Time for Things Labor, Leisure, and the Rise of Mass Consumption

THREE

Empirical Pattern in the United States

The last chapter presented in very general terms one of the most puzzling aspects of capitalism in the age of mass production and consumption: its characteristic combination of increasing productive power, ever growing output, and increasing consumption, at fixed or only slowly decreasing levels of work time. There is scant evidence that this combination makes much sense from the point of view of human well-being. The increasing disconnect between wealth and well-being is what Avner Ofer (2006) (among many others) has described as the “paradox of afluence.” The aim of this chapter is to fill out the bare-bones account introduced in Chapters 1 and 2 by analyzing large-scale developments in the economic history of the United States in the twentieth century. The United States is often taken to be the paradigmatic case of Fordist mass consumer driven capitalism and, thus, works well as a case study.

The history of the American economy between the late nineteenth and late twentieth centuries follows quite closely the general pattern outlined in the previous chapter. The economy experienced steady increases in productivity over this period. Working hours fell precipitously between the 1880s and the late 1920s, but thereafter, the rate of decrease slowed significantly and from 1940 onward further reduction in work time was very moderate indeed. As the century wore on, improvements in hourly productivity were increasingly converted into higher levels of output and (at least up until the 1980s) rising real wages at a more or less constant input of work time per full-time employee. Rising wages were for the most part funneled into increasing consumer spending, a significant part of which went to fund the acquisition of

consumer durables of various sorts. Indeed, Aglietta (2008) has argued that the demand for consumer durables formed a crucial stabilizing component of the Fordist mode of regulation. Drawing on empirical work in economics and economic history, and using data from the Bureau of Labor Statistics, as well as census materials collated in The Historical Statistics of the United States, the present chapter examines this pattern in some detail, analyzing trends in productivity, wages, work time, and consumption over the course of the twentieth century. Getting a clear picture of these trends and their interrelations requires some auxiliary analysis of savings rates, credit, changes in retirement, life span, time spent in education, and the efect of the deployment of labor-saving devices in the domestic sphere. In order to properly contextualize developments in the twentieth century, it is also helpful to look further back in time at the history of work, so the chapter touches in passing on patterns in earlier periods.

Before pressing forward with this analysis, some attention ought to be paid to the view that there is nothing deeply puzzling about the relation between work time, hourly wages, and productivity that emerged in the United States during the twentieth century. Gary Becker (1965) sets out this position clearly in an influential paper on time allocation, written in the mid-1960s. He takes it as given that consumers are guided in their decisions by an ongoing concern to optimize utility under changing conditions. Changes in work time are, according to Becker, simply expressive of the shifting balance in consumer behavior between the income efect, where higher hourly earnings lead to increased preference for leisure time, and the substitution efect, where the added utility of higher hourly earnings leads to a preference for more work. Becker acknowledges that in the US economy, the income efect ebbs after 1940, but he simply assumes that this must have been because, for rational economic reasons, the substitution efect increased in strength.1 Yet accounting for the change in that way involves circular reasoning. The argument has the following form: the decline in hours slowed to a creep because higher earnings yielded relatively more satisfaction. How do we know this? Because the decline in hours slowed to a creep. Becker simply takes it as self-evident that the decline in the rate of decrease in work time from 1940 onward reflects rational, optimizing choices on the part of individuals. He does not even attempt to provide historical evidence to substantiate this interpretation. My theoretical analysis sticks more closely to the rough ground of history, which seems entirely appropriate when analyzing historical trends.

Empirical Pattern in the United States

43

The chapter is organized into a number of sections, each of which deals, in an abbreviated fashion, with a particular set of developments. The sections examine productivity, wages, work time, and consumer spending, as well as saving. At the same time as giving a brief overview of developments in each of these areas, I consider various possible explanations for the overall pattern of the set of developments taken together.

