The Difference Between Optimization and Survival: An Argument for Learning More than Mainstream

Mads  Stolberg-Larsen

At any given point in time it will with great certainty NOT be the most fit that have most success. The implication is that learning what is not mainstream might have a higher value than currently thought.

At any given point in time it will with great certainty NOT be the most fit that are the most successful.

This is a manifestation of the difference between optimization and survival. 

When an individual or agent (that be an institution, country or animal) optimizes he or she does so given the prevailing environment in order to do well. Given that such agents solely use current circumstances as information to guide their decisions, such agents will in all likelihood do very well. They will do so because they fully utilize all resources available in an "efficient" manner where no redundancies exist in their behavior or bodies.

But; when circumstances change dramatically over a short period of time such optimizing agents  die. They die because they are fragile to rapid change that creates environments that in no way resembles the ones they optimized over. To rephrase this: when the world rapidly changes around them they die because they are not ready for the changes.

This hints at why many people misunderstand the concept of "survival of the fittest". When this term is thrown around is it often used to in relation to success. However, there is a great difference between survival and success (!). Success is a reference to someone or -thing doing well. Survival is someone or -thing existing. The latter has to do with being fit and surviving adverse shocks, i.e. surviving a spurt of rapid change, the former has to do with doing well under current circumstances. 

Now; a necessary condition of having success is having survived so long it is possible to have it. That is one needs to not have died before success is even possible. But in order to be the MOST successful at any point in time one needs to take full advantage of current circumstances - i.e. reduce redundancies that are irrelevant to the current environment. Reducing the redundancies leads to two things: success and fragility. Success because one now uses available resources without "wasting" them on things that are irrelevant to current circumstances, and fragility because smaller changes now can kill. As increased fragility decrease the chance of surviving adverse shocks it is clear how the most successful at any given point in time with great certainty is NOT the most fit.

The implication for economics is this: at any given point in time it will not be the most popular theories that stand the greatest chance of surviving the test of time. Learning what is not mainstream thus possibly have a much higher value than is currently thought.  

Tags: #debat  #studiet  #Kritiske Politter  #success  #survival  #optimization 

9 kommentarer


Christian Stassen

Christian Stassen @ d. 07. august 2014 #1

All the above rests on an assumption that the skills that make you successful in the current environment are not applicable to the future environment. That is not necessarily a good assumption for all types of skills.

Take math or coding skills - they will allow you to be successful in many different environments, but will be through applications in different areas. Good programmers could find jobs as a financial quant before the crisis, now the very same people find jobs in internet firms like google and facebook.


Mads  Stolberg-Larsen

Mads Stolberg-Larsen @ d. 11. august 2014 #2

No. It rests on the assumption that to be the MOST successful one needs to specialize/optimize to the current environment, which makes one fragile to change.

Now I do agree that some skills increases the change of surviving adverse shocks (for example a financial crisis) and allows one to continue to have success. But you forget that we are talking about having the MOST success. Prior to the crisis I highly doubt that it was the programmers and mathematicians that made the biggest buck in the financial organizations for which they worked. And I also doubt that many of the people making the biggest buck before 2007 are now the most successful in whatever industry they are employed in.


Christian Stassen

Christian Stassen @ d. 11. august 2014 #3

One is not optimizing apecific skills, one is optimizing lifetime income. So those who made the most before 2007 made the riggt choice even if they make less now (if they are even still working ). So the skills they learned while only being eg 80% transferrable was still optimal choice on average.

Obviously ur optimization might be inoptimal ex post, but there is nothing that shows that one should spend time on learning something "weird " just to have the off chance it might be valuable later. One still need to optimize the skills that give highest expected lifetime income ex ante.


Mads  Stolberg-Larsen

Mads Stolberg-Larsen @ d. 13. august 2014 #4

First off: No. One does not optimize lifetime income but expected lifetime utility (big difference as utility is affected by much more than income). That is if one believes that people (even economists) make decisions through optimization (an alternative being satisficing).

The statement "those who made the most before 2007 made the riggt choice even if they make less now. So the skills they learned while only being eg 80% transferrable was still optimal choice on average" is simply wrong in my view. They (the people making the most - not programmers and mathematicians) made decisions based on assumptions about risk and uncertainty that led to wrong decisions. I understand one has to look at their decisions as ex ante, but even doing so they did not protect properly against their models being wrong, which is unforgivable.

Furthermore there is something that might show one should spend time on learning something "weird" just to have the off chance it might be valuable later. It's called pay-off.
Bill Gates was a bright fella when he learned about programming (and lucky to be in the vicinity of somewhere to program at his young age), but at the time programming was something weird. Despite Bill Gates being bright I would argue that him learning the weird skill of programming must have something to do with him having a net worth a million times higher than the median household in the US (arguably network externalities also plays in).


Frederik Plum Hauschultz

Frederik Plum Hauschultz @ d. 14. august 2014 #5

Er der en translatør til stede?


Christian Stassen

Christian Stassen @ d. 14. august 2014 #6

(Frederik jeg ved ikke helt hvorfor det var skrevet paa engelsk men jeg holdt den bare der da jeg stolede paa at OP gjorde det med vilje med god grund)

Mads, your comment about the morality of the programmers who you now seem to blame for the credit crisis is off-topic, and also quite wrong (it is a topic you know absolutely zero about so you should probably not mix it with this one). The question is about whether they maximize their utility as you say, not about if they maybe took some wrong model assumptions.

