PRICING OF AIRPORT SERVICES ACCORDING TO PRINCIPLES OF
WELFARE ECONOMICS
Mathisen, T. A., Jørgensen, F. and Solvoll, G.
Bodø Graduate School of Business at University of Nordland
NO-8049 Bodø, Norway
1. INTRODUCTION
Airport infrastructure is, in most countries, financed by airport charges put on
the services produced at the airport together with revenues from commercial
activities. The international air transport organizations aim for standardization
of the airport charges (Martin-Cejas, 1997) and ICAO (2009) has a general
policy stating that charges, amongst other things, should be based on costs
and be non-discriminatory. Most airports charge both passengers (PAX) and
air traffic movements (ATM) but there are variations with respect to the
distribution of charges between PAX and ATM both between countries and
type of ownership.
The major part of the Norwegian airports is operated by the state owned
company Avinor. These airports are operated commercially and do not in total
require state subsidies, but there is some cross-subsidization between the
largest airports and those located in rural areas.
The Norwegian Ministry of Transport and Communication initiated in 2010 a
project to evaluate the charges that generate the airports’ traffic revenues.
This paper presents the main results from this project. The aim of the paper is
to analyze the Norwegian airports’ costs structure and subsequently derive
the marginal cost of serving passengers and airplanes. Taking the marginal
cost estimations as a starting point, the paper suggests a revised scheme for
aviation charges designed according to principles welfare economics.
The further organization of the paper is: Section 2 provides a brief
presentation of the Norwegian airport infrastructure with special focus on how
it currently is financed. Then, Section 3 discusses an econometric model
suitable for studying costs at Norwegian airports. Next, Section 4 gives details
about the data set, presents model results and derives marginal costs for PAX
and ATM. The results are applied in new schemes for airport charges in
Section 5 where the cost of raising public funds also is taken into account.
Finally, possible implications for air transport companies, passengers and the
authorities by implementing these airport charges are briefly discussed in
Section 6.
2. NORWEGIAN AIRPORTS AND AIRPORT CHARGES
2.1 The Norwegian Airport Structure
Today Norway is amongst the countries in Europe with the highest air
transport dependence (Williams et al., 2007). While Norway in 2003 had a
© Association for European Transport and Contributors 2011
1
domestic air trip rate per capita on 2.27, most European countries had less
than one third of this value. Moreover, Williams et al. (2007) shows that
Norway has the highest number of commercial airports with short runways (<
1 000 m) in Europe. 1
The state, through the wholly owned company Avinor, owns and operates 46
airports throughout Norway with a total annual traffic of about 40 million
passengers (PAX) in 2010 (Avinor, 2011). Avinor organizes the airports in
three groups where the main airport of Norway makes the first group, the
three other large airports located by the cities Bergen, Stavanger and
Trondheim makes the second group and the remaining 42 airports makes the
third group as illustrated in Figure 1. For further details regarding the airport
network in Norway see e.g. Lian (2010).
Figure 1 Norwegian airports owned by Avinor (Source: Avinor)
2.2 Current Airport Charges
The airport charges imposed on air traffic are compound of five elements; 2
starting, passenger, security, TNC (Terminal Navigation Charge) and in-flight.
Security and TNC are charged according to costs, while in-flight charges are
set by Eurocontrol. The two remaining categories, starting charge and
passenger charge which is the focus of this study, are set by Avinor according
to the directions from the Norwegian Ministry of Transport and
© Association for European Transport and Contributors 2011
2
Communication. Avinor also generates revenues indirectly related to air
transport from commercial activities such as rent from shops and tax-free
sales. Prices for the commercial activities are determined by Avinor. Total
revenues are almost evenly distributed between PAX and ATM charges on
one hand and commercial activities on the other hand.
The basis for this study is the airport charges for 2009 which is the year of
reference for all comparisons between new and old charges. Passengers
were charged NOK 36 3 and NOK 59 for domestic and international flights,
respectively. Air traffic movements were charged NOK 96 per tonne according
to the aircrafts’ maximum take-off weight (MTOW). For particularly large
aircrafts the charge is reduced to NOK 48 for each tonne exceeding 100. The
charges were updated in 2010 to match change in costs and the entire
scheme for ATM was revised in 2011.
