Description
Estimate which options people actually consider alongside which they choose, so that preference estimates are not distorted by options that were never in the running.
Standard choice models assume every option is considered, so an option that is often overlooked looks unattractive and its attribute sensitivities are misestimated. The independent availability logit treats each option's presence in a person's consideration set as a logistic function of covariates and sums choice probabilities over all possible sets. In a stated-choice study of jeans brands, availability constants were jointly significant (likelihood ratio 32.2 on 6 degrees of freedom), the weakest brand had an inclusion probability of about 0.3, and its price sensitivity doubled once consideration was modelled. With J options there are 2^J - 1 possible sets, which limits the number of options, and the independence assumption is strong.
Attributes
- Discipline
- Statistics & Modelling
- Loss type
- Unmigrated
- Stage
- Strategy, Definition
- Data
- Quantitative
- Effort
- Weeks
Steps
- Run a choice task with all options available and attributes varied by design.
- Fit a multinomial logit as a baseline.
- Fit the independent availability logit, modelling each option's inclusion probability with covariates.
- Test the availability constants jointly with a likelihood-ratio test.
- Compare attribute sensitivities between the two models.
- Report the distribution of set sizes, inclusion probabilities and the most likely consideration sets.
Sources
- Stated Choice Methods (Louviere et al., 2000)
Also known as
- Modelling consideration (choice set) formation
- Choice set generation model
- Independent availability logit
Categories
- Analytics & Behaviour
- Statistics & Modelling