elasticity of gaselone case study

24
Forthcoming in Journal of Transport Economics and Policy 1 The demand for automobile fuel: a survey of elasticities. Daniel J. Graham and Stephen Glaister Research Fellow and Professor of Transport and Infrastructure, Department of Civil Engineering Imperial College of Science, Technology and Medicine London, SW7 2BU UK. Abstract. This paper reviews the international research on the response of motorists to fuel price changes and assesses the orders of magnitude of the relevant income and price elasticities. The paper highlights some new results and directions that have appeared in the literature. The evidence shows important differences between the long and short run price elasticities of fuel consumption. Long run price elasticities typically fall between -0.6 and -0.8, short run elasticities are around 30% to 40% as large (between -0.2 and - 0.3). Price effects on traffic levels are smaller, about -0.30 in the long run and -0.15 in the short run. The long run income elasticity of fuel demand falls in the range 1.1 to 1.3 and between 0.35 and 0.55 in the short run. Introduction Road traffic levels and the volume of emissions generated by motor vehicles are amongst the most controversial issues faced by most governments. Taxation on fuel and building new road capacity are two important policy instruments that might be used to influence them or to respond to them. Transport tax is also a major source of government revenue. For forecasting and evaluation of policy four of the most critical considerations are the responsiveness of traffic levels to changes in fuel price, responsiveness of total fuel consumption to changes in fuel price and responsiveness of both of these to increases in general standards of living. These are characterised by four simple parameters, the price elasticity of traffic, the price elasticity of demand for fuel and the income elasticities of both. National transport policy cannot be successfully developed without a view on the magnitudes of these parameters. Yet considerable confusion was clearly displayed by the general public and senior politicians during the crises in United Kingdom and other parts of Europe, precipitated in 2000 by increasing crude oil prices combined with a long- standing policy of increasing rates of fuel tax. This review reports empirical evidence from a number of different countries, emphasising differences that are found between the long and short run price elasticities. The purpose is to provide an up-to-date survey of the international fuel demand literature, leading to an assessment of the general magnitude of the relevant elasticities. The paper is not a methodological review. It focuses on identifying the main themes in the literature and

Upload: aman-mittal

Post on 20-Jul-2016

12 views

Category:

Documents


1 download

DESCRIPTION

Case study on elasticity of gasoline

TRANSCRIPT

Page 1: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

1

The demand for automobile fuel: a survey of elasticities.

Daniel J. Graham and Stephen GlaisterResearch Fellow and Professor of Transport and Infrastructure,

Department of Civil EngineeringImperial College of Science, Technology and MedicineLondon, SW7 2BUUK.

Abstract. This paper reviews the international research on the response of motorists tofuel price changes and assesses the orders of magnitude of the relevant income and priceelasticities. The paper highlights some new results and directions that have appeared inthe literature. The evidence shows important differences between the long and short runprice elasticities of fuel consumption. Long run price elasticities typically fall between-0.6 and -0.8, short run elasticities are around 30% to 40% as large (between -0.2 and -0.3). Price effects on traffic levels are smaller, about -0.30 in the long run and -0.15 in theshort run. The long run income elasticity of fuel demand falls in the range 1.1 to 1.3 andbetween 0.35 and 0.55 in the short run.

Introduction

Road traffic levels and the volume of emissions generated by motor vehicles are amongstthe most controversial issues faced by most governments. Taxation on fuel and buildingnew road capacity are two important policy instruments that might be used to influencethem or to respond to them. Transport tax is also a major source of government revenue.

For forecasting and evaluation of policy four of the most critical considerations are theresponsiveness of traffic levels to changes in fuel price, responsiveness of total fuelconsumption to changes in fuel price and responsiveness of both of these to increases ingeneral standards of living. These are characterised by four simple parameters, the priceelasticity of traffic, the price elasticity of demand for fuel and the income elasticities ofboth.

National transport policy cannot be successfully developed without a view on themagnitudes of these parameters. Yet considerable confusion was clearly displayed by thegeneral public and senior politicians during the crises in United Kingdom and other partsof Europe, precipitated in 2000 by increasing crude oil prices combined with a long-standing policy of increasing rates of fuel tax.

This review reports empirical evidence from a number of different countries, emphasisingdifferences that are found between the long and short run price elasticities. The purposeis to provide an up-to-date survey of the international fuel demand literature, leading toan assessment of the general magnitude of the relevant elasticities. The paper is not amethodological review. It focuses on identifying the main themes in the literature and

Page 2: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

2

seeks to illustrate some of the new results and directions that have appeared in recentresearch.

Earlier extensive surveys of this literature are now well known (e.g. Drollas 1984, Oum1989, Dahl and Sterner 1991a, 1991b, Goodwin 1992). The most informative of thesesurveys are noted here to provide a general view about the orders of magnitude of theelasticities relevant to fuel demand. The paper then goes on to draw out some recentwork, which has added substantial content to the field by focusing on specific issues orby using innovative data or methodology.

Major review articles

Survey articles on the characteristics of fuel demand are noted here in chronologicalorder. In most cases these studies provided new empirical estimates as well as reviewmaterial. Where this is the case both contributions are reported. By focusing oncomprehensive reviews, which collectively cover hundreds of individual studies, thissection seeks to arrive at a balanced view of the likely orders of magnitude of fueldemand elasticities.

Drollas (1984) surveys a variety of academic and non-academic studies of gasolinedemand elasticities and also provides his own estimates for European countries in the1980s. The author cites price and income elasticities from previous studies predominantlyestimated for the US. His survey spans different modelling techniques including staticcross-sectional specifications and time-series and pooled cross-section time-series modelswith a variety of lag structures. While a range of estimates are found in the literature, theconsensus view is that the long run price elasticity of demand is around -0.80 while thelong run income elasticity is slightly below unity. Only some of the studies reviewed byDrollas distinguish short from long run elasticities. Those that do typically find short runprice elasticities to be one-third the magnitude of the long run, and short run incomeelasticities to range from a quarter to a half of the long run estimate. The review oflimited existing evidence on other countries suggests no substantial differences from theUS.

Drollas also provides his own price elasticity estimates for European countries over theperiod 1950 to 1980. The motivations for his empirical work are to extend the analysisbeyond the timeframe of the previous studies he reviews to include the mid 70s oil crisis,to incorporate a wider range of nations, and to implement a vehicle-stock adjustmentmodel which he believes to be well specified yet economical in data requirements.Specifically, Drollas estimates a vehicle stock adjustment model in its reduced formwithout explicit consideration of the vehicle stock itself. He estimates dynamic models inlog-linear form that relate gasoline consumption to income, the real price of gasoline, thereal price of other transport services, and the price of vehicles. The models are estimatedwith endogenous and exogenous lags according to geometric and inverted-V lag schemes.

The author’s results yield long run price elasticity estimates of approximately -0.6 for theUK, -0.8 to -1.2 for West Germany, -0.6 for France and -0.8 and -0.9 for Austria. These

Page 3: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

3

compare to short run figures of around -0.26 for the UK, -0.41 and -0.53 for WestGermany, -0.44 for France, and -0.34 and -0.42 for Austria.