3. THE MODEL
Airport charges are vital for the financing of airport infrastructure and make a
considerable part of operating costs for airliners. These charges to finance
airport infrastructure should, from a welfare economic point of view, be set to
reflect the costs imposed by the activities.
3.1 Previous Costs Studies
An important basis for the analysis of airports’ costs is to define which
services are provided by an airport and how they can be measured. Common
production measures to assess the activity at an airport are the number of
passengers (PAX), air traffic movements (ATM), work load units (WLU) and
commercial revenues, while the size (or capacity) often is measured by
runway length, terminal area (square meters), number of gates, etc (see e.g.
Martín and Voltes-Dorta, 2011).
One should, however, be aware that the choice of production measure and
model formulation will influence the cost estimates. A comparison of the
results using different production measures is given by Martín and VoltesDorta (2008) revealing large differences in estimated marginal costs
depending on model specification. Aiming to internalize external costs of air
transport such as noise, pollution and delays (see e.g. Santos and Robin,
2010) also make estimations of social marginal costs more complicated.
The most central production measures when studying airports are PAX and
ATM. However, a potential problem for econometric analyses is that these two
explanatory factors are highly correlated; when PAX increase the number of
ATM will increase to meet the increased demand and vice versa. A way to
avoid this problem is to use the above mentioned WLU-measure where
passengers for example are equivalent to 100 kg and counting together with
the weight of the airplane. This production measure has, however, shown to
be imprecise and vague and is rarely used in recent studies.
© Association for European Transport and Contributors 2011
3
Marginal costs can be derived using econometric analysis and form the basis
for designing charge schemes to finance airport infrastructure. An example of
marginal cost estimation in Scandinavia was carried out by Carlsson (2003)
using an exponential model with passengers as the only independent variable
and explained about 96% of variation in costs between airports. When aiming
to meet revenue restrictions the Ramsey-rule has been applied by MartinCejas (1997) and Hakimov and Scholz (2010) to design charge schemes
based on marginal cost estimates for Spanish and German airports,
respectively.
3.2 The Applied Cost Model
An econometric approach using a model with the production measures
number of passengers (PAX) and air traffic movements (ATM) as independent
variables is applied to study the costs structure at Norwegian airports.
Moreover, a dummy variable (INT) is introduced to indicate whether airports
have established infrastructure for international flights. The applied cost model
is presented in equation (1).
(1)
where
The dependent variable, costs ( ), in (1) is measured in Norwegian kroner
(NOK) at the 2009 price level. The accounting figures are reported by Avinor
and include all costs at the airport except security which is an outsourced
service and covered by a separate charge. Regarding the independent
variables, PAX is defined as the number of terminal passengers (sum of
arrival and departure and transfer) at both route- and charter flights and ATM
is the sum of arrival and departure flights including both route and charter.
The dummy variable, INT, is defined as 1 if international PAX at the airport
exceeds 5000, else
. The limit of 5000 PAX is set in accordance with
the airport owner Avinor. Such a lower limit ensures that an airport not really
dimensioned for international traffic is not treated as one in the analysis even
if international flights should appear on occasion. A-priori assumptions are
that all -parameters are positive. The error term
has the traditional
properties of normal distribution, expected value 0 and non-correlation with
the independent variables.
The model specification in (1) is linear. Such a model has properties in
accordance with reasonable a-priori assumptions regarding the cost structure
at Norwegian airports. For example, the linear specification implies that
marginal cost for an ATM is independent of PAX at the airport and vice versa.
Even if such a specification is less flexible than for example translog, it often
provides a good approximation for more advanced models (see e.g. Pels and
Rietveld, 2000). Moreover, the results from linear models are easy to interpret.
