This action might not be possible to undo. Are you sure you want to continue?
:
Evaluating the highspeed rail network in Spain
CAMPOS, Javier; HERNÁNDEZ, Aday.
1
Costbenefit analysis and demand
uncertainty: Evaluating the highspeed
rail network in Spain
Javier Campos (University of Las Palmas de Gran Canaria, Spain)
1
Aday Hernández (University of Las Palmas de Gran Canaria, Spain)
2
ABSTRACT
Highspeed rail (HSR) is undoubtedly regarded as one of the most significant
technological breakthroughs in passenger transportation developed in the second half
of the 20
th
century. However, building, maintaining and operating HSR lines is very
expensive, involves a significant amount of sunk costs and may substantially
compromise both the transport policy of a country and the development of its transport
sector for decades. For these reasons evaluating HSR projects deserves a closer look,
where the technological hype is adequately balanced against the inevitable uncertainty
that dominates demand figures. Under these premises the aim of this paper is to build
on the current CBA methodology for the appraisal of HSR projects, and develop some
new indicators about the circumstances under which such proposals might be socially
worthwhile. Our approach departs from traditional CBA but changes its perspective:
since demand uncertainty is the critical variable to accept or reject most projects, we
transform the traditional NPV>0 acceptance criterion into „demanddriven criteria‟ and
obtain different values for this critical demand under alternative simulated scenarios.
Our analysis is illustrated with examples from the Spanish HSR network, currently one
of the most ambitious projects in the world.
Keywords: Highspeed rail (HSR), CostBenefit Analysis (CBA), demand uncertainty
JEL Classification: L92 – L98
WCTRS Areas: (E) Transport Economics, Finance and Evaluation
(E3) Valuation of Internal and External Benefits and Costs
1
Email: jcampos@daea.ulpgc.es
2
Email: a.hernandezg@becarios.ulpgc.es
2
1. Introduction
The economic evaluation of public investment projects is one of the fields of applied
microeconomics which has gained larger relevance in the current context of budgetary
constraints. Pressures to find the most efficient allocation of resources are increasingly
being translated at every level of government into the need of selecting good projects
and investing public money where its social return could be more rewarding. There
exist several wellestablished techniques for conducting this assessment, but most of
them simply try to compare the benefits that the society as a whole accrues from
different types of public intervention in the markets, with the opportunity cost of the
resources used to finance such interventions.
Transport markets make an especially fruitful area for all sorts of costbenefit
analyses (CBA), since measuring and giving money values to traffic flows, travel time
and other transport costs seldom result as controversial as, for instance, evaluating the
impact of health or environmental policies. The economic evaluation of any public
intervention in transport markets particularly aims at quantifying in monetary terms the
change in social welfare generated by this intervention. To evaluate a transport project
is to identify, assess, aggregate and compare the costs and benefits generated by the
project once discounted over time. These costs and benefits can be solely valued both
from a financial point of view (just considering the revenues and costs generated by the
project), or from a more general, social, point of view, which is the focus followed in this
paper.
In the case of highspeed rail (HSR) infrastructure and services, the main social
benefits are primarily gained from time savings enjoyed by existing users, by users that
switch in from other modes, and from the willingness to pay of the newly generated
demand. Where appropriate, these benefits should also be weighed against those
experienced by users of alternative transport modes, either because congestion (in
roads or airports) is reduced and capacity is freed up, or because lives are saved (in
roads) due to the deviation of road traffic. On the opposite side, the main social costs of
HSR investments correspond to the expensive construction and maintenance costs of
the infrastructure, the operation and maintenance of services provided over it, and
other external costs (landscape effects, noise, pollution, etc.) whose components are
often more difficult to quantify. If the benefits outweigh the costs, investment in high
speed rail is, in principle, socially desirable.
However, even a positive net present value does not preclude the possibility
that there may be alternative projects that could provide a higher social benefit. When
the amount of the investment is large enough – for example in the case of huge
national or supranational investment programmes such as the Spanish PEIT
3
or the
European Structural and Cohesion Funds – the accuracy required when adopting
decisions on particular projects is also larger.
Although this is not a particular requirement for rail investment and our analysis
could be easily extended to other infrastructure areas, the current worldwide expansion
of HSR projects, jointly with the fact that most of these projects are very expensive and
particularly uncertain, sometimes with dubious social returns in a 2030 years horizon,
all make of this sector a very interesting case study.
3
The term PEIT (Plan Estratégico de Infraestructuras de Transporte) stands for Strategic Plan for
Infrastructure and Transport represents the dorsal spine of current Spanish public investment policy. It
compromises projects worth more than €200 billion over the 20052020 period, as explained in Section 2.
3
It is precisely the inevitable uncertainty associated with factors such as demand
predictions or the future technology developments in the rail sector
4
what leads us to
adopt in this paper an unconventional approach into costbenefit analysis. Instead of
conducting a standard assessment of a single highspeed line or corridor, using well
estimated demand and cost data, we will follow and upsidedown approach where
demand uncertainty plays a central role. We will not simply compare (social) benefits
and costs discounted (from an initial estimate of demand figures) to calculate a single
value of a social net present value (SNPV). On the contrary, we will try to figure out
what is the initial demand level that the project should have in order to provide a
positive (expected) social return. Although this methodological approach is not
completely new in the CBA literature, as it has been used, for example, in De Rus and
Nombela (2007) and De Rus and Nash (2008), our contribution adds a larger role for
uncertainty and explicitly shows that the reject/accept decision cannot be made based
upon a deterministic SNPV but on a complete probability distribution of SPNV values.
The main advantage of this methodology, that will be illustrated with data and
examples extracted from the Spanish PEIT and other international references, is that it
internalizes and isolates most of the demand uncertainty in HSR projects, making it
possible to determine, as a decision reference, the „minimum demand threshold from
which a project would be socially profitable‟. If this minimum amount is not met, the
society would gain more investing its resources on alternative projects.
After this introduction, the rest of this paper is structured as follows. Section 2
describes in larger detail the rail investment projects within the PEIT. Using the
