Division of labour is the outcome, not cause

I’ve written twice now on the spurious ideas in the division of labour story as an underlying cause of productivity growth in economics.

First I questioned whether Adam Smith’s observation of a division of labour in 18th century pin factories made much sense, given that the 18 tasks required to make a pin were undertaken by 18 men in some factories, but only 10 men or fewer in others. Clearly even in this iconic case it was not the division of labour tasks into more specialist roles that was the cause of the rapid gains to productivity in pin factories.

Second, I made the comment that the division of labour story has become an endless repository of ad hoc explanations for productive gains. The story has even captured the imagination of archeologists and anthropologists who saw the invention of tool-making as “gearing up for a clearly defined division of labor.” Yet the simple logic of increasing the number of possible production tasks following the invention of certain tools would mean that each person could do more, rather than fewer task, suggesting a rather strange interpretation of the division of labour. I also noted that the division of labour story rests on labelling conventions of roles in society rather than actual units of labour being devoted to fewer clearly defined tasks.

Here I want to be even more clear on this final point by presenting a minimal example of how roles and tasks can lead to confusion, and emphasise that productivity is the result of doing more tasks with fewer people – the opposite of specialisation.

Inspired by the ancient tool-making tribes of Jordan, my example is a 6-person tribe that undertakes 6 defined tasks, of which the two named roles undertake 3 each. Thinking in terms of roles there are 3 hunters and 3 gatherers, yet each person undertakes just one task within those roles.



Person in role

Track Hunter 1, 2, 3
Collect Gatherer 5, 5, 6

You might want to argue that the way I define tasks offers limitless ad hoc classification. Tracking an animal could be further divided into a team pursuit with specific sub-tasks for each member. Same with cleaning an animal. But this is kind of the point. Any defined task will be a bundle of sub-tasks. But in order to understand the division of labour we need to keep track of tasks at any one particular level of aggregation and not fall into the trap of calling something specialisation when it is just a different bundling of more tasks into one job.

One of the tribe members now invents the spear and woomera. Regular production of these tools requires 3 additional tasks to be undertaken by the new toolmaker role in the tribe.

Task Role

Person in role

Track Hunter 1, 2
Collect Gatherer 3, 4
Collect Tool maker 5, 6

Now we have more roles and fewer people in each of them! Exactly as predicted by the division of labour story.

But if we instead look at the tasks, we have more tasks per person. Each hunter, instead of being able to specialise in one task, like tracking, now must undertake more than one task on average as there are only two hunters available for three tasks. The same for our gatherers. What we see as specialisation in roles is the automatic result, not cause, of increasing productive capacities.

What has happened is that the invention of new production techniques has allowed more tasks to be undertaken by each person leading to fewer people in each role. It is not a case of dividing labour in a way so that each person completes fewer tasks, each requiring less training, in order to increase aggregate output, as is often argued. The most productive countries are not full of people doing repetitive narrowly defined non-skilled tasks, but highly educated people doing ‘specialist’ roles involving a hierarchy of complex and interrelated tasks that require years of training to master.

You might still be thinking that the joint production function from specialising on the task that each person has a relative advantage in rescues the division of labour story. Crusoe catches fish, Friday gathers coconuts, and their combined output can be greater than if they individually produced what they consumed. But the simplification in this story ignores the possibility that if Crusoe and Friday gather coconuts together, then fish together, that the complementarities in joint production for each task might increase their combined output by more than if they specialised and worked alone.

Maybe catching fish involves a number of distinct tasks that, when shared between them, would increase their their output beyond twice that of the most productive of the two men. Would that also count as division of labour? If so, then it appears that any joint production can be labelled as a division of labour without offering any insight as to why one division is better than another. What Crusoe and Friday actually need to grow their economy is to each achieve more tasks each in a given amount of time.

To cap off, the division of labour story to me is just an observation about the human roles in large scale production. It is not a causal story for increasing productivity. Productivity requires labour bundling, or given more tasks to each person with the help of new technologies, is the way to increase output.


  1. Very interesting. One quibble, Adam Smith was writing in the 18th century (1700s), not the 19th.

  2. Speaking of toolmaking and labour productivity, some people are questioning whether computing machines have been a net plus to labour productivity so far.

    I don’t think they have been, but will be when the robots penetrate further down the toolchain.

