Nominal and real wages for manufacturing workers, 1860–2007
Högre seminariet på ekonomisk-historiska institutionen, Göteborgs universitet, 15/4, 2009
The rise or fall in nominal wages is one of the most central aspects of an economy in which competitive labour markets determine the price of manpower. Nominal wages represent the marginal returns to a unit of labour input. In addition, the movement of nominal wages, deflated by a cost-of-living index, indicates progress or decline in living standards, the most important by-product of economic growth. Throughout most of the latter two centuries, and also in many of today’s less developed countries, the ratio between wages and prices has
refused to remain on a long-term upward track, preventing workers form gaining sustainable increases in the purchasing power of goods and services. It is only during era of modern economic growth that the expectations of perpetual rises in real wages have come true. From our vantage point, workers have gained enormously during the one and a half century this chapter reviews. Since 1860, real wages for male manufacturing workers have on average increased 2.2 percent annually implying that purchasing power thus measured has increased by a factor of 24. Apart from the gains they have reaped in increased purchasing power over an almost infinite variety of goods and services, they have also enjoyed shorter working hours per week, longer holidays and decreased years of work over the lifetime. Other benefits accruing to labour includes improved working conditions and greater security in the face of unemployment and sickness. The importance of real wages as a manifestation of the benefits of economic growth serves to justify the close attention social scientists devote to the
1measurement of nominal wages.
The purpose of this chapter is to document series of nominal and real wages for male and female manufacturing workers in 1860-2007. An attempt is made to construct nine wage series for groups of industries and an overall average series for manufacturing as a whole, spanning the era in its entirety and circumventing the problem with inter-temporal inconsistency. The first half of the chapter deals with the nature, sources and definitions of the historical wage sources available, with particular attention to the different wage sources before and after 1913. The second half contains a discussion of the results, with respect to growth rates of male real wages and differences in wages among industries and between female and male workers.
1 Scholliers (1989) gives a brief introduction to the significant role the measurement and interpretation of real wages have played in the historiography of economic history.
Quality of labour
Needless to add, employers have altered their minimum hiring standards many a time over the course of time. This project shares with all historical wage studies the intention of establishing wage series with as high degree of inter-temporal consistency as possible in the qualifications demanded of the workers. Commonly, attempts to document long-term series for wages seek to trace the evolution of wages for an unchanged professional characteristic, like unskilled, manual or plainly labourer. The idea is that it is possible to trace the nominal price of a particular kind of labour service the essential feature of which has remained, despite the many far-reaching implications of technological development, fairly unchanged over time. It is also sometimes suggested that this kind of labour service command no industry-specific formal skill or requires little or no on-the-job training. Therefore unskilled workers are subjected to similar demand and supply forces in the labour market which do not derive from industry-specific developments of new technologies (Williamson 1995).
Before the share of manufacturing industries reached significant importance in the economy typical workers labelled unskilled were mostly found in the construction sector, for
2instance helpers in the building trade or unskilled road and canal construction workers. For
Sweden, though, apart from an investigation of municipal workers’ wages in the construction
sector in 1860–1930, which was published in 1933, there is to date scarce supply of wage data for construction workers for the period under review. Even though the investigation of municipal workers is a numerically ample body of wage quotations, I have excluded it from this wage study which intends to include only private employments. Including the wage series for municipal workers would make the final average for manufacturing workers’ wages
inconsistent and more difficult to interpret.
In the initial stages of industrialisation it is perhaps doubtful whether unskilled workers were a commonly found labour service as most manufacturing industries were small-scaled and relied on handicraft-based technologies; they were what Sokoloff (1984) tellingly referred to as artisan shops. Some exceptions there certainly were, for instance the sawmill industry and iron and steal works, who employed a mass of helpers and transport workers, but the preponderance of unskilled workers in manufacturing industries probably awaited mass
3production technologies in the twentieth century. In their major contribution to the
2 The most classic example of long-term series of construction workers is that of Phelps Brown and Hopkins (1955, 1956). A more recent one is Allen (2001). For Sweden see Söderberg's contribution in this volume. 3 See Montelius et al. (1959) for insights into the type of labour iron works needed, and for the labour service performed at the sawmill industry, see Gustafsson (1965) and Cornell (1982).
documentation of wages before the annual publication of official wage statistics, Bagge et al.
