228
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
ASSESSING DEMAND AND
SUPPLY SUBSTITUTABILITY
IN BRAZILIAN
INTERNATIONAL AVIATION:
A NETWORK THEORY
APPROACH FOR ANTITRUST
ANALYSIS
1
Avaliação da substituibilidade de demanda e oferta na aviação
internacional brasileira: uma abordagem da teoria das redes
para análise antitruste
Eduardo Roberto Zana
2
Conselho Administrativo de Defesa Econômica (DEE/Cade) Brasília/DF, Brasil
Gabriel Oliveira de Alarcão
3
Conselho Administrativo de Defesa Econômica (DEE/Cade) Brasília/DF, Brasil
Tomás de Siervi Barcellos
4
Conselho Administrativo de Defesa Econômica (DEE/Cade) Brasília/DF, Brasil
1 Editor responsável: Prof. Dr. Victor Oliveira Fernandes, Conselho Administrativo de Defesa Econômica (Cade), Brasília,
DF, Brasil.
Lattes: http://lattes.cnpq.br/5250274768971874. ORCID: https://orcid.org/0000-0001-5431-4142.
1 Recebido em: 16/01/2025 Aceito em: 11/06/2025 Publicado em: 25/06/2025
2 Doutor em Economia da Indústria e da Tecnologia pela Universidade Federal do Rio de Janeiro (UFRJ). Especialista
em Regulação de Petróleo e Derivados, Álcool Combustível e Gás Natural. Coordenador-substituto de Análise de Atos de
Concentração no Departamento de Estudos Econômicos (DEE) no Cade.
E-mail: eduardo.zana@cade.gov.br Lattes: http://lattes.cnpq.br/4538077971564059
ORCID: http://orcid.org/ 0009-0004-3107-5096
3 Bacharel e mestrando em Economia pela Universidade de Brasília (UnB). Chefe de Serviço no Departamento de
Estudos Econômicos (DEE) do Cade.
E-mail: gabriel.alarcao@cade.gov.br Lattes: http://lattes.cnpq.br/4129959594619650
ORCID: http://orcid.org/0009-0008-3055-0632
4 Bacharel em Ciências Econômicas pela Universidade Federal de Santa Catarina (UFSC) e mestre em Ciências Sociais
pela Universidade de Brasília (UnB). Economista do Ministério da Agricultura, Pecuária e Abastecimento (Mapa). Coordenador
de Análise de Atos de Concentração no Departamento de Estudos Econômicos (DEE) no Cade.
E-mail: tomas.barcellos@cade.gov.br Lattes: http://lattes.cnpq.br/2660479477078148
ORCID: http://http://orcid.org/0009-0004-7041-6279
11
229
STRUCTURED SUMMARY
Context: The definition of the relevant market in the international aviation sector varies among
antitrust authorities, with some focusing on origin-destination (O&D) routes and others incorporating
broader or narrower criteria. While the European Commission emphasizes the O&D approach and
acknowledges the impact of the hub-and-spoke model on supply-side substitutability, debates
persist regarding the role of network eects and the substitutability of direct and indirect flights.
Objective: The aim of this article is to contribute to the debate by applying quantitative methods
based on network theory to identify the existence of demand and supply subs- titutability between
international routes departing from or arriving in Brazil, enabling a more accurate analysis in mergers
and acquisitions processes aecting the Brazilian international aviation market.
Method: The adoption of quantitative methods (utilizing network indicators) alongside qualitative
approaches to analyze substitutability on both the demand and supply sides within the international
aviation sector.
Conclusions: This study demonstrates the importance of adopting a network-based approach to
evaluate substitutability in the international aviation sector. By analyzing both demand and supply-
side factors, it provides insights into how route connectivity and airline behavior influence market
dynamics. The findings underline the necessity of incorporating network dynamics into antitrust
evaluations, oering a foundation for further research into the competitive implications of route
structures and airline strategies.
Keywords: aviation; network; antitrust; competition; hub-and-spoke; merger.
RESUMO ESTRUTURADO
Contexto: a definição do mercado relevante no setor de aviação internacional varia entre as autoridades
antitruste, com algumas focando nas rotas de origem-destino (O&D) e outras incorporando critérios
mais amplos ou mais restritos. Enquanto a Comissão Europeia enfatiza a abordagem O&D e reconhece
o impacto do modelo hub-and-spoke na substitutibilidade pelo lado da oferta, persistem debates
sobre o papel dos efeitos de rede e a substitutibilidade entre voos diretos e indiretos.
Objetivo: o objetivo deste artigo é contribuir para o debate aplicando métodos quantita- tivos baseados
na teoria de redes para identificar a existência de substitutibilidade pelo lado da demanda e da oferta
entre rotas internacionais com origem ou destino no Brasil, possibilitando uma análise mais precisa
em processos de fusões e aquisições que impactam o mercado de aviação internacional brasileiro.
Método: a adoção de métodos quantitativos (utilizando indicadores de rede) em conjunto com
abordagens qualitativas para analisar a substitutibilidade tanto do lado da demanda quanto do lado
da oferta no setor de aviação internacional.
Conclusões: este estudo demonstra a importância de adotar uma abordagem baseada em rede para
avaliar a substituibilidade no setor de aviação internacional. Ao analisar tanto os fatores de demanda
quanto de oferta, fornece insights sobre como a conectividade de rotas e o comportamento das
companhias aéreas influenciam a dinâmica do mercado. Os resultados ressaltam a necessidade de
incorporar a dinâmica das redes nas avaliações antitruste, oferecendo uma base para futuras pesquisas
230
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
sobre as implicações competitivas das estruturas de rotas e das estratégias das companhias aéreas.
Palavras-chave: aviação; rede; antitruste; competição; hub-and-spoke; fusão.
JEL Classification: [L40]
Summary: 1. Introduction; 2. Literature Review; 3. Data;
4. Demand Substitutability; 4.1 Methodology; 4.2 Results;
4.2.1 Link’s Importance and Eective Paths; 4.2.2 Edge
Removal; 4.1 Practical Example; 5. Supply Substitutability;
5.1 Qualitative Analysis; 5.2 Quantitative Analysis.;
Conclusion; References.
1 INTRODUCTION
The liberalization of international air transport markets and the ongoing consolidation
of the sector have brought renewed attention to the challenges of applying antitrust principles in
the aviation industry. Regarding the international jurisprudence applied to the aviation sector, the
Organisation for Economic Cooperation and Development (OECD)(2014) clarifies that there is no
consensus among antitrust authorities worldwide concerning the definition of the relevant market
in its geographical dimension. While some jurisdictions maintain the traditional definition based
on the origin and destination (O&D) approach, essentially focused on demand-side substitutability,
others have adopted broader definitions (including other modes of transportation) or narrower ones
considering the evolution of the civil aviation sector. However, there is no consensus on the need to
incorporate network eects in the process of competitive market analysis, with Australia considering
this factor for relevant market definition purposes (OECD, 2014), while other jurisdictions prefer to
incorporate network eects as part of the competitive eects analysis. As, unwittingly, observed
by the Directorate-General for Competition in the assessment of the Star Alliance case (European
Commission, 2013), a rigid demand-side definition of the relevant market when network industries
are in play may be myopic. Accepting out-of-market eciencies, even to the limited extent that they
accrue to the route of concern, in a way extends the boundaries of the relevant market. Ducci (2016),
in a paper on out-of market eciencies, two-sided platforms and consumer welfare, points out
the limitations of market definition when the economics of two-sided platforms are not accounted
for. Despite these considerations, there is debate over whether the best approach is to define on a
case-by-case basis rather than seeking clearer guidance and harmonization (OCDE, 2014). From the
perspective of the European Commission, as expressed in the Luthansa/Austrian Airlines case (Lenoir,
2016), the relevant market definition for passenger air transport services is understood inside O&D
approach, meaning that a particular route is seen as a separate market to be analyzed. However, to
determine whether a specific route between the origin and destination points should be considered,
the European antitrust body examines the various possibilities available.
Nonetheless, the European Commission considers the hub-and-spoke structure as an element
that, in practical terms, reduces supply-side substitutability. In theory, it is assumed that an aircrat
can be allocated to any route. However, in reality, given that traditional airlines use the hub-and-
spoke model to structure their flight operations, this means that traditional airlines decide to fly
almost exclusively on routes connected to their respective operating hubs (Lenoir, 2016). This tends
231
to be even more important on long-haul international routes, since this segment is dominated by
network air carriers (Lykotrafiti, 2023).
Moreover, according to Nannes (1999, p. 6-7), the antitrust analysis of entry conditions in
the commercial aviation sector has become “more sophisticated and substantially more factual”
over time. For example, in the 1980s, the Department of Transportation (DOT) approved a series of
transactions involving carriers with high market shares of city-pair trac, based on the reasoning that
other carriers could easily enter those routes and discipline fares, without adequately considering
whether new entry would be economically feasible given trac flows and hub economics (Nannes,
1999). In our present context, the delivery of wide-body aircrat by the major global manufacturers
(Boeing and Airbus) at a pace insucient to meet the post-pandemic surge in demand (Singh, 2024)
creates additional challenges for contesting the market in cases where incumbents exercise market
power on routes with high market concentration.
