WORKING TIME BEFORE 1813: MADDISON'S FIGURES

As a preliminary step in estimating employment and productivity measures for a group of OECD countries, Angus Maddison assembled an international data set on worktime that has become a point of reference in studies of convergence and in the new growth literature.¹⁰

With regard to the number of weeks worked, Maddison assumed that all countries before 1913 had the same number of public holidays as in 1950 and that there were no annual vacations in 1870. For 1913 he assumed one week's vacation across the board.¹²

At first pass, Table 1 appears to be an inappropriate starting point to compare labor market outcomes across countries. That said, can Maddison's underlying assumptions be justified? The answer is no. Since the publication of his estimates almost 40 years ago, labor and social historians have established the diversity of national work practices. Still, the relevant question is whether a new international worktime series would alter in any significant way our assessment of labor's experience and welfare before 1913. My answer is decidedly yes.

REVISED ESTIMATES OF WORKTIME: THE COMPONENTS

The Length of the Workweek

The Fifteenth Annual Report of the U.S. Department of Labor, under the supervision of Carroll Wright, published data on weekly workhours for the period 1850 to 1900.¹³

U.S. Department of Labor, Fifteenth Annual Report. Although the terminal year for non-U.S. data is in fact 1899, in the text and tables I use 1900. The Appendix gives a sample of the sources consulted in assembling the report. The citation in this paragraph is from page 12 of the report.

The project was far from modest. For 88 countries and territories, ranging from Algeria through Liechtenstein to Venezuela, the department consulted over 700 official publications. The coverage for the United States consisted of all the reports of the Federal and State departments of labor. The introduction to the report affirmed unequivocally that the “compilation may be considered exhaustive for the United States and nearly so for foreign countries.” Because the report included both manufacturing and nonmanufacturing sectors, it is more representative than series that rely on observations of a small sample of businesses, usually textile mills or mines, for shorter periods. The report distinguished as well between women's and men's workhours. Because of data limitations for the early years, I restrict myself to the period between 1870 and 1900, but there remain inevitable biases in the underlying sources. Many of the reports are based on census returns or surveys that ignored smaller firms. This would bias the results toward finding shorter workdays, because if the case of France was typical very small establishments such as workshops tended to have longer worktimes.¹⁴

I discuss the French case below.

The data for the Unites States pose another set of problems. They are based on retrospective information of a cross-section of firms that were in operation continuously throughout the period; concerns have been raised about their regional and industrial representation because of the different survival rates of firms across space and time. The direction of the bias is ambiguous, but some comfort can be taken from the research of Atack and Bateman who found very little evidence of regional differences in worktime; overall, the early reports used by the U.S. Department of Labor are in line with the more comprehensive estimates they culled from a representative sample of the 1880 Census of Manufacturing.¹⁵

Atack and Bateman, “How Long was the Workday,” pp. 803–04. I benefited from discussions with William Sundstrom on the reliability of the U.S. reports.

The department presented its results by individual occupations that I grouped into five categories or sectors defined in the Appendix. The classification serves two goals. Although heterogeneous in their composition, fixed capital investments varied across categories: iron and steel, and mining and construction being the most capital intensive; services and textiles the least. The occupational breakdown also reflected my objective to account for the effects of international forces on workhours. As designed, the textile sector comprises the most highly traded goods, followed by manufacturing, iron and steel, mining and construction, and services.¹⁶

The categorization by fixed capital and tradeables is meant to be suggestive. As defined, the manufacturing sector was heterogeneous in its capital stock and comprised some internationally traded goods. Coal and some iron and steel products were also traded internationally.

Table 2 gives some basic descriptive statistics of the data broken down by sex, sector, and country. The compilation is heavily weighted toward the United States, and, to a lesser extent, Great Britain. Of the 5,000 or so European observations, about 55 percent are from Britain and 15 percent from Germany and France each. The survey's breakdown by sector is roughly similar across countries (services account for between 5 and 8 percent of observations, and textile and manufacturing together about 45 to 50 percent). I also report mean and median workhours; they do not deviate greatly from each other. Where summer and winter hours were reported, I took the average of the two.¹⁷

Bienefeld (Working Hours, p. 142) surmised that in Britain winter hours were worked four months a year, implying that my procedure is biased against finding long hours.