characteristics provided by this description, we can then build an „ideal HSR project‟
that comprises most of these characteristics. This ideal is a standard highspeed rail
corridor whose technical and cost data can be easily extracted either from the PEIT or
the existing literature (see Campos et al., 2009). This is carried out in Section 3, where
the project (the construction and operation of a new highspeed line) is defined and the
assessment model is formulated to calculate the value of the initial demand figures that
guarantee a positive expected SNPV. Section 4 is then devoted to the practical
implementation of this model in full detail and result analysis, while in Section 5 the
main findings and lessons are discussed.
2. HSR projects in the Spanish PEIT program
The Strategic Plan for Infrastructure and Transport (PEIT) is an ambitious planning tool
developed by the Spanish Ministry of Public Works and Transport (Ministerio de
Fomento) to encompass all its medium and longterm policy actions within a single
framework, aimed at modernizing the country infrastructure in the 20052020 horizon.
Although the Plan includes actions in many different areas, rail transport concentrates
the most important government interventions, both in quantity (43.7% of total actions),
as in the overall amount of investments to be made in this sector (more than €108
billion). The Plan itself gives the railroad a „leading role‟ and establishes as one of its
specific objectives „(...) to achieve a complete restructuring of the current railway
network‟.
4
The MAGLEV (magnetic levitation) technology is increasingly viewed as the next standard for high speed
surface transportation. Note that, if successful, all current investments on HSR could become obsolete in
1020 years.
4
The reasons that explain this particular interest can be found in the initial
diagnosis made in the PEIT of the poor conditions of the Spanish railways at the
beginning of the century. The network, with a total length close to 13,500 km in 2005
(in a country of about 500,000 sq. km), only had 1,031 km with international (UIC)
width, 2,095 km of electrified double track, 3,628 km of electrified singletrack and still
5,494 km without electricity. This situation had created notable differences in the
performance levels across different lines of the network, had also produced difficulties
in their integration within the European network and led to the existence of corridors
and lines with very low utilization and unlikely survival.
The railway restructuring projects included in the PEIT were supposedly
designed to upgrade the Spanish network in order to progressively transform this mode
into „a central element for the articulation of intermodal transportation‟. The projects
affect more than 9,000 km of the network, which will be improved to support high
performance traffic (that is, HSR for passengers or improved speed for cargo). After the
PEIT the network will cover the entire country thanks to the construction of over 5,600
km of new tracks. Furthermore, in order to improve accessibility, the Plan also aims to
modify the current radial structure of the Spanish transport system (from the city
capital, Madrid to the provinces) by developing peripheral routes and getting all the
provincial capitals connected to the high performance network, so that 90% of the
population is situated within 50 km of a rail station.
These ambitious objectives are to be met in 15 years (20052020), but divided
into two different stages. Actions in the 20052008 period, almost completed at the time
of writing this paper, have focused on completing highspeed corridors that were
already under construction (most notably the MadridBarcelona one), on investment in
the conventional network to improve the operating conditions of rail freight services and
to facilitate exchange with the road and maritime transport, and on consolidating the
new institutional framework. Actions to be taken after 2009 are mainly designed to
improve services, gradually extending the HSR network across the country and
improving the interoperability with the French network. At this moment many of the
projects included at this stage have started, but many others are yet to be defined. In
particular, it is especially relevant for the purposes of this paper the fact that the PEIT
states that „(...) actions from 2013 onwards will be set out in accordance with the
results obtained in the previous period in terms of evolution of the modal share of
railways, that is, in terms of the real demand‟.
The guidelines and broad lines of action included in the PEIT have been
developed through sectorspecific plans, which are revised by various government
agencies (at least every 8 years) to redefine specific actions in each mode of transport.
In the case of rail, the exact investment estimate for the projects included in the 2005
2020 horizon amounts to €108,760 million, although this figure varies across three
different project types (Ministerio de Fomento, 2007):
1. New trunk lines and corridors exclusively for passenger services. These refer
essentially to sections or lines which exceed a given traffic threshold in the year
they come into service, ensuring the greatest possible social returns and territorial
impact, with greater timesaving to the destinations served. These lines clearly
make the future pattern of the rail network rigid, using parameters which do not
allow mixed traffic, so that the conventional network has to be used for goods, with
the risk of a possible underutilization of the two networks and of increased
administrative costs for the infrastructure this involves.
5
2. Lines and sections on routes varying substantially from existing ones, for mixed
traffic (passengers and freight). This second case relates to lines with medium
traffic, the prolongation of trunk lines, structural transversal axes and crossborder
links, designed in principle for mixed traffic, since the potential of the corridors they
serve does not seem at present to allow them to specialize in passengers, with the
resulting duplication of the network. In any event, the final design will depend on
the balance between the increased cost of construction of the mixed system and
the advantages of a concentration of traffic and enhanced occupancy, plus the
addition of a future option to close the conventional line, or to operate the two lines
on a specialized basis. Such actions would come in a second phase of
programming, except where the existing line is being used substantially. In such
cases, and particularly on lines with greater traffic, such as the Mediterranean
corridor or ValladolidBurgosVitoria, among others, the mixed traffic design
provided for in the PEIT has to be understood as applying to the entire ambit of the
corridor and not just of the specific line so that, if the traffic, functionality and
service conditions allow, parallel specialized lines could be created on part or all of
these axes.
3. Terminal sections, where the traffic is significantly less than in the other two cases,
for mixed traffic. This concerns sections or lines usually at the termination of routes,
with less passenger traffic and some goods traffic. The aim is to renew these
sections entirely, to increase speed, safety and service quality, seeking full
interoperability in the medium term with the rest of the European network. The
systematic introduction is therefore proposed of the multilateral sleeper on Iberian
gauge lines, to prepare for the gauge changeover.
