Issue 22(2), 2022, pp. 224-233
https://doi.Org/10.18757/ejtir.2022.22.2.5773
TU Delft OPEN
EJTIRISSN: 1567-7141
http://ejtir.tudelft.nl/
The border effect in European air transport
Zdeněk Tomeš1
Faculty of Economics and Administration, Masaryk University, Czech Republic.
Vlastimil Reichel2
Faculty of Economics and Administration, Masaryk University, Czech Republic.
Štěpán Veselý3
Kiwi.com, Czech Republic.
This article deals with the analysis of the border effect i n
European air transport. The border effect measures how trade or
transport flows are diminished w h e n they cross a national border.
This topic has attracted a great deal of attention within trade, but
it is still little studied within transport. Existing studies have
estimated that the border effect diminished air passenger
transport flows by a factor of five to six. However, there have been
many changes i n the European economy and transport since those
studies. Our estimate based on a new data set suggested the
existence of a border effect i n European passenger air traffic flows,
albeit with a lower value of around two. A possible reason for the
lower value is the growing integration of the European economy
and the development of low-cost carriers i n Europe between 2000
and 2019. Our econometric analysis also found differences i n
border effects among European countries. N o significant border
effect was detected for France, but we d i d find high and significant
effects for Germany, Spain, and Poland. These differences can be
attributed to different intensities of intermodal competition on
domestic routes.
Publishing history
Submitted: 28 lune 2021
Accepted: 13 lune 2022
Published: 24 lune 2022
Cite as
Tomeš, Z., Reichel, V . &
Veselý, Š. (2022). The border
effect i n European air
transport. European lournal of
Transport and Infrastructure
Research, 22(2), 224-233.
© 2022 Zdeněk Tom, Reichel
Vlastimil, Veselý Štěpán
This work is licensed under a
Creative Commons
Attribution 4.0 International
License (CC BY 4.0)
Keywords: border effect, air transport, gravity model, liberalization.
1
A : Department of Economics, Lipová 41a, 602 00 Brno-Pisárky, E: tomes@econ.muni.cz
2
A : Department of Economics, Lipová 41a, 602 00 Brno-Pisárky, E: reichel.v@mail.muni.cz
3
A : Lazaretní 925/9, 615 00 Brno - Zábrdovice, E: stepan.vesely@kiwi.com
EJTIR 22(2), 2022, pp.224-233
Tomes, Reichel and Vesely
The border effect i n European air transport
225
1. Introduction
The border effect is an empirical regularity that affects both international trade and transport.
National borders create an obstacle that significantly diminishes mutual exchange. The seminal
paper i n the analysis of the border effect was one by M c C a l l u m (1995), who analysed trade patterns
among Canadian provinces and US states. Controlling for G D P , population, and mutual distance,
he concluded that, on average, Canadian provinces were 22 times more likely to trade with other
Canadian provinces than with US states. This revelation attracted a great deal of attention because
it seemed hard to believe that such an open border could create such a trade obstacle. M u c h
subsequent research has tried to confirm or refute this result. Despite significant advances i n
methodology and estimation techniques, the border effect is an important and significant
determinant of international trade flows (Havránek - Iršová 2017).
The literature on the border effect i n the field of international trade is very rich. In contrast, there
is only scarce evidence of border effects i n international transport, although the impact of borders
on transport flows has generally been recognized (Gerondeau 1997, Nash 2015). The evidence on
the border effect i n land passenger transport is limited, with scarce evidence from Rietveld (2012).
More evidence can be found within air transport, where two specialized studies have emerged
(Klodt 2004, Hazledine 2009) with estimates for the border effect lying i n the range of five to six.
