The Baumol Puzzle
When you buy a car, should the price include a licensing fee to the ancestor who invented the wheel?
That question is implicit in William Baumol’s new book. I think that the answer is obviously “no,” but it is less obvious to Baumol. The puzzle is worth thinking through, and I think that it helps to illustrate the differences between what Nick Schulz and I call Economics 1.0 and Economics 2.0.
In Economics 1.0, if you have competitive markets and approximately constant returns to scale, then payments to the suppliers of factors of production tend to be “right” in two senses. First, they provide the correct incentives to supply hours of work, effort, capital, and so on. Second, factor payments tend to reflect the relative contribution of the various factors.
If you think of innovation as a factor of production in this sense, then Baumol argues that innovators are under-compensated. Most of the benefit of innovation accrues to people other than innovators–ultimately, it accrues to consumers. Baumol sees this as inefficient in theory but good in practice. He thinks that we would have a wildly unequal distribution of income if innovators were paid for the full value of their innovations. His resolution of the puzzle is that society has contrived institutions that pay innovators less than the value of their innovations in order to reduce poverty.
I lean toward a different resolution of the puzzle. I would emphasize that:
1. With innovation, the relationship between effort and value is unknown ex ante. You do not know whether that idea you are working on is going to turn out to be the Lisa (Apple’s first graphical-interface computer, which bombed) or the Mac. You do not know whether the problem you are trying to solve is going to require years of effort or yield to a moment of inspiration.
2. Useful innovations tend to be combinations of innovations, multiplied together. (Matt Ridley’s catch-phrase is “ideas have sex.”) The car requires the wheel, gearing, and the internal combustion engine (among other things). There is no way to compute the value of those innovations separately.
Combining (1) and (2), you do not know how valuable a current innovation will be in the future. Back in the stone age, who could know that the wheel would be useful in 2010 but that the hand axe would not?
(I should note in passing that it is physically impossible to pay a license fee to the inventor of the wheel today. The inventor is dead. Moreover, even if while the inventor was alive all of the wealth produced by the wheel could have been foreseen, there was not sufficient wealth available at the time to give the inventor the present value of his or her innovation.)
I think that the main reason that the wheel produced spillovers that far exceeded that which was needed to induce the effort required to invent the wheel is that there were unforeseen combinations of the wheel with other innovations. My guess is that this is true of much innovation. A transistor, by itself, is an interesting curiosity. But modified in many ways and combined with other innovations, it facilitates a revolution.
I believe that the “spillover effect,” in which the benefit of innovation far exceeds the compensation paid to innovators, reflects this multiplicative combinatorial character of innovation. For Baumol, innovation seems to exist as a factor of production in an Economics 1.0 world, and it just happens that innovation is not as well compensated as other factors. He thinks we would get more innovation if it were properly compensated, but we do the right thing by redistributing the benefits to reduce poverty.
I think that the relationship between effort and value in innovation is much less precise. Innovation has combinatorial characteristics, making it difficult to value a single innovation. And innovation has long-term benefits which cannot be forecast and for which it is not feasible to pay compensation.
I do not think we can put the square peg of innovation into the round hole of neoclassical production functions. Baumol’s puzzle is worth thinking about, but I do not think that Economics 1.0 is the right way to do it.