QWERTY, Lock-In and Path Dependence

16 Mar 2001 13:00

Look at the first line of letters on your keyboard: QWERTYUIOP. There's no real reason why just those letters should be sitting there in just that order: except that one of the early sorts of type-writers had that order, and became more popular than its competitors, and so fixed the pattern more or less permanently. QWERTY has become a general name for such "lock-ins" in technology and economics, also known as "path-dependence" (a mangled bit of physics jargon). It is held to result from "switching costs". These take two forms. One is what we might call pure switching costs; the opportunity cost of obtaining, installing and learning to use a new technology. (Once you've learned to touch-type, learning a new layout is a true pain.) The other form of switching cost is due to "network externalities," or "external increasing returns". Normally economic analysis assumes constant or decreasing returns, i.e. that each additional unit of resources devoted to a task brings the same return or less as the last unit. Increasing returns stand this on its head: unit of resources n+1 has a larger return than unit n. Many real-world production technologies show increasing returns, simply because they operate more efficiently at high volume, but what we're looking at here is different. The technology per se is not more efficient when it's more widely employed --- but it is more valuable. The more common QWERTY keyboards are, the more useful it is to learn them rather than, say, ABCDEF, or the Dvorak keyboard. In turn, the more people know how the QWERTY layout, the better it is to be selling that, rather than one of its competitors, and the better it is to by buying them for your employees. Hence there are increasing returns in the number of QWERTY keyboards in place, and this is due to the value of the "network" of such keyboards. Leaving that network imposes a switching cost, even if there is absolutely no difference in intrinsic merit between the product and its competitors. Note that the technology with the largest network will, all else being equal, grow at the expense of its rivals, since that is the one with the most valuable network externalities (and hence the highest switching costs to leave).

Similar ideas were not unknown to Adam Smith, and Alfred Marshall wrote about them at length in 1890, but after that they led a skulking, twilight life among the economic geographers and urbanists and development economists. (The more nobles (resp., geeks) live at Versailles (resp., in San Jose), the more important it is for the others to do so too.) About twenty years ago these ideas were re-introduced into mainstream economics by (among others) Avinash Dixit, Joe Stiglitz, Paul Krugman, Elhanan Helpman, Paul David and Brian Arthur. (The trick was to find ways of modeling increasing returns mathematically which are internally consistent, and consistent with standard economic decision theory.)

"All very well," you are saying, "but what does it have to do with beans?" What computer do you use? Probably an IBM clone. Why do you use it? Probably because everyone else has an IBM clone. Does everyone else have an IBM clone because it's a superior product? Manifestly no. Why then? Because it got an early initial advantage, and there were increasing returns. What was the initial advantage? Thousands of corporate purchasing agents saying, for forty years, "No one was ever fired for buying IBM". Then there is the case of VHS vs. Beta, and why Los Angeles has smog but no real public transit, and so on.

To put it a bit differently, lock-in can lead to market failure. Suppose that, if we could all agree to switch to some non-QWERTY keyboard layout, we would all, in the long run, be better off; so much better off that we still come out ahead after paying the switching costs. (It is claimed, for instance, that the "Dvorak" keyboard is measurably faster than QWERTY; see West below.) Normally, economists would expect that competition forces participants in markets to adopt the most efficient technology available, but that's not necessarily true in this case. If you switch, but none of your competitors do, you pay the immediate, short-run switching costs, but they don't, and you lose all the extra value from the increasing-returns-to-scale of employing the most common technology. In a word, you lose. (Technically: unless the discounted present value of the gain from adopting the new technology exceeds the switching cost, sticking with the old technology is a Nash equilibrium but not Pareto-efficient.) Some people, who have an unnatural affection for market allocation, find this prospect very troubling, and have devoted great efforts to arguing that, e.g., the QWERTY layout really is the most efficient, and indeed even that Microsoft Windows is the best of all possible operating systems.