• Population. Population has a direct effect on the size
of a market and may cause a bias in the estimates
if omitted. For example, a large increase in traffic
may reflect a sudden boom in population rather than
other effects. Population was tested in all model
specifications but the best results tended to be at the
city pair level, so it is only included in the US domestic
regression results.
• Route Distance (Trip Length). The use of route
distance is based on its ability to reflect the value of
travel time savings and availability of substitutes. As
distance increases, the viability of other transport
modes as a substitute decreases. The use of route
distance as an instrumental variable in 2SLS requires
that distance be uncorrelated with traffic. This is most
likely in the domestic US market.
• Substitute Goods. The inclusion of a substitute travel
price variable was tested on a subset of routes and
was found to increase price elasticities. Estimates
that exclude a meaningful substitute in the regression
model will produce more inelastic estimates than
a correctly specified model would produce. Route
substitutes can be defined as a different airport
serving the same catchment area (e.g. ChicagoO’Hare or Chicago-Midway) or a different destination
serving the same purpose (e.g. Las Vegas or Reno).
• Real Exchange Rates.10 In theory, as the foreign
country becomes more expensive (inexpensive),
leisure travellers will travel to the foreign country
less (more). However, the econometric research
was unable to obtain robust estimates using this
variable, possibly due to difficulties in obtaining an
accurate measure of the variable. Therefore, real
exchange rates were excluded from the final model
specifications.
• Time Variables. The use of several different forms
of time variables was explored, but only quarterly
(seasonal) time dummy variables were found to
increase the explanatory power of the model. This
suggests that both travel prices and demand are
inherently seasonal (at least on a quarterly basis).
• Route Dummies. These dummy variables were found
in many cases to increase the explanatory power of
the model, suggesting that some other route-specific
factors are important for demand (e.g. whether or
not non-stop service is available, whether there are
competing carriers operating on non-stop routes,
whether there is a low cost carrier, what types of
cultural or financial linkages there are between pairs
of cities). The use of route dummy variables controls
for these variables without the need to quantify them.
ECONOMETRIC RESULTS
The new econometric research supports the discussion
from previous chapters that the sensitivity of passengers
to the level of air travel prices depends significantly on the
level of the market being considered and its location. The
results are based on a synthesis of the new econometric
results and the review of previous research. Base
elasticity estimates are developed for the different levels
of aggregation (route, national and supra-national level).
Multiplicative estimates were then developed to adjust
the elasticities to reflect specific geographical markets.
i) Level of Aggregation
In summary, the econometric results found that at the
route level (where competition between airlines or citypair markets is high) the sensitivity of demand to price
is very high. However, at the national or regional level,
air travel is relatively price insensitive. The results support
demand elasticities of:
• Route Level: -1.4
• National Level: -0.8
• Supra-National Level: -0.6
Route Level. The review of previous research found route
level elasticities ranging from -1.2 to -1.5. Regressions
using the US DB1A data, which allows the use of route
dummies and variables to capture the price of route
substitutes, produced a similar air travel price elasticity
of -1.4. This elasticity estimate is applicable to a situation
where the price of an individual route changes (e.g.
higher airport charges at Paris CDG raising the price of
travel from London and diverting leisure traffic to another
destination, such as Frankfurt). Using distance as an
instrument variable within the 2SLS model produces
results that further support this elasticity, though there
still is some concern over the use of distance in this way
due to its perceived exogenous influence on demand.