The workday during the so-called second industrial revolution was strenuous, up to 12 or more hours per day, in most sectors and countries.¹⁸

Based on other accounts of the workday, I assume that reported hours were net of meal and rest times. Even if rest time has not been deducted entirely, there is a potential bias in the other direction. The report contains no mention of overtime hours. Contemporary sources indicate that at certain times of the year these may not have been negligible. Because of these opposing biases, I have left the initial series unadjusted. On these issues, see: Cross, Quest for Time, pp. 67, 234; Gårdlund, Industrialismens, p. 312; Heikkinen, Labour and the Market, pp. 154–56; Kaplow, “Saint-Lundi”; Pollard, “Labour in Britain,” pp. 157–58; Rist, La réglementation légale, pp. 177–88; and Rosenband, Papermaking, p. 108. Lars Svensson recommended the Swedish sources used in this article and kindly provided the translation.

Émile Vandeverlde, the respected Belgian social reformer, described the typical day in Brussels in the 1880s: Work started at 5:30 or 6:00, with an initial pause between 8:00 and 8:30 for le second déjeuner. Lunch at noon took up to two hours and there was another break at 4:00 PM. In summer the day ended as late as 8:00 PM.¹⁹

Vandervelde, Belgique ouvrière, p. 38. The workday in Sweden also began early and was punctuated by three or more pauses. Workers may have preferred to be in a warm factory than in their dark and cold homes. See Gårdlund, Industrialismens, pp. 312–14. On the implications of long workdays on sleep patterns, see Ekhirch, “Sleep.”

Some British factory workers had shorter worktimes on Saturday, but this practice was rare on the continent.²⁰

In 1850 Lancashire workers stopped at 2:00 PM on Saturdays; from 1875 until 1905, the day ended at 1:00 PM. Fowler, Lancashire Cotton, p. 65. These hour changes are consistent with the series reported in Table 3.

Contrary to the view that employers sought to stretch out the workday in more capital-intensive sectors, the work week was generally shorter in iron and steel, and mining and construction. Textile workhours were relatively long, giving some credence to the anxieties of contemporaries about international competition. There is no discernible difference in women's and men's worktime. Women who engaged in market work did so full-time.

The evaluation of time trends is not straightforward because observations by sector are spread unevenly across countries over the period. Figure 1 gives some idea of the unbalanced nature of the data set. For British manufacturing data are available annually, but observations for Belgian and French textiles are clustered at the end of the period, and those for German mining and construction in the early years. Calculating three- or five-year averages is one approach to solving the problem of missing observations. But this procedure is unsatisfactory because of the small number of observations in certain years. In effect, this assumes stationarity. I used the alternative strategy of fitting a time trend that controls for the likelihood of nonstationarity. Exploiting all the information contained in the reports, I estimated “fitted” values of workhours by country and by sector from 1870 to 1900. I separated Old and New World countries because specification tests indicated differences in the estimated coefficients in the two regions. Imposing restrictions improved the efficiency of the estimator.²¹

The division between regions meshes with contemporaries' statements. French textile workers voiced concern that the daily cross-border movement of Belgium workers thwarted their attempts to reduce workhours. Similarly, Canadian workers moved freely to the United States. As for Australia, it was connected, albeit indirectly, to North American labor-market developments because it was an alternative destination for British emigrants. On Belgian workers in France, see Strikwerda, “France and Belgium”; for North American labor-market integration, see Green, MacKinnon, and Minns, “Dominion or Republic”; on the New World connection, see Hatton and Williamson, Mass Migration, p. 54.

For the two groups of countries, I experimented with a quadratic trend but this proved to be insignificant.

HOURS OF WORK PER WEEK: ACTUAL AND PREDICTED VALUES

More formally, for Old and New Worlds individual observations on weekly workhours in country i, sector j, and in year k, were regressed on indicator variables for sex, country, sector, and year, and three sets of interaction terms, country and sector, country and year, and sector and year. I experimented with interaction terms for sex in order to derive women's workhours by sector and year, but there was very little difference in the trend and levels of worktime for women and men.²²

To illustrate, panel b in Figure 1 gives the fitted values for female and male workers in French textiles based on a regression with interaction terms between sex and time, sex and country, and sex and sector. The finding of little difference between women's and men's worktime implies that legislation intended to constrain the hours worked by the former also limited the latter's. This gives support to the frequent claim that men fought the battle for a shorter day “from behind women's petticoats.”

For ease of computation I omitted “sex” in estimating the fitted values used in this article. The fitted values used in this article thus correspond to male workers. In estimating Old World hours, I combined observations for Denmark and Sweden to get around the problem of small sample size. The New World regression contains indicator variables for U.S. regions that control to some extent for the problem of geographic representation alluded to earlier.²³

To arrive at a U.S. national average, the regions were weighted by their share of the population in 1900. Source: United States, Historical Statistics, A195, pp. 24–36.

