy, Angus Maddison showed that GDP per hour worked and GDP per employee were substantially higher in Britain than in Germany before World War II, projecting with times series from a benchmark close to the present.³

Maddison, Phases, Dynamic Forces, Monitoring, and World Economy.

Furthermore, Rainer Fremdling pointed out that the sectoral data underpinning the Maddison results also show labor productivity in industry to have been substantially lower in Germany than in Britain before World War I.⁴

Fremdling, “Productivity Comparison.”

Fremdling's study used data on nominal income in the commodity-producing sectors of the two countries in particular years, converted at purchasing power parities (PPPs) derived from agricultural and industrial products. Strictly speaking, Fremdling's results apply to a wide definition of industry, which includes construction and the utilities as well as manufacturing, but manufacturing was by far the biggest industrial sector in both countries. Recognizing the disjuncture between this result and the traditional literature, Fremdling records his doubts about Germany having lower industrial productivity than Britain on the eve of World War I and speculates that it may reflect an under-estimation of capital income in Germany by Walther Hoffmann.⁵

Ibid., pp. 39–42; and Hoffmann, Das Wachstum.

By projecting with the time series of Charles Feinstein and Hoffmann from a benchmark for circa 1935 produced by Broadberry and Fremdling, Broadberry then showed that labor productivity in manufacturing was slightly higher in Germany than in Britain in the pre–World War I period.⁶

Feinstein, National Income; Hoffmann, Das Wachstum; Broadberry and Fremdling, “Comparative Productivity”; and Broadberry, “Manufacturing.”

This work, which is perhaps best seen as taking an intermediate position between the traditional view of a large German superiority and the revisionist view of higher British productivity, has in turn been called into question by challenges to the Hoffmann index of industrial production and to the original Broadberry and Fremdling benchmark for circa 1935.⁷

Hoffmann, Das Wachstum; and Broadberry and Fremdling, “Comparative Productivity.”

Criticism of the Hoffmann industrial production index has come from Albrecht Ritschl and Burhop and Guntram Wolff, while doubt has been cast on the Broadberry/Fremdling benchmark by Fremdling and Reiner Stäglin's discovery of deliberate falsification of the records by officials to conceal the production of munitions by the Nazi regime.⁸

Hoffmann, Das Wachstum; Ritschl, “Spurious Growth”; Burhop and Wolff, “Compromise Estimate”; Broadberry and Fremdling, “Comparative Productivity”; and Fremdling and Stäglin, “Die Industrieerhebung.”

It would therefore be very helpful to have a benchmark estimate of the comparative Germany/U.K. labor productivity level in manufacturing for the pre–World War I period that is entirely independent of the HNA time series data. The first objective of this article is to provide such an estimate, based on the British industrial census for 1907 and official German data for a number of industries in 1907/08. As well as establishing more firmly the aggregate picture of broadly equal labor productivity in the pre–World War I period, this new benchmark provides for the first time disaggregated estimates of comparative labor productivity in different branches of industry for the pre–World War I period. The sectoral picture suggests that a substantial German labor productivity lead in heavy industry (chemicals, metals) was offset by a substantial British labor productivity lead in light industry (textiles and food, drink and tobacco). In addition, Britain had a labor productivity advantage in mining, an important nonmanufacturing industry.

A second objective is to experiment with alternative indices of German manufacturing output, to check for the consistency between time series projections and direct benchmark estimates. The two benchmarks for circa 1907 and circa 1935 are broadly consistent with the time series projections based on the Hoffmann index for Germany. A new series of Burhop and Wolff for the pre–World War I era can be incorporated in a way that is consistent with the two benchmarks, but suggests a somewhat larger German labor productivity lead between the 1870s and the 1890s than suggested by the Hoffmann series.⁹

Burhop and Wolff, “Compromise Estimate”; and Hoffmann, Das Wachstum.

A problem of inconsistency arises only in the case of Ritschl's new index of German industrial production for the interwar period, which incorporates a very major change to the output of the metal-working sector for the years 1925–1938.¹⁰

Ritschl, “Spurious Growth.”

A third objective is to rework Fremdling's income-based estimates of comparative Germany/U.K. industrial labor productivity, based on nominal net income per employee in British and German industry, converted at PPPs for agricultural and industrial products and converted at market rates.¹¹

Fremdling, “Productivity Comparison.”

We incorporate Burhop and Wolff's allowance for Hoffmann's underestimation of capital income in the pre-1914 period, restrict our attention to manufacturing and apply PPPs based only on manufactured products to these new data.¹²

Burhop and Wolff, “Compromise Estimate”; and Hoffmann, Das Wachstum.

We find broadly equal levels of manufacturing labor productivity in the two countries throughout the period 1875–1913, as suggested by the output-based estimates. Overall, then, we are able to reconcile most of the existing time series and benchmark estimates of comparative Germany/U.K. labor productivity in manufacturing before World War II.

