Review of Income and Price Elasticities in the Demand for Road Traffic
|
Mark Hanly, Joyce Dargay and Phil Goodwin ESRC Transport Studies Unit Tel: +44 20 7679 1586 |
Final Report March 2002
Executive Summary
1.1 Introduction
This is the non-technical summary of a literature review carried out at the request of the Department of Transport, Local Government and the Regions (DTLR). Its results are given in round terms, and without full discussion of the caveats necessary for their interpretation. It is followed by a moderately technical report with fuller results including discussion of various inconsistencies, discrepancies and caveats, and Appendices with mathematical and statistical details. Demand elasticity is a measure of the strength of the influence of changes in price and income. The main focus is the effect of changes in fuel price and income, on either the amount of fuel bought, or the total level of traffic (measured by vehicle kilometres). Some associated results concerning car purchase prices, the level of vehicle ownership, and effects on the efficiency of fuel use are also reported.
1.2 The Data
Published studies, confined to those carried out in the UK or other countries broadly comparable with the UK, were collected from academic journals, government reports, researchers and consultants (including, but not giving special attention to, studies carried out by the authors of this report). Altogether 69 different studies of this type were collected, after filtering to ensure that the same results were not included more than once as a result of repeated publication in different forms or minor variants. They were reinforced by a larger, and wider, literature adding other useful evidence, earlier reviews, etc. These 69 studies produced 175 different equations, containing 491 elasticities, based on data covering different periods spread over the 62 years from 1929 to 1991. Over 100 results dealt with fuel consumption, over 30 dealt with traffic levels, and others covered car sales and fuel efficiency. Nearly all were either for cars only, or for cars and lorries added together. At the aggregate level of interest to the review, there was very little evidence related to commercial traffic as a whole. Other data used included:
- Statistical evidence for the UK, from which we checked some of the findings of the literature review
- Other researchers' literature reviews, carried out earlier, from which we checked whether our conclusions were the same as theirs
- Some indirect evidence in other published work, which, while not producing elasticities, nevertheless provided evidence on related responses.
1.3 Methods of Analysis
After carefully reading each of the studies, we transferred all their relevant results to a computer database, and then used this to calculate their range of results, the average, and whether the results were different according to the methods, definitions and scope they had used. Using a similar approach, we then considered whether there were any definite patterns in the results, especially whether the elasticities had been changing over time, whether short term effects were different from long term effects, and whether differences in the elasticities found were themselves, in turn, influenced by combinations of other factors, and if so, which factors had the greatest influence. Finally, some of the patterns were checked by statistical analysis of the UK data, to see if the same effects were noticed.
1.4 The Main Results
All the following figures are the average of quite a wide range of different answers. Nearly all the studies are 'symmetrical', ie they assume that the effects of a reduction are equal and opposite to the effects of an increase, both for price and income. There is some evidence that this assumption may not be true.
Price Effects
Taking what we judged to be the best defined results, the overall picture they imply is as follows. (All the statements may, according to the assumption of symmetry, be reversed by replacing 'up' and 'down').
If the real price of fuel goes, and stays, up by 10%, the result is a dynamic process of adjustment such that:
a) The volume of traffic will go down by roundly 1% within about a year, building up to a reduction of about 3% in the longer run (about five years or so).
b) The volume of fuel consumed will go down by about 2.5% within a year, building up to a reduction of over 6% in the longer run.
The reason why fuel consumed goes down by more than the volume of traffic, is probably because price increases trigger more efficient use of fuel (by a combination of technical improvements to vehicles, more fuel conserving driving styles, and driving in easier traffic conditions). So further consequences of the same price increase are:
c) Efficiency of use of fuel goes up by about 1.5% within a year, and around 4% in the longer run.
d) The total number of vehicles owned goes down by less than 1% in the short run, and 2.5% in the longer run.
At face value, the results imply that the sensitivity of car ownership with respect to fuel price is rather large, constituting a larger part of the effect of price on traffic levels. We draw attention to a strong caveat: many studies only assess the effects on car ownership, or on traffic, or on use per car, but not at the same time or using the same data, therefore this conclusion is based on drawing together quite different studies. Considerations of sample sizes suggest that the two effects (c) and (d) are somewhat less well supported than (a) and (b). At this stage our view is that the results do give support for the idea that the effects of prices on car ownership are important enough to take seriously, but are not necessarily such an overwhelmingly large part of the overall effect.