I agree one should look at lifetime utility, but as you spoke about being the "most succesful" etc, i took it as if you looked for a direct link with income. maybe that is not what you meant, but it doesnt really change my argument (in fact it just makes your own weaker - is most succesful the one with the highest utility? if that isnt the case, then one shouldnt care at all about what skills one learn and rathre just study something fun)

ie, to maximize lifetime income, it is about learning a skill that does just that, which often is to learn "mainstream" things such as mathematics, physics, coding, etc - and "mainstream" economics being one of those things. For some it won't be the case, but again, there is nothing in your argument that "proves" that it is a good idea to devote time to learn about things such as "ecological economics" and things like that. To prove that, you need to show a sample of people, part of them having studied "only" mainstream (whatever that means) and the other part having studied other theories in the field to whatever extent you're trying to prove. then you must prove that a significant fraction of the latter group has higher lifetime earnings, after rinsing out other effects (e.g. a person who studied 5% "alternatives" could just be more curious, a trait which might in itself be the reason for the income bump, and hence not attributable to the learning of the other theory). a study which i am willing to place a very large bet on will come up empty handed.
--> if you want to then speak about lifetime utility, you will have to prove that the latter group is more "happy", a type of studies which is always rather dubious (which is again another reason why i somewhat "naturally" turned it towards a discussion about lifetime income)

(btw, when Bill Gates learned to program, it wasn't at all a weird thing, it was studied by many scientists all over the world. it was less commonplace, but not something that was "looked down" upon by "mainstream" scientists, which is what you're talking about here)


Mads  Stolberg-Larsen

Mads Stolberg-Larsen @ d. 14. august 2014 #7

@Frederik: Tog bacheloren på engelsk, så nogen gange falder det bare mere naturligt at formulere ting engelsk end på dansk, når det gælder ting der har lidt at gøre med teori og økonomi. Der er ikke nogen dybere grund end det, så fortsætter på modersmålet (selvom Google translate uden tvivl er svaret på dit spørgsmål!).

@Christian:

For det første går jeg ud fra, at du ved ualmindeligt meget om kredit/finanskrisen siden du ved hvor meget jeg ved.

For det andet er jeg rimelig eksplicit i parentesen lige efter "they" referer til de personer der tjente mest før 2007 - IKKE matematikere og programører (lost in translation..).

For det tredje tror jeg, vi er uenige om så mange grundlæggende ting, at vi diskuterer helt forbi hinanden.
Jeg brugte nytte istedet for indkomst fordi det koblet med prospect theory og en menneskelig tendens til at undervurdere halesandsynligheder, der aldrig er oplevet, kan bruges til at forklare at et negativt diskret hop i indkomst kan være meget ødelæggende ens nytte. Det fik jeg ikke med i ovenstående men det er værd at have in mente, når man vurderer folks valg ex ante.
I forhold til at maksimere livsindkomst ved at vælge at arbejde på at udvikle nogle bestemte evner, tror jeg ikke på, at det er sådan man tager beslutninger. Men det er ikke det, der gør jeg synes, det er en åndssvag måde at tage beslutninger på. Det er det domæne man forsøger at optimere over (fremtiden). Årsagen er såre simpelt den, at hale events dominerer så meget, du ikke kan kende gennemsnittet, fordi du ikke kender den underliggende distribution. Det gør optimering meget svært og det at beskytte sig mod downside mere attraktivt, hvis målet er at overleve på sigt. I forbindelse med teorier ser jeg således alternative teorier til mainstream som forsikring mod Mankiw måske tager fejl (hvilket i parentes bemærket selvfølgelig er en umulighed!).
Af samme grund er det dybt misforstået, når du og andre direkte ellers indirekte ligger økonomi i kurv med fysik, programmeringen og naturvidenskab. Vi har alt for få re-runs af eksperimenter af vores viden til at vi kan være i det selskab.


Christian Stassen

Christian Stassen @ d. 14. august 2014 #8

Ja i og med du har nul erfaring fra sektoren er det en rimelig antagelse.

Prospect theory snakket i øvrigt om tab, ikke indkomstændringer. Prospect theory har meget lille relevans her

Jeg er enig i at fremtiden er svær at spå om.

At hedge mod hale events gøres ikke ved at købe tilfældige haler som er det du anbefaler i og med du ikke kvalificerer en bestemt teori som værende bedre end andre. der findes en stor teori om tail risk hedging, du kan eventuelt referere til et papir der påviser din påstand?


Mads  Stolberg-Larsen

Mads Stolberg-Larsen @ d. 14. august 2014 #9

0 erfaring fra sektoren => 0 viden. Solid kausalitet.

Jeg er ret overbevist om, at det at miste sit job og gå ned i løn vil opleves som et tab. Så jo, prospect theory kan godt have relevans.

"Jeg er enig i at fremtiden er svær spå om" er ikke det jeg vil frem til. Det jeg prøver at sige er, at fremtiden er nem at spå om: Den vil overraske os. Og når den gør det, vil man se tilbage på nutiden og undre sig over, hvor dumme vi var (hindsight bias).

Til sidst: Hvor ser du mig anbefale at man bør hedge sig helt tilfældigt? Det vil jo være torskedumt (omend der er chance for, at det kan vise sig at være en bedre strategi). Der findes et utal af kvalifikationsmetoder. Min egen (uvidenskabelige) går på at læse de ting, jeg finder interessante.
Og papir: Ved ikke om den påviser eller påpeger noget, men jeg synes grundtesen i "Antifragility" er fantastisk (kend din downside og lad upside være open-ended). Indlægget ovenfor er til dels min fortolkning af nogle af Talebs skriblerier.


Tak for din kommentar!
Skriv venligst en kommentar der er længere end 5 tegn

Skriv en kommentar

Log ind for at kommentere - eller opret en bruger
Du vil måske også synes om