However, the European Commission’s mindset previously presented has been criticized
by network air carriers because, from their perspective, the traditional approach fails to consider
properly the network eects on demand side derived from the hub-and-spoke model structuring.
Several companies have emphasized the need for antitrust analysis to consider all airports that
are considered substitutes from the passengers’ perspective (European Union General Court, 2015).
Regarding substitutability between direct and indirect flights, the Commission found that it depends
on the flight duration. That is, the longer the flight duration, the greater the likelihood that indirect
flights exert competitive pressure on direct flights (European Union General Court, 2015). As for
the jurisprudence of Conselho Administrativo de Defesa Econômica (Cade), the relevant passenger
transport market has been geographically defined, like the European Union, based on origin and
destination routes
5
. Based on this analytical methodology, for long-haul routes, indirect flights that
increase travel time by up to 5 hours should also be considered part of the same relevant market.
Taking all of this into account, network structures can give rise to two distinct types of eects,
each with dierent implications for competition - eects that might be overlooked in traditional
analyses. On one side, a merger can intensify the dominance of a network carrier at key hub airports,
particularly those facing congestion. In such cases, the concentration of slots and routes linked to the
hub may strengthen the merged entity’s market power within the airport’s catchment area, reducing
consumer choice and harming competition. On the other hand, network eects mean that the outcome
of an airline merger goes beyond simply combining their existing routes. In this case, the whole is
much greater than the sum of the parts. Incorporating the target airline’s routes not only allows the
acquiring carrier to operate those services but also opens the possibility of launching new direct
or connecting routes. This can enhance the overall network, potentially generating eciencies that
might - though not necessarily - translate into lower prices and improved service quality (Olmedo
Peralta, 2020).
The aim of this article is to contribute to the debate by applying quantitative methods
based on network theory to identify the existence of demand and supply substitutability between
international routes departing from or arriving in Brazil, enabling a more accurate analysis in
mergers and acquisitions processes aecting the Brazilian international aviation market. This
5 Case No. 08700.004702/2023-81 (SEI 1377008). All Cade public proceedings mentioned in this article can be consulted
at: https://x.gd/0OMZL
232
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
article is organized as follows: Section 2 presents a literature review, summarizing key studies and
theoretical frameworks that underpin the research. Section 3 describes the data, outlining the data
collection. Section 4 presents the analysis from the demand side. It is presented the methodology
implemented (4.1), the general results (4.2) and then its application is shown with an example (4.3).
Section 5 presents a discussion on supply substitutability, incorporating both qualitative aspects
(5.1) and quantitative metrics (5.2) applied to understand patterns of network expansion before and
ater the pandemic crisis. Finally, Section 6 concludes the article, summarizing the contributions and
suggesting directions for future research.
2 LITERATURE REVIEW
Our work relates to the use of network to study the Brazilian air transportation sector. Rocha
(2009) studies the structural evolution of the Brazilian domestic airport network. Using data from 1996
to 2006, the results suggest that the companies have a tendency to invest in the most profitable routes
rather than in new routes, consequently, increasing the number of connections on specific routes. The
number of routes together with the number of airports decrease during the period, but the routes are
constantly changing and not necessarily within the most connected airports. This dynamic evolution
resulted in some airports becoming more central with time, while others become more peripheral.
The research in Couto et al. (2015) analyzes the main characteristics of the Brazilian air transportation
network, using national and international flight data using the complex network approach. The study
showed that the Brazilian network has small world characteristics with low average shortest path
length and high clustering coecient and the airport connections follow a power law distribution.
According to the authors, the main airports are Viracopos and Guarulhos, and travelers need to go
through an average of three connecting flights to reach their destinations. A resilience analysis of the
network’s robustness also identified that an interruption at Viracopos airport would divide the network
into six sub-networks, aecting 10% of passenger demand. Silva et al. (2022) investigates how the
structure of the air transportation network aects air ticket prices in Brazil. Using microeconometric
panel-data estimation, they find that the airports’ degree of substitutability (clustering coecient)
and peripheral location (authority score) are associated with lower average airfare prices. In contrast,
the convenience of transportation (degree), measured by the number of dierent cities that an airport
serves, and centrality (closeness) are attributes that raise the average price. Also, they find that
network’s topological characteristics can either amplify or mitigate the relationship between market
competition and airline ticket pricing.
Exploring the importance of links in transportation networks has a growing literature,
especially in road networks (Jenelius, 2010; Rupi et al., 2015; Li et al., 2020) and metro networks (Yang et
al., 2017; Jing; Xu; Pu, 2019). Jing, Xu, and Pu (2019) provides a new dimension of assessing metro network
performance—travelers’ route redundancy (or route diversity), which is defined as the number of
behaviorally eective routes between each origin-destination (O-D) pair in the network. Furthermore,
the results of route redundancy are compared with typical measures of topology network performance
in terms of measuring connectivity and accessibility of metro networks. Their dierences are attributed
to the fact that the route redundancy measure considers the travelers’ O-D-level route choice beyond
the pure network topology and the shortest path considerations of the existing measures.
233
While existing literature on the application of networks to the aviation sector has made
significant strides, most studies do not focus on competition issues. This article, therefore, stands out
by applying network metrics to understand the dynamics of flight networks, addressing both demand-
side and supply-side substitutability. It oers an innovative perspective for tackling antitrust issues,
contributing to the ongoing academic debate by filling this gap and providing analytical tools for
antitrust investigations.
3 DATA
The international air travel data utilized in this paper is collected from two dierent sources:
Brazilian National Civil Aviation Agency (Anac) and from the FlightsFrom
6
website. The Anac data
includes two separate datasets. The first one includes data on airfares of domestic and international
flights in Brazil. According to Anac, this accounts for about 50% of the total passenger movement
by year once the tickets acquired via frequent flyer programmes and specific agreements between
customers and airlines are not part of the database. The international airfare data is also restricted
to tickets for trips originating in Brazil and ending abroad (as well as round-trip tickets, as long as the
return is from the destination airport of the outward leg) fully operated by the airline that sold them
and only for your scheduled flights previously registered with Anac. Both datasets were extracted
using the flightsbr package in R (Pereira, 2024). Since our interest of study is the international flight
network, we only used the origin and destination pairs of the international airfare data. There are 398
unique origin-destination pairs, with 37 unique origins and 60 unique destinations. Only origin and
destination pairs that had more than 500 seats sold between 2022 and 2023 were considered.
The second Anac database includes detailed information on every international flight to and
from Brazil, as well as domestic flights within the country. The data include flight-level information
of airports of origin and destination, flight duration, operating airline, aircrat type, payload, and the
number of passengers, and several other variables (Dados […], 2024) . The flight-level information data
identifies the pairs of origin, where the passengers boarded, and destination, where they disembarked,
regardless of the existence of intermediate airports, served by a given flight. This dataset is utilized to
analyze the evolution of the network from the standpoint of the supply of routes. The data is grouped
quarterly and extends from the first quarter of 2017 to the last quarter of 2023. It includes data for the
number of take-os, passengers and seats available for each route and airline.
The second source of data is the FlightsFrom website. It encompasses data of flight
information for all airports, including all the direct flights leaving the airport and travel time
7
. This
becomes necessary as the Anac flight-level database has some limitations, as it does not include
the entire itinerary of the passengers, including potential connections, and is restricted to flights
that fly from Brazil only. With this data, we can build a larger and more connected network of flights,
linking domestic origins to a diversity of final destinations through direct and indirect paths. In this
database, we filter for the routes where the destination is either an origin or a destination in our OD
pairs database, built with Anac’s international airfare data, as said above. Therefore, our network
accommodates 100 airports, of which 37 are domestic ones.
6 https://www.flightsfrom.com/
7 This data can be found at https://github.com/Jonty/airline-route-data.
234
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
Although the expansion of the flights’ data allows for a better understanding of the network
topology and of which paths consumers can take, our dataset still has some limitations that curb
the possibilities of analysis. Our dataset lacks information on the flow of passengers on each route,
preventing any estimation of node importance or the application of capacity restriction. This happens
since Anac’s database does not track data at the passenger-level, not accounting for connections. With
that, it is not possible to observe which paths the passengers actually take and how they distribute
themselves amongst them. Another constraint in our dataset is the lack of data for flight tickets’ price,
preventing an estimation of elasticities and restraining the possibility to investigate the restrictions to
switch routes among consumer and the proper identification of alternative routes. Even though Anac’s
database has information on prices, the data has some methodological problems that prevent us from
realizing a robust analysis. The price data informs the price paid and the number of tickets bought
for each month and each route, but it has no information on when the passenger is travelling. This
means that the passenger flow database and the price data are not on the same page, therefore it is
not possible to know how much the passenger paid for a ticket to travel on a given flight. Given these
obstacles, our research focuses on the identification of possible alternatives paths, considering the
origins and destinations identified in the international airfare data from Anac, and which links are the
most important within this network, as they become important intermediaries in the paths identified.
