MerQuR+Statistics Glossary

Statistics Glossary → Alternative-specific constants (ASC)

Alternative-specific constants (ASC)concept

Constant terms in choice models such as conditional logit, defined relative to a base option, that capture each option's average appeal not explained by the measured attributes.

In choice models, an option's utility is tied to its measured attributes (such as price and travel time). But people may favour an option more or less on average for unmeasured reasons (brand image, comfort, habit). Alternative-specific constants capture this difference: with J options, the constant of a base option is fixed at zero and one constant is estimated for each of the other J − 1 options. A positive ASC means that, other attributes being equal, the option is more attractive than the base option.

With the constants included, the model reproduces the choice shares in the sample exactly. Person-specific characteristics that do not vary across options (age, income) cannot be estimated directly in conditional logit; they can enter only through interactions with the ASCs or option dummies. The constants are interpretable when the option labels are meaningful (for example, modes of transport); in unlabelled experimental designs they are usually not meaningful and pick up position effects.

An everyday example

In a city, people choose between car, bus and metro, with fare and travel time in the model. Even at the same fare and time, the metro's ASC is +0.6 relative to the bus: people prefer the metro, probably for unmeasured reasons such as punctuality and being unaffected by traffic.

In MerQur

In Conditional Logit, 'Additional output: Alternative-specific constants (ASC)' adds the alternative-identity dummies to the model; the base alternative can be chosen, and a warning is given when there is no alternative identifier.

Analyses where this concept appears

Conditional Logit (Choice Model)

Related concepts

Reference category (comparison level)