Journal of Air Transport Management 8 (2002) 165–175
Price elasticities of demand for passenger air travel: a meta-analysis
Martijn Brons, Eric Pels, Peter Nijkamp*, Piet Rietveld
Department of Spatial Economics, Tinbergen Institute, Free University Amsterdam, De Boelelaan 1105, 1081 HV Amsterdam, Netherlands
The demand for air transport is largely determined by the spending capacity of customers. This paper aims to offer more insight
into the determinants of price elasticities in the aviation sector. It seeks to identify both common and contrasting factors that
inﬂuence the price elasticities, on the basis of a comparative analysis among a large number of empirical studies in this ﬁeld. By
means of meta-analytical methods, the relative importance of several driving forces such as distance, type of ticket and the nature of
study is investigated. r 2002 Elsevier Science Ltd. All rights reserved.
1. Introduction The estimation of price elasticities in aviation can,
however, be rather difﬁcult, given the various problems
The continuing growth in the number of passengers concerning data availability on prices, number of
and aircraft movements necessitates a rise in investments passengers, etc. As an alternative, one can use research
in airport and aircraft capacity. But even with these new synthesis from various empirical studies undertaken
investments, peak congestion and the environmental elsewhere or in the past. Using existing research, one
impact of aviation remain problematic. Air transport is tries to ﬁnd common factors explaining potential
apparently a ﬁeld fraught with externalities. Another differences in, e.g., estimates of price elasticity. This
development in the aviation sector is the tendency to approach will be followed in the present paper.
form alliances. Although the literature shows that these
alliances may be beneﬁcial to passengers, they still need,
in one way or another, consent from aviation autho-
rities. In the deregulated aviation sector, aviation 2. Determinants of demand for passenger air transport
authorities therefore play a critical role in protecting
the population from excessive noise and in safeguarding In this section, a number of economic, demographic
the consumer against excessive usage of market power. and geographic determinants with respect to the demand
One potential instrument the government has at its for passenger air transport and its related price elasticity
disposal is the price. For example, the government can of this demand will be discussed.
put a price on the externality to reduce the negative
effect, but if the passengers are not very sensitive to price 2.1. Choice contexts in air transport demand
changes, this policy will have little effect; the airlines
simply pass the charge on to the passengers. The The price elasticity of the demand for a good, whether
government needs information on the price sensitivity a consumer good or a production input factor, is
of passengers in order to be able to estimate the likely directly related to the possibilities of substitution for
policy effects or to justify noise annoyance policy. This that good. A relatively large number of substitutes will
information is needed on different levels. A kerosine tax, imply a high price elasticity, whereas a lack of
for example, can only be justiﬁed in the context of an substitutes will likely force demand to become more
international policy arrangement and requires different rigid so that the demand for this commodity may
insights than a local noise charge. become inelastic. In the speciﬁc case of the demand for
passenger air transport, the structure and critical factors
of the demand are likely not different. Most of the
*Corresponding author. determinants of price elasticity discussed in this paper
E-mail address: firstname.lastname@example.org (P. Nijkamp). do not directly inﬂuence price elasticity, but rather affect
0969-6997/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 6 9 - 6 9 9 7 ( 0 1 ) 0 0 0 5 0 - 3
166 M. Brons et al. / Journal of Air Transport Management 8 (2002) 165–175
the level of supply of substitute modes and thus exert destinations with (nearly) the same utility results or for
their inﬂuence in various indirect ways. destinations within the same price class but which are
An important issue in aviation is that multiple levels of providing a higher level of utility. Whether or not the
substitution can be distinguished, as Fig. 1 shows. First, consumer will succeed in ﬁnding another, more favour-
different carriers may compete with each other on the able alternative after a price increase also depends on
same route, providing a case of intra-modal substitution. the degree to which the individual characteristics of the
In the case of homogeneous transport services, there will original destination can be substituted for. For instance,
be perfect competition implying a very high price an appointment with a business partner in New York
elasticity. However, when services of varying quality are cannot as easily be substituted for as a leisure trip to a
offered, the substitutability will be less. generic Mediterranean seaside resort, due to several
Next, on certain market segments, alternative trans- non-replaceable characteristics.
port modes may provide sufﬁciently similar qualities to Finally, non-transport goods may be regarded as
be regarded as substitute modes. Numerous factors, substitutes for transport goods since again budget
primarily of geographic, economic and demographic constraints, either a person’s disposable income or a
nature, determine the availability and the potential ﬁrm’s operational budget, force them into mutual
success of alternative modes as a substitute. It is obvious competition. This issue is closely related to the previously
that geographic components, such as seas, impenetrable mentioned phenomenon. It appears that immobility may
mountain ranges or even the mere distance of a trip, be regarded a substitute for transport as long as the
may complicate the presence or establishment of a given utility level derived from alternatively spending the
supply of adequate substitute modes. budget savings are at least equal to the utility provided
On the third place, different destinations with by the default consumption of transport. Note that these
sufﬁciently similar characteristics may assume the role trade-off effects are generally subject to mechanisms of
of substitutes for each other. A useful way to look at this diminishing rates of marginal return.
