Professional Documents
Culture Documents
com
Available online at www.sciencedirect.com
ScienceDirect
ScienceDirect
Transportation
Available Research
online Procedia 00 (2017) 000–000
at www.sciencedirect.com
Transportation Research Procedia 00 (2017) 000–000 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
ScienceDirect
Transportation Research Procedia 25 (2017) 17–29
www.elsevier.com/locate/procedia
World Conference on Transport Research - WCTR 2016 Shanghai. July 10-15, 2016
World Conference on Transport Research - WCTR 2016 Shanghai. July 10-15, 2016
An Econometric Dynamic Model to estimate passenger demand for
An Econometric Dynamic Model to estimate passenger demand for
air transport industry
air transport industry
Rafael Bernardo Carmona-Beníteza,* , Maria Rosa Nietoa, Danya Mirandaa
a
Rafael Bernardo Carmona-Beníteza,*, Maria Rosa Nietoa, Danya Mirandaa
Universidad Anáhuac México Norte, School of Business and Economics, Av. Universidad Anáhuac, No. 46, Col. Lomas Anáhuac, Huixquilucan,
a
Universidad Anáhuac México Norte, School of Business52786,
and Economics, Av. Universidad
State of Mexico, Mexico Anáhuac, No. 46, Col. Lomas Anáhuac, Huixquilucan,
52786, State of Mexico, Mexico
Abstract
Abstract
In this paper, we propose an Econometric Dynamic Model (EDM) to estimate pax demand. We apply the panel
dataIn Arellano-Bover
this paper, we propose
method an
to Econometric
calibrate the Dynamic Model
EDM, which (EDM) to by
is validated estimate pax demand.
the Sargan test and We
the apply the panel
Arellano-Bond
data Arellano-Bover method to calibrate the EDM, which is validated by the Sargan test and the Arellano-Bond
Autocorrelation test. We propose a forecasting approach using the EDM to foresee pax demand. Finally, we show an
Autocorrelation
application of the test.
EDM We by
propose a forecasting
proposing approach
an approach usingthe
to solve theAirline
EDM to foresee
Airport pax(AAH)
Hub demand. Finally,
location we showThe
problem. an
application
case study isofabout
the EDM by proposing
the Mexican an approach
air transport to AAHs
industry, solve the
are Airline Airport
identified Hub
for each (AAH) year.
predicted location problem. The
case study is about the Mexican air transport industry, AAHs are identified for each predicted year.
© 2017 The Authors. Published by Elsevier B.V.
©
© 2017
2017 The
The Authors.
Authors. Published
Published by
by Elsevier B.V.
Peer-review
Peer-review under responsibility
under responsibility of Elsevier
of WORLDB.V.
WORLD CONFERENCE ON
CONFERENCE ON TRANSPORT
TRANSPORT RESEARCH
RESEARCH SOCIETY.
SOCIETY.
Peer-review under responsibility of WORLD CONFERENCE ON TRANSPORT RESEARCH SOCIETY.
Keywords: econometric model ; dynamic model ; Arellano-Bover ; time series methods ; air transportation ; hub location problem.
Keywords: econometric model ; dynamic model ; Arellano-Bover ; time series methods ; air transportation ; hub location problem.
1. Introduction
1. Introduction
The air transport passenger (pax) demand is in continuous growth. The increment has a direct impact on airlines
andThe air transport
airports. passenger
The market (pax) demand
expansion has made is it
in possible
continuousfor growth.
airportsThe increment
to develop andhas a direct
grow, impact
because on airlines
airlines have
and airports.
developed by The market
opening newexpansion hashas
routes. This made it possible
changed for airports to
the airport/airline develop and
relationship. Ongrow, because
the airports airlines
side, they have
have
developed by opening
changed from generating new routes.
their This
profits hasairlines
from changed to the airport/airline
reducing relationship.
the airlines On thepax
fees to increase airports
demandside,
andthey have
generate
changed from generating
non-aeronautical revenues. their
Onprofits from airlines
the airline side, thetocompetition
reducing thebetween
airlines fees to increase
airlines pax demand
has increased. and generate
Carmona-Benitez
non-aeronautical
(2012) mentionedrevenues. Onhave
that airlines the improved
airline side, thebusiness
their competition between
developing newairlines has toincreased.
strategies Carmona-Benitez
reduce costs and fares, and
(2012) mentioned
to maximize profitsthat
andairlines
markethave improved their business developing new strategies to reduce costs and fares, and
share.
to maximize profits and market share.
________
________
* Corresponding author. Tel.: +52-55-5627-0210x8884; fax: +52-55-5627-0210x7117.
