Archival Methods

Methodology

An archival data analysis begins like most research in the behavioral sciences, namely with a hypothesis or at least conjecture about the way human beings behave in a given situation or set of situations. Yet from that point, archival analyses must depart from such standard methods as the experiment, survey, or interview. This departure occurs because the raw data already exist, requiring that the investigator only determine the data's location and the optimal manner of exploiting the data retrieved. The next steps after data retrieval are also distinctive in comparison to more mainstream methods. In the first place, the investigator must decide on the most appropriate unit of analysis. Whereas for most research the unit would be the individual research participant, archival data analyses have more options. For instance, the units might consist of consecutive years of national divorce statistics, film clips depicting scenes in which two characters fall in love, or newspaper comics whose main characters are parents, friends, or coworkers. The choice of analytical unit will often represent a compromise between the hypothesis being tested and the data that are actually available.

The next step is to define the sampling criterion or criteria. Here again there are abundant choices. The units might be confined to specific years and to geographical locations. The units might be randomly selected from a larger sample or even taken to exhaust the population (e.g., all Pulitzer-winning novels). Once the sample is defined, the raw data must be collected and prepared for variable coding and statistical analysis. Although sometimes archival data are already in a form suitable for data analysis, more often they must be first subjected to a well-defined coding scheme, most commonly some form a content analysis. These coding procedures will include not just the substantive variables of interest but also any control variables needed to avoid methodological artifacts. Once this process is complete, the investigator can apply various statistical techniques, such as multiple regression, factor analysis, structural equation models, and time-series analysis. Because archival data studies are inherently correlational in nature, the statistical analyses will often have to be more sophisticated than in laboratory data over which the investigator can exert experimental control.

...