The effect of newspaper endorsements and party identification on voting behavior : a controlled laboratory experiment

As political parties have declined as a factor in determining Americans' voting behavior, researchers have begun to study the roles of other variables in the voting process.

Researchers examining the influence of newspaper endorsements have found that the endorsement has a major impact on the voters' behavior. Nearly all of these studies, however, used surveys or content analyses to gather data, and as such, they can provide only a description of the endorsement's role in the voting process.

This laboratory experiment provided a setting in which the endorsement variable could be isolated and tested against party identification. In a hypothetical election in which 251 students at the University of Alabama in Birmingham played the role of voters, the students were divided into three groups and given information about two candidates in a state Senate race. Three separate treatments were administered at one-week intervals.

All three groups received the identical first treatment, a type set newspaper article describing the two candiates. The control group received additional superficial information about the two candidates during the second and third sessions. But, during the second session, the second group was given an endorsement of one of the candidates and third group was told the party affiliation of the two candidates. During the third session, treatments for these two groups were reversed.

Results showed that while members of the control group tended to remain with their original choice throughout all three sessions, respondents in the other two groups switched votes in considerable numbers based on endorsement and party identification. Despite a strong tendency among respondents to switch to the candidate representing their own party, the endorsement produced more statistically significant results, whether given to the subjects before or after party identification was known.

Despite the limitations to external validity found in any laboratory experiment, the study showed that when the endorsement is tested directly against party identification as a determinant of voting behavior among college students, the endorsement effect was strong enough to deny the non-endorsed candidate a majority of the vote among members of his own party.