Description
Compare packed and unpacked versions of likelihood, frequency or value questions, because listing components inflates their total and deflates the residual 'other' category.
Support theory holds that judged probability attaches to descriptions of events, not events. An implicit category ('natural causes') is judged less likely than the same category listed as explicit parts, and separately judged parts sum to more than the whole. Unpacking factors of about 1.4-1.9 were found, rising with the number of components, while complementary pairs still summed to about 1. Incentives do not remove the effect and experts show it too. Rottenstreich and Tversky separated implicit subadditivity (from unpacking and availability) from explicit subadditivity (from judging components one by one). Fox and Tversky found the same in prices: parts priced to about twice the whole. For surveys, detailed option lists inflate listed items and deflate 'other', and 'how often do you have problems?' gets a smaller answer than a list of named problems. Giving examples of the residual reduces the bias.
Attributes
- Discipline
- Survey & Psychometrics
- Loss type
- Latent
- Stage
- Definition, Evaluation
- Data
- Quantitative
- Effort
- Days
Steps
- Audit likelihood, frequency and option-list questions for differences in how finely categories are partitioned.
- Where totals matter, give packed and unpacked versions to random split samples, or ask for the total separately and compare with the summed parts.
- Compute the unpacking factor (sum of parts divided by packed estimate) and test the difference (for example Mann-Whitney).
- Collect complementary judgements as a check that the effect is not simple acquiescence.
- Populate residual 'other' categories with examples and keep partitions consistent across waves and comparisons.
- Report estimates with the partition used, and avoid comparing figures from differently partitioned questions.
Sources
- Heuristics and Biases (Gilovich et al., 2002)
- Preference, Belief, and Similarity (Tversky, 2004)
Also known as
- Partition-dependence check
- Subadditivity check
- Support theory unpacking
- Specificity premium test
Categories
- Analytics & Behaviour
- Survey & Psychometrics