Project Implicit.

[whyBish]

some of the proofs will work logically if the assumptions and initial conditions are crafted well enough, and variables are all accounted for.

...

The hardest thing, I think, is to recognize the assumptions you don't realize you are making. 

Occhi
[right][snapback]94906[/snapback][/right]

I work with economists far to often (Still working on a project for one of the big five banks here). The main problem with economists is what level of simplification they run models at.

An example is the question "What happens if the demand for power (electricity) goes up?"

level 1:
The supply of electricity is fixed, so price goes up.
level 2:
The supply is actually variable, but capped (i.e. there is some slack in the generation), so price may not change much at all, unless the demand exceeds the supply cap.
level 3:
The price goes up in the short term, and so more power stations are built.
level 4:
Power station building is lumpy so if the govt. environment (tax / regulation) is not predictable enough then it is too risky to invest in new power generation.
level 5:
Research into alternative energy sources increase.
level 6:
The substitution effect kicks in and people start using candles instead of lightbulbs (for example)
level 7:
Candles are made from oil, so the demand for oil increases.
level 8:
All of the above arguements about what will happen to the price of oil
level 9:
Since the price of oil increases, power generated from that source becomes more costly. However, substituation to other forms of generation + the reduction in demand affect the price in the opposite direction.

(Sorry I shouldn't have assigned numbers to these levels, I am trying to show additional levels of complexity to the model)

Result: Different answer depending on what model is chosen. Different answer depending on what timeframe was chosen. Different answer depending on a whole bunch of factors that are practically important, but the economist would call them 'confounding' factors (e.g. regulatory environment, substitutes etc.)

So they can prove that in *model* X when event Y occurs the outcome is Z, but then they apply that to the real world and it falls apart.