As shown in Table 1, this development of the highspeed rail network in Spain
has been carried out through five major radial corridors covering the Iberian Peninsula
and connecting the national network with French and Portuguese ones. Although the
characteristics of each corridor are very different, they all include the three types of
actions described above. In all cases, the most relevant projects involve the extension
of existing HSR lines, but there also are actions designed for the improvement of the
conventional network in order to equip it with high performance features, or even the
construction of new infrastructure in areas where the conventional grid did not exist or
did not meet the adequate technical conditions.
6
Corridor Description Operation starting year [distance]
5
Andalusia Will connect Madrid with the most
relevant Andalusian capitals through
the MadridCórdobaSeville line.
Madrid  Granada (+ 2010) [430 km]
Madrid  Málaga (2007) [540 km]
Madrid  Cadiz (+ 2012) [660 km]
Mediterranean
Axis
Will connect Madrid with the
Mediterranean coast (Valencia and
Alicante), then connecting with a
transversal axis from there to
France.
Madrid  Valencia (2010) [350 km]
Madrid  Alicante (2012) [420 km]
Northeast
Corridor
The MadridSaragossaBarcelona
France is almost completed and will
become the most relevant one in the
country. Connections with provincial
cities are also planned.
Madrid  Barcelona (2008) [620 km]
Barcelona  France (2009) [150 km]
North
Northwest
The MadridSegoviaValladolid axis
will be extended to the north coast
and Galicia.
Madrid  Valladolid (2007) [190 km]
Madrid  Bilbao (+ 2012) [395 km]
Madrid  Galicia (+ 2012) [600 km]
Extremadura 
Portugal
Will connect Madrid with Lisbon. Madrid  Caceres (+ 2012) [300 km]
Cáceres  Badajoz (+2012) [90 km]
Badajoz  Portugal (+2012) [50 km]
Source: PEIT (www.mfom.es)
Table 1. HSR corridors in Spain as designed in the PEIT
Indeed, most of the actions undertaken in the 20052009 period have consisted
on the conclusion of some lines that had been in planning and construction for several
years (MadridBarcelona, MadridValladolid, MadridMalaga), while actions in
subsequent phases of the Plan will materialize – as mentioned above – only if the
minimum demand criterion is met. It is on this type of future action where the rest of the
work done in this paper mainly focuses.
Therefore, in accordance with the above discussion and the information
provided by the PEIT, we can now define a „representative project‟ that could serve as
the unit of analysis for the following sections of this paper. It should include the building
of new HSR infrastructure (either by designing a new line or through a substantial
improvement into a existing conventional one) and its operation within a pointtopoint
corridor (i.e., between two major cities). This line should have an average length of
between 300 and 600 km, a period of planning and construction of about five years,
with a total estimated project life of 40 years. This line would operate in a context of
great uncertainty about demand and costs associated with it, although the latter can be
mitigated in part by using information extracted from other projects, as we immediately
will see.
5
These are estimated dates for the entire corridor. In most cases there are lines that start operations
before the full completion of the corridor, as it happened in the CórdobaAntequera (2006) line or in Lleida
Tarragona (2006). The MadridZaragoza route was inaugurated in 2003, and the pioneering one, Madrid
Seville, in April 1992.
7
3. Project definition and assessment model
The economic appraisal of a rail investment project traditionally consists in comparing
its discounted benefits and costs, by using – for example – the social net present value
(SNPV) as the decision criterion. If there is uncertainty about the variables or
parameters of the model, a sensitivity analysis is usually carried out ex post, showing
different results under different scenarios. In this paper we will follow an inverse
procedure: since the critical (uncertainty) variable is the demand, our objective will be
to determine the threshold value of the initial demand that equals to zero the
(expected) SNPV. Thus, our aim is to calculate „the minimum level of traffic in the year
of entry into service the new line that would guarantee a positive social return‟. Firstly,
we will begin with the technical definition of the project considering some reasonable
assumptions about it in order to facilitate its assessment and secondly we will illustrate,
step by step, the calculations to be done to produce the expected result.
3.1. Project definition
According to Section 2, our representative project consists in the construction and
operation of a highspeed line of L km long. As shown in Figure 1, the society makes a
gross investment equal to GI
t
euros, which includes both the costs of planning and
construction, during the construction period (between t = 0 and t = T ‟). During each of
the years of operation of the line (between t =T’+1 and t=T), the project contributes
annually to the society a change in the social net benefit (with respect to the situation
without project) equal to ASB
t
, which may be positive or negative. All these flows are
valued at 31th December each year.
Figure 1. Project structure
Once the flows associated to the project are discounted with the interest rate i,
the formal expression of the social net present value at t = 0 is traditionally given by:
0
' 1
(1 )
T
t
t
t T
SB
SNPV I
i
= +
A
= ÷ +
+
¿
, (1)
where I
0
summarizes, valued at t = 0, the initial investment net of its residual value
(RVI
T
):
6
6
Although it is not the most appropriate approach, we have considered that the residual value is just a
percentage of the initial investment, in order to simplify the model.
8
'
0
0
(1 ) (1 )
T
t T
t T
t
GI RVI
I
i i
=
= ÷
+ +
¿
(2)
On the other hand, the annual change in the net social benefit (ASB
t
) can be
decomposed into changes in consumers‟ surplus (ACS
t
), changes in operators‟ (or
producer‟s) surplus (APS
t
), changes in taxpayers‟ surplus (AGS
t
) and changes in the
indirect effects (AIE
t
). However, to simplify our analysis we will only consider that the
first two effects are relevant in our case, i.e. AGS
t
= AIE
t
= 0, so that our society only
includes transport users and transport producers:
ASB
t
= ACS
t
+ APS
t
(3)
3.2. An initial benchmark: the analysis of a static model
To illustrate the calculations associated to the assessment model, we will proceed step
by step, from a simple (static) framework to a dynamic one. Thus, let first assume that
the project has a total length of one period: all investment takes place at t = 0 and the
social costs and benefits associated to the operating period are valued at t = 1, i.e. (T’,
T) = (0, 1). Under these conditions, it is immediate to verify that expression (1)
becomes:
1 1
0
(1 )
CS PS
SNPV I
i
A + A
= ÷ +
+
(4)
3.2.1. Computing the (change in) consumers’ surplus
7
All the traffic in the new high speed line in period t = 1 comes from two possible
sources of demand: the diverted traffic from existing modes, which will be generically
denoted as: airplane (a), bus (b), car (c), or conventional rail (r),
8
or the newly
generated traffic, by users who did not previously on this route.
Diverted traffic
In the initial situation without the project (denoted by superscript 0) the users‟
generalized cost of travel for each of the existing modes (m = a, b, c, r) is defined as:
0 0 0 0 0
m m m m m
g p v = + t +c (5)
where:

0
m
p = monetary component of the cost in mode m (without project),

0
m
v = value of time in mode m (without project). Assuming that this value does not
change with the project, we will simply denote it as
m
v ,

0
m
t = total travel time in mode m (without project), and

0
m
c = monetary valuation of the travel quality in mode m (without project).
7
From now on, to simplify the notation, subscript 1 is dropped.
8
Conventional rail can be omitted when the project consists on the construction of a new line, where this
mode did not previously exist.
9
In the situation with project, after the investment (denoted by superscript I), the
change in the consumers‟ surplus of users diverted from other existing modes can be
expressed as:
9
0 0
( )
I
m m m
m
g g q ÷
¿
, (6)
where:

0
m
q = HSR trips made by travellers from other m modes.

I I I I
m m m
g p v = + t +c
= generalized cost of travel at HSR for travellers form other
modes.
 p
I
= average fare of a HSR trip, t
I
is the average duration of that trip, and ε
I
m
represents a monetary valuation of the HSR service quality, which is assumed
close to zero.
10
Generated traffic
The surplus associated with new trips (made by both old and new users) must be
added to the previous expression. It is conventionally determined by:
0 0
1
( )( )
2
I I
m m m
m
g g q q ÷ ÷
¿
, (7)
where
I
q
are total highspeed trips per year. These trips consist of the sum of all
diverted demand from other modes m,
0 0
m
m
Q q =
¿
,
and the generated traffic, that we will simply estimate as a proportion (o) of the diverted
traffic,
q
I
÷Q
0
= oQ
0
= o q
m
0
m
¿
,
so that:
0
(1 )
I
q Q = +o . (8)
3.2.2. Computing the (change in) the operators’ surplus
The change in producers‟ surplus is determined by the difference between the
operators‟ income with and without project, and the operation and maintenance costs
of the HSR line, plus avoidable costs associated with the reduction of service in each
mode m (because of the traffic that is deviated to the high speed line):
0 0 0
( )
I I I
m m m
m
p q p q C C ÷ + ÷
¿
, (9)
9
This expression, conventionally denominated „rule of one half‟ in CBA, implicitly considers that the
change in the social welfare of each traveller can be identified with the change that is produced in the last
traveller.
10
Alternatively, this valuation could be also incorporated to the value of travel time (v).
10
where
0
m
C are the avoidable costs in mode m and
I
C
the operating and maintenance
costs of the highspeed infrastructure.
In general, avoidable costs depend on the characteristics of each mode, the
level of competition and the volume of diverted traffic. It is therefore possible to express
them as:
0 0
m m m
C c q = (10)
where c
m
is the unit cost (per trip) saved in mode m, which we assume to remain
unchanged after the project (compared to the situation without the project). Moreover,
as an important simplification, we also assume that all operators in alternative transport
modes obtain normal profits and all their costs are recoverable, allowing us to omit the
calculation of surpluses in other modes different from m.
Therefore, operating and maintenance costs of a highspeed line could be
expressed as:
C
I
= I
M
+c
A
q
I
+(c
O
+c
M
)N, (11)
where:
 I
M
= infrastructure maintenance costs (usually fixed in terms of total traffic),
 c
A
= unit cost (per traveller) of acquisition of the rolling stock, net of its residual
value,
11
 c
O
= unit operation cost (per train) of the rolling stock,
 c
M
= unit maintenance cost (per train) of the rolling stock, and
 N = number of trains needed to satisfy the demand each year.
The number of trains in operation (N) is a crucial variable in order to determine
the supply of transport services, since it changes as the demand does. To compute N,
the capacity of trains (assumed to be constant) and the daily frequency of services
(assumed to be homogeneous throughout the year) are required. For example, if
e
q q =ì is the effective average occupation of each train (where ì is the load factor
and q is the number of seats) and H are the hours of operation per day,
12
then the
number of daily services during a year in each direction is obtained from the ratio
( / 365) /
I
e
q q , so that the frequency (F) is:
(( / 365) / )
I
e
q q
F
H
= .
Thus, the number of trains needed to serve the annual demand of travellers q
I
is given
by
I
F t , that is, using expression (8):
N[q
I
] =
t
I
365Hq
e
q
I
=
t
I
365Hq
e
(1+ o)Q
0
= N[Q
0
] . (12)
11
Although this cost should be formulated as a cost per train (equivalent to the acquisition price of each
train), the difficulty to obtain real values of the acquisition price (which are directly negotiated with the
producer and usually includes part of the maintenance costs) suggests the possibility of calculating
through the cost per passenger what simplifies the calculus and could be interpreted as a sort of leasing.
12
In this case, we consider q = 350, ì = 80% and H = 18. See Table 4 below.
11
3.2.3. Computing the (change in) social surplus
According to (3), the change in the social benefit is determined by the sum of changes
in the surpluses of consumers and producers, i.e. the sum of (6), (7) and (9):
(g
m
0
÷ g
m
I
)
m
¿
q
m
0
+
1
2
(g
m
0
÷ g
m
I
)(q
I
÷ q
m
0
)
m
¿
+ ( p
I
q
I
÷ p
m
0
q
m
0
+ C
m
0
÷ C
I
)
m
¿
(13)
Expanding the first term of the previous expression yields:
0 0 0
( ) ( )
I I
m m m m
m
p p v q ( ÷ + t ÷t
¸ ¸
¿
,
where it is assumed, for the sake of simplicity, that the project does not significantly
alter the monetary value of the quality associated to travel on different modes, i.e.
0 1
( ) 0
m m
m
c ÷c =
¿
.
Substituting the previous result in expression (13) and simplifying the outcome,
the change in the net social benefit can be rewritten as:
v
m
(t
m
0
÷ t
I
)
m
¿
q
m
0
+
1
2
(g
m
0
÷ g
m
I
) + p
I
¸
(
¸
(
(q
I
÷ q
m
0
)
m
¿
+ C
m
0
÷ C
I
m
¿
, (14)
showing that this change has got three elements: time savings in the diverted traffic,
the value of the new generated traffic and resource savings in the modes that lose
traffic.
Finally, given that
(q
I
÷ q
m
0
)
m
¿
= o q
m
0
m
¿
and developing the intermediate
term in (14), the previous expression leads to:
v
m
(t
m
0
÷ t
I
)
m
¿
q
m
0
+o
1
2
( p
m
0
÷ p
I
) + p
I
¸
(
¸
(
q
m
0
+
o
2
v
m
(t
m
0
÷ t
I
)q
m
0
m
¿
m
¿
+ C
m
0
÷C
I
m
¿
,
where, after incorporating the expressions of costs (10) and (11), together with (12), we
finally obtain as a benchmark:
ASB = ACS + APS =
= 1+
o
2