There has been a conjecture that the border effect may actually have become smaller over the past
two decades. The reason for this conjecture is the development of low-cost carriers, which have
opened and developed many new international connections (Calzada - Fageda, 2019). In addition,
the liberalization of air transport i n Europe (Mason et al. 2016) may have stimulated a decline i n
the border effect. Anecdotal evidence that can support this conjecture can be derived from Eurostat
figures, where from 2005 to 2017 the number of passengers at the top five international routes i n
European passenger air transport grew by 45% while the top five national routes declined by 11%
(EU 2005, 2017). The aim of the present paper is to estimate what the value of the border effect i n
European air transport was i n 2019. It further aims to analyse whether border effects can be
different for different countries.
The methodology of the paper utilizes the standard augmented gravity model. The explained
variable is the seat capacity on European air routes i n November 2019, regressed on G D P ,
population, distance, and a border dummy. In addition, other control variables were utilized, such
as common language and tourist destinations. The structure of this paper is as follows. It starts
with a literature review that analyses the border effect i n trade and transport. Section 3 describes
the methodology and data that we utilized. Section 4 includes results and interpretations, and
Section 5 concludes.
2. Literature Review
The question of h o w much national borders change the volume of trade has been widely analysed.
This flow of literature started with a seminal paper by M c C a l l u m (1995), who investigated the
border effect i n the context of the Canada-US border. M c C a l l u m utilized data on the trade of
Canadian provinces with other Canadian provinces and US states. His methodological approach
was to use a gravity equation i n log-linearized form, controlling for G D P and distance. A d u m m y
variable for domestic trade (Canadian interprovincial) was estimated as 3.09, which, after
exponentiating,4
generated a border effect of 22. This was an unexpectedly large value. M c C a l l u m
documented these effects through the example of the gravity model predicting that Ontario and
Quebec should export about 10 times as much to California as to British Columbia. The distance is
roughly the same, but Californian G D P is 10 times that of British Columbia. In reality, however,
4edom<stk = border effect; e3
-0 9
= 21.98
EJTIR 22(2), 2022, pp.224-233
Tomes, Reichel and Vesely
The border effect i n European air transport
226
both Ontario and Quebec exported more than three times as much to British Columbia as to
California (McCallum, 1995).
These results were surprising and i n contradiction to economic intuition. This paper started a long
stream of literature that attempted to verify, explain, or contradict the border effect. A review by
Havránek and Iršová (2017) stated that, based on an analysis of 61 studies dealing with the border
effect i n international economics, the average estimated value for the domestic d u m m y was 2.7,
generating a border effect as high as 15. This finding means that regions have been 15 times more
likely to trade with a region i n the same country than with foreign regions (controlling for G D P ,
distance, and other factors; Havránek - Iršová 2017).
A great deal of attention has been devoted to the issue of w h y the border effect exists at all. Head
and Mayer (2013) argued that imperfect information, very localized tastes, and distribution
networks are the main reasons. A different explanation has been offered by Guiso et al. (2009).
They concentrated on how cultural biases affected economic exchange with the help of data about
bilateral trust between European countries. They concluded that there were still important cultural
barriers i n Europe with a powerful influence from a shared language and religion. De Groot et al.
(2004) added that countries with similar institutional quality levels may have been familiar with
one another, which reduced transaction costs and the border effect.
Surprisingly, the border effect has not been analysed much within transport economics. Zijlstra
(2020) examined the border effect i n airport choice. He documented that air travellers were
reluctant to use a foreign airport to depart for their trips. A n analysis of the border effect i n
passenger land transport has been provided i n Rietveld (2012). H e analysed the barrier effects of
transport borders and the implications for transport infrastructure. He noted that borders
discouraged spatial interactions and that although the border effect i n Europe has declined over
the previous 15 years, it remained substantial. H e defined five barrier effects from borders i n
transport: preferences, public sector regulations, institutions, information, and transport costs.
Depending on the type of infrastructure, he estimated Europe's border effect for passenger land
transport i n the range of 1.4-3.3.