Figure 1 traces the fitted values from this procedure for some sectors for certain countries. The estimated values for Belgium textiles follow closely the average workhours F.-X. Van Houtte estimated for a sample of Ghent cotton textile firms, whereas the estimated values for German mining and construction correspond with the figures Walter Hoffmann gave for certain years.²⁴

For Belgium, see also Scholliers, Wages, Manufacturers, p. 130.

Table 3 compares changes in sectoral workhours across countries. From their initially high levels, textile hours declined rapidly, while service sector hours increased in some countries. The movement in mining and construction (as it was for iron and steel and manufacturing) hours was moderate. I will say more about sectoral patterns later in the article.

I next estimated the average length of the workweek by country on an annual basis using the fitted sectoral values. The employment weights used in this procedure are given in the Appendix; to save space Table 4 reports the actual values at five-year intervals. As calculated, the national series reflect the importance of the service sector in most economies in the sample, but, all told, differences between the weighted and unweighted national averages are not very great. Independent sources confirm my calculations. Although the Italian numbers are slightly higher than reported elsewhere, for eight countries (Australia, Belgium, Canada, France, Germany, Great Britain, the Netherlands, and the United States) the figures in the table correspond closely with other, albeit scattered estimates.²⁵

Australia: ILO, “Labour in Australia,” p. 1020; Belgium: Neuville, Évolution des relations industrielles, pp. 61–87; Rowntree, Diminuer la misère, pp. 574–76. Canada: Altman, “New Estimates,” p. 361; France: Marchand and Thélot, Travail en France, p. 145; Germany: Hoffmann, Wachstum der deutschen Wirtschaft, pp. 213–14; Great Britain: Bienefeld, Working Hours; Matthews, Feinstein, and Odling-Smee, British Growth, pp. 64–70, 566; Italy: Zamagni, History of Italy, pp. 201–05; Netherlands: Wintle, History of the Netherlands, p. 66; United States: Atack and Bateman, “How Long was the Workday,” p. 139; and Costa, “Wage and Length,” p. 58.

Scandinavian hours merit some reservation. They are an amalgam of Swedish hours principally from the 1870s, and Danish figures mainly for the 1890s. Still, the overall trend is similar to that found in a study by J. P. Christensen for Denmark.²⁶

Christensen, Lønudviklingen inden, p. 71. I would like to thank Gunnar Persson for helping with the translation. Christensen's figures are slightly less than those of a contemporary study of Copenhagen in the early 1890s that reported “89 percent of all factories had a working day of ten-and-a-half hours or less, whereas in the provinces a large number of establishments visited worked 11 hours or more.” UK PP 1893–94 vol. 39, Part 2, p. 780. For Sweden, ibid., pp. 778–79; and Tegle, “Den Ordinarie Veckoarsbetstiden,” p. 11.

The last column of Table 4 presents estimates for workhours in 1913 based on projections from the fitted values for 1870–1900, or, wherever possible, from independent estimates.²⁷

I prefer to use independent estimates (in italics) for 1913 wherever possible because the introduction of hours legislation in some countries after 1900 makes it hazardous to extrapolate from 1870–1900 time trends. That said, the predicted values for France, Great Britain, and the United States are very close to the independent estimates I use.

Trends in weekly workhours varied considerably across countries. Australia, Belgium, Germany, Ireland, Scandinavia, Spain, and Switzerland exhibited contractions in hours of more than 10 percent; France, the Netherlands, and the United States had more modest declines; Britain and Italy showed no trend, while for Canada time at work actually increased over the period. Comparisons between and within Old and New Worlds are ambiguous. On the one hand, the gap between Australia, “a sellers' labor market” and Italy, where the “elasticity of unskilled labor was very high, if not infinite,” widened over the period.²⁸

The quote for Australia is from Butlin, Australian Economic Development, p. 391; for Italy, Cohen and Federico, Italian Economy, p. 61.

On the other hand, North Americans labored as long as northern Europeans by the turn of the century. In subsequent sections of the paper I discuss in more detail these trends, but it would not be presumptuous to conclude that, although Maddis on's assumption of stable hours for Britain was appropriate, his inference that all countries followed this pattern does not stand up to scrutiny.

