Average value of time saved irrelevant to toll express lanes - Randall Pozdena ECONorthwest
By Peter Samuel
2013-08-30: It's not the average value of time saved (VoTS) that's relevant to users of toll express lanes running alongside free lanes. It's the higher time saved values of a small proportion of trips that choose the express lanes that determines how much the minority using express lanes will pay - the VOTS of the top decile or two of the distribution of total trips in the corridor.
Randall Pozdena, managing director and senior economist at ECONorthwest in Portland OR makes this point in comment on our reporting of a University of Minnesota study of motorist behavior on the 394 and 35W toll express lanes in the Minneapolis area by Michael Janson and David Levinson. We'd reported the researchers' surprise at the very high values of time saved they'd observed in the Minneapolis toll lanes - that drivers in express lanes regularly pay $1 to $2 per minute saved, or $60 to $120/hour.
That compares to full tollroad average driver payments of $15 to $30/hour.
The main thrust of our report was on the Janson-levinson discovery that toll rates can have a definite proxy effect - high rates being seen as a proxy indication of high congestion shifting the demand curve to the right.
But the other newsworthy findings of the study seemed to be the very high VOTS measured and the Janson-Levinson explanation that motorists may have an exaggerated sense of the time they save. They are behaving with imperfect knowledge, and systematically biased perception of what time saving they're buying.
Pozdena sees it quite differently.
He says people using toll express lanes are acting rationally and while their knowledge is imperfect he sees no systematic misperception.
"In a toll express lane setting, implicit values of time are very likely to be that high ($1 to $2/minute saved.) The reason is that the 'conventional' value of time is actually an average, but in the real world, there is a log-normal distribution around that average that reflects not only the distribution of income, but also the urgency of the trip at the instant. Thus, the users who self-select to pay to join the toll express lane are drawn selectively from the UPPER TAIL of the VoT distribution, and would not be expected to display average values of time."
He says ECONorthwest has developed a model for toll optimization model that is based on calculating the shape of the distribution of values of time.
"We take the regional model average VoT and form a distribution around that mean in a key step in the data flow called 'VoT Transform.'
He adds that at ECONorthwest they "routinely find our model predicting that the MARGINAL value of time of those who are skimmed off to the HOT lane is several multiples of the AVERAGE value of time, and the average of those above this cutoff would be even higher. Thus, I don't think that one can conclude that HOT lane users think they are getting 'greater time savings' than they really are. They are just the folks with long-tail values of time."
Pozdena says the proxy effect is not a problem if the pricing system is working as it should and prices according to density of vehicles. It is conveying good information to motorists about conditions they can expect.
Our report of Michael Janson and David Levinson study: http://www.tollroadsnews.com/node/6681
2p screed on ECONorthwest model:
Robert Bain, RBconsult Ltd comments:
Although I agree with both Janson/Levinson and Pozdena in the sense that:
(a) research I've conducted demonstrated that perceived toll road time savings were systematically higher than actual ones and
(b) the values-of-time-savings would be expected to be distributed as income is (ie. log-normal) and video surveys of toll road use tend to show an over-representation of new, high-end autos (high-income users)
...I do have to wonder if the high values of time that people are responding to are simply a result of model misspecification?
If you constrain the observed response (ie number of drivers using toll facilities) to be a function of cost and time saving alone, very high values of time will result. Perhaps if other explanatory variables were considered (journey time reliability, the urgency of a trip on a certain day, trip frequency etc.) we would find express-lane values of time closer to industry norms? Added model variables would effectively take the pressure off the need for value of time to explain everything.
This is off course inconvenient - but is nevertheless important.
It is inconvenient in the sense that, unlike time savings and costs, other explanatory variables are far more challenging to identify and quantify. And, in predictive mode, you have to forecast them! However it is important in the sense that simple explanations of toll road usage in terms of time savings and costs often fail to reflect observed behavior - and increasingly serve the industry poorly. This is likely to become more problematic as pricing becomes more sophisticated.
Investor Support Services