Dynamic Defaults

Participants complete a two-round investment task on Prolific, hosted in Qualtrics. In each round they trade six risky assets over five periods starting from a 1,000-point endowment; points not allocated to an asset stay in cash (the risk-free option). Round 1 is a non-incentivized practice round; only Round 2 earnings determine the bonus.

Three factors are randomized independently at the participant level: defaults, skill elicitation, and leverage choices.

The design tests whether experiencing a good default in the practice round causes participants to misattribute the resulting good outcome to their own skill — rating themselves as more skilled and then, in the incentivized round, investing more and choosing higher leverage.

Three factors are randomized independently at the participant level:

1. Round-1 default (group): In the good-default arm, Round 1 pre-fills the account with 100 units of the asset with the highest up-probability (the "good" asset). In the bad-default arm it pre-fills 100 units of the lowest-up-probability ("worst") asset. In the no-default arm nothing is pre-filled. Round 2 has no default in any arm.

2. Skill elicitation (show_skill): Half of participants rate their own task skill (7-point scale) before Round 1 and after Round 1; the other half do not. A post-Round-2 skill rating is collected from everyone.

3. Leverage tool (leverage_offered): Between the rounds, half of participants choose a leverage factor L from 0.5 to 1.5 (default 1.0) that scales the performance component of their bonus; for the other half L is fixed at 1.0.

The design tests whether experiencing a good default in the practice round causes participants to misattribute the resulting good outcome to their own skill — rating themselves as more skilled and then, in the incentivized round, investing more and choosing higher leverage.
Asset mechanics: each of the six assets has a per-period up-probability equal to exactly one of {0.25, 0.33, 0.45, 0.55, 0.66, 0.75}, assigned to positions at random and re-drawn each round. Each period a price moves up 5% (with its up-probability) or down 4%, over six steps from purchase to liquidation; the break-even up-probability is about 0.44. Participants are told only that an asset's up-probability "may be 25%, 33%, 45%, 55%, 66%, or 75%" — not which asset is which, nor the one-each structure.

In the good/bad-default arms the pre-filled 100 units of "Asset 1" (worth about the full 1,000-point endowment at the period-0 price of 10) is the good (0.75) or worst (0.25) asset respectively; in the no-default arm the good and worst assets fall at random positions.

Bonus = $3.00 + L x (Round-2 percent return / 10), floored at $0, with L = 1 for participants not offered the leverage choice (e.g., at L = 1, a +20% Round-2 return pays $5.00 and 0% pays $3.00).

Channel under test: good default -> good Round-1 outcome -> inflated self-assessed skill -> more aggressive Round-2 behavior (more invested, higher leverage) -> lower realized Round-2 profit.

2026-06-25

2026-07-31

- Round-1 final account balance (first-stage / manipulation check).
- Self-rated skill on a 7-point scale, elicited after Round 1 and after Round 2.
- Round-2 amount invested at the start of the round (1,000 minus cash).
- Chosen leverage factor L (participants offered the leverage tool).

Five pre-specified hypotheses, all two-sided (alpha = 0.05), estimated by OLS with HC3 robust standard errors and corrected as one family by the Holm step-down method. Predicted signs follow the skill-misattribution channel (good-default > no-default > bad-default).

H1 (first stage): round1_final_account ~ arm, full sample.

H2a (skill, primed): skill_post_R1 and skill_post_R2 each regressed on arm controlling for the pre-Round-1 skill baseline (skill_pre_R1); show_skill = 1 subsample.

H2b (skill, un-primed): post-Round-2 skill regressed on arm; show_skill = 0 subsample (this group's only skill elicitation, no baseline).

H3 (investment): Round-2 initial amount invested regressed on arm controlling for Round-1 initial amount invested; good- and bad-default arms only (both face the identical Round-1 default), good-default the reference.

H4 (risk appetite): chosen leverage L regressed on arm; leverage_offered = 1 subsample.