\

.

v
m
(t
m
0
÷ t
I
)q
m
0
+
m
¿
+
o
2

\

.

( p
m
0
+ p
I
)q
m
0
+
m
¿
+ c
m
q
m
0
m
¿
÷
÷ I
M
÷ c
A
+
(c
O
+ c
M
)t
I
365Hq
e

\

.

(1+ o) q
m
0
m
¿
(15)
12
3.2.4. Finding the minimum demand threshold: SNPV=0
The last step of the calculations is to solve the equation SPNV = 0 using expressions
(4) and (15). For a better understanding of the procedure, we first consider a simpler
case (where the diverted traffic comes only from a mode) and then we later generalize
this result. Thus, let us consider that there is only an alternative mode (e.g., the
airplane, m = a). In this case, expression (15) simplifies to:
ASB = (1+
o
2
)v
a
(t
a
0
÷ t
I
)q
a
0
+
o
2
( p
a
0
+ p
I
)q
a
0
+
+ c
a
q
a
0
÷ I
M
÷ c
A
+
(c
O
+ c
M
)t
I
365Hq
e

\

.

(1+ o)q
a
0
According to (4), SNPV
= 0 requires that
I
0
(1+ i) = ASB. Therefore, isolating q
a
0
from
the previous expression, we easily obtain:
q
a
0
=
I
0
(1+ i) + I
M
(1+
o
2
)v
a
(t
a
0
÷ t
I
) +
o
2
( p
a
0
+ p
I
) + c
a
÷ (1+ o) c
A
+
(c
O
+ c
M
)t
I
365Hq
e

\

.

, (16)
where this value of
0
a
q
determines the minimum value of the highspeed diverted traffic
from the airplane that makes this project socially profitable. Note that, from
0
a
q we can
immediately recover the volume of generated traffic (using the o parameter) and the
total demand.
In the general case, with m alternative transport modes, the problem is to
determine the value of
0
m
m
q
¿
in (15) so that
I
0
(1+ i) = ASB. To get it, it is necessary
to introduce some additional assumptions about the relation between different values of
the diverted traffic,
0
m
q , so the m values are reduced into just one.
13
Among the different alternatives we might consider, for example that the
diverted traffic can be expressed in relation to one of the modes. So, taking again the
airplane (a) as the reference mode, we would have:
Q
0
= q
a
0
+q
b
0
+q
c
0
+q
f
0
= q
a
0
+
b
q
a
0
+
c
q
a
0
+
f
q
a
0
=(1+
b
+
c
+
f
)q
a
0
(17)
where the  parameters (that reflect the relation between different diverted traffics) can
be determined endogenously in terms of the relative market share of each mode in
relation to the highspeed rail mode. Using this procedure, the change in social welfare
can be finally obtained as:
13
Alternatively, the problem would be unsolvable, since the expression SNPV = 0 would have infinite
solutions.
13
ASB =
= 1+
o
2

\

.

v
a
(t
a
0
÷ t
I
) + v
b
(t
b
0
÷ t
I
)
b
+ v
c
(t
c
0
÷ t
I
)
c
+ v
f
(t
f
0
÷ t
I
)
f
{ }
q
a
0
+
+
o
2

\

.

( p
a
0
+ p
I
) + ( p
b
0
+ p
I
)
b
+ ( p
c
0
+ p
I
)
c
+ ( p
f
0
+ p
I
)
f
{ }
q
a
0
+
+ c
a
+ c
b

b
+ c
c

c
+ c
f

f
{ }
q
a
0
÷
÷ I
M
÷ (1+ o) c
A
+
(c
O
+ c
M
)t
I
365Hq
e

\

.

(1+ 
b
+ 
c
+ 
f
)q
a
0
and solving again
I
0
(1+ i) = ASB we get:
{ }
0 0 0 0
0 0 0 0 0
0
( ) ( ) ( ) ( )
(1 ) ( ) ( ) ( ) ( )
( )
(1 )( (1 )
365
(1 )
2
2
I I I I
a a b b b c c c f f f
M I I I I
a a b b c c f f
O M I
A
a b b c c f f b c f
e
v v v v
q I i I p p p p p p p p
c c
c c c c c
Hq
¦ ¹
( t ÷t + t ÷t  + t ÷t  + t ÷t 
¦ ¦
¸ ¸
¦ ¦
¦ ¦
( = + + + + + +  + +  + +  +
´ `
¸ ¸
¦ ¦
¦ ¦ + t
( +  +  +  ÷ +o + + + +
¦ ¦
¸ ¸
¹ )
o
+
o
whose interpretation is similar to that of expression (16).
3.3. The general model of T periods
After this simplified introduction, we will now generalise the procedure assuming that
the change social benefit in (1) is distributed over T years. To repeat the same steps,
we first introduce a time subscript t from expression (15) so that the change in social
welfare in year t is given by:
ASB
t
=
= 1+
o
2

\

.

v
m
(t
m
0
÷ t
I
)q
mt
0
+
m
¿
+
o
2

\

.

