1/4 There used to be a clear definition of "economically rational."
This meant: maximize expected wealth.
When confronted with reality, this model fell flat on its face. People don't do it.
2/ Utility was introduced, but with it the meaning of "rational" became unclear.
Was utility a means of describing irrational behavior, or was maximizing expected utility the new rational?
Both notions exist in the literature.
3/ Making matters worse, "expected utility theory" exists in different forms: Bernoulli's original 1738 form is incompatible with von Neumann and Morgenstern's 1944 form.
When people say "expected-utility theory," some mean this, some mean that.
4/ As a consequence, when we say someone acts economically rationally or irrationally, this by itself is a meaningless statement.
Also, saying that expected utility theory holds or fails is meaningless unless we specify what we mean.
Economics has failed to define its terms.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
1/5 The Copenhagen experiment showed that Ergodicity Economics (EE), limited to its predicted utility functions for given dynamical settings, is a better fit to human behavior than classic expected-utility theory with a freely chosen single utility function.
2/5 I don’t find this terribly interesting. Classic expected-utility theory is conceptually flawed, and science is more than data-fitting. However well or poorly it fits observations, one would have to reject expected-utility theory anyway.
3/5 Here is what's interesting: I didn’t think EE would perform well in the Copenhagen experiment because I had bought into the narrative that the tested behavior was shaped by evolution over millions of years and cannot be re-learned on short time scales.
2/7 Early treatments of these problems, in the 17th century, assumed that people should optimize expected wealth (blue line). If they did, they would voluntarily play this game -- the blue line points up.
But real people didn't behave this way. They declined the offer to play.
3/7 This happened before it was known that expected wealth (blue line) is not what happens over time (long-time limit of any red line).
So people were puzzled.
The solution to this puzzle is called expected-utility theory (EUT), developed in 1738.
1/7 I disagree with the implicit statement by Mervyn King - wondering what @ProfJohnKay thinks - that maximizing expected utility is reasonable in the small world of a simple model.
My critique is more devastating and less palatable: it's a thought error to optimize this object.
2/ In other words, we don't need to make the model more realistic for expected utility maximization to become a bad idea. It's a priori the wrong object, namely: changes in utility are non-ergodic. Their expectation value has no physical meaning for an individual decision maker.
3/ I'm sorry if this ruins formal economics, but it's time we talked about it like grown-ups: the economic formalism is unacceptable from a quantitative-science point of view.
Before it's worth discussing how realistic it is, we have to correct the flaws in its implied physics.
Friedemann Schulz von Thun came up with a way of clarifying things people say.
In his scheme, any statement is dissected into four components
* factual
* self-revelation
* relationship
* appeal
2/9
The speaker, intentionally or not, sends messages on all of these levels.
The listener, intentionally or not, hears messages on all of these levels.
3/9
Communication problems arise when we confuse different levels. My intention may be to make a purely factual statement, and I may fail to consider its relationship content. That can be hurtful.
1/14
More on @soniasodha's Analysis program from last night.
@ReicherStephen was wonderfully clear. He says there's a "classic" view of humans now, according to which we are psychological frail, biased, faulty. We cannot cope with emergencies. bbc.co.uk/programmes/m00…
2/ This view, he says, is contradicted by the evidence. People don't tend to panic in emergency situations but act rather soberly and sensibly.
To me, that makes perfect sense: evolution would have swiftly got rid of us if our psychology failed us whenever we needed it the most.
3/ It also chimes with what we found in economics: the narrative is one of irrational humans, but when we check, the evidence had a tendency to evaporate, and models that take people's situations into account tend to predict actions correctly.
1/4 Like I said yesterday: it's a cliff-hanger. Let's see what @soniasodha has got for us on Monday.
My recollection: the UK's modeling at the time was catastrophically wrong - "UK 4 weeks behind Italy, and 5-6 days doubling time."
That's on the record (and it was wrong).
2/4 These numbers seemed to come out of @neil_ferguson's model. Then some model assumption was changed, and estimated deaths became more realistic (500,000). No one knew why because the model code was unpublished. It seemed like robustness hadn't been checked.
3/4 I remember someone claimed the change was due to previously underestimating hospitalizations by a factor 2. I found that strange because the pandemic was growing by a factor 2 every 2.5 days. So healthcare system collapse would just move back or forth by ~2.5 days.