The Age of Innovation

For not-so-surprising reasons, I’ve been thinking lately about lifecycles. My AFR op-ed today (partially written with a newborn babe in the crook of my arm) is on age and creativity. Full text over the fold.


Age No Bar to Brilliance, Australian Financial Review, 30 June 2009

Einstein’s major contributions to physics were published when he was aged 26. Mathematician Terence Tao won the Fields Medal (maths’ Nobel Prize) at age 31. ‘When Mozart was my age, he had been dead for two years’, quipped 37 year-old Tom Lehrer.

Yet at the other end of the lifecycle, examples abound. Frank Lloyd Wright designed the Guggenheim Museum from ages 76 to 91. Clint Eastwood directed Unforgiven at age 62. Paul Cézanne’s most valuable work was painted in the year of his death, aged 67.

Understanding the lifecycle of innovators is a puzzle with major implications for how we fund researchers and artists. Should we devote more towards early-career innovators, and risk wasting it on fizzling fireworks? Or is it better to look for established track records, at the risk of funding extinct volcanoes?

One researcher who has been studying the lifecycle of innovation across a variety of fields is University of Chicago economist David Galenson. Over more than a decade, Galenson and his co-authors have studied the careers of economists, poets, novelists, directors, architects, and artists. To identify the stars of each field, he gathers empirical data: rankings of ‘all time greats’, auction prices, prizes, citations, and even the number of times works appear in textbooks and anthologies.

Across a diverse set of fields, Galenson claims to have consistently identified two types of innovators: ‘conceptual’ innovators, whose work implements a single theory, and ‘experimental’ innovators, whose work evolves from real-life experience and empirical observation. Like Isaiah Berlin’s hedgehogs (who relate the world to a single vision) and foxes (who pursue many ends), Galenson’s dichotomy can be applied across many fields of creative endeavour. And the recurring pattern seems to be that conceptual innovators do their best work at an earlier age than experimental innovators.

What marks a conceptualist from an experimentalist? In art, Galenson distinguishes conceptual artists (Pablo Picasso, Edvard Munch) whose work aims to communicate specific ideas and emotions; from experimental artists (Edgar Degas, Wasily Kandinsky) whose ideas are vaguer, and often regard the artistic process as a journey.

Among architects, conceptualists are motivated by geometry (Renzo Piano, Walter Gropius), and experimentalists are inspired by nature (Frank Lloyd Wright, Frank Gehry).

For novelists, conceptualists have specific goals, and are best known for their plots (F. Scott Fitzgerald, Ernest Hemingway). Experimentalists are more focused on character development (Charles Dickens. Virginia Woolf).

In poetry, conceptualists (E.E. Cummings, T.S. Eliot) are technically sophisticated and grounded in literary history, while experimentalists (Wallace Stevens, Robert Frost) draw more from ordinary speech and observation.

Among great directors, conceptualists (Steven Spielberg, Stanley Kubrick) are those whose movies are carefully planned and animated by single ideas. Experimentalists (Robert Altman, Woody Allen) are generally less sure on their goals, and often make major changes to the movie during shooting.

And for Nobel-prize winning economists, conceptualists are theorists and methodological innovators (Paul Samuelson, Kenneth Arrow), while experimentalists make contributions that are principally empirical (Simon Kuznets, Robert Fogel).

Across these fields, a similar pattern can be seen – conceptual innovators tended to do their best work at a younger age than experimental innovators. Conceptual poet T.S. Eliot penned “The Love Song of J. Alfred Prufrock” at age 23. Conceptual novelist F. Scott Fitzgerald never regained the success of The Great Gatsby, published when he was 29. Among experimentalists, Wassily Kandinsky painted his best work around age 50, while economist Robert Fogel published his most cited work, Without Consent or Contract, at age 63.

Yet while Galenson’s research studies the relationship between age and the type of innovation, another approach is to ask whether the peak age has changed over time. Work by Benjamin Jones (Kellogg School of Management) analyses the age at which Nobel laureates made their prize-winning contribution.