( p
m
0
+ p
I
)q
mt
0
+
m
¿
+ c
m
q
mt
0
m
¿
÷
÷ I
M
÷ c
A
+
(c
O
+ c
M
)t
I
365Hq
e

\

.

(1+ o) q
mt
0
m
¿
14
The main simplifying assumption underlying this expression is that all the
parameters of the model remain constant over time with the exception of traffic, which
is assumed to grow at a constant rate (u). Thus,
0
mt
q represents diverted traffic from
mode m towards HSR in year t,
0 0
1
(1 )
mt mt
q q
+
= +u . (18)
Therefore, it is immediate to obtain the following:
ASB
t
(1+ i)
t
t =T ' +1
T
¿
=
= 1+
o
2

\

.

v
m
(t
m
0
÷ t
I
)
q
mt
0
(1+ i)
t
t =T ' +1
T
¿
m
¿
¸
(
¸
(
(
+
+
o
2

\

.

( p
m
0
+ p
I
)
q
mt
0
(1+ i)
t
t =T ' +1
T
¿
m
¿
¸
(
¸
(
(
+
+ c
m
q
mt
0
(1+ i)
t
t =T ' +1
T
¿
m
¿
÷
÷
I
M
(1+ i)
t
t =T ' +1
T
¿
÷
÷(1+ o) c
A
+
(c
O
+ c
M
)t
I
365Hq
e

\

.

q
mt
0
(1+ i)
t
t =T ' +1
T
¿
m
¿
(19)
where
q
mt
0
(1+ i)
t
t =T ' +1
T
¿
=
q
mT ' +1
0
(1+ i)
T ' +1
1+
1+ u
1+ i

\

.

+
1+ u
1+ i

\

.

2
+ ... +
1+ u
1+ i

\

.

T ÷( ' T +1)
¸
(
¸
(
(
Note that this final expression is just a sum of terms of a geometric progression with
common ratio (1+u)/(1+i) so that we can rewrite it as:
' 1
0 0
' 1
' 1
' 1
1
1
1
1 (1 ) (1 )
1
1
T T
T
mt mT
t T
t T
q q
i
i i
i
÷ ÷
+
+
= +
(
+ u  
÷
(

+
\ .
(
=
+ u ( + +  
÷

(
+
\ .
¸ ¸
¿
' 1 ' 1
0 0
' 1 ' 1 1
(1 ) (1 )
( )(1 )
T T T T
mT mT T
i
q q
i i
÷ ÷ ÷ ÷
+ + ÷
  +u ÷ +
= = ¸

u÷ +
\ .
and then replace it in (19).
15
Finally, from (1), we know that
I
0
=
ASB
t
(1+ i)
t
t =T '+1
T
¿
must hold to guarantee that
SNPV =0, i.e.:
I
0
=
ASB
t
(1+ i)
t
t =T ' +1
T
¿
=
= 1+
o
2

\

.

v
a
(t
a
0
÷ t
I
) + v
b
(t
b
0
÷ t
I
)
b
+ v
c
(t
c
0
÷ t
I
)
c
+ v
f
(t
f
0
÷ t
I
)
f
¸
(
¸
¸q
aT ' +1
0
+
+
o
2

\

.

( p
a
0
+ p
I
) + ( p
b
0
+ p
I
)
b
+ ( p
c
0
+ p
I
)
c
+ ( p
f
0
+ p
I
)
f
¸
(
¸
¸q
aT '+1
0
+
+ c
a
+ c
b

b
+ c
c

c
+ c
f

f
¸
(
¸
¸q
aT '+1
0
÷
÷
I
M
(1+ i)
t
t =T ' +1
T
¿
÷
÷ (1+ o) c
A
+
(c
O
+ c
M
)t
I
365Hq
e

\

.

(1+ 
b
+ 
c
+ 
f
)¸q
aT '+1
0
Thus, solving for
0
' 1 aT
q
+
, the value of the initial (first period of operation) demand
(diverted from the airplane) that equals the social NPV to zero is given by:
0 0 0 0
0 0 0 0 0
' 1 0
' 1
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
(1 )
( )
(1 ) (1
365
(1 )
2
2
I I I I
a a b b b c c c f f f
M T
I I I I
aT a b b c c f f t
t T
O M I
A
a b b c c f f b
e
v v v v
I
q I p p p p p p p p
i
c c
c c c c c
Hq
+
= +
( t ÷t + t ÷t  + t ÷t  + t ÷t  ¸
¸ ¸
¦ ¹
( = + + + + +  + +  + +  ¸ +
´ `
¸ ¸
+
¹ )
  + t
( +  +  +  ¸ ÷ +o + +

