From: Hal Finney (hal@finney.org)
Date: Thu Mar 13 2003 - 11:31:44 MST
Anders writes:
> The basic problem isn't that the allocation is unfair, but that
> political pressures can introduce more stuff as "public goods".
> Are puppet theaters public goods taxpayers should fund? At
> least in Sweden they are :-) So this means the government can
> introduce a nearly arbitrary number of new goods to vote about.
Bret Kulakovich made a similar point:
> Anyhow, you've begged the question: No one wants to spend for
> defence, so little is spent on defense. The next war comes around and
> - oh wait, there isn't a war - the first exobrate that showed up with
> a sharpened pencil took over the country.
Theoretically, the point of the mechanism is to find out how much people
truly value these various public goods. There is no particular problem
with new goods being created, as long as it doesn't make the mechanics
of the procedure become too unwieldy (and it's not clear whether the
whole thing would really be manageable today even with a modest set of
publi goods). If people like puppet theaters, they will be funded,
and likewise for national defense. The goal of the mechanism is to
solve the free rider problem so that the funding levels of these goods
accurately reflects the public sentiment.
Anders also wrote:
> And what would the default M be? If a new dial appears on all our
> taxameters set to M_gov, then I will find that lowering my dial will cost
> me, so I won't lower my dial very much even if I think the new good is
> useless. The same goes for everybody else, which means that the inertia
> in the system will lead to people voting for an allocation of nearly
> M_gov for something none might actually want. So maybe a better rule
> would be that any new public good should start out with an allocation
> of zero (after all, that was the state before the new dial anyway),
> but now we get the opposite effect: it is hard to raise the allocation
> to the level people might want, even if we all want it.
Yes, the details of the mechanism are not too clear to me either.
The one paper I saw which described an experiment had only five subjects,
but they got pretty rapid convergence to an optimal value. They used
iterated periods rather than the continuous mechanism I described, but
they got there within a few iterations.
It would be very interesting to try something on a larger scale, maybe
using the net. We could think of some public good, like paying some
person to post more or less, then run the GL mechanism and see what the
community consensus was. The hard part is probably getting people to
pay up at the end.
Hal
This archive was generated by hypermail 2.1.5 : Thu Mar 13 2003 - 11:39:34 MST