Thus, he finds that while gasoline demand may be inelastic in the short run it is less so inthe long run and these results are consistent with those of the previous studies he reviews.However, Drollas believes his estimates give evidence that the ‘true’ long run priceelasticities, particularly for European countries, may well be above unity. He attributesthese higher than expected long run elasticities to substitute types of transport fuels(diesel, liquefied petroleum gas), substitute forms of transport, and the fact thatconsumers can switch expenditure to activities or goods that compete with transport.Other important findings of this study are that similarity rather than diversity existsbetween countries in the characteristics of fuel demand, and that inertia in gasolineconsumption can be explained by the slowly changing vehicle stock and by thepersistence of inefficient habits.

Blum et al. (1988) review studies on aggregate time series gasoline demand models forWest Germany and Austria. The authors set out a typology of gasoline demand studiesbased on the formal econometric structure of the models used and provide a commentaryon the results obtained. Models are distinguished with respect to the form of the demandfunction, the treatment of time, the structure of the error component, and the estimationtechnique. The paper emphasises short-term elasticities. The studies they review forGermany and Austria give short run price and income elasticities over very large ranges,from –0.25 to –0.83 and from 0.86 to 1.90 respectively.

The authors express concerns over the demand specifications used to estimate theseelasticities. They argue that while many previous studies have interesting modelcharacteristics and estimation techniques, they are also typically characterised bydifferent restrictive functional forms, which have given rise to much of the variationbetween estimates.

Blum et al. go on to review some results for Germany by Foos (1986) which examine amuch larger number of variables than commonly found in gasoline demand studies,including important exogenous variables such as the level of economic activity, the pricesof other goods, weather conditions, and the availability of infrastructure. The data usedby Foos are for West Germany and are monthly from January 1968 to December 1983.

Foos’s results give a short-run price elasticity of –0.28 and income elasticity of 0.25. Theshort run price elasticity is of fairly typical magnitude but the income elasticity is smallerthan commonly reported. Blum et al. explain this result by pointing out that the modelalso contains variables reflecting the level of economic activity (i.e. employment, retailsales, industrial activity) – adding the elasticities of these variables to the elasticity ofincome gives a total elasticity of 1.22. Thus the authors argue, that by not explicitlyspecifying dimensions of the level of economic activity in gasoline demand models,which ultimately generates travel, previous studies have greatly over-estimated the pureincome elasticity.

Page 4: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

4

Other interesting results reported include the cross-elasticity of gasoline demand withrespect to the price of mass transit, estimated at 0.39, and the elasticity of fuelconsumption rate of cars, estimated at 0.61. Thus, a 10% increase in fuel efficiencybrings about a decrease in fuel consumption of 6.1% - motorists compensate by drivingmore. The authors also find that the availability of infrastructure and its quality has animportant bearing on fuel demand though they determine only a small impact fromweather conditions.

Sterner (1990) examines the pricing and consumption of gasoline in OECD countries. Hissurvey finds long run price elasticities falling in the interval -0.65 to -1.0 and for incomebetween 1.0 and 1.3. Using data for the OECD between 1962 and 1985 Sterner provideshis own set of estimates. He finds long run price elasticities of between -1.0 and -1.4using pooled data. The corresponding income elasticities vary from 0.6 to 1.6. Using timeseries data the price elasticities are between -0.6 to -1.0, and 1.1 to 1.3 for income. Theshort run elasticities for dynamic models appear to be around -0.2 to -0.3 for price, and0.35 to 0.55 for income.

Thus, the treatment of time and the particular methodological approach can have a crucialbearing upon the magnitude of elasticity estimates. Goodwin (1992) explores theseissues, updating previous work on gasoline price elasticities in his review of academicand non-academic studies undertaken in the 1980s and 1990s. His paper shows that morerecent work has generally revised the magnitude of elasticity estimates upwards. Theunweighted mean value of 120 elasticities of gasoline consumption with respect to fuelprices considered in the review is -0.48, compared with similar values from previousreviews of -0.1 to -0.4.

Goodwin highlights differences between recent studies by categorising estimates of theelasticity of gasoline consumption with respect to fuel price into cross-section or timeseries, and sub dividing this distinction into short term, long term, or ambiguous. The‘short term’ period generally refers to less than one year and the ambiguous categoryrefers to estimates obtained from models with no explicit consideration of the timedimension. Goodwin’s summary of results is reproduced in Table 1 below.

Table 1: Summary of Evidence from Studies of Elasticity of Gasoline Consumption withRespect to Price.

Explicit AmbiguousShort term Long term

Time-series -0.27(0.18, 51)

-0.71(0.41, 45)

-0.53(0.47, 8)

Cross-section -0.28(0.13, 6)

-0.84(0.18, 8)

-0.18(0.10, 5)

Note: Figures in parentheses are standard deviations and the number of quoted elasticitiesin the average.Source: Goodwin (1992)

Page 5: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

5

The results illustrate the difference in magnitude that exists between the short and long-term elasticities of fuel price increases on gasoline consumption. Long term elasticitiestend to be between one and a half and three times higher than the short term. However,having reviewed a wide range of studies, Goodwin also shows that differences inmethodological approach, in this case between time-series and cross-section methods,only marginally affected the magnitude of the elasticities.

The review also considers the effects of gasoline prices on traffic levels. An earlier paperby Dix and Goodwin (1982) hypothesised that the short run elasticities of traffic levelsand of gasoline consumption with respect to fuel price would be identical, but that theywould diverge over time as the long run gasoline consumption elasticity grew faster thanthe traffic elasticity. The reasoning here was that changes in trip rates, car ownership,destination choice, and location decisions would take some time to occur, and thatchanges in vehicle size and efficiency would have a strong effect on consumption whilepreserving mobility.

Goodwin’s evidence of elasticity effects of traffic levels with respect to fuel prices areshown in table 2.

Table 2: Summary of Evidence from Studies of Elasticity of Traffic with Respect toPrice.

Explicit AmbiguousShort term Long term

Time-series -0.16(0.08, 4)

-0.33(0.11, 4)

-0.46(0.40, 5)

Cross-section - -0.29(0.06, 2)

-0.5(N/A., 1)

Note: Figures in parentheses are standard deviations and the number of quoted elasticitiesin the average.Source: Goodwin (1992)

Table 2 does not support the Dix and Goodwin hypothesis. While it is the case that long-term elasticities are larger than short term, both short and long term effects of gasolineprices on traffic levels are much less than their effects on gasoline consumption.Goodwin notes that this is indicative of rapid behavioural responses that affect gasolineconsumption more than traffic. He suggests that they may be changes in driving style orspeed, or by modifying the least energy efficient journeys. If this is true, then it wouldseem that gasoline price manipulation might be a more effective tool where the objectiveis to decrease fuel consumption than to reduce road congestion.

With respect to the time effect in the magnitude of elasticities Goodwin draws threeimportant implications. First, behavioural responses to cost changes take place over timeand that this implies that time-independent estimates are subject to error. Second, therange of responses considered credible has to be extended to include changes in carownership, vehicle type, location decisions and the use of public transport. Third, policy

Page 6: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

6

options are wider than perceived by earlier studies and pricing has a powerful cumulativeeffect on the pattern of travel demand.

