Experiment A from "Neglect at Your Own Risk: Evidence on Risk-Taking Prevalence and Motives from the Field"

Participants complete an online survey in which they are introduced to one of four versions of a nested risky-choice problem. All versions present a menu of three options with increasing rewards and decreasing chances of earning the reward. The economic structure is held constant across the GoalQuest and RewardQuest conditions: Option 1 has the lowest reward and highest chance of success, Option 2 has an intermediate reward and chance of success, and Option 3 has the highest reward and lowest chance of success. The menu is calibrated so that Option 3 maximizes expected value.
The four experimental conditions are:
GoalQuest: Participants choose one sales goal from a menu of three goals. Each goal has a reward and a stated chance of being achieved.
RewardQuest: Participants choose one reward from an explicit lottery menu. RewardQuest points are determined by a random draw from 1 to 100, and the participant earns the selected reward if the points meet the selected reward threshold.
Incentivized RewardQuest: Participants make the same RewardQuest choice, but the selected reward may be implemented for real bonus payment. If the participant's coin flip is Heads, the displayed reward is converted to a bonus at 0.5 cents per $1 of displayed reward. If the coin flip is Tails, the participant receives no additional bonus from the RewardQuest choice.
Incentivized RewardQuest with pairwise information: Participants make an incentivized RewardQuest choice with the same bonus procedure and with additional pairwise conditional-probability information that makes the nested lottery structure more salient.
All conditions include instruction screens, comprehension questions, a choice menu, belief elicitation, process measures, demographic items, and feedback opportunities.

2026-07-09

2026-07-16

The primary outcome is conservative choice. Conservative choice equals 1 if the participant selects Option 1 or Option 2 and equals 0 if the participant selects Option 3. Option 3 is the expected-value-maximizing option in the experimental menu. We will report conservative choice shares by condition and estimate differences across the main adjacent condition comparisons.
The main condition comparisons are:
Condition 1 vs. Condition 2: GoalQuest vs. RewardQuest.
Condition 2 vs. Condition 3: hypothetical RewardQuest vs. incentivized RewardQuest.
Condition 3 vs. Condition 4: incentivized RewardQuest vs. incentivized RewardQuest with pairwise information.

1) Choice distribution: The share choosing Option 1, Option 2, and Option 3 in each condition.
2) Beliefs: Participants' reported conditional-probability beliefs about the chances of also reaching the next higher threshold conditional on reaching the lower threshold. The two main belief measures are the perceived chance of reaching/earning Option 2 conditional on reaching/earning Option 1, and the perceived chance of reaching/earning Option 3 conditional on reaching/earning Option 2.
3) Belief accuracy: Differences between reported beliefs and the Bayesian conditional probabilities implied by the menu. The relevant benchmark probabilities are approximately 89 percent for Option 2 conditional on Option 1 and 88 percent for Option 3 conditional on Option 2.
4) Process measures: Responses to process questions and open-ended text about how participants made their choices, especially whether they report making pairwise comparisons between options.
5) Gender differences: Conservative choice differences by gender, overall and by condition.

The study is an individually randomized online experiment. Eligible participants are randomly assigned to one of four conditions with approximately equal probability. Each participant sees only one condition.
Experimental design details
Participants first provide consent and confirm eligibility. They then receive condition-specific instructions and comprehension questions. Participants who answer the first comprehension question correctly proceed to the choice task; participants who miss the first comprehension question receive a second comprehension question before proceeding. The main analysis includes all valid completed responses regardless of comprehension status. Robustness analyses restrict the sample to participants who answer both comprehension questions correctly, and separately to participants who answer at least one comprehension question correctly.
After making a choice, participants report beliefs about conditional probabilities, answer process questions or provide open-ended decision explanations, and complete demographic and feedback items. In the incentivized conditions, bonus eligibility and bonus amounts are determined according to the coin-flip and exchange-rate procedure described to participants.

Not available

Randomization will be implemented by the survey platform / Qualtrics randomizer, with approximately equal allocation across the four conditions. The unit of randomization is the individual participant.

Individual participant

No

The target sample is 1,200 valid completed responses, with 300 participants per condition. The final sample may differ modestly due to platform availability, exclusions for invalid submissions, or budget constraints.

The target sample is 1,200 valid completed responses, with 300 participants per condition. The final sample may differ modestly due to platform availability, exclusions for invalid submissions, or budget constraints.

300 / condition

With approximately 300 participants per condition, the study is powered to detect moderate differences in conservative choice shares. For a baseline conservative-choice rate around 65 to 70 percent, 300 participants per condition provides roughly 77 percent power to detect a 10 percentage-point difference and near-complete power to detect differences of 15 percentage points or larger. The study is not designed to precisely reject very small differences of approximately 5 percentage points.

Experiment A - Analysis Plan