Direct answer
A worked example of sentiment surveys shows, step by step, how survey responses can be turned into a sentiment metric using explicit assumptions, and how that metric can be interpreted with clear limits. In this context, “worked example” means you can reproduce the arithmetic and logic from the stated inputs—without relying on hidden rules, live prices, or provider-specific behavior.
Definition and core mechanics
A sentiment survey is a structured questionnaire (or polling method) that asks people to express an opinion. The “sentiment survey” part refers to collecting attitudes (for example, bullish or bearish expectations, or confidence levels). The “worked example” part refers to a transparent scoring method that turns qualitative answers into a numeric value.
A typical pipeline has three mechanics:
- Question design and response scale: responses must map to categories (e.g., “more optimistic”, “neutral”, “more pessimistic”) or to an ordered scale.
- Scoring rule: assign numbers to each category (for example, +1, 0, −1). Define whether you use averages, weighted averages, or proportions.
- Aggregation over time or groups: combine responses within a date window or across segments, then compare results.
To keep this evergreen and independently verifiable, the example below uses only simple arithmetic on assumed counts.
Evidence or example (worked scenario with assumptions)
Assumptions (state everything before calculating)
- You survey N = 60 respondents.
- Each respondent chooses exactly one category:
- Optimistic: O
- Neutral: U
- Pessimistic: P
- You use a 3-point linear scoring rule:
- Optimistic = +1
- Neutral = 0
- Pessimistic = −1
- You compute sentiment score S as the average score per respondent:
- (S = (O\cdot 1 + U\cdot 0 + P\cdot (-1)) / N)
- You assume the categories are mutually exclusive and collectively exhaustive (every respondent picks one).
Scenario inputs
Assume the survey results for a given date window are:
- O = 28
- U = 20
- P = 12
- Check: O + U + P = 28 + 20 + 12 = 60 = N
Calculation
- Total score = 28*(+1) + 20*(0) + 12*(−1) = 28 − 12 = 16
- Average sentiment score (S = 16 / 60 = 0.2667) (approximately 0.27)
Interpretation (what the number can and cannot mean)
- A positive score (here ~0.27) indicates more optimistic than pessimistic responses, under the chosen scoring rule.
- The magnitude depends on the score design: if you changed the scale or weighting, the numeric value would change—even if the raw counts stayed the same.
Second snapshot to compare (still fully assumed)
To show how “worked example” comparisons work, assume a later window produces:
- O₂ = 18, U₂ = 26, P₂ = 16 (still N₂ = 60)
- Total score₂ = 18 − 16 = 2
- (S₂ = 2/60 = 0.0333) (about 0.03)
Change in sentiment under this scoring rule:
- (\Delta S = S₂ − S = 0.0333 − 0.2667 = −0.2334) (about −0.23)
This “drop” means the balance of responses shifted toward neutrality/pessimism in the survey data, not that any future outcome is implied.
Limitations and risks (material failure modes)
- Sampling and size: with a small N, one group of respondents can disproportionately affect the score. Even with N=60, the score can swing if a few answers shift.
- Question wording and scale effects: ambiguous questions, leading phrasing, or mismatched interpretation of categories can change results without any underlying attitude change.
- Aggregation rules can distort: using +1/0/−1 assumes the distance between categories is linear. If respondents view the categories non-linearly, the numeric conversion may not reflect how people actually differentiate answers.
- Time window and non-stationarity: the sentiment measured in one window may not be stable because people’s views can update for many reasons unrelated to any single economic variable.
- Selection bias and incentives: who chooses to respond (or how they are recruited) can correlate with attitudes in ways that the survey design does not capture.
Verification and next question
Because this is a worked example, verification is straightforward: take the assumed counts (O, U, P), apply the scoring rule (+1, 0, −1), and recompute (S) and (\Delta S). If you cannot reproduce the same arithmetic from the stated assumptions, the example’s method is not truly transparent.
A useful next question is: **what scoring rule and question design would still be robust if you changed the response scale or weighted respondents differently?