Presidential Election Polls, 1988: The Gallup/Conus Reports
About this Collection
Introduction: 1988 Presidential Election Polls: The Gallup/Conus Reports
The Gallup Organization Inc.
The election year of 1988 represented a milestone in modernpresidential politics. For the first time since state primaries and caucusesdecided the nominating process, the presidential primary season began with anopen race in both parties. Barred from seeking reelection by the 22d Amendment,President Ronald Reagan, the dominant figure in national politics during the1980s, chose to stay above the fray rather than to endorse George Bush, hisvice president for two terms, or any other candidate for the Republican partysnomination.
While Bushs tenure as vice president helped him become theearly front-runner for the Republican nomination, many political observersquestioned whether, after serving in a subordinate role for eight years, hecould become a convincing leader. Journalists noted that no vice president hadbeen elected president since Martin Van Buren defeated William Henry Harrisonin 1836.
There was also no clear front-runner for the Democrats.Neither Gary Hart, whose candidacy suffered after the Miami Heraldexposed his extramarital affair with model Donna Rice, nor Ted Kennedy, wholost the nomination to Walter Mondale in 1984, entered the race. Of the sevenDemocratic candidates, only Jesse Jackson was well known to voters across thecountry. But if there was one thing that most political observers agreed upon,it was that no black candidate, especially one as controversial as Jackson, waslikely to end up winning the Democratic partys nomination.
The Gallup Organization approached the 1988 election withover fifty years of presidential election polling experience since GeorgeGallups first poll in 1936. Since then the survey methods had been refined,but the way the polls were conducted and distributed remained much the same. In1984, as in previous elections, the Gallup Poll depended on national in-personinterviews for its reports on the presidential race, and the results werereleased in a nationally syndicated newspaper column.
All that changed in 1988. To cover the wide-open 1988presidential race, Gallups Andrew Kohut and Nancy Nygreen combined with ConusCommunications, which specialized in providing satellite news feeds to localtelevision stations. This resulted in the development of a preelection pollingpackage for news organizations that was unprecedented in its scope.
The tight deadlines of television news departments requiredmajor changes in how the polls were conducted. For the first time, Gallupsregular presidential polling was done entirely by telephone, which reduced theturnaround time between drafting survey questions and reporting results. Gallupalso offered state and regional surveys to subscribers to provide morelocalized information during the primary season.
The Gallup polls eventually found roughly twenty-five U.S.television sponsors who owned first broadcast rights in their local markets.WEEI Boston made the poll a regular feature in its coverage of the campaign,and NHK Tokyo was a major foreign broadcast sponsor.
Although Gallup broke new ground with the addition ofbroadcast outlets for its presidential polls in 1988, the polls continued to befeatured in close to one hundred domestic newspapers that either subscribed tothe polling package or received the results on a somewhat delayed basis in thebiweekly Gallup Poll column. The Los Angeles Times Syndicate distributed thereports to newspaper subscribers; newspapers that carried poll results regularlyincluded the New York Times, the Washington Post, the LosAngeles Times, and Japans Yomiuri Shimbun.
Aside from the broadcast and newspaper subscribers, thepolling service was also offered to political, governmental, academic, andbusiness groups. The polls had subscribers among all of these sectors,including research firms, political committees, university libraries, and majorcorporations. The costs of the reports varied by the type of subscriber, themarket, and the material provided. While a big city newspaper purchasing thepackage on a market-exclusive basis paid as much as $25,000 for the regularweekly reports, an institution interested in receiving data sets on a delayedbasis paid about $5,000. Subscribers interested in receiving only the weeklyreports, not for immediate publication or broadcast, paid less that $1,000.
The polls were designed to provide material for forty-fourweekly television and newspaper stories on the presidential race and to makethe information as timely and relevant to campaign coverage as possible. Thepolling schedule, completed in December 1987, planned for thirty-three pollsthrough the November election, including eighteen polls in specific states orregions.
Because of Gallups objective to provide in-depth coverageof the nominating primaries and caucuses, twenty-two of the polls wereconducted between the time of the Iowa caucuses, the first primary election,and the June California primary, the last major primary. The polling schedulealso reflected the nomination process. Thirteen of the polls were conductedbetween early January and Super Tuesday on March 8, when thirteen states,mostly in the South, held primaries or caucuses.
During the primary season, each survey consisted of 1,200interviews among registered voters, except for those projection polls thatwere conducted in the final weekend before a state primary. These final pollswere based on a sample of 3,000 registered voters per state. Once the primariesended, the sample size for the national polls was reduced to 1,000 interviews,with a few exceptions. The final November poll had a sample size of 3,000interviews.