¸ ¸
\ .
o
+
o
¿
)
c f
¦ ¹
¦ ¦
¦ ¦
¦ ¦
´ `
¦ ¦
¦ ¦
+ + ¸
¦ ¦
¹ )
As discussed below, once this initial value is found, the calculation of total
diverted demand, generated demand and all other variables of the model is immediate.
4. Model implementation and results analysis
Implementing the expressions just developed in the previous section and the formal
calculation of the minimum demand thresholds that guarantee the social profitability of
this HSR investment project requires an exogenous definition of the set of parameters
that complete the model. These are mainly technical parameters related to producers‟
and consumers‟ costs and the specific characteristics of each project. Although the
16
main point of this paper is a methodological one, the relevance of our simulation results
will be highly conditioned by the plausibility of the reference values that we use to
perform our simulations. Most of these come from other projects (see Campos and de
Rus, 2009) and the authors‟ experience in analysing HSR markets (Campos, 2009).
4.1. Parameters
4.1.1. Modelling the producers’ cost
In accordance to previous sections, our reference project is a highspeed line between
two cities (nonstop) with a total length of 500 kilometres (L), whose life is estimated in
40 years (T), out of which the first five ones (T’) correspond to the planning and
construction period. We assume that the same number of kilometres is built every year,
so that the annual gross investment (GI
t
) between t = 0 and t = T‟ (valued at the end of
each year) is given by:
'
t
L
GI ckm
T
= · ,
where ckm is the construction cost per kilometre in euros. The value of this unit cost of
construction varies across projects, because it is directly affected by terrain
circumstances. Campos et al. (2008) use a database of over 160 highspeed rail
projects around the world and obtain average values for the construction costs per
country. In the case of Spain, these values range between 9 and 17.5 millions of euros
per kilometre if costs of planning are included.
With respect to the residual value of the investment in infrastructure (RVI
T
), we
can simply consider that this is adequately represented by a proportion (o =10%, for
example) of the gross investment. Thus, the expression (2) for the initial investment
would be:
' ' '
0
0 0 0
1
(1 ) (1 ) (1 ) (1 ) (1 )
T T T
t t t
t T t T t
t t t
GI GI GI
I
i i i i i
= = =
     o o
= ÷ = ÷
  