- Round-2 realized return / final account balance.
- Average Round-2 amount invested (mean over Round-2 trading periods).
- Round-2 liquidation (binary: account reached zero before the round ended).
- Stated interest in trading after the task (7-point).
- Moderators: prior trading experience (yes/no), general trading self-assessment (7-point), and willingness to take risks in general and in financial matters (two 0-10 items).

Each secondary outcome is regressed on the treatment arm (OLS, HC3 robust SEs, two-sided, unadjusted p-values; reported as exploratory). Additional pre-specified secondary analyses: (a) within-arm OLS gradients of the leverage choice and Round-2 amount invested on realized Round-1 profit (round1_final_account - 1000), within the good- and bad-default arms; (b) descriptive splits of Round-2 amount invested by Round-1 gain vs. loss and by Round-1-outcome decile; (c) two-stage least squares estimates of the effect of realized Round-1 profit on later skill ratings, Round-2 amount invested, and the leverage choice, using the arm indicators as instruments for Round-1 profit (exclusion restriction: the default affects later outcomes only through realized Round-1 profit; first-stage F reported, with Anderson-Rubin weak-instrument-robust inference when F < 10). The post-task measures are also examined as moderators of the treatment effects.

Online experiment on Prolific (UK and US participants, English-fluent, age 18+), hosted in Qualtrics. Varying defaults in a trading task, and measuring market participation, self-rated skill, and leverage choices.

Each participant completes two five-period trading rounds: Round 1 is a non-incentivized practice round and Round 2 is incentivized. Three factors are randomized independently and orthogonally at the participant level: the Round-1 default condition (good / bad / no default, 1:1:1); whether self-rated skill is elicited before and after Round 1 (show_skill, 1:1); and whether a between-rounds leverage choice is offered (leverage_offered, 1:1). A post-Round-2 skill rating and a short post-task questionnaire (trading experience, interest in trading, risk tolerance) are collected from all participants. The bonus depends only on Round-2 performance.

Estimating equations (no-default the reference arm unless noted): three-arm contrasts Y = b0 + b1*good + b2*bad + e for H1 (round1_final_account, full sample) and H4 (leverage L, leverage_offered = 1). Baseline-adjusted OLS for H2a (skill_post_R1 and skill_post_R2 on arm + skill_pre_R1; show_skill = 1) and H3 (Round-2 initial amount invested on arm + Round-1 initial amount invested; good- and bad-default arms only, good-default the reference). H2b regresses post-Round-2 skill on arm in the show_skill = 0 subsample. All tests are two-sided with HC3 heteroskedasticity-robust SEs; the five primary tests form one family corrected by Holm. "Amount invested" = 1,000 minus cash, reconstructed deterministically from the per-period trade and price arrays.

Exclusions, applied in order: non-consent; incomplete Round 2; duplicate Prolific IDs (keep earliest); Prolific Bot/LLM authenticity-check "Low" flags (with a sensitivity analysis that includes them); and completion-time outliers below the 1st or above the 99th percentile.

Randomization is performed in the survey software (Qualtrics). The Round-1 default condition and the skill-elicitation factor are each assigned by a Qualtrics block randomizer with even presentation; the leverage-offered factor is assigned by a 50/50 draw in JavaScript at the start of the experiment. The three factors are mutually independent.

Individual participant.

No

No clusters; randomization is at the individual level (1,750 individual participants).

1,750 participants, retained after the pre-specified exclusions.

Target ~583 participants per Round-1 default arm (good-default / bad-default / no-default). Orthogonally, ~875 receive the skill elicitation and ~875 do not; ~875 are offered the leverage choice and ~875 are not. The skill (show_skill = 1) and leverage (leverage_offered = 1) subsamples therefore contain ~292 participants per default arm each; the H3 contrast (good + bad arms) uses ~1,167.

At N = 1,750 (two-sided alpha = 0.05, power = 0.80): - H1 (Round-1 final account): about 10 points. - H2a self-rated skill (post-R1 / post-R2, baseline-adjusted): about 0.29 / 0.34 scale points. - H2b un-primed post-R2 skill: about 0.43 scale points. - H3 (Round-2 amount invested): about 50 points. - H4 (chosen leverage L): about 0.08 leverage units.