Representation: Elections and Beyond

Chapter 5. Fairness and Bias in Electoral Systems

CHAPTER 5 Fairness and Bias in Electoral Systems Anthony McGann As scientists and citizens, we would like to evaluate objectively the fairness and bias of electoral systems. However, this is claimed to be impossible by many political scientists, as well as by various practicing politicians and the U.S. Supreme Court. The argument commonly made is that there are multiple , competing conceptions of fairness; and that once you exclude the most egregious and obvious violations, you cannot talk about one electoral system being fairer or less biased than another, but only lay out the choices or tradeoffs between competing goals. The claim that there are many definitions of fairness then slides very easily into the conclusion that they are all equally deserving of consideration. I wish to reject this relativism. I wish to do this not because it is normatively enfeebling, but because it is logically flawed. It is, in fact, possible to evaluate fairness and bias in electoral systems objectively, at least if you define fairness in terms of liberal equality, that is, treating each voter equally. And once we do this, the existing empirical literature tells us all we need to know to draw conclusions. In order to evaluate the fairness and bias of electoral systems, we need two things: a way of classifying electoral systems in terms of their mechanics , and a benchmark for evaluating fairness/bias. In terms of the first requirement, there is a well-developed literature classifying electoral systems that we can draw upon (see among others Rae 1967; Taagepera and Shugart 1989; Lijphart 1994; Katz 1997; Farrell 2001; Taagepera 2009). As Taagepera and Shugart (1989) argue, district magnitude (number of seats per electoral district) provides a first approximation for such a classification . This, of course, captures the distinction between single-member- Fairness and Bias in Electoral Systems 91 district plurality election (district magnitude of one) and proportional representation (multi-member districts). Furthermore, many of the other relevant characteristics can be expressed in terms of the equivalent district magnitude required to produce the same outcome. For example, Taagepera and Shugart (1989) introduce the concept of effective district magnitude, which combines the effect of district magnitude and electoral thresholds. Likewise, Taagepera (2009) shows that mixed electoral systems and differences in electoral formulas can be accommodated by the same measure. Of course, all characteristics of electoral systems cannot be reduced to a single unidimensional measure, so I will make use of other features (such as whether the system is open or closed list, for example) as necessary. The second thing we need is a benchmark against which to measure the bias of electoral systems. The principle of proportionality—seats awarded are proportional to votes—provides such a baseline. Not only is it mathematically simple, but it is logically implied by liberal political equality, as demonstrated formally by Hout and McGann (2009a, b). If seats are not allocated proportionally, then it must be the case that everyone is not being treated equally. This then poses the question of who is benefiting from this bias and who is losing. The main task of this paper is to classify and understand these various biases. An obvious objection is that I have just assumed a normative preference for proportional representation. This is actually not the case. The principle of proportionality can be shown to be logically necessary if all individual voters are to be treated equally—that is, if the electoral system is to be unbiased . Therefore, deviations from proportionality imply bias, which it is then our task to explain. Furthermore, it is important to note that proportionality is a principle, not an election system—proportional representation (PR) electoral systems are designed to produce proportionality, but vary considerably in how far they achieve this goal, especially when they have low district magnitude. Likewise, single-member-district plurality systems may sometimes approximate proportionality. Indeed, if we have a singlemember -district system, we are likely to be interested in what biases result from the way this system is implemented (for example, how the districts are drawn). Using proportionality as a benchmark for comparisons does not limit us to simplistic statement of the type “PR good, SMDP bad,” but actually allows us to build a framework in which we can make nuanced comparisons of small differences in electoral institutions. 92 Anthony McGann It is notable that the standard of treating all voters fairly is more demanding than that of treating all...