phenomenon is to adopt a hedonic point of view. The We conclude that prices play a role in various choice
speciﬁc characteristics of a destination may be regarded contexts. When the choice alternatives are very homo-
as attributes, each of which contributes to some degree geneous (case 1) we may expect that price elasticities are
to the perceived overall utility of the destination. The large whereas when the choice alternatives are very
rational consumer is, subject to given budget considera- different the price elasticities are more moderate.
tions, assumed to try to maximise overall utility, i.e. to
choose a destination that yields the highest level of 2.2. Economic determinants of demand and price sensitivity
utility. If now the relative fare level for this destination
increases, the consumer may reconsider his expenditure Various economic, demographic and geographic
choices and may start a search for less expensive factors can be identiﬁed which, according to their
Travel expenditure Non-travel expenditure
Travel expenditure 2 Travel expenditure
destination A destination B
Travel expenditure 3 Travel expenditure
Possibilities of substitution
destination A by destination A by
car 1 = Non-travel substitution
2 = Destination substitution
3 = Mode substitution
Travel expenditure 4 Travel expenditure 4 = Intra-mode substitution
destination A with destination A with
airline X airline Y
Fig. 1. Illustrative case of substitution in the aviation market.
M. Brons et al. / Journal of Air Transport Management 8 (2002) 165–175 167
formal theoretical speciﬁcation, may affect price elasti- towards a positive relationship between distance and
city estimate values. From Section 2.1, we may argue price sensitivity. Thus, the relationship between ﬂight
that price elasticities of demand and substitution distance and price elasticity of demand for air travel
possibilities are strongly related concepts from an appears to depend on a number of counteracting forces.
economic point of view. It is interesting then to observe It is not clear beforehand which effect will generally
that for several determinants, the substitute environ- prevail.
ment assumes an important role with respect to the ways In addition to mere travel distance, the geographic
these factors exert their inﬂuence on elasticity estimate location or continent in which a price elasticity is
values. estimated may be an important factor in determining the
If the initial demand for air transport would be value of these estimates. Cultural differences, relative
weakly related to the income level, the expenditures on abundance of ground transport networks or income
air transport would form a smaller share of the related factors can all play a role in the extent to which
disposable income for consumers with a higher dis- fare price changes affect demand levels for air transport.
posable income level than for consumers with less A distinction between business passengers and leisure
money to spend. Assuming decreasing marginal utility passengers can be based on two criteria resulting in the
returns on income, this would mean that utility loss as a institutionalised formal distinction between business
result of a fare increase will be lower for consumers with class travellers and economy class travellers and the
a higher disposable income. As a result, consumers with actual distinction according to travel motive. These
a higher disposable income would be less price sensitive, distinctions are generally assumed to be in mutual
implying a negative relation between income level and accordance, although this is not necessarily always true.
the magnitude of the price elasticity of demand for air Moreover, since passenger data on fare classes are
transport. generally more easily available than data on travel
However, higher incomes are generally associated motives, the latter often being the more relevant
with relatively higher demand for air transport (Mutti construct of demand theory, theory and empirics on
and Murai, 1977). Air travel is often a luxury good with the subject of demand sensitivity are not always
demand elasticities of income greater than unity, so that perfectly in consonance. However, as long as this is
income level and the share of air transport demand of accounted for, fare class data may well be used as a
disposable income are expected to be positively corre- reasonable proxy for travel motives.
lated (Crouch, 1992). If indeed the share of air transport Leisure travellers, essentially consumers, aim to
demand is higher for consumers with higher income maximise the utilityFor satisfactionFderived from
levels, this would suggest that despite a decreasing air travel and from the associated consumption of
marginal utility of income, the utility losses associated holiday experiences, subject to a given income or budget
with a fare increase are higher for this group of constraint. Characteristics of leisure travel demand are
consumers, which would imply that they may be more determinants such as travel costs, relative price of other
price sensitive than consumers with lower incomes. goods, income and socioeconomic characteristics. Busi-
As has previously been discussed, possibilities for ness travellers, who use business as an input to ﬁnal
transport substitution are likely to be directly related to production, are in general interested in minimising costs
the distance of a ﬂight. Long-distance ﬂights will for a given level of output. Business travel demand is
generally suffer from a smaller amount of substitute determined by factors such as travel costs, relative price
modes than short-distance ﬂights, particularly when of complementary production input factors and a ﬁrm’s
intercontinental ocean crossing ﬂights are concerned. output level. Because of this, leisure and business
The existence of such cases implies an inverse relation- travellers are likely to respond differently to changes
ship between distance and price sensitivity. On the other in certain socioeconomic factors inﬂuencing the de-
hand, long-distance ﬂights are usually more expensive to mand, and should therefore be modelled separately
begin with than short-distance ﬂights,1 so that an (Hooper, 1993).