It is clear from (1) that the marginal cost with respect to PAX and ATM are
and
, respectively. Since (1) do not include
capacity measures as independent variables and is estimated using crosssectional data, the marginal cost estimates must be interpreted as long run. 4
© Association for European Transport and Contributors 2011
4
3.3 On Weight based Marginal Costs for ATM
There are strong indications that the part of the charge scheme related to
ATM should be separated in one part depending on aircraft weight and
another part being independent of aircraft weight (see e.g. Graham, 2008).
The actual relationship between costs on one hand and ATM and departed
tonne on the other hand can be studied further by analyzing empirical data
using the specification in equation (2).
(2)
where
The variable WGT is introduced in (2) to represent average airplane weight,
measured in MTOW, for an ATM at the airport. The total weight departed at
an airport is thereby (ATM × WGT). The exogenous parameter
in (2)
represents how weight of airplane influence costs and is assumed to be
positive.
A differentiation of (2) with respect to ATM gives marginal costs
meaning that the marginal costs to the airports of a landing can
be divided in weight independent marginal costs (
and weight dependent
marginal costs
. The relationship between marginal costs and WGT
is concave, linear or convex for values of less, equal to or higher than 1,
respectively. In a study of 161 airports worldwide Martín and Voltes-Dorta
(2011) estimated a similar value to 1.1 which indicates increasing unit rates
with respect to weight. In contrast, the current charge scheme at Norwegian
airports has a decreasing unit rate as accounted for in Section 2.2.
Estimations using the Norwegian data set indicate that the statistical
properties are about unchanged for values of between 0.8 and 1.2. Hence,
since the model is easier to interpret when
, this value is applied and
implies that marginal costs increase linearly with WGT.
The weight based element (ATM
WGT) could not be incorporated in (1)
due to high multicorrelation. Hence, we have to estimate (2) separately. The
fact that model (1) incorporates both PAX and ATM while (2) only includes
ATM makes it necessary to adjust the estimates from (2) if they are used in
combination with (1). Hence, it can be assumed that the results from (2) is
valid in (1) for an average ATM on Norwegian airports which has a weight of
28 tonne. 5 More specifically, the estimated parameter
in (2) must be
adjusted to match
in (1) when airplane weight is 28 tonne. This will be
demonstrated in section 4.3.
4. DATA AND ESTIMATION RESULTS
4.1 Data
The data set consists of annual observations of production and accounting
figures for all 46 airports owned by Avinor for the years 2007, 2008 and 2009.
Two airports are omitted from the analysis. First, Oslo airport, Gardermoen
(OSL) has been omitted due to its large size compared to the other Norwegian
airports. If included OSL accounts for 40% of all airports’ costs and 30% of
© Association for European Transport and Contributors 2011
5
ATMs and is, thus, an extreme outlier for all variables. The relevance for the
results excluding OSL is handled by studying whether the estimates change if
this airport is introduced as a dummy variable. Second, the heliport (purely a
landing site for helicopters) at Værøy, (VRY) has been omitted since its cost
structure deviates somewhat from regular airports.
The observations for all three years are used as a pooled data set to increase
the number of observations in a cross-sectional analysis. To our knowledge
there have been no major changes in airport operations during these three
years that should make pooled data analysis inappropriate. Accounting figures
are adjusted to 2009 price level using the consumer price index provided by
Statistics Norway. Consequently, the data set contains a total of 132
observations over three years. Descriptive statistics for the pooled data set
are presented in Table 1.
Table 1 Annual descriptive statistics for the airports (N=132).
Variable
C
PAX
ATM
INT
Average
Std.dev.
Median
Minimum
Maximum
40 500 000 42 600 000 20 900 000 10 400 000 197 000 000
452 552
980 148
76 200
5615
4 657 634
9123
14 222
4887
711
70 502
0.26
0.44
0
0
1
Since the average values are significantly higher than the median values all
variables in Table 1 are right skewed which is due to the presence of few
airports with high traffic and many small airports with low traffic in the data set.
The average airport in the data set has annual operating costs of NOK 40.5
million, about 450 000 passengers and 9100 air traffic movements. If OSL is
included the average costs would rise to NOK 72.7 million.
Table 2 Correlation matrix.
