To appear in Transportation Research Part E: Logistics and Transportation Review 1 Passenger Facility Charge vs. Airport Improvement Program Funds: A Dynamic Network DEA Analysis for U.S. Airport Financing Young-Tae Chang 1 , Hyosoo (Kevin) Park 1 , Bo Zou 2 , Nabin Kafle 2 , 1 Graduate School of Logistics, Inha University, Incheon, Korea 2 Department of Civil and Materials Engineering, University of Illinois at Chicago, Chicago, USA Abstract: Passenger Facility Charge (PFC) and the Airport Improvement Program (AIP) are two major sources to finance U.S. airports. This paper develops a novel dynamic network DEA framework to investigate the substitutability between PFC and AIP funds. We find that the studied U.S. airports can substitute PFC for 8-35% of the current AIP funds and contribute significantly to the proposed plan of the US congress to cut AIP funding. In addition, the amount of PFC-for-AIP funds substitution negatively correlates with the productive efficiency of airports. The findings send an important message for future policy reforms on U.S. airport financing. Keyword: Passenger Facility Charge (PFC), Airport Improvement Program (AIP), Substitution, Dynamic network DEA, Airport efficiency 1 Introduction Passenger Facility Charge (PFC) and the Airport Improvement Program (AIP) are two major sources for airport financing in the U.S. PFC is a service fee charged to departing and connecting passengers at an airport, collected by airlines and then forwarded to the airport. AIP is a U.S. federal grant program administered by the Federal Aviation Administration (FAA) to provide funds to airports. PFC funds a broad range of capacity enhancing projects, for both airside and landside at airports. PFC can also be used for making payment for airport debt services. The use of AIP funds is similar to PFC, but usually limited to the planning and construction of projects related to aircraft operations improvement, covering runways, taxiways, aprons, noise abatement, land purchase, etc. Commercial revenue producing facilities (such as those related to shop concessions) are not eligible for AIP funding (Kirk, 2010). Together, the two financing sources account for about 46% in total airport funds in the U.S. (Dillingham, 2007). PFC and AIP funds have a complementary relationship. This is because PFC and the formula part of AIP funds, which accounts for 70-76% in total AIP funds (Kirk, 2009), are levied and allocated based on passenger enplanement. Moreover, according to the FAA rule, AIP funds allocated to large and medium airports will be foregone by 50% or 75% if those airports charge PFC at $3.00 or $4.50 per enplanement (Kirk, 2009). The complementary relationship has brought to the arena of policy debate the issue of potential substitution between the two types of airport financing sources, especially given
30
Embed
Passenger Facility Charge vs. Airport Improvement Program …translog.uic.edu/wp-content/uploads/sites/18/2016/09/TRE_Passenger... · Passenger Facility Charge vs. Airport Improvement
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
To appear in Transportation Research Part E: Logistics and Transportation Review
1
Passenger Facility Charge vs. Airport Improvement
Program Funds: A Dynamic Network DEA Analysis for
U.S. Airport Financing
Young-Tae Chang1, Hyosoo (Kevin) Park1, Bo Zou2, Nabin Kafle2,
1 Graduate School of Logistics, Inha University, Incheon, Korea
2 Department of Civil and Materials Engineering, University of Illinois at Chicago, Chicago, USA
Abstract: Passenger Facility Charge (PFC) and the Airport Improvement Program (AIP) are two major
sources to finance U.S. airports. This paper develops a novel dynamic network DEA framework to
investigate the substitutability between PFC and AIP funds. We find that the studied U.S. airports can
substitute PFC for 8-35% of the current AIP funds and contribute significantly to the proposed plan of
the US congress to cut AIP funding. In addition, the amount of PFC-for-AIP funds substitution
negatively correlates with the productive efficiency of airports. The findings send an important message
for future policy reforms on U.S. airport financing.
