On 18–19 February 2026 we ran a telephone pretest with a random sample of 101 respondents, split into two groups. The first rated trust in institutions and the likelihood of events on a 1-to-7 scale, the second on a −3-to-+3 scale. Both scales have seven points and identical end labels (in the trust battery, from “do not trust at all” to “trust completely”).
Formally it is the same scale. But to someone hearing the question over the phone it sounds different: in the bipolar format, zero clearly marks a neutral position, and the minus and plus signs mark the direction of the judgement.
Share of neutral and extreme answers in the trust battery
Pretest, 18–19 February 2026, 101 respondents
- 1…7 scale
- −3…+3 scale
Show data as a table
| 1…7 scale | −3…+3 scale | |
|---|---|---|
| Neutral answers (0) | 30.2% | 24.7% |
| Extreme ratings (±3) | 32.7% | 43.5% |
What the experiment showed
- The −3-to-+3 scale reduced the share of neutral answers in the trust battery from 30% to 25% and increased the share of extreme ratings from 33% to 44%. Respondents differentiated more clearly between institutions.
- The institutional trust battery became more internally consistent: Cronbach's alpha rose from 0.73 to 0.81.
- In the expectations battery, by contrast, the bipolar scale weakened consistency (from 0.53 to 0.40). We see this as an advantage: ratings of individual events are less reducible to general optimism or pessimism.
- The rank order of answer categories was largely preserved across formats (rank correlation 0.70 for trust and 0.56 for expectations), and interview length differed by about 16 seconds, which is not statistically significant. The format mostly changes the intensity of answers rather than their structure.
Based on the pretest, we adopted the −3-to-+3 scale for all trust and expectation batteries. The findings rest on a small sample and are preliminary; we continue to test them against the monitoring data.