There have been two econometric studies dedicated to this topic, namely Hazledine (2009) and
Klodt (2004). Hazledine (2009) analysed national and international departures from five Canadian
airports and estimated whether there were differences between destinations within Canada and
those abroad. His methodology was an augmented gravity model of passenger air travel that
utilized data for 212 non-tourist goals. Hazledine proxied the number of passengers by the number
of seats. After controlling for city population, G D P per capita, and distance, he estimated the
domestic coefficient's value as 1.8, yielding a border effect as high as 6. The second study dealing
with the border effect was by Klodt (2004), who analysed the ridership numbers from German
airports to domestic and international destinations. He estimated a border effect on the level of 3-
5. A n important factor was that a common language at origin and destination diminished the
border effect. The contemporary literature i n the area of air transport suggests that national borders
have remained an important factor i n the determination of transport flows (Dobruszkes 2021).
The development of the border effect may be able to help us to understand better the broader
changes i n the European and world economy. The ongoing liberalization of air transport has
stimulated air ridership. In comparison with the values from land passenger transport and trade,
the border effect i n air transport seems to be the lowest. This inference implies that international
air passenger travel has borne the lowest transaction costs and could be a primary source of
international trade initiatives. W e can summarize that borders have mattered significantly i n both
trade and transport. There is a great deal of evidence within trade, but the evidence on transport
has been limited. This paper aims to contribute to this topic with an analysis of European air data
from 2019.
EJTIR 22(2), 2022, pp.224-233
Tomes, Reichel and Vesely
The border effect i n European air transport
227
3. Methods and Data
The paper involves the standard augmented gravity model of transport. The gravity model is a
standard tool for estimating trade relationships (Anderson 2011, Shepherd 2013). The model has
been very successful i n empirical practice, but it has been criticized for having weak microfoundations.
This disadvantage has been recently overcome by advances i n theory (Head - Mayer
2014). There have also been improvements i n the estimation of gravity models (Matyas 1997,
Anderson - van Wincoop 2003, Silva - Tenreyro 2006). In addition, the gravity model is utilized
within air transport as a way to estimate origin-destination relationships, although usually without
explicitly controlling for the border effect (Grosche et al. 2007, Matsumoto 2007, Zhang - Zhang
2016).
In accordance with previous studies, the following model specifications were estimated:
Seatsi = B0 + £;
5
=1Bj cvjX + p6DOMESTICt + et, (1)
SeatSi = B0 + E;=i/?j cvjX + B6D0MESTICi + B7LANGUAGEi + eu (2)
SeatSi = B0 + £)= 1 Bj cvjX + 06DOMESTICi + B7LANGUAGEi + BJOUMSTi + eu (3)
Seatst = B0 + £ ) = 1 t y cvjX + B7LANGUAGEt + BsTOURlSTt + ZkUPk cdKi + et. (4)
The models were estimated i n a standard log-linearized form. A l l of our model specifications
included control variables (cvj) such as the logarithm of G D P per capita, the logarithm of the
population i n both the origin and destination cities, and the logarithm of the length i n kilometres
between city airports. The variable L E N G T H was defined as the geodetic distance between
airports. The geodetic distance between any two points is longer than the straight -line distance
between the same two points. The geodetic distance should therefore better represent the length of
the flight. This formed the gravity model's core with attracting (GDP and population) a n d
distracting (distance) forces. Next, a d u m m y for domestic flights was added to measure the border
effect (equations 1-3). The further controls that have been included i n the model are dummies for
a common language (equation 2), tourist destinations (equation 3), and country (cdf e ) to measure
the border effect for each country (equation 4). The variable L A N G U A G E was defined as 1 when
the official language of the country from the destination matched the official language of the
country from the origin. The variable TOURIST was defined as capturing discrepancies between
the population and the number of flights i n a given location. This definition corresponded to tourist
destinations not i n proximity to a large city. These locations are islands that are not accessible by
land transport (roads and railways are not possible). The country variables were 1 when the flight
was a domestic flight within the given country. The selection of the dependent variable was an
important choice. We used the variable SEATS as a proxy for the total number of passengers o n
the route, which aligns with the existing literature (Hazledine 2009, 2017). The number of seats can
be a suitable proxy for passenger numbers because it is less prone to fluctuations than the number
of passengers is. Table 1 provides a list of model variables.
