Combinations of technology in US patents, 1926–2009: a weakening base for future innovation?

In combinatorial models of innovations, new technologies are built from combinations of pre-existing technological components. Researchers learn which components work well together by observing previously successful combinations and the pool of ideas can be ‘fished out’, i.e. exhausted, if it is not ‘restocked’ by the discovery of novel connections. We first show US patents have made increasingly less novel connections among technological constituents since the 1950s, and that the number of technological fields to which these connections are applicable has stopped growing since the 1980s. We then estimate the parameters of an ideas production function, and find parameter estimates consistent with technology fields being fished out if not continually restocked by the discovery of novel connections between technological components. We use the ideas production function to estimate the number of new patent applications induced by each patent granted between 1926 and 2001, and show this number has trended downward since the 1940s.