* E-mail
Corresponding author.
address: Tel.: +52-55-5627-0210x8884; fax: +52-55-5627-0210x7117.
rafael.carmona@anahuac.mx
E-mail address: rafael.carmona@anahuac.mx
Nomenclature
The evolution of new airline business models has increased pax demand and therefore the need for more airport
infrastructure. Some airports have been bought by private firms, though most are owned by local and/or
state/regional governments (Neufville and Odoni, 2003). Some airlines have been subsidized by local governments
to fly to their airports and grow the economy by increasing the pax demand, air transport services, and business and
tourism activities (Papatheodorou and Lei, 2006). It is for these economic benefits that the main objective and
contribution of this paper is to develop an Econometric Dynamic Model (EDM) to estimate the pax demand based
on economic indicators. In this article, the EDM is calibrated using the panel data Arellano-Bover method to
estimate the pax demand at a State level. We decided to do it at a State level because usually economic indicators are
not reported locally. The Arellano-Bover method determines which economic variables are significant for estimating
the pax demand at a State level. The EDM is validated by the Sargan test and the Arellano-Bond Autocorrelation
test.
Normally, local governments, airports and airlines must invest large amounts of money to achieve the mentioned
economic benefits. Therefore, the pax demand must be forecasted to ensure that the demand is enough to overcome
investments. In this paper, we demonstrate how the EDM can easily be adapted to forecast the pax demand at a State
level.
Besides proposing the EDM and demonstrate how it can be used to forecast the pax demand at a State level, we
want to show an application of the EDM by proposing an approach to solve the Airline Airport Hub (AAH) location
problem. The solution to this problem is important because one of the airline strategies is to operate a hub-and-spoke
system. In such system, airlines consolidate traffic from different origins, sort it, and then send it to different
destinations, then spoke cities can have better services at lower prices while hub cities offer better services at higher
fares (Ehmer et al., 2008). Therefore, an AAH is where an airline centralizes its air traffic operations, pax and cargo
(Anton, 1992). For an airline, operating an airport as its hub means a business strategy, because it allows airlines to
offer more frequent flights and to expand the number of routes to more destinations (Kafani and Ghobrial, 1985;
Cook and Goodwin, 2008), the number of transfer pax increased the operational efficiency of the airline by reducing
costs through economies of scale (Tretheway and Kincaid, 2010), airlines can serve more routes and gain market
power at the hub (Ehmer et al., 2008), airlines increase aircrafts utilization given the connections, most pax are
Rafael Bernardo Carmona-Benítez et al. / Transportation Research Procedia 25 (2017) 17–29 19
R. B. Carmona-Benitez, M. R. Nieto, D. Miranda / Transportation Research Procedia 00 (2017) 000–000 3
transshipped at the AAH to their final destinations, the demand for flights is higher reducing unit costs per seat,
finally a centralized operation allows an airline to have maintenance facilities for both the aircrafts and the crew
(Cook and Goodwin, 2008). It is for these airport and airline benefits that we propose an approach to solve the AAH
location problem using the EDM.
The case study is first to apply the Arellano-Bover method to calibrate the EDM and determine the economic
indicators that explain the air pax demand in Mexico at each State; second, to use the EDM to forecast the pax
demand; and third, to determine in which Mexican State an AAH can be operated by one or more airlines and when.
This paper has the following structure: Section 2 describes previous works on economic indicators that have been
used to estimate the pax demand. Section 3 describes the experimental data. Section 4 develops the EDM to estimate
the pax demand per State. Section 4.1 describes how the EDM forecasts the pax demand per State over time. Section
4.2 describes the AAH location approach. Section 5 describes the results. Finally, Section 6 concludes the paper.
2. Literature review
Economic indicators determine the pax demand per route at a particular airport, region or city. The most common
economic indicators are Gross Domestic Product (GDP), GDP per capita, population (Pop), income, and income per
capita. There are many studies addressing the relation between the air pax demand with operating variables and/or
economic indicators. Airlines must know these indicators to predict the behavior of the pax demand, and then make
decisions such as whether it is profitable to establish their hub in a particular airport or not (Carmona-Benitez, 2012).
In this section, several studies to develop an EDM that find economic indicators that determine the pax demand in a
country, region or city are presented. Table 1 and Table 2 studies provide insight into some economic indicators that
have been used to estimate the pax demand.
The pax demand forecast is very important, and it can be done in the short, medium and long-term. A short-term
forecast is usually carried out for operational planning, while the medium and long-term are used to evaluate large
capital investments. When forecasting demand, there is no such thing as the best forecasting method. Multiple
forecasting methods are available and all are suitable for forecasting the pax demand, using any in particular depends
on the data structure.
