Subject: Re: Tom W. Bell paper
From: Thomas Lord <>
Date: Fri, 01 Sep 2006 13:11:58 -0700

Don Marti wrote:
>> I think Tom Lord and Hans Reiser have both made good
>> points that are relevant to the "why".  The existing
>> capital markets and patent system are optimized for
>> funding inventions that are already finished from a
>> research POV and well along in development.

Sure, that is a reasonable summary of some things I've said.
(And thank you for that end-to-end check!)

I don't think that prediction markets help very much and
here is why:

Prediction markets are zero-sum markets:  some traders
win by the exact amount that other traders lose.  They are
not gambling markets because skill is involved.  That
is to say that  they are more like poker than they are like
the lottery.  These markets do not give a reward for
discovering a new vein of gold -- they give a reward for
guesstimating when the next vein of gold will be discovered.
These markets do not allocate shares in the vein of gold --
they divvy up a pot of bets placed by gamblers who think
they can spot when that vein will be discovered.  A pair
19th century gentlemen in Atlanta could make a friendly
wager that huge amounts of Gold would be found in
California by 18xx and one would win, the other lose -- but
neither would have any claim to the gold.  They would
just be reallocating their collective fortune, not investing
in a speculative share of future wealth.

Specifically, prediction markets *do not allocate* any of
the economic growth that results from a research
success.   Rather, prediction markets allocate some of the
risk of investment in research.  (But they perform that
allocation in an odd way, see below.)

In other words, if we have a big bag of money to spend
on research and we are trying to decide how to divvy it
up among competing research proposals, then there is
an argument  made that prediction markets are the
right tool to manage the allocation.   What prediction
markets *do not* do is provide incentive for increased
investment in research.   Well, mostly.

Of course, *if* it is the case that a prediction market does
a significantly better job of allocating a fixed pool of
research funds than other methods (say, a panel of experts
operating an NSF-style bureaucracy) then, yes, the
introduction of prediction markets may help to incrementally
grow the size of the big bag of money to spend on research.
This isn't any better, though, than just changing the NSF
committee rules to get incremental gains:  no new incentives
to invest in research have been created;  existing incentives
may or may not have been slightly improved.

Even given all of that, even assuming that we really, really
desire that incremental gain, it isn't clear that prediction markets
are a wise idea.  I suspect they would be a corrupting influence:

Prediction markets characterize research, in essence, as
the activity of attempting to verify certain hypotheses --
does the question people are trading certificates in have a
"yes" or a "no" answer? 

That is a poor metric for the value of research.   For example,
the Michelson-Morley experiment was worth doing even
though (or perhaps because) it produced a negative result.
What would be the prediction market claim for that experiment?
"The aether will be measured by 1887?"   "The aether will
be disproven by 1887?"   Why would we want a market to
reward people making one of those predictions over the other?
There were no serious external reasons for anyone to invest
in hopes of a return from either outcome yet, as a society,
profound economic growth followed from investment in simply
addressing the question.   Not only did M-M establish an empirical
imperative that led to relativity, it had spin-offs of terms of the
propagation of more generally applicable experimental techniques.

(And please don't offer the alternative predictive market question:
"The hypothesis of the aether will be proved or disproved by
1887" because there was no question about that.   The answer
to that question was simply "yes, we know that."  Prediction
markets amount to a poker game for those already committed
to investing in research -- they are a way to split the bill and have
a little fun   They are not (aside, perhaps from the fun) a way to
bring new players to the table.)

In short, prediction markets imply a hopelessly reductionist
view of the value of research.

To sum up:

Prediction markets allocate risk, not rewards from research
and therefore create no new incentive to invest in research.

The means by which prediction markets allocate risk involves
an unrealistic view of the value of research.


Towards alternatives:

We can regard the field of potentially useful consequences
of performing research as an infinite supply of unexplored
territory -- in principle, anyone can go and discover a new
island or continent and, as side effect, identify new trade winds
as they explore the ocean of truth.

Unlike physical territory, the truths discovered in the course
of research are inherently non-rival -- there is little point in
trading in them.

To find ways to increase research investment and direct
it more intelligently we must form opinions about *process*
rather than predictions about *outcomes*.

Our incentives should NOT be

   I'll invest $X in research because I think outcome Y
   is likely.

Our incentives should reflect:

   I'll invest $X in research services from A because
   I think that useful outcomes are likely given A's
   systematic approach to exploration.

Externalities and transaction costs -- the ways in which
research, though  essentially non-rival -- does not propagate
evenly in the short term -- that is the essence of where to
find investment models.  

Which brings us back, but in a better light, to the general
idea of "exclusive rights."    We just ought to be more
creative about what those rights entail.

If I had the start-up capital, I'd start a lab that would
be funded by newsletter subscriptions and paid
on-site visits.   Indeed, if there are investors who
would be interested in such an approach, please
get in touch -- we can start quite small.   I'd be happy
to describe my own "systematic approach" in the
field of practical, open source R&D.