1. Productivity

Measuring productivity is a complex and technical matter, and there is much debate about methods and disagreement about exact numbers. Assessing changes in productivity over long stretches of historical time presents particular dificulties. Historical data on productivity get increasingly sketchy the further back in time we venture, and longer-term estimates of change in productivity by necessity involve creative extrapolation from often very imperfect indices. Constructing a picture of change over long periods of time usually also involves stitching together data from disparate and discontinuous archival sources. In addition, there are formidable conceptual problems with measuring productivity. While comparing quantities of material output per unit of input over time is, at least in theory, quite straightforward, it is very dificult when calculating productivity to take into account change in product quality.2 Furthermore the productivity of economic activity that does not produce a material output—knowledge production or the provision of services for example—is extremely hard to measure.

Despite these methodological and conceptual dificulties, it is clear that over the course of the twentieth century advanced capitalist economies experienced very steady improvements in productivity, which over the long term compounded into quite dramatic economic growth. For the United States, the rate of increase of multifactor productivity per year between 1929 and 2007 has recently been estimated to have been an adjusted average of 2.17 percent (Gordon 2010).3 The most significant increases in multifactor productivity took place between 1928 and 1950, described by economists as “one big wave” of productivity improvement (Gordon 1999). Labor output per hour increased by an average of well over 2 percent each year between 1900 and 2000, with a second big wave of improvement evident from 1940 to 1973 (Gullickson and Harper 1987). The trend of rising productivity meant that by 2007 each hour of labor produced 5.33 times more output than in 1929.4

44

Time for Things

There are reasons to think that the standard productivity figures substantially underestimate the real increase in productivity. The productivity figures, averaged across all economic activity, include productivity in the services sector, which economists have argued tends to be underestimated because of dificulties of measurement (Griliches 1992).5 As the US economy matured, there was a shift in the labor force, first from agriculture to industry and subsequently from industry to services.6 The proportion of the workforce employed in the service sector jumped from 38 percent in 1900, to 78 percent in 1999 (Fisk 2001). As the proportion of the population employed in the service sector increased over time, services took on relatively more weight in overall productivity calculations. If service sector productivity is systematically underestimated, and the service economy grew relative to other sectors as the economy developed, then it is probable that the true magnitude of the improvement in productivity over the course of the twentieth century is significantly greater than the oficial statistics suggest.

In addition, the standard productivity calculation fails to take into account suficiently the additional value generated by the creation of entirely new kinds of product. When new consumer goods get invented—things like personal computers and mobile phones—sources of utility come into existence that did not previously exist. Because national income accountants have dificulty valuing this added utility, it does not get registered in standard measures of changing productivity. Economist Bradford DeLong (2002) has come up with a very rough estimate for the added value of new products, suggesting that with this added value taken into account productivity in fact increased by between 14 and 25 times over the course of the twentieth century.7 It is not unreasonable, therefore, to conclude that the standard measures of labor and multifactor productivity should be regarded as providing a bare minimum estimate of changes in productivity. The real figures are likely to be higher, quite probably much higher.

In sum, there is little doubt that between the later nineteenth and early twentieth centuries, the economy of the United States, along with every other economy in the industrial world, experienced dramatic growth in productivity. The fact that per capita economic growth and improving productivity tended to move in tandem indicates that productivity improvements were predominantly used to increase aggregate output.

Empirical Pattern in the United States

45

1. Wages

Of course, raw productivity figures tell us little about the distribution of the benefits conferred by per capita economic growth. Increased productivity can be realized either as higher profits or as increased hourly real wages. The general pattern for the twentieth century is that at least some part of the increase in wealth generated by productivity improvements has been captured by wage earners, although the degree to which that was the case varied over the course of the century, and also between diferent parts of the economy. In the United States, real wages per hour in manufacturing increased on average by 1.43 percent per annum from 1900 to 1929, by 2.35 percent from 1948 to 1973, and by 0.46 percent after 1973 (Goldin 2000). The average hourly wage in 1909 was \$3.80 (in 1999 dollars). By 1999, it had risen to \$13.90. However, this underestimates the real increase in compensation because it does not take into account the added value of benefits, which barely existed in the early part of the century. Benefits as a fraction of compensation continued to rise over the course of the twentieth century. By 1999, they amounted to a further \$5.58 per hour. With benefits taken into account then, hourly compensation for manufacturing work in 1999 was \$19.48—a fivefold increase in real terms since 1909 (Fisk 2001).8 For all work, the average hourly wage, including benefits, was \$21.16 in 2000 (Moehrle 2001).

