All time references are in CEST
Contextualizing surveys: Incorporating digital trace and media data |
|
| Session Organisers | Mr Manuel Holz (University of Technology Chemnitz) Miss Britta Maskow (University of Technology Chemnitz) Miss Arianna Fay Zehner (University of Technology Dresden) Professor Natalja Menold (University of Technology Dresden) Professor Jochen Mayerl (University of Technology Chemnitz) |
| Time | Tuesday 18 July, 09:00 - 10:30 |
| Room | |
Contextualizing surveys: Incorporating digital trace and media data
Recent developments, methodological challenges and applications
The integration of survey and context data presents a significant opportunity to enhance analytical capabilities and provide more detailed insights into public opinion and social behavior. Traditionally, the contextual data used to complement survey data mostly consisted of time-invariant, country or geography-specific macro data. However, the advent of the digital age, characterized by new data forms like social media data and the digitalization of older formats (e.g., newspaper databases), offers innovative ways to contextualize probability sample surveys within their socio-political and historical contexts, thereby overcoming limitations associated e.g. with time-invariant contextual predictors.
By merging various data formats, researchers can achieve a more nuanced understanding of social phenomena. Nonetheless, this integration poses significant methodological and theoretical challenges. The goal of this session is to facilitate discussion on the practical implications of linking survey and contextual data formats within social science research.
We invite contributions that explore innovative methodological approaches to the collection and analysis of linked survey and contextual data. We particularly welcome submissions that utilize contextual data from traditional media sources (newspapers, television, etc.), digital media sources (social media platforms like Twitter, Instagram, Telegram, etc.), or digital trace data (web browsing data, mobile app usage, GPS location data, etc.). This list is not exhaus-tive, and we encourage contributions from other related fields as well.
Submissions should include studies that successfully combine both types of data to address important social science questions. Relevant topics include developing new tools, applying machine learning and computational methods for both data collection and data generation, integrating new data with traditional sources, harmonizing survey data and discussing statistical modelling techniques to accommodate the multitude of data sources.
Keywords: integration of survey and context data, digital media sources, digital trace data, statistical modelling techniques, machine learning, developing new tools
Ms Kata Számel (HUN-REN Centre for Social Sciences) - Presenting Author
Ms Anna Sára Ligeti (HUN-REN Centre for Social Sciences)
Ms Michelle Horváth (HUN-REN Centre for Social Sciences)
The phenomenon that cultural consumption varies between different social strata and classes has mostly been studied on survey data. The analysis of digital observational data for studying various social phenomena became more and more widespread in the recent years. These data are particularly useful for studying mobility and spatiality. Nevertheless, many social characteristics are difficult to infer from these new types of data. Complex characteristics that indicate social position are among these hard-to-measure characteristics as these are more effectively measured by surveys. In the present study, an investigation is conducted into the spatial characteristics of cultural consumption in relation to social class, using a data source that combines both survey and digital observational data to analyse this phenomenon.
The data has been collected in the framework of a donation-based digital data collection project that is representative for the Hungarian internet user population. 758 individuals have downloaded their entire social media activity and provided it for the researchers with consent for research purposes, while nearly 400 of these individuals also provided data on their mobility and location history, collected by Google since their registration to the platform. The donation was accompanied by a questionnaire that participants were asked to complete. The survey incorporated a series of questions on cultural consumption habits and social class position.
Using this combined data source, we seek to answer the question of how the spatial characteristics of cultural consumption differ between social classes, such as the types of places visited for cultural consumption (e.g. theatres, sports arenas, cafés), the temporal frequency and spatial clustering of visits, the distances travelled and the means of transport used. The study explores whether these patterns vary by social classes or if they are determined by other social-demographic factors, such place of residence or age.