A BENCHMARK FOR CIRCA 1907

Data Sources

Although there was no full industrial census for Germany before World War I, much high quality information was collected. Some of this information was published in the form of a complete employment census for 1907, while for a number of industries, further information on the quantity and value of output and material inputs appeared in various sources. The Imperial Office of the Interior (Reichsamt des Inneren) started industry surveys regarding output (value and volume), input (value and volume), employment, and wages for some industries already in 1897. After the turn of the century, such surveys were conducted several times and for some industries on an annual basis. The findings were finally published by the Imperial Statistical Office.¹³

Kaiserliches Statistisches Amt, “Ergebnisse”; and “Die Ergebnisse.”

Furthermore, production taxes for beer, sugar, and tobacco were levied in Germany, making collection of output data necessary. These output data were published annually by the Imperial Statistical Office and could be matched with employment data from the 1907 employment census.

In contrast to the British census of production, Germany's statistical office did not strictly apply the concept of value added. The modern national accounting standards use the term “gross value added” as the value of output less the value of intermediate goods consumed in the production process. The British census of 1907 used the term net output instead of gross value added. The German term Nettoproduktionswert basically corresponds to the modern concept of gross value added, but because the Imperial Office of the Interior did not collect comprehensive data on intermediate goods consumed, this measure is likely to be biased.¹⁴

Tooze, Statistics, pp. 56–57.

Nevertheless, this industrial census is a very useful source, covering nearly all firms in the surveyed industries. All firms listed by trade associations (Berufsgenossenschaften) were asked to submit anonymously information about the type of business, the number of establishments, the number of employees, the wage bill, the volume and value of intermediate goods consumed, and the volume and value of production and of goods sold.¹⁵

All employees of the firms listed by the Berufsgenossenschaften were counted for the industrial census, not only the insured employees.

Inputs were valued inclusive of transportation costs to the factory, whereas outputs were valued at factory gate prices (without transport costs). However, for all industries, no information was collected on several items of intermediate consumption. In particular, the costs of maintenance of machinery and buildings, heating and power generation, packaging and overheads are not included in the German data throughout. Therefore, comparing German Nettoproduktionswerte with British net output might overstate German labor productivity. The German census data are also less comprehensive in terms of industries covered than the British data. Whereas the British data cover total manufacturing, the German data focus on selected industries (see Appendix 1 for details). Furthermore, we can only use German and British data if we find the same information for the same industries in both censuses. The German data included in our 1907 benchmark cover more than 1.7 million employees, slightly more than 29 percent of Germany's total manufacturing employment in that year. The British data used for our benchmark include circa 1.56 million employees, about 27 percent of total manufacturing employment.

In our judgment, these sources are sufficient to allow us to make reasonably reliable estimates of physical output per worker in Germany in 1907/08, along the lines of those produced for 1935/36 by Laszlo Rostas and reworked by Broadberry and Fremdling.¹⁶

Rostas, Comparative Productivity; and Broadberry and Fremdling, “Comparative Productivity.”

These estimates can then be compared with estimates of physical output per worker in the United Kingdom, derived from the First Census of Production for 1907, which provide a comprehensive overview of British industrial performance.¹⁷

Board of Trade, Final Report.

Weighting Scheme

Although it is possible to estimate net output for some German industries, we do not have sufficient information to do this reliably for the whole manufacturing sector. Hence for Germany, we are forced to rely on employment weights, shown here in Table 1. For the United Kingdom, we have both employment and net output weights, which are also shown in Table 1. We shall see, using U.K. data, that the choice of net output or employment weights makes little difference, so that the lack of German net output weights is unlikely to be too serious a problem.

The structure of industry was quite different in the two countries, with German manufacturing more oriented towards heavy industry, particularly metal production, and British manufacturing more oriented towards light industry, particularly textiles and clothing. However, note that although Britain had quite a large share of net output in food, drink, and tobacco, this accounted for a smaller share of employment, because of high net output per worker. Britain also had a substantially larger share of the labor force in mining than did Germany.

The weighting scheme in Table 1 is derived in the following way. First, employment (and also net output for Britain) was allocated across major sectors (indicated in upper case letters) from census sources.¹⁸

Ibid.; and Kaiserliches Statistisches Amt, “Berufs- und Betriebszählung.”

Second, within major sectors, shares were allocated in proportion to the sectors for which we were able to make matches. For example, focusing on German employment weights, the major sector Chemicals & Allied Industries accounted for 188,073 employees out of a total of 5,758,891 in manufacturing, or 3.27 percent. Within this major sector, we were able to match data on two industries, general chemicals and coke. The general chemicals firms surveyed employed 4,578 persons, whereas in coke production 20,504 persons were employed in the surveyed firms. Total employment in the surveyed firms was thus 25,082, resulting in an employment weight of 82 percent for coke and 18 percent for general chemicals. Part A of Table 1 provides the weights within manufacturing, and part B provides the weights for industry including mining.