20 Rafael Bernardo Carmona-Benítez et al. / Transportation Research Procedia 25 (2017) 17–29
4 R. B. Carmona-Benitez, M. R. Nieto, D. Miranda / Transportation Research Procedia 00 (2017) 000–000
The decision to locate an AAH depends on the pax demand on the routes that the airline offers at an airport. The
AAH location problem is a long-term decision because it involves large capital investments. From a quantitative
approach, locating an AAH is in function of operating variables and economic indicators. The operating variables
restrict the ability of an airport to be operated as a hub for one or more airlines. The main operating variables are the
runway capacity, goods handling capacity, passenger terminal capacity, quality of service for passengers, total
number of destinations, and fares (Bruinsma, 2000). Table 3 shows some papers that aim to solve the AAH location
problem, all of them based on mathematical formulations for optimization. Aykin (1993) indicates that the AAH
location problem depends on the pax demand per route, and shows that the number of routes are the main
determinants of an airport hub. Campbell (1994) states that a hub is determined by the pax demand between airport i
and airport j, which depends on the distance (L) and the time or cost associated with the passenger’s connection at
the hub. Campbell (1994) uses the p-hub median model which aims to minimize transport costs between i and j. The
main disadvantage of the model is to assume that all passengers have to be transshipped at the airport hub and no
direct flights are considered. Jaillet et al. (1996) Introduce flow-based models for designing capacitated networks
and routing policies. They design a network concerning a single airline with a fixed share of the market. One of the
main issues to consider when solving the AAH location problem is airport congestion. Marianov (2003) developed a
model that optimized congestion based on climate, aircraft sizes, taking off and landing time, turnaround time,
creating a queuing system with Poisson distribution. Finally, Yang (2009) addresses the air freight AHH problem
and flight routes planning under seasonal demand variations. The models determine the number and location of hubs
and flight routes. As it can be noticed, optimization theory is commonly used to locate AAHs. Contrary to these
papers, we use the EDM to locate AAHs based on econometric theory and statistics.
3. Experimental data
In this paper, because the data is not available per city, the economic indicators are per Mexican State: Indicator
of Economic Activity (ITAEE), indicator of economically active population (PEA), the national consumer price
index (INPC), foreign exchange earnings from international arrivals (ME), the hotel occupancy index of domestic
and foreign visitors (TUR), and pax demand. ITAEE offers a particular view on the macroeconomic situation and
evolution of each State (INEGI, 2012). This indicator is prepared based on surveys, censuses and administrative
records (INEGI, 2012). PEA measures a country's workforce which is made up of people of working age (14 years
and older in Mexico) (INEGI, 2012). The PEA is the sum of the employed population (OC) and unemployed
population (DE):
PEA = OC + DE (1)
Where:
OC = Employed population [person]
DE = Total number of people within legal age to work but without a job [person]
INPC measures the change in prices of a basket of goods and services representative of Mexican families over
time (INEGI, 2012). ME is the incoming cash flow of the country through checks, bank transfers, bills of exchange,
money orders and special international accepted payments (Banxico, 2015). TUR gives an overview of the evolution
of services offered to tourists and tourism consumption in Mexico (INEGI, 2012). This index is calculated using
data for both domestic and foreign visitors coming from tourist accommodation establishments (World Bank, 2014).
In Mexico, it is calculated and published by the National Institute of Statistics and Geography (INEGI) in
coordination with the Secretary of Tourism (SECTUR). The index is the number of tourists who travel to a country
or within a country for a period not exceeding 12 months. Finally, in this paper, the pax demand is defined as the
total number of people using airports located in the same state for a period of 4 months. All the variables in this
paper are presented on a quarterly basis from the first quarter of 2006 to the first quarter of 2014 state-wide.
In this paper, the econometric method of the panel data is used to determine the economic variables that
significantly explain the pax demand. This method is chosen because we have data over time and by State. A panel
data is a combination of time series with cross-sectional units (countries, states, cities or regions, business, etc.)
(Cobacho, 2000). The method of panel data has some advantages: control over individual heterogeneity of each
variable, more variability, less collinearity among the variables, more degrees of freedom, greater efficiency,
detection of effects that are not observable by using pure cross-sectional data or time series.
The model to be estimated is based on economic variables described in the previous section: ITAEE, PEA, INPC,
freq, TUR, and ME per State. The model is as follows:
The distribution of certain economic variables is sometimes highly skewed, then, logarithmic transformations are
convenient for transforming a highly skewed variable into one that is more approximately symmetric. Logarithm
transformations are also used to eliminate the constant variability implicit in each of the variables in the model.