    • I’m not sure what you are guaging that on but I can say from a manufacturing perspective computers make a big difference.

      Take CNC machines: old days you would have one worker at a machine that required 100% concentration to make a part. A manufacturing worker today can easily multitask between 2 or even 3 cnc machines producing a near constant stream of parts each being made in a far shorter amount of time due to the computer’s ability to calculate and act so much faster. The initial setup time is roughly the same but with added programming and dry-run time. Its a crazy productivity gain.

      Above all else, humans are toolmakers. Strip us of everything that we think we need (apocalypse scenario) and the first things we will do is reach for the nearest stick and rock, and collaborate with the nearest other human through shared intent (computers help with that too).

      I don’t know what the precise measure of productivity is (in units or whatever) but I can say that never in history has one person been able to build more tools.

      • Excellent paper, although my read is that re: computers the author feels we got the lions share of productivity gains by the end of the ’80s:

        “Many of the inventions that replaced tedious and repetitive clerical labour with computers happened a long time ago, in the 1970s and 1980s. Invention since 2000 has centered on entertainment and communication devices that are smaller, smarter, and more capable,but do not fundamentally change labour productivity or the standard of living in the way that electric light, motor cars, or indoor
        plumbing changed it.”

    • If a computer can’t give a productivity increase, it’s more likely the role is a BS job.

    • Tassie TomMEMBER

      Computers certainly haven’t reduced the number of people employed in public health bureaucracies.

      They have however decreased the number of secretaries required to manage our private group.

  3. Excellent. Thank you, Cameron.

    Greater output comes from higher skills and increased work intensity, not task subdivision. GDP growth charts (I don’t have one at hand) show a consistent contribution from labour productivity while other factors flail around. I think you have turned a key economic precept on its head.

  4. It is interesting theme for discussion, but still not analyzed deeply. I have the impression that you do not make any difference between labour spesialization and work cooperation. Most of your examples, very simple indeed, are about cooperation, not just everyone doing everything. Cooperation is very different from lack of spesialization, e.g. everyone doing everything for himself. Specialization is more productive when it rises from cooperation and follows cooperation. This is the historical sequence of work activities separation and specialization.
    Today for example, the tendency of everyone doing everything on a project can be efficient and more productive in some industries, but very inefficient in others, beleive me. And generalization of this kind about productivity growth is fruitless. Technologies are not evenly productive in all industries and all kind of skills and human abilities replacement. Anyway, it is a good start for discussion on this matter.

  5. The capital hypothesis was overlooked by Smith as he emphasised labour. One area where division of labour yielded was essential was in the potteries. Capital investment was low, relatively, but each worker was assigned very boring and specific tasks from women operating as the treadmill labour and gaining larger right thighs; to those that laid out the clay for the potters to work with. This division of labour extended throughout the process. It was underpinned by low wages and owners that saw all the wages being spent back in factory shops and pubs.

    • Ronin8317MEMBER

      Reducing the skill needed for each process allow the employer to keep wages low. Lower wages however is not productivity, it is merely shifting more profit to the owner.

  6. george fripley

    The argument that multi-skilling people to do an array of tasks sounds beneficial, but my experience has been the opposite. Too often I’ve seen people achieve high levels of mediocrity, including in their primary skilled area, as tasks multiply. Obviously there needs to be balance for productivity gains, but it can be taken too far. You don’t want to lose excellence in areas as a result of an ideological push to multiskilling. My view is that tasks can be packaged in certain areas to gain benefits, but the line needs to be drawn to indicate where the boundaries are.

  7. Interesting in a compartmentalised view, but is it correlation or causation the author has touched upon? Probably a combination of both, but the key missing piece is the human factor. Where is the human payoff when basic human needs are met? Early models assume rational behaviour, yet humans are emotional beings with likes and dislikes that may override the rational choice. When basic needs are satisfied reward opportunity diminishes thus model outcomes are blurred. Simple models uncover complex questions in a feedback loop which fogs the original question. Simple early economic models were vehicles for observation and introspection. Adam Smith et al have probably rattled their graves over the decades as every subsequent crop of economic thinkers invests their intellectual capital in reshaping these early models to contemporary thinking. They were probably never entirely relevant, but presented pathways to understanding. Ironically the human condition that defrays the models also capitalises the models resulting in both optimal and suboptimal outcomes. I’m getting a headache, I’m reading Jorg – Meh!