Flood damages have increased in many regions around the world, and they are expected to continue to rise in the future due to climate change. To reverse this trend, awareness of flood risk among the population is required to support flood risk management policies and improve flood preparedness. However, empirical studies on the drivers of flood risk perceptions conducted thus far have reported mixed and contradictory results. The aim of this study is to provide insights into the factors that influence perceptions of various dimensions of flood risk to draw lessons to guide flood risk communication strategies. We test a variety of hypotheses of possible factors of influence on flood risk perceptions that are motivated by theoretical concepts and previous empirical studies, whilst also controlling for socio-demographic variables. A representative sample of 2,976 residents answered our survey assessing the role that past flood experiences and risk communication play in shaping flood risk perceptions. Besides exploring flood risk perceptions more robustly, this large sample also facilitates the systematic study of ‘don’t know’ answers, which are often dismissed as missing data in many studies. Rather in this study we analyze what ‘don’t know’ answers reflect in terms of knowledge about particular dimensions of flood risk. The study finds that older people, as well as those who have higher levels of income and education, are significantly more likely to express their flood risk perceptions, respondents who are unable to answer the questions on flood risk perceptions face a lower flood risk, report to have been living in their neighbourhood for a shorter period of time and have less first-hand flood experience. Previous studies might thus be biased by an implicit selection effect. Finally, we show that findings are highly dependent on other explicit choices made by researchers, including the apparently self-fulfilling impact of selecting one explanatory framework over another. New insights emerge from the role that information campaigns and social vulnerability play in the ability to answer the questions. Based on our findings, we offer recommendations for improving flood risk communication policies, specifically increasing the frequency of communication, ensuring that campaigns are focused in terms of the content they provide and the subgroups of the population they target.