Ridership forecast
7/8/2010
- CARRD offered the California High Speed Rail Authority Board concrete illustration of how the current ridership model failed to meet its objective of illuminating the differences between potential routes from San Francisco to the Central Valley. Presentation here
- Cambridge Systematics claimed that frequency didn’t matter because the frequency was so high. This was true for some routes, but not for all. The actual headways had dismal service to Warm Springs, the only East Bay station. Bay Area Headways (.xls)
1/31/2010 CARRD has obtained the actual models used in the ridership forecast from the High Speed Rail Authority. These were produced as per our request. This material as presented did not previously exist.”
**2/10/2010 The California High Speed Rail Authority has notified us that the freqency coefficient in the new business/commuter main mode choice model is off by a decimal. We have requested an updated table and will post this as soon as we have it.**
**3/1/2010 Very geeky spreadsheet that analyzes the unpublished changes to the frequency coefficients**
**3/11/2010 The California High Speed Rail Authority published a memo on 3/4/2010 defending the ridership model. We remain very concerned about the study and important information withheld from public view and find a number of areas of this memo quite misleading. We will publish a memo as soon as possible with our response. We have added new items to the timeline and are still waiting for updated corrected tables**
**4/21/2010 Want to see the headways and travel times used in the ridership forecast? CARRD has obtained this information. We have consolidated a number of files and done some basic comparisons in an excel file – Headways Pacheco and Altamont Peak and Offpeak (Excel file). Please read some important caveats (PDF) and information about this file.**
**4/29/2010 CARRD submitted comments as part of the environmental review process. You can read CARRD’s comments for Program Level EIR (PDF) that relate to the ridership forecasts.**
CARRD has taken an active role in providing an indepth review of the work done to produce ridership forecasts, as well as providing important information about the study to allow all interested parties to do the same. If your group would like a presentation on the issues CARRD identified with the study, please contact info@calhsr.com.
Trip frequency models, destination choice models HSR R&R Model – Coefficients & Constants File1
Access mode, egress mode, main mode choice model HSR R&R Model – Coefficients & Constants File2
These are available for the first time to the public and supercede those in the publicly available documentation.
The Three Page synopsis (“before” main model, “after” main model, transmittal letter)
Background
In conjunction with the MTC, the CHSRA hired a consultant to produce the ridership forecast that is being used to form operational plans, support the viability of private financing for the project and determine capacity requirements down to the level of the number of parking spaces required for Fresno (4,750).
Serious questions have been raised about the methodology and results of this study.
- Sample methods were highly biased towards those who are most likely to take a High Speed Rail train. 96% of the Californians surveyed to assess their interest in taking High Speed Rail for commuting were current train riders. [page 2-3, table 2-1 PDF]
- Documentation for the study is voluminous, yet confusing and missing key information required to assess the validity of the results.
- Drastic changes were made to the model used to produce forecasts, yet these changes were not included in the public documentation, the final project report or information provided to the peer review committee. According to the ridership consultant, this was a conscious decision on the part of MTC.[Transmittal Letter from Ridership Consultant]
- The final model reflects the data quality issues and fails industry standard primary validation checks. Producing forecasts from invalid models is problematic.
- It is concerning that the real model was never made public nor apparently distributed to the peer review panel, particularly as there were significant and obvious issues with model’s validity.
Timeline
January 2000 Charles River Associates releases a ridership and revenue report for use in the CHSRA’s 2000 Business Plan; forecast for intercity ridership in 2020 is 32 million (does not include service to Orange County)
January 2005 MTC hires Cambridge Systematics to conduct a new $1.5 million statewide ridership study to help determine what route High Speed Rail should use to get from the Central Valley to the Bay Area , with subsequent amendments to increase the contract by $69,500 in January 2007 and by $194,528 in May 2007
July 2005 First peer review panel meeting held with 19 people present: a team of 14 Panelists and 5 observers
December 2005 High Speed Rail survey documentation released
July 2006 Second peer review panel meeting held with 24 people present: a team of 9 Panelists, 4 observers and 11 other representatives
August 2006 Final model development report issued and ridership model (“August 2006 model”) detailed; this report is never updated to incorporate later significant changes to the model
August 2006- April 2007 Model undergoes various changes
March 2007 Cambridge Systematics presents ridership results to CHSRA board although intraregional travel in Southern California still to be finalized; forecast is 65 million interregional passengers via Pacheco in 2030, 69 million via Altamont
April 2007 Final changes made to the model (“April 2007 model”); these were unpublished until now
May 2007 Cambridge Systematics presents model with same forecast as March presentation at industry conference
July 2007 Cambridge Systematics presents their final study report to MTC and the CHSRA; the model presented in the August 2006 is repeatedly referenced and discussion of model results is limited to the now obsolete model. In several instances (pages 5-6,5-7), this discussion is contradictory with the April 2007 model used to generate the forecasts in the same report. According to Cambridge Systematics, MTC “elected” to include information from the obsolete August 2006 model and exclude information from the April 2007 model
Decision made to not hold previously planned third peer revenue (page 8-7). “Due to time constraints in the development of the forecasts, this review is just not being conducted.” Certain panel members will just be individually contacted
August 2007 Cambridge Systematics releases final forecast report; forecast is 70 million interregional passengers via Pacheco, 65 million via Altamont
September 2007 Cambridge Systematics releases third peer review report based on email comments from 3 peer review members, including one MTC consultant
October 2007 MTC issues memo responding to criticisms of ridership study. The response to all questions about model coefficients is a reference to the official documentation, none of which includes the final models. There is no hint that any other models other than the ones in official documentation exists.
June 2008 Presentation made by current and former Cambridge Systematics employees at industry conference titled “Communicating Complex Models: A Few of the Lessons Learned from the California Statewide Model for High-Speed Rail”
November 2008 CHSRA releases 2008 Business Plan, based on April 2007 model that fails validation checks; all publicly available documentation continues to reference the August 2006 model
December 2009 CHSRA releases 2009 Business Plan, based on April 2007 model that fails validation checks; all publicly available documentation continues to reference the August 2006 model
February 2010 Journal article describing the California High Speer Rail ridership and revenue forecasting model by Cambridge Systematics current and former employees is scheduled to go to print until CARRD points out that the models in the article are not the ones used to generate the forecasts and that descriptions of the frequency coefficients are inaccurate.
Resources
“Bullet train ridership numbers don’t add up, says watchdog“ San Jose Mercury News, February 6, 2010
Information on software implementation used for ridership study
A Self Instructing Course in Mode Choice Modeling: Multinomial and Nested Logit Models, prepared for U.S. Department of Transportation, Federal Transit Administration, F.S. Koppelman and C. Bhat, 2006. View Manual as PDF (Big file!)
Series of webinars and related documentation, offered by Cambridge Systematics, offered on behalf of U.S. Department of Transportation Full list at bottom of webpage
