This observation brings us to acrucial point: vote choice seems to coincidewith both partisan and non-partisanvariables (as shown by the NES, Holbrook,and Fair), but campaigns “matter” becausethey create and reinforce voters’ impressions of candidates, and thus help shape theirchoices.

As Just and her colleagues write,”One of the most striking conclusions thatone takes away from the in-depth interviewsis that it is impossible to categorize theindividuals in our panel as party voters,issue voters, personal voters, and so forth.Instead, individuals have a wide range (ifnot always a large amount) of informationabout the candidates, which they try tointegrate into a meaningful whole.”20 Inessence, they argue that voters accept hordesof information from local and cable news,the internet, advertising, and campaignevents, and they use it to infer one aspect ofa candidate from another. For example,some voters may deduce a candidate’s issuestances from his partisan affiliation, butothers critique a candidate’s abilities as acampaigner. Of particular importance forthese voters is a candidate’s commitment toa particular problem, such as the economy.From his willingness to campaign on anissue which, as we have seen, is salient formany voters, one might make judgmentsabout a candidate’s character and ability tomanage the economy to that voter’sbenefit.

21 Campaigns therefore provide theinformation necessary to make bothretrospective and prospective evaluations.In-depth interviews have proven that votersdo not simply “sense” things like thedirection of the economy, as Fair suggests,but that they actually depend on campaignsto help them make their decisions.22This argument rejects another ofFair’s assumptions, that campaigns arealways on equal footing and thus “canceleach other out.” In reality, campaignsconvey certain messages, which are filteredthrough the media and evaluated by voters.This argument rests on its own criticalassumption, that information-gatheringinfluences vote choice. Holbrook, inrejecting his own model, presents somecompelling evidence to suggest that it does.

Specifically, he cites Popkin’s work, whichsays that voters demand information fromcampaigns because they pay very littleattention to public affairs between elections.Holbrook also draws on Gelman and King’sargument that repeated swings in publicopinion indicate that voters continuallyanalyze campaign events, and that theirchoices are therefore informed, not simplysubconscious expressions of preexistingrealties. Because their choices are informed,they are somewhat predictable. Campaignevents, however, reduce elections’predictability because they can sway publicopinion inequitably, especially inbattleground states.23 As we have seen, inthe 2000 election, both individual andaggregate data indicated that Gore shouldhave won a close race. Instead, he lostdespite winning the popular vote, whichcannot be explained by forecasting modelsbut is explained by campaign and ElectoralCollege effects.

The fact remains, though, thatforecasting models are surprisingly accurate,and while they may neglect a few criticalpercentage points and some institutionalfactors, a stronger argument is required oncampaigns’ behalf. One of Holbrook’s finalconclusions explains the success of thesemodels, and why it does not disparage therole of campaigns. According to Holbrook,campaigns are as processes which generate aproduct, because, as we have seen, theyclearly influence public opinion and allowvoters to make choices based on rationalconsiderations.24 The positive, thoughimperfect, correlation of non-campaignfactors with electoral outcomes only provesthat campaigns are successful in forcingvoters to use these rational considerations,and that other, less-predictable factors are atplay. The work of Just and her colleagueshelps clarify this point, that presidential