# Episode 12: Mode Choice

Introduces utility, generalized cost, logit probability, market segmentation, and policy sensitivity, with a Torrance transit application.

## Sources and attribution

- [Transportation Land Use Modeling and Policy](https://uta.pressbooks.pub/oertransportlanduse/) by Qisheng Pan; Soheil Sharifi; published by Mavs Open Press, University of Texas at Arlington.
- [Torrance Transit Tomorrow Plan](https://transit.torranceca.gov/home/showpublisheddocument/61806/637371705858170000) by Torrance Transit System; published by City of Torrance.
- [Multi-Use Active Transportation Trail](https://www.torranceca.gov/our-city/community-development/planning-division/rfp-active-transit-trail) by City of Torrance; published by City of Torrance.

This episode is an independent study production based on the sources listed above. Source attribution does not imply affiliation, approval, sponsorship, or endorsement by any author, publisher, or institution.

## License

Adapted under CC BY-NC-SA 4.0; © 2024 Qisheng Pan and Soheil Sharifi; excludes third-party items noted in the source; No broad open license identified; public agency facts summarized in original language with attribution; No broad open license identified; public project facts summarized in original language with attribution

## Production notes

Independent educational adaptation. Not legal advice, engineering guidance, an official forecast, or official City of Torrance training. The narration is an original synthesis. Verify current law, guidance, course requirements, and source terms before relying on it.

Stable episode ID: `e383c2f9-41a1-5697-9981-fca5e3a32715`

## Transcript

Mode choice estimates how travelers divide among available ways of making a trip. The source chapter places this step after trip distribution in the traditional four-step model, while also recognizing that mode, destination, time, and other choices can interact. A mode is not truly available merely because it appears in a network file. The traveler must be able to reach it, afford it, use it safely, and complete the trip at the required time.

Analysts describe alternatives through generalized cost. This can combine in-vehicle time, waiting, walking, transfers, fares, tolls, parking, reliability, and other burdens. Different travelers value those attributes differently. Income, age, disability, household responsibilities, vehicle access, trip purpose, and time constraints can all affect choice.

Discrete-choice models represent each alternative with a utility function. A logit model converts relative utility into probabilities. The result is not a statement that every traveler makes an identical calculation. It is a structured estimate of how observed choices relate to measured attributes and unobserved variation.

The choice set is critical. If the model offers transit where service does not operate at the needed hour, the estimated probability is misleading. If walking access, transfers, parking search, or reliability are omitted, the model can favor the mode whose burdens are least completely represented. Market segmentation can improve interpretation, but it requires adequate data and must not turn demographic averages into stereotypes.

Calibration and validation should compare observed and modeled shares by purpose, time, geography, and relevant traveler group. Policy tests should change the actual service or cost variable, then report sensitivity instead of treating the result as guaranteed behavior.

Torrance application. This separate local segment cites the Torrance Transit Tomorrow Plan and the City's Multi-Use Active Transportation Trail project page. Consider a hypothetical scenario connecting neighborhoods, employment, Downtown Torrance, and the Regional Transit Center.

The choice set would use actual route frequency and span, transfer conditions, fares, roadway times, parking, and safe walking, rolling, and bicycling access. A trail connection could change first-and-last-mile time or comfort, while a service change could alter waiting and reliability. Those effects should enter as documented scenario assumptions, not as automatic mode-shift claims.

The analysis should distinguish residents, workers, visitors, people without vehicles, and travelers with disabilities where suitable data allow. It should retain a nondigital way to participate in surveys and should protect privacy. Results could be shown across several values of time or service assumptions to reveal uncertainty.

No ridership estimate, mode-share forecast, or benefit finding is made here. Current local data, calibrated parameters, project details, and agency review would be required. The exercise shows how mode choice connects service quality and complete-trip access to a transparent scenario.

This episode adapts chapter twelve of Transportation Land Use Modeling and Policy by Qisheng Pan and Soheil Sharifi, published by Mavs Open Press under CC BY-NC-SA 4.0. The source chapter includes derivative material attributed in SOURCES.md, including Creative Commons Attribution-ShareAlike material. All source figures, curves, and other third-party items are excluded. Torrance sources are independently summarized. No endorsement is implied.