Reason for the Failure to Replicate Results

Thursday, 19 July 2018: 08:30
Oral Presentation
Andrea BREITENBACH, University Marburg, Germany
Popper was of the opinion that a theory can only be falsified if "we discover a reproducible effect which refutes the theory" (Popper 2005: 66). Although the replication of empirical results has not been a high priority in empirical social research until recently, a replicability crisis has been recognized for some time. This crisis first began in other sciences and has now reached sociology.

For both research and teaching, replication of empirical analyses is an important tool. Thus, many authors, such as Diekmann or King, refer to the need to replicate studies. Diekmann puts forward the important argument that the error rate when using non-empirical surveys is high, but that it can be greatly reduced by replication. In addition, both researchers and students can learn a great deal from replication.

For these reasons, various studies were replicated in different areas of sociology, for example, family sociology, political sociology and environmental sociology, using alternative data sets and methods. It was found that, in many cases, the results did not match the original study. Often, this was attributable to a lack of information or transparency regarding the operationalization of the variables. If the operationalization is described in detail, results which are very close to those of the baseline studies can be obtained, even when different data or methods are used. The results of the replicated studies are summarized for this work and examples are presented.

The following approach will be proposed as a solution: For the sake of transparency in research, journals should be provided with information on data and data preparation when publishing empirical essays. This is not for the purpose of criticizing authors' errors, but to allow new perspectives on existing research, and to help reduce errors in the analysis of empirical data.