The principal source of data about traffic flows was a Kiwi.com database. Kiwi.com is an online
travel agency specializing i n finding itineraries i n air transport services to decrease the final price.
The Kiwi.com database of flights is quite extensive, covering almost all existing connections. W e
have utilized access to this unique database to obtain individual data about origin-destination
flows. W e chose all non-zero traffic flows that originated or terminated i n one of four selected
countries: Spain, France, Germany, and Poland. These countries were chosen because they form a
land block i n Europe, and there is therefore a realistic option for intermodal competition against
air. A t the same time, these are large countries that have significant domestic and international air
passenger traffic. The U K was excluded because of Brexit. The total number of unique destinations
was 217, and the total number of observations (connections) i n our data set was 1,574. Data were
EJTIR 22(2), 2022, pp.224-233
Tomes, Reichel and Vesely
The border effect i n European air transport
228
collected over three days i n November 2019, specifically from Thursday to Saturday (28-30
November). November was chosen as a typical non-holiday month. The seat capacity was
estimated as the frequency multiplied by the seat capacity of the plane used on the given route.
Figure 1 shows a unique map depicting the number of domestic and international seats for the
monitored airports. The figure captures the noticeable differences among countries. For example,
Spain showed a predominance of domestic flights, whereas Poland showed a predominance of
foreign flights. Interestingly, there were also areas with disproportionately more total seats than
their importance would suggest (e.g. the islands of Ibiza and Mallorca). These locations were
characterized by a pre-dominance of domestic flights. For this reason, we found it reasonable to
control for these factors (countries, specific locations) within the estimated regression models.
Our data set included only flights that originated or terminated i n one of our four selected
countries. Destinations were limited to the E U , Switzerland, Norway, and Iceland. Other n o n - E U
countries and overseas departments were not included. Collecting data i n November meant that
only a minimal number of tourist destinations were among the destinations. The data contains
some of them, including the almost exclusively Spanish airports of Lanzarote, Fuerteventura, Gran
Canaria, Tenerife, Ibiza, Malila, Menorca, and Mallorca. There was no set m i n i m u m flight length;
the data were unidirectional and only direct flights were included. W e acknowledge that there are
significant limitations to the data set. The data was collected only for three days i n the autumn. W e
did not have at our disposal origin-destination data that w o u l d enable us to deal with transfer
traffic or the issue of hub airports and connecting flights.
Table 1. Model variables
Description Unit Source
Dependent variable
SEATS Number of available seats on the route. Seats Kiwi.com
Independent variables - control variables
G D P 1 Regional GDP per capita at origin. USD; thousands Eurostat; O E C D
P O P 1 Population at origin. People; thousands Eurostat; O E C D
G D P 2 Regional GDP per capita at destination. USD; thousands Eurostat; O E C D
POP 2 Population at destination. People; thousands Eurostat; O E C D
L E N G T H Geodetic distance between origin and destination. k m Kiwi.com
Independent variables - dummy variables
DOMESTIC = 1 when flight was domestic. Binary
L A N G U A G E = 1 when the main language at destination was the
same as at origin.
Binary
TOURIST = 1 when there was a discrepancy between the number
of flights and the population in a location.
