Oisín Gilmore
University of Groningen
How Was Life? Volume II
New Perspectives on Well-being and Global Inequality since 1820
3. The working week in manufacturing since 1820
Abstract
Despite the substantial implications of working time for considerations of living standards and economic output, historical data on working hours are sparse. This chapter presents a new dataset on the length of the working week in manufacturing globally from 1820-20101. The dataset contains some 4 300 observations and covers 120 countries or political units. It shows that workers in manufacturing worked 60 to 90 hours per week in the 19th century, compared to around 40 hours today. This is a reduction of 20-50 hours, that is, 50-125% of today’s average working week. The data also show that weekly working hours declined rapidly after World War I, with the introduction of the eight-hour day, and again later in the 20th century, with the generalisation of the five-day week. This decline in weekly working hours stalled since the 1950s-60, and seem to have reversed in the most recent period.
Introduction
Working time and leisure time should be central concerns of any attempt to evaluate people’s well-being. As the OECD notes in the flagship How’s Life? publication, working hours and time off work are central to work-life balance. When considered globally, we continue to see substantial disparities in working time, with workers in some countries working significantly more than in others. This is equally true over time.
Amongst OECD countries the amount of time people spend working in a year has declined dramatically since the 19th century, and, with that, the amount of free time people have away from work has also substantially increased. This reduction in working time has two major components: first, more time spent on holidays and days off work, and second, a dramatic reduction in the length of the average working week (Huberman and Minns, 2007[1]).
This chapter presents a new dataset on the length of the working week in manufacturing globally from 1820-2010. The dataset contains some 4 300 observations and covers 120 countries or political units.2 This sectoral focus on hours per week in manufacturing allows us to ensure that the data are broadly comparable between countries and over time.
This chapter shows that workers in manufacturing worked 60 to 90 hours per week in the 19th century, as compared to roughly 40 hours today. That is a reduction of 20-50 hours, i.e. 50-125% of today’s average working week. The welfare implications of a reduction in working time of this scale are of course substantial. Both the reduction in the pain and toil of work and the increase in leisure time have substantial positive welfare implications (de Jong, 2015[2]). However, working time is often ignored in historical studies due to the inadequate quality of existing international datasets. This chapter advances our understanding of these issues by presenting a new international dataset on working hours in manufacturing since 1820.
The content of this chapter is as follows. The chapter starts by first describing the concept of “working time” and considering some of its historical and contemporary ambiguities. It then presents the existing datasets on working time in the long run and the historical sources used to build the new dataset presented in this chapter, and it assesses the quality of the data. The subsequent sections describe some of the main highlights of trends in working time between 1820 and 2010, and then some extensions of the measure of working time from a weekly measure to an annual measure for a restricted number of countries. The final two sections consider how working time correlates with real GDP per capita and what the priorities are for future research on this topic.
Description of the concepts used
Working time might appear a rather simple and uncomplicated concept. It would seem to be just the hours that you work. Unfortunately, it is a rather more complex concept with many ambiguities.
The first issue is what is “work” and what separates it from “non-work”. To a worker in a contemporary developed economy, this might seem like a straightforward question, but if we think historically and consider the different approaches to “work” over history, the complexity becomes clear. Likewise with “time”: while this may seem a simple uncontroversial concept, if we look at different attitudes to working time over the historical long run, it turns out that the idea of clearly delineated working time is a very modern concept, particular to our historical moment.
Marshall Sahlins’ acclaimed anthropological work, Stone-Age Economics (Sahlins, 1972[3]) on hunter-gatherer societies, describes how concepts of both work and time were fundamentally different in the !Kung hunter-gatherer society of southern Africa. First, he noted that the time the !Kung spent at work was dramatically less than the modern Western worker, but he also noted that the concept of work was something that researchers were statistically imposing on the activities of the hunter-gatherers. The clear distinction between work and non-work was not present in the !Kung people’s self-description of their activities (Sahlins, 1972[3]). This observation has been repeatedly confirmed in other studies of hunter-gatherer people. Again and again the average “working time” of hunter-gatherers, if it can be called that, has been found to be far lower than in advanced Western societies (Clark, 2005[4]).
Looking at the Middle Ages, Jacques Le Goff, in his classic essay on working time in medieval Western Europe (Le Goff, 1982[5]), argues that before the 14th century the approach to working time was fundamentally different to ours. He writes that “on the whole, labour time was still the time of an economy dominated by agrarian rhythms, free of haste, careless of exactitude, unconcerned by productivity – and of a society created in the image of that economy, sober and modest, without enormous appetites, undemanding, and incapable of quantitative efforts” (Le Goff, 1982[5]).
Clearly, across time and between different regimes of labour, conceptions both of work and of working time can be radically different, to the point that neither concept properly applies. Table 3.1 lays out an incomplete, schematic description of how regimes of working time have differed over the long run.