Additionally, the coefficient of the log-transformed data can be interpreted as an elasticity (an expected percentage
change in Paxit when some independent variable increases by one percentage). Those are the reasons why
Equation 2 variables are transformed to their logarithmic form and then, the EDM to be estimated will be the
following:
In literature, there are different methods for estimating a panel data model. The method is selected depending on
the nature of the independent individual effect of time ui. On one hand, the fixed effects model is used when in the
static model (β1 = 0) and regressors or independent variables are correlated with ui. On the other hand, random
effects model is used whether regressors are not correlated with ui and β1 = 0. Additionally, in literature, dynamic
models have been proposed (β1 ≠ 0) because of the autoregressive component of the dependent variable. These
models take into account past effects for future estimations (Bhargava and Sargan, 1983; Ahn and Schmidt, 1995;
Arellano and Bond, 1991; Arellano and Bover, 1995). As before, ui may or may not be correlated with the
independent variables. Whether static methods are used to estimate the dynamic effect, the estimates that would be
obtained would be inconsistent because the term Paxit-1 is correlated with the error εit due to the presence of the
component ui. Because of this, different methods are proposed. Anderson and Hsiao (1981) propose to use Equation
3 in first difference (Equation 4). Because the estimates calculated with Equation 4 would be inconsistent due to the
correlation between Δlog(Paxit-1) and Δεit, they propose to estimate it with instrumental variables (IV) using it as an
instrument for the variable Δlog(Paxit-1), the difference Δlog(Paxit-2)= log(Paxit-2)-log(Paxit-3) or the variable
log(Paxit-2). These IV are not correlated with the error Δεit. One problem is that these estimates are consistent but not
necessarily efficient. Arellano (1989) demonstrates that estimators that use variables in levels have less variance
than the variables in differences.
Arellano and Bond (1991) also use Equation 4 but with an IV method with all available instruments in levels at
every moment in time, and they propose an estimator using the Generalized Method of Moments (GMM). Their
method improves the efficiency of the resulting estimator. Subsequently, Arellano and Bover (1995) and Blundell
and Bond (1998) explain whether the uncorrelated instruments with the individual effects ui are available, these
instruments can be used in the levels of Equation 3. Thus, combining the above moment conditions for difference
equations with the new moment conditions in levels allows a GMM consistent and efficient estimator.
A condition that must be met is that there cannot be a second order autocorrelation in the first differences of the
errors. It is important that the first differences of the errors are correlated in first order; otherwise it would indicate
that there are no dynamic effects and the GMM estimator would not be appropriate. Arellano (1990) develops a
hypothesis examination to test autocorrelation in errors, the null hypothesis asserts no autocorrelation, whether it is
not rejected, the EDM is correct. Additionally, the Sargan Test is used because the GMM estimator uses IV. This
test measures over identification in the EDM. The null hypothesis asserts the correct identification of the EDM, and
whether it is not rejected, the model is well identified.
Rafael Bernardo Carmona-Benítez et al. / Transportation Research Procedia 25 (2017) 17–29 23
R. B. Carmona-Benitez, M. R. Nieto, D. Miranda / Transportation Research Procedia 00 (2017) 000–000 7
The Arellano-Bover method estimates the parameters for each of the EDM explanatory variables of the behavior
of the pax demand. The purpose of this section is to explain how the EDM can be used to forecast pax demand over
time.
After the economic indicators are determined by the EDM, these indicator values are forecasted using one or
more methods of forecasting demand. The increase of the pax demand is calculated with these factors and with the
estimates of the parameters calculated by the EDM described in Section 3.2.
It is noteworthy to mention that since the purpose of this paper is not to develop a model to forecast the economic
variables values, in this paper the Holt-Winters multiplicative method (Winters, 1960; Chatfield, 1975; Montgomery
and Johnson, 1976) is applied for each of the independent variables in the context of the panel data. We chose this
method because it is commonly used for forecasting. However, it is best to use the more accurate forecasting
method, which depends on each economic variable time-series values. The most accurate forecasting method is
determined by the Diebold-Mariano Test (Diebold and Mariano, 1995) and/or the Hansen test (Hansen, 2005).
The purpose of this section is to suggest an approach to solve the AAH location problem. In other words, to
determine in which State an AAH could be located in the next few years depending on the increase in pax demand
over time.