Results

The individual industry results reported in Table 2 are based mainly on comparisons of physical output per worker in the two countries, with a value above 100 indicating a German productivity advantage and a value below 100 indicating a British productivity advantage. The single figure for each sector is arrived at by taking the geometric mean of results using German and British employment weights. The figure for total manufacturing is 105.0, which is the geometric mean of 105.5 using German employment weights and 104.6 using British employment weights. Note that using British net output weights makes no real difference, yielding a comparative productivity ratio of 104.5. For total industry the geometric mean of 101.8 is derived from 102.9 using German employment weights and 100.8 using British employment weights, whereas the use of British net output weights yields a figure of 100.0.

One might argue that these comparative labor productivity measures are biased because the British data are for 1907 throughout, whereas German data are for 1907 (motor vehicles, cotton, silk, brewing, tobacco, sugar), 1908 (chemicals, coke, iron & steel, nonferrous metals, mining), and 1910 (leather, cement) respectively. Taking employment weights, 28 percent of the German data are for 1907, 68 percent for 1908, and 4 percent for 1910. Between 1907 and 1908 (1910) labor productivity increased faster in Germany than in Britain, thus our benchmark might overstate German productivity leadership. If projections are based on Hoffmann's classic figures, comparative labor productivity improved in favor of Germany between 1907 and 1908 (1910) by 8.2 percent (14.7 percent).¹⁹

Hoffmann, Das Wachstum.

Weighting this productivity increase by the employment shares for the different years yields a possible overestimation of Germany's productivity lead of 6.2 percent. Reducing our comparative labor productivity estimate by this leads to comparative labor productivity in manufacturing for 1907 of 98.9. Thus, even accounting for different years of census data did not change the basic results of a roughly equal labor productivity in manufacturing before World War I.

Just as important as the results for total manufacturing and total industry is the distribution of comparative productivity ratios across major sectors. The key finding here is that Germany had higher labor productivity in heavy manufacturing, particularly in ferrous and nonferrous metals, but also in chemicals, while Britain had higher productivity in light manufacturing, particularly textiles and clothing and food, drink, and tobacco. Britain also had a productivity advantage at this time in mining.

TIME SERIES PROJECTION, 1871–1938

Methodological Considerations

Benchmarks are available only for a small number of years. To build up a full picture of the evolution of comparative productivity over time, it is necessary to use time series to project forwards and backwards. However, this inevitably raises potential index number problems because of variation in the weighting schemes used in the output time series for the two countries, and differences in the proportion of output covered between the two benchmarks for each country. This has led to some controversy since the work of the Income Comparisons Project (ICP) for the post-1950 period, with H. Krijnse Locker and M. D. Faerber providing an early statement of the problem.²⁰

Krijnse Locker and Faerber, “Space and Time.”

Having initially provided a set of benchmarks that conflicted with time series projections, Alan Heston, Robert Summers and Bettina Aten have acknowledged that this is unsatisfactory in the following terms:

If per capita G.D.P. in India relative to Korea is 40% in a 1970 benchmark and 35% in a 1975 benchmark, then one inference is that the growth rate per capita in Korea must have been about 5% more than in India between 1970 and 1975. Often such an implied result is not consistent with the national growth rates in the two countries, posing a problem to users. ²¹

Heston, Summers and Aten, “Price Structures,” p. 2.

One solution offered by Summers and Heston to the problem of inconsistency between benchmarks and time series projections was to change the national growth rates as well as the benchmarks, in a process known as “consistentization.”²²

Summers and Heston, “New Set.”

However, Heston, Summers, and Aten have now acknowledged that it is difficult to persuade people of the need to change widely accepted growth rates to fit in with benchmark estimates:

It was hard to sell the idea of modifying country growth rates to countries, international organizations and to men or women of affairs. Therefore in our recent uses of consistentization we have not modified country growth rates, but only the different benchmark estimates.²³

Heston, Summers and Aten, “Price Structures,” p. 6.

The methodology adopted by Broadberry has been to recognize the need to tell a coherent story, but at the same time, to acknowledge the inevitability of some disagreement between constant price time series projections and current price benchmark estimates.²⁴

Broadberry, “Manufacturing,” “How Did the United States?” and “Relative Per Capita Income”; Krijnse Locker and Faerber, “Space and Time”; Nuxoll, “Differences”; and Dowrick and Akmal, “Contradictory Trends.”

If a projection is made from a single benchmark, other benchmarks can be used to provide a cross-check on the projection, with Broadberry suggesting that a reasonable aim should be to keep disagreements between direct benchmarks and time series projections to within 10 percent.²⁵

Broadberry, “Manufacturing.”

Applying this to the case of the United States and the United Kingdom during the period 1870–1990, Broadberry was critical of Marianne Ward and John Devereux's claim, on the basis of a new benchmark, that U.S. per capita income in 1872 was 18 percent higher than in the United Kingdom, compared with Maddison's result, based on time series projection, that the United States was at 78 percent of the U.K. level, a discrepancy of 40 percentage points, or more than 50 percent of the Maddison estimate.²⁶

Broadberry, “Relative Per Capita Income”; Ward and Devereux, “Measuring British Decline”; and Maddison, Monitoring.

