Determining how political parties should and/or do position themselves in “issue space” (that is, the combination of policy positions that they adopt) has long been a major preoccupation of political scientists. Just when I thought that the innovative work that could be done on this set of issues had already been done, along come James Fowler and Michael Laver, who, taking a page out of Robert Axelrod’s studies of cooperative behavior, hosted a “tournament” of competing strategic perspectives.
In “A Tournament of Party Decision Rules,” Journal of Conflict Resolution, February 2008), Fowler and Laver began with four base strategies: STICKER (never change positions), AGGREGATOR (position the party on each dimension at the mean position of the supporters of the party), HUNTER (if the party’s last repositioning produced increased support, make the same move again; otherwise, strike off in a different direction), and PREDATOR (if yours is the largest party, stay put; otherwise, move toward the largest party’s position). These base strategies were automatically entered in the tournament, but they were supplemented by submissions — 25 in all — from various scholars. Thus understood, the winner of the tournament would be the strategy that left the party in the strongest competitive position at the end of the competition.
To assess the performance of these strategies, Fowler and Laver conducted a complex set of simulations. In these simulations, several of the submitted strategies performed poorly; indeed, 16 of the 25 submissions proved to be even less effective than the totally immobile STICKER strategy. The tournament winner, by contrast, was more successful even than HUNTER, the most successful of the four base strategies. The winner, which was submitted by Kevin Quinn, was “in essence a satisficing parasite with a secret handshake, hardwired not to attack itself.” (Whew!)
bq. KQ [for Kevin Quinn] parties jittered when over the [survival] threshold, using tiny random moves with a very distinctive step size recognizable to other KQ-parties. When under the threshold, a KQ-party moved very (very) close to the position of a randomly selected other party over the threshold, provided this was not another KQ-party.
More generally, certain regularities stood out in the submissions and results. For one thing, successful strategies tended to locate their parties about one standard deviation away from the center of the space. For another, center-seeking strategies, as well as strategies, e.g., AGGREGATOR, that set policy positions at more extreme locations, fared much worse. Strategies that caused parties to move a considerable distance from one election to the next were also less successful. So were strategies that sought out a fixed point, such as the voter centroid.
Fowler and Laver caution, appropriately, that “The conclusions we can draw on the basis of the present tournament depend critically upon the rules that were submitted to it. And substantive inferences we might draw from the patterns we observe depend critically on the assumption that these patterns are not tournament artifiacts.”
For the future, Fowler and Laver expect to conduct more tournaments in order to play off various strategies against one another within a reasonable set of assumptions. Pretty cool stuff, I think.