Campaign Dynamics: The Race for Governor

Notes

Notes Chapter 2 1. Electoral considerations certainly continue to play a role in gubernatorial behavior after the election. Also, voters continue to observe and evaluate candidates after the election is over. However, during the campaign itself both candidates and voters are focused specifically on the election, and that time period serves as the focus of this analysis. 2. This rather simplified discussion assumes voting is costless. Downs deals with th idea that the costs of voting rarely outweigh its direct benefits if you weigh the poten tial benefits by the probability that any one vote will make the diference in an election. Thus, Downs acknowledges that many voters may not participate. However , this sort of failure to vote differs from my argument regarding activists intentionally not participating in one election because of a long-term focus on the electoral process. 3. This discussion does not ignore those infrequent cases in contemporary U.S. politics that have witnessed the success of independent candidates. These cases, however, do not alter the fundamental institutional structure that the two-party system creates and should not be considered in any way sufficient evidence to reject this theor . 4. This phenomenon opens the possibility for “campaigns” run by people other than the candidates to influence the salience of various factors or cleavages among voters. I particular, the efforts of interest groups can be melded into this theory of voting behavior . 5. In fact, some scholars may criticize the process by which presidential outcomes are predicted based on these models, several of which are cited by Gelman and King. Typically, such models include no more than a dozen previous elections (Gelman and King include eleven in their predictions). The number of independent variables begins to approach the number of elections in the model once dummies for outlier years, candidate home-state advantage, and the proportion of the population that is Catholic (for 1960) are included. To get around this degrees-of-freedom problem, these models disaggregate to the state level, produce predicted outcomes for each state, then reaggregate the state outcomes to produce a national estimate of who will win the election. 6. Such was the case in the 1990 gubernatorial election in Minnesota.The Republican nominee, Jon Grunseth, was eventually forced to withdraw from the race after the publication of charges that he swam nude with minors and had a long-running extra193 marital affair. In 1982 in Vermont, Democratic challenger Madeline Kunin repeatedly attacked Governor Richard Snelling for secretly allowing the transportation of highlevel radioactive waste across the state. In 1992 in Missouri, Democrat Mel Carnahan repeatedly accused his Republican counterpart, William Webster, of improper conduct as attorney general while distributing financial awards in workers compensation cases. Webster lost and later pleaded guilty to federal conspiracy and embezzlement char ges. Chapter 3 1. Anumber of works are available that provide more complete introductions to spatial voting models. Interested readers should begin with Downs 1957 and Black 1958, then explore Enelow and Hinich 1984 and Hinich and Munger 1997. Most of what is presented here has been developed in these works. 2. But see the directional theory developed by Rabinowitz and Macdonald 1989. 3. Using the square of the distance produces a measure of distance that is always positive. Using the absolute value of the distance would produce the same result. I employ the squared distance here because it facilitates later developments in the model. 4. Another common notation in the literature is to treat X as a vector of voters and use the xi to indicate the “ith” voter. The basic structure of what follows, however, does not change. 5. Beginning with (x ⫺ Pxa )2 , substitute in pa ⫹ vxa . This results in: (x ⫺ (pa ⫹ vxa ))2 This can be rewritten as: (x ⫺ pa ⫺ vxa )2 Expanding the squared term leads to: x2 ⫺ 2xpa ⫹ pa 2 ⫺ 2xvxa ⫹ 2pavxa ⫹ vxa 2 Because the expected value of vxa ⫽ 0 (see Alvarez 1997; Enelow and Hinich 1984), this expression simplifies to x2 ⫺ 2xpa ⫹ pa 2 ⫹ vxa 2 Collecting terms, the final expression is (x ⫺ pa)2 ⫹ vxa 2 6. SeeAlvarez 1997 for a more complete elaboration of these points and Alvarez and Franklin 1994 and Franklin 1991 for some empirical treatments of these issues. 7. The discussion of uncertainty due to incomplete information is readily transferable into the multidimensional model I explore, as is the term for nonpolicy factors introduced earlier. However, to present the issues raised by a multidimensional policy is194 Notes to Pages 24–28 [3.144...