It should be noted then that the tendency for improvements in productivity to be turned into increases in wage earnings greatly diminished toward the end of the twentieth century. Economists Lawrence Mishel and Heidi Shierholz (2011) found that between 1979 and 2009, while productivity increased by 80 percent, the hourly wage of the median American worker increased by only 10.1 percent, and most of this increase took place during the economic recovery between 1996 and 2002. From the late 1970s onward, growth in productivity and wages evidently became quite disconnected. In the last quarter of the twentieth century, the owners of capital captured the lion’s share of the increase in productivity. By contrast, during the Golden Age period of American capitalism, between the Second World War and the mid-1970s, real income per annum increased at a rate quite close to increasing labor productivity—2.35 percent for wages and 2.77 percent for labor productivity.

That productivity should be linked to increasing real wages is in line with basic economic theory. In the neoclassical theory of the production function the worker is a paid a wage equivalent to her added marginal product. As the productivity of labor improves, the theory predicts that wages should increase commensurately. The argument is quite straightforward. If workers’ wages are lower than their marginal product, then firms would gain by taking on additional workers, which increases demand for labor, leading to a rise in wages. If, on the other hand, workers’ wages are higher than their marginal product, then the firm gains by shedding workers, increasing the supply of labor, which puts downward pressure on wages.9 So at equilibrium, wages are set by marginal added product. In reality, of course, wage rates are determined not just by economic rationality narrowly conceived but also by a complex of sociocultural and institutional forces. Regardless of whether the theory itself is correct, however, the connection between wage levels and productivity under mid-century Fordist industrial capitalism became an institutional fact. In the postwar period, manufacturing wage levels were to a significant extent set by collective bargaining arrangements underwritten by the state (Zeitlin 1987). As the institutional and normative framework underpinning Fordism fell apart, so too did the linking of real wages to productivity.

46

Time for Things

Thus, for most of the century increasing productivity was at least to some degree turned into increasing real hourly wages. For the present inquiry, the question then arises of the extent to which increases in hourly wages were used to reduce work time as opposed to increasing real income.

1. Work Time

An increase in hourly wages, at least in theory, ofers wage earners a choice between either increasing their income for the same number of hours of work, or reducing hours for the same income. It is quite possible that increases in the average household income have a ratcheting efect on the normative standard of living, with the result that wage earners become very resistant to subsequent falls in income. However, it is not clear just why an increase in hourly wages should be used to increase household income rather than to reduce time at work at constant income. There are good reasons, discussed in the previous chapter, for thinking that as income rises, the relative value placed on free time should also increase. It follows from this that at income levels above those required for basic subsistence, an increasing proportion of marginal increases in hourly wages should be used to reduce the amount of time spent at work.

In line with this expectation, work hours did indeed fall as wages increased over the course of the twentieth century. Robert Fogel notes that work hours in the United States decreased by 50 percent between 1880 and 1995 (Fogel 2000). Hours for nonfarm workers dropped from 60 per week in 1900 to

