Robots are becoming more sophisticated and common, and consequently they will inevitably displace a large number of workers (they have already done so). But historically, while technological progress generally resulted in painful transitory periods, in the long run it was always for the better. The question is whether this will remain the case.
Recently, even middle class jobs were replaced by machines because of the IT revolution. This might exacerbate inequality as the upper classes and those who can adapt to the new technologies will be able to reap the benefits of higher productivity (thanks to machines), but others will lose their jobs or will be forced to lower paying positions. Until now, for the most part new jobs have been created for the displaced workers in various industries. But what if this ceases to continue? What if labor gets replaced by robots and no new labor-intensive jobs arise?
Intuition would suggest that overall the economy will be better off with more robotic technology, because robots are more productive so we can produce more. The issue that arises is whether the benefits of increased productivity will be widely distributed among the general population. If yes, then everyone can be made better off and we may even get to work less. If not, then inequality will worsen, but also the purchasing power of the masses may erode which can in the end put the economy on a downward trajectory.
Sachs, Benzell and LaGarda (2015) (ungated) examine this issue in a model where robots can basically replace labor. They reach a conclusion similar to my intuition. They consider a two-period OLG model with households who work, consume and save. When young, households work and save, when old, they consume their savings. Savings can take two forms: ordinary capital and robot capital.
Production is modeled by assuming there are two sectors producing two separate consumption goods. There is an automatable sector in which robots can replace humans and ordinary capital. And there is a traditional sector where labor (and ordinary capital) cannot be replaced by robots.
An increase in the productivity of robots has two important consequences.
Lower wages. Increased robotic productivity induces people to invest in robot capital instead of ordinary capital. There will thus be less ordinary capital for workers to work with, because people will reallocate their investments from ordinary capital to robots. Less capital to work with means workers’ marginal products (or equivalently their productivity) will be lower. This will mean labor will be less valuable, so wages will drop.
Increased return on savings. People can now invest in more productive assets (i.e. in the more productive robots as opposed to ordinary capital or less productive robots that we had before the technological advancement). More productive assets yield a higher return, so returns to savings will increase.
Clearly, the first effect is negative, the second is positive. When one is young, wages are a more important source of income, so the welfare of young people will drop. In old age, savings matter a lot more, so the welfare of old people will increase.
Which effect is stronger is ambiguous and in general markets cannot guarantee a welfare enhancement. But the authors show that government intervention can make everyone better off. Namely, taxing the old’s return on savings (via a capital gains tax) and redistributing the proceeds to the young can make everyone better off.
How can this work? The government would have to make the wage of the young equal to what it would have been without the advances in robotic technology. This will ensure the young are not worse off than before. Since the income of the young is now higher, they will invest more money in (various forms of) capital. This will make output grow. The positive effects of this growth in output will be high enough to offset the taxes paid by the old. So the old will be better off as well.
Let us now look at how the economy evolves over time. The economy is shown to have three possible phases.
No robots. Robotic productivity is too low for it to be profitable. So no robots are used. The economy is on a growth path. As capital stocks are growing, returns to capital are going down. If returns to capital go down enough so as to make robots profitable (or if robotic productivity grows), then we move to the next phase.
Mixed production. In this case, both robots and ordinary technology (of capital and labor) are used to produce the automatable good. The effects described above (lower wages, increased savings) are prominent in this phase. Government transfers financed by a capital gains tax can enhance welfare. With rises in robotic productivity, the economy can transition into the next phase.
Robots only. The automatable good is fully produced by robots in this scenario. There is a possibility for permanent growth because the automatable sector can become very productive. As this happens, capital is shifted from the traditional sector to robots. This makes traditional goods relatively more scarce. Therefore, their prices (and hence wages) will increase. The positive effect of rising wages (and the subsequent rise in savings/investment) can outweigh the negative effects of rising prices and the decrease of labor productivity (which is due to capital migrating from the traditional sector to the automatable sector). If this is the case, then there is permanent growth. Otherwise, there is contraction.
The possibility of contraction in the robots only case can give rise to cycles: we can alternate between the robots only and mixed production cases. It turns out that the higher robotic productivity is, the more likely we are to be in the permanent growth scenario. So this contraction-growth cycle can be escaped with further advances in robotic technology.
Below is a simulation of the model that illustrates these phases.
We start in the no robots phase in period 0. Then an increase in robotic productivity in period 5 (black) moves us to the mixed production phase. This increases utility (red) for the old, but subsequently the young suffer because of lower wages. Another increase in robotic productivity happens in period 10, it has similar effects. In period 15, we transition into the robots only phase. Robotic productivity is too low for now to deliver permanent growth, so we’re in the contraction-growth cycle as evidenced by the red curve. With a further increase in robotic productivity in period 20, we transition to permanent growth.
For the most part, utility without transfers is below the pre-innovation (i.e. no robots) level (dashed black line). We only grow above this rate once we enter the permanent growth stage of the robots only phase. And even then it takes several periods. This is largely bad news. But we can see that transfers (blue line) can ensure a much smoother transition, and it can keep us above pre-innovation utility levels at all times.
To conclude, let me say that I am somewhat skeptical as to whether robots will really make labor redundant. This has never happened in the past, we’ve always managed to create more work for ourselves, it’s like Parkinson’s Law is at work on a macro level. But clearly, technological advances are ever more sophisticated so the notion of robots taking over cannot be entirely dismissed.
This paper shows us two things. First, that a government transfer scheme is likely going to be necessary to ensure that robotic technology does no damage to welfare. Namely, transfers from people who mostly save or invest (the ‘old’ in the model, but could be ‘capitalists’ in the real world) to people who mostly rely on wages (the ‘young’ in the model, but could be ‘workers’ in reality) are necessary. This is kind of obvious, because the only way the population at large can take advantage of robots is if the benefits of robots are somehow distributed across the whole society.
The second message of the paper is that if robotic productivity increases a lot, we can hit a permanent growth path. This relies on the assumption that the production of certain “traditional” goods cannot be automated. These goods will become relatively scarce which will increase wages, and this can help move us on a permanent growth trajectory.
So altogether, robots can be awesome for our welfare. But we need to ensure proper redistribution is in place so as to distribute their benefits. I for one welcome our new robot overlords.