Behavioral finance FAQ / Glossary (Numeracy)
This is a separate page of the N-O section of the Glossary
Dates of related message(s) in the
Behavioral-Finance group (*):
Year/month, d: developed / discussed,
03/1i,2i,8i + see magical
thinking / numbers, small
Goodhart law, range
estimate aversion, quant,
overreliance on management
objectives, model, fuzzy logic
Eating numbers for breakfast, for lunch and for dinner.
Can all those delicacies be trusted?
The numeracy bias, in economic or financial analysis, is a strong reliance,
sometimes excessive, in "big data", statistical "indicators" and advanced
The numeracy bias phrase can be used with a positive or negative meaning.
So ...beware when you write your résumé!
On the positive side
On the negative side
In finance, business and
economics some taste for
numbers is recommended.
You do not go far without them,
as they are all over the place.
A blind trust in numbers might
For example some official
economic statistics might be
irrelevant or flawed and give the
illusion of science and truth.
Abstraction and numbers
vs. narration and fuzziness
Trusting hard numbers or soft descriptive words?
The inventions of numbers (from which writing derives, by the way)
and mathematics were fantastic jumps into progress for mankind.
Computers completed this evolution and more and more phenomenom can
now be measured in ways that allow very deep analyses.
Numbers bring precision and many activities would work poorly, or would
not even exist, without them.
For example, to neglect the base rate (see that word) probabilities could
cause serious mistake.
But in a world where nothing is totally clear-cut
to give precise value to unreliable numbers and fuzzy situations
would be a red herring and would distract from realities.
In some cases, softer tools might help
Some phenomena could be researched and explained better in narrative
terms than in abstract or digital ones.
Although narration also has its own limitation (see rationalization, story...).
Also fuzzy logic (see that phrase) brings a bridge between fully numerical
and more literal approximations.
Can statistics be trusted?
Is an abyss hidden behind the number mountain?
Are predictive models really predictive?
Huge progresses have been done in data mining and in scientific data
processing to find correlations, hidden trends, anomalies, antecedents (rare
This is what "big data" is about.
even so, statistics, as well as corporate accounts / management information
systems (see overreliance on management objectives and norms) should be
taken with caution because they might be
* imperfect, and for some of them fully misleading,
* misinterpreted. The role of humans is to understand what the raw results
really tell, how reproductive in the future they are and to what areas and
cases they can be really applied.
When using them for previsions and decisions, the data that are
Give only a limited view of reality. These limitations might lie in:
* the choice of what is measured, and the way it
* and the fact that reality does not lie only in numbers.
A broader culture, experience and aptitude at fact-finding and case
analysis and at imagining scenarios might be able to compensate an
overconfidence in mechanical prevision tools.
Be already obsolete.
Obsolete data are all the more dangerous when they
* reinforce mental anchoring in past situations,
* are considered as fully predictive and blur the fact that the future
is always uncertain.
Ignore rare occurrences that don't appear in too short statistics, or
which are totally new.
This is the case of "rare events" (see that phrase) for example
a sudden liquidity crisis (see liquidity squeeze).
This is also the case of totally new events or combination of
events, never seen before.
Hide errors in observation or calculation.
Or might not be bona fide by dissimulating accounting / statistical
manipulations (cooking the books).
See in this glossary: deception, manipulation, Goodhart law....
And what about correlations,
equations and modeling?
Is the answer a mathematical formula?
A clear-cut number?
Or something less precise?
Excessive trust in statistics is not the only numeracy bias, here are five other
types of too much love for numbers:
Seeing apparent but illusory patterns (simple coincidences)
in some number series (see representativeness heuristic).
Interpreting wrongly mathematical correlations as direct cause-effect
Reasoning on "magical numbers" (see numerology).
Believing that you can make precise predictions in an uncertain
The range estimate aversion (see that phrase) is a symptom of the
widespread dubious belief that precise estimates are possible or relevant.
Believing that you can make such clear-cut / reliable predictions by using
sophisticated quantitative model (see model, range estimate aversion)
Science is highly respectable and useful, much better
than magical thinking.
But beware of "scientific garbage".
Such false science is also a form of magical thinking,
myopia and routine.
Plain questioning might work better at ringing the alarm
bell or detecting opportunities!
Models based on historical probabilities can work rather well in
ordinary situations and be enough then to predict the usual risks
But they get obsolete in new states of affairs or when "rare events"
Then they exceed their "best before" date.
They are useless, if not counterproductive, when
uncertainty, which by definition does not stick to known
patterns, overrides plain mathematical risk.
The "big data" challenge
The new Grail? Or a perilous quest?
Big data crosses a mass of diverse information, until then seen separately,
or as too short statistical series.
It might spot hidden trends / correlations, rare events and weak signals
It can bring some order, a better understanding within the information
Thus it promises to help make much better previsions, to orientate
decisions in a more effective way.
On the other hand, human wisdom will stay more than ever crucial
* to make the sorting, to admit that the fundamental uncertainty will
not vanish even under an avalanche of diverse data, that many decisons
will keep being bets on an unknown future.
* thus to apply the findings wisely and creatively, by avoiding too
systematic / one-sided / over-confident (extreme numeracy bias).
(*) To find those messages: reach that BF group and, once there,
1) click "messages", 2) enter your query in "search archives".
Members of the BF Group, please
vote on the glossary quality at BF polls