Binary
F R A N C E = 1 when there was a domestic flight in France. Binary
P O L A N D = 1 when there was a domestic flight in Poland. Binary
G E R M A N Y = 1 when there was a domestic flight in Germany. Binary
SPAIN = 1 when there was a domestic flight in Spain. Binary
EJTIR 22(2), 2022, pp.224-233
Tomes, Reichel and Vesely
The border effect i n European air transport
229
4. Results and Discussion
Table 2 presents the results of our estimation. M o d e l 1 is the baseline regression, Models 2 and 3
controlled for a common language and tourist destinations. M o d e l 4 distinguished the border effect
according to country (France, Germany, Spain, Poland). The coefficients for G D P , population, and
EJTIR 22(2), 2022, pp.224-233
Tomes, Reichel and Vesely
The border effect i n European air transport
230
length had the expected signs, and their values were i n line with previous studies. The domestic
coefficient i n Models 1-3 was between 0.52 and 0.74, which after exponentiating yielded a border
effect i n the range of 1.7-2.1, which was both statistically significant from zero and less than had
been identified i n previous studies. The border effect was smaller w h e n the estimation controlled
for a shared language at the origin and destination (Models 2-4) and w h e n it controlled for tourist
destinations (Models 3-4).
Table 2. Regression results
M o d e l 1 Model 2 Model 3 M o d e l 4
OLS OLS O L S OLS
In SEATS In SEATS In SEATS In SEATS
In G D P 1 0.35*** 0.32*** 0.32*** 0.38***
(0.05) (0.06) (0.06) (0.06)
In POP1 0.45*** 0.46*** 0.47*** 0.46***
(0.03) (0.03) (0.03) (0.03)
In G D P 2 0.49*** 0.46*** 0.45*** 0.48***
(0.04) (0.05) (0.05) (0.05)
In POP2 0.28*** 0.29*** 0.30*** 0.30***
(0.02) (0.02) (0.02) (0.02)
In L E N G T H -0.15*** -0.14*** -0.15*** -0.12**
(0.05) (0.05) (0.05) (0.05)
D O M E S T I C 0.74*** 0.56*** 0.52***
(0.08) (0.11) (0.11)
L A N G U A G E 0.22** 0.23** 0.26**
(0.10) (0.10) (0.11)
TOURIST 0.28***
(0.10)
0.18*
(0.10)
F R A N C E 0.14
(0.14)
P O L A N D 0.84***
(0.18)
G E R M A N Y 0.61***
(0.17)
S P A I N 0.72***
(0.15)
const -1.30** -1.35** -1.37** -1.77***
(0.59) (0.59) (0.58) (0.60)
Observations 1,574 1,574 1,574 1,574
R squared 0.371 0.373 0.376 0.383
I n L -2,009 -2,007 -2,003 -1,994
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
W h y were our values lower than those from the previous studies by Hazledine (2009) and Klodt
(2004)? O u r lower value for the domestic coefficient (0.52-0.74) i n comparison with Hazledine's
(1.59-1.84) may have been due to differences between North American and European geography
and the air transport market where distant Canadian locations may stimulate the border effect
(Anderson - van Wincoop, 2003). The more relevant comparison is with the study by Klodt (2004),
which used departures from German airports to other European destinations. W e identified a
much lower domestic coefficient even compared to Klodt's values (1.02-1.60). Part of the variation
EJTIR 22(2), 2022, pp.224-233
Tomes, Reichel and Vesely
The border effect i n European air transport
231
between our results and previous results may be related to differences i n research methodology.
The differences may be partly due to the definition of the observed flights included i n the data set
(see Hazledine, 2009) or i n the slightly different specifications of the gravity models used. Although
there were differences i n the form of the specification and the selection of origin airports, we think
that the main reason for the lower values for the border effect was i n the real markets.
The enlargement of the E U and the Schengen Area has stimulated the growth of cross-border
traffic. Increased economic integration and the free movement of people has encouraged migration
and tourism within the E U . These developments could have stimulated a decline i n Europe's
border effect during 2000-2019. The growth i n demand was also reflected on the supply side when
low-cost carriers opened new air routes more often for international traffic than for European
national traffic (Calzada - Fageda 2019). Therefore, we interpreted our estimations as indicating an
actual decline i n Europe's border effect over the past 20 years. However, the border effect is
expected to change again as a result of the current COVID-19 pandemic and the military conflict
in Ukraine. Total air passenger transport has declined spectacularly after 2020. This fall has also
been unequal, with international traffic falling more sharply than national traffic. This
development has probably significantly changed the border effect after 2020. The structure and
dynamics of this development is a fruitful topic for further research.
