Recombinant uncertainty in technological search

From AcaWiki
Jump to: navigation, search

Citation: Lee Fleming (2001) Recombinant uncertainty in technological search. Management Science (RSS)
Internet Archive Scholar (search for fulltext): Recombinant uncertainty in technological search
Tagged: Economics (RSS) patents (RSS), innovation (RSS)


Technological innovation is frequently referred to as uncertain. Fleming's article tries to provide a model that can explain where uncertainty in technological innovation comes from. Building on a Schumpeterian concept of innovation as recombinations, Fleming suggests that technological uncertainty comes from combinations of unfamiliar components. Fleming's results show that although new combinations are also associated with less useful innovations on average, this lower mean is also associated with a higher variation which can explain breakthroughs as well as flops.

Fleming's framing synthesizes two perspectives. First, the classic innovation theory from Schumpeter's The theory of economic development: An inquiry into profits, capital, credit, interest, and the business cycle that innovation can treated simply as recombinations of existing components in novel ways. Second is the idea that innovations are uncertain and can be viewed as a search process that is either more (exploitation) or less (exploration) local building off March's (1991) classic Exploration and exploitation in organizational learning.

Fleming offers and tests five hypotheses:

  1. Recombinations of familiar components will increase an innovation's usefulness. (Supported)
  2. Recombinations of familiar components will decrease inventive uncertainty. (Not supported)
  3. Refinement of familiar combinations will increase an invention's usefulness. (Supported)
  4. Refinement of familiar combinations will decrease inventive uncertainty. (Supported)
  5. Cumulative use of a combination will decrease an invention's usefulness. (Supported)

Fleming, as he does in much of his other work, tests his hypotheses using patent data. He uses incoming citations as his measure of an invention's usefulness or important and as his dependent variable. He creates his independent variables using patent subclasses and takes advantage of the fact that the patent office usually assigns patents into multiple subclasses. He develops variables measuring component familiarity from the average degree to which the components (i.e., subclasses) have been recently and frequently used, combination familiarity and cumulative combination usage similarity.

Results (summarized in regards to the hypotheses above) also show that the reuse of components has a non-monotonic effect which is eventually positive. The story is consistent with the suggestion that the most effective innovations are the ones that are very new combinations of familiar and frequently used components.

Theoretical and Practical Relevance

Flemming's article has been cited more than 340 times since it was published 9 years ago in the literature on technological innovation.

For example, it has been cited by Eric von Hippel in Democratizing innovation to back up his support for the argument that innovators are likely to be users in one field with experience in another who are better able to make the type of high impact recombinations between disparate elements both of which are locally familiar to their two differen spheres of activity.