Sterner et al. (1992) examine the price sensitivity of transport gasoline demand. Theyreport results from earlier surveys (Dahl and Sterner 1991a, 199b) which stratify a widevariety of previous results by the type of model and data used, and calculated averageelasticities for each category. Results from dynamic models for OECD countries over theperiod 1960-85 show great degrees of difference in the short and long term magnitude ofprice and income elasticities. The short run price elasticity of gasoline demand variesbetween -0.10 to -0.24 depending on the model estimated. The equivalent long run figureis between -0.54 and -0.96. Averaging these estimates gives a short run value of -0.23and a long run figure of almost three and half times as large, -0.77. The average nationalincome short run elasticity is given as 0.39 and the long run as 1.17. Sterner et al. notethat the indication that the absolute value of the income elasticity is higher than for pricesuggests that gasoline prices must rise faster than the rate of income growth if gasolineconsumption is to be stabilised at existing levels.

Sterner et al. present the short and long run price and income elasticity estimatesgenerated from lagged endogenous variable models for 20 OECD countries. Thesefigures are shown in table 3.

Page 7: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

7

Table 3: Price and Income Elasticity Estimates of Gasoline Demand, OECD Countries,1960 - 1985.

Price Elasticities Income elasticitiesSR LR SR LR

Canada -0.25 (0.06) -1.07 (0.24) 0.12 (0.09) 0.53 (0.40)US -0.18 (0.03) -1.00 (0.15) 0.18 (0.07) 1.00 (0.38)Austria -0.25 (0.11) -0.59 (0.26) 0.51 (0.23) 1.19 (0.54)Belgium -0.36 (0.05) -0.71 (0.09) 0.63 (0.19) 1.25 (0.39)Denmark -0.37 (0.06) -0.61 (0.10) 0.34 (0.08) 0.71 (0.17)Finland -0.34 (0.15) -1.10 (0.47) 0.39 (0.24) 1.26 (0.76)France -0.36 (0.08) -0.70 (0.15) 0.64 (0.23) 1.23 (0.43)Germany -0.05 (0.07) -0.56 (0.82) 0.04 (0.16) 0.48 (1.92)Greece -0.23 (0.11) -1.12 (0.52) 0.41 (0.19) 2.03 (0.93)Ireland -0.21 (0.04) -1.62 (0.33) 0.12 (0.14) 0.93 (1.06)Italy -0.37 (0.13) -1.16 (0.40) 0.40 (0.17) 1.25 (0.52)Netherlands -0.57 (0.11) -2.29 (0.46) 0.14 (0.13) 0.57 (0.52)Norway -0.43 (0.13) -0.90 (0.28) 0.63 (0.20) 1.32 (0.42)Portugal -0.13 (0.07) -0.67 (0.34) 0.37 (0.18) 1.93 (0.94)Spain -0.14 (0.17) -0.30 (0.37) 0.96 (0.45) 2.08 (0.98)Sweden -0.30 (0.09) -0.37 (0.11) 0.51 (0.30) 0.99 (0.59)Switzerland 0.05 (0.16) 0.09 (0.28) 0.85 (0.29) 1.54 (0.53)UK -0.11 (0.07) -0.45 (0.27) 0.36 (0.20) 1.47 (0.81)Australia -0.05 (0.02) -0.18 (0.07) 0.18 (0.07) 0.71 (0.29)Japan -0.15 (0.03) -0.76 (0.17) 0.15 (0.01) 0.77 (0.06)Turkey -0.31 (0.06) -0.61 (0.11) 0.65 (0.16) 1.29 (0.32)

Mean -0.24 (0.09) -0.79 (0.29) 0.41 (0.18) 1.17 (0.62)Note: Standard errors are given in parentheses.Source: Sterner et al. (1992)

Given unweighted mean standard errors the 95% confidence interval for the average shortrun elasticity is approximately –0.06 to –0.42, and for long run –0.21 to –1.37. The longrun income elasticity is about 2.8 times as large as the short run. Excluding Germany,Spain and Switzerland, which have extremely low t-ratios, and re-calculating the figuresincreases the confidence intervals for average price elasticities. For short run elasticities,the confidence interval is approximately –0.12 to –0.44, and for long run elasticities from–0.38 to –1.38. The long run mean price elasticities for the OECD countries areapproximately 3.3 times as large as the short run elasticities. The difference in order ofmagnitude for the UK between the short and long run is, however, much greater, with anelasticity of about 4.1 times as large in the long term.

Sterner and Dahl (1992) extend the investigation into methodological issues, reviewing alarge number of different models that have been developed to explain how gasolinedemand is related to price, income, and other variables. They find that different modelspecifications can give very different estimates and they compare model results byapplying them to the same OECD data set (1960-1985). Long run elasticities can be

Page 8: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

8

estimated with either dynamic models on ordinary time series data or with static modelson cross-section data. The dynamic models give estimated price elasticities within therange -0.80 to -0.95, and income elasticities of between 1.1 and 1.3. Static models forcross-section data give roughly unitary elasticities for both price and income. Sterner andDahl also note that models using pooled data estimate price elasticities as high as -1.3 or -1.4. Short run estimates from dynamic models generally fall in the range -0.1 to -0.3 forprice and 0.15 and 0.55 for income.

Dahl (1995) reviews a number of previous gasoline demand surveys conducted since1977 and updates this work with evidence from the most recent US studies. The tablebelow summarises the results reviewed by Dahl.

Table 4: Demand Elasticity Estimates Reported by Dahl (1995).price elasticity income elasticity

short run long run short run long runTaylor (1977) -0.1 to -0.5 -0.25 to -1.0 - -Bohi (1981) -0.2 -0.7 - ≈ 1.0Kouris (1983) - -1.09 - -Bohi & Zimmerman (1984) 0.0 to -0.77 0.0 to -1.59 -0.18 to 1.20 -0.34 to 1.35Dahl (1986) -0.29 -1.02 0.47 1.38Dahl & Sterner(1991a,1991b)

-0.26 -0.86 0.48 1.21

Goodwin (1992) -0.27 -0.71 to –0.84 - - Source: Dahl (1995)

The studies reviewed were concerned with price elasticities in the industrialised worldand they generally found long run price elasticities between –0.7 and –1.0 and long runincome elasticity greater than 1.0. Dahl notes that these results suggest that taxes maywell be an effective means of reducing pollution from gasoline use, but to keep useconstant fuel prices would have to rise faster than income.

Dahl reviews 18 recent studies on gasoline demand from the US to explore how elasticityestimates have changed. For studies based on static models, she finds slightly long runprice and income elasticities of –0.16 and 0.46 from studies based on recent data,somewhat smaller in magnitude compared to previous estimates of –0.53 and 1.16.However, static analyses tend to produce intermediate run, rather than long run, priceelasticity estimates, and Dahl’s review of dynamic models shows no substantial reductionin the magnitude of the elasticity estimates. For instance, estimates based on laggedendogenous variable models shows short and long run price elasticities of –0.19 and-0.66 and short and long run income elasticities of 0.27 and 0.28. Those based on theinverted V model show long run price and income elasticities of –1.20 and 1.22.

Dahl believes on balance that elasticities have become smaller in magnitude over time,particularly for income. While previous studies show long run price and incomeelasticities of around –0.8 and 1.0, recent studies suggest a price response of around –0.6and a slightly inelastic income response. The reliability of these results, however, is

Page 9: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

9

tempered by the small number of estimates reviewed in Dahl’s update and by thepredominance of static models.