The Length of the Work Year

In 1870 the British trade office began collecting information on working conditions around the world in order to verify manufacturers' claims that longer hours abroad were a source of competitive advantage. The reports of the consular delegates, submitted under various guises into the early twentieth century, provide enough information to generate the series of workdays in 1870 and 1890 found in columns 1 and 2 of Table 5.²⁹

United Kingdom, PP 1870 vol. 66, PP 1871 vol. 68, PP 1872 vol. 62. The consular reports were superseded by the Abstract of Foreign Labour Statistics and the Commission on Labour, Foreign Reports, published occasionally before the First World War. PP 1893–94 vol. 39, PP 1899 vol. 92, PP 1901 vol. 73, PP 1906 vol. 113. Jeffrey Williamson suggested this source.

A limitation of the reports is that they did not cover all sectors of activity; most often the delegates cited days of work of textile workers or day laborers. The numbers assume no Sunday work and refer to full-time male production workers. No doubt seasonal work was still present, but by most accounts it was declining.³⁰

On the decline of seasonal work patterns for the United States, see Atack, Bateman, and Margo, “Part-Year Operation.” They report that in 1870 and 1880 a majority of establishments operated on a full-time basis and speculate that the number was increasing; nonetheless, part-time operation was still common. For France, see Bompard, Magnac, and Postel-Vinay, “Emploi, mobilité.”

Despite these shortcomings, independent estimates confirm the figures I have collected from the consular reports.³¹

In their calculation of worktime in the United States, Atack, Bateman, and Margo (“Productivity in Manufacturing,” p. 172) used 309 days of work; Rosenbloom (“Male Labour Supply”) assumed 306 days of work in 1880. The International Labour Organization began to collect data on days of work in the 1920s that are consistent with the evidence from the consular reports for 1890. ILO, “Legislation on Annual Holidays.”

The length of the work year was much longer in the New World. In some regions of Europe religious festivals were still common. For Russia, the consular delegate observed:

… [T]he number of festivals on which work is prohibited in all factories is 19, but in most cases 27 or 28 festivals are kept as holidays during the year. Thus the average number of holidays, including Sunday is 80 or even 100 days, so only 250 or 260 days are spent at work…. In Poland the number of working days is further reduced to 230 on an average, because in Russia religious festivals are observed according to the Julian calendar, while in Poland they are celebrated twice over, according to both the Gregorian and Julian calendars. Thus a Russian does at least 70 days less work during a year than a Frenchman in a similar position.³²

United Kingdom, PP 1893–94 vol. 39, Part 2, p. 864.

Protestant Northern Europe and Great Britain had fewer religious holidays. For the Netherlands, the correspondent found that Catholics celebrated seven holy days, whereas Protestant workers had only three days off in 1890.³³

United Kingdom, PP 1893–94 vol. 39, Part 1, pp. 12–13. I took the average value in Table 5.

Still, Lancashire textile workers had almost doubled their vacation days from 1870 until the War.³⁴

United Kingdom, PP 1909 vol. 80, p. 91. Fowler, Lancashire Cotton, pp. 63–65. Ireland seems to have kept pace. O'Grada, Ireland, p. 239.

On the continent, workers succeeded in refashioning traditional festivals or Saints' days into secular holidays. In Belgium, the introduction of May Day in 1895 in the iron and steel trades saw the end of the traditional holiday of St. Eloi, the patron saint of metal workers.³⁵

Strikwerda, House Divided, p. 262. For the “religious” origins of May Day elsewhere, see Panaccione, Memory of May Day, p. 15. In the Netherlands, Catholic and Protestant unions organized separate May Day celebrations. Harmsen and Karsten, “Le premier mai.”

There was a stronger preference to work more days in the New World. As in Europe, North American workers observed Saint Monday rituals, but Americans labored under different rules. Roy Rosenzweig found that American workers had just three to four official days off around 1870; Labor Day was the only additional holiday granted before World War I, a pattern repeated in Australia.³⁶

Rosenzweig, Eight Hours, pp. 69–70; and Gutman, “Work Culture,” pp. 547, 570. In Australia, national public holidays replaced those of individual colonies. Separation Day, formerly a Victorian public service holiday, became a “mere bank holiday” at the end of the century. Kingston, History of Australia, p. 184.

Borders were important. Biographies of French Canadian migrants to the textile factories of New England relate how they were unaccustomed to work on New Years' Day. As ever, Emerson's summarized most eloquently—and accurately—the American work ethic: “Works and days were offered us, and we chose work.”³⁷

Rodgers (Work Ethic, pp. 7–17) observed that the central premise of the American model was that “work was the core of moral life.”

