As I argued in my last post, there is good reason to believe that computers are substitutes for labor in general. Their apparent complementarity with skilled labor was the result of substituting skilled labor for unskilled labor rather than being based on a fundamental technological complementarity. The logical next question is what this means for wages.

To answer that we need a bit more theory first. How wages get determined has been hotly contested over the years but I think the basic answer turns out to be quite simple: through bargaining between labor and capital. Let’s start with the example of a single firm first and then work up to the economy as a whole. A firm produces value and how that value gets split up is the result of bargaining (I am saying value because value can be created even during a time when the firm does not yet have a profit). At the level of a single firm the bargaining involves not just the employees and the investors but also the suppliers (and even the customers if you have high customer concentration).

Let me explain more clearly what I mean by this by using suppliers as an example. If you are a music startup and you want to have licensed music on your platform you need to negotiate with the music labels. There happen to be only three of them left and you need a deal with each one of them to have all music (they are not separate suppliers that you can play off against each other). So it shouldn’t be surprising that music labels routinely take a huge bite out of the value created by online music services.

This kind of bargaining with suppliers (or customers) determines how value creation is split between firms but for the returns to labor and capital we are primarily interested in bargaining between those two factors of production and how that aggregates in the economy as a whole. So lets turn to this next. My basic argument has been that computers (plus skilled labor) have substituted heavily for unskilled labor so far. This explains the high unemployment rate. Important note: unemployment as measured in the government statistics has come down but that’s because a lot of people have stopped looking. Better measures, such as employment to population ratio and labor force participation are still way down, see my recent talk.

With a lot of people out of work bargaining will push down the take of unskilled labor to the reservation price, meaning the minimum wage for which someone is willing to work and which could be quite low. In many cases the only buffer here is the legally mandated minimum wage. Importantly this implies that there is no relationship between the marginal economic value created by this labor and their payment. This is why it is possible for companies such as Walmart or McDonalds to pay very little to their employees while making massive profits.

Now lets look instead at the type of skilled work that to date has been boosted by the use of computers. Since there are many firms hiring computer engineers and there are many computer engineers we should expect neither to have huge bargaining power. In that case the wages are likely to be a good reflection of the marginal revenue productivity of labor. In other words on the margin as firms expect that they can make more money by hiring more engineers they will do so and the wage to be paid for that has to be competitive as the engineers can choose to work elsewhere instead. That shows up quite well in the increase in starting salaries for computer science grads over the last decades. Important note: studies about average wages in IT are highly misleading as they tend to include data entry positions and other jobs that are already being rapidly eliminated.

This bargaining view of the world has a further benefit in that it also explains why management salaries have gone through the roof. As we have used computers to substitute for unskilled labor we have created massive additional profits. Skilled labor has been given some share of that but that share is limited to roughly the marginal product by competition between engineers. That leaves a large pool to be split between top management and capital. And in that bargaining situation it turns out that top management is surprisingly well positioned. Why? Because there is a large principal-agent problem between board members and the capital that they represent (btw, Piketty’s chapter on that is worth reading). It is a lot easier to vote for than against a raise for management especially if everyone else is doing it.

To the extent that I am right that much of the current demand for skills that work with computers are because they ultimately substitute against unskilled labor we should not expect the benefits for skilled labor to last forever. Instead, we will gradually see computers substituting for those as well and only top management (and capital) continuing to benefit. And so the current idea that we can somehow educate our way out of this divergence without the need for more profound changes in how we think about income is likely deeply flawed.