Dear UTSG listmember,
Phil Goodwin did not argue in his recent paper that the value of benefits depend on an underestimated value of time per hour. However, in an earlier email I did, which may have given the impression that Phil argued the same. In point of fact I only cited one conclusion in one part of his paper which I thought was consistent with Vierordt's law.
Hence I would like to clarify that when I said that Phil Goodwins conclusion "is consistent with Vierordt’s law" what I meant to say was: "is not inconsistent with the position I have derived from Vierordt’s law…", which depends, of course, on lots of other factors (see below) - but does not depend on the arguments being the same (which they aren't).
Several people have asked for Phil Goodwin’s paper in full, but it is not available except in very parge pdf format which doesn't attach very well to emails, so without wishing to misquote the author, his reply – which summarises the paper as well – is reproduced below (with permission).
In fairness to Tony Fowkes, there is literature which says that temporal processing leads to systematic underestimation (see Casini 1999 p85), which would mean that the sign of the 1% overestimation in Fowkes 1999 would be correct – a point also made by Phil Goodwin in the below – but then the same paper ALSO says:
“When the NON-TEMPORAL task involves more information or is more difficult [e.g. such as the judgement of the VALUE of delay], not only does the level of performance decrease dramatically in the temporal task but the perceived duration becomes shorter”
i.e. Casini et al (1999) indicate that there are links between temporal and non-temporal processing, which might extend to become a link between value of time (which is, apparently, a NON-temporal task) and duration perception (which obviously is a temporal task).
Casini. L. And Macar. F. (1999) 'Multiple approaches to investigate the existence of an internal clock using attentional resources' Behavioural Processes 45 p73-85.
So I think the magnitude and sign are both still open to question espescially for behavioural values of time, even if they do remain controversial for resource values of time.
Michael Nandris
On the duration / value debate, Phil Goodwin said:
Feel free to use this - or even reproduce entirely - if you want.
a) My ECMT paper, based on work that I had done for TfL before the start of congestion charging, did come to the conclusion that the economic benefits had been underestimated in various published studies by TfL and Government. This was for two main reasons (a) the elasticities had been wrongly revised downwards during the studies, and the traffic effect would be greater than estimated (this then has actually happened); (b) the second-round benefits of use of the revenue had not been included (of course, because of (a), this was less than it would have been if it had been included). But that paper took the official values of time used for granted, and did not challenge them at all.
So my conclusions about benefits really don't have any bearing on value of time estimation.
b) It does seem to me that Vierordt's law, which I had not heard of before, is sensible, and it is clearly very relevent to transport assessment where many of the millions of pounds worth of estimated time benefits are made up of very small time savings indeed: seconds, or fractions of seconds, not the 5 minutes versus 30 minutes that have often been referred to in the 'small time savings' debate over the years. BUT there is an extra complication, since most of the time values have been calculated not from real changes/differences in travel time, but on (i) times reported by travellers (in which the perception errors will already have been included before estimating a value) or (ii) will be responding to hypothetical time differences specified to them by an SP interviewer, in which case it is not at all clear at what stage the perception bias would enter the calulations. It seems to me that this could lead to an OVER-estimated value of time when applied to the 'real' (ie modelled)
time differences purported to arise from the scheme.
c) In any case, it is now clear that for very many schemes, the calculated time savings are an order of magnitude smaller than the day-to-day variation experienced by travellers, and I suspect that in these cases the issue is not that the travellers are overestimating the duration of small intervals of time (which they will do under V law in a controlled experiment), but not perceiving it at all because it would take more journeys than they make to experience an average to within the required degree of accuracy. Time savings in transport appraisal nearly always relate to statistical distributions, not precise changes, which we ignore.
d) The values are also statistical distributions, not fixed points. In recent work with David Hensher (Transport Policy, in the press), we come to the view that ignoring the skewed nature of this distribution is likely to overestimate the revenue from a fast tolled road: even if the mean value of time is correct, not as many people in the population will use the tolled facility as calculated.
So overall, it is my view that underestimation of the VALUE of small time savings is unlikely to be a material issue in transport evaluation, even if overestimation of the DURATION of small times is established at the level of the individual, as seems entirely plausible. I think the choice is still (as ever) between applying the standard unit time values, or lower ones, not higher ones. Until now I've always persuaded myself that standard values is the right answer, but with decreasing confidence.
Best regards,
Phil
---------------------------------
Do you Yahoo!?
The New Yahoo! Shopping - with improved product search
|