Keyword: Passenger Facility Charge (PFC), Airport Improvement Program (AIP), Substitution,
Dynamic network DEA, Airport efficiency
1 Introduction Passenger Facility Charge (PFC) and the Airport Improvement Program (AIP) are two major
sources for airport financing in the U.S. PFC is a service fee charged to departing and connecting
passengers at an airport, collected by airlines and then forwarded to the airport. AIP is a U.S. federal
grant program administered by the Federal Aviation Administration (FAA) to provide funds to airports.
PFC funds a broad range of capacity enhancing projects, for both airside and landside at airports. PFC
can also be used for making payment for airport debt services. The use of AIP funds is similar to PFC,
but usually limited to the planning and construction of projects related to aircraft operations
improvement, covering runways, taxiways, aprons, noise abatement, land purchase, etc. Commercial
revenue producing facilities (such as those related to shop concessions) are not eligible for AIP funding
(Kirk, 2010). Together, the two financing sources account for about 46% in total airport funds in the
U.S. (Dillingham, 2007).
PFC and AIP funds have a complementary relationship. This is because PFC and the formula part
of AIP funds, which accounts for 70-76% in total AIP funds (Kirk, 2009), are levied and allocated based
on passenger enplanement. Moreover, according to the FAA rule, AIP funds allocated to large and
medium airports will be foregone by 50% or 75% if those airports charge PFC at $3.00 or $4.50 per
enplanement (Kirk, 2009). The complementary relationship has brought to the arena of policy debate
the issue of potential substitution between the two types of airport financing sources, especially given
To appear in Transportation Research Part E: Logistics and Transportation Review
2
the pressure for the federal government to reduce outlay on AIP and the fact that the current PFC scheme
remains unchanged for many years. Indeed, AIP funds have experienced a decline in recent years (GAO,
2014; AMAC, 2014) and the $4.50 cap on the amount of PFC that airports can levy per enplanement
has not been increased since 2000, when the Congress passed the Wendell H. Ford Investment and
Reform Act for 21 Century (AIR-21).
There have been many advocates for lifting PFC by itself or in place of AIP funds. They come
down to mainly two arguments. The first argument relates to inflation: the Airport Council International
– North America (ACI-NA) believes that the $4.50 ceiling set back in 2000 should be raised to $7.00
(ACI-NA, 2015); Airports United, a national airport group, argues that PFC should be “modernized”
by raising it from $4.50 to $8.50 (Laing, 2015). The second argument for increasing PFC is that PFC
can reduce federal expenditures on AIP and thus help alleviate the possible funding shortage. In line
with this argument, the Congressional Research Services has considered partial defederalization as an
option for future airport financing, by allowing large and medium hub airports to opt out of the AIP
program in favor of unrestricted higher PFC (Kirk, 2009). Two additional evidences in support of the
second argument: the H.R. 608 FAA Reauthorization and Reform Act in 2011 proposed annual AIP
funding cut by $500 million from 2012 through 2014; in 2015 the White House’s budget proposal called
for elimination of AIP support for large hub airports in return for an $8.00 PFC (GAO, 2014).
Despite the advocates and arguments on potential reforms of airport financing policy – in particular
the substitution between PFC and AIP funds – there is almost no academic literature that investigates
the implications of PFC-for-AIP funds substitution for airport production efficiency. The only study we
are aware of is Zou et al. (2015), who use second-stage regression in a two-stage Data Envelopment
Analysis (DEA) approach and find positive and negative impacts of PFC and AIP funds on airport
production efficiency. The authors argue that the finding is consistent with the greater flexibility airports
have in utilizing PFC than AIP funds. However, the study does not provide answers to the question of
substitutability between PFC and AIP funds. In addition, the dynamic relationship between PFC and
AIP funds is not recognized: determination of AIP funds for an airport in the current year depends on
passenger enplanement in the previous year, which directly relate to the previous year’s PFC revenues.