Empirical Pattern in the United States

47

39.3 in 1990 (Lebergott 1996). Average hours per week for manufacturing workers dropped from 53 in 1900 to just over 40 at the turn of the twenty-first century.10 The National Bureau of Labor Statistics population survey found that in 2011 the average workweek for all workers was 38.3 hours, while the workweek for those who usually work full-time was 42.4 hours (for nonagricultural work, the numbers are 38.2 and 42.3, respectively). However, most of the decline in work hours since the late nineteenth century occurred before 1940. Economic historian Robert Whaples (2001), in an overview of historical research on work time in the twentieth century, notes that, although there are a variety of series measuring the workweek over time, using inconsistent measurements, they all show more or less the same pattern, “with weekly hours falling considerably during the first third of the century and much more slowly thereafter.” The length of the working week fell precipitously between 1900 and 1920, from about 60 hours to about 50 hours (a full workday).11 It subsequently continued to fall but at a significantly reduced rate, such that between 1920 and 1929, it dropped only a further 2 hours, to 48 hours per week. William Sundstrom’s (2006) analysis finds that the length of the workweek fell an average of 3.4 hours per decade from 1900 to 1950, thereafter declining only very slightly until 1980, after which it increased again to what it had been in 1950. By 1940, the full-time workweek had more or less stabilized at around 40 hours per week, and although it continued to decrease for the rest of the century, the rate of decrease became very gradual. While some recent analysts disagree with Juliet Schor’s (1991) contention that work hours actually increased toward the end of the twentieth century, the evidence shows that since the Second World War, hours have not fallen by very much, if at all (Lebergott 1996).12 Moreover, a significant number of Americans have always worked two or more jobs—according to one study, as much as 6 percent of the population (Owen 1988). This means that the average length of the workweek per job underestimates the average amount of time spent at work for each working individual.13 There has been some increase in vacation days per year, which obviously has the efect of reducing the average amount of time spent at work each week. However, economist John Owen (1988) estimates that even when vacation is taken into account, the reduction in the working week for nonstudent males between 1950 and 1986 was a mere twenty-four minutes per week (from 39.9 to 39.5 hours). Owen makes the point that nonstudent men are a good measure of underlying trends in work time because, unlike students and women, they are less likely to have other responsibilities. A decline for that population of twentyfour minutes per week between 1950 and 1986 is very small relative to the impressive increases in productivity during those years. Over the century as a whole, there was an increase in output per hour of labor of approximately 850  percent, while during the same period the working week declined by only 30 percent. Increases in labor productivity were most significant during the “big wave” period, between 1940 and 1973 (Gordon 2010). As noted, work time did fall rapidly between 1900 and 1940, but from 1940 onward, further decline in the length of the working week was minimal, despite continued increases in productivity. Across all sectors of the economy, annual work hours in the United States declined by about 10 percent between 1950 and 2014, while in the same period productivity increased by 350 percent.14

48

Time for Things

The oddness of this pattern is particularly evident if we narrow our focus to manufacturing. It was the organized workers of that sector who, in the later nineteenth and early twentieth centuries, pushed for a reduction in work time, first militating for the ten-hour day and then for the eight-hour day. Examining the change in average full-time manufacturing hours per week across the entire century we see a drop from fifty-three hours in 1900 to forty-two hours in 1999—a 20 percent reduction (Fisk 2001). But once again, almost of all of this reduction in hours took place between 1900 and 1940. Between 1950 and 1999 manufacturing labor productivity increased by an average of 2.9 percent per year, for a total increase of 417 percent (Cobet and Wilson 2002). Yet over this period the length of the working week in manufacturing barely budged at all. There was thus a marked deceleration in the rate of decline of manufacturing working hours in the postwar years at the same time that there was a pronounced acceleration in both manufacturing productivity and hourly earnings.

The extent to which increasing productivity was converted into more output, rather than fewer hours, is made clear by the following calculation. As we have seen, productivity in the United States increased by at least 433 percent (5.33 times) between 1929 and 2007. In 1929, work hours averaged forty-eight per week, while by 2007 this number had fallen to about forty. If a quite moderate proportion, say 20 percent, of the increase in productivity experienced by the United States over this period had been converted into less working time, rather than higher output, then between 1929 and 2007, there would have been close to a 50 percent reduction in the length of the working week—such that by 2007 Americans would work for twenty-four hours per week, rather than close to forty. Even a very modest 10 percent conversion of increased productivity into fewer working hours over that period would have yielded a 33 percent reduction in the working week, from forty-eight hours (in 1929) to thirty-two hours.15 If we take 1940 as a base year, rather than 1929, the results are even more striking. By 1940, the working week was just over forty hours per week. This followed the passing of the Fair Labor Standards Act of 1938, which made time-and-a-half pay compulsory for hours worked in excess of forty per week. Once hours stabilized at around forty per week in 1940, they simply failed to fall much further. Yet labor productivity increased by about 5 times between 1940 and 2007. A 20 percent conversion of increased productivity between 1940 and 2007 into fewer hours rather than more output would have reduced the workweek to twenty hours, while a 10 percent conversion would have reduced it to twenty-eight hours.

