|
|
Conclusions and open questionsIn this paper, we have investigated situations in which price wars spontaneously arise in a population of competing information brokers. With certain reservations, our results suggest that price wars are a basic and undesirable feature of agent-based societies in which agents perform at ``myopically optimal'' levels, in the sense of having perfect information but no memory. In the simplest price wars, two or more brokers offering the same product (set of categories) successively undercut one another until some or all of the brokers jump discontinuously to other, less competitive niches. The short-sighted optimality of the myoptimal strategy causes an immediate resumption of the price war, and a never-ending cycle of fairly regular price wars ensues. Less optimal but equally myopic policies may actually lead to better collective behavior in the sense that both the brokers and the consumers have higher average utilities overall. However, the underlying myopia makes the system inherently unstable, and periods of relative calm and prosperity will necessarily be punctuated sporadically with price wars. During the course of a price war, the product offered by the brokers may be highly volative. During these more complex price wars, brokers offering different products may still find themselves in partial competition with one another, and the downward evolution of the prices cannot be understood merely as successive undercutting. In any case, the results presented here suggest several avenues of further research. As developed in section 2, the model made no provision for the cost of switching categories. In any real situation, there is likely be a nonzero cost of adding or dropping a category -- e.g., capital outlays for additional bandwidth, storage, or processing power. The three-broker, three-category price wars we observed crucially depended on brokers being able to freely switch categories in order to get the ``best'' profit. Some early investigations into the effects of switching costs show that they may or may not quench price wars, depending on the relationship between the magnitude of the switching cost and the asymmetry among various choices of interest vectors. Note, however, that an excessive switching cost can also prevent brokers from switching out of a price-war situation. Similarly, in any real business there are a number of ``friction'' forces opposing change in general, which may help prevent price wars. These include costs associated with changing the price of a good and less-than-perfect responsiveness of the consumer population. Randomizing effects such as incomplete or false information on the part of the brokers would tend to decohere brokers' behavior. Another approach to the problem would endow the agents with the capacity to avoid price wars by looking ahead further. One approach would be to try to predict the future behavior of the system by simulating it, but this almost certainly requires too much knowledge of system state and the strategies of the other agents. A more feasible approach might be to use a machine-learning algorithm that learns to avoid price wars, but the question remains whether this is possible, and if so how. Of course, the dynamics of learning itself is likely to cause other emergent behaviors, some of which may themselves be undesirable. If all else fails, and competition among independent brokers inevitably leads to prices wars, a third category of solution is still available. In such a case, the demands of profitability would give a strong motivation for brokers to form cartels and set prices across the board. It is an open question whether efficient, effective cartels can be created, and their rules enforced, in an open population of agents.
|
|