Nassim Taleb

Factual · Economics and markets

“Practically all economic and informational variables have been shown since the 1960s to belong to the D2 class”

Nassim Taleb, paper author · 2019 · On the Statistical Differences between Binary Forecasts and Real World Payoffs (arXiv)

Since the 1960s, practically all economic and informational variables have been shown to belong to the D2 class, or at least the intermediate subexponential class, which includes the lognormal.

Mostly false False in substance, right in a detail.

The second judge would not resolve this claim; the verdict shown is the first judge's.

What would count as true or false

Judge as of 2019. True: the empirical literature since the 1960s (e.g. Mandelbrot 1963, Fama, and later studies of returns, income, wealth, firm size, city size, web traffic) has shown that nearly all the economic and informational variables studied are subexponential, meaning power-law or heavy-tailed, including lognormal, rather than thin-tailed. Mostly true: this holds for the great majority of variables studied, with a few well-known exceptions or ongoing disputes about whether some are power law or lognormal, which does not matter here since the claim covers both. Mixed: heavy tails are well established for a substantial share of economic variables (e.g. financial returns, wealth, firm size) but many major variables have not been shown to be subexponential or are found to be thin-tailed. Mostly false: heavy tails are well documented only for some variables (e.g. asset returns and wealth), so 'practically all' overstates a narrower finding, or the evidence is inconclusive. False: the literature shows that most economic and informational variables are thin-tailed (Gaussian or exponential class).

Evidence

1. Power Laws in Economics: An Introduction American Economic Association (Journal of Economic Perspectives) · February 2016

   “Many of the insights of economics seem to be qualitative, with many fewer reliable quantitative laws. However a series of power laws in economics do count as true and nontrivial quantitative laws—and they are not only established empirically, but also understood theoretically.”

2. Was Pareto right? Do the US wealth and income distributions follow power laws? Greenwich Papers in Political Economy (University of Greenwich) · 2023

   “Nevertheless, a sizeable empirical literature fails to find support for them in the data.”

3. The Variation of Certain Speculative Prices The Journal of Business (University of Chicago Press), via JSTOR; hosted at web.williams.edu · October 1963

   “It is my opinion that these facts warrant a radically new approach to the”

4. Was Pareto right? Do the US wealth and income distributions follow power laws? University of Greenwich, Greenwich Papers in Political Economy (GPERC92) · 2023

   “Nevertheless, a sizeable empirical literature fails to find support for them in the data.”

5. Was Pareto right? Do the US wealth and income distributions follow power laws? University of Greenwich, Greenwich Papers in Political Economy (GPERC92) · 2023

   “We find strong support for asymptotic power laws beyond the 91st percentile”

6. Was Pareto right? Do the US wealth and income distributions follow power laws? University of Greenwich, Greenwich Papers in Political Economy (GPERC92) · 2023

   “Teulings, C. N. & Toussaint, S. (2024), ‘Top Wealth Is Distributed Weibull, Not Pareto’, SSRN”

7. Stretched exponential distributions in nature and economy: «fat tails» with characteristic scales arXiv (cond-mat), by Jean Laherrère and Didier Sornette

   “To account quantitatively for many reported «natural» fat tail distributions in Nature and Economy, we propose the stretched exponential family as a complement to the often used power law distributions.”

8. Stretched exponential distributions in nature and economy: «fat tails» with characteristic scales arXiv (cond-mat), by Jean Laherrère and Didier Sornette

   “The plot does not show any straight line whatsoever but rather a systematic downward curvature (very close to a parabola) implying a much ``thinner'' tail than predicted by a power law.”

9. Stretched exponential distributions in nature and economy: «fat tails» with characteristic scales arXiv (cond-mat), by Jean Laherrère and Didier Sornette

   “found that a power law truncated by an exponential provides a reasonable fit at short time scales (much less than one day), while at larger time scales the distributions cross over progressively to the Gaussian distribution which becomes approximately correct for monthly and larger scale price variations.”

10. Stretched exponential distributions in nature and economy: «fat tails» with characteristic scales arXiv (cond-mat), by Jean Laherrère and Didier Sornette

    “Historically, the central limit theorem led to the first paradigm in terms of Gaussian pdf’s that was first put in doubt by Mandelbrot [41] when he proposed to use Lévy distributions”

Reasoning

E1 (JEP 2016) says several power laws in economics are established empirically, and E3 and E10 show that Mandelbrot challenged the Gaussian model for speculative prices in 1963. That supports heavy tails for some variables, but no item shows that practically all economic and informational variables are subexponential. Other items point to qualifications: E2/E4 report a sizeable literature that finds no support for power laws in wealth and income data. E5 finds strong support for asymptotic power laws above the 91st percentile, and E9 reports that price changes become approximately Gaussian at monthly and longer horizons. Since the evidence documents heavy tails only for certain variables and is otherwise inconclusive on the broader "practically all" claim, the criteria place this at mostly false.

Cite this claim

Nassim Taleb, “Since the 1960s, practically all economic and informational variables have been shown to belong to the D2 class, or at least the intermediate subexponential class, which includes the lognormal.” (said 2019). Verdict: mostly false. Reliability API, claim c_df81f1de46ef, method 0.1.0+28da144b. https://epistacklabs.com/claims/c_df81f1de46ef