Broadberry's case was strengthened by the provision of additional economy-wide benchmarks for 1910, 1937, and 1950, plus reconciliation of benchmarks and time series projections at the level of ten subsectors, at the previously mentioned dates plus additional benchmarks for the majority of sectors at 1924, 1930, 1968, and 1990.²⁷

Broadberry, “How Did the United States?”

In most cases, benchmarks fell within 10 percent of the time series projections. Given the widespread acceptance of the time series produced by Feinstein and by John Kendrick, together with the coherence between these times series and the benchmarks, there must be a strong presumption in favor of the Maddison estimates.²⁸

Feinstein, National Income; Kendrick, Productivity Trends; and Maddison, Monitoring.

We now consider the reconciliation of time series projections and direct benchmarks in the case of German and British manufacturing industry.

Choice of Benchmark

The first issue that must be decided is the choice of benchmark on which to base the time series projections. As we have constructed a new benchmark for 1907, this obviously provides one option for the base. However, we could alternatively use either the Broadberry and Fremdling benchmark for 1935 or the Fremdling, Herman de Jong and Marcel Timmer benchmark for 1935/36.²⁹

Broadberry and Fremdling, “Comparative Productivity”; and Fremdling, de Jong and Timmer, “British and German Manufacturing.”

The latter draws on the findings of Fremdling and Stäglin, that the published 1936 census was distorted for military-strategic purposes.³⁰

Fremdling and Stäglin, “Die Industrieerhebung” and “Eine Input-Output-Tabelle”; and Reichsamt für wehrwirtschaftliche Planung, Die deutsche Industrie.

Fremdling and Stäglin recalculate the input-output table for 1936, using archival records. This leads to different estimates of net output and employment in individual branches of industry, although the effects on industry as a whole are quite small. We have opted for the 1935 benchmark of Broadberry and Fremdling for the following reasons.³¹

Broadberry and Fremdling, “Comparative Productivity.”

First, there is much less information available for 1907 than for 1935/36, so a later benchmark is likely to be more reliable. Second, the Broadberry and Fremdling benchmark is on the same conceptual basis as the 1907 benchmark presented in this article, working in terms of quantity indicators rather than nominal net output compared using industry-specific price ratios.

A comparison between the Broadberry and Fremdling and the Fremdling, de Jong, and Timmer benchmarks for circa 1935 is provided in parts A and B of Table 3. Reassuringly, there is a strong measure of agreement amongst the different benchmarks, despite the conceptual differences and the distortions introduced by military planners. First, note the narrow range for the total manufacturing comparative productivity ratio in the different studies, with Fremdling, de Jong, and Timmer arriving at a ratio of 105 or 107, compared with the Broadberry and Fremdling ratio of 102.³²

Fremdling, de Jong and Timmer, “British and German Manufacturing”; and Broadberry and Fremdling, “Comparative Productivity.”

These differences are well within the 10 percent bounds suggested by Broadberry for this type of work.³³

Broadberry, “Manufacturing.”

Second, note that if we compare the pattern of comparative productivity performance across sectors in 1935 from Table 3 with the pattern for 1907 in Table 2, there is a high degree of persistence. Germany had a labor productivity lead in heavy industries such as chemicals and metals in both years, while Britain had a persistent labor productivity lead in light industries such as textiles and clothing, and food, drink, and tobacco. Third, note that this pattern of stronger German productivity performance in heavy industry and better British productivity performance in light industry is as clear in the Fremdling, de Jong, and Timmer benchmark as it is in the Broadberry and Fremdling benchmark.³⁴

Fremdling, de Jong, and Timmer, “British and German Manufacturing”; and Broadberry and Fremdling, “Comparative Productivity.”

Revisions to the German Time Series Data

The British historical national accounts for this period are uncontroversial and we employ Feinstein's index of production for manufacturing, as in Broadberry's previous work.³⁵

Feinstein, National Income; and Broadberry, Productivity Race.

Employment data for Britain are also derived from Feinstein, but with some interpolation for the pre-1920 period, using the cyclical pattern in Feinstein's total employment series.³⁶

Feinstein, National Income.

Germany's historical national accounts are rather more controversial, however. A widely used industrial production index was calculated by Hoffmann.³⁷

Hoffmann, Das Wachstum.

This index is based mainly on physical output series for 12 industries, but includes construction and the utilities as well as manufacturing. To be comparable with Feinstein's index for Britain, Hoffmann's index of total industrial production thus needs to be recalculated to cover only manufacturing, as in Broadberry's earlier work.³⁸

Feinstein, National Income; Hoffmann, Das Wachstum; and Broadberry, “Manufacturing.”

Hoffmann's industrial production index was calculated from time series of physical output for 11 industries and a labor income series for one industry (metal processing).³⁹

Hoffmann, Das Wachstum, pp. 344–95.

These 12 series were aggregated into a comprehensive industrial production index using net output weights. However, because reliable estimates of industrial net output for Germany are available only for 1936, Hoffmann came up with a procedure to create proxy net output weights for earlier years.⁴⁰

Ibid., pp. 389–95.

