Plan Revisions in Dynamic Portfolio Choice

All participants complete the same incentivized dynamic portfolio choice task. The study contains two randomized design features.

First, the risky asset's return at each stage is independently drawn to be up or down with equal probability. The resulting return path determines which states of the participant's previously submitted plan become relevant.

Second, participants are randomly assigned with equal probability to one of the two incentive schemes: between-subject random incentive scheme (BRIS) or mixed random incentive scheme (MRIS). Under BRIS, the participant's potential performance payment is based on the terminal wealth reached along the realized return path, using the decisions applicable after any permitted revisions. Under MRIS, one of the three submitted decision snapshots is randomly selected, and the potential performance payment is based on the terminal wealth generated by that snapshot along the realized return path.

The display of previous choices is identical for all participants. At each revision stage, the participant's previous choices for still reachable states are pre-filled as editable defaults. Participants can either keep these choices or actively revise them.

The return path is central to the primary hypothesis. The comparison between BRIS and MRIS is secondary and addresses the persistence and depth of plan revisions.

2026-08-03

2026-08-16

The primary outcome is the within-participant change in the equity share at a reached decision node relative to the participant's most recent prior plan for that same node.

The equity share is the amount allocated to the risky asset divided by the money available at the relevant node. It is bounded between -1 and 2.

At the first revision stage, the outcome is the equity share chosen at the reached node minus the equity share specified for that node in the participant's initial contingency plan.

At the second revision stage, the outcome is the equity share chosen at the reached node minus the equity share most recently specified for that node at the first revision stage.

The study has two co-primary outcomes:
1. The change in the equity share after an up return of the risky asset.
2. The change in the equity share after a down return of the risky asset.

Observations from both revision stages contribute to the confirmatory evaluation of these outcomes. Results will also be reported separately by revision stage to describe how plan revisions develop over the course of the task.

We predict that participants will increase the equity share relative to their most recent prior plan after an up return and after a down return.

The participant's own previously submitted choice for the same state provides the comparison point. The outcome therefore captures an unplanned revision rather than the planned variations in equity shares across states or stages.

For participants with positive exposure to the risky asset, up and down returns generate increases and decreases in wealth, respectively. Positive revisions following these returns resemble house-money and break-even patterns. Reference dependence is one possible explanation for such dynamic inconsistency, particularly if reference point is stochastic and endogenous. However reference points are neither introduced as an experimental feature nor directly observed, and the primary hypotheses do not depend on identifying a particular reference-point mechanism.

A cumulative measure comparing later revisions with the original contingency plan will be reported as a complementary outcome.

The principal secondary outcome is whether the development of plan revisions across stages differs between the BRIS and MRIS incentive schemes.

In particular, we will compare the persistence and magnitude of revisions at first revision stage across the two incentive schemes. We expect positive revisions at first revision stage to be more pronounced at all still reachable nodes under MRIS, whereas revisions under BRIS may be concentrated around immediate reached node after observing the first return. We do not make a directional prediction about the difference between the incentive schemes at second revision stage.

Additional descriptive or exploratory outcomes include:
1. The number and identity of still reachable decision nodes revised at each stage.
2. Measures of risk attitudes, including the implied relative risk aversion and certainty equivalents, recovered from participants' allocation choices where feasible.

Under BRIS, decisions made after a return is observed directly determine the participant's terminal payoff along the realized path. Under MRIS, an earlier or later decision snapshot may be selected for payment. This difference may affect how extensively participants revise their plans and whether revisions persist into all still reachable nodes of the tree.

The payment scheme comparison is secondary. The expectation is that revisions at first revision stage will be made at all still reachable decision nodes under MRIS than under BRIS.

Detailed variable construction and statistical specifications will be provided in a separate Analysis Plan.

Participants complete a computer-based investment task consisting of three investment periods and a terminal payoff stage. Each participant begins with 100 ECUs and allocates money between one risk-free and one risky asset.

The risk-free asset has a gross return of 1. At each period, the risky asset independently produces an up gross return of 2 or a down gross return of 1/2, each with probability 1/2. Wealth compounds across the three periods. Consequently, earlier allocations affect the money available at later periods, and the task represents a single compounding investment rather than a sequence of separately endowed gambles.

Before any return is realized, participants submit a complete contingency plan. This plan specifies an allocation for the initial decision and for every future state that could arise over the three-period binomial return tree.

After the first return is realized, participants observe which node has been reached. Their previous choices for the still reachable states are displayed as editable defaults, and they can either keep or revise them. The same process is repeated after the second return. After the third return, the terminal wealth associated with the applicable choices is calculated.

Participants are independently assigned with equal probability to the BRIS or MRIS incentive scheme. The primary outcomes concern revisions relative to participants' own prior plans following randomized returns. The incentive scheme comparison is secondary.

All participants receive 10 CHF for completing the experiment. Exactly three participants are randomly selected to receive an additional performance payment based on their decisions, their realized return path, and the rules of their assigned incentive scheme.

Not available

Randomization is performed by computer, independently of participant behavior. Two randomizations are carried out within the experimental software (oTree) using its pseudo-random number generator: the sequence of up/down risky-asset returns each participant faces (each draw fair, p = 0.5), and assignment to one of the two payment schemes with equal probability. A third randomization, administered by the laboratory after the session, selects the small subset of participants who are paid their potential payoff for real.

Individual participant, for all three randomizations (return path, payment-scheme assignment, and selection of paid participants). There is no clustering.

No

Not clustered. The individual participant is the unit of randomization and observation. Target 500 participants; a minimum of approximately 300 is anticipated.

Approximately 300–500 individual participants (equal to the number of clusters, as the design is not clustered).

Participants are independently assigned to Basic or WRIS with equal probability. Therefore, approximately one-half of the final sample is expected in each incentive scheme, although the realized arm sizes need not be exactly equal.

At 300 participants, the expected allocation is approximately 150 Basic and 150 WRIS participants. At 500 participants, it is approximately 250 Basic and 250 WRIS participants.

Because each of the first two returns is independently up or down with equal probability, approximately one-half of the participant-stage observations are expected to follow an up return and one-half to follow a down return. The exact numbers will depend on the realized random draws.

The minimum detectable effects are expressed as changes in the equity share, where 0.10 represents a change of 10 percentage points of available wealth. A conservative calculation uses one relevant revision observation per participant and therefore does not credit the additional precision that may be provided by the R_2 observations. It uses standard deviations observed in design-matching, non-preregistered pilot experiments: approximately 0.11 following an up return and 0.27 following a down return. At a total sample of 500 participants, the approximate minimum detectable mean changes are 0.020 following an up return and 0.048 following a down return. At a total sample of 300 participants, the approximate minimum detectable mean changes are 0.025 following an up return and 0.062 following a down return. These calculations assume 80% power and account for the two co-primary hypotheses. They are conservative with respect to the use of observations from both revision stages.