Thompson (2007) attributes most of the Liberty-ship literature's inflated forgetting estimate to a named aggregation bias in Fischer's (1949) standard shipyard dataset

economicslearning-curveorganizational-forgettingliberty-shipspeter-thompsonfischer-1949shipbuildingdata-quality

Thompson identifies a specific reason the standard data source behind the Liberty-ship forgetting literature — Fischer's (1949) official USMC shipbuilding summary, the same yard-level data underlying Argote, Beckman & Epple (1990) and Leonard Rapping (1965) — produces an inflated forgetting signal. Many yards periodically switched labor onto non-standard vessels (tankers, colliers, troop-ship conversions) while continuing to report that labor against Liberty-ship output: "output data derived from Fischer (1949) assign all this labor to Liberty ships." Because reported output falls while reported labor holds roughly constant during these switches, "the net effect is to give the appearance of a dramatic decline in productivity as the rate of Liberty ship production declined" — a pattern indistinguishable, in the standard specification, from organizational forgetting, but actually a mislabeled product-mix change. Thompson generalizes the point in his own conclusion: "unobserved changes in a firm's product mix can produce spurious evidence for organizational forgetting."

His new dataset — built from unit-level National Archives records rather than Fischer's yard-level monthly aggregates — is what lets him strip the bias out, and doing so drives roughly half of the reduction from his own unadjusted loglinear estimate of 8.4 percent down to 4.2 percent once product mix is corrected for (see the full six-specification range). This is the mechanism behind Thompson's much lower re-estimate relative to Argote, Beckman & Epple's 25 percent: not a different statistical technique applied to the same numbers, but a different, more granular dataset that avoids attributing another vessel type's labor to Liberty ships.