Variable
Costs (C)
Passengers (PAX)
Air traffic movements (ATM)
International airport (INT)
C
1
.96*
.95*
.76*
PAX
ATM
1
.98*
.67*
1
.65*
INT
1
* Significantly correlated at 1% level (2-tailed).
The matrix in Table 2 shows that correlations between costs and the
independent variables are very high. High correlation is also the case between
the independent variables, in particular between PAX and ATM. This is
expected from earlier studies of this industry (e.g. Hakimov and Scholz, 2010).
4.2 Estimation Results
Estimation results by applying the empirical data in (1) are presented in Table
3. The model explains about 96% of the variation in costs between airports
and the F-test indicates good model fit. All estimated parameters have signs
© Association for European Transport and Contributors 2011
6
in accordance with the a-priori assumptions and are significantly positive at
1% level, except
which is significant at 5% level (2-tailed).
Table 3 Estimation results – airport cost model
Variable
Coefficient
Std. dev.
Constant
17 230 851
1 166 577
PAX
26.98
4.17
ATM
604
283
INT
21 382 929
2 334 910
N
132
2
R
0.96
F-test
995
t-value
14.8
6.5
2.1
9.2
Further tests of statistical properties indicate near extreme multicollinearity
with a mean VIF value of 19.4. This is an expected consequence of the model
specification due to the close relationship between the independent variables
PAX and ATM. This leads to high standard errors but is not a violation of OSL
regression assumptions. The Breusch-Pagan/Cook-Weisberg test and Whites
test returned p-values of 0.45 and 0.38, respectively, which fails to reject
homoskedasticity at the 5% significance level (e.g. Wooldridge, 2006). A study
of the residuals shows mean value about 0 and a slightly skewed and highly
peaked distribution. The error term is not significantly correlated to any of the
independent variables. Hence, the statistical properties are generally good
and indicate that the estimation results from the OLS regression can be
trusted.
The estimation in Table 3 shows long-run marginal costs for an extra
passenger (PAX) of NOK 27 and for an extra air traffic movement (ATM) of
NOK 604. Values must be doubled if only departures are charged. Fixed costs
are about NOK 17.2 million and additional costs if the airport is prepared for
international flights are about NOK 21 million per year.
Table 4 Estimation results – weight-based cost model
Variable
Coefficient
Std. dev.
Constant
14 985 590
1 206 577
ATM
848
429
ATM × WGT
0.048
0.013
INT
18 701 570
2 908 037
N
132
2
R
0.95
F-test
819
t-value
12.4
2.0
3.6
6.4
Estimation results by using the weight based cost model in (2) are presented
in Table 4. The parameters are significantly positive at 5% level or higher and
the model generally has good statistical properties. Multicollinearity is present
due to the close relationship between ATM and (ATM × WGT).
The coefficients in Table 4 shows that for an ATM the weight independent
marginal cost is NOK 848 and the weight dependent cost increase by NOK 48
© Association for European Transport and Contributors 2011
7
when airplane weight increases by 1 tonne. This indicates that the ATM
charge should not increase proportionally with respect to weight. That is; a
weight increase by X% should increase charges by less than X%. For
airplanes of 20 tonne and 50 tonne, it follows from Table 4 that the weight
dependent part of the ATM marginal costs makes 53% and 74%, respectively
4.3 Marginal Cost Estimations
Since (1) includes both PAX and ATM and (2) only includes ATM, a calibration
is required if the estimates from the weight dependent model in Table 4
should be used together with the estimates of ATM in Table 3. This enables
marginal costs for ATM using weight independent and weight dependent
elements and PAX to be applied simultaneously. Let us assume a fixed
relative proportion between the values of the weight independent
and
weight dependent
marginal costs. An average aircraft of 28 tonne
with marginal costs, according to the -estimate in Table 3, of NOK 604 form
the basis for the calculations. An adjustment factor to match Table 4 estimates
to the Table 3 estimates can then be calculated according to equation (3).