For each industry, this involved taking net output per employee in 1936 and multiplying it by employment in the same industry in 1882, 1907, and 1933. The product of 1936 labor productivity and 1882 (1907, 1933) employment then gives the weight of each industry for the years 1871–1895 (1895–1913, 1925–1938). Because we are interested in manufacturing, we will work with the ten industries excluding construction and the utilities.

We will report first the results of time series projection using the Hoffmann industrial production index recalculated on to a manufacturing-only basis by Broadberry.⁴¹

Hoffmann, Das Wachstum; and Broadberry, Productivity Race.

We will then construct two alternative projections to take account of recent criticisms of the Hoffmann index.⁴²

Hoffmann, Das Wachstum.

One of the most important output series used by Hoffmann has been criticized recently. Ritschl presents a new time series for metal processing, the most important branch of German industry, for the years 1913 and 1925–1938, which suggests that the level of output in the interwar period was much lower relative to the prewar period than in the Hoffmann series.⁴³

Ritschl, “Spurious Growth”; and Hoffmann, Das Wachstum.

Ritschl believes that the Hoffmann index fails to reflect a substantial shift in the distribution of income in favor of labor, which occurred across World War I. Because the production index for metal processing is calculated by dividing labor income by a constant labor income share, the hypothesis of a nonconstant labor income share in metal processing is quite important.⁴⁴

It is not clear how Ritschl treats the territorial losses of Germany after World War I. Territories covering about 35 percent of German 1913 steel output were lost after the War. Ritschl fixes output of metal processing industry to 1913 = 100, but it is not clear if he uses 1913 territory or 1921 territory to do this.

As we shall see later, however, Burhop and Wolff's income-based estimates suggest the alternative possibility that Hoffmann underestimated the pre–World War I level of capital income.⁴⁵

Burhop and Wolff, “Compromise Estimate.”

A further criticism of the time series used by Hoffmann is made by Burhop and Wolff, who derive a new index for industrial output for the years 1851 to 1913 based on re-estimated industrial income data.⁴⁶

Ibid.

Two indices of manufacturing output in Germany between 1871 and 1938 are plotted in Table 1. First, there is the original Hoffmann industrial production index adjusted on to a manufacturing-only basis by Broadberry.⁴⁷

Hoffmann, Das Wachstum; and Broadberry, Productivity Race.

Second, there is an index including Ritschl's correction for the interwar period, as well as Burhop and Wolff's correction for the prewar period.⁴⁸

Ritschl, “Spurious Growth”; and Burhop and Wolff, “Compromise Estimate.”

For the prewar period, the Burhop and Wolff index is slightly above the level displayed by Hoffmann's original data, especially for the 1870s to 1890s.⁴⁹

Ibid.; and Hoffmann, Das Wachstum.

Large differences can be noted between Hoffmann's original data and the manufacturing output index based on Ritschl's modification. Generally, Ritschl's index is well below the level displayed by Hoffmann's original data.⁵⁰

Ritschl, “Spurious Growth”; and Hoffmann, Das Wachstum.

FIGURE 1 INDICES OF MANUFACTURING PRODUCTION IN GERMANY

Results

Figure 2 displays the time series projection from the 1935 Broadberry/Fremdling benchmark estimate for comparative Germany/U.K. labor productivity in manufacturing.⁵¹

Broadberry and Fremdling, “Comparative Productivity.”

The first projection, labeled Broadberry, uses the Hoffmann index of German manufacturing output.⁵²

Broadberry, Productivity Race; and Hoffmann, Das Wachstum.

Note first that the projected values for 1907 and 1908 are 107.8 and 116.7 respectively, reassuringly close to our circa 1907 benchmark value of 105.0. If, alternatively, the 1907 benchmark is used as the basis of the projections, this gives projected values for 1935 and 1936 of 100.5 and 99.3, respectively. This is again reassuringly close to all the circa 1935 benchmarks ranging from 102 to 107.

FIGURE 2 COMPARATIVE GERMANY/UNITED KINGDOM LABOR PRODUCTIVITY IN MANUFACTURING, 1871–1938 (United Kingdom = 100)

The second projection in Table 2, labeled “Burhop-Wolff-Hoffmann,” uses Burhop and Wolff's data for the prewar period and Hoffmann's data for the interwar period.⁵³

Burhop and Wolff, “Compromise Estimate”; and Hoffmann, Das Wachstum.

This results in a similar picture to the Broadberry estimate for the period from the turn of the century to 1938. However, between the 1870s and the 1890s, this series suggests that German labor productivity was already ahead of the British level, although the difference is not that great. Reconciliation of this time series finding with cross-sectional evidence would require a new benchmark for the 1870s. With the data available at the moment, that does not seem feasible, so this must remain an open issue.

Finally, we calculated an index using Burhop and Wolff's data for the prewar period and Ritschl's data for the interwar period.⁵⁴

Burhop and Wolff, “Compromise Estimate”; and Ritschl, “Spurious Growth.”

Using this time series to extrapolate from the 1935 benchmark backwards to the 1907 benchmark in Table 2 leads to a German labor productivity lead over Britain of more than 50 percent in 1907–1908, about 40 percentage points above the 1907 benchmark estimate, and a rather unlikely finding. It thus seems that Ritschl's modifications to Hoffmann's index of industrial production need further adjustment.⁵⁵

Ibid.; and Hoffmann, Das Wachstum.