4 DEMAND SUBSTITUTABILITY
4.1 Methodology
In this section, we present the methodology implemented to assess route redundancy in
the aviation network. To do that, an index is built, ranking the routes according to their importance
in the O-D network. Based on the work of Jing, Xu, and Pu (2019) and Xu et al. (2018) two factors are
considered in the construction of the index: ecient path and not-too-long path.
Dial (1971) defined an ecient path as one which does not backtrack. As it progresses from
node to node it always gets further from the origin and closer to the destination. A path is ecient if
every link in it has its initial node closer to the origin than is its final node and has its final node closer
to the destination than is its initial node. All links should satisfy:
where tail
a
and head
a
are the tail and head of the link a; l
r
(tail
a
) and l
r
(head
a
) are the cost of
the shortest route from the origin r to the tail and head of link a, respectively; A is the set of directed
links.
Typically, passengers do not prefer too long routes or are willing to accept a certain amount
of extra time when their primary or secondary path alternatives are not available. With that in mind,
following Leurent (1997), a length constraint is introduced to ascertain that every link is reasonable
enough relative to the shortest path:
Where l
a
is the cost (length of free-flow travel time) of link r is an allowable/acceptable
235
elongation ratio for link a with respect to the origin r. τ
a
may be set to 1.6 for inter-urban studies or
between 1.3 and 1.5 for urban studies (Tagliacozzo; Pirzio, 1973; Leurent, 1997). By summarizing all links
on route k:
Therefore:
where l
k
(or l
p
) is the cost of the route k (or p); l
r
and l
s
are the shortest costs from origin r to r
and to destination s. This ensures that the cost of the route does not exceed (1+τ
max
) times the cost
of the shortest path. In our study, the cost of the route is the travel time in minutes. However, in order
to account for waiting time in connections, we add an extra 60 minutes for each domestic route and
an extra 150 minutes for international routes. In addition, to reduce computational burden, only the
20 shortest paths are calculated for each origin-destination pair.
To build the redundancy index, the following is considered. For each O-D pair, the eective
paths are calculated according to the method above. The paths are ordered by their length, and for
each link in each eective path is attributed a value. So, for a link i in a O-D pair p and rank r, the links’
redundancy index is:
where N
p
is the number of passengers recorded in the O-D pair p and rank
p,r
is the rank of
the path p in the path p. In that way, links in the shortest path receive higher values than links in
the second shortest path and so forth. So, for each link i in the network, the Redundancy Index is
measured as the sum of the redundancy index for all O-D pairs:
where P is the set of all O-D pairs and R is the set of the ordered eective paths for each O-D pair.
236
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
4.2 Results
To analyze the results, verify which links are more important and identify path alternatives,
we run two analyses. In the first one, we estimate the redundancy index for each edge in the network
and the number of eective paths for each O-D pair. This way it is possible to identify key routes
in the network, considering it is a lane for many pathways and the number of passengers that it
probably transports. It is important to point out that the redundancy index value has no direct
interpretation, as it is considered only the volume of passengers between the origin and destination
of which the edge is part, not the actual volume of travelers on the route. This index also allows for
the identification of destinations that are access constrained than others, as they present a lower
number of eective paths connecting them. Recognizing points with limited access is also interesting,
because any negative impacts in the routes leading to those may result in significant backlashes in
passengers’ welfare.
Meanwhile, in the second setup, the international routes (the ones where its origin is in
Brazil) are removed, one by one, from the network and then the number of alternative paths and
travel time are re-estimated for each origin-destination pair. The goal with this is to understand which
routes play a crucial role in connecting final destinations or, at least, conducting passengers to out
of the country, but also to see the routes that become more important and, therefore, can be seen as
alternatives to the removed route. Finding substitute routes can help antitrust analysis when it comes
to establishing the geographically relevant market.
In both experiments, the value of τ
α
is set at 0.3, i.e. passengers are willing to elongate their
travel time in 30% compared to the shortest path connecting the origin to the destination. This value
is lower than the one observed in urban studies (Leurent, 1997). Although some may think that the
willingness to extend the travel could be higher for air transportation, we opt for a more conservative
approach. Adding the waiting time for connections supports the use of a lower value for the parameter.
Also, since we have no data on flights’ frequency outside of Brazil, our estimation of connection time
is probably underestimated, making it reasonable to use a smaller value of the parameter. The chosen
value would also not extrapolate the travel time extension for indirect flights adopted either by the
European Commission
8
or Cade
9
, therefore making it a reasonable value.
4.2.1 Link’s importance and eective paths
Table 1 presents the 10 routes with the highest redundancy index. As can be seen, out of
the six first values, five belong to routes leading either to Santiago or to Buenos Aires. This outcome
is intuitive since Santiago and Buenos Aires are the main destinations of Brazilian passengers. It is
interesting to point out the values obtained for the routes Curitiba (CWB) to Santiago (SCL) and Foz
do Iguaçu (IGU) to Santiago (SCL). Even though these pathways don’t transport a high volume of
passengers, their redundancy index value is prominent, due to the fact that these routes are viable
short paths to connect travelers to Santiago. This example may be interpreted from two opposite
points of view. First, it could mean that these routes, although not highly traveled, are important to
8 See Case No COMP/M.3280—Air France/KLM, 11.02.2004.
9 See Opinion No. 6/2024/CGAA4/SGA1/SG. Process No. 08700.004702/2023-81 (SEI 1377008), Applicants: International
Consolidated Airlines Group and Air Europa Holding, S.L. Brasília: CADE, 2024.
237
the network, and a traditional approach would miss that. On the other hand, this could be due to the
lack of data on passenger flow for each route, which downplays the significance of the airports and
exposes one of the limitations of the current study.
Table 1 – Link Importance - Most Important Links
Origin
Airport
Origin
City
Origin
Country
Destination
Airport
Destination
City
Destination
Country
Redundancy
Index
Number
of Seats
Available
GRU São Paulo Brazil SCL Santiago Chile 300580.6 1019358
CWB Curitiba Brazil SCL Santiago Chile 222818.3 31536
GRU São Paulo Brazil MIA Miami USA 192325.8 860656
IGU Foz do
Iguaçu
Brazil SCL Santiago Chile 174838.6 36051
GRU São Paulo Brazil EZE Buenos
Aires
Argentina 158022.7 882914
GRU São Paulo Brazil AEP Buenos
Aires
Argentina 150009.2 1053591
NAT Natal Brazil LIS Lisbon Portugal 148660.9 93867
CNF Confins Brazil LIS Lisbon Portugal 144665.9 180208
GRU São Paulo Brazil LIS Lisbon Portugal 144543.4 734379
VCP Campinas Brazil LIS Lisbon Portugal 141940.7 369314
Source: Own Elaboration (2025).
The redundancy index, along with the approximated network built in this study, enables
us to speculate which international routes might play an important role in conducting Brazilian
passengers abroad. Table 2 consolidates the results for the top 10 international routes, according to
their redundancy index.
Table 2 – Link Importance - Most Important Links
Origin
Airport
Origin City Origin
Country
Destination
Airport
Destination
City
Destination
Country
Redundancy
Index
LIS Lisbon Portugal LHR London United
Kingdom
30219.27
LIS Lisbon Portugal MXP Milan Italy 222818.3
MAD Madrid Spain LHR London United
Kingdom
21712.71
ZRH Zurich Switzerland MXP Milan Italy 21,572.81
LIS Lisbon Portugal ORY Paris France 20086.97
SDQ Santo
Domingo
Dominican
Republic
JFK New York USA 18073.51
MAD Madrid Spain MXP Milan Italy 17586.95
238
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
CDG Paris France MXP Milan Italy 16310.04
ZRH Zurich Switzerland FRA Frankfurt Germany 12255.34
BCN Barcelona Spain MXP Milan Italy 11026.16
Source: Own Elaboration (2025).
The main finding is the dominance of European routes. The reason for that is twofold. In the
first place, destinations in Europe represent 20% of all unique destinations and 34% of all O-D pairs.
The second reason is related to distance. Out of the fiteen furthest destinations, on average, only
five
10
are not located in Europe. Besides that, the fact that European countries are relatively close to
each other makes connections between them more suitable. Additionally, it is fundamental to point
out Lisbon’s role as a bridge to the rest of Europe. The Portuguese capital has eleven
11
dierent direct
flights from Brazil, oering great access to Europe.
4.2.2 Edge removal
In this section, the goal is to understand how the removal of some routes aects the network,
identify which other routes are impacted and which edges become more important in the absence
of the removed one. Considering the objective of the current study, we solely remove international
routes whose origin is in Brazil. Those come from the flight-level data from Anac, introduced in the
Data section. We selected the flights from 2022 and 2023 that had more than 50 departures in both
years combined. Removing routes that are no longer operational, we have 109 active routes. The idea
here is the following: every edge is removed one by one, and, ater the removal, the redundancy index
is re-estimated for all remaining edges in the network. Given this setup, we can calculate for which
edges the redundancy index increased and, consequently, its importance, and estimate an average
increase in travel time in all paths that contained the disconnected edges.
