The Seat Product Model matters to electoral and party systems specialists in what it is able to predict, and to all political scientists as one example of how to predict. The seat product (MS) is the product of assembly size (S) and electoral district magnitude (M, number of seats allocated). Without any data input, thinking about conceptual lower and upper limits leads to a sequence of logically grounded models that apply to simple electoral systems. The resulting formulas allow for precise predictions about likely party system outputs, such as the number of parties, the size of the largest party, and other quantities of interest. The predictions are based entirely on institutional inputs. And when tested on real-world electoral data, these predictions are found to explain over 60% of the variance. This means that they provide a baseline expectation, against which actual countries and specific elections can be compared.
To the broader political science audience, this research sends the following message: Interconnected quantitatively predictive relationships are a hallmark of developed science, but they are still rare in social sciences. These relationships can exist with regard to political phenomena if one is on the lookout for them. Logically founded predictions are stronger than merely empirical relationships or predictions of the direction of effects. Finally, isolated equations that connect various factors are nice, but equations that interconnect pack even more predictive punch. Political scientists should strive for connections among connections. This would lead to a more scientific political science.
Predicting the Number of Parties: A Quantitative Model of Duverger's Mechanical Effect
