Existing literature suggests that, if anything, larger constituencies make reelection more challenging for legislative incumbents. Yet theoretical arguments commonly are unclear about why more populous constituencies should disadvantage incumbents relative to challengers. Additionally, most empirical research has focused on congressional races, none of which involve small populations. I argue that there is good reason to believe that smaller size enhances challenger viability and thereby lowers incumbent safety. I also provide an empirical analysis of the relationship between constituency size and incumbent safety, focusing on individual contests featuring an enormous variance in district population and other research advantages: races for positions as California county supervisors. Controlling for both candidate and demographic variables, and examining both likelihood of winning and incumbent vote percentage, I find that there is a modest but consistent positive relationship between constituency size and incumbent safety.
Various studies have considered the question of whether constituency size affects the reelection prospects of incumbent American legislators. Some scholars have found that larger constituencies produce more electoral losses for incumbents and/or lower victory margins (see especially Hibbing and Brandes 1983; see also Abramowitz 1988; Carey Niemi, and Powell 2000; Lee and Oppenheimer 1999, ch. 4,) Others have found a negligible or nonexistent relationship between constituency size and measures of incumbency safety (Black 1974; Krasn 1994: ch. 3; Westlye 1991: ch. 7: Bledsoe 1993: 142-43). To my knowledge, no empirical study has argued that larger constituency size enhances electoral safety.
Yet there is reason to be suspicious about whether previous research has captured the critical variance in the constituency size variable. Scholars generally have focused on elections to the U.S. Senate, taking advantage of the "natural experiment" created by the American Founders (i.e., the ability to compare electoral success rates for the same office across states of very different sizes). Such research has sometimes been bolstered by consideration of U.S. House races. Of course, Senate and House races are most important in terms of potential impact on public policy, and therefore intrinsically worthy of study. At the same time, in comparison to the electorates for other American legislative races (and indeed, legislative electorates in most other democratic countries), all of the congressional constituencies are quite large. Consider statewide congressional contests in Alaska and Wyoming, with populations of 627,000 and 494,000 in 2000, respectively. While sparsely populated in comparison to other states, Alaska and Wyoming have constituencies that vastly exceed in population the constituencies for virtually all American state and local electoral contests as well as the constituencies of other legislatures using the single district, first-past-the-post electoral system such as the British House of Commons, Canadian Parliament, Canadian provincial legislatures, etc. Alaska and Wyoming (and to a lesser extent many other states) are also quite dissimilar from other entities in terms of the unusual challenges they pose for interacting with voters in states with land masses of more than 570,000 and 97,000 square miles, respectively.
Methodologists generally recommend examining the full range of an explanatory variable to enhance certainty about causal inferences (King, Keohane, and Verba 1994: 215-16). Accordingly, it is desirable to examine American subnational elections to understand the impact of district size on incumbent safety. For example, as Carey, Niemi, and Powell (2000: 684) indicate, state legislative districts vary in size from a few thousand to hundreds of thousands of residents. Yet aside from the work by Carey and his colleagues, little systematic attention has been given to the effect of variance in district size at the state or local levels. …