(3)
The adjustment factor, , in (3) states that Table 4 estimates should be
multiplied by 0.28 if used together with PAX. Hence, weight independent
marginal cost for an ATM is NOK 237 (NOK 848 × 0.28) and the increase in
marginal cost when the aircraft weight increases by one tonne is NOK 13.40
(NOK 48 × 0.28). These estimates will be combined with the PAX marginal
cost amounting to NOK 27.
5. MARGINAL COST BASED CHARGE SCHEMES
The Norwegian Ministry of Transport and Communication has stated three
main objectives for the revised charges for airports owned by Avinor. First, all
requirements of the EU (2009) directive for airport charges must be met.
Second, the new charges should to a larger extent than the current ones be
based on the principles of welfare economics. Finally, there are budget
(revenue) restrictions. Additionally, it has been common practice that charges
are equal for all airports throughout the country, but this condition must be
relaxed if e.g. Ramsey pricing or peak-load pricing should be implemented.
Other considerations are that charges must be perceived as “fair” for both
carriers and passengers and meet regional policy objectives.
5.1 Charges Based on Marginal Costs Only
Based on the marginal costs estimates for PAX in Table 3 and adjustments
according to equation (3) for ATM gives the charges presented in equation
(4). The charges are set for departures only, implying that estimated values
are multiplied by 2.
(4)
© Association for European Transport and Contributors 2011
8
It is evident from (4) that marginal cost based airport charges imposed on the
air transport companies increase by NOK 54 for each passenger and by a
fixed element of NOK 474 and a variable element increasing by NOK 26 per
tonne according to the maximum take-off weight (MTOW) for the airplane.
A comparison between the marginal cost charges suggested in (4) and
current charges shows that, on average for domestic and international flights,
ATM charges are reduced by NOK 52 for an average aircraft of 28 tonne and
PAX charges are increased by NOK 10. This corresponds to a reduction of
54% for ATM and an increase of 23% for PAX. Consequently, marginal cost
pricing implies a transfer in taxation from ATM to PAX. That is; even though
both elements should still be taxed, it is indicated that a considerable shift
towards charging PAX should be implemented.
A consequence of marginal cost pricing is, however, that total revenues are
reduced by about NOK 700 million compared to the current charge scheme
for Norwegian airports. Several approaches can be chosen to meet a budget,
or revenue, restriction. The simplest way is to raise all marginal costs
elements by the same factor, relatively spoken, that is sufficiently high to
match revenues generated by the current charge scheme. Such an
adjustment would raise the airport charges by 57% to NOK 85 for PAX and
NOK 748 + NOK 42 × MTOW for ATM. It could, of course, be questioned
whether it is reasonable to use the same percentage adjustment factor for all
three elements of the airport charges in (4), but we have no indications of the
direction of such differentiations of the adjustment factor.
Another well recognized approach to meet revenue restrictions is to use the
differences in air travelers’ price elasticity throughout the country by the
Ramsey-rule. Charges could, for example, be differentiated at the airport
level. The diversity in airport location from rural areas to large cities enables a
differentiation according to the presence of alternative transport for
passengers. The price elasticity will be lower (less price sensitive) if
alternative transports are few. Hence, such airports should have a higher raise
in marginal costs than airports located in regions with many transport
alternatives. Ramsey pricing according to these criteria could, however, be
problematic from a regional policy perspective since the group with few
alternatives (which would get the highest raise) usually is located in rural
areas where the airport is regarded an important mean for continued
settlement and further economic development. An example of the
implementation of the Ramsey-rule at Norwegian airports according to these
criteria is given by Jørgensen et al. (2011).
5.2 Taking into Account the Costs of Raising Public Funds
The deficit arising from marginal cost pricing must be financed by the state.
Hence, it should be addressed whether the raising of public funds to finance
the revenue loss has any welfare economic consequences. The costs for
general taxation have been studied empirically and The Norwegian Ministry of
Finance (2005) has concluded that an add-on factor of 20% should be used
as a rule.
© Association for European Transport and Contributors 2011
9
The welfare optimal charges, given that the loss of revenues must be financed
by the state, can be derived by maximization of social surplus under budget
constraint. When solving the Lagrange function the expression for optimal
price, P*, can be rephrased as in equation (5).
