Harnessing digital trace data to make inferences on public opinion

My main tasks as a Research Assistant at IAS concern the acquisition, preparation, and analysis of textual data from the world wide web. Put differently, I do web-scraping, data wrangling, and text mining. My “weapon of choice” in doing so is R. The overarching goal that I am trying to contribute to is mapping the discourse regarding asylum seekers in Sweden over the past decade. Given that the Swedish right-wing party, the Sweden Democrats, have gained more ground over the past ten years, one might assume that the discourse regarding migration has changed as well.

During an internship at a private institute, I have gained some experience in opinion polling in general, and concerning party choices in particular. However, we were using rather traditional methods such as surveys and telephone interviews. Digital trace data might bear some advantages over those “classic” surveys. One of them is that they are non-reactive, an asset that may eliminate response biases, among which the best-known is probably the so-called social desirability bias. Hence, I aim to leverage these data to draw inferences about things one may not be able to assess using “classic” survey methods.

Before coming to LiU

I acquired a bachelor’s degree in Political Science at the University of Regensburg. My thesis assessed how German politicians communicated on Twitter, revealing, for instance, that retweets mainly happen within party lines, while mentions are more likely to cross these lines.

I also worked there as a teaching assistant and provided students with an introduction to R.