In response to:
The Golden Arm from the October 24, 1985 issue
To the Editors:
Roger Draper’s argument in The Golden Arm [NYR, October 24, 1985] that robots will cause immense unemployment misunderstands how employment is affected by technological advance. “The one great truth of the matter,” he says, then going on to quote one of the authors under review, is that “unlike other technologies…which increase the productivity of a worker, the robot actually replaces [italics in original] the worker.”
The distinction he draws does not exist. All increases in productivity replace workers. More goods being produced by the same number of workers is the definition of a productivity increase. Before 1900 between 80 and 90 percent of American workers were farm workers. The percentage is now 3 percent. Hybrids, fertilizers, and farm machinery “replaced” the rest.
Productivity increases always offer a choice. Society can choose either greater leisure (using fewer work hours to produce the same number of goods) or more goods. The increased productivity of the American economy over the past century was used partly to provide more leisure. The normal workweek is now forty hours, whereas it used to be sixty-plus. But mostly, we have chosen more goods. Per capita consumption in the United States is now many times what it was a century ago, and we support the world’s most expensive military establishment on top of this.
What is to prevent greater leisure from becoming involuntary unemployment? Mr. Draper seems to think that the only defense is the new technology itself providing additional jobs to replace those it eliminates. This can happen, but it is not necessary, and it is not how new jobs have been created by productivity increases in the past, for the most part. Rather, the productivity increases have the immediate effect of reducing the costs of the goods they produce. This in turn eventually reduces their prices, which leaves consumers with more to spend on other goods, and that, finally, provides more jobs in the industries which produce the other goods. Neither the other goods nor the new jobs generally beat any other connection than that just stated to the productivity increases or the new technologies which brought them about. The millions of people in our economy who currently do not work on farms, for example, are employed in a wide variety of tasks which have no connection with what their ancestors did or their contemporaries are still doing on farms.
The choices which the increased productivity provided by robots will force upon us are likely to be difficult, but not different in kind from the choices increased productivity forced upon us in the past. One difficulty is that only societies which have already reached a certain level of sophistication can make use of them. They are therefore likely to be of no immediate help to the poorer 50 percent of the population of the world. There is nothing unique about robots in this respect. The technology which increased the productivity of the American farmer more than fifty-fold, for example, hardly touched the poorer farmers of the world. Except in certain areas where hybrids have been successfully introduced, farming methods in the underdeveloped countries have scarcely changed at all.
On the other hand, the technology of artificial intelligence, of which robots are a part, seems to be more transportable than were the more conventional mass-production technologies of the industrial revolution. This fact creates opportunities for countries like Taiwan and South Korea, where labor is still relatively cheap although sophisticated enough to do precise mechanical tasks, but problems for high-wage societies like our own. Technological advances that can be as easily implemented abroad often result in exporting jobs. On the other hand again, there is an obvious limit to how far this can go. It will stop when the living standards in these other countries catch up with ours.
That the problems which robots will pose will not be new does not mean that they will not be severe. The economic dislocations created by the first industrial revolution were bad enough. With his evidently very good managerial and technical understanding of robots, Mr. Draper is in a favorable position to try to avoid such conditions occurring again.
University of Southern California
Roger Draper replies:
It is quite true that “More goods being produced by the same number of workers is the definition of a productivity increase.” Yet neither in logic nor in fact does this definition imply, as Professor Slawson suggests, that “All increases in productivity replace workers.” For the most part, the labor-saving devices that sparked the first Industrial Revolution were designed to help manufacturers raise production, not to cut their total labor bills. The cost of labor per unit of output did fall, of course, but since output rose dramatically as well, there was no saving on actual expenditures. Not did the actual number of workers who made cotton yarn and cloth, the most important products affected by the first Industrial Revolution, decline as a result of it; on the contrary, employment in those industries boomed. The very much smaller number of men and women forced out of their livelihoods as handicraft yarn spinners and loom operators could usually find other kinds of manual work—in the cotton factories, for example.
Professor Slawson might nonetheless claim that the cotton industry vindicates his fundamental point: the power of technology to raise employment indirectly by stimulating output and consumption. In the case of the first Industrial Revolution, that is precisely what happened. But it will not happen in future, and we can see why if we apply Professor Slawson’s analysis to the robotics-generated productivity increases that lie ahead.
First, the use of robots, like the use of other labor-saving devices, has “the immediate effect of reducing the [manufacturing] costs of the goods they produce.” True. “This in turn eventually reduces their [the goods’] prices, which leaves consumers with more to spend on other goods….” Less certain, but probably correct. Finally, falling real prices create “more jobs in the industries which produce the other goods.” The third point is the doubtful one. For if robots become so common that real consumer prices fall significantly, is it likely that these “other goods” will be manufactured chiefly by human beings?
In short, as Harley Shaiken truly says in Work Transformed: Automation and Labor in the Computer Age, “unlike other technologies…which increase the productivity of a worker, the robot actually replaces the worker.” The labor-saving methods of the past—such as hybrids, fertilizers, and farm machinery, to cite those mentioned by Professor Slawson—were restricted to particular industries or to a narrow range of functions. Not so robots, which can be used in almost any industry and for a wide range of functions. That is why robots threaten manual labor as such, not this or that kind of manual labor.