(2002) develop an experimental design to determine the risk attitude of an individual. They use their observations to argue that increased incentives appear to change risk attitudes, leading to greater risk aversion. Popular utility functions that do not allow for such effects are therefore misspecified. Building on this finding, they estimate a flexible utility function that characterizes their aggregate data well, but that does not assume constant (absolute or relative) risk aversion.The basic Holt and Laury (2002) (HL) design should become an important tool for the interpretation and design of experiments in which risk attitudes could play a role. However, their most important result-showing the effect of scaling up the stakes of the lottery choice task-is confounded by a possible order effect. An order effect occurs when prior experience with one task affects behavior in a subsequent task. The primary methodological contribution of experiments in economics is to enhance control. Ideally, such control makes the explanatory variables of interest, in this case scale, orthogonal to other explanatory variables such as order, allowing clearer inferences about behavior than one could get from field econometric data. We argue that not controlling for order effects results in a misspecification of utility functions that is as important as that of scale.The subjects in the HL design were given sequences of three or four tasks, as shown in Table 1. Task #1 involved choices over lotteries with a baseline level of prizes, which we will refer to as the "1ϫ scale." Task #2 involved hypothetical choices over prizes with a scaled-up level of prizes, either 20ϫ, 50ϫ, or 90ϫ. Task #3 repeated task #2, but with choices that involved real payoffs. Task #4 was a return to the baseline task with real 1ϫ prizes. In some sessions, subjects were not given task #2 or task #3. 1What could one infer from the comparison of measured risk attitudes in the 1ϫ and 20ϫ tasks in HL? Unfortunately, any observed difference could be due either to the scale of the prizes involved or the task order, or some combination of both. Thus, the effect of scale is intrinsically confounded with the possible effect of task order.2 This is a logical flaw in their design, which 1 HL recognize the possibility that wealth effects could also confound the effects of scale in an in-sample design. To handle this they use a clever device: when the subjects proceed from task #1 to task #3, they are asked if they are willing to give up their earnings in task #1 in order to play task #3. Since the stakes are so much higher in task #3, all subjects chose to do so. This means that the subjects face tasks #1 and #3 with no prior earnings from these experiments, although they do have experience with the type of task when facing task #3. No such trick can be applied for task #4, since the subjects would be unlikely to give up their earnings in task #3 in this instance. Thus the responses to task #4 have no controls for income built into the design.2 HL are aware (p. 1647,...

It is widely accepted that individuals tend to underinsure against low-probability, high-loss events relative to high-probability, low-loss events. This conventional wisdom is based largely on field studies, as there is very little experimental evidence. We reexamine this issue with an experiment that accounts for possible confounds in prior insurance experiments. Our results are counter to the prior experimental evidence, as we observe subjects buying more insurance for lower-probability events than for higher-probability events, given a constant expected loss and load factor. Insofar as underinsurance for catastrophic risk is observed in the field, our results suggest that this can be attributed to factors other than only the relative probability of the loss events.

No abstract

We propose and test a new method for eliciting curvature-controlled discount rates that are invariant to the form of the utility function. Our method uses a single elicitation task and has the advantage of obtaining individual discount rates without knowledge of risk attitude or parametric assumptions about the form of the utility function. We compare our method to the Andersen et al. (2008) double elicitation technique in which the utility function and discount rate are jointly estimated. We use a laboratory experiment to perform a within-subjects comparison of discount rates from these two methods and find consistent results, which is reassuring given the wide range of estimates in the literature. In addition, the estimated discount rates in our study are "plausibly low" in contrast to the vast majority of discount rate studies. Our results are robust to relaxing the expected utility assumption of linearity in the probabilities, as we find little evidence of probability weighting in our data. In a second experiment, we find that discount rates are not sensitive to the length of the horizon, but are sensitive to the length of the front-end delay, suggesting present bias. We estimate average discount rates to be 12.2 percent in the first experiment and 11.3 percent in the second experiment when the front-end delay is at least two weeks.

This study examines the influence of malpractice claims on the practice behavior of a panel of obstetricians in Florida during the period 1992–1995 to determine whether physicians respond to malpractice events by performing more cesareans, consistent with the notion that cesarean sections are employed as “defensive medicine.” Findings indicate that clinical events resulting in claims that lead to substantial indemnity payments have a significant, modest effect on physician practice behavior: physicians experiencing those claims increase their risk-adjusted cesarean rates by about one percentage point. Malpractice experience does not appear to affect patient mix, but claims with large payouts may affect patient volume.