Table 3.1. Different patterns of time use in hunter-gather, agrarian, proto-industrial, industrialising and industrialised countries
When we try to develop estimates of working time historically, we find radical differences in the organisation of time in relation to work. In agrarian societies, work is highly seasonal and dependent on land quality and on the type of farming being done.3 Several studies, primarily looking at the United Kingdom, have found that prior to the Industrial Revolution urban workers appear to have worked full days (i.e. 10-12 hours) but far fewer days a year than subsequently. Prior to the Industrial Revolution, workers worked in the region of 150-200 days per year. During the Industrial Revolution this increased to over 300 days per year.4 This contrasts with industrialised economies in the 20th century when restrictions on hours of work with the introduction of the eight-hour day and the emergence of substantial paid holidays emerges as the norm. A clear distinction between working time and free time emerged in that period.
The fact that the work time/free time distinction is a highly modern one is not the only conceptual problem related to working time. Even in developed economies where this distinction is widely understood, the concept of work time contains other ambiguities. One obvious problem is that much work is unpaid. For example, when calculating the work time of a mother, what consideration should we give to unpaid domestic work? Another problem is whether all time devoted to work should be considered work time. Time spent commuting to and from work is not generally considered as work time, but it is, effectively, time devoted to working. But what about time moving to and from work sites while on the job? For example, in the case of coal miners, should the time spent travelling from the opening of a mineshaft to the coal face be considered work? Or what about meal times? Whilst a two-hour break for a siesta might not be considered part of work time, what about a 30-minute break for a coffee and sandwich? The assessment of these issues is ultimately contingent on social convention (Gershuny and Sullivan, 2019[11]).
These issues have been of concern to labour statisticians for a long time. The first International Conference of Labour Statisticians (ICLS), convened by the International Labour Organization (ILO) in 1923, agreed on a set of guidelines for the collection of statistics on working time (1923, 1924). These guidelines were then revised in 1962, at the tenth ICLS, and yet again more recently at the 18th ICLS in 2008. Under the latest guideline, working time is no longer confined to hours in paid employment and includes self-employment. And a distinction is made between two broad categories of working time data (International Labour Organization, 2008[12]).
The first establishes the preferred form of data: hours actually worked. This includes three elements: hours spent on an activity directly related to work (time spent working, moving between work locations, in-work training, etc.); in-between time (the inevitable interruptions to work processes that occur); and resting time (coffee breaks etc.). It explicitly excludes commuting time, long breaks such as meal breaks, annual leave, holidays, time spent in education or in training that is not part of the job and other reasons for not working (maternity and paternity leave, parental leave, slack in business, bad weather, etc.).
The second category of working time data is slightly broader and includes measures of working time calculated in the employer/employee relationship. These include in particular the concepts of normal hours, (usual) working hours and hours paid. Normal hours are the hours above which overtime would be paid, and are often set down in legislation, collective agreements, etc. Usual hours include in addition regular overtime. Hours paid simply refers to the hours for which a worker is paid. Note that the primary difference between these is overtime that is not paid. And as the 1962 guidelines note, “because of the wide difference among countries with respect to wage payments for holidays and other periods when no work is performed, it does not seem feasible at this time to adopt international definitions of hours paid for” (International Labour Organization, 1962[13]).
Historical sources
This chapter presents a newly collected global dataset on working time in manufacturing. It improves the three existing international datasets on working time (see below) in a number of ways. First, it covers a longer time period. Second, it has a far greater number of observations. Third, unlike all three previous datasets on working time, it does not rely on a large number of interpolated estimates. Finally, unlike previous datasets, it focuses exclusively on manufacturing: as explained below, it does so to ensure that the data are of the highest possible quality and broadly comparable over time and between countries.
The first of the three previously developed international datasets on working time is the dataset developed by Angus Maddison and presented in slightly different forms in a series of publications (Maddison, 1964[14]; 1995[15]; 1982[16]; 2001[17]). In these publications, Maddison takes decadal observations for a small number of countries from the ILO Yearbooks of Labour Statistics, and from a single observation for the average working week in the United Kingdom in 1860 he constructs very rough estimates for the years between 1860 and 1938. The second dataset is the Total Economy Database, produced by the Conference Board (2019[18]), which provides estimates of annual (rather than weekly) hours worked for 64 countries from 1950 onwards. This dataset draws on data from the OECD, Eurostat, the Asian Productivity Organization and a small number of research papers (Crafts, 1997[19]; Hoffman, 1998[20]). A substantial amount of the data in the Total Economy dataset, especially for earlier years, is based on interpolations and extrapolations. The third historical dataset is the one compiled by Huberman and Minns (2007[1]), perhaps the best of these previously available datasets, although it covers only 15 countries. From 1929 onwards, it takes most of its data from the prior two datasets, adding to them a new set of data for 1870-1900 collected by Huberman (2004[21]; 2012[22]), from the Fifteenth Annual Report of the US Commissioner of Labor (U.S. Department of Labor, 1990[23]).