Across physics, chemistry, medicine and economics, Jones finds that the age at which laureates made their greatest contribution has shifted upwards over time. In the early-twentieth century, the typical prize was given for work when the winner was aged in their late-thirties, but these days it is typically given for work done in the forties.

Looking carefully at lifecycles, Jones concludes this is due to the increasing educational burden that each generation of innovators imposes on their successors. While the great minds of the early-twentieth century became research-active at age 23, those of the late-twentieth century only became research-active at age 31.

As the knowledge frontier moves out, it takes longer for the next generation to attain it. Using patent data, Jones also shows that as science has become more complex, there are more collaborations and fewer ‘renaissance men’ making contributions across different sub-fields. In the absence of a paradigm shift, the normal process of scientific accumulation steadily moves the knowledge frontier outwards.

While lifecycle economics has policy implications, it also naturally leads to introspection. I couldn’t resist asking 37 year-old Jones how studying age-creativity patterns has made him view his own research. He responded: ‘My studies suggest my current age is one of peak productivity – and soon there will be a decline – so I’d better get back to work!’

Andrew Leigh is a 36 year-old economist at the Australian National University.

(To avoid example-overload, the paragraph on movie directors didn’t appear in the AFR version.)

I found this a fascinating literature to delve into. The way that Galenson marshals evidence to create his ranking lists is creative, though invariably imprecise. From interviews, I’ve heard that experts in the field generally hate the notion that there can be any unidimensional ranking of talent, and are horrified by the idea of counting mentions in anthologies to derive that ranking.

For those who’d like to read more, Galenson’s hyperlink takes you to the dozen or so NBER papers that he’s written on this theme (available only to those at edu or gov sites, sorry). Galenson has also published a book summarising his work. Ben Jones’s hyperlink takes you to the papers on his website.

I also had an email exchange with Ben Jones (who has a PhD in economics from MIT and a bachelor’s degree in aerospace engineering at Princeton). I couldn’t do justice to Ben’s reply in the article, but he’s given me permission to post it below.

AL: Do you have any theories for the cross-sectional differences in age at great innovation between disciplines. For example, in Fig 7 of your ReStud paper, it looks like medical Nobellists now get the prize for work done around age 40, but chemists for work done around age 45. Also, people frequently talk about great maths innovators being young. Can we say anything systematic across the sciences (as Galenson does across the creative arts) about why young people tend to dominate some fields but not others?

BJ: There are two likely answers to this question. First, there are cross-sectional differences in the amount of foundational knowledge. A field with less foundational knowledge makes it easier for younger scholars to reach the knowledge frontier and actively produce great ideas. One example is early 20th century physics, when several empirical anomalies appeared that undercut classical physics. These anomalies, which ultimately resulted in the quantum mechanics revolution, opened the door to younger scholars, who did not require much pre-requisite knowledge to make signature contributions at young ages. Heisenberg, for example, made his key contributions at ages 23 and 26, even though he had nearly failed his doctoral exams on account of his poor knowledge of classical theory.

The ease of access to the knowledge frontier can also help explain some contemporary age phenomena. For example, we have seen very young entrepreneurs make leading, early contributions in computer software, internet search, and social networking, and more generally in .com enterprises. But we have not seen this so much in chemical synthesis, solid state physics, genomics, nanotechnology, or other high-tech fields. A key distinction is that the early days of software (Bill Gates and Microsoft), internet search (Sergey Brin, Larry Page, and Google) were in new fields with relatively little foundational knowledge, making it easy to be at the frontier at young ages.

Second, there appears to be a bias toward younger scholars when theoretical reasoning is relatively important, as is suggested in Galenson and Weinberg’s work. For example, the prevalence of theory, in tandem with measures for the amount of foundational knowledge, are substantial predictors of the age at which Nobel Prize winners have produced their great ideas in physics, chemistry, and medicine.

Lastly, it is worth nothing that, cross-sectional differences across fields are not stable. For example, in the early 20th century, great achievement in physics came earlier in life than great achievements in chemistry, but the end of the 20th century this pattern had reversed. The amount of foundational knowledge and the relative preponderance of theory are always evolving – differentially within fields – and shift the age-creativity relationship dramatically over time.