Income Effects
If real income goes up by 10%:
a) The number of vehicles, and the total amount of fuel they consume, will both go up by nearly 4% within about a year, and by over 10% in the longer run.
b) But the volume of traffic does not grow in proportion: 2% within a year and about 5% in the longer run.
Taken together, these would imply that use per car declines as income increases, though (as with the price effect above) this depends on comparison of different studies, and is not yet well supported by direct evidence. (A small number of studies show a direct hint of this in the short run, but not in the long run). It is possible that as incomes increase, successive new car owners are attracted into the car market who have less inclination to drive much. An additional effect implied is that rising income has generally been associated with a fall in the efficiency of use of fuel, for which a possible reason might be that as incomes grow people buy newer, but larger, vehicles.
Does it make a difference what method is used to calculate the elasticities?
One strong, repeated and consistent result is that studies using methods which allow explicit estimation of short run and long run elasticities separately, nearly always find that the long run effect is substantially higher than the short run effect, both for price and income, and for all measures of demand. Beyond this, there are many indications of results which are significantly different in a statistical sense, but with no obviously useful or consistent pattern: we cannot say that there is any obvious pattern for certain methods to produce consistently higher or lower elasticities than other methods. The most obvious difference was in the case of studies using methods which do not allow for the difference between short and long run effects, ie the effect refers to an end-state with no reference to the process or time scale of getting there. We call these 'static', and they are often different from the short and long run results, though not consistently in the same direction. Many (but not all) of the income elasticities produced by this method seem to be high, and rather similar to long run effects. Many (but not all) of the price elasticities produced by this method seem to be low, and more similar to short run effects, or between the short and long run effects. There are theoretical reasons why this puzzling difference might be true, but testing for these reasons is no easier than calculating the explicit difference between short and long run effects directly. We advise use of the dynamic estimates wherever possible. Even where different methods produce rather similar figures at average levels of price or income, they will not do so when used for forecasting the effects of much higher or lower future levels, or for changes much bigger than the observations on which they are based. There is no obvious pattern such that results which are more credible for forecasting outside the observed range, will necessarily be shown to perform better within the range of data available.
Are the effects the same for cars as for trucks?
We did not have sufficient information in the studies to calculate the freight transport effect eparately, but there are two pieces of relevant evidence: First, the effects of a price increase for diesel plus petrol causes a smaller reduction in the total amount of fuel bought, than for petrol alone. Secondly, the effect of an overall fuel price increase has a smaller effect on the total traffic level (including lorries) than petrol prices have on the private car traffic. Although not all goods vehicles use diesel and not all cars use petrol, these two results taken together suggest that goods traffic is less sensitive to price, and private cars more sensitive. The difference is large enough to be important, but not well defined enough in the data to provide a definite figure, because the proportion of lorries and cars varies greatly, but is not recorded in most of the studies. The same is not true for effects of changes in income, for which the effect on personal transport and goods transport seems to be rather similar in size.
Are the results for the UK the same as for other countries?
The UK appeared in a few studies on its own, and more in combination with other countries, eg EU members. For the studies which included the UK, the results are slightly smaller than the overall average quoted above for the effect of price on fuel consumption, and slightly bigger for the effects of price on traffic (hence with a lower fuel efficiency effect than the data base as a whole), and the income effects were slightly smaller. For a few UK-only studies, the effects of price are slightly smaller, and at face value suggest that there has been no fuel efficiency impact at all, though this is at odds with other evidence and is not strongly supported.
Have the elasticities been changing over time?
There is a widespread practice, used in DTLR forecasts, to assume that both price and income elasticities would get absolutely smaller over time, ie the market becomes progressively less sensitive to price and income changes. Separately, it is assumed that price elasticities will go down if the price level in real terms goes down. For price elasticities, there is no strong evidence that this has happened. Studies in which the price elasticity has grown over time, or stayed constant, are more evident as those showing a fall, especially for the longer-term effects which are most relevant for the forecasts. One exception to this is a common (but not universal) finding that price elasticities were higher during the period of very high prices following 1974. Prima facie, this pattern supports the assumption that price elasticities are related to price level, but does not support the idea that they are (via value of time) inversely proportional to income and therefore decline over time. A wide variation of results is also seen in the pattern of income effects, but in this case the balance of evidence does suggest that they have been declining over time, as income increases.
The full report in PDF format is available as a download at the foot of this page.
For related documents, pages and internet links, see the column on the right.
