shows the distribution of investment in the two probability treatments, where, in line with EUT and CPT, the timing
of the resolution of risk is neglected for the moment. Maximal investment (20 euro) turns out to be the mode in each treatment.
However, a strong heterogeneity in investment choices, spread over the whole investment space, can be observed.12
Furthermore, investment is higher under high than low probability (Mann–Whitney–Wilcoxon test: p < 0.01; Kolmogorov–
Smirnov test: p < 0.01).13 Specifically, under high probability mean investment is nearly 5 euro (about 50%) higher and maximal
investment is more frequent (48%) than under low probability (23%). Also, investment below 10 euro is very rare for high probability.
These results are compatible with EUT predictions but are incompatible with CPT, using standard parameter estimates.14
To further investigate the predictive power of EUT concerning investment behavior we will use the individual data from
the risk preference task. We obtain risk aversion estimates from the choices made in this task by following the procedure
employed by earlier studies, using a CRRA utility function.15 The mean value of the estimated risk aversion parameter is
0.539 (standard deviation: 0.397), while the median value is 0.595, which is between the estimates reported by Holt and Laury
(2002) and Harrison et al. (2007). Thus, it seems that the fact that participants had to bring their own money did not lead to a
(substantial) selection bias. Using these estimates we can generate predictions at the individual level. Fig. 3 presents the frequency
distributions of investment predicted by EUT for both probability treatments. In line with actual investment we observe