The data presented in this chapter represent a significant improvement on previously available long-term series on working time. The dataset covers a much longer time frame and involves no interpolations or extrapolations except those in the source datasets. However, unlike the above three datasets, the data in this chapter focus exclusively on manufacturing, rather than attempting to estimate the average working week across the entire economy. By focusing exclusively on manufacturing, we can ensure that the data are broadly comparable over time and between countries.
While it would be preferable to have data on average working time for the entire economy, there are significant difficulties in estimating this over the long run. For much of the economy, little data on working time was recorded. Despite agriculture accounting for the vast majority of working time expended in most economies over history, we have reliable data on working time in agriculture for only a handful of countries. There is also very little data on working time in the service industry. While there is reasonably good data on working time in construction, mining and transport, country coverage for these series varies greatly.
Some estimates of the average working time outside agriculture do exist. However, the industries included in these estimates often differ by country, especially for earlier years.
Ideally, we would have reliable data on average working time across the entire economy, or at least data on the average working week outside agriculture. But neither of these datasets is achievable given the historical data available. Therefore, a choice needs to be made. On the one hand, it would be possible to present data that cover the greatest possible proportion of the workforce. However, because different parts of the workforce would be included at different times and in different countries, this data would not be comparable across time and between countries. On the other hand, we could choose a specific section of the workforce to focus on. This would allow for the data to be comparable across time and between countries, but it would be less comprehensive. In this chapter I choose the latter strategy by focussing on working time in manufacturing. For nearly all countries, the best time series on working time refer to manufacturing. Focussing on manufacturing thus ensures that the data is of the highest possible quality.
The new dataset presented in this chapter has been collected from three historical sources. The first and most important source is the series of ILO Yearbooks of Labour Statistics. A version of the Yearbook has been published since 1934 covering countries from 1927 onwards. All data on working time from each annual publication for 1934-70 has been extracted. Unlike Maddison, who took decadal observations for a small number of countries from the ILO Yearbooks, I take all the annual data available from every single Yearbook from 1934 to 1970. From 1970 onwards, I drew on the data that is available from the ILO website. However, this data oddly does not contain all the data of the ILO Yearbooks. I therefore completed these data by adding data from ILO Yearbooks for 1980, 1990, 1995, 2000 and 2009, covering all years from 1970-2008.
The second historical source is country-level studies. I collected these for 16 countries: Australia, Belgium, Canada, China, Denmark, Finland, France, Germany, Ireland, Italy, the Netherlands, Spain, Sweden, Switzerland, the United Kingdom and the United States. These studies are primarily in two forms. First, there are government publications from the respective country. Second, there are academic studies on the history of working time in the respective country. These data sources are fully described in Gilmore (forthcoming[24]). For this chapter, this type of data is used only for the period 1820-1939. After that, all the data come from the ILO.
The third source is the data from the Fifteenth Annual Report of the US Commissioner of Labor collected by Michael Huberman (2004[21]; 2012[22]), covering the years 1870-1900. It should be noted that this presents estimates for the entire economy, not exclusively manufacturing. However, these estimates are dependent largely on data on working time in manufacturing. Further, for the countries where Huberman presents data for both manufacturing and the entire economy, the figures are broadly similar to each other (Huberman, 2004, pp. 970-971[21]).
The data from the various historical sources were combined in the following manner. From the ILO Yearbooks, if available, I used the data series on the average working week for total manufacturing. If this was not available, I constructed an unweighted average of working time in various manufacturing industries. If no data from the ILO was available and the year was 1939 or earlier, I used data from national sources. Once again, if data for the average working week for total manufacturing were available, I used those. If not, I constructed an unweighted average of working time in various manufacturing industries. If none of this data was available and the years were between 1870 and 1900, I used data from Huberman.
Data quality
Generally, the data used in this chapter are of good quality and refer to actual hours worked by workers in manufacturing, although some data refer to hours paid for or normal hours worked. Almost all data come from establishment surveys or household and labour force surveys where employers or workers are asked how many hours were worked over the previous week or month.
The most significant problems that might affect the consistency of the data are in relation to, first, whether or not paid non-working time (e.g. breaks and holidays) is included; and second, whether or not unpaid working time (e.g. unpaid overtime) is included. As noted above, measurement of actual working time according to ICLS guidelines measures only hours of actual work. It does not include breaks, holidays and other forms of paid non-working time, and it does include unpaid working time such as unpaid overtime. For later years, most data were collected in line with the ICLS guidelines and are therefore highly comparable. For earlier years, most data are also on hours of actual work, but as the ICLS guidelines did not always exist, actual working time might be measured differently in different countries. However, as workers generally received fewer paid holidays in the past, estimates of actual working time are likely to include overtime and unlikely to be biased by the inclusion of substantial paid leave.