(1933, 1935) presented separate unskilled and skilled wage series for some industries. The average skilled-unskilled wage ratio remains without any significant tendency to either increase or decrease between 1860 and 1913 which makes the issue, skilled or unskilled wages series, unimportant. In Social Boards official wage statistics, which came into being in 1913, there is no specific occupational category like common labour or unskilled labour. A mere distinction is made between workers on the one hand and technical personnel (engineers, chemists, and supervisors), administrative staff and shop assistants on the other. Thus the label worker is very heterogeneous. Statistical agencies in the twentieth century have in general not responded to economic historians’ infatuation for the prefix unskilled (David and
Solar 1977). It is a great paradox that whereas the rise of mass production technologies in the twentieth century turned the manufacturing industry into a great employer of the typical unskilled worker, the label unskilled fell out of favour among statistical agencies. When the twentieth century drew to a close, and the forces of technical development and capital accumulation served to contract the labour force in manufacturing industries, the share of unskilled workers started to decline. Perhaps the market service sectors, some of which are characterised by a multiplicity of quite monotonous labour services, have instead become the most important employer of unskilled workers.
Historical wage sources
Today’s researchers with a keen eye on the historical dimensions of wages depend crucially on past generations concern with the detailed documentation and compilation of wage records. Various possible wage sources are at our disposal, which makes the enterprise to patch some or all of them together, to capture as long a time span of history as possible, particularly challenging. Sweden is known for its rich source of statistical material from earlier times, for instance a census from 1749 and annual publications of industrial statistics from the middle of the nineteenth century. With wage statistics one is however inclined to take a less benign view on the supply of useful source material, in particular for the nineteenth century. For instance, we have nothing even close to resembling the Norwegian Vedergang’s arkiv, which supplies a
wealth of wage quotations back to the mid-eighteenth century (Grytten 2007). We also lack the type of government surveys which were conducted in the US post-bellum years, the so-called Weeks and Aldrich reports (Margo 2006), or the wage material which A. L. Bowley
and G.H. Wood used to compiled their pioneering index of average earnings in the UK during the nineteenth century (Feinstein 1990).
In relation to the US and the UK, the overwhelmingly agrarian nature of the Swedish economy in the latter half of the nineteenth century may explain the low priority Swedish authorities gave to the systematic collection and presentation of wages. Not until the first decade of the twentieth century, when the first major labour conflicts threatened to disrupt the social order, were a series of investigations into the working conditions and wage remunerations in manufacturing industries carried out, which paved the way for the establishment of Social Board’s official wage statistics in 1913. Our knowledge of wage
movement for the pre-World War I era relies instead on the large-scale project Wages in
Sweden, which contains the results of a major research project intended to trace the evolution of nominal and real wages (Bagge et al. 1933, 1935). After 1913 the official wage statistics
will be used. The presentation of the sources and methods used to construct series of Swedish wages thus falls into two parts. The first covers 1860–1913, and deals with Wages in Sweden,
and the second covers 1913–2007, coinciding with Social Board’s, first, and Statistcs Sweden thereafter, annual publications of official wage statistics.
Wages in Sweden, 1860–1913
It may be useful to distinguish between wage rates and earnings, although the boundary between the two is frequently blurred. Many historical studies, including the present one, are forced to depend on a mixture of these two types of wage measures. Wage rate is the stipulated return for a given time unit of labour input, most often hourly or daily. It excludes payment for overtime work, premiums bonuses and other forms of payment by result as well as payment in kind. Studies intending to establish the movement of wage rates and capture a lengthy time span of history ultimately rely on old firm records where the stipulated wage rate is given, or on wage rates established by labour unions. Earnings are a more inclusive measure which incorporates the effect of changes in the composition of the labour force between industries and of changes within each industry by age sex, skill and region. It also allows for the effect of payments for overtime work, and for premiums, bonuses and other forms of payment by results as well as any rise in hourly rates paid to compensate for a reduction in standard hours worked. Dividing the wage bill by the number of working hours derives a measure of average hourly earnings.
In the 1920s, Gösta Bagge, at the Institute for Social Sciences at Stockholm University, launched a large-scale project with the explicit intention of tracing the evolution of wages,
4 Two of the final five cost of living and national income for Sweden between 1860 and 1930.
volumes which were published in the 1930s deals with wages, the first in the manufacturing and mining sector and the second in the municipal service sector and the agriculture and forestry sectors (Bagge et al. 1933, 1935). The originality of their research project concerns the wage series between 1860 and 1913, which were based on thus far unused archival sources. For the period 1913 to1930 the wage data were gathered from Social Board’s official
wage statistics thus belonging to the next section.
Weak unionisation in Swedish manufacturing industries pre 1900 precluded Bagge et al.
from using union wage rates. Instead they relied on wage rates paid to individual workers based on information from about 100 firms whose pay rolls had survived. The method used to construct wage series in 1860–1913 can be summed up as a four-stage procedure: First,
collecting pay rolls from firms whose wage records had survived. Second, in each firm selecting workers from a few representative occupations. One of the selection criteria was that the worker’s employment exceeded 11 months, thus excluding temporary workers. A worker's
wage was not recorded for more than 5–10 years, circumventing the influence of age on the
movement of wages. Payment in kind, probably a more important component in the latter half of the nineteenth century than it is today, was generally omitted. The most important benefit was access to free housing. The authors discussed the matter at length yet approached it unambiguously. In most instances the value of free housing was not incorporated into earnings. In other cases, when some workers received a cash allowance for rent while others had free housing, the value of free housing was included. Third, computing wage averages for each firm. Fourth, when the different wage series of the firms showed a similar movement over time, they were combined to form industry-specific wage series, and the different series assigned weights based on the quality of the series, the size of the firm and the relative importance of the industry.
