3. RESULTS
Table 1 shows the summary statistics of the different samples used in the analysis. Out of the
1,211 participants in the survey, item non-response leads to 184 missing cases, which
corresponds to 15% of the entire data, leaving 1,027 valid cases. Of these, 264 individuals
report that they cannot donate blood because of medical reasons, so they are excluded from
the analysis. This leaves 763 usable cases, of which 509 are free-riders (i.e. 67% of usable
cases). As can be seen, the distribution of background characteristics across the whole sample
and the smaller samples used for the analyses are similar. With respect to the group of interest
(i.e. free-riders; n=509), mean age is 44 and about half of them are female; regarding the
reasons to free-ride, 22% choose because others already do it, 10% aversion to needles, 20%
have not thought about it, and 48% give no reason (see last column of table 1).
Table 2 reports (in odds ratios) the results of the logit regression that describes individuals’
propensity to free-ride. The model passes the Reset specification test, indicating that there is
no evidence of functional form problems.
Results show that females have a significantly higher propensity to free-ride than men: the
relative risk of free-riding is 45% higher in females than in males (p