[Image: iStock]
Qualitatively, a ferment of patent activity is a clear suggestion that a technology is on the move. But can patent data provide any quantitative clues about which technologies are improving fastest? Chris Benson and Chris Magee of the Massachusetts Institute of Technology think they have an answer, based on a recently published analysis of key metrics from half a million patents across 28 domains (PLOS One, doi: 10.1371/journal.pone.0121635)
The punch line: A handful of early-stage metrics from patents do seem highly correlated with subsequent rapid improvement in the technology at hand—and could, in the researchers’ view, form a toolkit for investors, venture capitalists and labs looking to identify potential breakout areas.
One challenge of working on patents is the sheer size and complexity of the data set. Benson and Magee attempted to make the problem more tractable by focusing on 28 broad technological domains—including a number of areas tied directly or indirectly to optics and photonics, such as 3-D printing, LEDs, optical storage and transmission, photolithography and solar energy. The 28 areas they selected represented, according to an earlier analysis by the same team, more than 10 percent of the U.S. patent database.
Next, the team had to mine that database for the patents most relevant to each specific technology domain—a potentially laborious task, given the diversity of patents and overlaps across technological areas. Benson and Magee fashioned a novel approach to identifying these relevant patent sets that involves analyzing the overlap of identified U.S. and non-U.S. patent classes for a given patent. The result was a non-overlapping set of patents to represent each technology domain.
With the data set in hand, the MIT team next framed ten separate hypotheses potentially tying specific patent metrics to the speed of subsequent progress in the technology. The metrics ranged from the simple number of patents within the technological domain, to more complex or subtle notions such forward and backward citation rates and the fraction of patents from other domains. They developed numerical methods to calculate each metric from the patent data, plotted the results against the technology’s documented rate of improvement (calculated using a well-established technique), and attempted to quantify the correlation.
Benson and Magee found that, looking at technological improvement over the past 12 years, three metrics in particular had a strong correlation with the technology’s rate of progress. The metrics were: (1) the average number of citations to within the specific domain; (2) the number of times the patent is cited by subsequent patents within the first three years of its lifetime; and (3) the proportion of relatively recent patents within the technology domain.
The method correctly identified optical and wireless communications, 3-D printing, and MRI technology, from their patent data, as areas that are rapidly improving. Domains such as batteries, wind turbines, and combustion engines appear, according to the analysis of Benson and Magee, to be improving at slower rates.
The MIT team believes that the method they’ve outlined could result in a Standard & Poor’s-like rating system for technologies, allowing for a more informed idea of how fast new areas are likely to mature. But Benson cautions that the method, as a prediction tool, has a degree of nuance.
“I don’t see it as something to hand out to the masses to play with,” he says, instead envisioning it as a tool that the researchers would use to help investors and others understand how future technological capabilities are evolving. “We're probably more like a real estate agent, and less like Zillow.”