Table 3 presents the 10 routes which removal result in the highest traveling time increment.
The flight between Belém and Pamaribo leads the way, with a 260% rise in travel time due to its
withdrawing. Although the route is not a relatively significant one in terms of volume
12
, the exercise
shows how its removal has a high impact in the network. We can also see that the flight Panama
(PTY) to Pamaribo (PBM) ends being the one which importance increases the most. It also interesting
to point out how the extraction of the route São Paulo (GRU) - Johannesburg (JNB) impacts a whole
path, augmenting the role of the São Paulo (GRU) - Luanda (LAD) - Johannesburg (JNB) pathway. In
some cases, the route that sees its role expand might be a domestic one, given that demand may be
redirected to other airport within the domestic network. One example of that is the route Porto Alegre
(POA) to Santiago (SCL), where the Porto Alegre (POA) to Curitiba (CWB) ends up being the one with
the highest increase in importance.
10 These are Dubai, Doha, Istanbul, Addis Ababa and Los Angeles.
11 Belém, Brasília, Belo Horizonte, Fortaleza, Rio de Janeiro, São Paulo, Natal, Porto Alegre, Recife, Salvador and Cam-
pinas.
12 Its 91th out of 172 in terms of seats in 2023.
239
Table 3 – Routes with the highest increase in travel time due to removal
Origin
Airport
Origin City Destination
Airport
Destination
City
Time
Increase (%)
Benefited
Route
Benefited
Route Origin
Benefited
Route
Destination
BEL Belém PBM Pamaribo 260% PTY →
PBM
Panama Pamaribo
GRU São Paulo LAD Luanda 60% GRU →
JNB
São Paulo Johannesburg
GRU São Paulo PDP Pamaribo 56.2% AEP →
PDP
Buenos Aires Punta del
Este
MAO Manaus PTY Panama 52.7% CNF →
PTY
Belo
Horizonte
Panama
GRU São Paulo JNB Johannesburg 48.6% GRU →
LAD
São Paulo Luanda
GRU São Paulo JNB Johannesburg 48.6% LAD →
JNB
Luanda Johannesburg
IGU Foz do
Iguaçu
SCL Santiago 48.3% FLN →
SCL
Florianópolis Santiago
POA Porto Alegre SCL Santiago 43.3% POA →
CWB
Porto Alegre Curitiba
GRU São Paulo VVI Santa Cruz 40.8% ASU →
VVI
Asuncion Santa Cruz
FLN Florianópolis AEP Buenos Aires 39.65% CWB →
AEP
Curitiba Buenos Aires
Source: Own Elaboration (2025).
When the analysis shits to routes which removal exercises the lower impact in terms of
extension of travel time, a few patterns stand out. Table 4 sums up the results. Excluding the route
Brasília (BSB) to Lisbon (LIS), all the remaining ones its origin is in the Southeast region of Brazil. This
is due to this region being the one with the more central airports (Couto et al., 2015) and, since these
places are somewhat close, alternatives to the removed routes are not time costly. Another point is
that São Paulo (GRU) and Rio de Janeiro (GIG) are, most of the time, the closest alternative to each
other. The proximity
13
and the fact that these are the largest international airports in the country
supports this fact.
13 Its, on average, an one hour flight.
240
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
Table 4 – Routes with the lowest increase in travel time due to removal
Origin
Airport
Origin City Destination
Airport
Destination
City
Time Increase
(%)
Benefited
Route
Benefited
Route
Origin
Benefited
Route
Destination
GIG Rio de
Janeiro
FCO Rome 4.6% GRU → FCO São Paulo Rome
GRU São Paulo LHR London 4.6% GRU → LHR São Paulo London
GRU São Paulo FRA Frankfurt 4.6% GIG → FRA Rio de
Janeiro
Frankfurt
VCP Campinas MCO Orlando 4.9% BSB → MCO Brasília Orlando
GRU São Paulo JFK New York 5.3% GIG → JFK Rio de
Janeiro
New York
VCP Campinas LIS Lisbon 5.3% CNF → LIS Belo
Horizonte
Lisbon
GIG Rio de
Janeiro
FRA Frankfurt 5.5% GRU → FRA São Paulo Frankfurt
CNF Belo
Horizonte
FLL Fort
Laudardale
5.8% MAO → FLL Manaus Fort
Lauderdale
BSB Brasília LIS Lisbon 5.9% NAT → LIS Natal Lisbon
GRU São Paulo CDG Paris 6% SSA → CDG Salvador Paris
Source: Own Elaboration (2025).
Table 5 presents the increase in travel time and the benefited route from the exclusion of
the 10 routes with the higher volume of passengers in 2023, considering flight-level data from Anac.
This table shows how the methodology here implemented can help in the identification of relevant
airports in the network. Although Curitiba (CWB) is not a very important airport
14
it is a significant
alternative to passengers traveling to destinations in South America, such as Buenos Aires (EZE, AEP)
and Santiago (SCL).
Table 5 – Routes with the highest volume of passengers in 2023
Origin
Airport
Origin
City
Destination
Airport
Destination
City
Time
Increase (%)
Benefited
Route
Benefited
Route Origin
Benefited
Route
Destination
GRU São
Paulo
SCL Santiago 10.6% GRU → CWB São Paulo Curitiba
GRU São
Paulo
AEP Buenos
Aires
16.2% GRU → CWB São Paulo Curitiba
GRU São
Paulo
MIA Miami 7.3% BSB → MIA Brasília Miami
14 It was the 13th in terms of volume of passengers in 2023
241
GRU São
Paulo
EZE Buenos
Aires
26.2% CWB → EZE Curitiba Buenos Aires
GRU São
Paulo
LIS Lisbon 12.8% GRU → CNF São Paulo CNF
GIG Rio de
Janeiro
EZE Buenos
Aires
26.8% FLN → EZE Florianópolis Buenos Aires
GRU São
Paulo
MAD Madrid 6.1% SSA → MAD Salvador Madrid
GRU São
Paulo
CDG Paris 6% SSA → CDG Salvador Paris
GIG Rio de
Janeiro
SCL Santiago 21.1% CWB → SCL Curitiba Santiago
GRU São
Paulo
PTY Panama 17.4% BSB → PTY Brasília Panama
Source: Own Elaboration (2025).
4.3 Practical exampleTo show the full extension of the methodology presented in Section (4.1),
we consider an example based on the failed proposed merger between International Consolidated
Airlines Group (IAG) and Air Europa Holding, S.L. (Air Europa)
15
. According to the European
Commission and Cade, the transaction could reduce competition on long-haul routes between
Madrid and South America, on which both parties oer a direct connection and face competition
from only a few other competitors with a nonstop connection. In its analysis, SG/CADE defined the
relevant geographical market as the routes: São Paulo/Guarulhos - Madrid/Barajas (GRU-MAD); São
Paulo/Guarulhos – Barcelona/El Prat (GRU-BCN); São Paulo/Guarulhos – Londres (GRU-LON); Rio de
Janeiro/Galeão – Madrid/Barajas (GIG-MAD); Rio de Janeiro/Galeão – Londres (GIG-LON); Salvador –
Madrid/Barajas (SSA-MAD). The routes through London were considered indirect. SG/CADE outlined
competitive concerns on the GRU-MAD route. The criteria for the consideration of indirect routes
were routes whose travel time would add up to 5 hours. Our methodology diers from this as it
considers a time increase compared to the shortest path, not a fixed time increase.
Looking at the flow of passengers between Brazil and Madrid, Figure 1 presents the whole
network connecting Brazil to Madrid. For this calculation, only the origin-destination pairs whose final
destination was Madrid were used.
15 Case M.11109 in the European Commission and Process No. 08700.004702/2023-81 in Cade.
242
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
Figure 1 – Brazil - Madrid network
Source: Own Elaboration (2025).
The reason why Salvador is the one with the highest value is its distance from Madrid. While
São Paulo and Rio de Janeiro are 8355 and 8124 kilometers away from the Spanish capital, 6915
kilometers separate Salvador from Madrid, resulting in a reduction of travel time of about 100 minutes.
Another reason is that Salvador is relatively close to other cities in Brazil, making the city a feasible
connection. Although an interesting result, this also highlights a limitation of our methodology, since
distance (travel time) is the only factor taken into account.
Additionally, we only observe Portugal (especially Lisbon) as a connection to Madrid within
Europe, not also London, as indicated by Cade. First, this could be due to a low selected value for . A
higher value could make other cities in Europe, such as London, Paris, and Frankfurt, viable options but
at the same time could increase the number of domestic connections, which might not be reasonable.
Although Lisbon might present itself as an alternative, the limitation of our study does not allow for
any extrapolation of the results. A major factor that is not considered in our analysis is the capacity
of the airports, which could curb any deviation in demand.
Expanding the analysis, we can observe the impact of the route when we consider the full
network, not only the one leading to Madrid. For that, we simulate the removal of the GRU-MAD route
and then recalculate the redundancy index.