Table 3.2 describes the data quality using the four levels of quality described in How Was Life? (van Zanden et al., 2014[25]).5 From the 1930s onwards, nearly all data come from the ILO, i.e. “an official statistical agency …[that]… uses techniques that ensure equivalent credibility” (Ibid.). Prior to that year, the data come from a range of national level sources. But again, nearly all are measures of actual hours of work and are based on survey data. Little data are available prior to the 1870s, and the data are of poorer quality.
Table 3.2. Quality of data on working time by region and benchmark year, 1820-2010
Note: 1. High quality: the product of official statistical agency (national or international); 2. Medium quality: the product of economic-historical research using the same sources and methods as applied by official statistical agencies; 3. Moderate quality: economic historical research, but making use of indirect data and estimates; and 4. Low quality: estimates based on a range of proxy information. In case of multiple sources, the lowest quality source is given.
Some issues affecting data quality should be flagged. First, although almost all of the data were reported in the original sources in the form of hours per week, this was not always the case. A small amount of data was given in other forms and had to be adjusted (e.g. data reported in the form of hours worked per fortnight were divided by two; those given in the form of hours per month were multiplied by 12/52; and data given in the form of hours per day were converted to hours per week by assuming that workers worked 5, 5 ½, or 6 days a week – depending on the time period considered). This application of these assumptions should not be particularly controversial. Frequently, the normal number of days in the working week could be easily calculated from the available data. Occasionally, it was assumed that the number of days in the working week was similar to neighbouring countries.6
A second larger problem of data quality is the problem of representativeness. In one case, this problem is very stark. Most of the data on working time in sub-Saharan Africa prior to the 1980s refer to the average working week of white male South Africans. For the entire dataset a major problem is how representative this dataset is of women workers. In most cases the original sources did not state whether the data are for male or female workers. But where data are not given for both sexes, unless a figure for all workers is given, I have used the data for male workers. I made this decision because I often do not have reliable consistent data on the gender composition of the workforce in manufacturing, so I am unable to generate an estimate of the average working-time weighted by sex. Fortunately, from available data we know that during this entire period most workers in manufacturing were male, and therefore the average working time for male workers should be broadly representative of the average worker (Mitchell, 2013[26]).
For series early in the dataset, there are further problems of representativeness. First, the data often refer to a few manufacturing industries, which may not be representative of all manufacturing workers. Second, especially in the 19th century, there was substantial seasonal variation in hours of work, which are not fully accounted for in this dataset due to data constraints.
Despite these problems of representativeness, the data presented in the chapter are deemed to be of high quality for historical data and provide significant new insights into the history of working time.
Main highlights of historical trends in working hours
Figure 3.1 presented below shows a dramatic decline in the average working week in manufacturing across the world. In the late 19th century, working hours were in the region of 60 per week while, by the start of World War I, working hours were around 55 hours per week. After World War I there was a dramatic decline in working hours in the West, with the introduction of the eight-hour day and the forty-eight-hour week in both Western Europe and in the Western Offshoots (including the United States, Canada, Australia and New Zealand). By the start of the second half of the century, working hours in manufacturing were generally somewhere between 40 and 48 hours a week. Since then, the decline in working hours has to a substantial degree stalled. In the most recent period, there is some evidence that they are beginning to increase again. Figure 3.1 also shows the decline in working hours outside the West. This generally followed a similar trend as in the West but condensed into a shorter time period.
Figure 3.1. Average working hours in manufacturing across world regions, 1870-2010
Note: Dots indicate regional averages with population coverage above 40%.
Table 3.3 below allows us to look more closely at these changes in working time by presenting developments at a national level since 1820. In both Western Europe and the Western Offshoots, the early 18th century was characterised by an increase in working time from high levels to even higher levels, followed by a decline. This suggests a familiar Kuznets (inverted U) shape to the development of working hours relative to GDP (Spoerer and Streb, 2008[27]), i.e. one where the initial stages of industrialisation bring a number of negative side effects but, as economic development proceeds, those negative effects diminish. As with most other types of Kuznets curve, the relationship appears to break down in the 1980s and 1990s, with the decline in hours plateauing in the West and increasing in medium-income countries like Turkey, Egypt and China.