Miss Daria Szafran (University of Mannheim) - Presenting Author
Dr Ruben Bach (University of Mannheim)
Mr Frieder Rodewald (University of Mannheim)
Professor Florian Keusch (University of Mannheim)
Previous findings reveal that Instagram negatively affects its users’ body image. However, existing studies often rely on small and selective samples, which considerably hinders the generalizability of findings and comparisons between subgroups. Instagram use is often measured via self-reports, which are prone to recall and social desirability biases. Experimental studies investigating the impact of Instagram abstinence on body image use app trackers to monitor treatment compliance. However, these trackers do not provide deeper insights into users' detailed Instagram behavior or the content they consume.
To address these limitations, we conduct a two-wave experiment examining how a one-week Instagram abstinence affects respondents’ body image. We recruit 2,000 German Instagram users aged 18 to 65 through a probability-based online panel, randomly assigning them to either treatment (Instagram abstinence) or control group in the first survey wave. In addition to survey questions on body image and Instagram use, respondents are asked to donate their Instagram data in the second survey wave. This approach enables us to objectively measure treatment compliance and explore whether specific subgroups (e.g., defined by gender or age) are more or less affected by Instagram abstinence. Our findings are broadly generalizable by drawing on a large, probability-based sample. At the same time, the combination of survey and donated data offers a more accurate and detailed picture of Instagram use than self-reports alone.
Our study contributes both to the applied literature on Instagram’s impact on body image and the methodological literature on data donation. The enhanced generalizability and nuanced understanding of Instagram usage enable us to provide more robust insights into how body image of different population subgroups is influenced by Instagram use. Ultimately, these insights help inform interventions to mitigate social media’s detrimental effects on body image and well-being.
Ms Arianna Zehner (Dresden University of Technology) - Presenting Author
Professor Natalja Menold (Dresden University of Technology)
Dr Manuel Holz (TU Chemnitz)
Ms Britta Maskow (TU Chemnitz)
The question is addressed as how media discourse is related to the public opinions. We implemented Latent Semantic Analysis (LSA) to categorize German media discourse on foreigners for 2006, 2016, and 2021. Based on semantic meaning and co-occurrence, LSA organized a list of “Ausländer” (“foreigner”) terms into four categories: Administration & Policy, Social Integration, Xenophobia, and Limiting Migration. This approach allows us to see which categories are more prevalent for specific years. Utilizing a set of German General Social Survey (GGSS) variables, we then compared the national opinion towards foreigners to the categorical trends of media discourse for each year.
Our results demonstrate how media text can contextualize trends in public opinion. 2006 was the most “anti-immigrant” survey year and reflects the most “Xenophobia” discourse relative to other years. 2016 demonstrates more “Limiting Migration” discourse comparatively, although “Xenophobia” discourse still comprised the majority for this year. 2021 (the most recent and most liberal surveyed year) demonstrated the least amount of “Xenophobia” discourse and the most “Social Integration” discourse. We discuss: Applications for LSA in media text categorization, the categorical comparison between years for the term “Ausländer,” as well as the comparison between public opinion trends and trends in media discourse. We establish these preliminary results within the theoretical relationship between mass media and public opinion.
Dr Jen Schradie (Sciences Po)
Dr Isabelle Langrock (Sciences Po ) - Presenting Author
While surveys provide important insight into people’s opinions and perspectives on news and information practices, they are severely limited by the nature of self-reported data and in capturing both online and offline flows. Computational tools that track passive web browsing can present a more complete picture of people’s online viewing habits. However, these methods provide little insight into what online sources might mean to participants and how any information might be shared with others. We provide a more complete picture by also adding a third form of data collection: a daily diary study, which asks open-ended questions every day over a short time period. Diaries can provide more granular insight into the sources, content, and discussions that are capturing people’s attention. In this paper we review a study we conducted in the United States and France to demonstrate the power of these three data collection strategies. In particular, we show how together they provide deeply contextual insight into people’s everyday information environments and information seeking behavior, particularly how this information can move from online sources to group chats and offline conversations. Our findings present a more robust depiction of the information environment and the ways that disinformation spreads. We demonstrate how the design of our study – that linked self-reported survey responses with browsing history and open-ended questions – is a much-needed triangulation across three very different forms and types of data that does not rely on only social media posts or survey responses.