Lindy Effect
Under some survival models, the expected remaining life of a book, idea, or technology increases with how long it has already lasted. The Lindy effect uses age as evidence of possible persistence, subject to assumptions about the kind of thing and its failure process.
THE IDEA
The Lindy Effect
A practice can survive for centuries even when its purpose is difficult to explain. Taleb argues that such survival deserves attention when estimating how long the practice might persist. Longevity is evidence about survival; it does not establish that a belief is true or that a practice is beneficial.
The Lindy effect describes cases in which expected remaining lifetime increases with age. Taleb connects the idea with Mandelbrot’s work and distinguishes perishable individuals from potentially long-lived works, ideas, and technologies. But being nonperishable is not enough to establish the effect. A process can have no fixed maximum lifespan and still have a constant expected remaining life.
One model that produces Lindy behavior has a power-law survival tail. For a Pareto survival tail with exponent α greater than one, expected remaining lifetime equals current age divided by (α − 1). With other exponents, the mean may not be finite. An exponential model instead has constant expected remaining lifetime, regardless of age. Choosing between these models requires evidence; age alone does not settle the choice.
The definition of the item matters too. A “convertible car,” a “car,” a “bound book,” and a “book” refer to different things. A physical copy can deteriorate while a work continues through new editions. A technology can persist while individual devices wear out. Estimates depend on which of these is being tracked and what counts as its disappearance.
Waiting-time examples, including earthquakes, also depend on the model and the events being counted. Increasing expected waiting time is possible under some distributions; it is not a general rule that every earthquake becomes more distant the longer people have waited.
The argument does not justify dismissing newly recognized diseases or psychological categories. Labels such as “Type A personality” or “passive aggressive” need evidence appropriate to the claims made about them. How recently a condition was named does not establish whether it is real, harmful, or treatable.
Printed books and electronic files illustrate another distinction. Some old books remain readable while some much newer files become inaccessible. That comparison raises questions about preservation and dependence on formats, but it does not by itself estimate the survival distribution of books or files.
As a heuristic, Lindy reasoning asks what past survival tells us about future persistence. Its usefulness depends on a relevant comparison group, a clear definition of survival, and reasons to believe that the conditions supporting past survival will continue.
When to use it
Evaluating the longevity of a new technology or business model.
Deciding whether to invest in a new or established idea.
Assessing the robustness of cultural practices and traditions.
Assessing whether a longevity argument is being used beyond its scope, such as to dismiss a newly named condition.
Scientific ideas: distinguish a theory’s continued acceptance from the evidence that supports its claims.
Company analysis: examine what has supported past survival and whether those conditions persist; longevity alone does not demonstrate antifragility.
Deciding which cultural works to study: works that have survived centuries are more likely to remain relevant.
Understanding the distribution of waiting times in complex systems (e.g., earthquakes, wars).
FROM THEORY TO EVERYDAY LIFE
In practice
Limits of Survival Evidence
An anecdote about barefoot running does not test the Lindy effect. To use longevity as evidence, define the practice or technology whose survival is being measured and compare it with relevant alternatives.
Television comedian's career
A television comedian’s working career is constrained by a human lifespan. A recorded performance may continue to attract an audience after that career ends. The career and the work are different objects for a longevity estimate.
Books vs electronic documents
A five-hundred-year-old book remains readable, while electronic files from a decade ago may be irretrievable.
Newly Named Conditions
A condition’s recent name does not show that it was recently invented or that it is a fad. Assess the evidence for the condition directly; the Lindy effect cannot establish a diagnosis’s validity.
Book Longevity
In a simplified Lindy illustration, a book in print for 50 years might have an expected remaining print life of another 50 years. That equality depends on the chosen model; it is not a guarantee for an individual book.
Technologies and Companies
A long-lived institution might be modeled as having an expected remaining life that grows with age. In a simplified illustration, an institution that has lasted 100 years might be expected to last another 100 years. Applying this to a particular company or university requires evidence that the model fits and that relevant conditions have not changed.
A MOMENT TO REFLECT
Put it to work.
What technologies or ideas have persisted for a long time, and how might the Lindy effect predict their future?
Why is the age of a diagnostic label insufficient to judge whether the condition is real or clinically useful?
How does the Lindy effect change your view on modern inventions versus ancient traditions?
How can the Lindy effect guide your selection of books, technologies, or ideas to invest time in?
Why might a long-lasting institution be more likely to survive future shocks than a younger one?
What are the risks of applying the Lindy effect to perishable items (like humans) as if they were non-perishable?
Often confused with
References
- The New Republic article (1964)
- Mandelbrot, B. (1997). Fractals and Scaling in Finance — Mandelbrot (1997)
- Gott, J. R. (1993). Implications of the Copernican principle for our future prospects — Gott (1993)
- Gott, J. R. (1994). Future longevity of the human species — Gott (1994)
- Caves, C. (2000). Predicting future duration from present age — Caves (2000)
- Rees, M. (2003). Our Final Hour — Rees (2003)
- Bostrom, N. (2002). Anthropic Bias — Bostrom (2002)
- Pigolotti, S., et al. (2005). Stochastic processes and power-law distributions — Pigolotti et al. (2005)
- Arbesman, S. (2011). The life expectancy of scientific ideas — Arbesman (2011)
- Khmaladze, E., et al. (2007, 2010). Survival of empires — Khmaladze et al. (2007)
- Taagepera, R. (1978, 1979). Size and duration of empires — Taagepera (1978)
- Turchin, P. (2003, 2009). Historical dynamics — Turchin (2003)
- Sornette, D. and Knopoff, L. (1997). The paradox of the waiting time for earthquakes — Sornette and Knopoff (1997)
CONTINUE THE CONVERSATION
Source texts
- Antifragile (Nassim Nicholas Taleb)
- Antifragile (Nassim Nicholas Taleb)