Project Connect’s urban rail forecasting methodology — Inflating ridership with “fudge factor”?
20 June 2014
Graphic: Watts Up With That blog
♦
By Susan Pantell
Recently Project Connect posted a Technical Memorandum dated 13 June 2014 from Alliance Transportation Group discussing what it describes as “Central Corridor Initial LPA Transit Ridership Forecasting Methodology and Summary Ridership Forecasts”. In this posting, researcher Susan Pantell provides a critical analysis of this memo.
Screenshot of page 1 of Alliance Transportation Group’s Technical Memorandum on Project Connect’s ridership forecasting methodology.
This memo does not really provide data on their methodology since the model is secret. Beyond that, their documentation is largely hand-waving.
1. Most importantly, they did analysis only for 2030. FTA now requires current year ridership analysis. “Current year” is the most recent year for which data on the existing system and demographic data are available. An applicant may choose to also evaluate a 10-year or 20-year horizon, and, in that case, the current-year and future-year estimates will each count 50%. Current year ridership would be a lot lower because there is not the development around Highland or the eastern side of UT, but they did not do it.
2. They estimate 15,580 daily trips using the model, which they round up to 16,000. Then they say that on game or event days, ridership could be 20,000 or higher. So they conclude “the project team believes that the median value of 18,000 is a reasonable preliminary estimate of 2030 ridership.” At the end of the memo they explain that this is not based on their calculations, but on their assumption of a 10-15% increase in ridership based on future development (18,000 is a 15% increase).
Lyndon Henry says that is a reasonable assumption, and it may be, but it is not based on data or adequately documented in this memo. They don’t say how many days they predict ridership will be 20,000 or over. There are a lot of events in Austin, but not a lot with high ridership — only 8-9 game days for football and about 10 days for SXSW. If I assume 40 days with 21,000 ridership and 15,600 on the other days, the average comes to 16,300.
They are also accounting for the special event days by adding 25 to the annualization factor of 300 that FTA uses. In addition, they add 103,000 to the annual ridership figure to account for special events.
3. Note that they estimate that total trips for the Capital Metro system will increase by 10,700 in 2030, which is lower than the ridership estimate above because bus ridership will be reduced along the route. Based on that figure, bus ridership will go down by almost 5,000 trips.
4. “Transit fares were set at the equivalent Capital Metro fares for premium transit modes discounted to 2005 model base year dollars.”
Why are they estimating 2030 ridership based on 2005 fares? Because ridership is higher with lower fares. They are assuming $1.50 fare. Using an online calculator, $1.50 is $2.78 in 2030 dollars for a 2.5% inflation rate. (For 2020 it would be $2.02 – $2.34.) That’s assuming they don’t raise the rates beyond the inflation rate.
I calculated the ridership based on a 2030 fare of $2.78 and assuming a 0.4% decrease in transit ridership for every 1% fare increase [TCRP, Report 95, Transit Pricing and Fares, 2004, Chapter 12, p. 12-6. TCRP RRD #61, Traveler Response to Transportation System Changes, 2003, p.19]. I come up with a 2030 ridership of 12,300, as compared with their 15,580. If you add their 15% fudge factor, it comes out to 14,000. If you decrease the base ridership of 12,300 by the same percentage as they do to come up with the total system trips, it comes to 8,500 new trips for the system.
Is that worth $1.4 billion? ■
_______________Austin Rail Now 
Umm, I’m pretty sure you have it exactly backwards.
They use 2005 fares because they’re using the 2005 _model_. Everything in the model (incomes, disposable incomes, price sensitivity, etc.) is calibrated to _2005 dollars_.
..
No such income-related parameters are mentioned in the Technical Memorandum released by Project Connect (per Alliance Transportation Group) and reviewed in Susan Pantell’s analysis. There remain legitimate questions as to not only why 2005 fare levels were used to produce ridership estimates for 2030, but also why a 9-year-old forecasting model was used for travel demand forecasting and public transit ridership forecasting in 2014.
— ARN editor
..
“but also why a 9-year-old forecasting model was used for travel demand forecasting and public transit ridership forecasting in 2014.”
I thought that was pretty clear in the memo, they use a 9 year old model because that’s the most recent model (based on the most recent data, circa 2005 and 2010).
“As part
of the model refinement, CAMPO calibrated the transit component of the model using data
obtained from Capital Metro transit onboard surveys conducted in 2005 and 2010.”
Even if the model was up to date today, why would you plug a fare in 2030 dollars into a current-year model (which would be expecting 2014-dollars)?
Novacek writes
==I thought that was pretty clear in the memo, they use a 9 year old model because that’s the most recent model (based on the most recent data, circa 2005 and 2010).==
We can be pretty sure that modeling technology has advanced in 9 years, so the question remains why this model has not been updated.
==Even if the model was up to date today, why would you plug a fare in 2030 dollars into a current-year model (which would be expecting 2014-dollars)?==
One would expect that a model that uses estimates of population, employment, economic activity, tax revenues, etc. projected to 2030, would also use fare level estimates projected (i.e., inflation-adjusted) to 2030. And how do you know that the CAMPO/Project Connect model “expects” 2014 dollar values?
