In this paper we assess overall accuracy in survey self-reports on giving to charitable organizations, direction of bias in self-reports, and the influence of this bias on relationships. We compare donations to one specific health charity reported in the Giving in the Netherlands Panel Study 2003 with donations recorded in the database (n = 191). We find that (a) reported donations are significantly higher than recorded donations; (b) reported amounts contributed are correlated very strongly with recorded contributions; (c) differences between amounts reported and amounts recorded are positively related to education, religious affiliation, and the tendency to social desirability, and negatively to household income. This suggests that effects of education are overestimated and effects of income and religious affiliation are underestimated using self-reports on donations rather than archival records. © 2010 Springer Science+Business Media B.V.