The present paper intends to fill these gaps and contributes to the literature on airport efficiency
modeling and financing in six ways. First and foremost, we look into the substitutability of PFC for AIP
funds, by taking into explicit consideration the existing complementary relationship between these two
airport financing sources. Second, we specifically account for the inter-temporal dynamics of airport
passenger enplanement, PFC, and AIP funds allocation, i.e., the fact that passenger enplanement
determine both PFC revenue in the current year and AIP funds in the following year. Third, we specify
sub-structures while modeling the airport production process, which allows for more detailed
characterization of how airports use inputs to generate intermediate and final outputs. The three
contributions are integrated into a dynamic network Data Envelopment Analysis framework, which is
the first time in the airport efficiency literature. Fourth, the results from our modeling provide important
policy insights that support the growing arguments for PFC-for-AIP funds substitution in the US airport
sector. Fifth, using random effect Tobit regression, the relationship between the amount of substitutable
AIP funds and airport efficiency scores is estimated. Finally, while focusing on PFC-for-AIP funds
substitution, the methodological framework proposed in this study can be generalized to investigate
substitutability between any two complementary resources in transportation and non-transportation
production systems.
To appear in Transportation Research Part E: Logistics and Transportation Review
3
The paper continues with a review of existing literature on airport efficiency modeling in Section
2. The dynamic network structure, and subsequently three models on PFC-for-AIP funds substitution,
airport efficiency measurement, and the substitution-efficiency relation are presented in Section 3.
Section 4 presents the data, followed by modeling results and discussions in Section 5. Conclusions and
directions for future research are given in Section 6.
2 Literature review The field of airport efficiency modeling has an established body of literature. Two most popular
methods are DEA and stochastic frontier analysis (SFA). Gillen and Lall (1997) are among the earliest
researchers introducing DEA to measure airport efficiency and derive relevant performance indices.
Both airport terminal efficiency and movement efficiency are quantified. Sarkis (2000) uses a variety
of DEA models to find the effect of airport hub status, being in a snowbelt, and being part of a multi-
airport system on the operational efficiency of U.S. airports. Later Sarkis and Talluri (2004) apply
clustering analysis to DEA results to identify the benchmarking airports with similar resource utilization
structure. Different from DEA which draws frontiers using deterministic mathematical programming,
SFA incorporates random errors in deriving the airport production frontier line. Pels et al. (2003) use
SFA to estimate the efficiency and investigate the returns-to-scale properties for air transport
movements and passenger movements at 34 European airports. SFA is also employed by Oum et al.
(2008) to explore the relationship between airport ownership structure and airport efficiency, and by
Scotti et al. (2012) to assess the impact of airport competition on the technical efficiency of Italian
airports. All these studies consider airport production as an “all-in-one” process, i.e., inputs are
converted to outputs without dealing with what occurs inside the production process. On the other hand,
airport production can be described in more detail, for example, one can consider that airports first use
labor and capital to attract flight traffic, which then produces passenger enplanement and cargo
throughput.
The theoretical DEA literature has evolved recently into incorporating internal structures and inter-
temporal dynamics of production processes. DEA models with internal structures is frequently termed
“network DEA” and models with inter-temporal dynamics named “dynamic DEA”. Network DEA
models are first introduced by Färe and Grosskopf (2000), who enrich traditional all-in-one DEA
models by enabling the characterization of sequential or parallel processes in production. Liang et al.
(2006) develops a network DEA model to measure efficiency of firms in a supply chain. The model
considers the case that a supplier and a retailer are under coordination or in a leader-follower
relationship. The supplier’s improvement influences that of the retailer’s, and the way the influence
works depends on the relationship specified. Yu and Lin (2008) evaluate the performance of European
railways using a network DEA model. The model divides the railway operation into passenger, freight,
and consumption processes, and allocates common inputs (e.g., the number of employees and length of
line) to different operations. Tone and Tsutsui (2009) develop a network model using slacks-based
measures (SBM), which has the advantage of not requiring all inputs or outputs to be improved equi-
proportionally.
Dynamic DEA models consider how production of a firm in one period affects its production in
the next period while measuring efficiency. Nemoto and Goto (1999) first extend traditional DEA to a
dynamic framework by incorporating the adjustment cost of investment and inter-temporal substitution.
