What drives economic progress? The answer to that question remains something of a mystery.
to your inbox.
We’ll send you one email a week with content you actually want to read, curated by the Insight team.
Since the 1800s, productivity—the amount of value a company creates per dollar of investment or hour of worker labor—has increased in the U.S. Indeed, this long-term trend has been largely responsible for economic growth nationwide over the last two centuries, allowing incomes to rise even as the workweek has grown shorter and production less capital-intensive.
But it is not clear why productivity grew throughout this period—let alone why it grew faster in some years than others, or why different industries have experienced particular ups and downs. That’s a puzzle that Kellogg professor of finance Dimitris Papanikolaou has thought a lot about.
“What’s driving productivity differences over time and across firms and sectors?” Papanikolaou asks. “People say, ‘Oh, it’s innovation, technological change.’ But we don’t have great measures of those things. That’s what motivated our research.”
The researchers measured the quality of each patent by comparing how similar its text was to earlier and later patents—the idea being that highly innovative patents should be unlike anything that came before, but influential enough that later inventors would adapt and expand on them. By counting how many exceptionally innovative patents are filed in a particular year, they could estimate how much breakthrough innovation happened that year.
The researchers found that changes in the level of breakthrough innovations were associated with changing productivity levels across time periods, industries, and firms. Moreover, spikes in patent quality successfully predicted watershed inventions as well as individual firm profits, suggesting that innovation might indeed be key to understanding the last two centuries of economic growth.
“We provide robust evidence that technological progress is correlated with productivity,” Papanikolaou says, “by creating the first direct measure of technological progress that is comparable across time and space.”
How to Measure Innovation
Innovative new technologies can improve productivity by making work faster, cheaper, and easier. But innovation has typically been a difficult thing for researchers to measure, according to Papanikolaou.
Most studies of innovation have relied on patents. “Not all innovations are patented, but patents by definition are related to invention,” he says. “That seemed like a useful starting point.”
“The first patent that mentions the word ‘electricity’ is going to be important. The thousandth patent that does this, less so.”
But the challenge is that many patents are “frivolous or useless,” Papanikolaou notes. Moreover, since laws and regulations governing intellectual property change over time, there may be reasons unrelated to invention that explain why different numbers of patents are issued each year. This means simply counting patents is not a valid measure of innovation.
Papanikolaou’s team “wanted to develop a metric to weigh different patents differently” by focusing on their innovativeness, he says—what could be thought of as a patent’s quality, or “breakthrough factor.”
To understand the innovativeness of an invention, previous studies have typically measured how many times its patent is cited by other, subsequent patents. Unfortunately, simply counting citations is, once again, a flawed approach: citation patterns have changed over the decades, Papanikolaou says, “and it’s not obvious how to compare patents from different times.”
“An analogy would be if you measure an online article’s reach or popularity by how many clicks it gets,” he says. “It’s easy to compare two articles published on the same day. But how can you compare an article published in 1990 to one published today? One has been around much longer, and the other might benefit because more people access the Internet now.”
Moreover, the U.S. Patent Office did not even record citations prior to the mid-1940s, making it impossible to use this measure to draw conclusions about the distant past.
That meant the authors had to find a new way to measure a patent’s quality.
Analyzing Patent Text
The answer Papanikolaou and colleagues devised lay in the actual text of patents.
The researchers reasoned that the words used in a groundbreaking patent should be quite unlike that of any previous patent.
“For example,” Papanikolaou says, “the first patent associated with electricity is going to be very important. Electricity itself is not patented. So the first patent that mentions the word ‘electricity’ is going to be important. The thousandth patent that does this, less so.”
Furthermore, if an invention was a true breakthrough, then subsequent inventors should build on it, meaning later patents will have similar text. Thus, the researchers assigned higher quality scores to patents with text that did not resemble earlier patents but did resemble subsequent ones.
The researchers analyzed the text of more than nine million patents filed with the U.S. Patent Office since 1836, drawing on state-of-the-art approaches. After compiling every word used in every patent, they filtered out extremely common words like “the” as well as technical jargon. That left them with roughly 1.7 million terms, which they then weighted by relative frequency. The idea is that words like “petroleum” and “electricity,” which are less common than words like “process” or “invention,” also tend to say more about a particular invention—thus, they were weighted more heavily.
“We found that, over time, periods where there were a lot of influential patents were followed by periods of high productivity growth.”
The researchers also accounted for how language changes over time. “Innovation is often associated with the creation of new words,” Papanikolaou says. “Two patents are going to be very related if they both used the word ‘electromagnetic’ back in the 19th century. But if they both use that word now, they’ll be considered less related than in the past.”
The researchers first validated their measure of patent quality. As expected, the measure successfully predicted how many citations a patent would receive. What’s more, the quality measure performed even better than number of citations in predicting watershed inventions of the 19th and 20th centuries, from the telegraph to television to plastics to genetics-based innovations.
Next, the authors used their measure to test the link between innovation and productivity.
“We found that, over time, periods where there were a lot of influential patents were followed by periods of high productivity growth,” Papanikolaou says.
Clusters of productivity-boosting patents correlated closely with three major waves of U.S. technological progress: the late 19th century’s so-called Second Industrial Revolution (which saw new railroad technology and the birth of electricity), the 1930s (new applications of electricity and the development of chemicals, including Bakelite, the first fully synthetic plastic), and the 1990s (revolutions in computers, communication, and genetics).
The link between innovation and productivity held across sectors of the economy. For example, during periods when a given sector saw 30 percent more innovation than average, that sector tended to see up to 11% higher productivity.
And individual companies reaped the benefits of productivity boosts, too. A typical company that patented a breakthrough innovation could enjoy about 5% greater future profitability than peers without any breakthroughs.
The Future of Patents and Innovation
Papanikolaou believes that the innovation metric that the team created opens up many new avenues of research. “Now we can use this measure to explore other patterns, such as how much productivity is driven by changes in technology versus regulatory changes, shifts in market power, or other factors,” he says.
But the research can also help make sense of the present era. “There’s a big debate right now about whether productivity is slowing down or not,” Papanikolaou says. “If so, we want to reverse it, if possible. But in order to do so, we first have to understand the causes of the slowdown. ”
Sachin Waikar is a freelance writer based in Evanston, Illinois.
Bryan Kelly, Dimitris Papanikolaou, Amit Seru, and Matt Taddy. 2018. “Measuring Technological Innovation over the Long Run.” NBER Working Paper No. 25266.
Coworkers can make us crazy. Here’s how to handle tough situations.
Plus: Four questions to consider before becoming a social-impact entrepreneur.
Finding and nurturing high performers isn’t easy, but it pays off.
A Broadway songwriter and a marketing professor discuss the connection between our favorite tunes and how they make us feel.