Memories of the Stock Market

How to cite

welch, ivo. 2026. "Memories of the Stock Market." AEA RCT Registry. AEARCTR-0018258. https://doi.org/10.1257/rct.18258-1.0

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

ivo welch*

* Primary investigator

Registry ID
AEARCTR-0018258
DOI
https://doi.org/10.1257/rct.18258-1.0
First registered
Last updated
Status
In development
Trial dates
2026-03-31 – 2026-05-31
Countries
United States of America ()
Registry record
http://www.socialscienceregistry.org/trials/18258

Abstract

This study replicates and extends an earlier survey (AEARCTR-0018154) that documented how recalled past S&P 500 returns drive expectations of future returns. The original study found that (a) recalled returns predicted expected returns (standardized beta = 0.36--0.41), (b) showing actual returns produced modest upward revisions (+1.9% at 12m, +1.2% at 5y), (c) self-reported surprise amplified the revision, (d) the BGS norm-surprise mechanism was rejected, (e) order and nudge manipulations had no effect, and (f) market recall dominated personal portfolio returns 2.5:1.

The new survey (v6) introduces two major innovations. First, a deception treatment randomly shows some respondents an artificially low 2025 return (5%) instead of the actual return (18%), followed by a debrief that reveals the truth. This separates information anchoring from genuine memory-error correction. Second, a post-reveal "typical annual return" question tests whether the subjective norm itself shifts in response to revealed information. Additional new measures include self-reported optimism and risk aversion.

The pre-registration below codifies the original findings as directional replication hypotheses (R1--R12), formulates new hypotheses for the v6 features (N1--N7), and specifies estimation methods.

(The earlier data will be used also, but the paper plans to report separately if anything changes between earlier data and the to-be-performed data.)

Intervention

1. Deceptive 2025 Return (between-subjects, 50/50 or similar split):
- Truth condition: Participants see the actual 2025 S&P 500 return of 18% on the information reveal page.
- Deception condition: Participants see a fabricated 2025 return of 5%. The time series chart is truncated to Dec 2024. A debrief page after the post-expectation questions reveals the true 18% return and asks a 4-option reaction question (already_knew / suspected_strong / suspected_weak / deceived).

2. Order Randomization (between-subjects, 50/50):
- Recall-first: Recall sections (B--E) precede pre-information expectations (F).
- Expect-first: Pre-information expectations (F) precede recall sections (B--E).
Same as in the original study.

3. Post-Reveal Typical Return Question:
The same "typical annual return" question (D1) is re-asked after the information reveal, allowing measurement of norm shift.

4. New Demographics:
Self-reported optimism (1--5) and risk aversion (1--5) are added after the political leaning question. Demographics question order is randomized (except the attention check, which is always last).

Experimental design

Prolific

Design details

Randomization method
RNG
Randomization unit
by prolific. in the survey itself, simply random number generators
Observations
300-500 individuals, depending on prolific availability, with incomes of $100k or more
Primary outcomes
rimary Outcomes (end points) A. Replication Outcomes (same as original study) Variable Description Type ExpctPre.1Y (F1) Pre-information 12-month expected return numeric (%) ExpctPre.5Y (F3) Pre-information 5-year expected annual return numeric (%) ExpctPost.1Y (H1) Post-information 12-month expected return numeric (%) ExpctPost.5Y (H3) Post-information 5-year expected annual return numeric (%) ExpRev.1Y (H1 - F1) Belief revision: 12-month numeric (pp) ExpRev.5Y (H3 - F3) Belief revision: 5-year numeric (pp) OthExpctPre.1Y (F2) Perceived consensus 12-month expectation numeric (%) OthExpctPost.1Y (H2) Post-information perceived consensus numeric (%) OthExpRev.1Y (H2 - F2) Revision in perceived consensus numeric (pp) Mkt.Too.High (F4/H4) "Is the market too high?" categorical Regret (F6/H6) Investment regret (pre and post) ordinal Recall.1Y (C1) Recalled 2025 S&P 500 return numeric (%) Recall.5Y (B2) Recalled 2020--2025 annual return numeric (%) Typical.Y (D1) Perceived typical annual return numeric (%) OwnPerf.1Y (E1) Own portfolio return in 2025 numeric (%) MmryErr.1Y (Actual - C1) Memory error numeric (pp) Surprise (G1) Self-reported surprise at actual returns ordinal (1--3) B. New Outcomes (v6) Variable Description Type Typical.Y.Post (D1 post) Post-reveal typical annual return numeric (%) NormShift (D1.post - D1.pre) Change in perceived norm numeric (pp) shown_return_1y Displayed return (5% or 18%) treatment indicator debrief_reaction Whether deception was detected categorical (4 levels) optimist Self-reported optimism ordinal (1--5) risk_averse Self-reported risk aversion ordinal (1--5)
Secondary outcomes
Secondary Outcomes (end points) - Mkt.Too.High switching (F4 != H4) - Regret revision (H6 - F6) - Surprise x Revision interaction - Confidence x Recall interaction - PCA factor structure (replication of 2-factor solution) - Source attribution and recall accuracy - Narrative theme associations - Optimism and risk aversion as predictors of recall and expectations - Demographics order effects

Cited by

No citation data available for this registration yet.

Keywords