Doing Archival Research in Political Science By Scott A. Frisch, ...

if members could vote on a scale from “strongly agree” to “strongly disagree.” Some tough votes are conducted using voice votes, yet those are excluded from the roll call data record. Finally, some votes must be understood contextually and sequentially. As John Kerry admitted on the campaign trail in 2004, when he was asked about his vote against spending for the occupation of Iraq and Afghanistan, he remarked, “I actually did vote for the $87 billion before I voted against it.”¹² Though one may read Kerry’s remarks as a politician trying to have his cake and eat it, too, they actually highlight the difficulty of understanding an individual’s position on an issue by looking at one vote alone. Given all of these shortcomings, should scholars not use roll call voting data in their research?

Survey research depends on random sampling to measure the attitudes of a population. On average, random sampling does a good job of generating data that represent the attitudes and opinions of a much larger population. But, it is also possible that a random sample can produce data that are entirely unrepresentative of the population. Because random sample surveys can produce misleading data, leading to faulty inferences, one should not rely only on random sample surveys. However, no serious social scientist would argue that such surveys should be rejected altogether because they are occasionally faulty.

All data sources have shortcomings. The question is how the researcher deals with those shortcomings. Given that there will be gaps of all kinds in the archival record, how does one deal with the problem of missing data? Given that politics is, by its nature, a collective enterprise, it is unusual for written communications to vanish entirely. Researchers can explore multiple archival collections for evidence. It may be that data missing in one collection exist in another collection that is closely related.

Conducting interviews is invaluable for the archival researcher. First, it is an excellent method for deepening one’s understanding of process. It is also a very good way to establish the completeness and validity of one’s data collection. For instance, when they are collecting data regarding congressional appropriations earmarks, Frisch and Kelly