There you have it – the two most interesting findings from an article about the accuracy of self-reported donations, taken to their extremes. Of course the headline is phrased such that it captured your attention. But if you actually read the article you will find that the headline is a little white lie (from a fairly secular university graduate). The conclusions that can be drawn from the research are much more specific: the accuracy of self-reported donations to a major health research and prevention charity (the Dutch Cancer Society) in the course of the calender year 2003 varies systematically with some of the predictors of actual giving, including the level of education and religious affiliation.
Here’s why this matters. In survey research we usually rely on the information that respondents provide about themselves. If people are inaccurate in their responses about their donations – for instance because they forget about donations they have made or because they exaggerate their giving – a bias is introduced in our predictive models of giving. We may think that getting a university degree makes people more generous, but in fact university graduates are simply more likely to say they donate larger amounts.
Or do they? Yes, university graduates donate more than people with lower levels of education, it turns out. But they also report higher donations than have actually been received by the Cancer Society. Relying on university graduates’ self-reports of donations we overestimate their generosity, we found in a comparison with their actual donations. Donations by respondents with tertiary education are self-reported to be €29 higher than donations by respondents with primary education, though the average difference in donations as recorded in the charity’s database is only €17.
For religion, we found the opposite pattern. As a rule, donations by members of all religious groups distinguished in this study are recorded to be higher than reported. Reformed Protestants for example report donations that are somewhat higher (€8) but do not differ significantly from donations reported by the non-religious. According to the database, however, the average donations recorded by Reformed Protestants are significantly higher (€15) than donations by the non-religious.
Whether the inaccuracy is motivated or due to cognitive problems is difficult to determine. We did rule out that higher educated respondents were more likely to forget about smaller donations as a result of making a larger number of donations over the course of the calender year. We did find support for the case of motivated responding. We found that respondents who agreed with the statement that they “do everything to make others feel more comfortable” have a higher probability of overreporting. But controlling for this difference did not reduce the bias in the effect of tertiary education on overreporting. The search continues…and any cues are appreciated.