Table 3.3. Average working week for manufacturing workers in selected countries, 1800-2010
The early stage of this curve has been well studied by historical research. The increase in working time during industrialisation has been a major issue in both social history (Thompson, 1967[28]; Reid, 1976[29]) and economic history (Bienefeld, 1972[30]; Clark and Van der Werf, 1998[31]; Voth, 1998[8]; Allen and Weisdorf, 2011[6]; Humphries and Weisdorf, 2017[9]). It should be noted that the very high working times observed during this period often involved a different rhythm of work to what might be considered as normal today. The early 19th century in the United Kingdom saw the transition from the “workshop system” – where workers had substantial autonomy around their work time, being free to enter and leave work at their leisure, but often spending their entire working day with their family at their workplace – to the “factory system” in the early 19th century, with workers subject to “factory discipline” and having little control over work time, set times for meals, and fixed hours for starting and ending work.7
The high levels of work time led to workers’ demands for their reduction relatively early in the Industrial Revolution. A major concern at the time was the impact of long working hours on children and family life. As early as 1802, legislation was introduced in the United Kingdom restricting the working time of children serving as “parish apprenticeships”8 to 12 hours a day excluding breaks. In 1819, working hours for all children under 16 were restricted to 12 hours in cotton mills. In 1833, this was reduced to eight hours for children under 13, with a requirement that children attend two hours of schooling a day. In the 1840s, similar restrictions were placed on the working time of women. The result of these legislative changes was that by the end of the 1840s women and anyone under 18 could work no more than 12 hours per day. The introduction of restrictions on the working time of men however progressed much more slowly.
In other industrialising countries such as France, Germany, the United States and the Benelux, the process of industrialisation happened later and faster, as did the introduction of restrictions on working time, with many basic reforms coming in the 1890s and early 20th century (Huberman, 2012[22]).
By the start of World War I, manufacturing workers in most Western countries were working around 55-60 hours per week. However, immediately after the war, legislation introduced the 8-hour day across most of the Western world, leading to a sudden reduction of working time to around 48 hours per week. The 48-hour week persisted in most countries until World War II, with a few exceptions.9 By the early 1970s, most developed economies had shifted from the 6-day week to the 5-day week, resulting in a 40-hour week. Simultaneously, there was an increase in paid holidays (a pattern described in the section below on Holidays and annual working time). Since the early 1970s, the decline in hours has somewhat stalled, with full-time manufacturing workers generally continuing to work roughly around a 40-hour week, although the average working week for manufacturing workers for some countries in Western Europe declined by 1-2 hours. However, the growing numbers of workers working part-time or flexi-time makes the analysis of this decline hard to interpret. Uniquely in France, there has been a greater reduction in hours with the shift to a 35-hour week.
Outside of Western Europe and the Western Offshoots, the shift to a 48- or 45-hour week in the post-war period has also stalled, albeit at a higher level. The transition to a 40-hour week has been achieved in relatively few developing countries, and where it was achieved it did not persist for long. Indeed, in some developing countries not only has the reduction in hours stalled but over the last 20-30 years the average working week in manufacturing has increased in length. Conversely, other countries such as Ukraine and Moldova saw substantial reductions in hours of work in the late 1990s. This decline was presumably driven by the economic distress and poor economic performance experienced in those countries. The low value of 36 hours per week as the average working week for Eastern Europe and the former USSR for the 1990s in Table 3.4 below is partially explained by these developments.
Generally, the population-weighted regional averages shown in Table 3.4 tell the same story as above. For Western Europe, Eastern Europe and the former USSR and the Western Offshoots: a gradual decline until the introduction of the eight-hour day after World War I, the shift to a five-day week over the subsequent 40 years, followed by a lengthy period since the 1970s with little if any reduction in the average working week. In Latin America and the Caribbean, the average week seems to have stabilised earlier and at a slightly higher average level. In the rest of the developing world, average working hours in manufacturing appear to have persistently remained at higher levels than in the West.
Table 3.4. Average working week in manufacturing by region and decade, 1820-2010
Hours per week
Note: Square brackets indicate estimates based on a population coverage below 40%.
Of course, caution should be exercised when comparing these regional averages. As described above, the influence of Ukraine and Moldova in skewing the regional value for “Eastern Europe” should caution against overinterpreting these averages when they rest on a small number of observations. Special caution should also be paid with the estimates for sub-Saharan Africa, the Middle East and North Africa, and East Asia. As Table 3.5 shows, the number of observations for these regions is rather low. Data coverage for sub-Saharan Africa is especially poor. The average for the East Asia region includes only four countries, although one of these countries is China; unfortunately, data coverage for China is very poor with almost no observations before 1990.
However, while the regional data might be based on a small number of observations, they are normally quite reliable. And when considered along with the entire dataset presented in this chapter, which includes around 4 300 observations on working in manufacturing across over 120 countries or political units and over a period of nearly 200 years, we can be reasonably confident about these estimates. Considered together, this dataset allows us to draw some broad insights into the history of working time in manufacturing.
Table 3.5. Number of observations in working-time dataset by region and decade, 1800-2010
Holidays and annual working time
There are several reasons to be interested in extending the analysis of weekly working hours to annual working hours. First amongst these is that any consideration of the welfare implications of a reduction in working time would need to take into account the increase in the numbers of days of annual leave and holidays enjoyed by workers. Further, the levelling off in the reduction of weekly working time since the 1970s described above has been partially compensated for, at least in Western Europe, by a substantial increase in the number of days of leave and holidays over the same period.