The approach is divided into three steps: 1) The EDM is used to find the economic variables that determine the
pax demand per State; 2) A time-series method is chosen to forecast the economic variables values, the most
accurate according to the time-series values; 3) Based on the EDM forecasts and the airport classification of Federal
Aviation Administration (FAA) (Table 4), future AAHs are identified for each predicted year. The location of an
AAH is determined in the State whose forecast pax demand is equal to or greater than 1% of the total number of pax
transported nationwide per year, which is consistent according to the airport classification of Federal Aviation
Administration (FAA) (Table 4). We used the FAA airport classification because it is the only civil authority
providing an airport classification into hubs based on pax demand.
5. Results
As explained in Section 3, Table 5 shows the results of applying the Arellano and Bover (1995) model. The
variables logPEA and logME are removed from Equation 3 because logPEA is not significant and although logME is
significant, it is negative and therefore not consistent with the economic theory. We can notice that the variable
logfreq is simultaneous with the dependent variable logPax, meaning that we face an endogeneity problem. The
methodology developed by Arellano and Bover (1995) considers this problem and handles it using the IV procedure
explained in Section 4. Furthermore, as for the other explanatory variables, their p-values indicate that all are
significant, including the constant term. Thus, the model is as follows:
Table 8 shows the pax demand forecast by State for the years 2015, 2016 and 2017 using the EDM model. Table
8 demonstrates the percentage of the pax demand for each State based on the total annual demand in Mexico for the
years 2015, 2016 and 2017, and it indicates the type of hub airport that can be located at each State in accordance
with the airport classification of the FAA (Table 4) based only on the pax demand.
In Table 8, it can be observed that the proposed approach indicates that 16 States have enough pax demand to
classify their biggest airport as LH; therefore, according to the proposed approach, 16 AAHs can be located for the
year 2015, one AAH at each of these 16 States. It is because air pax forecasting demands are large enough for one or
more airlines to operate a hub and spoke network system as business strategy, at least at the biggest airport located
in one of these 16 States. In the same year, 10 States have a pax demand equivalent to MH airports, and 3 States
have a pax demand equivalent to SH airports. By 2016, it can be observed that the proposed approach locates 15
AAHs, because 15 States have enough pax demand to classify their biggest airport as LH. In this year, the pax
demand proportion of the State of Chihuahua decreases; therefore, the State of Chihuahua may be degraded from
LH to MH airport classification. The air pax forecasts indicate that the State of Morelos may have enough demand
to classify its airport as SH. In the same year, 11 States have a pax demand equivalent to MH airports, and 4 States
have a pax demand equivalent to SH airports. Finally, in 2017, the proposed approach predicts that the State of
Querétaro may have enough pax demand to operate their biggest airport as AAH. On the contrary, the results
indicate that the State of Mexico may be degraded from LH to MH airport classification. Consequently, 15 States
have enough pax demand to classify their biggest airport as LH, which means only 15 airports can be operated from
one or more airlines as AAH. In the same year, 11 States have a pax demand equivalent to MH airports, and 4 States
have a pax demand equivalent to SH airports.
From 2015 to 2017, the results of the application of the forecasting approach indicate that in Mexico the pax
demand is falling only in 4 States (Michoacán, Sonora, Tabasco and Zacatecas), and in 26 States the demand is
growing. Mexico City is perhaps the State with the highest pax demand growth followed by the State of Quintana
Roo, but in percentage, the States of Morelos and Baja California are perhaps the States with the highest pax
demand percentage growth. Mexico City is already an AAH for different airlines and its airport is congested.
Consequently, it may be difficult for a new airline to operate an AAH in Mexico City. The first option to locate an
AAH may be in the State of Quintana Roo because this State presents the second largest increment in pax demand
growth. Hence, Quintana Roo is perhaps the State that offers better possibilities for one or more airlines to operate
its biggest airport as AAH. Quintana Roo’s biggest airport is Cancun, and it is possible to conclude that Cancun is
then the airport that represents the first option, from one or more airlines, to operate a hub and spoke system. It is
noteworthy that nowadays, at least one Mexican airline is operating this airport as its AAH. The second option to
locate an AAH is in the State of Baja California because this State may have the third position in pax growth. Baja
California’s biggest airport is Tijuana, and it is possible to conclude that Tijuana may be the airport that represents
the second option, from one or more airlines, to operate a hub and spoke system. It is noteworthy that nowadays, at
least one Mexican airline is already operating this airport as its AAH.