Färe et al. (2007) and Tone and Tsutsui (2010) propose dynamic SBM models to capture inter-period
“carryovers”. Applications of dynamic DEA abound, in fields such as electric utilities (Nemoto and
To appear in Transportation Research Part E: Logistics and Transportation Review
4
Goto, 2003), agriculture production (Jaenicke, 2000), and retail department stores (de Mateo et al.,
2006). Recently Lu et al. (2014) use dynamic SBM to evaluate life-insurance firms and recognize the
fact that firm debt and equity in one year will be transcended to the next year.
The DEA literature has further seen combining the network and dynamic aspects of firm
production into a single DEA framework. Färe and Grosskopf (2000) and Färe et al. (2007) lay the
stepping stones of the dynamic network DEA, with adapted definitions of production possibility sets
and efficiency measures. Bogetoft et al. (2009) develop a dynamic network DEA model to examine the
optimal investment path for U.S. manufacturers. The authors decompose total outputs of the chosen
U.S. manufacturers into final outputs (consumption), private investment, and public investment.
Production in one period generates final outputs and investment, which are used as inputs along with
capital stock for the next period. The model then finds the amount of investment that maximizes the
final outputs throughout the investigation periods. Dynamic network DEA has also been considered
lately in the context of SBM (Tone and Tsutsui, 2014). Interested readers may refer to Cook et al. (2010)
and Kao (2014) for detailed review on network and dynamic DEA models.
The literature reports several applications of network DEA to measuring airline performance. Zhu
(2011) uses a virtual price network DEA model to analyze the efficiency of 21 airlines in the world.
The production network comprises two sub-processes: first operational costs producing load factor and
fleet size, which then generate revenue and revenue passenger miles. Tavassoli et al. (2014) develops a
two-stage network SBM model with shared inputs to assess performance of 11 domestic airlines in Iran.
The first stage evaluates airline technical efficiency in relation to capacity building. The second stage
measures the service effectiveness, i.e., how efficiently a given capacity generates passenger and cargo
flows. Mallikarjun (2015) uses an unoriented network DEA model to measure the efficiency of U.S.
domestic airlines, with a three-stage production process: operation, service, and sales. The consideration
of the sales stage is particularly interesting because the ultimate goal of an airline is maximizing revenue
and all other airline network DEA papers confine their analysis up to the service stage. Li et al. (2015)
follow the network structure in Mallikarjun (2015) and develop a virtual frontier SBM model to assess
the performance of 22 international airlines. A hypothetical frontier is constructed to discriminate
efficiency scores among efficient airlines, by using maximum output and minimum input sets. We note
that none of the above studies consider inter-temporal features of the airline production, although this
naturally exists in airline businesses. For instance, airline debt can be carried over from one period to
another.
Network DEA has also been applied to the airport sector. Yu (2010) uses network SBM model to
assess the efficiency of domestic airports in Taiwan. In the paper airport operations are divided into
“production” and “service” stages. The service stage is further disaggregated into airside and landside
services. The author finds that service efficiency is far lower than production efficiency for Taiwanese
airports. Lozano et al. (2013) develops a directional distance function DEA model with a network
structure. Airport operations are divided into “aircraft movements” and “aircraft loading” stages. The
model incorporates the possibility that decreasing airport flight delay restricts the number of aircraft
movements, and shows that network DEA model has more discriminatory power than traditional, all-
in-one DEA model. A virtual price network DEA model is applied to Brazilian airports by Wanke
(2013). It delineates airport operations into two stages pertaining to “physical infrastructure” and “flight
consolidation”. Flight regularity, location, and international status are found to have positive impacts
on the efficiency of flight consolidation. Some of the major features of the above three airport DEA
studies are summarized in Table 1. We note that the network structures in Lozano et al. (2013) and
To appear in Transportation Research Part E: Logistics and Transportation Review
5
Wanke (2013) are largely consistent, with the first stage producing aircraft movements and the second
stage generating passenger enplanement and cargo throughput. Yu (2010) put aircraft movements as an
output in the second stage. The choice of inputs differs in the three studies, which may be because of
data availability. Again, it should be noted that none of these studies consider inter-temporal dynamics
in the airport production process.