Atack and his coauthors have recently studied the relationship between days at work and workhours per week for the Unites States. Because of diminishing returns using the technology in place, they argue that it was cost efficient for an employer to shorten the length of the workweek, but extend the work year. Although my data do not permit a thorough analysis of this relation for other countries, there does appear to be an inverse relationship between days at work and the workweek across the sample. Relative to Belgians, Canadians worked fewer hours per week but more days. It is not clear, however, that employers alone initiated adjustments along these margins, because if they had concerns about the diminishing returns of the long workweek they would have as easily reduced the length of breaks and pauses during the workday. It was the rare continental employer who adopted the “English working day” of one-half hour for lunch.³⁸

Milhaud, “Unbroken Working Day,” p. 796.

Supply-side factors may have been involved in the choice between days off and the length of the workweek.³⁹

My argument borrows from Hamermesh's (Workdays, Workhours) study of the choice between workdays and workweeks in the United States and Germany in the 1990s. The existence of fixed costs along a particular dimension can produce “the unusual result of large jumps in labor supply along that dimension when the worker's wage rate changes (p. 14).” Among other factors, Hamermesh identified changes in commuting times that alter the tradeoff between working fewer days or a shorter workday.

The pattern for Belgium is consistent with workers' commuting patterns, many of whom traveled up to 50 kilometers to their place of employment as early as the 1880s.⁴⁰

In 1900 Belgium had the most extensive network in Europe with over 325,000 commuters daily, or 10 percent of its labor force. See Mahaim, Abonnements ouvriers, pp. 204–59; Rowntree, Diminuer la mis ère, pp. 291–302; Polasky, “Transplanting.” On walking distances in Canada, see Lewis, Manufacturing Montreal, p. 146; for the United States, Liepmann, Journey to Work, pp. 64–65.

A Belgian may have preferred to stay home on more days than a Canadian who lived in walking or cycling distance to his factory. The implication is that any reduction in the time of travel (including fewer days of work) must have increased worker welfare that is not captured in conventional measures of GDP, a theme to which I return below.⁴¹

GDP calculations treat commuting times as leisure. For a discussion, see Mokyr, Gifts of Athena, p. 156.

From these estimates of days worked, I deducted the number of days absent due to illness, using information from the U.S. Department of Labor's study of health insurance.⁴²

U.S. Department of Labor, Twenty-Fourth Annual Report and Workmen's Compensation Laws. I would like to thank John Murray for recommending these sources. See also, Varlez, “Assurance,” p. 8.

I have not made any deductions for average days lost due to unemployment or for work stoppages. Although some national series exist for these measures, further research is needed to make them comparable across countries. The resulting series in columns 4 and 5 in Table 5, therefore, represent full-time weeks of work in 1870 and 1890. Compared to Maddison's approximations, the new figures amount to between one and one and a half fewer workweeks.

ANNUAL WORKTIME, 1870–1913

Table 6 presents estimates of annual worktime for 1870 to 1913 with comparative data for 2000. These figures can be thought of as the average annual contracted hours of work of a male employee.⁴³

After 1870 the movement in worktime was anything but uniform. Certain European countries had caught-up to British levels by the end of the period. In 1870 Danes toiled almost 700 hours more than the British, but in 1913 the gap had collapsed to 70 hours. The decline in Britain was restricted almost exclusively to more days off, but northern and some western European countries benefited from shorter and fewer days. As a result, the variance (coefficient of variation) of annual hours for the set of European countries falls over the period, although much of the change seems have taken place after 1900 when national authorities became more active in legislating the length of the workweek. The narrowing of worktime between the Old World and North America was even more dramatic. By the end of the period, the United States was actually working longer than eight European states, thus giving a historical dimension to the claim of the “overworked American.” But the gap between Australia and Italy actually widened from 200 to 600 hours. Because of these offsetting examples, the variance for the entire set of countries was roughly stable between 1870 and the war; remarkably it was the same order of magnitude as in 2000. All together, the raw numbers suggest that, although rankings may have altered and the continental gap had disappeared by 1913, national work practices resisted any tendency toward convergence despite strong forces of economic integration.

