@RogerBohn complains that The Economist praises a dangerous and obsolete management concept - namely the experience curve or learning curve.
The theory of learning-by-doing was introduced by economists to explain the macroeconomic observation that productivity typically increased during an extended period in which the production processes and technologies remained the same. The theory suggests that these productivity improvements can be related to cumulative production volumes. Economists use the theory to predict that aggregate productivity levels will increase under certain circumstances. However, management consultants have generally used the theory at a microeconomic level, apparently believing that it predicts productivity improvements in specific production units, and that (as Bohn complains) "improvement is inevitable and the same for everyone in an industry".
Bohn points out how productivity in the car manufacturing appears to run entirely counter to the microeconomic interpretation of learning-by-doing. Productivity at Toyota improved far faster than at General Motors, even at a time when it produced (and had cumulatively produced) far fewer cars than GM.
Now here's the twist that may save the theory. Toyota's competitive advantage was in JIT and “The Toyota Production Process,” which Bohn describes as "a system for making more rapid improvement". So Toyota core process wasn't manufacturing-cars, it was improving-manufacturing-cars. Toyota was making improvements at a staggering rate, which left its competitors standing. It's more difficult to count improvements than to count cars (because improvements can be understood in different ways), but it is not hard to believe that Toyota's cumulative number of improvements has been higher than that of GM for a long time now. If we reframe the production system in this way, perhaps the theory of learning-by-doing could apply to this example after all.
But this interpretation of the theory of learning-by-doing is quite different from the conventional interpretation, and would make its use as a predictive tool or as a consultancy tool much more problematic.