increase in costs will require a larger share of a In general, demand for business travel tends to be less
passenger’s budget. This second aspect seems to point sensitive to changes in air fare than demand for leisure
travel. Leisure travel is generally regarded as discre-
In today’s complicated aviation market, distance and fare price are
tionary expenditure. Many goods and services compete
not as closely related as one would expect. Landing rights, alliances, with leisure travel for obtaining a share of the
tariff classes and other market factors often dominate the cost consumer’s discretionary budget. Thus, even while the
structure of fares and distort such a close relationship (see, for number of perfect substitutes for air travel may not be
example, Marsan and Kostoris, 1993; Cooper and Maynard, 1971).
overwhelming as has been pointed, leisure travel,
Ample examples of not only decreasing marginal prices of ﬂight
distance but even decreasing average prices of ﬂight distance suggest compared to business travel, has some additional
that negative relations between distance and fare should not substitutes inside as well as outside of the transport
necessarily be considered as anomaly. sector and therefore tends to be more sensitive to
168 M. Brons et al. / Journal of Air Transport Management 8 (2002) 165–175
changes in airfares, implying a higher absolute price to mind. However, complex time-dependent behavioural
elasticity. Moreover, the absolute elasticity of business patterns may induce sufﬁcient distortion effects to
demand with respect to airfares is likely to be lower than prevent us from simply adopting this long-run adjust-
that for leisure travellers because of two other factors. ment rule as a rule of thumb. First of all, sudden cost
First, the total cost of travel includes a value of time changes may easily lead to an exaggerated behavioural
component. Since business travellers generally value response in the short term, which may subsequently be
time higher than leisure travellers, airfares seem to form perceived erratic and corrected for in the long run. All
a smaller part of their total travel costs. This would other things being equal, this latter aspect may point to
mean that an increase in airfares leads to a smaller rise an inverse relationship between time horizon and price
in total travel costs for business passengers, compared to sensitivity which is quite the opposite from the predic-
leisure travellers, and therefore a lower willingness to tion of the former aspect. It seems as if the ability and
substitute monetary for time-saving advantages. It is the willingness to adjust the demand show divergent
important, though, to point at the initial difference in paths. The relation between price elasticity of demand
fare levels between business class and economy class for air transport and time horizon seems to be rather
trips. Because of the fact that business class fares are complex and to depend on various partial effects.
generally much more expensive, the airfares still form a An important issue here is the fact that the degree to
large part of the total travel costs for business class which either one of these effects prevails under the given
passengers and the monetary disadvantages of an circumstances may be directly related to the degree to
increase in fares will be higher than those for leisure which substitution is available. For the speciﬁc case of
travellers. The willingness to substitute monetary for the demand for passenger air transport, a number of
time advantages may thus be higher than expected for aspects are relevant in this context. First, there is a lack
business class passengers. The difference in price of sufﬁcient substitution transport modes for the air
elasticity values between business class and economy transport sector. The mere travel speed of the mode is as
class passengers therefore seems to depend on the yet still unmatched, whereas intra-modal substitution
difference between the ratio of the business class can hardly be regarded as a cost-evasive substitution due
versus the economy class passengers’ time valuation to differential fare structures and competition-related
and the ratio of business class versus economy class characteristics inherent in the air transport sector.
initial fares. Secondly, there are factors complicating possible long-
Secondly, business travellers are more likely to be run adjustment strategies. Relocation costs, both
concerned with maximising their productivity while pecuniary and non-pecuniary, as a means to evade
travelling. Therefore, they are most likely willing to increased transport costs tend to be relatively high due
pay for a higher quality service that allows last-minute to the average ﬂight distance and the often trans-cultural
bookings and changes to travel plans, and provides nature of the ﬂight.
better check-in and on-board facilities. Additionally, These aspects may imply that as far as possibilities for
any rises in airfares tend to be absorbed by the ﬁrm adjustment strategies are concerned, the use of long-run
rather than the individual traveller. time horizons may not as obviously result in higher price
sensitivity estimates compared to short-run time hor-
2.3. Study descriptive determinants of price elasticities izons as expected, while at the same time expressions of
short-run price sensitivity will to a large degree be
In addition to the economic and geographic determi- limited to demand changes on an aggregate modal level.
nants, a number of moderator variables can be The overall relationship between time horizon and price
distinguished that may exert inﬂuence on the precise sensitivity is, among other factors, determined by the
level of elasticity estimates from a certain study. Also ratio of leisure travellers to business travellers, since the
here, plausible relationships between substitution possi- former group generally disposes of more possibilities for
bilities and some moderator variables do exist, in demand adjustment on an aggregated level of the mode
particular, the time horizon that is used in the original than the latter.
case study. Most empirical research on airline demand has used
The distinction that exists between short-run and cross-section data, notably a sample of city-pair data.
long-run elasticity estimates is an important one. This offers the advantage of larger samples that are
Generally speaking, in the long run, consumers and often available in time series analysis, which is essential
ﬁrms are better able to adjust to price signals than in the for studying dimensions of consumers’ travel demand
short run, implying that the long-run demand tends to such as time valuation and quality of service or for
be more elastic than the short-run demand (Oum et al., studying the modal choice behaviour of consumers.