Table 7 gives another perspective on the significance of the new worktime estimates for evaluating labor market outcomes during the late nineteenth century. Maddison's assumption of common and stable worktimes across countries implies that movements in wages or per capita income capture all adjustments in well-being. But a different picture emerges if the assumption of stable workhours is dropped. It is well known that GDP per capita underestimates workers' welfare because it does not include gains to workers owing to increased leisure time. Fewer hours worked by the employed implies greater leisure, which presumably adds to workers' utility. This is not true for the unemployed or for those working fewer hours than they would like to. Still, standard neoclassical analysis suggests that adults in countries with fewer hours worked will be better off relative to those in countries with more hours worked at the same level of GDP per capita. To get an idea of how much better, I follow a standard approach and measure leisure-augmented income in period tas the sum of income in period tand the difference in hours worked between tand the base year 0, valued at the wage rate in t.⁴⁶

These findings bring a new perspective to debates about the impact of globalization on worker welfare. The closing of the continental gap in worktime sits well with Williamson's view of the benefits of global market integration. As incomes rose in the relatively labor-abundant Old World, its worktime fell faster than in the labor-scarce New World. Williamson may have in fact underestimated the depth of the global labor market, because continental movements in worktime meant that broader measures of worker welfare that include leisure converged as well. The fit between openness and changes in hours lends further credence to the received wisdom on the positive gains of globalization. Recall from Table 3 that textiles, the sector comprising the most competitively traded goods, had the deepest cuts in worktime over the period. This result is also found at the more aggregate level. Figure 2 traces the relation between reductions in worktime from 1870 to 1913 and trade shares in 1913, defined as the value of exports and imports divided by GDP. Small open economies, such as Belgium, Switzerland, Denmark, and Sweden, exhibited large contractions in annual hours from 1870 to 1913, but in protectionist Canada, France, and Italy worktime hardly changed.⁴⁹

TRADE AND HOURS, 1870–1913

THE RELATION BETWEEN INCOME AND HOURS ACROSS COUNTRIES

Albert Rees, in his classic textbook, asserted that the decision to labor 72 hours per week “in the early days of the industrial revolution” was neither implausible nor impractical. At very low real wages, “with habituation and a less intense pace of work, output could continue to rise with increases in the workweek well beyond any number of hours we could now regard as reasonable.”⁵⁰

Figure 3 shows an inverse relation between hours and income for a number of countries between 1870 and 1913. The relation was anything but uniform, however. Countries worked different hours at the same levels of income. The evidence offers some support to Hans-Joachim Voth's claim that late industrializers worked fewer hours than Britain at comparable stages of development because of the gradual substitution of human capital for labor time.⁵²

ANNUAL WORKHOURS AND GDP, 1830–1913

Table 8 reports results of regressions of hours on income by country. For this purpose, I used the observations on weekly hours from the U.S. Department of Labor Survey multiplied by estimates of the number of weeks worked (Tables 2 and 5). The dependent variable is then annual hours for the set of observations from the department's survey. This setup permits controls for sector and sex. The value for per capita income corresponds to the year of the worktime observation. The estimated coefficients—the income elasticity of the demand for leisure—indicate substantial differences in behavior across the sample of countries.⁵³

THREE WORLDS OF WORKTIME

The relation between hours and income varied across the sample because counties worked by different rules. Workhours are a public good and when individual workers wanted to exchange some of their higher earnings for more leisure, they needed a collective voice to negotiate on their behalf. But at the most fundamental level, the nature and form of collective-decision differed across the sample of countries.⁵⁵

The failure of hours to contract as incomes rose in Britain has puzzled historians because unions were relatively strong and had participated in earlier rounds of worktime legislation. But after the establishment of the nine-hour workday in selected sectors in the early 1870s, British unions chose not to press for further reductions in worktime.⁵⁶

French unions were clearly weaker than those in Britain, but the pattern in worktime in the hexagon must be sought elsewhere. Not only did hours remain long, but the dispersion of French hours across sectors was generally greater than that found in Britain, excluding services. Small- and medium-sized enterprises, which dominated French industry, had a tendency to work long and often irregular hours compared to larger units.⁶²

Compared to the British and French cases, regulation was the key to Belgium's success in reducing worktime. The country in 1870 was in Marx's phrase a “capitalist's paradise,” and in Vandervelde's eyes a “workers' purgatory.”⁶⁷

The form, timing, and success of collective-decision making varied across and within Old and New Worlds. The Belgium model of regulation echoed earlier changes in Switzerland and predated those in the Netherlands. Spaniards appear to have reduced the workweek in the absence of any substantive legislation, whereas in Scandinavia the decline in hours preceded regulation. Key textile unions in Barcelona may have led the way to a shorter hours in Iberia, and in Denmark collective-decision making may have been facilitated, to borrow a phrase from Kindleberger, by its celebrated “group behavior.”⁷¹