Opinion Pieces: The one thing we know about forecasts is that they are wrong: but by how much?
Opinion Pieces: since 2007, Prof. David Hensher has written an opinion column in the Australasian Bus and Coach magazine, where he monthly discusses a lot of different transport-related hot topics. In this section we are revisiting these columns.
August 2010
I have just returned from New Zealand where I gave three addresses on various aspects of transport reform. Over the total of 5 hours of talks it occurred to me that we have not progressed very far in improving our ability to forecast patronage and project costs of new major transport investments. Don Pickrell in 1992 blew the lid off of the USA Federal Government subsidy program associated with capital intensive projects since the program did not require state and local governments to be accountable for their forecasts and hence preference by States was given to such projects over other projects. The big errors occurred in ridership forecasts and capital cost estimates.
In the mid 1970, Dan McFadden (who received the Nobel prize for economics in 2000) demonstrated that a major reason for forecast errors in ridership of public transport is in the nature of the transport models used to study demand for travel (by mode, destination and frequency). Essentially, the models that contribute to significant errors back in the 1970s, known as four stage models, typically using highly aggregate data (at a traffic zone level – as if the zone travels) such that much of the explanatory power (or variability in travel behaviour response) is assumed away through working with average people and average behaviour. Essentially, the models that contribute to significant errors back in the 1970s, known as four stage models, typically using data describing the travel activity of the average traveller living in a defined physical area (like a postcode or a traffic zone) in terms of the average income, household size, age, and average trip time etc at a traffic zone level as if the physical zone travels; such that much of the explanatory power of the forecasting models in explaining travel demand is assumed away by working with average people and average behaviour.
Disappointingly all the consultants in Australia (with very rare exception) still use essentially the same methods as developed and applied in the 1960’s. Back then, McFadden and his team at Berkeley California showed the limitations of such transport models. So here we are in 2010, observing what has almost become a “law” of errors – if you want to get your forecast closer to reality, then halve public transport patronage forecasts and double capital costs. This also applies to toll roads.
Given we have known this for many years, why is the practice still blind to the evidence? Some have suggested it is because the numbers on ridership look on the low side to get Treasury support and indeed to enable an ultimate healthy benefit-cost ratio when subject to an environmental impact statement assessment. Others have suggested that if we ever want to get public transport built then we must “exaggerate” the evidence since no one really knows the truth anyway. This is known as strategic misrepresentation (which colloquially is known as lying). Well what to do given those in the know are fully aware of this.
Maybe we are safer simply buying more buses because no one apparently asks the question – how many more bums of seats will this deliver? Have you ever seen a benefitcost analysis or a patronage estimate associated with any request or announcement to increase the number of buses by 1 or 100 or 300?
Food for thought
¿Comments? ¿Opinions? ¿Similar News? Send them to us!
Tweet
