Economic Models and Experiments

Economic Models and
Experiments
Francesco Guala
Università degli Studi di Milano
A common complaint:
Economists spend too much time and effort
constructing and analysing mathematical
models
Such models are unrealistic, idealized,
inadequate representations of real economies
Modelling is a source of bias, leads to mistaken
explanations and misleading policy advice
“As I see it, the economics profession went astray
because economists, as a group, mistook beauty,
clad in impressive-looking mathematics, for truth.”
“the central cause of the profession’s failure was the
desire for an all-encompassing, intellectually
elegant approach that also gave economists a
chance to show off their mathematical prowess.”
(Krugman 2009)
«To put it bluntly, the discipline of economics has yet
to get over its childish passion for mathematics and
for purely theoretical and often highly ideological
speculation, at the expense of historical research
and collaboration with the other social sciences.
Economists are all too often preoccupied with petty
mathematical problems of interest only to
themselves. This obsession with mathematics is an
easy way of acquiring the appearance of scientificity
without having to answer the far more complex
questions posed by the world we live in.»
(Piketty 2014: 32)
How modelling works:
1. Imagine (or build) a system that is in some ways
‘similar’ to the target (real-world)
2. Describe its characteristics (assumptions)
3. Manipulate an aspect of the model (e.g. change
some assumption)
4. Derive conclusions (describe other characteristics of
the model that were not obvious initially)
5. Apply them (to a target)
Model vs world:
“The automobile market is used as a finger exercise to
illustrate and develop these thoughts. It should be
emphasized that this market is chosen for its
concreteness and ease in understanding rather than for
its importance or realism.”
(Akerlof 1970: 489)
Suppose there are 4 kinds of cars:
 New and used
 Good and bad (“lemons”)
Owners know their quality, buyers do not
(asymmetry of information)
There must be just one price for all cars
What is it?
Suppose that
Quality = x (0 ≤ x ≤ 2)
Buyers value cars more than sellers
U1 = Σxi
U2 = (3/2) Σ xi
μ = average quality
What price?
μ=p/2
0
If μ=p/2, then p=2μ
i.e. p>3μ/2
But then, D=0 !!
p
2
Surprising conclusion:




In Akerlof’s market, there is no trade: bad cars
drive good cars out of the market
These are the ‘economic costs of dishonesty’
But in the real world there is a market of used
cars!
So what does Akerlof’s model explain? Does it
explain anything at all??
1. Quality is uniformly distributed
2. Utility is linear in the number of cars
3. Traders do not make mistakes
4. There are just 4 kinds of cars
5. Quality can be measured on a continuous scale from 0 to 2.
6. There are just two types of traders
7. All cars belong to group 1 traders
8. Group-1 traders know the quality of each car
9. Group-2 traders don’t know the quality of each car
10. Group-2 traders know exactly the average quality of cars
11. Group-2 traders derive 1.5 times utility from enjoying a car
than group-1 traders
12. Prices are known by every trader
…
14. All traders maximise their own utility (there’s no altruism in
the model)
15. Utility is a function of the quality of the car only (no ‘pure
joy’ of buying a car)
16. Demand depends on quality and price only
17. Supply depends on price only
18. There are no counteracting institutions (no trademarks,
guarantees, etc.)
…
-> Conclusion: no trade (or less trade)
Problem:
• Models are useful only if they are simpler than
their targets
• Simplifying assumptions mis-describe the
target
• So the model gives a false account (or is a false
analogue) of the target
The ‘explanation paradox’
a. Economic models are false
b. Economic models are nevertheless
explanatory
c. Only true accounts can explain
(Reiss, Philosophy of Economics, Chapter 7, p.
127)
Solution 1:
a. Economic models are false
b. Economic models are explanatory
c. Only true accounts can explain
A good model need not be true:
It must be ‘credible’
Bob Sugden
Problem:
Credibility is not a substitute for truth
“Why has John ordered a salad?”
1. Beause he’s vegetarian
2. Because he’s on a diet
1. is ‘good for me’, 2. is ‘good for you’ (given our
background beliefs)
But we want to know which one is right!
Solution 2:
a. Economic models are false
b. Economic models are explanatory
c. Only true accounts can explain
A good model must be useful for
prediction. Explanation and truth are
unnecessary.
Milton Friedman
Julian Reiss
Problem: prediction is not a
reliable guide for intervention



For example: cumulative rainfall predict
prices (Hendry)
Why? They are highly correlated
But global warming does not cause deflation
(probably)
Solution 3:
a. Economic models are false
b. Economic models are explanatory
c. Only true accounts can explain
Nancy Cartwright
Dan Hausman
Uskali Maki
Good models must tell true causal stories
For example:
Akerlof (1970)
1. Shows that market exchange takes place with
symmetric information
2. Shows that it disappears with asymmetric
information
The falsity in the model is functional to make a causal truth
emerge clearly
“The example of used cars captures the essence of the
problem” (p. 489)
But how do we know that the causal story
is true?
‘Supers’ equilibrium
p=300
‘Regulars’ equilibrium
p=165
Lynch et al (1986) “Product Quality, Consumer Information and ‘Lemons’ in Experimental Markets”
“Numerous institutions arise to counteract the effects of
quality uncertainty. One obvious institution is guarantees […]
to ensure the buyer of some normal expected quality. […] the
risk is borne by the seller rather than by the buyer.
A second example of an institution which counteracts the
effects of quality uncertainty is the brand-name good. Brand
names not only indicate quality but also give the consumer a
means of retaliation if the quality does not meet expectations.
For the consumer will then curtail future purchases. […]
Chains – such as hotel chains or restaurant chains – are
similar to brand names. […]
Licensing practices also reduce the quality of uncertainty. For
instance, there is the licensing of doctors, lawyers, and
barbers.”
(Akerlof 1970, pp. 499-500)
Introduce
warranties
Objection: it is true ... of what?
• A laboratory market is not a real market
– In real markets there aren’t only 6 sellers and 8
buyers
– There aren’t only ‘Supers’ and ‘Regulars’
– The warranties are not perfectly enforceable
(etc etc)
So, are we back to square one?
NO!
Important distinction:
Internal vs External Validity
Internally valid = captures the true causal story
in the experimental setting
Externally valid = holds also in other
circumstances
There is usually a trade-off: the more confident
we are about IV, the less confident about EV
But this is true always, in all science
For example: genetics, medical experiments,
nuclear physics, etc.
• Distinguish pure science from application
• There’s no reason to believe that scientific
knowledge is always easily applicable
For example: penalty kicks
q
p
Prediction
0.42
0.38
Actual
frequency
0.43
0.40
Palacios-Huerta, I. (2003) “Professionals Play Minimax”, Review of Economic Studies 70: 395415.
The deepest thought ever:
“After all, it is not our stupidity which hampers us, but chiefly our lack
of information, and when one has to make do with bad guesses in
lieu of information the success cannot be great. But there is a
significant difference between the natural sciences and the social
sciences in this respect: experts in the natural sciences usually do
not try to do what they know they cannot do; and nobody expects
them to do it. They would never undertake to predict the number
of fatalities in a train wreck that might happen under certain
conditions during the next year. They do not even predict next
year’s explosions and epidemics, floods and mountain slides,
earthquakes and water pollution. Social scientists, for some strange
reason, are expected to foretell the future and they feel badly if
they fail.”
(Machlup 1961: 14)
30
[email protected]
users.unimi.it/guala