Breakthrough Innovations and Productivity Over the Long Run

After the 2008 financial crisis, many modern economies have suffered from persistent economic stagnation. Developed countries have not been as effective as previously in promoting long‐term productivity growth, There is a mounting evidence that the slowdown is due to the difficulty of finding new ideas and a more conservative research strategy.

OECD, Labour Productivity Growth 2001-2021
OECD, Multifactor Productivity Growth, 1985-2020

The slowdown might be rooted in the availability of potential breakthrough innovations and the incentives for creating them. This is the idea explored in a new paper1 (my emphasis).

There seems to be a gap between the production of ideas and the ability to translate these ideas into productivity growth in advanced economies. Most research projects fail to produce groundbreaking results, and only a few generate impactful breakthrough scientific discoveries. This may lead to insufficient scientific funding (…) Scientists pursue progressively less risk, as riskier research leads to more failures, and this reduces publications. However, the authors find that publication of failures leads to speed of discovery and in that way, breakthrough innovations. Thus, science policies could improve discoveries by subsidizing riskier strategies, encouraging the publication of failed experiments, and incentivizing strategy diversity, as a record of failures provides an understanding of research behavior and improves its efficiency.

Breakthrough innovations represent distinct improvements in the technological frontier and which become the new foundation upon which subsequent inventions are built. But how do we identify potential new breakthrough ideas?

If citation data were objectively determined and consistently available, a breakthrough innovation would receive a large number of future citations. Given the absence of consistently available citation data, Bryan Kelly et. al. instead propose a measure that is similar in spirit that can be constructed by analyzing the text of patent documents2.

Based on text analysis and machine learning, patents are identified that differ from previous patents (novel) but are broadly similar to later patents (impactful). The most relevant patents represent clear improvements in technology, and innovations have been built on them. Therefore, it is justified to talk about breakthrough patents.

We use advances in textual analysis to create links between each new invention and the set of existing and subsequent patents. Specifically, we construct measures of textual similarity to quantify commonality in the topical content of each pair of patents. We then identify significant (high quality) patents as those whose content is distinct from prior patents (is novel), but is similar to future patents (is impactful). Since our indicators of the significance of a patent require no other inputs besides the text of the patent document, they are consistently available for the entire history of US patents spanning nearly two centuries of innovation (1840 – 2010).

we define a `breakthrough’ patent as one that falls in the top 10% of the quality distribution (among all patents in all years).

Figure displays the similarity network for four patents. Op. Cit.

The data show that there is a significant drop in the number of breakthrough patents in the early 2000s. After strong growth in previous decades, the drop is seen in particular in the ICT sector, although it can be observed in other sectors.

Based on an econometric analysis, the slowdown in innovation activity has a clear temporal connection with the later slowdown in productivity in the 2010s in sectors where the number of breakthrough patents decreased.

Based on this analysis, one breakthrough patent is linked to an increase of about 2% in labor productivity in the corresponding industry after 10 years.


(1) Kuusi, Tero, and Jenni Nevavuo. ‘Breakthrough Innovations and Productivity: An International Perspective’. ETLA Working Papers, ETLA Working Papers, 31 March 2023.

(2) Kelly, Bryan, Dimitris Papanikolaou, Amit Seru, and Matt Taddy. ‘Measuring Technological Innovation over the Long Run’. Working Paper. Working Paper Series. National Bureau of Economic Research, November 2018.

Featured Image: scientific and technological breakthroughs be accelerated to improve the human condition, Midjourney @carliwi

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.