Risk aversion (a 2nd order risk preference) is a time-proven concept in economic models of choice under risk. More recently, the higher order risk preferences of prudence (3rd order) and temperance (4th order) also have been shown to be quite important. While a majority of the population seems to exhibit both risk aversion and such higher-order risk preferences, a significant minority does not. Rather than simply dismissing this behavior, we show how both risk-loving as well as risk-averse behaviors might be generated by a simple type of basic lottery preference for either (1) combining “good” outcomes with “bad” ones, or (2) combining “good with good” and “bad with bad.” We further show that this dichotomy is fairly robust at explaining higher order risk attitudes in the laboratory. In addition to our own experimental evidence, we take a second look at the extant laboratory experiments that measure higher order risk preferences and we find a fair amount of support for this dichotomy. Our own experiment is the first to look beyond 4th order risk preferences and we examine risk attitudes at even higher levels. The consistency of these results with expected utility theory and with a few non-expected utility theories is also examined.
