And It Ain't Shinola... II

by the Sandwichman

Wherein the Sandwichman assesses the validity of the claim by Lanoie, Raymond and Shearer that:

...little empirical work has been done to measure the consequences of work sharing. ...data that would permit the direct measurement of the productivity effects of work sharing has generally not been available....

Counterpoised to the above statement is the observation made by Chris Nyland in Reduced worktime and the management of production (1989) that a "vast mass of [empirical] research has been undertaken into how the relationship among effort, efficiency and time manifests itself within the production process (page 45)."

(see also And It Ain't Shinola, Part I)

To be sure, "work sharing" is only a particular application of reduced worktime and, for the most part, the "vast mass" of research deals with circumstances that differ in important ways. However, that broader research also bears on the narrow issue of work sharing in important ways. Most important is the issue of isolating the effects of the worktime change on output. This is difficult to do at the best of times. When the matter is clouded by poor implementation, short time horizons and a simultaneous increase in the length of the working day, the results have to be taken as fundamentally inconclusive.

To Lanoie, et al.'s credit, they do acknowledge several of the confounding factors. However, their cursory theoretical discussion and even shallower grounding in the specialized literature lead them to understate the ambiguity of their empirical findings and thus to vastly overstate the significance of their results: "the results suggest that work sharing has induced a significant reduction in productivity." The results suggest no such thing. On the contrary, what the results suggest to someone with a grounding in the working time literature is that it may be a misnomer to even call the schedule change "work sharing". Why not call it a 10% pay cut with an attempted speed-up?

The "classical" empirical studies of working time and output are Abbe's analysis of the introduction of an eight-hour day at the Carl Zeiss optical works in 1900, Vernon's investigation of work schedules in British munitions factories during World War I and Kossoris's study in the U.S. during World War II. All of these studies found significant gains in productivity from shorter work days and work weeks and even larger productivity losses from increased days and weeks. Another important finding was the time lag of several months before the productivity gains would take hold. In Abbe's research and several of the Vernon and Kossoris case studies reductions in working time resulted in actual gains in aggregate output per worker, not just gains in hourly productivity.

In 1924, Otto Lipmann of the Institute of Applied Psychology in Berlin collected all of the then known material on the relation between the hours of work and volume of production. The volume presented "about 700 separate data on the effects of changes in hours of work on output... and a bibliography of about 400 works." Sounds more like Nyland's vast mass than Lanoie's little empirical work. However, Lipmann pointed out, "A comparatively small portion only of the data collected and compiled by the Institute is scientifically satisfactory." Lipmann ventured a general impression of the then already vast mass of data:

The more exact the observations on which a report is based the more frequently do they represent the effects of a reduction in hours as good, or explain the absence of good effects by other causes. On the other hand information based on estimates very frequently reports bad effects of reductions in hours of work.

By 1965, Lipmann's vast mass of data would seem only half vast by comparison with the "huge empirical literature on the relationship between working hours and productivity" reviewed by D.G. Brown (cited by Scott and Spadavecchia): 1,233 firm level studies from the U.S. and U.K., "52 per cent of observations involving a decrease in hours indicated that this did not significantly reduce output."

This is not to say that reducing the hours of work always causes higher productivity. The circumstances of the change and how it is implemented can be as important -- sometimes more important -- than the particular change in hours. In particular, "work sharing" may not conform to the productivity boost because of workers' fear of a shortage of demand. They may, consciously or not, slow down in an effort to "make the job last." Of course, if there is an actual shortage of work, there may also be more involuntary down time during the course of the working day. An increased capacity for work may not be matched by an increased opportunity to use that capacity. Nevertheless, a fall in output isn't necessarily evidence that the change in schedule caused that decline. Again, Lipmann,

The relation between hours of work and hourly output is by no means simple or direct, that is to say, a change in hourly output occurring simultaneously with a change in hours of work is not necessarily (post hoc ergo propter hoc) an effect of the latter change.

In addition to Nyland's book, cited above, another text that surveys the empirical work on hours of work and output is Stress and performance effectiveness by Alluisi and Fleishman, some of which is taken from earlier contract research Alluisi did for NASA. They state:

From the very earliest studies, improvements in industrial productivity have generally been found following reductions in the total hours of work in both the workday and the workweek.

This broad consensus contradicts Ohanian's assertion that the SINGLE study by Lanoie et al, "is evidence that worksharing that reduces the number of days an employee works, even keeping the length of the workweek fixed, also reduces output per hour." The error of such a conclusion is amplified by Lanoie et al.'s evident (self-reported) lack of familiarity with the (vast) empirical literature related to their study.