Stephen Rees's blog

Thoughts about the relationships between transport and the urban area it serves

Archive for February 25th, 2011

How they got it so wrong

with 11 comments

The title is in response to a post on Pricetags which links to the story of a private sector tunnel operator in Australia who has gone bust because the traffic through their tunnel was less than expected. This is in contrast to the Golden Ears story, where Translink failed to transfer the revenue risk, and is now on the hook for the shortfall in toll revenue there, with consequent impacts on transit service. Which is also one of the reasons why questions are now being asked about the forecast toll revenue on the new Port Mann Bridge.

I spent much of my career involved in forecasting what could happen if various things happened on transport systems. New facilities, and how to pay for them being just one of the variations. And the tool most frequently used is one I have talked about here quite often – the four step transportation model. This takes someone else’s forecast of land use – population and employment distribution – and applies it to observations of recent travel behaviour, and then looks at the capacity of the existing system, and tries to allocate the resulting traffic to the network in a way that seems reasonable.

All models are to a greater or lesser extent partial views of reality. Mostly they use comparative statics: they take a snap shot of the recent past (a travel survey, traffic counts) to get an idea of how the present system is being used – usually just at the morning peak hour on weekdays as that has typically been the peak load on most urban systems, as everyone seems to get up at about the same time, leave home and go to work or school. This is then compared to the future scenario when the populations has grown, and there are more people and hence  more trips to be made. Not surprisingly, given the assumption (not unreasonable in our case) that there will be more trips it is usually the case that greater system capacity appears to be required.

In cities which are functioning well, the traffic pretty much fills up the space available when the system is working properly. Even brief interruptions – usually caused by events like collisions – will cause significant delays in a system that is operating close to capacity. And people will continue to add themselves to the traffic flow because they have less than perfect information. If the congestion is intolerable it takes a while for people to learn that and adapt, but adapt they do. There are many ways that people work out what is best for them, and some constraints on what they can do. The first option is to change the route to one that is longer but less congested. It may be a greater distance but the hope is that it will not take too much longer. Another option is to change the time of day of travel – but while we have seen some peak spreading, there is a practical limit as humans need to sleep and most of them do it at night for around 7 to 8 hours. So the hours between say 10pm and 6am are usually the quietest – except in the entertainment district, of course.

The way that trips are measured is known as “generalized cost” – a combination of time and money. Motorists are notoriously “bad” at estimating marginal trip costs. That is because much of the cost of owning and operating a vehicle occurs in one big lump. Having incurred the cost of buying, insuring and licensing the vehicle, there is a great incentive to use it as depreciation is rapid especially in the first few years and only poorly correlated to use. Motorists are very sensitive to costs that apply to trips like the cost of parking (but not if they get a “free” parking space at work as part of their remuneration) and, to some extent, fuel costs. But as people were saying on last night’s CBC news, when gas prices spike due to events in the Middle East, they have little option but to pay. If there is an expectation that as prices – or parking charges – or tolls – will continue to increase people have to start to look at other routes and other modes.

In Greater Vancouver, we have, on the whole, not done a great job of providing alternatives to car use. Much more has always been spent on increasing road capacity than transit capacity but when transit does get better for some people they do use it – the Canada Line, UPass and the brief period of the winter Olympics show that. Some of the more obvious alternatives – like car pooling – simply have not been very successful here, and the required vehicle occupancy for HOV lanes has been declining to avoid the “empty lane syndrome”. Although if the traffic in the general purpose lanes is not stopped, or very much slower, and the HOV lane clear, then it has very little attraction. Certainly not enough to overcome the “disutility” of having to share your car.

Transport economists have got steadily better at understanding how people see price signals, and the models can be “calibrated” to produce results which reflect our local realities. For instance, the model left to itself simply works by “gravity” – the number of trips between two places being a function of size and distance.  In Greater Vancouver that had to be modified as people do not locate themselves to be near their job – and drive past jobs on the way to the one they have. So the travel survey data on origins and destinations, and the cordon survey of traffic crossing key boundaries and crossings is used to make the distribution of the trips in the model closer to what is observed.

