Does knowledge acquired through learning by doing persist through time or does it decay or depreciate? The classic learning curve model described in Chap. 1, which uses cumulative output as the measure of experience, assumes that knowledge acquired through learning persists indefinitely through time. More recent research indicates, however, that this characterization might not be accurate.
Several case studies reported that when production was resumed after an interrup- tion at a manufacturing firm such as a strike, unit costs were higher than the level achieved before the interruption (Baloff, 1970; Hirsch, 1952). In addition, De Holan and Phillips (2004) found evidence of forgetting in their case studies of hotels in Cuba. These case studies suggest that organizations might not retain all the knowledge they acquire indefinitely through time and that organizational “forgetting” can occur. Organizational forgetting has important consequences for organizational perfor- mance. If forgetting occurs, organizations will not be as productive in the future as they anticipate. That is, if there is forgetting, forecasts of future production based on the classic learning curve will overestimate future production. Failure to achieve expected levels of productivity can lead to large problems for organizations. Delivery commitments might not be met. Customers can become dissatisfied. Significant financial penalties for late deliveries might be incurred. Inaccurate forecasts of future productivity make it very difficult for organizations to plan and organize their internal operations. Further, strategic analyses based on inaccurate forecasts of future productivity can be very misleading. In extreme cases, an organization’s actual productivity is so far below its expected or forecasted productivity that the organization is not competitive. Thus, if forgetting occurs, it is very important for the organization to allow for this forgetting in forecasts of its future productivity. Further, the organization should consider strategies for minimizing forgetting. These strategies for retaining knowledge are discussed in the following chapter on organi-zational memory.
Theoretical papers (e.g., Batchelder, Boren, Campbell, Dei Rossi, & Large, 1969; Carlson & Rowe, 1976) and simulation results have developed the theoretical implications of forgetting for forecasting, planning, and scheduling (e.g., see Smunt, 1987; Smunt & Morton, 1985; Sule, 1983). Yet empirical studies of organizational learning typically assume that there is no forgetting and use cumulative output as the measure of organizational experience [e.g., see the many studies in the reviews by Dutton and Thomas (1984) and Yelle (1979)]. Similarly, forecasts of future pro- ductivity based on the learning curve also do not account for forgetting. Concerns about the possibility of forgetting coupled with the absence of its consideration in empirical studies and forecasts led my colleagues and me to embark on a research program on the extent of forgetting in organizations. Such knowledge would advance our understanding of organizational learning at a theoretical level as well as lead to important information that would enable organizations to improve forecasts and ideally increase their productivity.
My colleagues and I wanted to test empirically whether organizational knowl- edge was cumulative, as the classic model implied, or whether it depreciated, as the Lockheed example suggested. Dennis Epple, Sara Beckman, and I developed a method for generalizing the classic learning curve to analyze empirically whether knowledge decays or depreciates (Argote, Beckman, & Epple, 1990). The conven- tional measure, cumulative output, implies that there is no forgetting or deprecia- tion: experience obtained from a unit of output produced in the distant past is as useful as experience obtained from a unit produced yesterday. My colleagues and I tested this assumption by developing a measure of knowledge that embeds the con- ventional measure as a special case. We introduced a parameter, lambda, to form a geometric weighting of an organization’s past output. Estimates of lambda equal to one correspond to the classic cumulative output measure, whereas estimates less than one imply forgetting or depreciation because output in the distant past receives less weight in predicting current productivity than recent output. We have used this approach to test empirically whether depreciation occurs in different industries, most notably shipbuilding, automotive, and fast food.
Before turning to the empirical studies, a much publicized production program that appeared to evidence forgetting—Lockheed’s production of the L-1011 TriStar—is described. My discussion of the TriStar program is based on informa- tion publicly available in newspapers, trade publications, the company’s annual reports, and the like. My intention in discussing the TriStar program is not to criti- cize Lockheed. Indeed, the L-1011 is widely regarded as a major technological success. My intent, rather, is to demonstrate that predictions of Lockheed’s produc- tivity based on the classic learning curve were dramatically off the mark. Thus, the classic learning curve was inadequate to describe the Lockheed production pro- gram. The pattern of costs Lockheed experienced is consistent with a model in which depreciation occurred. Background information about the Lockheed case is presented to illuminate reasons why knowledge depreciated there. Some of the background information is described to provide context for the example; other information is more suggestive of causes of knowledge depreciation.
Source: Argote Linda (2013), Organizational Learning: Creating, Retaining and Transferring Knowledge, Springer; 2nd ed. 2013 edition.