243
Figure 2 – GRU-MAD impact network
Source: Own Elaboration (2025).
As can be seen from Figure 2, the removal of the GRU-MAD route has a worldwide impact. 157
routes available in our dataset suer an increase in their redundancy index. At the same time, 39 routes
are negatively aected, most of which Madrid is the origin. This simple exercise exemplifies how the
adoption of a network approach can enhance the analysis in the aviation sector and how a traditional
analysis might not paint the full picture as it misses the direct and indirect network eects. Taking the
example above, an increase in concentration on the selected route could have a negative impact on
many other routes and, consequently, passengers. Domestic routes, which play an important role in
feeding international routes, could see higher prices due to the deviation in demand from passengers.
This simple example shows how a traditional analysis might miss some of the ripple eects of a change
in conditions in a route. Companies which operate on important routes may have a higher dominant
position than what is captured by their market share, in such a way that an O&D approach would
underestimate the merger eects. On the other hand, if the company operates on a limited network, its
position might be reduced, even though it has a high market share in a route. This goes for the analysis
of a merger’s eciencies as well. Increasing returns due to network eects (Arthur, 1994) could justify
an operation, as the airline may become better positioned to optimize its network.
Once more, it is important to stress the limitations of the results. The goal here is to implement
network theory as an alternative to discussions involving the aviation industry. The absence of prices
in our analysis curtails any stricter interpretation of alternative routes or paths as actual substitutes.
It can only show paths the passenger would be willing to take, say the price was the same for all
routes. Therefore, with detailed price data for each route, the analysis could be expanded to account
for a more realistic passenger behavior and to understand substitution not only as a factor of the time
of travel, but as the price as well. In this case, the metric would cover not only the amount of time that
passengers would be willing to elongate their travel time, but also how much they would be willing to
spend on a longer flight.
244
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
5 SUPPLY SUBSTITUTABILITY
In the antitrust context, the ability of dominant players to exercise market power on routes
under their control should not be analyzed solely from the demand perspective. This is because,
even if consumers consider alternative routes unfeasible — leading to significant and non-transitory
increases in air ticket prices on the route under review — there would still be the possibility, from
the supply side, that competing airlines might start operating on the route in question due to the
increased profitability initially generated by this fare increase. Thus, this section aims to analyze
the substitutability of the supply of airline seats in the post-pandemic context from two distinct yet
complementary perspectives: one qualitative and the other quantitative.
In the qualitative analysis, evidence is presented regarding the challenges airlines face in
increasing capacity in the post-pandemic scenario, highlighting the implications for the sector’s
competitive dynamics. As discussed in the next subsection, the contraction in demand for new aircrat
during the pandemic crisis - due both to uncertainties about sector recovery and to the substantial
increase in airlines’ leverage - combined with the diculties faced by major global wide-body aircrat
manufacturers (Boeing and Airbus) in meeting backlogged orders, has brought this issue to the
forefront of antitrust authorities’ attention.
The quantitative analysis employs empirical exercises based on a route network construction
methodology to illustrate recent patterns of network expansion in the post-pandemic period and
assess their potential impact on the competitive dynamics of the sector.
5.1 Qualitative analysis
From a supply perspective, antitrust authorities and academics in their analysis of merger
cases and or competition analysis involving commercial airline sector have historically focused
their analysis on airport infrastructure availability and regulatory mechanisms for slot allocation at
coordinated airports – the approach taken in Bilotkach and Lakew’s work (2014) – without adequately
considering whether new entry would be economically feasible given trac flows and hub economics
(Nannes, 1999).
However, the fact that the European Commission’s ex-post assessment concluded that the
divestiture of slots alone has not been sucient to remedy competitive harm (Richen, 2024; Ibitoye,
2023) has shed light on the need for additional antitrust remedies that take into account network
eects and, implicitly, the availability of the assets necessary to ensure the timely and eective entry
of a new competitor on the routes under antitrust scrutiny. According to Chen (2024, p. 331), “[...] if
there is not enough perfect transfer network and passenger flow, even if some slots are released,
other airlines will not take over, but in a short time, the choice of passengers will be reduced, and
the fare will not be lower.” Following this new guidance, in the case of the merger between Luthansa
and ITA, approved by the European Commission on July 3, 2024, the Commission imposed as one of
the conditions for the approval of the transaction that the parties must make available to one or two
rival airlines the necessary assets to enable them to start non-stop flights between Rome and Milan
and certain airports in Central Europe (European Commission, 2024). Similarly, in the merger between
Korean Air and Asiana Airlines concluded in 2024, the European Commission imposed remedies
245
extending beyond traditional slot divestitures. These included the sale of Asiana’s cargo division to
Air Incheon and the transfer of five Airbus A330 aircrat along with approximately 100 pilots to T’way
Air, a South Korean low-cost carrier. This support enabled T’way to commence direct flights between
Seoul and key European cities such as Paris, Rome, Frankfurt, and Barcelona, thereby preserving
competition on these routes (Min-Hyung, 2023).
In Brazil, the most recent and emblematic case analyzed by antitrust authority Cade was
Gol-Webjet merger case
16
, in which has been imposed for the applicants a performance commitment
agreement (TCD) for a four-year period establishing a requirement to maintain 85% eciency in the
use of slots at Santos Dumont Airport in Rio de Janeiro (SDU). This aimed to prevent slot idleness
granted to the company by Anac
17
.
However, the current post-pandemic period indicates that there is no shortage of slots at
Western airports, and according to (Transforming […], 2023), this overall situation is not expected
to change until 2030. Therefore, it can be stated that the scarcity of airport infrastructure does not
currently represent a significant obstacle to competition, which means that the mandatory allocation
of slots would most likely be an ineective antitrust remedy to promote competition
The recovery of the international aviation sector starting in 2023 has led airlines to face
substantial challenges in expanding their routes (as networks links) or increasing flight frequencies
(corresponding to an increase in the intensity of existing routes) due to a limited supply of new
aircrat. This supply crisis is mainly due to a combination of two factors: (i) a decrease in demand
for new aircrat during the pandemic crisis, driven by uncertainties about the sector’s recovery and a
substantial increase in airlines’ leverage; and (ii) diculties faced by major global wide-body aircrat
manufacturers (Boeing and Airbus) in meeting backlogged orders.
Regarding factor (ii), it is worth highlighting the specific challenges faced by each aircrat
manufacturer. In the case of Boeing, the American company has been experiencing an internal crisis
since 2018, both reputational and operational, caused by the accidents involving the Boeing 737
Max, which have impacted its production and delivery rates even before the pandemic. The crisis
resurfaced in early 2024 when an emergency door on a Boeing 737 Max 9 jet detached during takeo
on a flight operated by Alaska Airlines (Relatório [...], 2024). Following the incident, the Federal Aviation
Administration (FAA), the U.S. aviation regulatory agency, opened an investigation into the case and
set a monthly production limit of 38 new aircrat for the model, below the delivery rate of 45 units
reached before the latest incident (in November 2023) and the target of 57 aircrat set by Boeing for
July 2025 (Crumley, 2024). Given this situation, the average delivery time for Boeing aircrat reached
16.9 years in the first quarter of 2024, almost double the average time of seven years observed in 2018.
The crisis at the American manufacturer, combined with the post-pandemic recovery of
the aviation sector, has benefited competitor Airbus. However, the manufacturer is encountering
diculties in meeting aircrat delivery targets and reducing the average order lead time. In the first
quarter of 2024, the company had an order backlog of 8,626 aircrat, with annualized deliveries of
16 Process No. 08012.008378/2011-95.
17 Similarly, in the LATAM-IAG Merger Case (Process No. 08700.004211/2016-10), Cade held that various remedies should
be imposed on the applicants to ensure competition on relevant international routes and pass any operational eciencies on
to passengers. As a structural remedy, the ACC (Merger Control Agreement) required applicants to make slots available, free of
charge, to a potential market entrant at a London airport of the entrant’s choice for a period of 10 years, for use in daily nonstop
f lights departing from Guarulhos Airport in São Paulo.
246
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
568 during the period, resulting in an estimated average delivery time of 15.2 years (Faury; Toepfer,
2024), compared to an average of 8.7 years observed in 2018. Since the delivery pace was weaker at
the beginning of 2024, if the target of delivering 800 aircrat is met this year (and without considering
the addition of new orders), the indicator would drop from 11.7 recorded in 2023 to 10.8 in 2024, still
above the level observed before Boeing’s crisis.
Given the scenario presented, the main concern from an antitrust perspective is that new
competitors’ entry into air routes could be hampered by the scarcity of new aircrat, not only due to
the backlog and delays in deliveries by major manufacturers but also due to the increased bargaining
power of leasing companies, which tend to raise leasing contract costs. This has asymmetric impacts
on potential new entrants, which typically have a smaller scale of operations.
Evidence suggests that aircrat delivery delays have already compromised the expansion of
the aviation sector. At the end of October 2024, Luthansa announced the suspension of its São Paulo-
Munich route, which was set to start in December 2024, due to a lack of aircrat (Pólo Júnior, 2024)
18
.