1992). In the case of demand for transport, decisions However, the disadvantage is that it does not always
with respect to geographic relocation and asset holding permit accurate estimation of price and income elasti-
as long-run responses to increased transport costs come cities since cross-section data generally exhibit relatively
M. Brons et al. / Journal of Air Transport Management 8 (2002) 165–175 169
little variation in air fares per unit of distance within a elasticities in order to be able to draw some general or
given fare class (Straszheim, 1978). transferable conclusions.
Time series data are more useful in estimating price
and income elasticities, primarily since price (and
income) changes have been dramatic during the last 3.1. Data collection and variables used
decades. But price changes, however signiﬁcant they
have been, are also relatively infrequent due to govern- Since the objective of our meta-analysis basically
ment regulation, whereas changes in service variables consists of a comparative reevaluation of previous
such as schedule frequencies, speed of aircraft, and research on price elasticites for passenger air transport,
density of seating also occur. we collected a number of 37 studies in which one or
Multicollinearity pervades clearly both cross-section more price elasticities for passenger air transport were
and time series estimates. In cross-section models with estimated, leading to a total number of 204 observa-
gravity variables, fares tend to be strongly correlated tions. An Appendix lists the studies examined. From
with distance variables. This problem is most severe in each study the elasticity estimates as well as information
time series studies. Variables such as price and income on certain geographic, economic and demographic
tend to be tightly correlated with a time trend. sample variables and moderator variables were gathered
Parameter estimates are, therefore, generally sensitive and entered into a database.
to changes in model speciﬁcation and sample In our analysis, we treat the price elasticity estimates
coverage. Moreover, the multicollinearity, coupled with as the dependent variable whereas the study descriptors
data limitations, leads to a persistent tendency to and the other variables are included for their respective
underspecify or oversimplify the model, with con- explanatory power with respect to the variance among
sequent biases in regression coefﬁcients (Jung and Fuji, price elasticity estimates. In doing so, possibilities for
1976). mutual interdependencies among the explanatory vari-
ables will also be considered. For an overview of the
variables employed we refer to Fig. 2a.
The geo-economic variables used for explaining the
3. Meta-analysis variance among price elasticity estimates are transfer
distance; fare class; geographic location. Moderator
In this section, the evidence offered by empirical case variables employed here are research method; time
studies on the price elasticities of demand for passenger horizon used and period of data collection.
air travel will be examined by means of a meta-analysis. The majority of the data are all directly or indirectly
The objective is to test whether price elasticity estimates gathered from the original studies. Most of the
encountered in the literature are statistically more or less explanatory variables have been subdivided into various
equal, and if not, to explain the variation in these subcategories. Each elasticity estimate observation
Explanatory variables Dependent variable
Geo-economic Moderator variables: Effect size:
- Income - Elasticity time horizon - Price elasticity of
- Transfer distance - Method of analysis demand
- Geographic scope - Data collection period
- Fare class
Fig. 2. (a) Dependent and independent variables used. (b) Distribution of price elasticity estimates.
170 M. Brons et al. / Journal of Air Transport Management 8 (2002) 165–175
belongs to exactly one of the categories for each The categorisation decision as to whether an observa-
explanatory variable. tion should be considered a long-run or a short-run
Some discrete variables such as fare class, method of estimate is decided upon by following the author’s
research used or elasticity time horizon are generally statement about it. In the literature on which this paper
mentioned explicitly in the original studies. Other is based, sometimes different opinions were encountered
variables have been calculated and categorised based as to how a long-run elasticity estimate is deﬁned by the
on information in the original study. In Table 1, the author. Despite these differences, a common underlying
possible categories of each explanatory variable an assumption could be discerned that a long-run estimate,
elasticity estimate observation can belong to are as opposed to a short-run estimate, does not only signal
presented. the direct budgetary effects of a fare change on demand,
Gross domestic product (GDP) values are determined but also takes into account adjustments relating to
by averaging the values for the origin and destination relocation and asset ownership.
country of any given ﬂight. These values are calculated The variables with respect to fare class and method of
by taking the average of the GDP values for every year analysis are categorised based on direct information in
of the period in which the data used in the original study the original study.
has been collected. GDP values obtained from Summers
and Heston (1991), are corrected for price level.
Values for the period of data collection are deter- 3.2. Descriptive statistical results
mined by calculating the average year of the period in
which the data used in the original study has been Before we will proceed with the use of meta-regression
collected. techniques to statistically explain the variations among
The values for the distance variable are calculated as price elasticity estimates, it may be interesting to look at
the estimated average distance of every origin–destina- some descriptive statistics on the set of price elasticity
tion combination that enter in the original study’s estimates we used and how this relates to several other
calculation of a given price elasticity estimate and are sample characteristics and moderator variables we
then categorised. employed in our analysis.
The geographic scope variable consists of the Fig. 2b visually represents the distribution of the set
categories North America; Europe; Australia and of elasticity estimates we collected from the case studies.
intercontinental. The ﬁrst three categories consist of The overall mean price elasticity, based on our set of 204
intra-continental ﬂights within the concerned continent. observations, is below unity at À1.146, implying that
Flights between different continents are categorised as price changes will result in more than a proportional
‘intercontinental’. change in demand. The standard deviation of the
elasticity distribution is 0.619.
