Disclaimer. Don't rely on these old notes in lieu of reading the literature, but they can jog your memory. As a grad student long ago, my peers and I collaborated to write and exchange summaries of political science research. I posted them to a wiki-style website. "Wikisum" is now dead but archived here. I cannot vouch for these notes' accuracy, nor can I even say who wrote them. If you have more recent summaries to add to this collection, send them my way I guess. Sorry for the ads; they cover the costs of keeping this online.
Fowler. 2006. Habitual voting and behavioral turnout. Journal of Politics 68 (May): 335-344.
Scholars in this specialized area develop formal models to best fit actual empirical results (behavior of voters). Scholars formalize behavioral assumptions and then evaluate whether or not predictions match up with actual data on voter behavior. The debate is over which model fits better.
Fowler's paper is a critique of BDT (Bendor, Diermeir, and Ting)'s learning model. BDT had a learning model in which a successful outcome reinforces the behavior (i.e. if you vote and your guy wins, you'll vote again; if you vote and your guy loses, you'll abstain next time). BDT's model successfully predicts a realistic (aggregate) level of turnout (roughly 50%).
The BDT paper's main problem is in the feedback mechanism. The model's monotonic feedback moderates the model's predictions by definition, causing every voter's probability of voting to converge to 0.50. (See pg. 337 for a definition of "moderating feedback.") Thus, the finding that (aggregate) turnout is high (under BDT's model) is merely artifactual--it is a product of their feedback mechanism. As such, changing the input parameter's in BDT's model does not affect turnout like it should. Note Table 2, right columns: Even if the cost of voting skyrockets, the BDT model continues to predict turnout near 50%.
Moreover, although the BDT model predicts the right aggregate level of turnout, it makes incorrect conclusions about individual turnout behavior. It concludes that the average voter turns out with probability of 0.50. But this ignores evidence that voting is habitual (i.e. that some people vote with probability close to 1 and others vote with probability near 0). Casual voting conflicts with "well known empirical phenomenon of habitual voting."
Another way of stating this is that the BDT model is not good at explaining the sequential behavior of individuals, which in turn affects the model's computation of turnout. While the BDT model is strong in predicting overall turnout, it cannot explain why individuals' habitual behavior. Fowler's alternative model (without feedback) "matches observed data better because it can generate both habitual voting and high levels of aggregate turnout" (p. 342)--that is, Fowler explains both individual behavior (habitual voting) and aggregate behavior (high levels of turnout).
Fowler proposes a different learning mechanism, which is only a minor change to the BDT model but one which eliminates the convergence to 0.50. With Fowler's change, the model still predictions high turnout (as it should), but it also allows for there to be habitual voters and habitual abstainers.
Page 343 shows the results of Fowler's model (after running a simulation), BDT's model, and actual data from the South Bend study. Fowler's predictions fit much better by predicting that many people won't turn out in any year and some will turn out in every year--unlike the BDT model, which predicts that most people will turn out in half the years being studied.
Research by the same authors
Research on similar subjects