Far from being isolated cases, capacity saturation can be observed through the seat
occupancy rate of flights operated in the Brazilian market. According to Anac’s 2023 Annual Report,
annual load factor (RPK/ASK) reached 85.4% last year, the highest level at least since 2005. According
to a BNDES study (Gomes; Fonseca, 2014, p. 135), a rate above 85% indicates that ‘[...] as this indicator
is an average, the company will already be leaving people on the ground or losing passengers to the
competition (saturation point, spill)’.
This is thus a case of ’expansion under scarcity’, which should serve as a warning to antitrust
authorities regarding the timeliness of increasing seat supply on evaluated routes, even by incumbents
themselves. Although this indicator above 80% was observed in several years in the last decade,
the long waiting list for aircrat delivery from major global manufacturers should be considered,
indicating that this saturation cannot be resolved in the short term.
5.2 Quantitative analysis
At first glance, one might say that airlines have full flexibility in managing their respective
fleets, since, unlike most capital goods, aircrat can be moved to dierent airports in a matter of hours
to achieve higher returns on more profitable routes. However, this theoretical possibility does not
take into account various operational, regulatory, and strategic constraints that significantly reduce
this flexibility in practice.
Given the practical diculty of obtaining profitability levels per route for each airline - and
thus identifying the eects on aircrat reallocation - the issue was reformulated from a network
perspective, aiming to understand the extent to which airlines reallocate their flights intensively
(through the reallocation of the existing fleet) and extensively (through the addition of new aircrat),
regardless of the underlying motivations. The initial hypothesis of this study is that, for various
reasons, airlines do not tend to substantially change their fleet in response to short-term changes;
therefore, new destinations would largely be added through the introduction of new aircrat.
18 American Airlines, in turn, announced at the end of June this year that it decided to pause hiring new pilots until the
end of 2024 due to delayed aircrat deliveries by Boeing (American […], 2024). Similar announcements had been made months
earlier by Delta Air Lines and United Airlines (Schlangenstein; Beene, 2024).
247
From a conceptual standpoint, in this case, the network is formed by the connection (routes)
between origin and destination airports (network nodes), with edges represented by A and the weight
P, which reflects the strength or importance of the connection between network nodes, represented
by the number of flights or the quantity of seats oered. Based on these measures, it is possible to
derive network evolution indicators.
The first indicator pertains to the variation in the number of edges (∆A), which consists of
the absolute expansion of the network in an extensive manner. The variation in the number of edges
weighted by seat oer gives (∆PA). The weight variation (∆P) represents the change in the total weight
of the network’s edges. Based on these metrics, it is possible to construct indicators to verify the
network’s evolution pattern. Figure 1 shows a hypothetical case of extensive network expansion, with
the addition of destination E with weight 1.
Figure 3 – Network expansion with an inclusion of new destination
Source: Own Elaboration (2025).
As can be seen, the addition of the new destination to the network results in ∆A = ∆PA = 1 and
∆P = 1. With this, the extensive network expansion indicator (x) can be obtained, calculated by the
following algebraic expression:
248
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
On the condition that the denominator is not equal to zero:
where A
i
N
is the i-th added edge; A
j
S
is the j-th removed edge; P
i
N
represents the i-th weight of
the added edge i; P
j
S
represents the j-th weight of the removed edge j; P
k
represents the weight of the
edge k; PA represents the contribution of the weights of the new edges (net) to the network flow; z is
the number of edges added to the network in period t + 1; g is the number of edges removed from the
network in period t + 1; n is the total number of edges (or weights) in the network and t is the initial
or reference period.
In this way, it is possible to determine the extent of a network expansion in a selected period
through the allocation of flows (weights) between the new edges (in net terms) and the pre-existing ones.
It should be noted, therefore, that the weights of the new edges are considered to explain
extensive expansion, not just the absolute variation of the edges. This means that the emergence
of a new route, which from the beginning has a higher frequency of flights during the evaluation
period (t), is classified by indicator x as extensive growth. Any subsequent intensification of the new
route’s usage in a following period (t + 1), thus, is considered intensive expansion. This dierentiation
is justified because, in most cases, there is a minimum ecient scale for operating a given route,
depending on the commercial strategy and passenger demand profile. For example, on a route with
a predominantly business-travel profile, it is assumed that there is a higher preference for airlines
with a higher flight frequency, since in the event of a flight cancellation, there are greater chances of
reallocating passengers to another flight with the least possible delay. Another scenario occurs with
an airline oering a particular destination as a connection from a longer route, with the decision on
the number of flights to be oered being conditioned to the latter.
In practical terms, the indicator x for the hypothetical case in Figure 3 is calculated as follows:
This means that the entirety of the new flow (route or seat availability) has been allocated to
the new edge (in net terms). If the new route were to emerge with a weight of 2, the same value for the
indicator x would be obtained:
Additionally, another way to assess the network’s evolution pattern is related to the network
density variation indicator (d), obtained through the following algebraic expression:
Taking the example listed in Figure 1, we obtain that:
249
Based on the Anac database, it was possible to perform the calculations of the presented
metrics. To reduce the chances of ∆P = 0, which would make the relative extensive network expansion
indicator (x) unsolvable, the seat oer of the routes was considered as the weight variable for the
network edge, rather than the flight frequency on each route.
Regarding the time frame, due to the limitation of information on fleet size and type, as well
as the use of aircrat during the COVID-19 pandemic period for the restoration of supply, the analysis
considered the 2022-23 biennium and the isolated year of 2023 for this purpose.
Based on the Figure 4, changes in the evolution of the air trac network during the period
from 2017 to 2023 can be observed. In the second quarter of 2020, the first period under the full
impact of the crisis in the aviation sector caused by the COVID-19 pandemic, it is observed that all
metrics were negative (and significantly, based on absolute values), indicating a contraction in both
the number of destinations and/or origins (∆A) and the additional seat oer (PA), as well as the total
seat oer (∆P), which constitute the weights of each network edge. In this case, there was a decrease
in network density, given the disproportionate reduction in the total seat oer relative to the number
of edges. Similarly, during the recovery periods (2020-Q2 and Q3; and 2021-Q4 to 2022-Q1), all metrics
displayed positive values, reflecting both extensive expansions, marked by the addition of new edges
to the network, and intensive expansion, characterized by an increase in network density.
250
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
Figure 4 – Results of the variation in selected international flight network metrics – quarterly
variation compared to the immediately preceding period (2017 to 2023)
Source: Own Elaboration (2025).
Note: Green circle with a check mark inside: non-negative values; red circle = negative values.
Figure 5 shows the accumulated quarterly variations for the selected periods of the dierent
network metrics. The objective of this empirical exercise is to compare the period immediately following
the pandemic crisis with the most recent recovery phase of the air sector (ater the most pronounced
recovery phase of air trac). For this purpose, the following longer periods (highlighted in bold) were
selected: (i) before the pandemic crisis: from the second quarter of 2017 to the first quarter of 2020; and
(ii) the recent recovery phase: from the first quarter of 2022 to the fourth quarter of 2023.
Figure 5 – Results of the variation in international flight network metrics for selected periods - Anac data
Source: Own Elaboration (2025).
251
Regarding period (i), an initially counterintuitive result is observed, as there is a simultaneous
increase in the number of net edges (∆A = 34) and a reduction in the contribution (in terms of seat
oer) of these same edges. The explanation for this is that, even though two new edges with weight
1 are added, the fact that one edge with weight 4 is simultaneously removed is enough to make PA
negative (-2), even though ∆A is positive (+1).
Due to the negative result in period (i), and in order to compare the previous growth profile
of air trac with the more recent period, the same metrics were calculated for periods (i.1) Q3 2018
to Q1 2019 and (i.2) Q3 2019 to Q1 2020, which present a more consistent growth scenario (before the
pandemic) for analytical purposes. The results show dierent intensities of contribution from the new
net edges to network expansion, with period (i.1) showing an extensive network expansion indicator
(x) of 13%, while in period (i.2) the weighted contribution of new net edges was 43%. However, it should
be noted that in absolute terms (PA), the contribution of the new net edges was significantly higher
in period (i.1) compared to period (i.2), which may indicate a percentage closer to typical scenarios of
air sector expansion or a specific growth pattern during the analyzed period.
In the 2022-23 biennium, it was found that the indicator x reached 8.3%, a percentage lower
than that of periods (i.1) and (i.2) before the pandemic. However, when calculating the same metrics
only for 2023, it was found that the indicator x for that period was 11.8%, which is very close to the
value observed for the periods before the pandemic. Thus, it is understood that, based on the data
available up to 2023, it can be concluded that no significant shit in the growth pattern of the network
in the international air sector, concerning flights departing from or arriving in Brazilian territory, can
be identified considering the current scenario.
To deepen the understanding of the dynamics of the international flight network, dispersion
measures were calculated for the network metrics ∆P and ∆PA to identify possible breaks in the
pattern in the post-pandemic period, as shown in Figure 6.
