As a simple story, the learning by doing hypothesis does a nice job, but it does not tell the whole story. A piece of research by Lanier Benkard established this empirically in 2000. In the paper (published in the American Economic Review), Benkard rejected a simple model of learning in favor of one where organizations learn, but that learning depreciates over time. In other words, some learning is "forgotten" as time passes.
There are two reasons that organizations might forget. First, employees leave the company intermittently, and as they leave, they take their knowledge with them. Second, employees might just forget how to do the tasks as effectively as a couple of months back.
To test for the existence of learning and the extent of forgetting, Benkard used data on the labor costs for producing commercial aircraft. Benkard found evidence in the data that both learning and forgetting were prevalent:
- At the introduction of a new line of aircraft, the firm's costs would start high and then decrease dramatically.
- If a new line of aircraft is very similar to a line that was produced in the past (i.e., many of the same techniques or parts can be reused), Benkard found that there was not much loss in the gains from learning how to produce the earlier models.
- Finally, Benkard found that costs increased as drastically different new aircraft were introduced.