Framing, Social Norms, and Household Investment: A Randomized Vignette Experiment in a Japanese Panel

All respondents read a short hypothetical scenario about receiving unexpected extra income of 300,000 yen and deciding between keeping it in a bank account and investing part or all of it in a globally diversified investment fund. A randomly selected quarter of respondents reads, in addition, one truthful sentence about the long-run performance of such funds; a second quarter reads one truthful sentence about the possibility of large gains and losses of such funds; a third quarter reads one truthful sentence about how Japanese households allocate their financial assets; the remaining quarter reads no additional sentence. All statements are factually accurate and sourced. The exact wording is documented in the hidden fields and the attached analysis plan, embargoed until trial completion.

2026-08-29

2026-11-30

Amount allocated to the investment fund (numeric, 0-300,000 yen; elicited in units of 10,000 yen).

The allocation is a hypothetical choice, deliberately simple (one amount, no percentages or probabilities), asked identically in all four arms immediately after the scenario. The confirmatory contrasts are each information arm versus the no-information arm on this amount. The two directional contrasts (returns-information arm versus no statement; norm-information arm versus no statement) are separate confirmatory hypotheses about distinct mechanisms, each adjudicated by its own two-sided test at the 5 percent level; there is no composite success criterion, and the third contrast carries no directional prediction. An indicator for allocating any positive amount (the extensive margin) is a secondary outcome.

(1) Indicator for allocating any positive amount (extensive margin).
(2) Item nonresponse on the allocation question.

The extensive margin separates "whether to invest at all" from "how much"; in a population where deposit-holding is widespread and many respondents hold no risky assets, the two margins can respond differently to information. Item nonresponse (the panel's questions are not mandatory) is analyzed as an outcome to detect differential drop-off by arm. Pre-treatment portfolio holdings and wealth, measured by survey questions asked before the scenario in the same wave and in earlier waves, serve as covariates and pre-specified moderators, not outcomes. Effects on actual portfolio holdings measured in later panel waves are exploratory.

The experiment is embedded in wave 15 of the NIRA-Okubo Japanese national online panel (NIRA is the Nippon Institute for Research Advancement; fielded in the second half of 2026 by a commercial survey company; roughly 9,000 expected completes). Continuing panelists are invited first and are not screened on employment status, and newly recruited respondents, who fill the remaining cells of the survey company's stratified design, are in employment at recruitment. The analysis sample is restricted to respondents aged 18 and over. All respondents answer the panel's standard household-portfolio questions (types of financial assets held; total financial assets), then read the hypothetical windfall scenario described in the intervention field. Respondents are randomly assigned (individual level, one quarter each) to one of three additional one-sentence statements or to no additional statement, and then state the amount of the windfall they would put into the investment fund. The windfall and the allocation are hypothetical and no money changes hands. Randomization is independent of a separate information experiment fielded elsewhere in the same questionnaire and registered separately. The assignment indicator (which of the four versions was shown) is stored permanently as a respondent-level variable, including for the group shown no additional sentence.

Not available

Randomization is carried out by the survey company's platform at the individual respondent level, at the moment a respondent begins the questionnaire. Each respondent receives an independent random draw over the eight combinations of the two experiments fielded in this wave (four conditions of this experiment, crossed with two information conditions of a separate experiment registered separately), each combination with probability one eighth. Assignment is therefore one quarter per arm for this experiment and independent of the other experiment by construction. The draw is made when the respondent enters the main survey, and a number is drawn for every respondent who enters, including those who later break off; the pattern does not change if a respondent pauses and resumes. The assignment indicator is stored permanently as a respondent-level variable. The assignment involves no blocking, no stratification and no quota balancing.

Individual (respondent)

No

Approximately 9,000 (clusters are individuals; individual-level randomization).

Approximately 9,000 wave-15 completes (about 2,250 per arm). Respondents under 18, whom the panel admits and who make up about 0.2 percent of a wave, are excluded from the analysis sample.

4 arms: approximately 2,250 each (long-run returns / price risk / deposit norm / no statement).
The wave also carries a separate two-arm information experiment, registered separately, and the survey platform assigns respondents over the eight combinations of the two (4 x 2). Assignment to the four arms of this trial is therefore approximately one quarter each, and approximately one eighth of respondents falls in each of the eight combinations.

Wave 15 refreshes the panel with new respondents. Wave 14 completed 8,215 interviews, the smallest wave to date, and on the three previous occasions when a wave without a refresh was followed by a refreshed wave the sample grew by 9, 10 and 14 percent; power is therefore planned at 9,000 completes, about 2,250 per arm, with 8,000 as the pessimistic case. The survey organization's stated target is about 10,000, which would improve each figure below by roughly 5 percent. Using MDE = 2.8 x SD x sqrt(2/n per arm) at 80 percent power and a 5 percent two-sided test, with n = 2,250 per arm: any pairwise arm contrast on the allocation amount has an MDE of approximately 8.3 percent of a standard deviation of the outcome (the outcome is newly fielded, so no historical standard deviation exists; effects are expressed in standard-deviation units). On the extensive margin, assuming a baseline share of 35 percent allocating a positive amount, the pairwise MDE is approximately 4.0 percentage points. For interactions of a pairwise contrast with a standardized moderator, the MDE is approximately 8.3 percent of a standard deviation per one standard deviation of the moderator.