Introduction and Aims. This study assessed the comparability of estimates of alcohol's harm to others across different administration modes in Swedish general population surveys. Harm was categorised as harm from strangers' drinking and harm from heavy drinkers known to the respondent. Design and Methods. Three surveys were conducted in 2011/2012 (n=6841), including identical questions. One was based on self-administered postal or Web questionnaires, and two were based on computer-assisted telephone interviews of which one included a more ambitious procedure in terms of for example monetary incentives to the respondents. Pearson (2)-tests were used to compare differences in the prevalence of harm. To estimate potential effects of survey mode, the samples were pooled, and multivariate Poisson regression models with mode as explanatory variable were used, adjusting for socio-demographic and behavioural factors. Results. Respondents in the two computer-assisted telephone interviews were more likely to report harm from strangers' drinking compared with respondents in the self-administered postal or Web questionnaires. However, no significant differences were found between survey modes concerning reports of harm from known people's drinking. Discussion and Conclusions. A survey mode based on interviews seems to facilitate reports of harm from strangers' drinking. This does not apply to reports of harm from known people's drinking. Therefore, the comparability of estimates of alcohol's harm to others between survey modes depends on the type of harm being studied.