Learning by Doing Technical Note

This report describes the benefits delivered to the US Census Bureau as part of the work undertaken in Project LA-444 titled “Learning by Doing: An Investigation†. The project delivered three important benefits. The first benefit of Project LA-444 was estimating learning rates by industry, which were then used to characterize inter-industry heterogeneity in learning by doing, and correlated to aspects of industry competition such as industry profitability, and differences between incumbents and entrants. Scholars could use these industry-by-industry learning rates as another industry-level variable in their studies, much like they use industry R&D intensity or capital intensity. Second, the project created a linked LRD-NBER Patent Dataset, along with an SSEL-NBER Patent Dataset bridge, both of which will help the Census Bureau with furthering their objective of developing a better understanding of innovation. This linked dataset significantly enhances the coverage of patent data -firm data links from about 4,000 unique patent assignees in the original patent dataset linked to Compustat data to more than 48,000 unique assignees, representing almost two-thirds of all assignees during 1975 to 1997. In addition, this dataset allows the patent data-firm data link to vary over time unlike the original patent dataset that is based on the 1989 ownership structure. Finally, this linked dataset was used to generate a number of new descriptive statistics linked to patenting.