Since I began thinking about portfolio optimization and option pricing, I've struggled to get an intuition for the risk premium, i.e. that investors are only willing to buy risky instruments when they are compensated by an add-on above $r$ (the risk free interest rate).
On the one hand this is understandable and backed by most empirical data.
On the other hand there is this divide between the risk-neutral world of derivatives pricing and the real world with real world probabilities.
The bridge between both worlds goes via a hedging argument.
I wonder (just to give one possibility of debunking the classic risk premium argument) if there really was something like a systematic risk premium $\mu-r$ nothing would be easier as to squeeze it out by simply buying the underlying and selling the futures contract against it. The result should be a reduction of the risk premium until it becomes $0$ and the trading strategy wouldn't work anymore.
The result should be that even in a risk-averse world probabilities become risk-neutral (i.e. the growth rate $\mu$ equals $r$) like in option pricing.
Is there something wrong with this reasoning?
OK, my original argument doesn't work. Anyway: I'm still feeling uncomfortable with the idea. My reasoning is that the whole idea of an inherent mechanism for compensating people for taking risk within a random process nevertheless (!) seems to be build on shaky ground. If it was true it would also be true for shorting the instrument since you are also taking risk there. But in this more or less symmetrical situation one side seems to be privileged. And the more privileged one side is the more disadvantaged the other side must be. It all doesn't make sense...
I think the only compensation that makes sense in the long run is $r$.
Please feel free to comment and/or give some references on similar ideas. Thank you again.
After a longer quiet period on that issue see this follow-up question and answers:
Equity risk premium and risk