Table 1. Application of network DEA in airport analysis
Yu (2010) Lozano et al. (2013) Wanke (2013)
Data Taiwanese airports Spanish airports Brazilian airports
First stage Production Aircraft movement Physical infrastructure
Second stage Service Aircraft loading Flight consolidation
Input Number of employees, runway
area, apron area, terminal area
Runway area, apron
capacity, number of
boarding gates,
number of baggage
belts, number of
check-in counters
Terminal area, aircraft
parking spaces, number of
runways
Intermediate
output between
stages
Runway capacity, terminal
capacity Aircraft movements
Number of aircraft landings
and take-offs
Output Aircraft movements, passenger
movements, cargo volume
Passenger movements,
cargo handled,
number of delayed
flights, accumulated
flight delay minutes
Number of passengers, cargo
throughput
For airport finance in the U.S., as mentioned before academic research to support and inform the
heated policy debate on whether and ways to reduce federal government aids is almost non-existent.
Specifically for the potential substitution of PFC for AIP funds and its implication for airport production
efficiency, we are only aware of Zou et al. (2015) as a relevant study. By developing a two-stage DEA
model, the authors find a positive linkage between airport production efficiency and PFC use, but a
negative linkage between airport production efficiency and the use of AIP funds. Although the study
investigates indirectly the potential change in airport efficiency under hypothetical PFC-for-AIP funds
substitution scenarios, it does not examine the substitutability between the two airport financing sources
conditional on production feasibilities. In addition, the study does not capture any inter-temporal
interactions nor considers the sub-structures of the airport production process.
3 Model This section presents the model formulation. We first specify in subsection 3.1 the dynamic
network structure of the airport production and the choice for production inputs/outputs. Then
To appear in Transportation Research Part E: Logistics and Transportation Review
6
subsections 3.2 and 3.3 present the mathematical models for PFC-for-AIP funds substitution and airport
efficiency measurement. The PFC-for-AIP funds substitution is from the FAA perspective, whereas
airport efficiency measurement is from the standpoint of individual airports. Building on these models,
subsection 3.4 further specifies a random effect Tobit regression model to estimate the relationship
between the amount of substitutable AIP funds and airport efficiency.
3.1 Dynamic network structure
The dynamic network structure of the DEA model considered in our study is shown in Figure 1.
In terms of the network feature, we divide airport production into two stages, termed “aircraft movement”
and “aircraft loading”, based on the previous airport efficiency studies using network DEA (Lozano et
al. (2013) and Wanke (2013), as in Table 1). For the dynamic feature, we explicitly consider the fact
that airport production in year t affects airport production in t+1, through passenger enplanement.
Figure 1. The dynamic network structure of the DEA model
The first stage in the dynamic network structure, “aircraft movement”, uses labor, materials and
capital (net assets) inputs to produce aircraft operations. In addition, the first-stage production generates
flight delay as an undesirable output. The second stage, “aircraft loading”, describes the process of how
aircraft movement produces passenger and cargo flows. At the second stage, promotion is further
considered as an external input, which is defined as the amount of money spent on marketing and
advertising to attract passenger and cargo demand. The two-stage network structure implies that airports
do not generate passengers and cargoes directly from using labor, materials, and capital inputs; rather,
aircraft movements mediate between use of these inputs and passenger/cargo flows.
In terms of inter-temporal dynamics, passenger enplanement in year t determines PFC in year t
and AIP funds in year t+1. Since PFC is levied on a per passenger basis, the number of passengers
enplaned in year t times the amount of PFC levied per passenger gives total PFC in year t. By contrast,
To appear in Transportation Research Part E: Logistics and Transportation Review
7
AIP funds are allocated by the FAA to airports in year t+1 based on the passenger enplanement in year
t. This is why we position “PFC determined” at the end of year t, and “AIP determined” at the beginning
of year t+1 in Figure 1. It should be noted that both AIP funds and PFC are included in capital (net
assets) in the FAA data reporting (see Section 4). Thus there is no need to consider AIP funds and PFC
separately as two additional inputs.