INCOME BASED ESTIMATES

We can compare these findings based on output data with alternative estimates of comparative Germany/U.K. labor productivity in manufacturing based on income data. This is important because Fremdling found labor productivity in German industry to vary between around half to two-thirds of the British level between the 1870s and World War I, which appears to conflict strongly with our findings of broadly equal labor productivity.⁵⁶

Fremdling, “Productivity Comparison.”

The most important adjustment follows from the use of Burhop and Wolff's allowance for Hoffmann's under-estimation of capital income in the pre-1914 period.⁵⁷

Burhop and Wolff, “Compromise Estimate”; and Hoffmann, Das Wachstum.

This was noted by Fremdling, who argued that Hoffmann assumed far too low a rate of return on capital, but Fremdling did not produce an alternative estimate.⁵⁸

Fremdling, “German National Accounts”; and Hoffmann, Das Wachstum.

The second adjustment arises from the fact that our estimates are restricted to manufacturing, although this makes only a small difference. The third adjustment arises from the fact that Fremdling used PPPs for commodity output, including agricultural as well as industrial commodities. We calculate the PPPs solely on the basis of manufactured products, which makes a big difference because protection in Germany ensured that the price of agricultural products did not fall in line with agricultural prices in Britain, where markets remained open to cheap grain and meat from the New World.⁵⁹

In addition, some heavy industrial products were also protected by import duties.

But even here, it is worth noting that the particular industrial products for which Fremdling was able to obtain data are biased towards finding a higher price level in Germany.⁶⁰

Fremdling, “Productivity Comparison.”

Fremdling provides series for the prices of pig iron, cotton yarn, and flour to represent the metals, textiles, and food sectors, respectively.⁶¹

Ibid., p. 30.

However, Germany imported pig iron from Britain to make steel that was competitive on world markets, and also imported cotton yarn from Britain to produce cotton cloth that was competitive on world markets. Also, the principal ingredient of flour was grain, the price of which was kept artificially high in Germany through agricultural protection.

In Table 4, we derive British net output per employee from information in the historical national accounts combined with the 1907 Census of Production. This involves using Feinstein's time series of NNP from the income side to project the value of net output in manufacturing from the 1907 Census, and the time series of employment in manufacturing from Broadberry to project the 1911 manufacturing labor figure from Feinstein.⁶²

Feinstein, National Income, table 1; Broadberry, Productivity Race, p. 43; and Feinstein, National Income, table 60.

The assumption of a constant share of manufacturing in current price national income is consistent with the figures of Phyllis Deane and William Cole.⁶³

Deane and Cole, British Economic Growth, p. 166.

For Germany, the 1907 share of manufacturing output in total industrial output is calculated using the weights given by Hoffmann.⁶⁴

Hoffmann, Das Wachstum, p. 390.

This is then multiplied with the 1913 total industrial output from Burhop and Wolff to yield an estimate of manufacturing net output that is substantially higher than that implicit in Hoffman's income side data.⁶⁵

Burhop and Wolff, “Compromise Estimate,” p. 642; and Hoffmann, Das Wachstum.

This yields a 1913 manufacturing labor productivity of M 2,413. The income-based 1913 benchmark for Germany is projected backwards using the nominal NNP data together with the assumption of a constant share of manufacturing in Germany. Combined with data on manufacturing employment, this yields an income-based estimate of manufacturing labor productivity. These data series for the two countries can be compared using market exchange rates or PPPs, as in Table 4.

The whole period is divided into four sub-periods, 1875–1884, 1885–1894, 1895–1904, and 1905–1913. The average PPPs for manufactured products move more closely in line with the exchange rate than Fremdling's PPPs for a wider range of commodities, which were heavily influenced by protected agricultural products.⁶⁶

Fremdling, “Productivity Comparison,” p. 33.

Using these PPPs, German manufacturing output per worker was slightly below British labor productivity throughout the period. However, using the market exchange rate instead of PPPs, labor productivity was slightly higher in Germany. The PPPs most likely understate true German productivity, for the reasons outlined previously, while the market exchange rate probably overstates German productivity. Taken together, the two measures suggest broadly equal labor productivity in German and British manufacturing before World War I. Correcting the Hoffmann income data for underestimation of capital income therefore removes a large part of the discrepancy between the income-based and output-based measures.⁶⁷

Engel, “Die erwerbstätingen juristischen Personen,” p. 468, gives some support to our hypothesis of an equal manufacturing labor productivity in Britain and Germany around 1880. According to our extrapolation, nominal manufacturing labor productivity was 1,428 M between 1875 and 1884. Engel estimates a gross manufacturing output per worker of 3,600 M for the early 1870s. A reasonable assumption is that net output per worker (labor productivity) is about 40 percent of gross output, or 1,440 Marks, a figure quite close to our extrapolation.

