Method, ethics and data rules

This page gathers in one place the method, ethics and data rules that apply to the institute’s own programme and to every commissioned study. The rules rest on published documents: the Measurement Framework (IRISP/2026/005) and the Survey Ethics and Anonymisation Protocol (IRISP/2026/002). Where a study departs from them, the departure and its reason are stated in that study’s method annex.

The page is versioned. Every change is recorded with a date and a version number; earlier versions are available on request.

Each research question measures exactly one of the concepts below, and the report states which. Concepts are not forced onto the same true–false scale; where a comparison is not appropriate, the limit is stated.

1. Perception — a belief shared within a population: the prevailing view of how things are. Measurement: the respondent’s own view and their estimate of what the majority thinks are asked separately; the difference is the perceived-majority gap.
2. Attitude — an evaluative judgement about an object: approval or disapproval. Measurement: rating scale. An attitude cannot be true or false and is not compared with the record.
3. Trust — willingness to rely on an institution, a source or a person. It belongs to the attitude family; it is not confused with factual knowledge about the institution and is reported separately.
4. Experience — an event that happened to the respondent. Measurement: self-report. It can be placed beside the record; it does not replace evidence but is its trace at the level of the person.
5. Factual estimate — a numerical estimate of a verifiable quantity. Measurement: estimation question. The source, date and scope of the comparison indicator are stated in every table; without an indicator the estimate enters no gap calculation.
6. Perception gap — the distance between what is perceived and what is measured or recorded. It is calculated only for factual-estimate and perceived-majority questions; there is no gap for attitude or trust.

The concept definitions and the measurement approach rest on the sources below; citing the institute’s own documents is not considered sufficient on its own. Each reference has been verified individually.

1. Lippmann, W. (1922). Public Opinion. Harcourt, Brace. — the distance between public perception and fact.
2. Ross, L., Greene, D. & House, P. (1977). The “false consensus effect”: An egocentric bias in social perception and attribution processes. Journal of Experimental Social Psychology, 13(3), 279–301. — perceived majority.
3. Noelle-Neumann, E. (1974). The spiral of silence: A theory of public opinion. Journal of Communication, 24(2), 43–51. — the effect of majority perception on behaviour.
4. Prentice, D. A. & Miller, D. T. (1993). Pluralistic ignorance and alcohol use on campus. Journal of Personality and Social Psychology, 64(2), 243–256. — pluralistic ignorance.
5. Zaller, J. R. (1992). The Nature and Origins of Mass Opinion. Cambridge University Press. — the formation of attitudes and opinions.
6. Duffy, B. (2018). The Perils of Perception: Why We’re Wrong About Nearly Everything. Atlantic Books. — factual-estimate gaps.
7. Groves, R. M., Fowler, F. J., Couper, M. P., Lepkowski, J. M., Singer, E. & Tourangeau, R. (2009). Survey Methodology (2nd ed.). Wiley. — sampling, measurement error, weighting.
8. AAPOR. Code of Professional Ethics and Practices; Best Practices for Survey Research. — professional standards; implies neither membership nor certification.

Every measurement is published with a method annex carrying the headings below; a result without an annex is not published. The first completed annex appears with the report of the first measurement round (Q4 2026).

1. Target population and scope (country, age, language)
2. Sampling frame, method and size; groups not reached
3. Mode and dates of data collection; field organisation
4. Full text of the questionnaire and the translation protocol
5. Pilot study and the changes made
6. Weighting variables and their source
7. Quality controls: consistency, duration, re-contact
8. Measure of uncertainty and the assumptions behind it
9. Limitations and cases where comparison is not appropriate
10. If artificial intelligence was used: its task, its limits and the human oversight

Records are de-identified at collection: direct identifiers are separated from the record, and the record is held under a code. The key that links code to identity is kept in a separate environment; access is limited to persons authorised in writing by the data controller, and every access is logged.

The key is destroyed within ninety days of the close of fieldwork. While the key exists the data are not anonymous but pseudonymous, and are named as such. Destruction alone is not considered sufficient: indirect identifiers such as age, locality and occupation are assessed for re-identification risk and coarsened where necessary.

Individual-level records never leave the institute — not to the public, and not to the commissioning institution.

Three processes are kept separate, and all three are explained to the participant when consent is obtained.

Leaving the interview: at any moment, without reason. On request, the answers given up to that point are deleted.

Withdrawing consent: possible until the key is destroyed. The request is matched by the code given to the participant or by the interview date, and the record is deleted. After the key is destroyed the record can no longer be matched to anyone, so withdrawal cannot be applied.

A request to locate and delete a record is subject to the same limit. No promise is made that contradicts what is technically possible; “withdrawable at any time” holds only for the key period.

Every publication’s imprint states its review status with one of three labels: Internal review (scientific board, before publication); External review (a named external reviewer, only where it took place); Unreviewed (working papers, and documents published before the board was appointed). Review that did not take place is not claimed.

Reviewers are chosen by the scientific board. A member with a tie that could affect a review recuses; the recusal is recorded.

When a significant error is found, a dated correction note and a new version number enter the publication’s imprint; the earlier version remains accessible. Because the scientific board has not yet been appointed, all documents published so far count as unreviewed and say so in their imprint.

The institute’s own programme is open access: report, method annex and aggregate data are published together.

In a commissioned study the findings belong to the commissioning institution and are not published without its permission. When the commissioning institution makes a finding public, the method annex and the identity of the commissioning institution are also made accessible; this condition is part of the contract. The same measurement framework and the same ethics protocol apply; a study that asks for a departure from the published protocol, or whose result is fixed in advance, is not undertaken.

Shared with every study: aggregate tables, the questionnaire, the data dictionary and the analysis code. Every shared package carries a version number and terms of use; aggregate tables are released under an open licence (CC BY 4.0).

For microdata, the access level is set after assessing re-identification risk and is stated with its reason: open (coarsened), on application (for researchers, under a use agreement), or not shared (where the risk is high). Personal data are never placed in open access at any level.

No empty data area is published while there are no data; the first package comes with the report of the first measurement round.