This article notes a fact about current wages in the data science industry and speculates a bit about possible causes.
U.S. data scientists are currently experiencing strange returns to experience. Here is the source of that claim, from PayScale. Below is the key image from the source article:
Here is a chart from the same source for a Web Developer’s pay:
Both figures were retrieved on 5/29/15. The web developer pay chart is more expected because it has clear decreasing returns to experience demonstrated by a decreasing or concave slope.
These charts could mean many different things. I will discuss 3 interpretations but there are others.
First, let’s talk a bit about the X-axis. I think it would be more precise to use units of years of experience rather than categories of experience. On such a chart we might continue to see the current pattern of increasing returns to experience, but I doubt it. This leads us to our first and most obvious interpretation of the chart.
The first interpretation of the chart is that there is a bubble in data science wages because the wages exhibit increasing returns to scale when this is inconsistent with efficient market theory. The sky may be falling because this could signal a wider bubble in business intelligence and so on.
As just discussed, however, I think this pattern would vanish if properly charted. However, even if there are not increasing returns with respect to years of experience, it still might be the case that such returns are higher than their counterpart levels under an efficient allocation.
The second interpretation is that there are no increasing returns but there is a bubble. The wages simply seem to high for experienced data scientists. Why might this be?
Modern data science is a somewhat new field. We can speculate that experienced data scientists are rare enough that they may exercise some oligopsonic wage bargaining power due to lower than efficient levels of competition.
The market can produce workers of high productivity, but this may be hard to measure or signal to employers. It could be that the industry employers are not skilled at differentiating workers by productivity, so they use years of experience as productivity signal when determining wages.
This interpretation might lead us to believe that the chart is only a temporary artifact and that over time the market will solve the issue.
Unlike watches or widgets, the market can’t rapidly produce laborers with many years of experience. For now the professionals with more experience have bargaining power but they will gradually be displaced.
The third interpretation is that there is no bubble at all. It could be that investing in data science, IT, or intelligence in general, is somewhat akin to investing in TFP rather than investing in more normal kinds of labor or capital. If this is the case then we would not be surprised by a more flat or even increasing curve on a graph with experience on the horizontal axis and wages on the other.
My conclusion is that each of these explanations play some role. I think data scientists receive decreasing returns to experience, but their rate of decrease is small compared to other professionals. I think this is explained in part both by a shortage or labor, which can also be viewed as high demand, and also in part by the innovative nature of the field. Finally, I do think there is a bit of wage bargaining power enjoyed by the most experienced data scientists. I think this is probably also best explained by a shortage of such laborers.