CONCLUSION

Labor productivity in manufacturing was about the same in Germany and Britain from the late nineteenth century to World War II, supporting the earlier findings of Broadberry.⁶⁸

Broadberry, Productivity Race.

Benchmark estimates for 1907 and 1935 show Germany ahead in heavy industry (chemicals, metals), whereas Britain had higher productivity in the light industries, especially in textiles, and food, drink, and tobacco. Our new 1907 benchmark estimate thus offers qualified support to the traditional view of Williams, Gerschenkron, and Landes, who saw Germany as modernizing before World War I on the basis of heavy industry.⁶⁹

Williams, Made in Germany; Gerschenkron, Economic Backwardness; and Landes, Unbound Prometheus.

However, it also encompasses the newer historical national accounting viewpoint, which notes that the high-productivity modernized parts of the German economy co-existed with low-productivity traditional parts, so that whole-economy productivity was substantially higher in Britain than in Germany.

Starting from Broadberry and Fremdling's 1935 benchmark estimate of comparative labor productivity in manufacturing, we project backwards to 1871 and forwards to 1938 using new time series data of Germany's manufacturing output.⁷⁰

Broadberry and Fremdling, “Comparative Productivity.”

The two benchmark estimates are broadly consistent with the Hoffmann and Burhop and Wolff indices of industrial output.⁷¹

Hoffmann, Das Wachstum, p. 390; and Burhop and Wolff, “Compromise Estimate.”

However, Burhop and Wolff's industrial output data suggest a German labor productivity lead in manufacturing already during the 1870s. Future research should check this hypothesis with a new benchmark estimate. So far, however, the available data are not sufficient to calculate such an early benchmark estimate.

We support our finding for the pre–World War I period using new income data from Burhop and Wolff.⁷²

Ibid.

Using these data, we rework Fremdling's income-based estimate of comparative labor productivity in manufacturing.⁷³

Fremdling, “Productivity Comparison.”

We find broadly equal levels of manufacturing labor productivity in the two countries throughout the period 1875–1913, as suggested by our output-based estimates. Overall, we are able to reconcile most of the existing time series and benchmark estimates of comparative Germany/U.K. labor productivity in manufacturing before World War II.

Appendix 1: Underlying Data for 1907 Benchmark

General Note: individual figures in the tables have been rounded, whereas the calculations use unrounded figures. This results in a few very small apparent discrepancies.

Appendix 2: Data For Time Series Projections

Appendix 3: Comparative Productivity in Manufacturing

We would like to thank participants in the Berlin Colloquium, the Second Conference on German Cliometrics, and the referees for helpful comments. Remaining errors are our responsibility. The financial support of the European Commission Marie Curie Actions, (Contract no. MRTN-CT-2004-512439) as well as the support of Juliane Schrader is gratefully acknowledged.

References

Board of Trade. Final Report of the First Census of Production of the United Kingdom (1907). London: HMSO, 1912.

Broadberry Stephen N. “Manufacturing and the Convergence Hypothesis: What the Long Run Data Show.” This JOURNAL 53, no. 4 (1993): 772-95.Google Scholar

Broadberry Stephen N.. The Productivity Race: British Manufacturing in International Perspec-tive, 1870-1990. Cambridge: Cambridge University Press, 1997.CrossRef

Broadberry Stephen N.. “How Did the United States and Germany Overtake Britain? A Sectoral Analysis of Comparative Productivity Levels, 1870-1990.” This JOURNAL 58, no. 2 (1998): 375-407.Google Scholar

Broadberry Stephen N.. “Relative Per Capita Income Levels in the United Kingdom and the United States since 1870: Reconciling Time-Series Projections and Direct Benchmark Estimates.” This JOURNAL 63, no. 3 (2003): 852-63.Google Scholar

Broadberry Stephen N., and Rainer Fremdling. “Comparative Productivity in British and German Industry, 1907-37.” Oxford Bulletin of Economics and Statis-tics 52, no. 4 (1990): 403-21.Google Scholar

Burhop Carsten, and Guntram B. Wolff. “A Compromise Estimate of German Net National Product, 1851-1913, and its Implications for Growth and Business Cy-cles.” This JOURNAL 65, no. 3 (2005): 613-57.Google Scholar

Chandler Alfred D., Jr. Scale and Scope: The Dynamics of Industrial Capital-ism. Cambridge, MA: Harvard University Press, 1990.

Deane Phyllis, and William A. Cole. British Economic Growth, 1688-1959. (second edition). Cambridge: Cambridge University Press, 1969.

Dowrick Steve, and Muhammad Akmal. “Contradictory Trends in Global Income Inequality: A Tale of Two Biases.” Review of Income and Wealth 51, no. 2 (2005): 201-29.Google Scholar

Engel Ernst. “Die erwerbstätigen juristischen Personen im preussischen Staate, insbesondere die Aktiengesellschaften.” Zeitschrift des Königlich Preussischen Statistischen Bureaus 15 (1874): 449-536.Google Scholar

Feinstein Charles H. National Income, Expenditure and Output of the United Kingdom, 1855-1965. Cambridge: Cambridge University Press, 1972.