The lowest elasticity we found is À3.20, while the
Table 1 highest is positive at 0.21, the latter being the only
Categorisation of variables positive price elasticity we found in the original studies.
The curious double top formation in Fig. 2 largely stems
from the general difference in price elasticity estimates
Gross domestic product Continuous variable between case studies that focused on business class
Data collection period Continuous variable
travellers and other case studies. The distributions in
Distance Close distance (o500 miles) Figs. 3a and b clearly illustrate that price elasticities
Medium distance (500–1500 miles) among case studies focusing on business class passengers
Long distance (>1500 miles) generally tend to be higher (closer to zero) than elasticity
estimates gathered from other studies. Figs. 4a–c which
Geographic scope North America
represent the elasticity distribution within the close-
Australia distance, medium-distance and long-distance subcate-
Intercontinental gories, respectively, show equally interesting patterns.
The relative positions of these distributions seem to
Elasticity time horizon Short-term elasticity imply that long-distance ﬂights generally correspond to
higher price elasticities than short-distance ﬂights.
Fare class Economy class Medium-distance ﬂights tend to correspond to medium
Business class level elasticities.
In Table 2, a number of simple correlation coefﬁcients
Method of analysis Time series analysis between the set of elasticity estimates and several classes
of sample characteristics and study descriptors are
M. Brons et al. / Journal of Air Transport Management 8 (2002) 165–175 171
Fig. 3. (a) Distribution of price elasticity estimates from studies on business class travellers. (b) Distribution of price elasticity estimates from other
Fig. 4. (a) Distribution of price elasticity estimates from short-distance ﬂights. (b) Distribution of price elasticity estimates from medium-distance
ﬂights. (c) Distribution of price elasticity estimates from long-distance ﬂights.
172 M. Brons et al. / Journal of Air Transport Management 8 (2002) 165–175
Table 2 3.3. Meta-regression analysis
Correlations of determinants with price elasticity
Coefﬁcient Standard error The meta-regression equation to be estimated has
already been discussed.2 The positive and signiﬁcant
Year of data 0.255 0.000
Distance 0.304 0.000 coefﬁcient of the year of data variable (Table 3) points
Intercontinental 0.466 0.000 at a lower price elasticity over time. Apparently,
North America À0.350 0.000 consumers have become less price sensitive in time.
Europe À0.062 0.382 The distance variable enters the regression analysis with
Australia À0.124 0.077
a negative, but insigniﬁcant coefﬁcient. Apparently,
Short-term estimates 0.199 0.004
Long-term estimates À0.199 0.004 there are multiple factors exerting their inﬂuences here,
Economy class À0.303 0.000 thus distorting an unambiguous relationship between
Business class 0.341 0.000 the distance of a ﬂight and the price elasticity of demand
Time Series studies 0.438 0.000 for a ﬂight. Arguably, the theoretically predicted
Pooled studies À0.250 0.000
Cross-section studies À0.133 0.058
decreasing effect on price sensitivity due to a relative
GDP À0.206 0.006 lack of substitute modes on long-distance ﬂights may
prove insufﬁcient to dominate the theoretically pre-
dicted increasing effect on price sensitivity because long-
distance ﬂights demand a larger share of the disposable
income than short-distance ﬂights. Apparently, this
The price elasticity level appears to be positively seems to hold even if, as has been pointed out earlier,
correlated with the year data variable. Apparently, fare and ﬂight distance are not as perfectly positively
travellers have become less price sensitive over the years. correlated as would be expected from a cost perspective.