The telephone survey methodology used in the preelectionpolls was consistent. Registered voters were screened from a random sample ofresidential telephone numbers for the geographic area being sampled, whether itwas a state, a region, or the nation. Nonvoters were asked three or fourdemographic questions so that the sample could be weighted according to thelatest census information about adult population, including sex, age, race,education, and, in some cases, geographic region and Hispanic origin.
Most of the interviews were conducted over a three-dayperiod beginning on a Friday evening and ending on Sunday. If the first telephonecall to an eligible respondent failed to make contact, two more attempts weremade. During surveys immediately before an election, refusal conversions wereattempted by telephone interviewing supervisors who called back thosehouseholds where someone initially declined to be interviewed.
The final national survey and the final projectionpre-primary polls in individual states were designed to provide daily resultsover the three-day period such that any late shifts in candidate support couldbe detected. For these final polls, 1,000 interviews were conducted each day ofthe three-day period. Each set of results was then weighted separately in orderto minimize demographic differences between each days results.
Polls conducted more than a week before an election day werebased on all registered voters interviewed or, in the case of the primaries andcaucuses, based on all who intended to vote. Polls conducted in the final daysbefore an election included a set of turnout questions so that results couldalso be reported on a likely voter base. The turnout questions used were thosePaul Perry and others at Gallup had developed thirty years earlier, and thequestions had demonstrated their accuracy in later studies of Gallups personalinterview preelection surveys.
The format of the reports sent to subscribers each weekduring the campaign was standard. Each report began with a description of thesurvey methodology, followed by a short report highlighting the results to afew key questions - generally the trial heat or horse race results. Thisfirst section was compiled for television subscribers and was followed by alonger analytical piece that Andrew Kohut and I wrote for print subscribers.The weekly reports also included selected tables that broke out demographicdifferences to key questions, plus a technical appendix that included the exactquestion wordings, question-by-question results, and information about thesample.
The reports were written under tight deadlines. They weretypically put together on a Monday and sent out to clients that same day;television stations were embargoed from releasing the results until 5 oclockon Tuesday, while newspapers were embargoed from publishing the results untilWednesday morning. The schedule for reporting the final projection pollresults was even tighter. Gallup processed and analyzed the results from thefirst two nights of interviewing on Sunday morning for release on televisionthat evening and inclusion in Monday morning newspapers. Subscribers thenreceived the full results on Monday for immediate release, based on the fullthree days of interviewing.
Unlike exit polls, where the final analysis is releasedafter the results of the election are known, the Gallup polls had to becompleted before the election outcome was certain. In retrospect, there are afew cases where the analysis clearly missed the mark, such as the finalreporting on the New Hampshire Republican primary. But for the most part thepolls held up well.
Gallups reports provided a perspective on the race that wasoften missing from other election polls. When Michael Dukakis led early in thecampaign, Gallup cautioned against reading too much into the early lead becausehe lacked a well-defined image. Gallups polls also debunked the notion thatGeorge Bush had such a negative wimp image among voters that his candidacywas doomed to failure. In July, prior to the Democratic national convention andwell before the Willie Horton ads aired, the Gallup polls identified theMassachusetts prison furlough issue as a key factor behind the drop inDukakiss popularity. The polls also stated correctly that the Quayle factor(Bushs selection of Dan Quayle as his running mate) was unlikely to have muchof an effect on the outcome of the election. Despite the bad reviews ofQuayles performance in his debate against Democratic vice presidentialcandidate Lloyd Bentsen, later polls showed that the proportion of voters whosaid Quayle was qualified to be president had actually increased since the debate.
These reports and the supporting data represent the singlebest source of preelection poll findings and analysis for historians, politicalscientists, and graduate students. Historians can find a wealth of informationabout how the public viewed the candidates at various stages of the campaign.Particularly striking is the improvement in George Bushs image immediatelyafter the Republican national convention. Political scientists and graduatestudents can also find information about how the voters in 1988 viewed thepolitical parties, the presidential nominating process, and the candidatestelevision advertisements.
The 1988 presidential election may end up as a milestone ofa different sort. Election reforms, such as shortening the campaign and restrictingthe amount and types of advertising, became an issue in the later stages of thecampaign. Noted more for its war of made-for-television sound bytes andnegative advertising than for contrasts in political ideologies, the campaignmay prove later to have been the one to spur substantive political reform.Whatever the eventual judgment of history on the 1988 campaign, the Gallupreports will remain one of the best sources of information about how Americansviewed politics during that year.