(5)
where
In (5)
is optimal price,
is marginal costs, X is the demand for airport
services,
is the price elasticity and is the Lagrange multiplier. Since
is negative it is required that
to ensure that
. The
factor represents the reduction in social surplus when the requirement for
profit is increased by NOK 1. Hence, can be interpreted as the welfare loss
by taxation. If
then
and optimal charge will, according to (5), be
equal to marginal cost. A rephrasing of (5) can be made to show more clearly
the adjustment factor, F, for marginal costs under revenue restrictions as
presented in (6).
(6)
where
,
and
It is evident from (6) that F increase when λ increase (higher taxation costs)
and when
is reduced (lower price sensitivity). It can be derived from (6)
that the value of F is 1.26, 1.20 and 1.16 for price elasticities of -0.8, -1.0 and 1.2, respectively, given a welfare loss by taxation of
. Hence, assuming
a price elasticity of -1.0 and The Norwegian Ministry of Finance (2005)
recommendations of taxation costs, the optimal charges based on long-run
marginal costs and considering cost of raising public funds are NOK 65 for
PAX and NOK 569 + NOK 31 MTOW for ATM.
6. CONCLUSIONS AND IMPLICATIONS
This paper presents the results from a project initiated by the Norwegian
Ministry of Transport and Communication aiming to derive airport charges that
are more in line with principles of welfare economics. A multiple regression
model is applied to derive marginal costs for passengers (PAX) and air traffic
movements (ATM) being the two main production measures at airports. The
empirical data consists of accounting and production figures for all 46
Norwegian publicly owned airports for the years 2007 to 2009.
When disregarding the social cost of raising public funds, it is well recognized
that marginal cost pricing produces the highest social surplus and therefore is
desirable from a perspective of welfare economics. Under reasonable
assumptions regarding travelers’ price elasticity and using the Ministry’s
recommendation for costs of raising public funds, airport charges should be
raised by about 20% based on marginal costs when aiming to maximize the
airports’ social surplus. If the charges aim to generate the same revenues as
the current charge scheme the values should be raised by an even higher
factor. The derived values for the three different marginal cost based charge
© Association for European Transport and Contributors 2011
10
schemes are presented in Table 5 along with the current charge scheme. For
all marginal cost based charge schemes using raised values the adjustment
factor is similar for PAX and ATM. This simplification is made because there is
no evidence of how the adjustment factor should be differentiated.
Table 5 Charge schemes based on long-run marginal costs (MC) and current
charges.
ATM
PAX
Fixed
MTOW
Marginal cost (MC) pricing
54
474
26
MC pricing adjusted for welfare loss
65
569
31
MC pricing adjusted to 2009 revenue
85
748
42
Current charges (2009) *
44
0
96
* Valid for aircrafts with MTOW less than 100 tonne.
Two main changes are suggested by the marginal cost based charges
compared to the current scheme. First, there should be a shift towards
relatively higher charge put on PAX compared to ATM. For example, if
revenues should be held at the 2009 level but based on marginal costs the
ATM charge is NOK 1924 (748 + 28 × 42) compared to NOK 2688 (28 × 96)
based on the current charge scheme for an average departure of 28 tonne.
This is in line with IATA (2010) recommendations that, wherever possible and
practical, airports should transition to passenger based charges. The benefits
are related to improved accountability for passengers and lower risk for airline
companies. In isolation, such a change will make a higher proportion of the
carriers’ costs variable and, thereby, reduce the cabin factor required to
establish new routes.
Second, the ATM charge should be divided in two parts of which one element
depends on airplane maximum take of weight (MTOW) and another fixed
element is independent of flight weight. Such a separation is reasonable since
many factors related to serving planes at airports do not vary with MTOW.
Even though the data set contains objective information which is considered
reliable, there is potential for further improvements in the analysis if a more
detailed data set is made available. This particularly regards separating costs
related to the inside area (e.g. terminal) and the airside area (e.g. runway
system). Such data would enable us to run econometric regressions on inside
costs using PAX as independent variable and on airside costs using ATM as
independent variable. Analyses would then be more accurate and the problem
of multicorrelation would be omitted.