— ARN editor
“We can be pretty sure that modeling technology has advanced in 9 years, so the question remains why this model has not been updated.”
They _are_ updating it. And it says that in the memo. It’s just not ready yet.
“. Because the CAMPO 2040 demographic forecasts are not final, the
license allows the project team to use non-conforming demographic scenarios pending
approval of the CAMPO forecasts.”
“CAMPO is updating its demographics in preparation for
development of its 2040 metropolitan transportation plan.”
“One would expect that a model that uses estimates of population, employment, economic activity, tax revenues, etc. projected to 2030, would also use fare level estimates projected (i.e., inflation-adjusted) to 2030. ”
No, you wouldn’t. The memo outlines exactly what they changed. They used the 2005 model and the _only_ thing they updated was the demographics (population and employment ).
“Because of its focus on compliance with FTA New Starts guidelines, this version of the CAMPO
model was selected for use in analyzing high capacity transit alternatives in the central corridor.
CAMPO issued a license for use of the selected version of the travel demand model for the
duration of the analysis. Because the CAMPO 2040 demographic forecasts are not final, the
license allows the project team to use non-conforming demographic scenarios pending
approval of the CAMPO forecasts. ”
“The travel demand models used demographic data such as population and employment to
determine the number and location of trips that will be generated in the region during a given
analysis year. Building upon CAMPO demographic estimation methodologies adapted by the
City of Austin in partnership with CAMPO and Travis County staff, the project team developed
forecast demographics for a 2030 study horizon year.
Using the established CAMPO control totals for the city of Austin, the project team allocated
population and employment at the parcel level based upon attractiveness factors and Imagine
Austin density goals for each parcel. Factors considered include local land use plans, existing
zoning, development potential of the parcel, parcel accessibility and the level of development
of surrounding parcels. ”
That’s the _only_ thing they changed. No changes to “economic activity, tax revenues”. Other than that change, it’s the 2005 model. Which means it would be calibrated to 2005 fares. Plugging in 2030 dollars is erroneous.
Novacek, you seem to be making inferences and claims based on a misreading and misinterpretation of the Technical Memorandum, and not based on factual evidence. Alternatively, you may be privy to information about these issues in a way that you haven’t revealed. If the latter is the case, this is information that should be disclosed to the general public. Project Connect and CAMPO seem to have gone to great pains to avoid doing exactly that.
— ARN editor
..
“Novacek, you seem to be making inferences and claims based on a misreading and misinterpretation of the Technical Memorandum, and not based on factual evidence.”
What misreading and misinterpretation? They _explicitly_ state that they used the CAMPO models (which they _explicitly_ state are calibrated to 2005 and 2010 data) with changed population and employment numbers. Nothing in the memo says anything about forward-inflating consumer sentiment to 2030 levels.
“Alternatively, you may be privy to information about these issues in a way that you haven’t revealed.”
Nope, I’m just reading the memo. Oh, and I’m on CAMPOs emailing list, so I know they’re preparing the CAMPO 2040 estimates. Maybe that’s the misunderstanding on your part. They really only do major updates every 5-10 years.That would be why there’s no more recently updated model.
Novacek, you may be “just reading the memo” but you’re misreading and misinterpreting the information in it, making claims apparently drawing on your own suppositions and inferences.
This is curiously similar to the modus operandi of Project Connect and its surrogates, who create smoke and confusion under the pretext of “scientific” or “professional” inquiries, then reach unsubstantiated conclusions on behalf of a preset agenda.
— ARN editor
That’s just it, there is no interpretation or inference. They told you what the model was, and what data they plugged into it, and what they plugged in makes sense because _that’s how math works_. It’s you who are somehow inferring that when they tell you the model is calibrated to 2005, that actually means it’s calibrated to something else.
Your claim basically requires that they’re “trying to put one over on you” and then they _immediately tell you about it_. If they were trying to inflate the numbers, why not omit just that one sentence? Or why use 2005 ticket prices, why not 2000? Or free fares?
Novacek persistently invents fictitious content for Project Connect’s Technical Memorandum, which essentially consists of vague generalities, and reveals nothing about calibration of fare levels to income levels or other economic-related parameters. Readers can check Novacek’s claims for themselves — the link to this memo is included in the article we have posted. We repeat it here:
http://www.projectconnect.com/connect/sites/default/files/reports/20140613_Urban%20Rail%20Ridership%20Analysis%20-%20Summary%20methodology%20%28final%29.pdf
— ARN editor
Who is inventing fictitious content?
It’s you all who decided, on no basis, that the model must be calibrated to 2030 (even though that directly contradicts what they _explicitly_ said in the memo).
Readers are encouraged to review these exchanges of comments and the Technical Memorandum (previously linked), and decide for themselves whether there is any substance or merit to Novacek’s contentions.
— ARN editor
[…] […]