Unfortunately, however, data on the number of days of leave and holidays enjoyed by workers is not widely available, especially for the entire panel of countries reviewed above. Therefore, this section relies on the data in Huberman and Minns (2007[1]) to describe the experience of a restricted panel of exclusively Western countries.10
Table 3.6 presents the number of days of leave and holidays enjoyed by workers in ten Western countries between 1870 and 2000. As can be seen, between 1870 and 1940 the number of days of leave and holidays increased in every country, with the 1920s seeing the emergence of paid leave. In the mid-20th century there was little change. Then in the later 20th century the number of days of leave and holidays enjoyed by workers in Western Europe increases substantially, while in the Western Offshoots there is almost no change in the number of days of leave and holidays, except for a small reduction for workers in Canada and the United States.
Table 3.6. Annual days of leave and holidays per year for manufacturing workers in selected countries 1870-2000
As described in Ward, Zinni and Marianna (2018[32]), the methods in use today to measure annual working time are not currently standardised. Our estimates of the number of annual working hours are based on the method currently used by Eurofound (2019[33]) and Eurostat (2018[34])11. These are only rough estimates of annual working time, as some significant components of annual working time are not considered, such as extra hours worked (i.e. overtime) and hours not worked due to sickness absences, maternity/paternity and parental leave, strikes and lock-outs, etc. These estimates are not comparable in any case with those made by the OECD for the entire economy. The method used here involves calculating the number of annual hours of work by multiplying the number of weekly hours times 52, and then subtracting the number of days of public holidays and paid annual leave, converted into hours under varying assumptions about the length of the working week across different periods.12
The resulting estimates for annual working hours of manufacturing workers are given in Table 3.7. According to this estimate, the number of hours worked per year in Western Europe continued to decline from 1980 to 2000, rather than plateauing as in the series of hours worked per week. In the Western Offshoots, on the other hand, the end to the reduction in weekly working time observable in Table 3.5 remains visible when looking at annual working hours.
Table 3.7. Annual working hours for manufacturing workers in selected countries, 1870-2000
Note: The calculation of the annual working hours figure for 1940 is based on data for days of leave and holidays in 1938.
Correlation with GDP per capita
We might expect that the relationship between GDP growth and working time would be a relatively simple one: as income increases, people will choose to work less, and therefore hours worked will decrease. This would imply a simple inverse linear relation between GDP and working time. However, the above description of developments of working time in manufacturing since 1820 does not support the view that working time has declined smoothly over time. Rather, there are both identifiable turning points and periods when, despite changes in GDP, there is relatively little change in the length of time that workers spend working.
Figure 3.2 presents the relation between working time and real GDP per capita across all countries included in our dataset. The data on real GDP per capita are taken from the Maddison Project (Bolt et al., 2018[35]). As can be seen, the relation is a quite tight non-linear relation.
Figure 3.2. Hours of work and GDP per capita since 1820
The figure suggests that hours worked decline quite rapidly with increases in GDP when GDP per capita is below USD 20 000. When GDP per capita is above that point, the average working week clusters quite closely around 40 hours per week regardless of the country’s level of income.
The concave positive relation between hours worked and GDP per capita visible in the bottom left corner of this figure reflects a small number of observations, all of which are from the period after 1970. These are primarily from countries such as Ethiopia, where the average working week in manufacturing in the 1990s and 2000s dropped to very low levels, which is presumably related to poor economic performance.
Figure 3.3 sheds further light on the historical dynamics of the relation between GDP per capita and weekly working hours in manufacturing. All blue dots are observations from before 1945, while all black dots are observations from after 1945. In both time periods, the relationship is almost linear, i.e. an increase in GDP per capita is associated with a reduction in working hours. But in the pre-1945 period, even a small increase in GDP per capita was associated with a substantial reduction in working time. After 1945, almost all the observations are clustered around 40 hours per week, implying that an increase in GDP is associated with only very small reductions in working hours. These figures hence provide some support to the idea that there are identifiable turning points in the relation between working time and GDP.
Figure 3.3. Hours of work and GDP per capita since 1820
Note: Data before 1945 are displayed in blue dots, those after 1945 in black dots.
Priorities for future research
While the data presented in the chapter represent a substantial improvement in our knowledge of long-run changes in working time, there are several ways in which this evidence could be improved in the future.
First, the data coverage for the early period of the dataset can be improved. It is highly likely that further national data series could be constructed for the early 20th century.
Second, this chapter has ignored both women’s working time in manufacturing and working time outside manufacturing. It is highly likely that working time across the non-agricultural economy has tracked manufacturing closely, but that does not mean that the development of working time outside manufacturing is not worthy of further study. There are still many large unanswered questions here.