Empirical Pattern in the United States

49

The length of the working week is, of course, only one dimension of the onerousness of labor. Other dimensions are the intensity of labor during the workday and general working conditions, both of which greatly complicate the picture. Conditions of labor in industrial capitalism in many ways improved over time, as the state increasingly stepped in to regulate things like safety in the workplace. Yet at the same time, as capitalism advanced, more efective managerial systems intensified the extraction of labor during work. Intensity of work is clearly not indicated by the length of the working day, and the extensive and intensive dimensions of labor can vary independently.

A comparison with an earlier stage of capitalism is instructive on this point. Although hours were long in the nineteenth century, social historians have provided much evidence that in the early period of capitalism, work was not clearly diferentiated from nonwork. As E. P. Thompson (1967, p. 60), in his famous paper on industrial work time, noted, before the advent of the factory, artisans worked irregularly over the course of the working day, with “no great sense of conflict between labor and passing the time of day.” Eric Hobsbawm (1964), in a classic overview of the topic, describes the way in which work across the capitalist world in the earlier nineteenth century was informally broken up by nonwork activities, even including frequent interruptions of the working day for drinking sessions. Labor historian David Montgomery (1980) maintains that the history of the American working classes broadly conforms to the pattern described by Hobsbawn. Herbert Gutman (1973), writing about work patterns in nineteenth and early twentieth century, concurs with Montgomery’s assessment. In just one of a great many examples, Gutman quotes a cigar factory owner in 1877 complaining about his workers: “The dificulty with many cigarmakers is this. They come down to the shop in the morning, roll a few cigars and then go to the bar or saloon and play pinochle or some other game . . . working probably only two or three hours a day” (p. 558). Roy Rosenzweig (1985) describes a similar workplace culture in late nineteenth century America in his account of the transformation of work and leisure in Worcester, Massachusetts. Within this older culture of the workplace, work efort was dictated by custom and habit rather than by the exigencies of production.16

50

Time for Things

A large part of the improvement in productivity in modern capitalist economies has been the result of capitalists fighting against these lackadaisical work habits, exerting greater control of the labor process, in order to increase the intensity of work. For example, Joel Mokyr, examining productivity in the British economy between 1856 and 1973, notes that GDP per hour worked increased over that period by an average of 1.7 percent per year. Yet once the “quality” of labor is taken into account, adjusting for education levels and, most importantly, for increases in the intensity of efort, the increase in productivity over this period drops to 0.7 percent (Mokyr 2003). If a substantial proportion of increasing productivity is accounted for by the increasing intensity of the labor process then, as productivity advances, work per hour will tend to become more draining on the worker. Under these conditions we might expect a strengthening rather than an attenuating preference for fewer hours on the part of workers. The fact that the tendency for hours to drop diminishes in the latter two-thirds of the twentieth century then becomes even more perplexing.

One complicating factor when considering changes in productivity, wages, and hours is the impact of the entry of women into the workforce. In 1900, roughly 20 percent of women participated in paid work. By 2000, that figure had risen to 60 percent (Dora Costa 2000b). The result of this change has been an overall increase in the labor force, on one estimate accounting for a 50 percent rise in the working population aged between twenty-five and forty-four over the first four-fifths of the century (Goldin 1986). Much of the growth in female labor took the form of part-time work, particularly after 1950.17 According to Dora Costa (2000b, p. 109), “in 1950, 23% of all working married women in couples where both spouses were between ages 25 to 39 were working fewer than 35 hours a week. By 1970 this proportion was 35% and in 1998 was still 33%.” The tendency for women to work part-time hours meant that the efect of more women entering the labor force was to depress average hours worked per employee.18A decrease in overall average hours caused by the mobilization of women as part-time workers clearly should not be interpreted as an increase in leisure.19 Indeed, according to Avner Ofer’s analysis, hours devoted to wage labor per household with children increased significantly between the late 1960s and the end of the century. On his estimation, “families with children had a combined 53 hour week working for pay in 1968, and a 64 hour working week in 2000” (Ofer 2006, p. 298).20

Empirical Pattern in the United States

51