3.2 PFC-for-AIP funds substitution model
3.2.1 Defining production possibility set
The first step to specify the PFC-for-AIP funds substitution model is to define the production
possibility set (PPS), in which airports are assumed to be operable. Defining PPS is important because
airports should be ensured that the optimal PFC-for-AIP funds substitution path is attainable.
As mentioned above, the airport production process is decomposed into two stages. At stage k
(k=1,2) and in year t, an airport o uses a combination of inputs ( )t t t
k1o k2o kIox ,x ,...,xt
kox to produce
outputs ( ).t t t
k1o k2o kRoy , y ,..., yt
koy In particular, an intermediate output 1
t
oz , i.e., aircraft operations, is
produced from stage 1, which is used as an input for stage 2. Then PPS for the airport is the set of
vectors 1( , , )t
ozt t
ko kox y satisfying the following constraints:
1
1
1 2
1 1
(i) , , , ,
(ii) , , , ,
(iii) ,
nt t t
kj kij kio k
j
nt t t
kj krj kro k
j
n nt t t t
j j j j
j j
x x i I k t
y y r R k t
z = z t
(1)
where t
kj is a weight imposed on airport j = 1,…,n to construct the frontier space. n is the number of
airports in the dataset. kI and
kR represent the sets of inputs and outputs in stage k. For inequalities
(1)-i and (1)-ii, the left-hand side (LHS) corresponds to the sets of airports that lie on the production
frontier. The crucial difference between traditional all-in-one DEA models and the network DEA model
presented here is the role of intermediate outputs described in equality (1)-iii. The equation indicates
that the amount of intermediate outputs produced in stage 1 on the frontier should be equal to the amount
used as inputs in stage 2 on the frontier, although in general the intermediate output could be lost or
added from external sources (Tone and Tsutsui, 2009; 2014). In our study, the intermediate output from
stage 1 is aircraft operations, which are not possible to change once produced. Thus the equality
relationship (1)-iii is used. Note that (1)-iii does not require the intermediate outputs to be lower or
higher than the observed level. Tone and Tsutsui (2009; 2014) name this “free link.” If the intermediate
output is restricted to be equal to the observed amount, it will be “fixed link.” To specify the fixed link,
(1)-iii will be substituted by the following constraint:
1 2
1 1
, .n n
t t t t t
j j j j o
j j
z = z z t
(2)
where toz is the observed amount of the intermediate output at time t. As we do not want aircraft
operations to change by PFC-for-AIP funds substitution, the fixed link assumption is more appropriate
To appear in Transportation Research Part E: Logistics and Transportation Review
8
than the free link assumption. In other words, we replace constraint (1)-iii by constraint (2) in
characterizing the PPS for each airport.
3.2.2 Model formulation
The objective of the PFC-for-AIP funds substitution model is to find the maximum amount of AIP
funds that can be replaced by PFC at each airport, while guaranteeing that the airport production is
feasible. In this study, we view that the substitution is from the FAA perspective, and this substitution
departs from the existing complementary relationship between PFC and AIP funds stipulated by the
FAA formulas and rules. The mathematical expressions for the existing complementary relationship
between PFC and AIP funds are rather complex. Details about the relationship is provided in Appendix
A.
To be more specific about the departure from the existing complementary relationship between
PFC and AIP fund through substitution, we define receivable and desirable PFC (AIP funds).
Receivable PFC and AIP funds are PFC and AIP funds that an airport can collect under the current FAA
formulas and rules. In other words, they denote the amounts that we currently observe. Desirable PFC
and AIP funds are PFC and AIP funds when PFC-for-AIP funds substitution were in place. We use the
term “desirable” since we consider it to be the policy direction that the FAA wants to pursue. In sum,