Abstract: Brentel, I. (2018, 28.06.). How to Make Experiments Replicable – A Case Study of Online Political Stock Markets (OPSM) in Germany. WAPOR 71st Annual Conference, Worl Association for Public Opinion Research (WAPOR). Marrakech, Marokko. [Tandem 1]

Relevance & Research Question:
Election research studies have asked whether surveys influence the electorate’s expectations of election results and what effects have risen from these expectations of voting decisions. Theoretical assumptions have been empirically examined with different experimental designs. In this context among others the following points are critically seen:

  1. Methodological aspects: the electorate’s expectations of election results via questioning; the measurement of the electorate’s expectations of election results takes place only once, but is a dynamic process.
  2. Intersubjectivity of experiments: problems in documentation (esp. the intention of the theoretical base regarding operationalisation, measurements and treatments); This leads to a lack of replication and comparative studies.

With this paper we suggest a solution for these problems. We will discuss how we can make experiments replicable and accessible for other researchers and further for replications as well as comparative studies. This will be done by showing an alternative data resource for public opinion data. The case study’s questions whether experimental designs within OPSMs open up further methodological possibilities to investigate the impact of surveys on the expectations of election results using a longitudinal data source. Thereby, OPSMs operate according to the following principle: The convergence of many participants leads to realistic prices of objects (candidates and parties) which can be compared with poll results and used as election forecast.

Methods & Data:
In order to evaluate the research question of our case study, we conducted an experiment on the influence of published poll data on voting behaviour using an OPSM on the Bundestag elections 2013 provided by a German national newspaper. Between May 1 and July 17, 2013, 413 active participants were randomized and divided into four groups. These groups were shown different presentations of the parties’ survey results (short and long survey trend, daily cross section, no survey). The daily number of trading activities, the kind of transaction, and the transaction price generated the dependent variables.
In making experimental design (measurement considerations, treatments, etc.)  visible and thus data accessible for academic research we used a highly innovative digital documentation and harmonization software, CharmStats, to reach the second goal of our project. CharmStats allows us to document experiments and other data in high quality standards of data documentation.

One target of the project is to make the complex process of data with experimental designs most transparent and replicable. CharmStats offers the possibility to fulfil the project´s goals to enable researchers to reproduce the procedure of an experimental design with the development of definitions to operationalisation towards measurements and treatments. Additionally, CharmStats outputs a report for documentation. Moreover, syntaxes for harmonisations, indexes and aggregation of variables can be compiled for different statistical software.
Finally, the results of the study show that the trading activities differ between the groups and trading activities depended on changes in polls. With OPSM expectations of election outcome could be measured in form of real actions as a result of knowledge about survey results. Besides, temporal dynamics of different expectations could also be mapped longitudinal.