Guessing the Odds: Subjective Probabilities and Investment Decisions Under PRIIPs

How to cite

Ramos, Sofia; Loban, Lidia; McCourt, Maurice; Sutan, Angela; Yadav, Vijay. 2026. "Guessing the Odds: Subjective Probabilities and Investment Decisions Under PRIIPs." AEA RCT Registry. AEARCTR-0018402. https://doi.org/10.1257/rct.18402-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Sofia Ramos* , Lidia Loban , Maurice McCourt , Angela Sutan , Vijay Yadav

* Primary investigator

Registry ID
AEARCTR-0018402
DOI
https://doi.org/10.1257/rct.18402-1.0
First registered
Last updated
Status
In development
Trial dates
2026-04-20 – 2027-01-31
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/18402

Abstract

This study examines how financial disclosure formats affect belief formation and investment decisions. Participants complete a two-round incentivized investment task in which they observe information about a risky fund and allocate €10,000 between a risky and a safe asset. The treatment varies the format of return information: (i) a scenario-based disclosure consistent with EU Packaged Retail and Insurance-based Investment Product (PRIIP) regulation, presenting labeled outcomes without probabilities; (ii) a past-performance disclosure based on historical returns; and (iii) a combined display including both formats (IS3).

The experimental design combines between-subject variation in the first round with within-subject variation across rounds through randomized disclosure sequences. The empirical analysis focuses on three linked objects: beliefs, their internal structure, and investment choices. In particular, the study tests whether disclosure format affects expected returns, the distribution of subjective probabilities, and the mapping from beliefs to portfolio allocation. The design further allows for an examination of whether treatment effects operate through changes in beliefs and whether they are consistent with salience and ambiguity-based mechanisms.

Intervention

Participants complete a two-round incentivized investment task. In each round, they observe a disclosure describing the performance of a risky investment fund, report their beliefs about future returns, and allocate a €10,000 endowment between a risky fund and a safe asset paying a fixed return of 2%.
The treatment varies the format in which return information is disclosed. Three information settings (IS) are used: (IS1) a PRIIPs-style scenario table presenting four labeled outcomes — stress, unfavourable, moderate, and favourable — without associated probabilities; (IS2) a table of historical realized returns; and (IS3) both formats presented simultaneously.
Participants are randomly assigned to one of five disclosure sequences across the two rounds: Scenario performance →Past performance (IS1 then IS2), Past performance → Scenario performance (IS2 then IS1), Scenario performance → Scenario performance (IS1 repeated), Past performance → Past performance (IS2 repeated), and Combined →Combined (IS3 repeated). Crossover sequences enable within-subject identification of format effects; retest sequences serve as the comparison group to isolate format switching from generic second-round trends.

Experimental design

Randomized controlled experiment with a two-round crossover design. Participants are recruited from the laboratory subject pool at a Business School in France, consisting predominantly of students. Assignment to disclosure sequences is fully randomized at the individual level using the Qualtrics randomizer, independent of participant characteristics. The design supports identification through both between-subject comparisons in Round 1 and within-subject changes across rounds. The experiment may be conducted in multiple waves of data collection. The primary analyses are pre-specified as described, while additional analyses, particularly those exploring heterogeneity, belief structures, and behavioral mechanisms, are considered exploratory.

Design details

Randomization method
Individual-level randomization using the Qualtrics randomizer tool. Assignment to one of five disclosure sequences (Scenario performance →Past performance, Past performance → Scenario performance, Scenario performance → Scenario performance, Past performance → Past performance, Combined→Combined) is fully automated given pre specified percentages and independent of participant characteristics.
Randomization unit
Individual participant.
Clusters
300 participants (same as observations; design is not clustered).
Observations
300 participants 600 participant-round observations (300 participants × 2 rounds).
Arms
Participants are allocated across five sequence arms (Scenario performance →Past performance, Past performance → Scenario performance, Scenario performance → Scenario performance, Past performance → Past performance, Combined→Combined) according to pre-specified proportions, with approximately 35% assigned to S→P, 35% to P→S, 10% to S→S, 10% to P→P, and 10% to C→C, subject to random variation at assignment. This corresponds to roughly 210 participants per main crossover arm and smaller subsamples in the retest conditions.
Primary outcomes
Expected returns, measured as: (1) a stated point estimate of the fund's most likely one-year return, expressed as a percentage; (2) an implied expected return constructed as the probability-weighted mean across eight predefined return bins each spanning a 5-percentage-point range.
Secondary outcomes
Risky allocation (share of €10,000 invested in the risky fund); integration gap (difference between point estimate and implied expected return); belief concentration (Herfindahl index over probability bins); tail risk perception (probability assigned to the bottom two return bins); belief updating (within-participant change in expected returns across rounds); allocation change (within-participant change in risky allocation across rounds).

Cited by

No citation data available for this registration yet.

Keywords