© Association for European Transport and Contributors 2011
11
REFERENCES
Avinor. (2011) Traffic statistics, http://www.avinor.no/en/avinor/traffic.
Carlsson, F. (2003) Airport marginal cost pricing: Discussion and an
application to Swedish airports. International Journal of Transport Economics,
30 (3) 283-304.
European Union. (2009) Directive 2009/12/EC of the European Parliament
and of the Council of 11 March 2009 on airport charges Official Journal of the
European Union, 52.
Graham, A. (2008) Managing Airports: an international perspective (3rd ed.).
Butterworth-Heinemann, Oxford.
Hakimov, R., and Scholz, S. (2010). Calculation of Ramsey Prices for German
Airports. GAP report, Berlin School of Economics and Law.
IATA. (2010) Passenger Based Airport Charges,
http://www.iata.org/whatwedo/airport-ans/charges/Documents/Passengerbased-airport-charges.pdf.
ICAO. (2009) ICAO’s Policies on Charges for Airports and Air Navigation
Services. Doc 9082.
Jørgensen, F., Mathisen, T., and Solvoll, G. (2011). Lufthavnavgifter i Norge takstsystemets struktur og betydning for tilbud og etterspørsel. (Airport
charges in Norway). SIB report 2/2011, Bodø Graduate School of Business,
Norway. In Norwegian.
Lian, J. I. (2010) Network dependency and airline competition Consequences for remote areas in Norway. Journal of Air Transport
Management, 16 (3) 137-143.
Martin-Cejas, R. R. (1997) Airport pricing systems in europe and an
application of Ramsey pricing to Spanish airports. Transportation Research
Part E: Logistics and Transportation Review, 33 (4) 321-327.
Martín, J. C., and Voltes-Dorta, A. (2008) International Airports: Economies of
Scale and Marginal Costs. Journal of the Transportation Research Forum, 47
(1) 5-22.
Martín, J. C., and Voltes-Dorta, A. (2011) The econometric estimation of
airports' cost function. Transportation Research Part B: Methodological, 45 (1)
112-127.
Pels, E., and Rietveld, P. (2000) Cost functions in transport. In K. Button & D.
A. Hensher (Eds.), Handbook of transport modelling, Vol. 1, 321–333.
Pergamon, Amsterdam.
Santos, G., and Robin, M. (2010) Determinants of delays at European
airports. Transportation Research Part B: Methodological, 44 (3) 392-403.
© Association for European Transport and Contributors 2011
12
The Norwegian Ministry of Finance. (2005). Veiledning i samfunnsøkonomiske
analyser. (Guide for welfare economic analyses). Oslo. In Norwegian.
Walters, A. A. (1965) The long and the short of transport. Bulletin of the
Oxford University Institute of Economics and Statistics, 27 (2) 97–101.
Williams, G., Fewings, R., and Fuglum, K. (2007) Airport provision and air
transport dependence in European countries. Journal of Airport Management,
1 (4) 398-412.
Wooldridge, J. M. (2006) Introductory econometrics: a modern approach (3rd
ed.). Thomson South-Western, Mason, OH.
© Association for European Transport and Contributors 2011
13
NOTES
1
Short Take-Off and Landing (STOL) planes such as Bombardier Dash-8 100
(or DHC-8) operate on local airports with runways as short as 800 meters.
Local airports with sufficiently long runways are operated by jet planes.
2
All taxes are charged once per trip. A trip (without transfer) includes two
flight movements and visits at terminals both at departure and landing. It is
however, not common to charge these services separately for the origin and
destination airports.
3
€ 1 ≈ NOK 8
4
See e.g. Walters (1965) for a further discussion of the application of
marginal cost concepts in transport.
5
Average MTOW is weighted by the traffic amount at the airports and,
thereby, calculated by dividing the sum of departed tonne for the whole
industry by the total number of departures. The average for 2007, 2008 and
2009 is 28349 kg which is approximated to 28 tonne in the analysis.
© Association for European Transport and Contributors 2011
14