Jumping to Hasty Experience Curves

Abstract: The "experience curve" used to be treated as a rule of thumb in strategic management theory. However, it would be hasty to conclude that an approximately 80% log-linear experience curve is a rule of thumb regardless of the industry, company, or product. According to the fundamental research on learning curves, generally, the new model of a product is partly composed of old model parts and we cannot observe the progress of these parts from the beginning of the processes. Therefore, in the first product of the new model, the learning rates of old model parts cannot be identical logically, even if the parts' learning rates are identical at the beginning. In fact, empirical data repeatedly deny identical learning rates for all products. Applying search theory, we obtain an approximate log-linear learning curve and show the curve's initial concavity, that is, if we started observing the progress midstream, and not from the beginning of the process. Thus, both varied learning rate proper to each product and initial concavity of the learning curve are phenomena induced by observing the progress midstream.Keywords: learning curve, experience curve, search theory, initial concavity, progress function(ProQuest: ... denotes formulae omitted.)IntroductionThe continuing production of multiple units of the same product often shows the phenomenon of decreasing production cost per unit. The pattern in which the production cost per unit (or the direct labor hours per unit) decreases at a given reduction rate is known and researched as the "progress function" or "learning curved An early observation of this pattern emerged in aircraft manufacturing (Dutton & Thomas, 1984; Dutton, Thomas, & Butler, 1984). In 1936, T. P. Wright, aeronautical engineer for Curtiss-Wright Corporation, published a widely referenced paper (Wright, 1936) on the learning curve.The learning curve is generally expressed by the progress function f: the production cost2 of the nth unit of the product is formulated as...where the n of the nth unit of the product is the cumulative number of units produced, and a and b are constants (a > 0, -1With respect to the progress function for the log-linear model, when the cumulative unit number doubles, the production cost per unit declines 2b times:...where p = 2b is the learning rate, and 1 - p = 1 - 2b is the progress ratio.3 For example, when b = -0.322, the progress ratio (1 - p) is 20% and the learning rate (p) is 80%, which means an 80% learning curve. Following the progress function, the production cost per unit declines to 100 p % with each doubling of cumulative volume. During World War II, progress functions were applied not only to the aircraft industry but also to the shipbuilding industry in an attempt to use the data created by government shipbuilding contractors (Searle, 1945; Yelle, 1979).From Learning Curve to Experience CurveAfter World War II, Andress (1954) advocated that the application of learning curves developed in the aircraft industry during World War II would result in more accurate and easier forecasting of production costs. Thus, he insisted that their application is valid in industries other than aircraft, namely, electronics, home appliances, residential home construction, shipbuilding, and machine shops. Further, Hirschmann (1964), whose basic argument relies on Andress (1954), advocates that the learning curve is applicable in petroleum refining, to which Andress disagreed. He insisted on the application of the learning curve at the industry level in the petroleum, electric power, and steel industries.The progress of assembly work is much faster than that of machine work, and approximately 75% of the total direct labor input in the aircraft industry is assembly work. …