I enjoyed Dr. Martin Wachs' brief article
Planning for High Speed Rail (HSR) in California. It centers around planning and forecasting: "The current debate is divisive precisely because improved data and models cannot provide a better glimpse into the future" Wachs says.
After nearly three decades in planning and policy, and six years involvement in local politics I have come to believe that forecasting is unnecessary as a primary decision making tool. Why?
Because transportation infrastructure deployment lags demand by decades. Why do we need to forecast 20+ years into the future? The need is present (or not.) Current data are the best quality data we have. So instead of making numerous uncertain assumptions about a future we do not control, let's assume that the project is built overnight and assess its benefits at the present time.
In the case of Hawaii, when I first relocated here in 1990, traffic congestion was the No. 1 issue and has remained in the top 10 ever since. Twenty three years later, roadway capacity addition has been marginal and most major bottlenecks have not been addressed. So why does the city's planned rail project have a 2030 horizon?
I think the answer is this: The proposed rail does not generate enough ridership with the current population so artificial demand balloons for population and jobs in 2030 are "forecast" to justify the system.
Similarly California has a population of 38 million with huge concentrations in Los Angeles and San Francisco metro areas. If HSR does not work for 38 million, then it won't work. Why is a 20 to 50 year projection needed?
Worse yet, the typical planning models lack credibility because they are never tested with back-casting. If Parsons Brinkerhoff's model is trusted to predict transit ridership (bus and rail) in Honolulu with year 2006 as baseline and 2030 as the horizon, does it also predict transit ridership for 1982? In other words, we should check the model's ability to project forward by using the 24-year period from 1982 to 2006; for which we have perfect information. How does the model do?
I think the answer is this: It does very poorly and for this reason back-casting is never applied.(1)
Obviously, since the inception of a project, several years go by for environmental studies, design, engineering, funding, bidding and construction. So a 10 years-out plan and forecast is needed. But multi-billion dollar energy and transportation infrastructure projects should be justified by their "now" value and not by future demand balloons.
If the projects are beneficial "now" then long range forecasts can be performed for selecting the proper size for them. For example a reversible highway sized at two lanes for sufficient congestion relief may need to be built with three lanes to accommodate future demand. Similarly a 1,000 MW power plant may be engineered with a 1,200 MW capacity for the future.
Projects funded with private capital, fully or partly as in Public-Private Partnerships, sophisticated risk analyses that protect investment from foreseeable risks are conducted. These meticulous and carefully inspected forecasts of project costs and revenues have little in common with the manipulated forecasts for taxpayer funded and subsidized systems, primarily transit systems.
For taxpayer funded projects
forecasting is unnecessary and indeed misleading for decision making. It is commonly used a
tool for deception, particularly for rail projects.(2)
Endnotes
(1) My students and I investigated the accuracy of traffic volume forecasts primarily in the agriculture-to-residential mega developments in Ewa between 1976 and 2002. We found this:
The study compared forecast traffic levels from Traffic Impact Analysis Reports prepared between 1976 and 2002 to actual traffic volumes recorded by the Hawaii State DOT in the city and county of Honolulu. The information extracted uniformly from 11 reports included year of report, consultant, type of project, location, movement, forecast horizon, forecast traffic volumes and forecasting method.
This study focused on road and residential developments and examined the accuracy of traffic demand forecasting, the conservative or optimistic tendency in traffic forecasts and the potential factors affecting accuracy.
The results revealed that traffic forecasts are on average overestimated by 35% and there is a clear tendency to overestimate future traffic volume. Errors ranged from -40% to +200%.
SOURCE: Caroee, Maja, Panos D. Prevedouros and Alyx Yu, Volume
Forecasts for Environmental Impact Statements and Traffic Impact Analysis
Reports: Accuracy of Road and Residential Developments in Honolulu, Paper 13-1398,
92nd Annual Meeting of Transportation Research Board, Washington, D.C., 2013.
(2) Flybjerg, Bent, et al.,
Delusion and Deception in Large Infrastructure Projects (2009)