Nested Logit Discrete Choice Model
Also known as: Tree Logit Model, Hierarchical Logit Model, Generalized Extreme Value Logit, İç İçe Logit Modeli
The Nested Logit model is a discrete choice framework that groups mutually exclusive alternatives into hierarchical nests, allowing correlated unobserved utilities within each nest while maintaining independence across nests. Introduced formally by Ben-Akiva and Lerman (1985) and grounded in McFadden's Generalized Extreme Value (GEV) theory, it extends the standard Multinomial Logit by relaxing the restrictive Independence of Irrelevant Alternatives assumption within predefined groups of similar alternatives.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use Nested Logit when decision-makers face a set of discrete alternatives that can be meaningfully grouped into clusters sharing correlated unobserved factors — making the IIA assumption of standard Multinomial Logit implausible within groups. It is appropriate when sample sizes support hierarchical maximum likelihood estimation and when the nesting structure is theoretically justified rather than data-driven. Limitations include sensitivity to the imposed tree structure, the requirement that lambda_m lie in (0, 1] for consistency with utility maximization, and identifiability issues when nests contain few alternatives. Mixed Logit or Probit models should be considered when the correlation pattern is more complex or unknown.
Strengths & limitations
- Relaxes the IIA assumption within nests, allowing realistic substitution patterns among similar alternatives.
- Grounded in Generalized Extreme Value theory, ensuring a theoretically consistent probability model derived from random utility maximization.
- The inclusive value parameter lambda provides an interpretable measure of within-nest similarity and cross-elasticity.
- Computationally more tractable than Full Information Maximum Likelihood probit models for moderate numbers of alternatives.
- The nesting structure must be pre-specified by the analyst; misspecification of nests leads to biased parameters and incorrect elasticities.
- Requires lambda_m in (0, 1] for global consistency with utility maximization; values outside this range indicate model misspecification.
- IIA still holds across nests, which may be unrealistic when alternatives in different nests share unobserved attributes.
- Estimation complexity grows with the number of nests and levels; deep hierarchies can cause identification problems with small samples.
Frequently asked
How does Nested Logit differ from Multinomial Logit?
Multinomial Logit assumes all alternatives have independent and identically distributed errors (IIA holds globally), implying proportional substitution across all pairs. Nested Logit allows correlated errors within nests, so adding an alternative draws disproportionately from other alternatives in the same nest rather than equally from all alternatives, yielding more realistic cross-elasticity patterns.
What does the inclusive value parameter lambda_m represent?
Lambda_m is the dissimilarity or scale parameter for nest m. It ranges from 0 to 1 under utility maximization: a value near 1 means alternatives in the nest are barely correlated (similar to standard logit), while a value near 0 means they are near-perfect substitutes. It also equals the ratio of the within-nest scale to the upper-level scale.
Can Nested Logit have more than two levels?
Yes. A three-level (or deeper) nested structure is theoretically permissible within the GEV framework. Each additional level introduces its own dissimilarity parameters and inclusive values. However, identification becomes challenging quickly, and deep hierarchies require large samples and strong theoretical justification for each branching level.
Sources
- Ben-Akiva, M., & Lerman, S. R. (1985). Discrete Choice Analysis: Theory and Application to Travel Demand. MIT Press. ISBN: 978-0-262-02217-0
How to cite this page
ScholarGate. (2026, June 2). Nested Logit Discrete Choice Model. ScholarGate. https://scholargate.app/en/econometrics/nested-logit
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Mixed LogitEconometrics↔ compare
- Multinomial LogitEconometrics↔ compare
- Spatial Interaction ModelSpatial analysis↔ compare