Verifying reference-dependent utility and loss aversion with Fukushima nuclear-disaster natural experiment

We verify prospect theory with natural experimental data by adopting regression kink design. Our data were collected in 2013 and 2014 from residents displaced by the Fukushima Nuclear Disaster in 2011. We examine how a sudden gain/loss affects stress/utility in four dimensions/resources: family size, health, house size, and income. We find that (i) there is a higher sensitivity to losses from a reference point than to gains (i.e., loss aversion) in health, and possibly in income as well, (ii) the reference point may change over time, and (iii) value function is not separable in the four dimensions/resources. These findings have a few implications. First, in view of the loss aversion, a sufficient—apparently more than enough—compensation should be provided to those who lost so that they can regain the original utility. Second, if the reference point is lowered, the victims must be over-compensated for their loss to recover the original utility. Third, separable value functions should be used with caution.