The method describe above is sometimes called kinetic since it tracks the movement of
wages better than it establishes the average wage level of all workers in a particular industry, a
methodology which is attributed to A. L. Bowley (1895). Cross-sections were therefore used to compare the wage levels of the selected workers in relation to the average level of all workers in a particular concern. These cross-sections were made for 1865, 1885 and 1905 and
4Carlsson (1982) gives all the fascinating details about the whole project.
it was found that in most cases the wage level of the selected workers was on average five percent higher than the average of all workers. In addition, the authors considered carefully the problems of heterogeneity with respect to age, gender and professional qualifications. Today we stand to benefit greatly from that careful consideration because it made possible the documentation of homogenous professional characteristics. For most of the industries there are separate series of skilled, in a few cases also semiskilled, and unskilled workers. Furthermore, their industry-specific wage series excludes the potential influence of changes in the composition of the labour force. The disadvantages concern for instance the low coverage and the unequivocal approach to the benefits in kind. It is also doubtful whether their wage levels can be used to address issues of standards of living since they do not reflect the levels of incomes that workers had at their disposal.
The wage data for manufacturing represent the most comprehensive part of the project. The authors documented 7 industry-specific wages series in the main section of the volume, while the appendix contains the original wage data from the different firm records. Besides these 7 manufacturing industries there are also wage series for sugar industries and railway workers, but these industries cannot be linked to the post-1913 series and thus omitted here. In relation to the 9 industries which this chapter includes, Bagge et al. fail to provide wage series
for the quarrying industry and the chemical industry. Based on the wage material presented in their monographs section, I have attempted to establish wage series for these industries, though admittedly these two series rest on a more tenuous empirical foundation. For female workers there are two industry-specific wage series, textile and food, but no aggregate series.
What do we know about the reliability of the wage evidence presented in Bagge et al.
(1933)? Admittedly, not so much. Gustafsson (1965) has shown that there are problems with the wage series of sawmill workers in 1890–1913. Based on an investigation into the
movement of wages for sawmill workers in Norrland, 1896–1913, he argued that Bagge et
al.’s series makes too little progress and that the wage levels are too high. He has also raised concerns about the omission of temporary workers. Berglund (1982) has criticised the methodoly and the wage data for workers in the matches industry. However, no major new attempt has been made to gather new wage data in the extent similar to which Bagge et al. did.
Prado (2008) has cast some doubt on the weighting scheme applied in Bagge et al. The lack
of transparency in the description of how the different wage series were combined into an aggregate measure makes it difficult to replicate their result. Schön (1988) has, within the context of Swedish historical national accounts, raised the level of gross output for several industries, including sawmills. We therefore have reasons to suspect that weighting with
respect to these higher levels of gross output, and thereby employment, will modify Bagge et
al.’s average wage series for manufacturing workers. Weighting the industry-specific series
by recognition to the large number of workers employed in the sawmill industry gives a more slow growing series of wages for manufacturing at large pre-1890 compared to Bagge et al.’s
aggregate series, as table 2 shows. The reason is the sluggish growth of sawmill workers’
wages. They grew steeply in the wake of booming conditions in the world markets for sawmill products at the first half of the 1870s, but collapsed when the boom turned to bust in the second half of the 1870s. In fact, sawmill workers’ wages did not resume a sustained
upward course like that of manufacturing until the beginning of the 1890s. Thus, the aggregate series of manufacturing wages offered here makes slower progress than Bagged et
al.’s original series.
The official wage statistics, 1913–2007
Apart from the wage statistics of day workers in agriculture which were reported in the Swedish Agricultural Statistics, the first official attempt to document wage conditions covered the manufacturing sector. The reason was that the first far-reaching labour conflicts in Sweden occurred among industrial workers. The authority felt a pressing need for more information on labour and wage conditions in several manufacturing industries. In 1899–1909
the Board of Trade (Kommerkollegium) therefore undertook a number of special investigations into the following industries: bakeries, tobacco industry, mechanical engineering and book printing. These investigations were carried out with meticulous care, impressive in terms of scoop and coverage, providing wage information on separate workers
5classified by education, sex, occupation and region. However, the inflationary conditions in
the aftermath of World War I, when money wages increased dramatically, first, and then plummeted in the deflation that ensued, required that wage statistics be published annually. Otherwise the wage statistics would be outdated and of limited use once they saw the light of day.
