We investigate how people allocate a limited set of resources between multiple risky prospects. We found that only a small percentage of decisions followed some form of naïve diversification or mean-variance optimization. In general, people were less mean-variance optimal than a naïve 1/N heuristic. Aspects of choice sets, such as domain, skew, and second order stochastic dominance, affected resource allocation decisions in a similar manner to their influence on single choice gambles. Individual traits traditionally linked to risk propensity seem to manifest in terms of the degree to which people are inclined to diversify. Lower risk aversion is linked to increasing diversification. Risk congruency, the degree to which peoples' self-reported and elicited risk aversion matches, moderates how susceptible people are to cost framing nudges. We find evidence for heterogeneous clusters where people either under-weight or over-weight segregated costs, leading to the same nudge producing opposite behavioral results.