On the basis of the surveys reviewed in this section, which have assimilated manyhundreds of studies, we get a clear indication that despite variation in elasticities of fueldemand there are fairly narrow ranges within which the values typically fall. Short-termprice elasticities tend to be between –0.2 and –0.3 while the long run elasticities typicallyfall between –0.6 and –0.8. For income, the long run elasticity is usually estimated asslightly higher than unity (1.1 to 1.3) and the short run elasticity in the range 0.35 to 0.55.

However, while the overwhelming evidence points towards values within these rangesthe review articles do not categorically account for the variation in the estimates. Thefollowing sections attempt to shed some light on this issue. They draw upon recentstudies that have added substantially to our understanding of elasticity estimates byexploring specific themes, or by explicitly setting out to explain the variation in elasticityestimates.

Micro-level data: individual and household demand studies.

One important issue surrounding gasoline demand elasticity estimates is the analyticaldifferences permitted by the use of dissaggregate as opposed aggregate data. Most of theestimates reviewed above, and the vast majority of gasoline demand studies in general,are based on aggregate level data at the country or sub-national level. Thus, these studiesconsidered both commercial and consumer demand. Some authors have recently shownthat the use of micro-level data, which reflects individual and household behaviour moreclosely, can add detail to our understanding of the temporal nature of consumer response.

Eltony (1993) uses household data to quantify the behavioural responses that give rise tonegative price elasticities of demand for gasoline. He estimates household gasolinedemand in Canada using pooled time-series and cross sectional provincial householddata. His model recognises three main behavioural responses of households to changes ingasoline prices: drive fewer miles, purchase fewer cars and buy more efficient vehicles.Eltony estimates five separate equations that attempt to explain: gasoline demand per car,the stock of cars per household, new car sales per household, new car fuel efficiency andthe choice of vehicle type. Using pooled time-series and cross-section data on theCanadian provinces from 1969-1988 he estimates short run gasoline price elasticities percar, holding fuel economy constant, of -0.21 and a short run income elasticity of 0.15.

From these estimates Eltony goes on to determine dynamic price elasticities of gasolinedemand for Canada by simulating the model over the period 1989 - 2000. He assumes abase case in which real household income, the unemployment rate, the real price of newcars, the interest rate, and the real price of gasoline per gallon in Canada and the US areequal to 1988 values and remain constant for the rest of the time horizon. In analternative solution to the model the real prices of gasoline in Canada and the US areassumed to increase by 10%. The two model solutions are obtained and the percentagechange in gasoline consumption computed.

Page 10: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

10

His results for short term (one year) and long term (two to ten years) are given in table 5below.

Table 5: Dynamic Price Elasticities of Gasoline Demand in Canada.year1 -0.31202 -0.46733 -0.53704 -0.59815 -0.69846 -0.81327 -0.89358 -0.94789 -0.983910 -1.007311 -1.019212 -1.0239 Source: Eltony (1993)

Table 5 demonstrates a number of important points about short run and long run effectsof increasing the price of fuel. The short run dynamic own price elasticity of gasoline isestimated at -0.31. He finds that almost 75% of household response to price changes inthe first year can be attributed to driving fewer miles. A further 10% results from analteration in the composition of the fleet to more fuel efficient vehicles and the remaining15% can be attributed to changes in the size of the fleet. Eltony finds intermediate term (5year) price elasticities ranging from -0.689 to -0.709 and the long term elasticities from -0.975 to -1.059. Table 5 also shows a rapid response to price increases within the firstfour years. Eltony interprets these results as pointing to the importance of improving fuelefficiency as an effective means of reducing household gas consumption.

Rouwendal (1996) seeks direct verification of the validity of short-term behaviouralresponses to fuel price increases using individual consumer data. The author obtainedinformation about fuel use per kilometre driven from the Dutch Private Car Panel, arotating panel in which car drivers participate for three months. Rouwendal seeks toinvestigate the relationships between fuel use and other recorded information about carsand their drivers in the short run. With respect to cars, he is able to observe weight,cylinder volume, year of construction, and type of fuel. Known driver characteristicsinclude sex, classifications of age and income, total number of kilometres driven eachyear by the main car user, information about business, whether the driver receivescompensation for the cost of the car, and for employed people, the distance betweenresidential and work location. Monthly information about fuel prices in Holland isavailable.

The author presents OLS estimates for specifications that are linear in parameters withthe logarithm of the number of kilometres driven per litre of fuel as the dependant

Page 11: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

11

variable. His results show heavier cars to be less fuel-efficient than others and diesel carsto be more fuel efficient. Gender effects are not found but age is important with thedriving style of the elderly generally being less fuel-efficient. As regards the gasolineprices, Rouwendal estimates that a 10% increase in fuel price will induce drivers toincrease the average distance per litre of fuel by 1.5%. Rouwendal regards this centralresult as verification of the significant effect of gasoline prices on fuel use in the shortrun. Surprisingly, the income of the main driver is found to be insignificant, although thetype of employment is not. Rouwendal points out that this result conflicts with thecommonly held belief that there are short run income effects. It is, however, perhapsconsistent with the finding of Blum et al. (1988) that some explicit consideration of‘economic activity’ in gasoline demand models substantially reduces the magnitude ofthe income elasticity.

Short-term response is also investigated by Hensher et al. (1990) in an earlier study, butin this case with respect to vehicle use and fuel price. The authors develop a model toexplain vehicle kilometres per annum for households in the Sydney metropolitan area interms of a range of vehicle characteristics as well as household price and incomeattributes. They are able to distinguish elasticities on the basis of household carownership characteristics. Their data covers the period 1981 to 1982 for 1,172households. Hensher et al. start from the premise that households face a set of alternativevehicle technologies and select the one that is consistent with the maximisation of thejoint utility of vehicle choice and use. Parameter estimates are presented in the absence ofselectivity of vehicles, and in the presence of selectivity where that is derived from thenon-linear specification of the type choice model.

Heshner et al.’ s results are consistent with Rouwendal’s findings on short-termresponses. They show a substantial price effect on vehicle use but only small andinsignificant effects from household income in the short term. The estimated short runprice elasticities of vehicle use are –0.26 for 1 vehicle households, -0.33 for 2 vehiclehouseholds and –0.39 for 3 vehicle households. However, the authors find that income isnot confirmed as an important empirical influence on vehicle use, except for two vehiclehouseholds, with an estimated elasticity of 0.14.

Puller and Greening (1999) provide a recent example of the use of micro level data toidentify the intricacies of temporal response to short run gasoline price changes. Theyreview short run estimates of price elasticities of gasoline demand from a number ofprevious studies based on dissagregated household data. A summary of this review isprovided in table 6 below.

Page 12: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

12

Table 6: Estimates of Short-Run Price Elasticities from Studies Based on Household Datashort run price elasticity

Archibald and Gillingham (1980) -0.43Greene and Hu (1986) -0.5 to –0.6Walls et al. (1993) -0.51Greening et al. (1995) 0.00 to –0.67Dahl and Sterner (1991a) -0.52Source: Puller and Greening (1999)

Puller and Greening examine household adjustment to changes in the real price ofgasoline using a panel of US households over 9 years. They believe their work differsfrom the studies they review in two ways. First, they allow household vehicle stock tochange over time and therefore are able to capture long-run adjustments. Second, theydecompose demand into a vehicle usage and a vehicle stock component. The authorspresent a basic demand framework that explains the household demand for gasoline interms of contemporaneous and lagged real prices of gasoline, the real income of thehousehold, and a vector of household demographic characteristics.