Figure 6 – Results of the dispersion measures for the network metrics P and x for the selected
periods - Anac data
Source: Own Elaboration (2025).
As observed, there is a notable change in the dispersion measures during the post-pandemic
period. The coecient of variation (CV) of ∆P decreased to 0.9 in the 2022-23 biennium, compared
to 17.6 in the 2017-Q2 to 2020-Q1 period. Furthermore, with the reduction of the CV, this indicator,
which was previously much higher than that of PA (5.1 in absolute terms) in the initial period, is
now lower than the CV of 1.6 for PA in the post-pandemic period. To explain these results, several
252
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
hypotheses can be outlined, such as asymmetric shock eects (oil prices, geopolitical factors, etc.)
aecting the dynamics of the sector’s operations. However, one plausible explanation is a lower risk
appetite among airline companies - many of which are financially more fragile - for taking on the
risk of increasing seat capacity during the recovery phase of the sector. Another factor is that the
unavailability of aircrat might be discouraging airlines from entering new routes (even though the
subsequent result could be an exit later), which explains the reduction of the CV of PA as well as ∆P.
From an antitrust analysis perspective, the main contribution of this article is to highlight that there
may have been structural changes in the post-pandemic scenario that could negatively aect the
competitive dynamics of the sector, which would require more detailed and in-depth analyses within
the scope of concentration analysis in the sector.
6 CONCLUSION
In relation to demand-side substitutability, the analysis focused on identifying key routes and
assessing alternatives in Brazil’s international aviation network. Two main evaluations were conducted:
one on the redundancy index, which highlighted important routes based on their contribution to
multiple paths, and another on the impact of removing international flights from Brazil on alternative
routes and travel times.
The redundancy index revealed that routes to Santiago and Buenos Aires, even from less
tracked cities, were crucial, while European routes, particularly to Lisbon, were dominant. Most
origin-destination pairs (91.4%) had alternatives, suggesting good connectivity, though limited pricing
data prevented a deeper understanding of route substitutes.
As a result of the edge removal analysis, it is evident that the elimination of specific
international routes originating in Brazil can have significant ripple eects across the entire network.
Routes that may appear to be of lesser importance in terms of passenger volume can still play a crucial
role in the overall connectivity, with some removals leading to substantial increases in travel time. This
highlights how interconnected the aviation network is with certain flights, despite low trac, acting as
critical links that aect the redundancy of other routes. Conversely, the removal of high-trac routes
oten results in only marginal increases in travel time, especially when alternative routes from central
hubs like São Paulo and Rio de Janeiro are available. These insights underscore the importance of
considering network dynamics when assessing the impact of route discontinuations. Additionally,
the findings have relevance for the antitrust context, where the European Commission’s approach of
defining relevant markets by point-of-origin and point-of-destination pairs aligns with the analysis
presented here, emphasizing the strategic significance of specific routes within the broader network.
Further research could expand this framework to include a wider range of aected routes and explore
their implications on network structure and competition. Another important finding is that, when
accounting for network eects, the impacts of a merger might be better estimated, as it considers
not only the impact in competition for a route, but for a larger scope. The same goes for analyzing
network synergies. A traditional approach might miss some of the positive impacts of the airlines’
network optimization.
In relation to supply-side substitutability, the initial hypothesis of this study posited that
airlines would not substantially alter their fleets in response to short-term changes; therefore, new
253
destinations would primarily be added through the introduction of new aircrat. However, our findings
suggest that the evolution of the air trac network in the post-COVID-19 period has not yielded
conclusive results. A comparison of pre- and post-pandemic growth periods reveals that the extensive
expansion indicator (x) reached 8.3%, which is lower than the values observed in various pre-pandemic
periods. However, when the same metrics were calculated for 2023 alone, the indicator x for that period
was 11.8%, closely aligning with values seen in the pre-pandemic periods. Thus, it is not possible to
discern any significant change in the growth pattern of the network within the international air sector
for flights departing from or arriving in Brazilian territory under the current conditions.
Conversely, the dispersion measures applied to certain network metrics reveal a clear shit in
behavior patterns in the post-pandemic period. While numerous hypotheses can be considered, one
plausible explanation is a reduced risk appetite among airlines, many of which are financially more
fragile, making them less inclined to increase seat availability during the sector’s recent recovery
phase. Another contributing factor could be the unavailability of aircrat, which may be discouraging
airlines from entering new routes, even if this leads to eventual exits, as pointed out in the qualitative
analysis section.
From an antitrust analysis perspective, the data suggests the potential for structural changes
that could negatively impact the competitive dynamics of the air sector. Therefore, ongoing studies
and detailed analyses are necessary to assess from the demand side the potential influences of the
hub-and-spoke model on consumers’ route selection, given that flight frequency is a significant factor
in choosing an airline. On the supply side, there is a need to deepen the analysis of metrics behavior,
aiming to link them with other relevant variables, such as aircrat occupancy rates and profitability
changes of airlines, which can be influenced by fluctuations in oil prices.
From a regulatory perspective, understanding competitive dynamics through network theory
applied to the airline sector enables regulators to act in ways that prevent strategic behavior related
to slot usage. For instance, an incumbent airline could transfer flights from one airport to another with
the sole purpose of maintaining its dominant position at the latter, where slots are more contested.
This could lead to the creation of artificial hubs and harm the competitive process.
From an academic standpoint, the application of network theory to the analysis of competition
in the airline sector opens new avenues for understanding market dynamics beyond traditional
frameworks. By capturing the structural interdependencies between routes - and their recent
transformations in the post-pandemic period - this study sheds light on the multiple dimensions
that can influence competitive outcomes. Future research is needed to integrate both demand- and
supply-side aspects into a unified framework, including operational constraints (such as airport
capacity and slot availability) as well as dynamic games of entry deterrence (Aguirregabiria; Ho, 2010),
to simulate the potential impacts of mergers and acquisitions in the commercial aviation sector. To
this end, it is essential to improve the public database of international flights maintained by Anac,
in order to incorporate all passenger connections and enable integration with the airfare pricing
database published by the regulatory agency.
254
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
REFERENCES
AGUIRREGABIRIA, Victor; HO, Chun-Yu. A dynamic game of airline network competition: Hub-and-
spoke networks and entry deterrence. International Journal of Industrial Organization, [s. l.], v. 28, n.
4, p. 377–382, 2010. DOI: https://doi.org/10.1016/j.ijindorg.2010.03.003. Available at: https://x.gd/nDjya.
Accessed on: Apr. 15, 2025.
AMERICAN Airlines halts training for new pilots for the rest of 2024. The Dallas Morning News, Dallas,
Jun. 24, 2024. Available at: https://tinyurl.com/yuhecvzv. Retrieved: 14 nov. 2024.
ARTHUR, W. Brian. Increasing Returns and Path Dependence in the Economy. Ann Arbor, Michigan:
University of Michigan Press, 1994. Available at: http://www.jstor.org/stable/10.3998/mpub.10029.
Retrieved: 3 jan. 2025.
BILOTKACH, Volodymyr; LAKEW, Paulos Ashebir. On sources of market power in the airline industry:
panel data evidence from the US airports. Transportation Research Part A: Policy and Practice, [s. l.], v.
59, p. 288–305, Jan. 2014. Available at: https://doi.org/10.1016/j.tra.2013.11.011. Accessed on: Apr. 15, 2025.
CHEN, Shiqi. Examining Airlines Merger Control in the Post-pandemic Context: The Case of Korean-
Asiana Airlines Merger. US-China Law Review, Valley Cottage, v. 21, n. 7, p. 321-335, jul. 2024. DOI:
10.17265/1548-6605/2024.07.003. Available at: https://x.gd/kKIFdf. Retrieved: 15 abr. 2025.
COUTO, Guilherme S. et al. Structural properties of the Brazilian air transportation network. Anais
da Academia Brasileira de Ciências, Rio de Janeiro, v. 87, i. 3, p. 1653–1674, 2015. DOI: https://doi.
org/10.1590/0001-3765201520140155. Available at: https://x.gd/79Ohs. Retrieved: 19 dez. 2024.
CRUMLEY, Bruce. Boeing Woes Mount with More Losses and Production Curbs. Inc., [s. l.], May 24, 2024.
Available at: https://tinyurl.com/cf7z6n5h. Retrieved: 14 nov. 2024.
DADOS Estatísticos do Transporte Aéreo. Agência Nacional de Aviação Civil (Anac), Brasília, 2024.
Available at: https://x.gd/P1WSz. Retrieved: 19 dez. 2024.
DIAL, Robert B. A probabilistic multipath trac assignment model which obviates path enumeration.
Transportation Research, Pergamon, v. 5, i. 2, p. 83–111, 1971. Available at: https://x.gd/1Zh3j. Retrieved:
14 nov. 2024.
DUCCI, Francesco. Out-of-Market Eciencies, Two-Sided Platforms and Consumer Welfare: A Legal
and Economic Analysis. Journal of Competition Law and Economics, Oxford, v. 12, i. 3, p. 591-622, 2016.