M o d e l 4 included four individual country variables to measure the differences i n the border effect
among France, Germany, Poland, and Spain. The domestic d u m m y was insignificant i n France and
highly significant i n Germany, Poland, and Spain. W e thus d i d not find any border effect at all for
France. H o w could this be possible? W e think that the intense competition of high-speed rail with
airlines on France's domestic lines effectively nullified France's air transport border effect. H i g h speed
rail has also been developed i n Spain and Germany, but the frequencies for major
destinations have been lower for Spain and Germany than they have for France. This development
can be documented i n many examples where national air transport passenger flows diminished
significantly due to the opening of national high-speed rail lines (Dobruszkes et al. 2014). The broad
development of the high-speed rail network i n France could be a major source for not finding a
significant border effect for France.
Table 3. Border effect results
M O D E L / P A P E R G E O G R A P H Y DOMESTIC BORDER
D U M M Y EFFECT
Model 1 FR+PL+DE+ES 0.74 2.10
Model 2 FR+PL+DE+ES 0.56 1.75
Model 3 FR+PL+DE+ES 0.52 1.68
F R A N C E - P
O L A N D 0.84 2.32
G E R M A N Y 0.61 1.84
Model 4 SPAIN 0.72 2.05
To control for idiosyncrasies i n our estimations, we performed some robustness checks. W e have
experimented with other variables, checked the results for sub-samples, and played with different
specifications. The results of these robustness checks were satisfactory because they d i d not alter
the basic results of our estimations. A s for the limitations of our approach, the most significant are
the following. The first is the concentration on the number of available seats, which was chosen as
the explained variable. It w o u l d be interesting to compare the results with passengers as the
dependent variable. Such data could be extracted from the International A i r Transport Association
or Eurostat and could be an interesting topic for further research. Secondly, we have not controlled
for many other factors that can influence the border effect and affect its value. O n the other hand,
EJTIR 22(2), 2022, pp.224-233
Tomes, Reichel and Vesely
The border effect i n European air transport
232
we have used a similar specification as i n the previous studies by Klodt and Hazledine. Thirdly,
we were unable to differentiate among direct and connecting flights due to limitations i n our data
sets, and we therefore could not work with origin-destination data. Fourthly, due to our data
limitations, we were not able to take into account the issue of cross-border airports that can attract
passengers from neighbouring countries.
5. Conclusion
This article aimed to estimate the border effect i n European air passenger transport i n 2019. W e
estimated its level as 1.68-2.32, meaning that national air passenger flows i n Europe were roughly
two times as high as the international flows for matching pairs of cities. Another significant result
from our estimation was that the border effect was very different for different European countries.
We found no border effect for France and significant border effects for Germany, Spain, and
Poland. W e have interpreted this as the result of differences i n the level and intensity of intermodal
competition on the national market, especially between air and high-speed rail. The current
COVID-19 pandemic w i l l undoubtedly influence the level of the border effect. The spectacular fall
in air transport traffic has been disproportional between domestic and international traffic, and
very probably the COVID-19 crisis w i l l have a lasting impact on travel behaviour. Therefore, it can
be expected that the border effect w i l l go up i n the upcoming period.
Acknowledgements
This article was supported by the project N e w Mobility - High-Speed Transport Systems and TransportRelated
Human Behaviour, Reg. No. CCZ.02.1.01/0.0/0.0/16_026/0008430, co-financed by the
Operational Programme Research, Development and Education.
References
Anderson, J. E. (2011). The gravity model. Annual Review of Economics, 3(1), 133-160.
Anderson, J. E. and van Wincoop, E. (2003). Gravity with gravitas: A solution to the border puzzle.
American Economic Review, 93(1), 170-192.