Puller and Greening apply a variety of estimation techniques and lag structures to theirdata. Using one year lags, as previous studies have, the short-run price elasticity ofgasoline demand is estimated to be around –0.35, a figure they believe to be consistentwith estimates from the literature. However, when they use different specifications ofquarterly lagged prices they estimate a much larger price elasticity of –0.8. This, theyargue, indicates that the initial immediate response of consumers to a price rise involves amuch larger decrease in gasoline consumption compared to the total annual short-runelasticity.

This section has looked at how the gasoline demand studies using disaggregated datahave been used to shed more light on the temporal nature of behaviour response. Theconsensus from these studies is that short term price elasticity effects do exist and are ofthe order of magnitude suggested by the main survey articles reviewed above. There isevidence, however, that income effects are more difficult to determine in the short runusing disaggregated data. However, the models used at the micro level tend to be muchless restrictive in exogenous variable specification than the aggregate studies and, asBlum et al. (1988) suggest, this may well account for the absence or reduction of theincome elasticity.

Vehicle technology and fuel efficiency

Many recent studies have investigated fuel efficiency and vehicle technologycharacteristics in gasoline demand models. Typically, the gasoline elasticities studies, andparticularly those using aggregate data, have either not explicitly modelled fuel efficiencyor have accorded the issue inadequate attention. Interest in the role of fuel efficiency hasgrown in recent years as researchers try to understand the implications of fiscal policy fortraffic levels, vehicle emissions, and environmental externalities (e.g. Hall 1995,Koopman 1995, Small and Kazimi 1995, Crawford and Smith 1995, Eyre 1997,

Page 13: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

13

McCubbin and Delucchi 1999, Delucchi 2000). This section draws together someprominent research from the elasticities literature that has considered this particulardimension of fuel demand.

Baltagi and Griffin (1983) provide an early example of the explicit treatment of fuelefficiency effects in gasoline demand estimation. They are interested in the magnitude ofthe price elasticity of demand for gasoline and review earlier studies that show widevariation in the magnitude of price elasticity estimates. For instance, Houthakker et al.(1974) in a study of the US indicate very low price elasticities of demand ranging from -0.04 to –0.24 using quarterly data for a cross-section of states. Sweeney (1978), on theother hand, using a model that incorporated the efficiency characteristics of theautomobile fleet, finds a higher long run price elasticity of –0.73.

Baltagi and Griffin are unhappy with such a wide range in estimates, believing them to besymptomatic of the methodology and data used. They wish to obtain more consistentestimates and to understand the implications for estimates of the method and data used.Applying 8 alternative estimation techniques to pooled cross-section time series data theyset out to quantify the magnitude of the price elasticity of gasoline demand in OECDcountries for the period 1960 to 1978. The model they propose explains gasolineconsumption per vehicle by income per capita, gasoline prices, the stock of cars percapita, and a proxy variable reflecting the level of vehicle efficiency.

Following the application of these different estimation methods Baltagi and Griffin findthat the long run price elasticity of gasoline demand typically falls within the range -0.6and -0.9 – a range consistent with the orders of magnitude given in most survey articles.However, in contrast to previous studies (i.e. Houthakker et al. 1974, Ramsey et al. 1975,Mehta et al., 1978) they find a slow adaptation rate with the major response being due tothe efficiency characteristics of the automobile fleet. Approximately 60% of theadjustment to the long run equilibrium takes place within the first five years – previousstudies had claimed it was almost instantaneous. Thus they find that adaptations in thegasoline efficiency of the fleet and driving conditions require long periods foradjustment.

Broader aspects of fuel efficiency are considered by Espey (1996b). She analyses the roleof fuel prices, income, government taxation and technological change in influencing theconsumers’ choice of fuel economy. The study uses an international data set thatcomprises observations on eight countries; USA, Japan, France, Germany, the UK,Norway, Sweden, and Denmark; between 1975 and 1990. The equation estimatedexplains the demand for fuel economy (average fleet fuel efficiency, km / litre) by fuelprices, per capita income, an automobile purchase and registration tax index, and a timetrend which is thought to reflect technological change.

Espey’s results indicate a price elasticity of fuel economy of around 0.20 but an incomeelasticity not significantly different than zero. The time trend in the model is also foundto be statistically significant implying a 2.8% annual increase in fuel efficiency over timethat is not explained by changes in fuel prices and income. The influence of time declines

Page 14: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

14

over time from 5% in 1975 to under 2% by 1990. Espey makes clear that the time trendcaptures a combination of pure technological improvements in fuel economy and theimpact of implicit and explicit environmental standards. The elasticity of fuel economywith respect to vehicle taxation is estimated at 0.09, and the coefficient on the laggeddependent variable is 0.94, indicating that only 6% of the effect of a change in fuelprices, income, or vehicle taxation takes place in the first year.

Espey considers the implications of her results for transport policy in the US. She arguesthat fuel prices account for around half the differences in fuel economy between the USand other countries in her study. There is however, no strong relationship betweenincome and fuel economy. The author also believes that purchase and registrationtaxation regimes have an important bearing on differences in fuel economy.

The issue of how fuel efficiency affects gasoline demand is explored directly by Oraschand Wirl (1997). Their investigation is motivated by a desire to explain the asymmetry ofgasoline demand with respect to energy prices. For the US, they note that the dramaticreduction in gasoline prices after 1986 did not have an effect on demand comparable tothe previous price increases of 1974 and 1979 / 1980. The authors investigate the effectof technical fuel efficiency on gasoline demand for the UK, France and Italy. Theyestimate an energy demand model with efficiency explicitly treated within an asymmetricframework and a second model excluding efficiency. They find that the explicitconsideration of energy efficiency proves less important than previously thought withlittle noticeable difference in price elasticity effects. The income elasticities are found todiffer – being higher with efficiency included in the model. The authors are scepticalabout the importance of technical efficiency for fuel demand. They conclude that energyand environmental taxes are unlikely to give rise to R & D efforts in efficiency unlessthey are very high. Otherwise, any response will be modest and come about only throughconsumer adjustments.

Johansson and Schipper (1997) examine aspects of car fuel in relation to decreasingoverall travel and increasing fuel efficiency for 12 OECD countries over the period 1973to 1992: US, UK, Japan, Australia, Germany, France, Italy, The Netherlands, Sweden,Denmark, Norway and Finland. Their fuel use data are disaggregated in such a way that itallows them to conduct separate estimations for vehicle stock, mean fuel intensity, andmean annual driving distance. Using a variety of different estimation techniques andmodels, the authors use their results to obtain estimates for long run car fuel and traveldemand.

The results confirm the importance of increasing fuel efficiency in gasoline demand.They calculate a long run fuel price elasticity of approximately -0.7, in which the largestfraction, just under 60%, is due to changes in fuel intensity. The gasoline demand figureis more than double the estimated price elasticity of travel demand. The long run incomeelasticity of fuel demand is approximately 1.2, almost all due to the number of cars, andis of identical magnitude with respect to travel demand. The fuel efficiency effect isfound to arise both through increased technical efficiency and the imposition ofenvironmental standards.

Page 15: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

15

Johansson and Shipper also consider the effects of different taxation measures on fueland travel demand. They find that a fuel tax increase will reduce overall long run fuelconsumption much more than an increase in the other car related taxes, for example,taxing car ownership.

