Some survey researchers find use for data from non-asking sources. This is a positive development! Three cheers for survey researchers! Reliance on survey research is weakening and this erosion of confidence in the asking method will lead them to accept and use other methods of research, such as direct observation, observation of behavioral traces, and content analysis. Excerpts from Newsletter No. 102 (below) describe the use of content analysis of administrative records. I have put the most relevant parts in bold. (I received this newsletter from a university survey research center.)
Using Survey Data in Combination with Data from Other Sources: Data from Administrative Records
Over the last several bulletins, types of data that are often combined with survey data have been described. In SNB #102, we finish this series with a discussion of incorporating data from administrative records with survey data. Administrative records include any data that is kept by a business, government agency, or other organization for the purposes of their work or function. For example, it might include school records of students’ grades, medical records, or financial or tax records for an organization or business. . . .
As with many other external sources of data, data from administrative records can also be used to estimate nonresponse bias. For example, in a survey of the parents of children in a specific public school district, researchers could use administrative records to assess whether response rates are different across school or grade level. Similarly, given access they could compare the grades, test scores, or attendance records of students whose parents did and did not participate.
Administrative records can also contain useful substantive information. Medical records checks can be used to assess the accuracy of survey self-reports or to incorporate additional specific medical information with survey data. Similarly, administrative records of voter turnout, for example, are sometimes used as an objective measure of whether a respondent voted in an election. These records are used both to check the accuracy of respondents’ self-report of voting (because respondents tend to overreport turnout) and as a measure of turnout. . . .”
For a further discussion of the need for non-asking data to check or verify self-reports, and for a complete statement of the weaknesses of survey research, see The Problem with Survey Research.