Calzada, J. and Fageda, X. (2019). Route expansion in the European air transport market. Regional
Studies, 53(8), 1149-1160.
De Groot, H . L., Linders, G. J., Rietveld, P. and Subramanian, U . (2004). The institutional determinants
of bilateral trade patterns. Kyklos, 57(1), 103-123.
Dobruszkes, F., Dehon, C. and Givoni, M . (2014). Does European high-speed rail affect the current level
of air services? A n EU-wide analysis. Transportation Research Part A: Policy and Practice, 69, 461-
475.
Dobruszkes, F. (2021). A global business? Mapping the densest passenger airline routes. Journal of
Transport Geography, 92.
E U (2005). EU transport in figures 2005.
E U (2017). EU transport in figures 2017.
Gerondeau, C. (1997). Transport in Europe. Artech House.
Grosche, T., Rothlauf, F. and Heinzl, A . (2007). Gravity models for airline passenger volume estimation.
Journal of Air Transport Management, 23(4), 175-183.
Guiso, L., Sapienza, P. and Zingales, L. (2009). Cultural biases in economic exchange?. The Quarterly
Journal of Economics, 224(3), 1095-1131.
EJTIR 22(2), 2022, pp.224-233 233
Tomes, Reichel and Vesely
The border effect i n European air transport
Havranek, T. and Irsova, Z. (2017). Do borders really slash trade? A meta-analysis. IMF Economic Review,
65(2), 365-396.
Hazledine, T. (2009). Border effects for domestic and international Canadian passenger air travel, journal
of Air Transport Management, 25(1), 7-13.
Hazledine, T. (2017). A n augmented gravity model for forecasting passenger air traffic on city-pair
routes, journal of Transport Economics and Policy (JTEP), 52(3), 208-224.Hazledine (2014)
Head, K. and Mayer, T. (2013). What separates us? Sources of resistance to globalization. Canadian
Journal of Economics/Revue canadienne d' economique, 46(4), 1196-1231.
Head, K. and Mayer, T. (2014). Gravity equations: Workhorse, toolkit, and cookbook. In Handbook of
international economics (Vol. 4, pp. 131-195). Elsevier.
Klodt, H . (2004) Border effects in passenger air traffic. Kyklos, 57(4), 519-532.
Mason, K., Morrison, W . G . and Stockman, I. (2016). Liberalization of air transport i n Europe and the
evolution of low-cost' airlines. In Liberalization in Aviation: Competition, Cooperation and Public
Policy; Forsyth, P., Gillen, D., Huschelrath, K., Niemeier, H M , Wolf, H . eds. (pp. 141-156).
Routledge.
Matsumoto, H . (2007). International air network structures and air traffic density of world cities.
Transportation Research Part E: Logistics and Transportation Review, 43(3), 269-282.
Matyas, L. (1997). Proper econometric specification of the gravity model. World Economy, 20(3), 363-368.
McCallum, J. (1995). National borders matter: Canada-US regional trade patterns. The American Economic
Review, 85(3), 615-623.
Nash, C. (2015). When to invest in high speed rail. Journal of Rail Transport Planning & Management, 5(1),
12-22.
Rietveld, P. (2012). Barrier effects of Borders: Implications for border crossing infrastructures. European
Journal of Transport and Infrastructure Research, 22(2), 150-166.
Silva, J. S. and Tenreyro, S. (2006). The log of gravity. The Review of Economics and Statistics, 88(4), 641-
658.
Shepherd, B. (2013). Gravity model of international trade: A user guide. United Nations Economic and Social
Commission for Asia and the Pacific.
Zhang, Y. and Zhang, A. (2016). Determinants of air passenger flows i n China and gravity model:
Deregulation, LCCs, and high-speed rail. Journal of Transport Economics and Policy (JTEP), 50(3),
287-303.
Zijlstra, T. (2020). A border effect i n airport choice: Evidence from Western Europe. Journal of Air
Transport Management, 88,101874.