Available at: https://x.gd/wx0aT. Retrieved: 14 nov. 2024.
EUROPEAN COMMISSION. Case COMP/AT.39595 – Continental/United/Luthansa/Air Canada. Decision
of 23 May 2013. Available at: https://x.gd/56ETY. Retrieved: 2 apr. 2025.
EUROPEAN COMMISSION. Case M.11071 – DEUTSCHE LUFTHANSA / MEF / ITA. Commission Decision of
3.7.2024 declaring a concentration to be compatible with the internal market and the EEA agreement,
2024. Available at: https://x.gd/GRj58. Retrieved: 12 apr. 2025.
EUROPEAN UNION GENERAL COURT. Judgment of the General Court (Eighth Chamber) of 13 May
2015. Niki Lutfahrt GmbH v European Commission. Case T-162/10. 2015. Available at: https://tinyurl.
com/3mwyv9rx. Retrieved: 14 nov. 2024.
255
FAURY, Guillaume; TOEPFER, Thomas. Q1 Results 2024. Airbus: Blagnac, France, 2024. Available at:
https://x.gd/pTzrG. Retrieved: 14 nov. 2024.
GOMES, Sérgio Bittencourt Varella; FONSECA, Paulus Vinicius da Rocha. Análise Econômico-Operacional
do Setor de Transporte Aéreo – Indicadores Básicos. Aeronáutica, Brasília, i. 40, 2014. Available at:
https://tinyurl.com/2e5cd7z4. Retrieved: 14 nov. 2024
IBITOYE, Victoria. EU adopting “rigorous approach” to airline remedies, Guersent says. GCR, London,
Sep. 15, 2023. Available at: https://x.gd/jldLw. Retrieved: 15 abr. 2025.
JENELIUS, Erik. Redundancy importance: Links as rerouting alternatives during road network
disruptions. Procedia Engineering, [s. l.], v. 3, p. 129–137, 2010. 1st Conference on Evacuation Modeling
and Management. DOI: https://doi.org/10.1016/j.proeng.2010.07.013. Available at: https://x.gd/MiBYr.
Retrieved: 16 set. 2024.
JING, Weiwei; XU, Xiangdong; PU, Yichao. Route Redundancy-Based Network Topology Measure of
Metro Networks. Journal of Advanced Transportation, [s. l.], v. 2019, n. 1, p. 4576961, 2019. DOI: https://
doi.org/10.1155/2019/4576961. Available at: https://x.gd/OZ5X4. Retrieved: 11 jul. 2024.
LENOIR, Nathalie. Research for TRAN Committee: Airport Slots and Aircrat Size at EU Airports: In-
Depth Analysis. Brussels: European Parliament, 2016. Available at: https://x.gd/AIw2z. Retrieved: 14
nov. 2024.
LEURENT, Fabien M. Curbing the computational diculty of the logit equilibrium assignment model.
Transportation Research Part B: Methodological, v. 31, n. 4, p. 315–326, 1997. DOI: https://doi.org/10.1016/
S0191-2615(96)00035-5. Available at: https://x.gd/NHvVR. Retrieved: 16 set. 2024.
LI, Feiyan et al. Identification of critical links in a large-scale road network considering the trac flow
betweenness index. PLOS One, [s. l.], v. 15, n. 4, p. 1–23, abr. 2020. DOI: 10.1371/journal.pone.0227474. DOI:
https://doi.org/10.1371/journal.pone.0227474. Available at: https://x.gd/Iinht. Retrieved: 16 set. 2024.
LYKOTRAFITI, Antigoni. Regulatory convergence between U.S. antitrust law and EU competition law in
international air transport - taking stock. Journal of Competition Law & Economics, Oxford, v. 19, i. 1, p.
146–176, 2023. DOI: https://doi.org/10.1093/joclec/nhac013. Available at: https://x.gd/ImkuN. Accessed
on: Apr. 14, 2025.
MIN-HYUNG, Lee. Korean Air to transfer control of 5 airplanes, 100 pilots to T’way for Asiana takeover.
The Korea Times, Seoul, Oct 16, 2023. Available at: https://x.gd/mN83r. Retrieved: 15 abr. 2025.
NANNES, John M. The importance of entry conditions in analyzing airline antitrust issues. Washington,
D.C.: U.S. Department of Justice, 1999. Available at: https://x.gd/iSeSaU. Retrieved: 12 abr. 2025.
OLMEDO PERALTA, Eugenio. Market definition and conditions to ensure competition in the European
market for air transport services: an analysis of past and upcoming merger cases. Market & Competition
Law Review, Lisboa, v. 4, i. 1, p. 87-113, 2020. Available at: https://x.gd/L0cFR. Retrieved: 4 April, 2025.
ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT (OECD). Executive Summary of
the Discussion on Airline Competition. Paris: OECD, 2014. Available at: https://tinyurl.com/4bm36ucz.
Retrieved: 14 nov. 2024.
PEREIRA, Rafael H. M. Package ‘flightsbr’: Download Flight and Airport Data from Brazil. [S. l.]: OSF
preprints, 2024. Available at: https://x.gd/FyvI1. Retrieved: 14 nov. 2024.
256
ZANA, Eduardo Roberto; ALARCÃO, Gabriel Oliveira de; BARCELLOS, Tomás de Siervi. Assessing
demand and supply substitutability in Brazilian international aviation: a network theory approach
for antitrust analysis. Revista de Defesa da Concorrência, Brasília, v. 13, n. 1, p. 228-256, 2025.
https://doi.org/10.52896/rdc.v13i1.1932
POLO JÚNIOR, Alberto. Luthansa suspende rota São Paulo-Munique e reduz voos entre Rio e
Frankfurt por falta de aeronaves. Airway, [s. l.], Oct. 31, 2024. Available at: https://x.gd/oZHs9.
Retrieved: 14 nov. 2024.
RELATÓRIO sobre acidente com Boeing 737 MAX 9 da Alaska Air diz que faltavam 4 parafusos no
plugue de porta. O Globo, Rio de Janeiro, Feb. 6, 2024. Available at: https://tinyurl.com/bd3vdrxx.
Retrieved: 14 nov. 2024.
RICHEN, Kaya. Lessons to learn from airline mergers. Antitrustpolitics.com, 12 jul. 2024. Available at:
https://x.gd/aaFSt. Retrieved: 15 abr. 2025
ROCHA, Luis E. C. da. Structural evolution of the Brazilian airport network. Journal of Statistical
Mechanics: Theory and Experiment, [s. l.], v. 2009, abr. 2009. Available at: https://dx.doi.org/10.1088/1742-
5468/2009/04/P04020. Retrieved: 16 set. 2024.
RUPI, Federico et al. Ranking links in a road transport network: a practical method for the calculation
of link importance. Transportation Research Procedia, [s. l.], v. 5, p. 221–232, 2015. DOI: https://doi.
org/10.1016/j.trpro.2015.01.003. Available at: https://x.gd/Nf8ql. Retrieved: 16 dez. 2024.
SCHLANGENSTEIN, Mary; BEENE, Ryan. Boeing Crisis Blows Up Airlines’ Growth Plans as Output Stalls.
Bloomberg, [s. l.], Mar. 12, 2024. Available at: https://tinyurl.com/3bs5btn4. Retrieved: 14 nov. 2024.
SILVA, Thiago Christiano et al. The role of network topology in competition and ticket pricing in
air transportation: evidence from Brazil. Physica A: Statistical Mechanics and its Applications, [s.
l.], v. 601, 2022. DOI: https://doi.org/10.1016/j.physa.2022.127602. Available at: https://x.gd/H4ybS.
Retrieved: 16 set. 2024.
SINGH, Rajesh Kumar. Aircrat shortages turn into cash bonanza for some airlines. Reuters, [s. l.], Mai.
31, 2024. Available at: https://tinyurl.com/2nxtck. Retrieved: 14 nov. 2024.
TAGLIACOZZO, F.; PIRZIO, F. Assignment models and urban path selection criteria: results of a survey of
the behaviour of road users. Transportation Research, v. 7, i. 3, p. 313–329, 1973. Available at: https://
doi.org/10.1016/0041-1647(73)90020-8. Retrieved: 16 set. 2024.
TRANSFORMING the Travel Experience: How Airports Are Getting Congested and What Are the Possible
Solutions. Iata, [s. l.], Jun. 21, 2023. Available at: https://tinyurl.com/478s8k7a. Retrieved: 14 nov. 2024
XU, Xiangdong et al. Transportation network redundancy: complementary measures and computational
methods. Transportation Research Part B: Methodological, [s. l.], v. 114, p. 68–85, 2018. Available at:
https://doi.org/10.1016/j.trb.2018.05.014. Retrieved: 11 jul. 2024.
YANG, Xin et al. Measuring route diversity for urban rail transit networks: a case study of the Beijing
Metro Network. IEEE Transactions on Intelligent Transportation Systems, [s. l.], v. 18, i. 2, p. 259–268,
2017. Available at: https://doi.org/10.1109/TITS.2016.2566801. Retrieved: 16 set. 2024.