Travellers are also less price sensitive, as ﬂight distance The Europe dummy shows an insigniﬁcant positive
increases. This may be due to the relative lack of sign. This is rather surprising, since one would expect a
substitution modes on longer distance ﬂights. From the somewhat higher price elasticity among European
geographic location variables, the intercontinental vari- travellers due to the fact that the intra-European
able is positively correlated to price elasticity, while each passenger surface transport network offers better sub-
of the continent variables is negatively correlated with stitutes than its North American and Australian
the price elasticity. The ﬂight distance of intercontinen- counterparts. Moreover, the fact that the European
tal ﬂights is generally longer than intra-continental income level is low compared to North American and
ﬂights, and therefore this ﬁnding may partially be Australian values would lead one to assume a somewhat
explained by the same arguments as given above with higher price sensitivity among European travellers. Yet
respect to the distance coefﬁcient. we ﬁnd that passengers within Australia are more price
The correlation coefﬁcients between price elasticity sensitive than other passengers. It is also interesting to
and short-term estimates is negative, whereas that for see that the time series dummy is rendered insigniﬁcant
long-term estimates is positive. This implies that by the inclusion of the region-speciﬁc dummies. One
demand responses to fare price changes are higher when might expect multicollinearity here, especially when
a longer time horizon is used. The negative correlation there is a relatively large number of time series studies
coefﬁcient between price elasticity and the economy for a speciﬁc region.3 However, the time series dummy
class dummy and the positive coefﬁcient between price has the opposite sign of the Australia dummy, and the
elasticities and the business class dummy point at a same as the insigniﬁcant US dummy, where one
lower price sensitivity among the latter segment. The typically would expect a relatively large number of
dummy for time series studies is positively correlated cross-sectional studies. Note that the sign of the time
with price elasticity, while both the dummy for cross- series dummy is positive and signiﬁcant, indicating that
section studies and the dummy for pooled studies are estimates from time series data yield lower price
negatively correlated with price elasticity. According to elasticities (i.e. more price sensitive passengers). This
the correlation coefﬁcient, GDP is negatively correlated could indicate that, as some authors observe, time series
with price elasticity, which means that travellers with a estimates without extensive lag structures typically yield
higher income tend to be more price sensitive than
travellers with a lower income. The fact that travel, 2
Instead of using GDP values in the regression equation, we used a
and speciﬁcally air travel, may be regarded as a dummy variable for case studies that correct for income while
relatively luxury good which would provide for a estimating demand. Due to the high number of such studies in our
set, the use of GDP value as an explanatory variable would lead to
positive relationship between income level and the
share of income allocated to air transport, may explain 3
Correlation results indeed indicate a strong correlation between the
this higher price sensitivity among high income time series dummy and intercontinental ﬂights and, to a lesser extent,
travellers. between the time series dummy and the North America dummy.
M. Brons et al. / Journal of Air Transport Management 8 (2002) 165–175 173
Table 3 other modes, the results show that a long-run adjust-
Meta-regression coefﬁcients ment time causes the price elasticity to decrease.
Estimation 1 Estimation 2 Finally, the coefﬁcient for the dummy for observa-
tions which are estimated by a demand equation in
Coefﬁcient T-value Coefﬁcient T-value
which an income related explanatory term is entered is
(Constant) À21.972 À2.200 À18.793 À1.885 positive and signiﬁcant, meaning that the inclusion of an
Year of data 0.011 2.146 0.009 1.852
income-related term in the demand equation draws the
Distance À0.005 À0.066 À0.023 À0.306
Europe 0.201 0.962 0.112 0.425 absolute (negative) coefﬁcient of the fare price related
North America 0.022 0.130 termFthe price elasticity of demandFtowards zero. In
Australia À0.493 À2.194 other words, if the results of the original study are
Long run À0.644 À2.555 À0.906 À3.278 corrected for income level, this yields a lower price
Business class 0.552 5.414 0.622 5.913
sensitivity. In theory, omission of the income variable in
Cross-section 0.181 1.594 0.008 0.060
Time-series 0.439 3.022 0.154 0.830 the demand equation causes the price parameter to be
Dummy GDP 0.589 4.151 0.761 4.347 biased. When only the price and income are arguments
in the demand equation and if the income is falsely
R2 0.416 0.435 omitted, the bias is given by the income parameter
multiplied by Cov(price, income)/Var(price) (see
Greene, 2000). When this covariance is negative (for
short-term elasticity estimates, although we have already example, because in the course of time income increases
included a long-term dummy.4 in real terms and fares decrease in real terms as a result
The dummy coefﬁcient for observations derived of deregulation), the bias has a negative sign and the
exclusively from business class data has a positive sign estimated price sensitivity is too high. Thus, inclusion of
and is signiﬁcant. Business class travellers’ price the income would yield a lower price sensitivity.
elasticities are lower on average than economy class
travellers’ price elasticities. This may be explained by the
lower price sensitivity in the business class segment, 4. Conclusions
which might be primarily due to the higher valuation of
time among these travellers. Other explanations may be When formulating an environmental policy concern-
that business class passengers generally have less ing aviation, the authorities also need information on
substitutes than leisure passengers. Business class the price sensitivity of passengers to predict the
travellers may also have a higher degree willing to pay effectiveness of the policy. For example, if the airlines
more for ‘higher quality’ services such as allowance for or airports can charge all extra costs to the passengers
last-minute bookings and changes in travel plans, better without decreasing demand, the policy has no other
check-in and on-board facilities and that tickets are effect than increasing the authorities’ revenues. In this
usually paid by the ﬁrm, rather than the individual paper, price elasticities of demand for passenger air
traveller. Apparently, these characteristics of demand transport were analysed. This research synthesis pro-
for business class are important enough to result in a vides valuable information on the ‘‘expected’’ level
signiﬁcant coefﬁcient with the expected sign. Again, it is of price sensitivity in a speciﬁc setting. It can also
important to note that travellers with a business motive provide information on the design of new research/a
do not necessarily travel business class. An analogous questionnaire.
story holds for travellers with a leisure goal, although From the meta-regression analysis, the following
the discrepancy will be of less importance. results are obtained. Long-run price elasticities are
The coefﬁcient of the dummy for observations in higher in absolute value, as was to be expected from a
which long-run elasticities are estimated is negative, theoretical point of view. Hence, basing long-run policy
implying that a longer response time indeed enables instruments on short-run elasticities leads to distortions.