Fremdling Rainer. “German National Accounts for the 19th and Early 20th Century: A Critical Assessment.” Vierteljahrschrift für Sozial- und Wirtschafts-geschichte 75, no. 3 (1988): 339-57.Google Scholar

Fremdling Rainer. “Productivity Comparison between Great Britain and Germany, 1855-1913.” Scandinavian Economic History Review 39, no. 1 (1991): 28-42.Google Scholar

Fremdling Rainer, and Reiner Stäglin. “Die Industrieerhebung von 1936: Ein Input-Output Ansatz zur Rekonstruktion der volkswirtschaftliche Gesamtrechnungen für Deutschland im 19. und 20. Jahrhundert - ein Arbeitsbericht.” Vierteljahrschrift für Sozial- und Wirtschaftsgeschichte 90, no. 4 (2003): 416-28.Google Scholar

Fremdling Rainer, and Reiner Stäglin. “Eine Input-Output-Tabelle für 1936 als Grundlage einer neuen volkswirtschaftlichen Gesamtrechnung für Deutschland.” Conference paper, Annual conference of the Verein für Socialpolitik, 2004.

Fremdling Rainer, Herman de Jong, and Marcel Timmer. “British and German Manu-facturing Compared: A New Benchmark for 1935/36 Based on Double Deflated Value Added.” unpublished paper, University of Groningen, 2006.

Gerschenkron Alexander. Economic Backwardness in Historical Perspec-tive. Cambridge, MA: Harvard University Press, 1962.

Heston Alan, Robert Summers, and Bettina Aten. “Price Structures, the Quality Fac-tor, and Chaining.” Organisation for Economic Co-operation and Development, 2001. http://www.oecd.org/dataoecd/23/40/2425050.pdf.

Hoffmann Walther G. Das Wachstum der deutschen Wirtschaft seit der Mitte des 19. Jahrhunderts. Berlin: Springer, 1965.

Kaiserliches Statistisches Amt. Statistisches Jahrbuch für das Deutsche Reich 30 (1909).

Kaiserliches Statistisches Amt. “Berufs- und Betriebszählung vom 12. Juni 1907. Gewerbliche Betriebsstatistik.” Statistik des Deutschen Reichs 213 (1910).

Kaiserliches Statistisches Amt. “Ergebnisse der deutschen Produktionserhebungen.” Vierteljahrshefte zur Statistik des Deutschen Reichs 22, no. 3 (1913), Ergänzungsheft.

Kaiserliches Statistisches Amt. “Die Ergebnisse der deutschen Produktionserhebungen.” Vierteljahrshefte zur Statistik des Deutschen Reichs 23, no. 3 (1914): 108-35.

Kendrick John W. Productivity Trends in the United States. Princeton, NJ: National Bureau of Economic Research, 1961.

Krijnse Locker H., and M. D. Faerber. “Space and Time Comparisons of Purchasing Power Parity and Real Values.” Review of Income and Wealth 30, no. 1 (1984): 53-83.Google Scholar

Landes David S. The Unbound Prometheus: Technological Change and Indus-trial Development in Western Europe from 1750 to the Present. Cambridge: Cambridge University Press, 1969.

Maddison Angus. Phases of Capitalist Development. Oxford: Oxford Uni-versity Press, 1982.

Maddison Angus. Dynamic Forces in Capitalist Development: A Long-Run Comparative View. Oxford: Oxford University Press, 1991.

Maddison Angus. Monitoring the World Economy, 1820-1992. Paris: Organisation for Economic Co-operation and Development, 1995.

Maddison Angus. The World Economy: A Millennial Perspective. Paris: Organisation for Economic Co-operation and Development, 2001.CrossRef

Nuxoll Daniel A. “Differences in Relative Prices and International Differences in Growth Rates.” American Economic Review 84, no. 5 (1994): 1423-36.Google Scholar

Reichsamt für wehrwirtschaftliche Planung. Die deutsche Industrie - Gesamtergebnisse der amtlichen Produktionsstatistik. Berlin: Verlag für Socialpolitik, Wirtschaft und Statistik Paul Schmidt, 1939.

Ritschl Albrecht. “Spurious Growth in German Output Data, 1913-1938.” European Review of Economic History 8, no. 2 (2004): 201-23.Google Scholar

Rostas Laszlo. Comparative Productivity in British and American Industry. Cambridge: Cambridge University Press, 1948.

Summers Robert, and Alan Heston. “A New Set of International Comparisons of Real Product and Price Levels: Estimates for 130 Countries, 1950-1985.” Review of Income and Wealth 34, no. 1 (1988): 1-25.Google Scholar

Tooze J. Adam. Statistics and the German State, 1900-19145: The Making of Modern Economic Knowledge. Cambridge: Cambridge University Press, 2001.

Ward Marianne, and John Devereux. “Measuring British Decline: Direct Versus Long-Span Income Measures.” This JOURNAL 63, no. 3 (2003): 826-51.Google Scholar

Williams Ernest E. Made in Germany. London: Heinemann, 1896.