Dictionary of Polling: The Language of Contemporary Opinion Research

TANDEM SURVEYS Two identical surveys. Tandem surveys are taken at the same time, from the same population, but by separate polling organizations. Tandems resemble controlled experiments because they allow the two surveys to be directly compared. And like experiments, tandems control for most extraneous factors that can influence poll results. They do not control, however, for house effects--the term used to describe the idiosyncrasies that distinguish one survey organization from another ( T. W. Smith, 1978:443-463).

Researchers prize tandem surveys because they provide replication--the sine qua non of scientific proof. Replicated surveys produce the strongest evidence and inspire the greatest confidence in the scientific accuracy of a study. If the results obtained are comparable, then reliance on both studies is increased. Unfortunately, tandem surveys are uncommon. Turner and Martin report only five "instances in which two survey organizations cooperated in the fielding of a survey." These range from the Stouffer study on civil liberties conducted in 1954 by Gallup and NORC to the National Health Care Expenditures Survey conducted in 1977 by NORC and the Research Triangle Institute. Interestingly, most of the differences observed in these tandems were attributed to house effects. Otherwise, differences between the compared surveys were small ( Turner and Martin, 1984:149-154). See alsoHOUSE EFFECTS.

TDM Popular acronym for Total Design Method, a widely known set of procedures for carrying out mail surveys. TDM was described by Don Dillman in Mail and Telephone Surveys: The Total Design Method ( 1978). Dillman's book, which has been influential, argues that mail surveys need not be the black sheep of polling methods. Mail's generally notorious reputation may have been deserved, but nothing inherent in the method itself made this inevitable.

According to Dillman, response rates of 90 percent or better are possible

-225-