https://orcid.org/0000-0002-5613-760X
![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
ABSTRACT
In a time when freedom of movement is being challenged by an increasing number of European Union member states, and where immigration has been dominating public debate for years, this study investigates the effects (i.e. frame salience and frame sentiment) of migration news on individuals’ attitudes about free movement. We are taking into account respondents individual media diet by linking a two-wave online survey in five European countries (n = 7,794) with an automated content analysis of online news coverage in these countries (n = 26,696). Findings indicate that overall the salience of specific frames (i.e. labour market and security), as well as sentiment, positively influence free movement attitudes among citizens. However, there are country-specific differences for both salience and sentiment effects. These findings have implications for our understanding of media effects on immigration attitudes and policy preferences as well as for comparative media effects research in general.
Acknowledgements
We thank Rachel Edie for her helpful comments. This work was supported by Horizon 2020 Framework Programme [grant number 727072].
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 The multilingual dictionaries reached an acceptable level of validity with an average precision of 0.89 and an average recall of 0.83.
2 The F1-score is a harmonic mean between the validation criteria of precision (the share of relevant document of all retrieved documents) and recall (the fraction of relevant documents that was successfully retrieved). See Song et al. (CitationForthcoming) for a discussion of validation procedures in automated dictionary approaches.
3 Sample keywords from the dictionary for the labour frame are, for example, ‘workplace’, ‘wage’, ‘employers’, ‘employement’, ‘hired workers’ etc. Sample keywords from the dictionary for the security frame are, for example, ‘lawbreaker’, ‘police officer’, ‘borders’, ‘public safety’, ‘prosecution’ etc.
4 To increase validity of the results, the number of keywords per article that were required to determine that a frame will be coded as present may vary between 1 and 3 keywords.
5 Some articles may include several frames. Since sentiment is always measured on the level of individual articles, all frames present in the same article will be assigned the same sentiment score.
6 The term stacked analysis refers to the fact that each respondent is included in our models twice. Once with the dependent variable, the lagged dependent variable, frame salience, and frame sentiment in connection to the labour frame and once with these measures in connection to the security frame. In the models, standard errors are then clustered by respondents.
7 Please note that substantial results remain largely the same when excluding all respondents who do not use any of the online news sources from our sample (see Online Appendix Table A1). When looking at the country-specific models however, sentiment becomes only significant in Poland (see Online Appendix Table A2)
8 Please note that substantial results remain largely the same when not including the lagged dependent variable (see Online Appendix Table A3).
9 Correlations in the dependent variable between wave 1 and wave 2 remain relatively high within each country. In Spain, r = 0.48 for labour-specific free movement policy preferences and r = 0.49 for the security-specific free movement policy preferences. In the UK, the correlation is at r = 0.63 and r = 0.60, respectively. In Germany, it is at r = 0.52 and r = 0.53, respectively. In Sweden, it is at r = 0.57 and r = 0.54, respectively. In Poland, it is at r = 0.47 and r = 0.42, respectively.
10 Analyses without the lagged dependent variable show largely similar results. Tables are available in the online appendix (Tables A3 and A4).
11 Please note that our models control for self-selection effects concerning specific news outlets as we include the lagged dependent variable from wave 1.