The focus on fuel efficiency in gasoline demand studies, although yielding some quitedifferent results, does indicate that increasing efficiency is crucial in explaining the longrun price elasticity. Most studies show a slow rate of adaptation, but none the less, astrong and identifiable effect. An important and consistent implication of these studies isthat the impact of fuel price changes has a greater impact on fuel demand and vehicleemissions than on vehicle use and congestion, particularly in the long run.

Non-stationary data and the cointegration technique

The appropriateness of different data types (cross-section, time-series, pooled) and themethodologies applied to each has proved a source of constant debate in gasoline demandresearch. Many recent studies have expressed concern over the customary treatment oftime-series data and particularly the lack of recognition of the non-stationary nature ofthese data. This has given rise to the widespread use of cointegration techniques that seekto model the non-stationary nature of time-series data explicitly. The use of this methodis employed both as a means of distinguishing the short from the long run gasolinedemand characteristics, and for calculating the speed of adjustment towards the long runvalues. The results obtained in this way often give estimates that are outside the rangereported in the major reviews.

If the dependent and independent variables are trending variables the time series data issaid to be non-stationary, and if there is a long term relationship between them then theyare cointegrated. Then the mean and variance of the time-series are non-constant overtime and the value of the process at any point depends on the time period itself. Thecointegration technique is designed to distinguish the long run relationship, the manner inwhich the two variables drift together, from the short run effect, the relationship betweendeviations of the dependent variable from its long run trend and deviations of theindependent variables from their long run trends.

The cointegration method typically follows three basic steps. First, the time series underconsideration are examined to determine if the variables are non-stationary. Second, if thevariables are found to be non-stationary the cointegration of the variables is investigated.If the variables do indeed possess a long run relationship the long run elasticities may beestimated from the cointegrated regression. Third, the short run elasticities and the rate ofadjustment towards the long run equilibrium can be estimated by means of an ErrorCorrection Model (ECM).

Bentzen (1994) estimates short and long run elasticities of gasoline demand for Denmarkusing annual time-series data for the economy covering the period 1948-1991. The model

Page 16: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

16

estimated explains gasoline consumption per capita by the price of fuel, vehicle stock percapita, and increasing fuel efficiency which is represented by a time trend.

The author finds a stable long run relationship between the variables in his model andgoes on to estimate the error correction model to distinguish short and long run effects.The estimated short run price elasticity is –0.32 and the long run, –0.41. The short runvehicle per capita income elasticity is 0.89 and the long run 1.04.

The short run price elasticity estimated by Bentzen is of similar magnitude to valuesreported in other studies, the long run value, however, is somewhat lower. Besidesdifferences in data and models, the author believes that the lower value can be at leastpartly explained by the particular statistical technique used, with explicit treatment of thenon-stationary properties of the variables.

Samimi (1995) uses cointegration techniques to examine the short and long runcharacteristics of energy demand in Australia’s road transport sector. He has quarterlydata for the Australian road transport sector from 1980 to 1993. The model estimated hasa lag endogenous structure. The dependent variable is road transport energy demand,which includes gasoline and diesel oil. The independent variables are fuel prices, the lagof road transport energy demand, and road transport output, which is measured as therevenue generated by carrying goods and passengers for hire and reward and provision ofother road transport services.

The cointegration estimates yield price elasticity estimates of –0.02 in the short run and –0.12 in the long run. The estimated income elasticities are 0.25 in the short run and 0.48in the long run.

Samimi notes that the long run income and price elasticities for Australia are of muchlower magnitude than found previously. The author explains the difference in the longrun price effect by hypothesising that more efficient vehicle technology is built into hislong run estimate. But he also argues that use of different time periods or differenteconometric specifications would yield different estimates, mainly due to changes inmarket structure. On this basis the author questions the existence of stable priceelasticities.

Eltony and Al-Mutairi (1995) estimate the demand for gasoline in Kuwait for the period1970-1989 using a cointegration and error correction model. The model they estimate,which is identical to that of Bentzen (1994), explains per capita gasoline consumption inKuwait by the real price of gasoline and real per capita income. Their cointegrated resultsshow a short run price elasticity estimate of -0.37 and a long run price elasticity of –0.46.The estimated short and long run income elasticities are 0.47 and 0.92 respectively.Again the long-run price elasticities are outside the range typically reported in theliterature

Gasoline demand in India is examined by Ramanathan (1999) using a cointegrationmethodology to analyse long and short run behaviour. The model estimated in the paper

Page 17: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

17

explains national per capita gasoline consumption (in tonnes) as a function of real percapita GDP and the price of gasoline. Time-series data are used for estimation coveringthe period 1972-1973 to 1993-1994.

The author’s results for India estimate a short run price elasticity of gasoline demand of -0.21 and a short run income elasticity of 1.18. The cointegration model indicates that theadjustment of gasoline consumption towards its long run equilibrium occurs at arelatively slow rate with 28% of the adjustment occurring within the first year. The longrun price elasticity of demand estimated is -0.32 and the long run income elasticityestimate is 2.68.

Ramanathan thus derives a very high long run income elasticity and a rather inelasticprice effect. The author believes that the low level of gasoline consumption in India andthe gradual increasing economic growth can explain the differences between his resultsand those obtained elsewhere. He concludes that over-pricing of gasoline as a policyinstrument is unlikely to have an influential effect on gasoline demand in India.

The cointegration studies of time series data estimate long run price elasticities that areoften substantially lower than those reported in the major reviews. Researchers adoptingthis particular technique frequently state that this is due to the application of a moreappropriate treatment of the non-stationary nature of time series data. However, thegenerality of these results is still open to question because it is not clear why the use of along time series, regardless of treatment, yields lower price elasticity estimates. Certainly,as is illustrated in the next section, there may be reason to believe that price elasticitieshave grown over time at least partly as a result of increased fuel efficiency, a factor thatoften has received insufficient attention in many of the cointegration studies.

Meta-analysis of gasoline demand elasticities

Espey (1998) carries out ‘meta-analyses’ of international gasoline demand elasticities toexplain the variation in the magnitude of estimated price and income effects. This workforms a particularly important and novel contribution to the literature because it examinesempirically why variation in estimates exists. Thus while the major reviews identify thevariation, Espey’s work seeks to explain it. The paper extends and up-dates earlier workwhich focused on variation in elasticity estimates of gasoline demand for the UnitedStates alone (Espey, 1996a).

Espey’s study is based on an extensive review of articles published between 1966 and1997 which gave 277 estimates of long run price elasticity, 245 estimates of long runincome elasticities, 363 estimates of short run price elasticity, and 345 estimates of theshort run income elasticity. The author’s analysis provides four models that seek toexplain separately variation in the short and long run income and price elasticities. Thebasic hypothesis is that variation in elasticity estimates can be explained by demandspecification, data characteristics, ‘environmental’ characteristics (i.e. the level of thedata, the setting, time span analysed etc.) and the estimation method.

