Time spent on TikTok and endorsement of discrimination against men and attitudes toward sex work

This working paper examines associations between self-reported daily time spent on TikTok and strong endorsement of discriminatory or dehumanizing statements about men (P88–P91), as well as attitudes toward sex work (P70–P73). It distinguishes recognizing prostitution, escorting, and publishing erotic content on platforms such as OnlyFans as “work like any other” from personally encouraging a woman close to the respondent to undertake these activities. The analysis uses unweighted data from a cross-sectional computer-assisted web survey conducted by SW Research for the Dwaplus Foundation among 2,001 heterosexual respondents in Poland aged 18–35, including 982 women and 1,019 men. Binary logistic regressions estimate the odds of selecting the strongest approving response, adjusting for sex, age, education, settlement size, liberal–conservative orientation, and feminist self-identification. Four ordinal specifications examine the full five-category response distributions: proportional odds, partial proportional odds, location–scale, and proportional odds with categorical time spent. Longer self-reported use was associated with higher odds of the strongest approving response in all eight items. Adjusted odds ratios ranged from 1.72 to 2.00 for P70–P73 and from 1.81 to 2.17 for P88–P91 per standard deviation of the ordinal time indicator. All eight associations remained significant after Benjamini–Hochberg correction. Full-distribution models showed that associations differed across response thresholds in P73, P88, P90, and P91. Only P91, concerning a nighttime ban on men’s presence in public spaces, also showed a significantly stronger association at the highest threshold than at the remaining thresholds on average: ratio of odds ratios = 1.31, 95% CI [1.10, 1.56]. Location–scale models provided the lowest AIC and BIC for P90 and P91. These findings indicate variation in the shape of the association rather than a uniformly stronger relationship confined to the most approving category. The study is exploratory, was not preregistered, and has not been peer reviewed. Its cross-sectional, self-report design cannot establish causal effects, temporal ordering, or respondents’ exposure to particular content. The odds ratios concern a standardized ordinal time indicator, not an additional hour of use. The accompanying public-use SPSS (.sav) dataset contains 2,001 respondents and 32 variables. Direct identifiers and unrelated fields were removed, records shuffled, and rare demographic combinations masked. Substantive responses are preserved, allowing reproduction of unadjusted frequencies, distributions, and correlations; masking control variables limits exact replication of adjusted models. Bilingual documentation and a variable dictionary accompany the data.