consumers to adjust better to changes in fare prices. Passengers become more price sensitive over time; this
Even if such long-run demand adjustment effects are also needs to be acknowledged in the design of long-run
expected to show less severe in air transport than in policy instruments. Business passengers are less sensitive
to price; this is a common ﬁnding in the literature. The
In a plot of a sufﬁciently long time series, convergence to difference is about 0.6, ceteris paribus. This fact gives
equilibrium might be visible. However, a simple regression (of trafﬁc the airlines the opportunity to charge extra costs
on fare) would weight the data point representing the initial impact of (resulting from a price-based policy instrument) to the
the fare change the same as the data point representing the long-run
passengers, where the business passengers can be
equilibrium. The estimated fare coefﬁcient is therefore biased towards
the initial impact. Because the cross-sectional variance is larger, the charged more than proportional without decreasing
estimation results from a cross-sectional study are therefore more demand. If this is not acknowledged by the authority,
likely to be indicative of the long-run effect (Abrahams, 1983). environmental policy may have little success, although
174 M. Brons et al. / Journal of Air Transport Management 8 (2002) 165–175
the authority gets the revenues. Surprisingly, it is found sectional data in the US domestic market. Cicil
that European passengers are not more price sensitive Aeronautics Board, paper presented at the 47th annual
than US passengers and Australian passengers, while meeting.
one would expect that the availability of more sub- Bureau of Transport and Communications Econom-
stitutes in Europe would result in a higher price ics. Demand for Australian domestic aviation services:
sensitivity within Europe. If in the underlying study, forecasts by market segment. Occasional paper 79,
the income variable is falsely omitted, the price elasticity AGPS, Canberra.
is higher in absolute value, i.e. the price sensitivity is Bureau of Transport and Communications Econom-
higher. When the results of the study are to be used, it is ics. Trends and prospects for Australian international
important that one is aware of this bias. This ﬁnding air transport. Occasional paper 88, AGPS, Canberra.
also suggests that while setting up a new study, income Bureau of Transport and Communications Econom-
should not be left out. ics, 1995. Demand elasticities for air travel to and from
The research agenda that follows from this paper is as Australia, working paper 20.
follows. First, attention should be paid to the question De Vaney, A.S., 1974. The measuremental cost of
whether research synthesis allows for the determination airport noise. Environmental Quality Program, Texas
of conﬁdence intervals for price elasticities in speciﬁc A&M University.
cases. In order to do this, one needs accurate data on a Fridstrom, L., Thune-Larsen, H., 1989. An econo-
wide range of study characteristics. For example, the metric air travel demand model for the entire conven-
ﬁnding on price sensitivity of European passengers tional domestic network: the case of Norway.
necessitates additional analysis, using more and more Transportation Research 23B(3), 213–223.
speciﬁc characteristics. Second, methods of analysis Furuichi, M., Koppelman, F.S., 1994. An analysis of
other than simple regression analysis should be ex- air travelers’ departure airport and destination choice
plored. For example, inclusion of extra explanatory behavior, Transportation Research 28A(3), 187–195.
variables in the analysis in this paper if they were to Haitovsky, Y., Salomon, I., Silman, A., 1987. The
become available, could cause endogeneity or multi- economic impact of charter ﬂights on tourism to Israel:
collinearity problems. In general, given the relatively an econometric approach. Journal of Transport Eco-
small number of observations and the fact that some nomics and Policy 21(2), 111–134.
observations may be obtained from the same study, one Hartley, P., Trengrove, C., 1991. East–West Airlines’
should take notice of the asymptotic properties of the Sydney–Melbourne service: market expansion or market
estimator. Finally, research synthesis does not take away diversion? The Economic Record 66(194), 203–220.
the need of speciﬁc case studies. As explained above, Hensher, D.A., Louviere, J.J., 1983. Identifying
research synthesis may be necessary because performing individual preferences for international air fares. Jour-
new case studies is, for example, too expensive or there is nal of Transport Economics and Policy 17, 225–245.
too little time for it. But, of course, research synthesis Ippolito, R.A., 1981. Estimating airline demand with
can only be carried out on the basis of a sufﬁcient quality of service variables. Journal of Transport
number of similar case studies. Economics and Policy 15(1), 7–16.
Jud, G.D., Joseph, H., 1974. International demand
for Latin American tourism, Growth and Change
Appendix A. The case study literature examined 25–31.
Jung, J.M., Fuji, E.T., 1976. The price elasticity of
Abrahams, A., 1983. A service model of air travel demand for air travel. Journal of Transport Economics
demand: An empirical study. Transportation Research and Policy 10, 257–262.