The forecast process is surprisingly simple minded. It just takes present circumstances and says suppose we had a new link or a wider road, how would the traffic respond to that? Then it says: and how would that work in some future date when there are more trips? The algorithms to do that have steadily become more complex as new information about behaviour becomes available. But there are always “known unknowns”. For instance, how much will people have to spend on transportation in future? And how will the price of oil – and the expectation of future oil prices impact that decision?

There is a “time budget” as well as a money budget. There are only so many hours in the day, and North Americans have steadily been increasing the amount of time they travel and work at the expense of sleep, with evident results in poorer health. We have also seen a steady erosion of real disposable household incomes. Wages have not kept pace with inflation – except at the top end of the income scale – and governments have shifted to ever more regressive revenue collection methods. Income taxes have been reduced but all sorts of fees and charges imposed. So far as I am aware, that has not been incorporated into many traffic and transportation models. I think that people must have been getting steadily more sensitive to changes in prices. They are borrowing more to maintain their life style, but are hitting the limit of what they can stretch to. So it really is unsurprising to me that assumptions made in recent years about the willingness to pay a toll for a faster journey have been shown to be overly optimistic. I do not recall ever having a conversation over a model scenario about a populace that in real terms has been getting steadily worse off, in most cases. And with no prospect of that getting better and a very real prospect of increasing energy costs.

The great assumption in all economic models is always “other things being equal” (ceteris paribus) but that is about as silly as assuming that everyone has perfect information (which underlies most of what is taught in Econ101 – and seems to be about all that anyone remembers).

Forecasts cannot be made without assumptions – and mostly modellers try to make those assumptions explicit and given time they will test the sensitivity of the model results to different scenarios. This is all very technical and complicated, and gets done at relatively lowly levels of the hierarchy. The eyes of managers and politicians quickly glaze over if you talk about this stuff to them. It was not uncommon to be given instructions – or in later years, thanks to the FoI Act, a clear understanding without anything traceable – that a certain outcome was required and that the figures in the final report should reflect that. There have been studies that examined transportation project outcomes against earlier forecasts and nearly all have been shown to have been overly optimistic. (Here is a link to one – I know there are more out there.) And there are plenty of well known projects which turned out to be quite wrong about financial viability. Concorde has always been one of my favourite examples.

Much has been made of the impact of greater private sector discipline in this area. Modellers started talking about “investment grade” forecasts – and much stress placed on “due diligence”. But it seems to me that there is a sad record of the private sector producing wildly unreliable forecasts as way of getting funds out of investors. And thanks to the invention of the limited liability “joint stock” corporation often avoiding completely the consequences of over optimism. The over provision of trunk fibre optic cables being one glaring example. We now get cheap long distance, but only because many investors lost their shirts in companies that no longer exist. It seems very odd indeed to me that here we trumpet the supposed benefits of private sector finance, but seem unable to transfer any of the real risk. People who buy shares know – or ought to know – that there are no guarantees. The potential return ought to be greater because it puts a value on the risk that the whole of the investment may be lost. Unlike secured loans or bonds which have some recourse against assets that can be sold in the event of default. What we seem to be doing is paying a higher price for capital and at the same the shouldering much of the risk too.  Great for the corporations, not so good for the rest of us.

A risky investment has to promise possible high returns to get investors. Just talk to anyone – as I did recently – involved in the BC mining industry. Or people offered a “share of the profits” on a show or film: somehow the accountants make sure that there is little or nothing for the angels, no matter what the box office takings. It seems to me to be the height of irresponsibility to introduce this approach to infrastructure and facilities on which we will depend. I think it needs to be recalled that most public enterprises came into being to correct the excesses (and abuses) of the private sector. After the first fine flourish of modern capitalism came a period of increasingly stringent regulation – and, in many sectors, public provision being the preferred delivery method.  The early provision of railroads in Canada for instance which lead to the creation of CN.  Health care is another good example in Canada where, despite the rhetoric, the data is clear that public sector provision is much more efficient and effective than the private sector model in the US. The great shift that started in the 1980s to deregulation and privatization in general ignored the warnings of those who knew their history. And the results were often exactly as predicted and in some cases (like British Rail) horribly worse.

UPDATE March 1

It seems modellers here are not the only ones who failed to predict the decline in car traffic across the Pacific North West as reported by the Sightline Institute

Written by Stephen Rees

February 25, 2011 at 12:13 pm

Posted in Transportation