The next speaker in this session at the 2026 International Communication Association conference in Cape Town is Ernesto de León, whose interest is in who donates digital trace data. Data donation studies have become increasingly popular in recent times, of course, but continue to be plagued by sample bias: some users are simply much more likely to be willing to donate data to researchers.
The focus of criticism has been on the willingness to debate, the lack of data on non-participants, and the limited efforts to scope the downstream effects from biased data donation samples. The present project addressed these by comparing data donation results with revolts from an existing true probability sample in the Netherlands. This enables a full comparison of demographic and attitudinal patterns within this representative sample with the results of data donations. The focus is here on personality, beliefs, and skills.
This, then, enables a move from representational bias to inferential bias: a sample can be compositionally biased but still produce unbiased relationships if effects are stable across subgroups; this can be addressed using a mechanism called difference-in-coefficient tests. Key activity patterns to compare across the two populations here are social media use, search engine use, news readership, and e-commerce behaviours.
The project invited some 4,600 people from the representative sample to donate their data; 500 did so. These were more likely to have specific personality traits: more agreeability, less conscientiousness, less extraversion, more openness; more trust in science, and less left-leaning; more digital skills, more education, higher age.
For online news use, social media, use, and search engine use, attributes in the two samples were often similar, though some showed a direction shift in correlations. So overall, people who donate are different in terms of personality, ideology, and digital skills – but they are not different enough to drastically alter our inferential ability. This observation makes data donation studies a more defensible and robust method than we have assumed in the past.
