A prominent American economist and Nobel Prize winner, George J. Stigler, developed an intriguing method for estimating long-run costs.53 Stigler believed that the use of accounting data, with all their distortions and subsequent need for adjustments, made the validity of cost estimation based on such data questionable. His method was to observe an industry over time, categorize the firms in the industry by size (measured as a percent of total industry capacity or output), and then arrive at a conclusion regarding cost efficiency based on the relative growth or decline of these size categories. His results for the steel industry (using data for 1930, 1938, and 1951) showed that medium-size firms (defined as between 2.5 percent and 25 percent of industry capacity) appeared to have gained in share of total industry output over the 21-year period, from 35 percent to 46 percent of total, whereas small firms (less than 2.5 percent of capacity) and large firms (actually just one firm, with more than 25 percent of capacity) lost market share. Stigler concluded the existence of a U-shaped long-run average cost curve whose path first showed net economies of scale, than constant returns, and finally diseconomies of scale.