In this paper, we investigated the dynamical behavior of a market
consisting of simple automated sellers and buyers of a vertically
differentiated product or service. We introduced a simple family of
buyer utility functions that allowed for tradeoffs between
price and quality, and studied two different populations of such
buyers. The first population was extremely sensitive to quality,
while the second was extremely sensitive to price. For each of these
populations, we explored the dynamic collective behavior resulting
when the sellers employed five different price-setting strategies
ranging widely from perfect knowledge and unlimited computational
power to almost zero knowledge and capability.
For the quality-sensitive buyer population, all pricing strategies
eventually led to the same price equilibrium. However, for the
price-sensitive population, we found that most pricing strategies
led to large-amplitude cyclical price wars. It is possible to explain
these price war dynamics (and their absence in the case of the
derivative-following strategy) as resulting from the underlying
topology of the profit landscape, a concept that was introduced
in earlier papers [1, 2, 3].
Preliminary explorations indicate that multi-peaked landscapes and
price-war dynamics can occur in cases more general than were reported here.
We have observed these phenomena in hybrid populations of buyers
with 
, and in which 
and 
are not strictly correlated. In these cases, the nature of the
price war is slightly different: a seller offering a higher quality
can ``undercut'' a lower-quality seller even by offering its product
at a higher price. Price wars consist of general downward trends in
the sellers' prices, with what appear to be
roughly constant non-zero gaps between the prices.
Another fruitful avenue for further research is to explore the
effect of inhomogenous pricing strategies. In preliminary work along
these lines, we have watched a single myoptimal take advantage of
four derivative followers.
Finally, we expect the dynamics to get considerably more interesting
when we permit sellers to vary their quality and their price
simultaneously in an effort to maximize their profit. Previous work
on an information filtering model [1, 2, 3, 4]
has revealed the existence of very
complex price and niche wars, in which the sellers in the economy
all attempt to grab the same niche at the same price, leaving a large
segment of the buyer population unsatisfied. It will be interesting
to see whether sellers in a vertically differentiated market will
all attempt to fight over some perceived optimal quality level,
leaving other quality levels abandoned. If so, it will be very
important to understand this phenomenon well enough to devise
mechanisms for thwarting it, as it jeopardizes the viability of
this and other automated agent economies.