To render an annual publication possible several modifications, in practice simplifications, were necessary (SOU 1931). As with most official wage statistics the method used by the Social Board from 1913 onward was designed to provide a measure of average hourly earnings, thus fundamentally different from Bagge et al.’s attempt to provide a more
specific and homogenous measure of wages. The Social Board collected data by requesting
5 A recent doctoral dissertation, Lundh Nilsson (2007), is based on the investigation of the mechanical engineering industry.
6 It extracted firms to provide information on wages and other closely related issues in surveys.from a single concern the total sum paid out as wages during one year and divided this sum by the number of working hours, days or workers (the average number of workers on a number of pay days, for instance once a week). By doing this it obtained a measure of average hourly, daily and annual earnings. There are three types of hourly wages, called a, b and c. Category a
and b represent payment for regular working hours, where a stands for time-wages and b
piece wages. Piece wages are generally higher. Label c is the total wage sum paid out divided by the number of working hours. The wage sum includes all kinds of payment, whether they consist of piece wages, time wages, benefit in kind, holiday pay or overtime. The official statistics uses the c-wage to illustrate movements of hourly wages over time, and it is the only measure which is accessible throughout the whole period. It is therefore the measure of average earnings which is used in this study.
Bagge et al. levelled damning criticism against the way the official wage statistics were constructed. A series of earnings so constructed makes no distinction between changes in the composition of the work force, except holding constant for the impact of female, males and under age, for which the official wage statistics provided separate information. The inherent problem with this method, as the authors correctly pointed out, is that a change in the computed average may be caused by a number of factors other than changing wage rates. For instance, if the composition of skilled and unskilled workers changes in favour of skilled workers the increase in the average wage mirrors not only changing wage rates but also the compositional effect of higher wages for skilled workers. Likewise, if the average age of the working population in a specific firm increases the measure of average earnings will increase too. Further problems include the composition between full and part-time workers. A substantial share of temporary workers will drive a wedge into the measure of average annual and hourly earnings. In many cases the actual number of employed workers is larger than the average number of workers given by the statistics, which leads to too high an estimate of the annual income per worker. This is especially problematic in industries with pronounced seasonal characteristics or those employing casual labour. The method does not cause any problem with the number of hours, however, which means that in the official statistics hourly wages are more useful than annual.
Even though Bagge et al. painted the method of the official statistics in a quite
derogatory light, they acknowledged their usefulness for indicating the movement of wages
6 Socialstyrelsen and SCB used the same methodology.
post-1913. In their final index the official post-1913 series is linked to their pre-1913 series to provide a wage series covering the entire 1860–1930 era. They also admitted that the official
statistics covered a much broader segment of the manufacturing sector. In 1930, it encompassed 67 percent of the industries, and while that share admittedly was smaller in 1913, it rose in the course of the twentieth century to include the lion’s share of all firms.
The years before 1921 represent the most serious lacuna in the official statistics. It is sometimes wrongly asserted that the Social Board began to publish wage statistics annually in 1913. In fact, it did not start publishing annually until 1919. In 1918 the Social Board requested firms to give, not only wages for that year, but also for 1913–1917. This wage
material is less complete, and less reliable, than for post-1919 years, and there are no wage data for 1914 and 1915. Besides, no information on average hourly wages was given between 1913 and 1920. Later on, The Social Board estimated hourly wage series for male and female workers for manufacturing as a whole on the basis of information on working hours, but the gap in our knowledge as to the evolution of hourly wages in different industries still exists. There are two ways out of this dilemma, none of which is very satisfactory. First, to extrapolate backwards from 1920 to 1913 by drawing on the series of daily wages. However, the substantial number of working hours which were cut in the two years preceding the Hours of Labour Act in 1920 makes daily wages a bad indicator of hourly wages. The rate at which reductions took place varied widely by type of activity and location which means it is difficult to ascertain the movement of hourly wages for different industries. And there are still gaps in 1914 and 1915. Second, to fill the gap with Social Board's aggregate hourly wage series, a solution which imparts the misleading impression that all industries grew in tandem? The preferred solution here is to leave out the years between 1913 and 1921 for all the industry-specific wage series.
The labels of the nine different manufacturing industries is conditioned by a
7, which was put into practice in 1993. In fact, SNI 92 classification scheme called SNI 92
severely limited the number of wage series that could possibly be constructed. The official statistics provide more disaggregated wage data from 1921 to 1985, but the only way to carry
8the series forward to the present day is to base the classification of industries on SNI 92.
Some of the major changes which SNI 92 brought about cannot be dealt with satisfactorily, for instance, wood industries do not include furniture after 1993. The most far-reaching
7 Svensk näringslivsindelning 8 SNI 92 turned into SNI 2002 but that implied only minor modifications.