Skip to content

JordanD67/swissmetro-mnl-analysis-in-R

Repository files navigation

Swissmetro SA: Multinomial Logit Model for Modal Choice Analysis

Project Context

This repository contains the midterm report analysis for the Urban Transportation Planning and Analysis (UTPA) course at Tokyo University.
The project assesses the potential impact and demand for a major, hypothetical transport innovation: the Swissmetro SA, an underground mag-lev (magnetic levitation) system designed to connect major Swiss cities at speeds up to 500 km/h.


Feature Details

Feature Details
Data Source Stated Choice Survey Data
Survey Period March 1998, Switzerland
Sample Size 1191 respondents
Transport Alternatives Rail (Train), Swissmetro (SM), Car (for car owners only)
Key Variables Cost (CO), Travel Time (TT), Demographics, General Abonnement (GA) pass ownership

Key Concepts & Methodology

Multinomial Logit Model (MNL)

  • The core modeling technique is the Multinomial Logit Model (MNL).
  • MNL is used to analyze discrete choice (selecting one alternative from a set) by estimating parameters ($\beta$) of a utility function explaining why a traveler chooses a specific mode.

General Abonnement (GA) Pass

  • GA pass allows unlimited travel on most Swiss train lines.
  • Its inclusion helps understand how a pre-paid, non-recoverable travel cost influences the choice between public transport (Rail/SM) and private transport (Car).

Elasticities

  • Measure the sensitivity of demand (probability of choosing a mode) to changes in variables like cost or travel time.
  • Direct Elasticity ($\epsilon^D$): Effect of a change in a mode's attribute on its own choice probability.
  • Cross Elasticity ($\epsilon^C$): Effect of a change in one mode's attribute on another mode’s choice probability.

Value of Travel Time Saving (VTTS)

  • Calculated as the ratio of the travel time coefficient to the travel cost coefficient:
    VTTS (Value of Travel Time Saving) = β_TT / β_CO
  • Represents the monetary amount travelers are willing to pay to save one unit of travel time.

Core Analysis Findings

1. Model Goodness-of-Fit and Coefficients

  • The initial MNL model demonstrated a good fit, with aggregate choice probabilities closely matching observed sample ratios (approx. 57% for Swissmetro choosing SM).
  • Including the GA pass variable was necessary; excluding it resulted in counter-intuitive, positive cost coefficients for rail and Swissmetro.

2. Demand Sensitivity (Elasticities)

Metric Result Interpretation
Direct Elasticity for SM Travel Time Approx. -0.55% A 1% increase in SM travel time decreases the probability of choosing SM by 0.55%.
Cross Elasticity (SM TT → Car) Approx. +0.95% A 1% increase in SM travel time increases the probability of choosing Car by 0.95%, showing a substitution effect.

3. Value of Travel Time Saving (VTTS)

  • Rail (Train): Highest VTTS (~14.54)
  • Swissmetro (SM): High VTTS (~13.65)
  • Car: Lowest VTTS (~2.97)

4. Impact of Luggage

  • Including the LUGGAGE variable improved model goodness-of-fit ($\rho^2$ increased).
  • The Car coefficient with luggage had the highest negative value, suggesting luggage strongly reduces Car choice probability compared to Train or Swissmetro.

Conclusion

The analysis confirms a strong potential demand for the Swissmetro, especially among current public transport users.
Advanced econometric models like MNL are essential for accurately forecasting demand by accounting for pre-paid passes, trip-specific needs (e.g., luggage), and other complex factors.

About

This repository contains the analysis and report for a midterm assignment in Urban Transportation Planning and Analysis (UTPA) class at Tokyo University. The project focuses on modeling the modal choice between Rail, Swissmetro (SM), and Car alternatives, based on a Stated Choice (SC) survey conducted in Switzerland in 1998.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors