Wednesday 27 July 14:00 to Friday 29 July 14:00
University of Hohenheim, Stuttgart
In this workshop, experts will teach the theoretical and practical skills necessary for analysing massive amonts of serial historical sources.
Furthermore, hands-on exercises will introduce the use of the statistics package "R" for the implementation of media analyses.
Members of individual projects will discuss possible synergies for their research, as, for example, a joint digitization (pdf and OCR) of historical journals.
The first of a series of workshops on methodological and theoretical questions concerning the extraction of economic experience and expectations from historical sources took place from 27 to 29 July at the University of Hohenheim and was organized by Prof. Dr. Sibylle Lehmann-Hasemeyer (Hohenheim) and Prof. Dr. Ulrich Fritsche (Hamburg). A combination of expert lectures, presentations from the PP 1859’s researchers projects, and a hands-on tutorial on statistical analysis with R, the workshop proved a fruitful kick-off event and was helpful in refining the participants’ research tools.
The workshop created greater awareness for the possibilities and especially the limits of content analysis in capturing complex meaning. The necessary expenses for a systematic extensive content analysis are only justifiable if the method promises insights going beyond what can be gleaned by more traditional research methods. In the absence of serial, standardized sources, which is the case for many historical projects, its applicability seems limited. Measuring the frequency of expressions used in serial media can show trends in economic communication, but is not without pitfalls. For the programme’s purposes, it is necessary to include the temporal dimension of statements (referring to the past, present, or future) into the coding in order to reveal changes in behavior as the result of learning, which means that automated coding probably is not possible. A central insight won was the observation from an earlier project that economic actors are influenced asymmetrically by media information, with bad news having a distinct effect on individual’s expectations of inflation and good news hardly affecting inflation expectations. This example highlighted the necessity to construct a plausible model of recipients’ processing of information and the translations of expectations into economic action. Learning models appear especially suitable, since they allow tracing how recipients’ behavior is modified by repeated confirmation or frustration of their expectations. The change of language through history poses a particular challenge for automated content analysis. So far, automated methods have only yielded disappointing results when trying to recognize connotations and tonality of statements. Only in the case of perfectly defined expressions, these methods prove reliable. The most promising approach so far seems to be a mix of quantitative and qualitative methods.
The problems posed by the extraction of meaning from highly diversified source material that could reveal economic actor’s related formation of experience and expectations will be further discussed at the workshops on archival work in Berlin and the theory of economic expectations in Düsseldorf. On a more practical note, the workshop also spread useful knowledge about methods of digitalization and the preparation of data, for example how to successfully use optical character recognition solutions for old scripts. Future work shall include refining methods to prepare structured data from historical documents, the intensification of data-sharing and co-operation between projects, and exploring the accessibility of existing data collections behind paywalls. The participants agree to compare results and methods systematically at the Priority Programme’s yearly conference in February 2017.