guest post by Sandwichman
Five Million Jobs?
“It is remarkable, in view of the virtual unanimity of opinion among economists as to the general shape of the relationship between hours and output, that the effect of hours shortening has received so little attention in published projections. It is often completely ignored even in their description.” – Edward F. Denison, “Employment and Hours of Work: Their Contribution to Past Growth and a Projection of the Future” in The Sources of Economic Growth, 1962.
In the early 1960s, Edward Denison, the founder of growth accounting, estimated that roughly one-tenth of the economic growth that occurred between 1909 and 1958 in the U.S. could be attributed to the “effect of shorter hours on [the] quality of a man-hour’s work.” That would have translated into approximately three million jobs created — except Denison also assumed that the optimal working time for output was 48.6 hours a week, 52 weeks a year (the 1929 annual average)! So he subtracted “the effect of shorter hours on [the] quality of a man-year’s work” from his estimate of growth, leaving an implied net loss of around 200,000 jobs over the 48-year period. What those calculations show, more than anything, is how sensitive any employment projections are to the framing assumptions. We will soon have occasion to tweak these assumptions to generate a range of employment estimates, but first a word about the long-term trend in hours of work.
Joseph Zeisel, writing in the Bureau of Labor Statistics Monthly Labor Review in 1958 called the long-term decline in the industrial workweek, “one of the most persistent and significant trends in the American economy in the past century.” That is, it had been a persistent and significant trend up until around 1940. After a brief spike during World War II, however, the length of the average workweek in manufacturing quickly receded to pre-war levels and then remained essentially unchanged for 65 years, with only minor, short-lived fluctuations reflecting the ebb and flow of the business cycle. By March 2010, the average manufacturing workweek lasted six minutes longer than it had in August 1945.
The chart above shows the decline in weekly manufacturing hours from 1850 to 1950, the levelling off of hours after World War II and a trendline extrapolating from the 1850-1950 rate of decline. In the broader economy, annual hours of work continued to decline after World War II, but at a slower rate than they had previously.
Furthermore, the increased participation of women and students in the workforce, a sectoral shift of employment away from manufacturing and toward services and the expansion of part-time work contributed to the post-war decline in annual hours. Annual hours for full-time workers showed even less movement. The chart below shows “potential” and actual annual hours from 1909 to 1958 as reported by Denison and annual hours from 1950 to 2009 as reported by the Conference Board’s Total Economy Database.
Had the annual hours of work continued to decline at the rate they did from 1909 to 1958, average annual hours in 2009 would have been about 14% lower than they were. What might the effect on job creation have been? Recall that Denison’s estimates were based on his assumption that the optimal length of the workweek for total output was 48.6 hours, 52 weeks a year. He also supposed that maximum productivity would occur at 33.9 hours a week. Below that latter figure, output was assumed to fall faster than hours as hours declined. But is it reasonable to assume that the optimal hours for both output and hourly productivity would remain the same from 1909to 2009? No.
Not unless we are prepared to defend the proposition that technology has remained unchanged or that the optima are unaffected by technology.
To estimate the job creating potential of shorter hours, I assume that the trend of hours of work from 1909 to 1958 approximates the amount of working time that would be optimal for output. This suggests that in 1909 the optimal workweek was indeed 52 hours and in 1958 it was 39.6 hours. Projecting that trend indicates an optimal workweek of 29 hours in 2009 (or, alternatively, 32 hours a week with five weeks’ vacation). Since a longer than optimal week subtracts output from the optimal potential, we can estimate that about six and a half percent of potential output was wasted by excessive hours of work.* That is to say, people were too worn out to produce as much as they otherwise could have. In terms of jobs lost, this represents over five million jobs.
I’ve used some fairly conservative benchmarks to anchor the trendline. If I had chosen 1940, instead of 1958, as the endpoint, then the trend would have indicated an optimal working year of 1400 hours and around 11 million potential jobs foregone. Also, Sydney Chapman’s theory of the hours of labor shows that hours determined in a hypothetical competitive market tend to be longer than optimal with regard to output. So even the pre-1940 trend may overstate the length of working time that would be optimal. The assumption that the given hours of work in 1909 and 1958 were optimal is, in effect, unrealistic. But to produce any estimate at all it is necessary to make such an unrealistic assumption. In the period since World War II, a significant payroll tax and regulatory bias against reducing the standard full-time hours of work has emerged, more or less “freezing” weekly hours at the forty-hour standard established in 1938 by the Fair Labor Standards Act.
Remember, these are not work-shared jobs, created by redistributing a given number of hours of work. They are productive jobs associated with an expanded output. Whether or not that additional economic growth would be a good thing is another matter I will examine in due course. For the present, however, what I want to call attention to is the seeming indifference of conventional economists to a rather considerable untapped potential source of job creation.
*An explanation of how maintaining excessively long hours over the long run actually detracts from total output can be found in Sydney Chapman’s 1909 Economic Journal article, “Hours of Labour.” Initially, an extra hour of working time (beyond the hypothetical long run optimum) adds an increment of output to the daily total, but over time, say a few months, the accumulated fatigue leads to a diminished pace throughout the day and a lower total daily output. The optimum is thus determined by the amount of time and effort can be sustained over the long run.
To estimate how much output is subtracted by excessive hours, we use an idealized, smoothed work curve that assumes a rise in hourly productivity through the early hours of the day followed by a plateau and decline of productivity as fatigue sets in. These changes can be represented by a sine curve, which is convenient for calculation of the area under each segment of the curve, representing the variation in hourly productivity. For example, the diagram below represents work curves for 1909 and 1958. As indicated, the output per worker in 1958 was a bit more than twice the output per worker in 1909. That greater total output, however, occurred in a much-reduced number of hours of work: 2704 hours in 1909 and 2060 hours in 1958. If we assume that annual hours in both 1909 and 1958 were optimal, given the respective levels of technology, then increasing annual hours in 1958 to the 1909 level would have reduced total output in 1958 by around 22%.