Third, while data on working time in the average working week are very useful for understanding the burden of work experienced by workers, they do not provide a complete picture. Historical research would gain from an improved understanding of holidays, breaks and leisure time, from data on annual hours of work and from knowledge of work patterns and work intensity during hours of work.
Finally, this chapter has been highly descriptive, with a focus on what has happened since 1820. Based on the data presented here, further analytic work on the causes and consequences of changes in working time can and should be pursued in the future.
References
[6] Allen, R. and J. Weisdorf (2011), “Was there an ’industrious revolution’ before the industrial revolution? An empirical exercise for England, c. 1300-1830”, The Economic History Review, Vol. 64/3, pp. 715-729.
[30] Bienefeld, M. (1972), Working Hours in British Industry: An Economic History, London School of Economics and Political Science, Weidenfeld and Nicolson.
[35] Bolt, J. et al. (2018), “Rebasing ‘Maddison’: New income comparisons and the shape of long-run economic development”, GGDC Research Memorandum, Vol. 174.
[4] Clark, G. (2005), “The condition of the working class in England, 1209-2004”, The Journal of Political Economy, Vol. 113/6, pp. 1307-1340.
[36] Clark, G. (1994), “Factory discipline”, The Journal of Economic History, Vol. 54/1, pp. 128-163.
[31] Clark, G. and Y. Van der Werf (1998), “Work in progress? The industrious revolution”, The Journal of Economic History, Vol. 58/3, pp. 830-843.
[19] Crafts, N. (1997), “Economic growth in East Asia and Western Europe since 1950: Implications for living standards”, National Institute Economic Review, Vol. 4/97.
[2] de Jong, H. (2015), “Living standards in a modernizing world – A long-run perspective on material wellbeing and human development”, in Global Handbook of Quality of Life, Springer, Dordrecht.
[33] Eurofound (2019), Working time in 2017-2018, Publications Office of the European Union, https://op.europa.eu/en/publication-detail/-/publication/2d79f703-0a84-11ea-8c1f-01aa75ed71a1.
[34] Eurostat (2018), Quality issues regarding the measurement of working time with the Labour Force Survey (LFS), Publications Office of the European Union, Luxembourg, https://ec.europa.eu/eurostat/documents/7870049/8734970/KS-FT-17-005-EN-N.pdf/0eb7518a-d0ca-45c4-942e-49107724e09a.
[11] Gershuny, J. and O. Sullivan (2019), What We Really Do All Day: Insights from the Centre for Time Use Research., Pelican, UK.
[24] Gilmore (forthcoming), Working Time in Industry: A New Dataset.
[20] Hoffman, A. (1998), Latin American Economic Development. A Causal Analysis in Historical Perspective, GGDC monograph series, No.3, Groningen.
[22] Huberman, M. (2012), Odd Couple: International Trade and Labor Standards in History, Yale University Press, New Haven.
[21] Huberman, M. (2004), “Working Hours of the World Unite? New International Evidence of Work-time, 1870-1913”, The Journal of Economic History, Vol. 64/4, pp. 964-1001.
[1] Huberman, M. and C. Minns (2007), “The times they are not changin’: Days and hours of work in Old and New Worlds, 1870-2000”, Explorations in Economic History, Vol. 44/4, pp. 538-567.
[9] Humphries, J. and J. Weisdorf (2017), Unreal Wages? Real Income and Economic Growth in England, 1260-1850.
[12] International Labour Organization (2008), Measurement of working time, Report II, 18th International Conference of Labour Statisticians (Geneva), https://www.ilo.org/wcmsp5/groups/public/---dgreports/---stat/documents/publication/wcms_099576.pdf (accessed on 11 January 2020).
[13] International Labour Organization (1962), Resolution concerning statistics of hours of work, 10th International Conference of Labour Statisticians (Geneva), https://www.ilo.org/wcmsp5/groups/public/---dgreports/---stat/documents/normativeinstrument/wcms_112455.pdf (accessed on 11 January 2020).
[7] Isett, C. (2016), The social history of agriculture from the origins to the current crisis, Rowman & Littlefield.
[5] Le Goff, J. (1982), Time, work & culture in the Middle Ages, Chicago: The University of Chicago Press.
[17] Maddison, A. (2001), The World Economy: A Millennial Perspective, Development Centre Studies, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264189980-en.
[15] Maddison, A. (1995), Monitoring the World Economy, 1820-1992, OECD Development Centre, Paris, http://www.ggdc.net/maddison/Monitoring.shtml (accessed on 11 January 2020).
[16] Maddison, A. (1982), Phases of Capitalist Development, New York: Oxford University Press.
[14] Maddison, A. (1964), Economic Growth in the West, New York: Twentieth Century Fund.
[37] Mattesini, F. and B. Quintieri (2006), “Does a reduction in the length of the working week reduce unemployment? Some evidence from the Italian economy during the Great Depression”, Explorations in Economic History, Vol. 43/3, pp. 413-437.
