Measuring individual risk attitudes when preferences are imprecise

There is widespread interest in measuring risk attitudes and incorporating such measures into broader econometric analyses. We consider three elicitation procedures currently in use. We find considerable variability within – and even more, between – the results they produce. We suggest that this reflects the way that different instruments interact with imprecise underlying preferences. The short‐run implication is that such procedures need to be used with caution and are likely to be highly context‐specific. The longer run implication is that adding ‘white noise’ to deterministic models is inadequate: we need to develop models that allow for imprecision and procedural variation.