17A (5), 385–393. Liew, C.K., Liew, C.J., 1979. Airline deregulation and
Agarwal, V., Talley, W.K., 1985. The demand for its impact on intercity travel demand. Transportation
international air passenger service provided by US air Research Record 732, 34–39.
carriers. International Journal of Transport Economics Marin, P.L., 1995. Competition in European aviation,
12(1), 63–70. pricing policy and market structure. The Journal of
Alperovich, G., Machnes, Y., 1994. The role of wealth Industrial Economics 43(2), 141–159.
in the demand for international air travel. Journal of Melville, J.A., 1998. An empirical model of the
Transport Economics and Policy 28(2), 163–173. demand for international air travel for the Carribean
Andrikopoulos, A.A., Teriovitis, T., 1983. An ab- region. International Journal of Transport Economics
stract mode model, a cross-section and time series 25(3), 313–336.
investigation. International Journal of Transport Eco- Milloy, H.B., Douglas, L., Sullivan, S.M., 1985.
nomics 10(3), 547–555. Market response to discount fares on selected domestic
Brown, S.L., Watkins, W.S., 1974. The demand for trunk routes. Papers of the Tenth Australian Transport
air travel: a regression study of time series and cross- Research Forum, ATRF, Melbourne, pp. 1–18.
M. Brons et al. / Journal of Air Transport Management 8 (2002) 165–175 175
Morrison, S., Winston, C., 1989. Enhancing the market. Journal of Transport Economics and Policy
performance of the deregulated air transportation 14(1), 19–35.
system. The Brookings Institution, Washington, DC. Taplin, J.H.E., 1997. A generalised decomposition of
Morrison, S., Winston, C., 1986. The economic effects travel-related demand elasticities into choice and gen-
of airline deregulation. The Brookings Institution, eration components. Journal of Transport Economics
Washington, DC. and Policy, 183–191.
Mutti, J., Murai, Y., 1977. Airline travel on the North Thompson, A., Caves, R., 1992. The projected market
Atlantic. Journal of Transport Economics and Policy share for a new small airport in the north of England.
11(1), 45–53. Regional Studies 27(2), 137–147.
Okeahialam, C.C. An empirical study of cross- Walters, A.A., 1978. AirportsFan economic survey.
channel route choice. Channel Tunnel Research Unit, Journal of Transport Economics and Policy 12, 125–160.
University of Kent, Canterbury. Wheatcroft, S., 1956. The Economics of European Air
Oum, T.H., Gillen, D.W., 1983. The structure of Transport. Manchester University Press, Manchester.
intercity travel demands in Canada: theory tests and
empirical results. Transportation Research 17B(3),
Oum, T.H., Gillen, D.W., Noble, S.E., 1986. Demand
for fareclasses and pricing in airline markets. Logistics Abrahams, A., 1983. A service model of air travel demand: An
and Transportation Review 22(3), 195–222. empirical study. Transportation Research 17A (5), 385–393.
Oum, T.H., Zhang, A., Zhang, Y., 1993. Inter-ﬁrm Cooper, M.H., Maynard, A.K., 1971. The price of air travel, Hobart
Paper 53. Institute for Economic Affairs.
rivalry and ﬁrm-speciﬁc price elasticities in the deregu-
Crouch, G.I., 1992. Effects of income and price on international
lated airline markets, Journal of Transport Economics tourist demand. Annals of Tourism 19 (4), 643–664.
and Policy 27(2), 171–192. Greene, W.H., 2000. Econometric Analysis, Fourth edition. Prentice
Saad, M., 1983. Demand for air passenger travel in Hall, New Jersey.
Australia. Papers of the Eighth Australasian Transport Hooper, P., 1993. The elasticity of demand for travel: a review.
Research Forum, Vol. 2, Perth, pp. 605–623. Research Report, Institute of Transport Studies, Sydney.
Jung, J.M., Fuji, E.T., 1976. The price elasticity of demand for air
Smith, A.B., Toms, J.N., 1987. Factors affecting travel. Journal of Transport Economics and Policy 10, 257–262.
demand for international travel to and from Australia. Oum, T.H., Waters, W.G., Yong, J.S., 1992. Concepts of price
Bureau of Transport Economics, Canberra. elasticities of transport demand and recent empirical estimates.
Straszheim, M.R., 1978. Airline demand functions in Journal of Transport Economics and Policy 26 (2), 139–154.
the North Atlantic and their pricing implications. Mutti, J., Murai, Y., 1977. Airline travel on the North Atlantic.
Journal of Transport Economics and Policy 11 (1), 45–53.
Journal of Transport Economics and Policy 12, Marsan, M.A., Kostoris, F.P.S., 1993. Competition through fares and
179–195. fare accesses on the air transport markets. Discussion Paper Series
Talley, W.K., Schwarz-Miller, A., 1988. The demand 776, CEPR, London.
for air services provided by air passengerFCargo Straszheim, M.R., 1978. Airline demand functions in the North
carriers in a deregulated environment. International Atlantic and their pricing implications. Journal of Transport
Economics and Policy 12, 179–195.
Journal of Transport Economics 15(2), 159–167. Summers, R., Heston, A., 1991. The Penn World Table (Mark 5): an
Taplin, J.H.E., 1980. A coherence approach to expanded set of international comparisons, 1950–1988. Quarterly
estimates of price elasticities in the vacation travel Journal of Economics 106, 327–368.