Page 18: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

18

Espey’s results indicate that elasticity estimates are sensitive to a number of differentaspects of model structure. In terms of price effects, the inclusion of vehicle ownershipand fuel efficiency variables serves to lower estimates of the short, but not the long run,price elasticity. Static models tend to produce larger short run price elasticities and lowerlong run price elasticities, indicating that perhaps these models produce intermediate-runelasticities. No differences are found for price elasticities across different dynamicspecifications, nor are differences in long run price elasticity estimates found among timeseries, cross-sectional, and cross-sectional-time series studies. The paper does show,however, that the short run price elasticity has tended to decrease over time while thelong-run elasticity has tended to grow. The author believes this temporal effect is due toincreased fuel efficiency. “As prices rose during the 1970s and people made some initialadjustments in driving habits and bought more fuel efficient vehicles, there were feweroptions for further short run responses to price changes. However, as automobile fuelefficiency improved during the late 1970s and early to mid-1980s, long run responses tofuel price changes were larger than before 1974.” (Espey, 1998, 290)

As regards income effects Espey’s analysis finds that the inclusion of vehicle ownershipand vehicle characteristics substantially influences results. Models that include somemeasure of vehicle ownership estimate significantly lower short and long run incomeelasticities. No statistically significant differences are found for long run estimatesbetween static and dynamic models, or between different dynamic specifications. Nor areany differences found for long run estimates in studies based on cross-sectional, time-series, or cross-sectional-time series data. Finally, the author finds that the short runincome elasticity has remained fairly constant over time while there is evidence to showthat the long run elasticity may be declining.

The author concludes that the exclusion of vehicle ownership in demand models wouldbe expected to bias results, particularly short run effects. The finding that elasticityestimates are changing over time prompts Espey to warn against using elasticity estimatesfrom the 1970s or even 1980s to extrapolate into the future. But the author also arguesthat in many ways price elasticity estimates are relatively robust, having a fair degree ofconsistency across data types and across functional forms and estimation techniques.

Conclusions

On one level, our survey shows that there are a range of different views about themagnitude of price elasticity effects on gasoline consumption and private travel demand.Figure 1 below illustrates differences in magnitude, showing estimates of long and shortrun price elasticities of gasoline consumption from various studies. These estimates varygreatly both between and within geographical areas of study for long and short runelasticities. For instance, long run price elasticity estimates range from –0.23 in the US to–1.35 in the OECD countries, and within the US itself from –0.23 to –0.8, and within theOECD from –0.75 to –1.35. Short run price elasticities range from –0.2 to –0.5

Page 19: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

19

Sources: Houthakker et al. (1974), Sweeney (1978), Baltagi and Griffin (1983), Drollas(1984), Sterner (1990), Goodwin (1992), Sterner et al. (1992), Sterner and Dahl (1992),and Eltony (1993).

The Figure illustrates the important influences that particular data and methods ofestimation can have on the results obtained. Whether the data used for estimation arecross-section, time series, or pooled, has an influence on the magnitude of the estimatesobtained. For this reason, discussion of individual gasoline price elasticity estimates hasto be based on a clear understanding of the method used and of the empirical context forestimation.

But while the use of specific data or methodological approaches can create crucialdifferences in the magnitude of elasticity estimates, the overwhelming evidence from oursurvey suggests that long run price elasticities will typically tend to fall in the –0.6 to –0.8 range. This order of magnitude is indicated by those papers we have reviewed that arethemselves extensive surveys and which have considered hundreds of individualestimates across a range of empirical contexts (Drollas, 1984, Sterner, 1990, Goodwin,1992, Sterner and Dahl, 1992). In many cases authors explicitly claim to find similaritiesand not differences between countries in the size of long run price elasticities. Individualstudies, which apply a variety of different estimation techniques to the same data (Baltagi

Figure 1: Gasoline price elasticities.

US

US

US

OECD

OECD

OECD

OECD

OECD

UK

France

Austria

Germany

Canada

Various Countries

Various Countries

-1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0

Short-ru n Long Ru n

Page 20: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

20

and Griffin, 1983, Eltony, 1990) also produce long run estimates within the same range.These same studies show that short run price elasticities normally range from –0.2 to –0.3. In other words they tend to be around 30% to 40% as large as the long run effects.Again, this is fairly consistent across different empirical environments.

Thus, concentrating on evidence that has proved to be consistent across studies, we candraw out three central conclusions from our survey of the literature and highlight some oftheir implications.

i. There are differences between the short and long run elasticities of fuelconsumption with respect to price. Typically, short term elasticities are in theregion of –0.3 and long term of between –0.6 and –0.8. Therefore, it may be rightto say that “it won’t make much difference” or “people will use their cars just thesame”, but only in the short run. The evidence is clear - and remarkably consistentover a wide range of studies in many countries - that in the long run there is asignificant response, albeit a less than proportionate one.

ii. Both long and short term effects of gasoline prices on traffic levels tend to be lessthan their effects on the volume of fuel burned. The short term elasticity of trafficwith respect to price is about –0.15 and long term about –0.30. So motorists dofind ways of economising on their use of fuel, given time to adjust. Raising fuelprices will therefore be more effective in reducing the quantity of fuel used thanin reducing the volume of traffic.

iii. The demand for owning cars in heavily dependent on income. The long runincome elasticity of fuel demand is typically found to fall in the range 1.1 to 1.3.Short run income elasticities are between just less than a third and just over a halfthat size in magnitude: elasticities normally estimated in the range 0.35 to 0.55.

Our overall assessment is that, given the wide variety of the countries providing evidenceand the several methods used to analyse the data, there is a reassuring consistency in theindications of the magnitude of the four critical parameters of interest. These offer soundcomponents in the design of national transport policy. As fuel prices rise, abstractingfrom other changes, fuel consumption will fall by a less than proportionate amount.Traffic will also fall, but it is less responsive than fuel consumption. As economic activityand real incomes increase, abstracting from other changes, both traffic and fuelconsumption will increase by a slightly greater proportion. In all cases it takes time forpeople to adjust, so the initial impact effects are smaller than effects in the long term.Improvements in the fuel efficiencies of vehicles have significantly changed fuelconsumed and the fuel costs of travelling a given distance.

The implications for policy are that fuel taxation can play a significant part in moderatingtraffic growth, fuel consumption and the volume of emissions, especially over the longterm. However, as several of the authors surveyed point out, an implication of the relativemagnitudes is that real fuel prices would have to rise faster than real incomes in order tooffset their effect. On several occasions governments have found it impossible to sustain

Page 21: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

21

that. If this continues we will inevitably be faced with having to deal with theconsequences of more traffic and, possibly a higher volume of emissions. Improvedvehicle design and promotion of different fuels will offer a way of mitigating some of thedamage from the emissions. Preventing growth in traffic looks much harder and unlesssome alternative and realistic technique can be implemented the choice is betweentolerating increasing congestion and providing more highway capacity.

References

Archibald R and Gillingham R (1980) ‘An analysis of the short-run consumer demand forgasoline using household survey data’, Review of Economics and Statistics, 62, 622-628.

Baltagi B and Griffin J (1983) ‘Gasoline demand in the OECD: an application of poolingand testing procedures’, European Economic Review, 22, 117-137.

Bentzen J (1994) ‘An empirical analysis of gasoline demand in Denmark usingcointegration techniques’, Energy Economics, 16, 139-143.

Blum U, Foos G and Guadry M (1988) ‘Aggregate time series gasoline demand models:review of the literature and new evidence for West Germany’, Transportation ResearchA, 22A, 75-88.

Bohi D (1981) Analysing Demand Behaviour: A Study of Energy Elasticities, Publishedfor Resources for the Future by John Hopkins Press, Baltimore, MD.

Bohi D and Zimmerman M (1984) ‘An update on econometric studies of energydemand’, Annual Review of Energy, 9, 105-154.

