Missing, biased and unrepresentative : the quantitative analysis of multisource biographical data.

With the growth in interest in collective biography as a historical technique, many predominantly qualitative historians find themselves faced with large amounts of information. These data, collected from a variety of sources, are often highly irregular, making statistical analysis extremely problematic. Current practice is to ignore these problems and proceed with quantitative analysis suitable only for much more regular data. It is argued that a more satisfactory approach is to ascertain and directly confront the difficulties of analyzing such information. The three central problems are identified as missing data, systematic bias, and the lack of a representative sample. Using a practical example, the author explores the relationship between gender, the family, and political socialization within the Communist Party of Great Britain and shows how each of the issues can be dealt with in turn. The author first distinguishes truly missing data from "negative information," which commonly appears to be missing in historical sources. He then stratifies the data to remove systematic biases relating to the issue at hand. Finally, he divides the sample into different populations, on the basis of the sources from which individuals are known, and compares the results obtained to examine whether his conclusions appear to depend on quirks of populations contained in the sources. These ideas open a new range of sources to quantitative analysis and raise the possibility of allowing new types of evidence to count in historical inquiry