In lieu of an abstract, here is a brief excerpt of the content:

APPENDIX A Models and Methods for the Case Studies In chapter 4, I described the campaign process as a dynamic interaction between the candidates and voters, with the media playing a key linking role. Here I present a model of that process. While presented as a statistical model, the data necessary for an actual empirical analysis exceed what is available in this project . I specify the model as a set of three simultaneous equations: (A.1) (A.2) (A.3) where SAt and SBt are matrices containing i measures of campaign strategies for Candidates A and B, respectively, directed at potentially salient cleavages among voters. SVjt is a matrix containing j measures of the level of salience of these cleavages among voters. Mjt is a matrix of k measures of the content of SV a SA b SB c SV d M t i it i I i it i I j jt j J k kt t k K = + + + + − = − = − = − = ∑ ∑ ∑ ∑ 3 1 1 3 1 1 3 1 1 3 1 3 1 e SB a SA b SB c SV d M t i it i I i it i I j jt j J k kt t k K = + + + + − = − = − = − = ∑ ∑ ∑ ∑ 2 1 1 2 1 1 2 1 1 2 1 2 1 e SA a SA b SB c SV d M t i it i I i it i I j jt j J k kt t k K = + + + + − = − = − = − = ∑ ∑ ∑ ∑ 1 1 1 1 1 1 1 1 1 1 1 1 1 e 177 the media coverage of the election. For all three equations,a is an i ⫻ 1 vector of coefficients relating the strategy of candidateA at time t ⫺ 1 to each dependent variable, b is an i ⫻ 1 vector of coefficients relating the strategy of candi date B at time t ⫺ 1 to each dependent variable, c is a j ⫻ 1 vector of coeffi cients expressing the relationship between the salience of each cleavage at time t ⫺ 1 and each dependent variable, and d is a k ⫻ 1 vector of coefficients re lating the content of the media at time t ⫺ 1 to each dependent variable. The subscripts t and t ⫺ 1 refer to points in time during any one election. Equations A.1 through A.3 essentially model the dynamic process of the campaigns as a first-order vector autoregressive ( AR) process between candidate strategy and the salience of cleavages among voters because each equation includes lagged values of all three of the dependent variables as regressors. The relationship between the strategies of the two candidates along with the salience of issues to voters as modeled here is also consistent with what Finkel (1995) calls a crosslagged model. The initial values for the salience of each cleavage within the electorate could be determined by public-opinion polling prior to the beginning of the campaign. Much of the political context within which campaigns take place is exogenous to this system of equations. For example, the partisan makeup of a state defines opportunities an /or obstacles for candidates and provides a backdrop for the formation of public opinion. Such factors remain relatively constant over the short duration of a campaign. Candidate strategy may influenc the salience of such factors but not their existence. Exogenous shocks to the system (for example, revelation of a previously unknown scandal) can enter the system through media coverage and/or the error terms of the equations.The error terms of these three equations would likely be contemporaneously correlated with each other. As mentioned previously, detailed (possibly weekly) measures needed to estimate this model for the two cases included in chapters 5 and 6 do not exist. Instead, I have five polls forVirginia conducted from September 1 through the end of October and six polls for New Jersey conducted from June through elec tion day. I also have a poll conducted in February 1993 in New Jersey that asked a limited number of questions about the popularity of Governor Florio. Two of the Virginia polls were conducted by the Survey Research Lab at Virginia Commonwealth University, while the other three were conducted for the Richmond Times-Dispatch by the research department of Media General. Four of the preelection polls from New Jersey and the February poll measuring the popularity of the governor were conducted by the Eagleton Institute of Politics .1 One of the preelection polls and the final poll available from New Jerse, an exit poll, were conducted by CBS /New...

Share