26 Rafael Bernardo Carmona-Benítez et al. / Transportation Research Procedia 25 (2017) 17–29
10 R. B. Carmona-Benitez, M. R. Nieto, D. Miranda / Transportation Research Procedia 00 (2017) 000–000
State Forecast FAA Hub Forecast FAA Hub Forecast FAA Hub
2,015 % Type 2,016 % Type 2,017 % Type
Aguascalientes 403,789.67 0.46% MH 427,809.08 0.44% MH 452,282.99 0.44% MH
Baja California 4,020,445.49 4.60% LH 5,178,606.50 5.36% LH 6,568,263.74 6.32% LH
Baja California Sur 2,005,641.76 2.29% LH 2,098,149.19 2.17% LH 2,133,878.98 2.05% LH
Campeche 354,182.54 0.41% MH 396,663.17 0.41% MH 396,942.86 0.38% MH
Chiapas 864,674.71 0.99% MH 937,098.99 0.97% MH 1,003,615.73 0.97% MH
Chihuahua 907,224.32 1.04% LH 935,339.86 0.97% MH 1,002,279.28 0.97% MH
Coahuila 737,483.52 0.84% MH 816,944.53 0.85% MH 830,067.04 0.80% MH
Colima 321,569.05 0.37% MH 349,426.52 0.36% MH 355,324.07 0.34% MH
Distrito Federal 33,453,676.73 38.26% LH 36,543,294.78 37.84% LH 38,355,124.46 36.93% LH
Durango 178,965.23 0.20% SH 181,215.26 0.19% SH 184,230.04 0.18% SH
Estado de México 950,750.27 1.09% LH 995,436.04 1.03% LH 977,956.23 0.94% MH
Guanajuato 1,390,209.30 1.59% LH 1,521,608.01 1.58% LH 1,571,182.33 1.51% LH
Guerrero 1,322,972.18 1.51% LH 1,394,090.13 1.44% LH 1,355,920.13 1.31% LH
Jalisco 8,269,314.92 9.46% LH 8,510,761.01 8.81% LH 8,659,078.29 8.34% LH
Michoacán 501,667.24 0.57% MH 506,460.67 0.52% MH 470,021.34 0.45% MH
Morelos 40,110.62 0.05% - 57,484.76 0.06% SH 66,526.52 0.06% SH
Nayarit 150,008.35 0.17% SH 165,370.05 0.17% SH 168,657.86 0.16% SH
Nuevo León 4,478,432.31 5.12% LH 4,713,275.30 4.88% LH 5,061,801.59 4.87% LH
Oaxaca 1,059,647.17 1.21% LH 1,123,508.69 1.16% LH 1,160,501.83 1.12% LH
Puebla 413,153.42 0.47% MH 456,848.35 0.47% MH 461,703.01 0.44% MH
Querétaro 745,181.13 0.85% MH 927,182.99 0.96% MH 1,050,473.67 1.01% LH
Quintana Roo 14,308,205.57 16.37% LH 16,620,710.72 17.21% LH 19,130,023.02 18.42% LH
San Luis Potosí 482,473.76 0.55% MH 531,980.42 0.55% MH 552,226.87 0.53% MH
Sinaloa 2,008,608.12 2.30% LH 2,129,933.45 2.21% LH 2,185,724.81 2.10% LH
Sonora 1,519,606.67 1.74% LH 1,546,501.22 1.60% LH 1,500,231.55 1.44% LH
Tabasco 739,356.68 0.85% MH 733,174.25 0.76% MH 733,631.87 0.71% MH
Tamaulipas 1,671,085.59 1.91% LH 1,871,040.46 1.94% LH 1,989,718.39 1.92% LH
Veracruz 2,687,113.67 3.07% LH 3,199,329.78 3.31% LH 3,481,212.42 3.35% LH
Yucatán 1,227,490.07 1.40% LH 1,490,581.65 1.54% LH 1,791,424.25 1.72% LH
Zacatecas 215,952.36 0.25% SH 213,327.23 0.22% SH 204,425.72 0.20% SH
TOTAL PAX Nationwide 87,428,992.42 96,573,153.06 103,854,450.89
6. Conclusions
The main contribution of this paper is the proposition of an Econometric Dynamic Model to estimate pax demand
per Stat, but the EDM is not restricted to forecast pax demand per state, it can forecast per city, per airport, per
airline, etc. Due to the nature of data, we applied the panel data Arellano-Bover method to calibrate the proposed
EDM, and we validated it by applying the Sargan tests and the Arellano-Bond Autocorrelation test. The panel data
Arellano-Bover identifies the economic variables that are significant to air pax demand per location and these
variables linearly explain the behavior of the pax demand. The Arellano Bover chooses the model variables, among
other variables frequency is chosen. An airline chooses its hub at the airport where it can operate more frequently,
according to the hub and spoke network business model. The EDM considers the total airlines frequencies at the
airport or the total number of flights as explanatory variable for the pax demand. Hence, the forecast of the pax
Rafael Bernardo Carmona-Benítez et al. / Transportation Research Procedia 25 (2017) 17–29 27
R. B. Carmona-Benitez, M. R. Nieto, D. Miranda / Transportation Research Procedia 00 (2017) 000–000 11
References
Ahn, S., Schmidt, P., 1995. Efficient estimation of a model with dynamic panel data. Journal of Econometrics, 68, 5–27.
