Integration Research 2.0—Harnessing the Power of New Data Sources to Advance Knowledge on Behaviour and Attitudes of Migrants and Natives

To overcome the limitations of traditional survey data, social scientists have recently turned to new forms of data and novel approaches to data collection that promise faster, more frequent, and potentially more accurate information for social science research in general and studies on immigration and integration in particular. This project, supported by the Fritz Thyssen Foundation, examined three examples of new data sources: (1) passively collected data from smartphone sensors and apps, (2) aggregated internet search queries, and (3) responses obtained from voting advice applications (VAAs).

For each of the three approaches, we conducted systematic reviews of the existing literature, discussed their benefits and limitations, identified issues that need further clarification, and provided best-practice examples and user guidelines for future research in these areas. We identified 41 studies using smartphones for data collection in the context of migration and integration research. The review revealed concerns about passive data collection regarding privacy and data security as the main obstacle, leading to low participation rates and missing data. For Google Trends, we coded 360 social science studies published between 2010 and 2021. The results show that the large majority fails to assess the internal validity of their Google Trends measure, does not consider whether the data are reliable across samples, and does not discuss the (lack of) generalisability. Similarly, we examined 175 VAA studies and identified several open fields for future research, e.g., in the area of political information of immigrants. We also found important methodological issues limiting the external validity of the existing findings due to the reliance on volunteer samples in VAA studies.

Additionally, we worked on two of the open issues identified in the systematic reviews. We developed a systematic approach to selecting and validating keywords from Google Trends to measure xenophobic attitudes. The results for Germany show that when initially selecting a long list of potential keywords, only very few terms pass the steps of internal validation. Thus, our findings make us skeptical about the measurement of attitudes in general and anti-immigrant attitudes in particular with Google Trends data. We also conducted an experiment to measure the effect of providing feedback in a VAA on participants’ answers compared to a traditional survey. The outcome suggests that the feedback mechanism does not change the answers themselves but leaves open the possibility that it changes the composition of the sample.