Crawford I and Smith S (1995) ‘Fiscal instruments for air pollution abatement in roadtransport’, Journal of Transport Economics and Policy, 29, 33-51.

Dahl C (1986) ‘Gasoline demand surveys’, The Energy Journal, 7, 67-82.

Dahl C (1995) ‘Demand for transportation fuels: a survey of demand elasticities and theircomponents’, The Journal of Energy Literature, 1, 3-27.

Dahl C and Sterner T (1991a) ‘Analysing gasoline demand elasticities: a survey’, EnergyEconomics, 13, 203-210.

Dahl C and Sterner T (1991b) ‘A survey of econometric gasoline demand elasticities’,International Journal of Energy Systems, 11, 53-76.

Dargay J and Vythoulkas P (1998) ‘Estimation of dynamic transport demand modelsusing pseudo-panel data’, 8th World Conference on Transport Research, Antwerp,Belgium, 12-17 July 1998.

Page 22: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

22

Dargay J and Vythoulkas P (1999) ‘Estimation of a dynamic car ownership model: apseudo-panel approach’, Journal of Transport Economics and Policy, 33, 287-302.

Deaton A (1985) ‘Panel data from time series of cross-sections’, Journal ofEconometrics, 30, 109-126.

Delucchi M (2000) ‘Environmental externalities of motor vehicle use’, Journal ofTransport Economics and Policy, 34, 135-168.

DETR (1997) National Road Traffic Forecasts (Great Britain) 1997, London: HMSO.

Dix M and Goodwin P (1982) ‘Petrol prices and car use: a synthesis of conflictingevidence’, Transport Policy and Decision Making, 2 (2).

Drollas L (1984) ‘The demand for gasoline: further evidence’, Energy Economics, 6, 71-82.

Eltony M (1993) ‘Transport gasoline demand in Canada’ Journal of Transport Economicsand Policy, 27, 193 - 208.

Eltony M and Al-Mutairi N (1995) ‘Demand for gasoline in Kuwait: an empiricalanalysis using cointegration techniques’, Energy Economics, 17, 249-253.

Espey M (1996a) ‘Explaining the variation in elasticity estimates of gasoline demand inthe United States: a meta-analysis’, The Energy Journal, 17, 49-60.

Espey M (1996b) ‘Watching the fuel gauge: an international model of automobile fueleconomy’, Energy Economics, 18, 93-106.

Espey M (1998) ‘Gasoline demand revisited: an international meta-analysis ofelasticities’, Energy Economics, 20, 273-295.

Eyre N, Ozdemiroglu E, Pearce D, and Steele P (1997) 'Fuel and location effects on thedamage costs of transport emissions', Journal of Transport Economics and Policy, 31, 5-24.

Foos G (1986) ‘Die determinanten der verkehrnachfrage’, Karlsruher Beiträge zurWirtschaftspolik und Wirschaftsforschung, 12, Loper Verlag: Karlsruhe.

Glaister S and Graham D (1999) ‘The incidence on motorists of petrol price increases inthe UK’, Mimeo, Imperial College, 1999.

Goodwin P (1992) ‘A review of new demand elasticities with special reference to shortand long run effects of price changes’ Journal of Transport Economics and Policy, 26,155-163.

Page 23: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

23

Greene D and Hu P (1986) ‘A functional form analysis of the short-run demand for traveland gasoline by one-vehicle households.’ Transportation Research Record 1092.Transportation Research Board, National Research Council, Washington D.C., 10-15.

Greening L, Jeng H, Formby J, and Cheng D (1995) ‘Use of region, life-cycle and rolevariables in the short-run estimation of the demand for gasoline and miles travelled’,Applied Economics, 27, 643-656.

Hall J (1995) 'The role of transport control measures in jointly reducing congestion andair pollution', Journal of Transport Economics and Policy, 29, 93-103.

Houthakker H, Verleger P and Sheehan D (1974) ‘Dynamic demand analysis for gasolineand residential electricity’, American Journal of Agricultural Economics, 56, 412-418.

Johansson O and Schipper L (1997) ‘Measuring the long run fuel demand of cars:separate estimations of vehicle stock, mean fuel intensity, and mean annual drivingdistance’, Journal of Transport Economics and Policy, 31, 277-292.

Koopman G (1995) 'Policies to reduce CO2 emissions from cars in Europe: a partialequilibrium analysis', Journal of Transport Economics and Policy, 29, 53-70.

Kouris G (1983) ‘Energy demand elasticities in industrialised countries: a survey’, TheEnergy Journal, 4, 73-94.

McCubbin D and Delucchi M (1999) 'The health costs of motor vehicle-related airpollution', Journal of Transport Economics and Policy, 33, 253-286.

McKay S, Pearson M and Smith S (1990) ‘Fiscal instruments in environmental policy’,Fiscal Studies, 11, 1-20.

Mehta J, Narasimham G and Swamy P (1978) ‘Estimation of a dynamic demand functionfor gasoline with different schemes of parameter estimation’, Journal of Econometrics, 7,263-269.

Orasch W and Wirl (1997) ‘Technological efficiency and the demand for energy (roadtransport)’, Energy Policy, 25, 1129-1136.

Oum T (1989) ‘Alternative demand models and their elasticity estimates’, Journal ofTransport Economics and Policy, 23, 163-187.

Puller S and Greening L (1999) ‘Household adjustment to gasoline price change: ananalysis using 9 years of US survey data’, Energy Economics, 21, 37-52.

Ramanathan, R (1999) ‘Short and long run elasticities of gasoline demand in India: anempirical analysis using cointegration techniques’, Energy Economics, 21, 321-330.

Page 24: Elasticity of Gaselone Case Study

Forthcoming in Journal of Transport Economics and Policy

24

Ramsey J, Rasche R and Allen B (1975) ‘An analysis of the private and commercialdemand for gasoline’, Review of Economics and Statistics, 57, 502-507.

Rouwendal J (1996) ‘An economic analysis of fuel use per kilometre by private cars’,Journal of Transport Economics and Policy, 30, 3-14.

Samimi R (1995) ‘Road transport energy demand in Australia: a cointegrated approach’,Energy Economics, 17, 329-339.

Small K and Kazimi C (1995) ‘On the costs of air pollution from motor vehicles’, Journalof Transport Economics and Policy, 29, 7-32.

Sterner T (1990) The Pricing of and Demand for Gasoline, Swedish Transport ResearchBoard: Stockholm.

Sterner T and Dahl C (1992) ‘Modelling transport fuel demand’, in T Sterner (ed)International Energy economics, Chapman and Hall, London, 65-79.

Sterner T, Dahl C and Franzén M (1992) ‘Gasoline tax policy: carbon emissions and theglobal environment’, Journal of Transport Economics and Policy, 26, 109-119.

Sweeney J (1978) ‘The demand for gasoline in the United States: a vintage capital model’in Workshops on energy supply and demand (International Energy Agency, Paris) 240-277.

Taylor LD (1977) ‘The demand for energy: a survey of price and income elasticities’, inInternational Studies of the Demand for Energy, (ed) W Nordhaus, North Holland,Amsterdam.

Walls M, Krupnick A and Hood C (1993) ‘Estimating the demand for vehicle milestravelled using household survey data: results from the 1990 National PersonalTransportation Survey, Resources for the Future Discussion Paper ENR 93-25,Washington D.C.