Alekseev, K., Seixas, J., 2002. “Forecasting the air transport demand for passengers with neural modelling”. In: The VII Brazilian Symposium on
Neural Networks IEEE Computer Society Washington: Recife, Brazil, 86-91.
Anderson, T., W., Hsiao, C., 1981. Estimation of dynamic models with error components. Journal of the American Statistical Association, 76,
598–606.
Anton, B., F., 1992. “El sistema hub and spoke en el transporte aéreo”. In: Anales de Geografía de la Universidad Complutense. Madrid.
Arellano, M., 1989. A Note on the Anderson-Hsiao Estimator for Panel Data. Economics Letters, 31, 337-341.
Arellano, M., 1990. Testing for Autocorrelation in dynamic random effects models. Review of Economic Studies, 57, 127-134.
Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The
Review of Economic Studies, 58, 277–297.
Arellano, M., Bover, S., 1995. Another look at the instrumental variable estimation of error-components models. Journal of Econometrics, 68,
29–51.
Aykin, T., 1993. The hub location and routing problem. European Journal of Operational Research, 83, 200-219.
Banxico, 2015. Glosario (Retrieved from http://www.banxico.org.mx/divulgacion/glosario/glosario.html)
Beckmann, M., Bobkoski, F., 1958. Airline demand: An analysis of some frequency distributions. Naval research logistics quarterly, 5 (1), 43-51.
28 Rafael Bernardo Carmona-Benítez et al. / Transportation Research Procedia 25 (2017) 17–29
12 R. B. Carmona-Benitez, M. R. Nieto, D. Miranda / Transportation Research Procedia 00 (2017) 000–000
Bhargava, A., Sargan, J., D., 1983. Estimating dynamic random effects models from panel data covering short time periods. Econometric, 51,
1635-1659.
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87, 115-143
Bruinsma, F., Rietveld, P. Brons, M., 2000. Comparative study of hub airports in Europe: ticket prices, travel time and rescheduling costs.
Tijdschrift Voor Economische En Sociale Geografie. Journal of Economic & Social Geography, 91 (3), 278.
Campbell, J., 1994. Integer programming formulations of discrete hub location problems. European Journal of Operational Research, 72, 387-405.
Carmona-Benitez, R., B., 2012. “The Design of a Large Scale Airline Network”. Thesis, Delft University of Technology. In: Thesis Series T
2012/2, Delft.
Carmona-Benitez, R., B., Carmona-Paredes, R., B., Lodewijks, G., 2013. Damp trend Grey Model forecasting method for airline industry. Expert
Systems with Applications, 40, 4915-4921.
Carmona-Benitez, R., B., Nieto, M., R., 2015. “Bootstrap estimation intervals using bias corrected accelerated method to forecast air passenger
demand”. Lecture Notes in Computer Science, 9335, 315-327.
Chatfield, C., 1975. The Holt-Winters forecasting procedure. Applied Statistics, 27, 264-279.
Cobacho, T., M., B., 2000. Contrastes de hipótesis en datos de panel. XIII Jornadas de ASEPUMA.
Coldren, G., Koppelman, F., Kasturirangan, K., Mukherjee, A., 2003. Modelling aggregate air travel itinerary shares: logit model development at
a major US airline. Journal of Air Transport Management, 9 (6), 841-851.
Cook, G., N., Goodwin, J., 2008. Airline networks: A comparison of Hub-and-Spoke and Point-to-Point Systems Airline Networks: A
Comparison of Hub-and-Spoke and Point-to-Point Systems. Journal of Aviation/Aerospace Education & Research, 17 (2) (Retrived from
http://commons.erau.jaaer/vol17/iss2/1).
Diebold, F. X., Mariano R., S., 1995. Comparing Predictive Accuracy. Journal of Business and Economic Statistics, 13, 253-263.
Ehmer, H., P., Berster, G., Bischoff, W., Grimme, E., Grunewald, Maertens, S., 2008. Analysis of the European Air Transport Market, Airline
Business Models: Topical Report Prepared for the Directorate-General for Energy and Transport in the European Commission, DLR-German
Aerospace Center, Cologne.
