"Paul G. Silva" <paulgsilva@...> writes:

> The above brain dump makes me wonder if prediction markets are well
> suited to answering "big" questions. After all, isn't it so much
> harder to disentangle cause and effect with big problems?
>
> If someone were 2 years ago were to propose "This infrastructure
> project will increase GDP by 10% in five years"... and then this
> current economic crisis hits us. Maybe the GDP did increase BECAUSE OF THE
> INFRASTRUCTURE by 10%, but the economy as a whole ends up
> shrinking by 5% so... the project WORKED but GDP (or GDP+) is a
> measure too prone to other factors for us to be able to use it for any
> but LARGE IMPACT items.
>
> Thoughts?

You ask good questions.

Some approaches:

* The "side-bets" approach. In this approach, investors are expected
to arrange their portfolios to control what risks they are exposed
to.

For instance, suppose that Proposal X is a proposal in a futarchy
and that I as an investor believe the following:

* I believe Proposal X will increase average health in 10 years;
other people disagree.

* I assume as conventional wisdom that an increase in average
health 10 years from now will increase GDP+ in 20 years. I know
no more about this than everyone else.

* I don't know much about other factors affecting GDP+

* And for the sake of simple exposition, let's say that a low
mortality rate (say, below a threshhold of 1% per year in a
certain population) is the sole measure of health.

How can I bet on X but I avoid getting whomped when, even if I'm
right, something else lowers GDP+ (say, the banking crisis)? I
could invest as follows:

* Long on proposal X

* Long on a futures issue that, contingent on the mortality rate
10 years from now being low, pays off OPPOSITE to GDP+ 20 years
from now.

Note that I'm just neutralizing an implied short position in
it. So if measure X is enacted and it lowers mortality as I
believe, I become indifferent to GDP+. Since other investors
don't believe X will lower mortality, presumably I can buy this
issue plus an issue of X for less than the pair of them is
worth if I'm right.

* Long on a futures issue that, contingent on the mortality rate
10 years from now NOT being low, pays off according to GDP+ 20
years from now. Similar reasoning to the above, for the case
when X is not enacted.

I'm sure the logic is familiar to everyone here: Any losses I take
on one are offset by gains in the other, except for my intended
exposure. I gain or lose only according to the effect of proposal
X on mortality.


* The "layers upon layers" approach. In this approach, proposals
often propose to create other markets. For clarity, I'll call them
sub-markets and the first one the "main" market.

A sub-market, if created, has its own utility function and commands
certain specified resources. Let's make that a bit clearer: In the
situation I'm talking about, there's a proposal X to create a
decision sub-market S. The proposal X specifies:

* what resources S is to be able to disburse. For instance, $10
million over its lifetime. (Assume that the main market
disburses resources of at least that amount.)

* what S's utility function is to be. For instance, it might be
the lowering of the mortality rate 10 years from now.

So X is in effect a proposal to spend $10 million according to the
decisions of sub-market S. If X is not enacted, nothing special
happens and S doesn't come into being.

The expectation is that X will be enacted if both:

1. Whatever it measures tends to cause higher GDP+

* Of course it will never correlate as well as GDP+ itself, so
the next item is crucial.

2. Its payoff function contributes more by increasing the quality
of the decisions in S than it detracts by being a proxy for
GDP+. That is, because it measures a smaller, simpler
situation, S is easier for investors to predict than the main
market.


What's neat is that neither approach has to be written into the
futarchy rules. An enterprising individual could just do either.

Tom Breton (Tehom)