Online versus traditional data: Advantages and challenges for the study of anti-immigration attitudes

Abstract

In the context of the digital society, the vast amount of information from online traces can offer a valuable resource for social researchers, and thus even sparked hopes that these data sources could replace traditional ones like surveys. However, such expectations probed to be overly optimistic. Online organically generated data (i.e., produced more or less accidentally through digital interactions) often lacks essential elements for social research such as controlled procedures for extrapolating to population parameters, detailed information on individual characteristics of research subjects, or the operationalization of theoretical constructs (cf., Section 2). The real challenge lies in whether and how organically generated online data can be repurposed to complement traditional sources for research objectives. While this has succeeded in studying disease spread, consumption patterns, or migration flows, it is far more difficult for attitudes in general, and particularly for attitudes towards immigrants and immigration (hereinafter, ATII); which do not constitute observable behavior (at least not by themselves) but abstract thoughts positively or negatively evaluating a phenomenon that involves complex social, political, economic, and cultural considerations. This thesis tackles the ambitious challenge of evaluating the advantages and challenges of organically generated online data sources as opposed to traditional ones in generating knowledge about attitudes towards immigrants and immigration (ATII). To accomplish this general goal, we analyze ATII using four distinct tools, two online -Google and Twitter- and two traditional sources -survey and focus groups-, each pair comprising both qualitative and quantitative empirical information, with the Spanish case as the empirical setting. By leveraging the unique strengths of each tool, we explored their potential to complement one another. This exploratory research design, presented in Section 3, intends to evaluate each data source's analytical performance, on the one hand, and to integrate their empirical findings for a more comprehensive understanding of ATII in Spain, on the other. These assessments are based on the six publications that constitute the main contribution of the thesis. Our empirical results illustrate how each data source excels in different areas. Section 4 discusses how Twitter allows to map real-time reactions to a migration crisis. It revealed clusters of discussion topics and identified potential culprits, helping trace the generic contours of users' opinions during a migration emergency. In Section 5 we detail how focus groups uncover the interaction between anti-establishment and anti-immigration discourses, although with a limited scope. They provided a detailed view of how hostile discourses can be countered through shared grievances and alternative scapegoating rather that direct confrontation. Google Trends, while considerably less precise, provided insights on the evolution of searches about immigrants from 2010 to the present with regional breakdowns, as covered in Section 6. Finally, in Section 7 we detail how the data collected through a survey specifically directed at this topic allowed not only to neatly distinguish two different constructs in this realm, attitudes towards immigrants and towards immigration, but also further sub-dimensions that turned out to be substantively decisive. Furthermore, respondents can be classified based on their attitudinal profile, also enabling the examination of their individual characteristics -though with its well-known limitations in terms of extrapolation and response bias. Ultimately, Twitter and focus groups demonstrate a strong performance with respect to the outlined prospects, whereas Google Trends did worse, and survey data exceeded our expectations. As for substantive results, this thesis obtains relevant substantive findings regarding ATII in Spain. Despite the differences in their concrete objects of study and their temporal and geographical coverage, the results from the four data sources converge in highlighting the nuanced and multifaceted nature of these attitudes. Specifically, the survey data indicate that it is necessary to distinguish at least between cognitive-conative evaluations of immigration, on the one hand, and affective predispositions towards immigrants, on the other, each with specific sub-dimensions. This finding, which contradicts much of the specialized literature, urges researchers to acknowledge the complexity of ATII and redouble their efforts to pursue facet-specific explanations. Among the sub-dimensions, restrictive migration policy preferences -including flow regulation and redistributive policies- were particularly prevalent in Spain. Perceived politicians' responsibility in this regard might contribute to reduce anti-immigrant hostility, but widespread grievances suggest a scenario of calm grounded in a rather precarious balance. According to our results, transparency regarding eligibility criteria for accessing public resources could be particularly useful in mitigating perceptions of conflict of interest, while direct confrontation might not be as effective in countering hostile discourses as more assertive strategies. It is up to future research to determine the extent to which these findings are shaped by the specifics of the case study and the research design employed. These substantive results are mainly -though not exclusively- based on the two traditional data sources considered, as online tools proved to be more advantageous in terms of cost and (temporal and geographical) coverage, while the former provided considerably more detail on the focal construct. To put it succinctly, it could be said that the contrast of traditional versus organically generated online data ultimately involves a trade-off between quality and quantity, respectively. Among other source-specific features, this observation is what lies behind the integrated research design that we propose in Section 9, which suggests combining various tools whose strengths can complement one another to enhance this field of study. In short, the development of such a toolbox with diverse characteristics is not only valuable, but indispensable for uncovering all sides of the prism and understand ATII in its full complexity.