FAA Airport Classification. http://www.faa.gov/airports/planning_capacity/passenger_allcargo_stats/categories/
Gardner, E., Mckenzie, E., 1989. “Seasonal exponential smoothing with damped trends”. In: Management Science, 35, 372-376.
Grosche, T. Rothlauf, F., Heinzl, A., 2007. Gravity models for airline passenger volume estimation. Journal of Air Transport Management, 13,
175-183.
Grubb, H., Mason, A., 2001. Long lead-time forecasting of UK air passengers by Holt-Winters methods with damped trend. International Journal
of Forecasting, 17, 71-82.
Hansen P., R., 2005. A test for superior predictive ability. Journal of Business and Economic Statistics, 23, 365-380.
Hsu, C., I., Wen, Y., H., 2000. Application of Grey theory and multiobjective programming towards airline network design. European Journal of
Operation Research, 127 (1), 44-68.
Hwang, C., Shiao, G., 2011. Analyzing air cargo flows of international routes: an empirical study of Taiwan Taoyuan International Airport.
Journal of Transport Geography, 19, 738-744.
INEGI, 2012. Indicadores Trimestrales de la Actividad Turística.
(Retrieved from http://www.inegi.org.mx/est/contenidos/proyectos/cn/itat/doc/SCNM_Metodologia_22.pdf )
Jaillet, P., Song, G., Yu, G., 1996. Airline network design and hub location problem. Location Science, 4 (3), 195-212.
Janic, M., 2007. “The sustainability of Air Transportation, A quantitative analysis and assessment”. 1 Ed., Ashgate Publishing Limited, England.
Kanafani, A., Ghobrial, A., A., 1985. Airline hubbing – Some implications for airport economics. Transport Research Part A: Policy and Practice,
19, 15-27.
Marianov, V., 2003. Location models for airline hubs behaving as M = D = c queues. Computers & Operational Research, 20, 983-1003.
Montgomery, D., C., Johnson, L., A., 1976. “Forecasting and time series analysis”. New York, McGraw –Hall.
Neufville, R., Odoni, A., 2003. Airport Systems Planning, Design and Management. McGraw Hill, New York.
Papatheodorou, A., Lei, Z., 2006. Leisure travel in Europe and airline business models: A study of regional airports in Great Britain. Journal of
Air Transport Management, 12, 47-52.
Profillidis, V., A., 2000. Econometric and Fuzzy models for the forecast of demand in the airport of Rhodes. Journal of Air Transport
Managmenet, 6, 95-100.
Samagio, A., Wolters, M., 2010. Comparative analysis of government forecasts for Lisbon airport. Journal of Air Transport Management, 16 (4),
213-217.
Shen, G., 2004. Reverse-fitting the gravity model to inter-city airline passenger flows by an algebraic simplification. Journal of Transport
Geography, 16 (4), 213-217.
Srinidhi, S., 2009. “Development of an airline traffic forecasting model on international sector”. In: The 2009 IEEE International Conference on
Automation Science and Engineering, IEEE Computer Society Washington, 322-327.
Swan, W., M., 2002. Airline route developments: a review of history. Journal of Transport Management, 8, 349-353.
Tretheway, M., Kincaid, I., 2010. “Competition between airports: Occurrence and strategy”. In: Airport competition: The European Experience,
Forsyth, P., Gillen, D., Muller, J., Niemeier, H. (eds.), Aldershot: Ashgate.
Wei, Z., Jinfu, Z., 2009. “Passenger traffic forecast based on the Grey-Markov Method”. In: IEEE International Conference on Grey Systems,
630-633.
Winters, P., 1960. Forecasting Sales by exponentially weighted moving averages. Management Science, 6, 324-342.
World Bank, 2014. http://www.bancomundial.org/
Rafael Bernardo Carmona-Benítez et al. / Transportation Research Procedia 25 (2017) 17–29 29
R. B. Carmona-Benitez, M. R. Nieto, D. Miranda / Transportation Research Procedia 00 (2017) 000–000 13
Xiaowen, F., Tae Hoon, O., Anming, Z., 2010. Air Transport Liberalization and its impacts on Airline Competition and Air Passenger Traffic.
Transportation Journal, 49 (4), 24-41.
Yang, T., 2009. Stochastic air freight hub location and flight routes planning. Applied Mathematical Modelling, 33, 4424-4430.
Zeni, R., H., 2001. “Improving forecasting accuracy in revenue management by unconstraining demand estimates from censored data”. In:
Published by Dissertation.com, PhD in management program, New Jersey.