[26] Mitchell, B. (2013), International Historical Statistics., Palgrave Macmillan Publishers Limited.
[38] Neumann, T., J. Taylor and P. Fishback (2013), “Fluctuations in weekly hours and total hours worked over the past 90 years and the importance of changes in federal policy toward job sharing”, National Bureau of Economic Research, Vol. w18816.
[39] Pollard, S. (1965), The Genesis of Modern Management: A Study of the Industrial Revolution in Great Britain, Penguin Books.
[29] Reid, D. (1976), “The decline of Saint Monday 1766-1876”, Past & Present, Vol. 71, pp. 76-101.
[3] Sahlins, M. (1972), Stone Age Economics, Chicago: Aldine-Atherton.
[27] Spoerer, M. and J. Streb (2008), “Leben, um zu arbeiten, oder arbeiten, um zu leben? Warum uns der Rückgang der Jahresarbeitszeit in den letzten 125 Jahren nicht beunruhigen sollte”, Geschichte Und Gesellschaft, Vol. 34/1, pp. 116-128.
[10] Stephenson, J. (2018), “Real wages? Contractors, workers, and pay in London building trades, 1650-1800”, The Economic History Review, Vol. 71/1, pp. 106-132.
[18] The Conference Board (2019), Total Economy Database™, https://www.conference-board.org/data/economydatabase/TED1.
[28] Thompson, E. (1967), “Time, work-discipline, and industrial capitalism”, Past & Present, Vol. 38, pp. 56-97.
[23] U.S. Department of Labor (1990), Fifteenth Annual Report of the Commissioner of Labor: Wages in Commercial Countries. 2 vols. Washington, DC: GPO..
[25] van Zanden, J. et al. (eds.) (2014), How Was Life?: Global Well-being since 1820, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264214262-en.
[8] Voth, H. (1998), “Time and Work in Eighteenth-Century London”, The Journal of Economic History, Vol. 58/1, pp. 29-58.
[32] Ward, A., M. Zinni and P. Marianna (2018), “International productivity gaps: Are labour input measures comparable?”, OECD Statistics Working Papers, No. 2018/12, OECD Publishing, Paris, https://dx.doi.org/10.1787/5b43c728-en.
Notes
← 1. Oisín Gilmore would like to thank Herman de Jong, Rick Veldkamp, Pedro Miguel, the editors of this volume and other participants in the OECD workshops on historical well-being held in Utrecht in September 2018 and in Paris in June 2019. This chapter was supported by the Institut für die Geschichte und Zukunft der Arbeit (IGZA) and the Netherlands Organization for Scientific Research (NWO).
← 2. Almost all the 120 political units are sovereign nations. However, a small number of other political units (such as the Saar Protectorate, West Berlin, the Palestinian State, etc.) are also covered by the database.
← 3. See Isett and Miller (2016, p. 70[7]), for a description of some of these differences.
← 4. It should be noted that while there is a broad consensus on this point, a number of researchers (Bienefeld, 1972[30]; Voth, 1998[8]; Allen and Weisdorf, 2011[6]; Humphries and Weisdorf, 2017[9]; Stephenson, 2018[10]; Clark and Van der Werf, 1998[31]) disagree.
← 5. All the figures in this chapter as well as in Table 3.4 and Table 3.5 are based on data for all 120 political units included in our database.
← 6. A very small number of observations were adjusted in other manners, as described in Gilmore (forthcoming[24]).
← 7. For a discussion of this transformation see Clark (1994[36]) and Pollard (1965[39]).
← 8. Parish apprenticeships were a means through which poor, illegitimate and orphaned children could be put to work and provided with some training in how to work. Generally, these “apprenticeships” were for very low status work.
← 9. The 40-hour work week was introduced in France in 1936. Around the same period, significant reductions of working time also occurred in Italy and the United States. See Mattesini and Quintieri (2006[37]); Neumann, Taylor and Fishback (2013[38]).
← 10. These figures are for the entire economy, not just manufacturing.
← 11. In Eurofound (2019[33]) the figure for weekly hours is normal hours not actual hours as is the case in the estimates presented in this chapter. Eurostat (2018[34]) discuss some aspects of the difference between these two figures.
← 12. For this calculation, days of leave have to be converted into annual hours of leave. This can be done using the following equation:
The Eurofound (2019[33]) and Eurostat (2018[34]) studies assume 5 working days per week. However, prior to World War I, manufacturing workers typically worked 6 working days per week, with the 5-day week introduced in most Western countries only after 1960. By 1980, the 5-day week was a well-established norm across Western Europe and the Western Offshoots (Gilmore, forthcoming[24]).Therefore, for the above calculation, I assume a working week of 6 days in 1870 and 1910, and of 5.5 days in 1940 and 1950.