Most quantitative studies in the social sciences suffer from missing data. However, despite the large availability of documents and software to treat such data, it appears that many social scientists do not apply good practices regarding missing data. We analyzed quantitative papers published in 2017 in six top-level social science journals. Item-level missing data was found in at least 69.5% of the papers, but their presence was explicitly reported in only 44.4% of all analyzed papers. Moreover, in the majority of cases, the treatments applied to missing data were incorrect, with many uses of deletion methods that are known to produce biased results and to reduce statistical power. The impact of missing data and of their treatment on results was barely discussed. Results show that social scientists underestimate the impact of missing data on their research and that they should pay more attention to the way such data are treated.