+ + + + +
\ . \ .\ .
¿ ¿ ¿
.
Once the infrastructure has been built, the provision of highspeed rail services
involves two types of costs: those related to the maintenance of the infrastructure itself
and those associated to the provision of the services (i.e., acquisition, operation and
maintenance costs of rolling stock).
According to (11), annual maintenance costs of infrastructure are denoted by I
M
which are considered independent of the traffic volume, because they only depend on
the line length,
M I
I cmk L = · , where cmk
I
is the maintenance cost per kilometre.
According to Campos et al. (2008), a reasonable value of this parameter for Spain
would be around 35,000 euros per kilometre and year, making the annual cost for a
line of 500 kilometres line around 17.5 million euros.
In the case of the acquisition costs of the rolling stock (c
A
) raised in expression
(11), it must be taken into account – as discussed in footnote 11 – that they are being
considered the unit cost per traveller and not by train; therefore, it is not easy to obtain
appropriate references on its value. In France, according to UIC data (UIC, 2005),
17
there is reliable information about total acquisition values for different models of train
(TGV Réseau, TGV Dúplex), these values are around 50,000 euros per seat, what
implies that a train of 350 seats costs a minimum of 17 to 18 millions euros. However,
factors such as the inside design, technical specifications and other characteristics may
change the price dramatically. Thus, and although the range of variation can be high,
an initial estimation of the unit cost c
A
, can be between 25 and 40 euros per passenger
and year for an intensive use of between 300,000 and 500,000 passenger per train and
year.
The estimation of unit costs for operation and maintenance of rolling stock is
less difficult due to the existence of better information. Thus, Campos et al. (2008)
shows that the average operative cost per seat is around 53,000 euros, with little
dispersion between countries, so that the cost per a train of 350 seats (c
O
) would be
18.55 million euros per year.
14
The equivalent figure for the maintenance costs (c
I
)
would be 1 million euros per train and year, assuming an average use intensity of
500,000 kilometres per year.
4.1.2. Modelling the users’ cost
As discussed above, the users‟ generalized cost (g) includes the price of each trip (p),
total travel time (t) valued by each passenger and a set of additional factors (c) globally
related to the „quality‟ of each mode, which will not be considered in this case. In
principle, since our analysis is based on a hypothetical PEIT corridor it is difficult to
estimate a value for the generalized cost; however, using as a reference information
from MadridBarcelona corridor (620 km) and adjusting it to 500 km line, Table 2 shows
reasonable estimates for p
m
and p
I
(HSR), taking into account differences in vehicle
types.
15
Mode Vehicle
(Company)
Price (620 km) Estimated price (500
km)
Airplane (a) Boeing 757
(IBERIA)
30 – 160 € 24 – 130 €
Bus (b) Eurobus (ALSA) 25 – 45 € 20 – 36 €
Car (c) Mediumsized car 85 € (*) 68.5 €
Conventional rail
(r)
Alvia Train
(RENFE)
60 – 100 € 48 – 80.5 €
High speed rail AVE Train
(RENFE)
70 – 150 € 56 – 121 €
Source: AENA (2006) and own elaboration. (*) Includes tolls and fuel consumption.
Table 2. Price estimates per mode (one way trip, 2006)
14
This value primarily includes labour costs and energy, as well as the proportion of the overall cost of
administration and service management.
15
Recall that, according to the avoidable costs assumption in (10) (i.e., all operators in alternative modes
have normal profits in the corridor and all their costs are recoverable), we consider that pm = cm, i.e. a
100% degree of cost coverage in all the alternative modes.
18
With respect to total travel time – including time spent in the vehicle, access
time and waiting time – it can be estimated using a similar approach, although in this
case we have in Table 3, more detailed information on several Spanish corridors and
their exact travel time for different modes. This allows us to calculate in the last two
rows, the average speed of each mode and extrapolate it to a line of 500 kilometres.
16
Line [length] AVE Plane Road*
Madrid  Granada [430 km]
Madrid  Malaga [540 km]
Madrid  Cadiz [660 km] (Jerez)
Madrid  Valencia [350 km]
Madrid  Alicante [420 km]
Madrid  Barcelona [620 km]
Madrid  Bilbao [395 km]
Madrid  A Coruña [600 km]
Madrid  Badajoz [390 km]
3:25
2:40
3:30
1:40
2:00
2:45
2:25
3:55
3:00
2:37
2:24
2:24
2:14
2:22
2:30
2:22
2:29
2:20
4:18
5:24
6:36
3:30
4:12
6:12
3:57
6:00
3:54
Average Speed (km/h)
Estimated time for 500 km
180 km /h
2.77
hours
170 km /h
2.94
hours
100 km /h
5 hours
Source: AENA (2006).
* In road transport (car) is assumed a constant speed of 100 km/h.
** An hour and a half is added to bus (6.5 hours).
*** Total time in AVE is multiplied by 1.5 times for conventional rail.
Table 3. Total travel time by mode: different corridors
Once the travel time for each mode has been estimated it is finally necessary to
define the value of time (v
m
) that will be used in the valuation of time savings from
diverted traffic to HSR. Although the value of time is one of the critical parameters in
the appraisal of any transport project, it cannot be obtained straightforwardly, because
it depends on specific factors such as type of user, travel purpose (work, leisure,…),
frequency, period of the day, year, etc. For this reason, and in order to simplify our
analysis we initially consider that v
m
is an average value, common to all transport
modes and users and it has an (initial) constant value of €15 per hour. This
simplification is removed later in order to introduce uncertainty in our model.
16
Although it is generally assumed that HSR services run at speeds exceeding 250 km/h, the average
commercial speed is much lower because of intermediate stops and traffic restrictions. In the case of the
airplane, we must include access time to the airport and hence, the average speed is further reduced.
19
4.1.3. Demand parameters
The final set of parameters that are needed to solve model are those related to the
demand characteristics. We must first determine what percentages of diverted traffic to
HSR come from each alternative mode, i.e., the parameters in expression (17). For
example in the MadridSeville corridor, the introduction of the highspeed rail in 1992
reduced the demand of air transport by 3040%, decreasing the load factor and flight
frequencies. The conventional rail transport was also affected by the introduction of
new services (MadridSeville, MadridMalaga and MadridCordoba), that were among
the rail operator (RENFE) main lines before 1990. Today, these lines have lost most of
its traffic, while the effects on the road transport have been less significant. Globally
speaking, and taking a conservative approach using the data from other existing high
speed lines (MadridSaragossa or MadridBarcelona), it could be considered that
diverted traffic ratios percentages may be safely given by 75% (from conventional rail
to HSR), 20% (from airplane), 4% (from car users) and 1% (from bus passengers).
Using these shares (s
m
), and selecting the airplane as the reference mode, 
m
parameters are determined by the ratio:
m
m
a
s
s
 = .
With respect to the generated traffic, this is estimated – according to (8) – as a
proportion of diverted traffic. In particular, we consider a reasonable o=0.10. Finally,
the annual growth rate of demand (u) is assumed constant throughout the life of the
project, and equal to 3%. The following table summarizes the values assigned to the
main parameters of the model.
20
Parameter Notation Value Unit of measure
Line length
Last year of the construction of the project
Last year of the project
L
T’
T
500
5
40
Kilometres
Year
Year
Average capacity of each train
Load factor
Hours of operation
q
ì
H
350
80%
18
Seats
%
Hours
Unit construction cost per kilometre
Proportion of the residual value of the infrastructure
Unit maintenance cost of the infrastructure
Unit acquisition cost of the acquisition of rolling stock (per passenger)
Unit operative cost of rolling stock (per train)
Unit maintenance cost of rolling stock (per train)
ckm
o
ckm
I
c
A
c
O
c
I
20,000,000
10
35,000
30
18,500,000
1,000,000
€ /km
%
€ / km
€ / passenger
€ / train
€ / train
Average price of a high speed trip (one way)
Average price of a plane trip (one way)
Average price of a bus trip (one way)
Average price (operative cost) of a car trip (one way)
Average price of a conventional train trip (one way)
p
I
p
a
p
b
p
c
p
f
88.5
77
28
68.5
64.25
€
€
€
€
€
Avoidable average costs (plane)
Avoidable average costs (bus)
Avoidable average costs (car)
Avoidable average costs (conventional train)
c
a
c
b
c
c
c
f
77
28
68.5
64.25
€
€
€
€
Table 4. Summary of the values of the main parameters of the model
21
Parameter Notation Value Unit of measure
Total travel time on high speed rail
Total travel time by plane
Total travel time by bus
Total travel time by car
Total travel time on conventional rail
t
I
t
a
t
b
t
c
t
f
2.77
2.94
6.5
5.0
4.15
Hours
Hours
Hours
Hours
Hours
Value of time (high speed rail)
Value of time (plane)
Value of time (bus)
Value of time (car)
Value of time (conventional train)
v
I
v
a
v
b
v
c
v
f
15
15
15
15
15
€ / hour
€ / hour
€ / hour
€ / hour
€ / hour
Diverted traffic from plane
Diverted traffic from bus
Diverted traffic from car
Diverted traffic from conventional rail
s
a
s
b
s
c
s
f
20
1
4
75
%
%
%
%
Ratio of generated traffic with respect to diverted traffic o 10 %
Annual rate of growth of demand u 3 %
Table 4. Summary of the values of the main parameters of the model (cont.)
22
4.2. Evaluation results
4.2.1. Calculating the initial demand thresholds
From the values of the parameters in Table 4 it is straightforward to compute the
expressions derived in Section 3 using a standard worksheet. In particular, with the
proposed values and a discount rate of 5% we get the following results with respect to
the initial demand (when it starts the operation period t = 6) that at least guarantees a
SNPV equal to zero.
Diverted traffic: 16,702,817 passengers
 from the airplane: 3,340,563 passengers
 from the bus: 167,028 passengers
 from the car: 668,113 passengers
 from the conventional train: 12,527,112 passengers
Generated traffic: 1,670,282 passengers
Total required demand: 18,373,098 passengers
Table 5. First year demand requirement for a socially profitable HSR line
These figures show that if the total annual demand in the first year is less than
18.3 millions of passengers, the project will not be socially profitable. It is obvious that
this is just an approximate figure conditioned by the assumptions about the parameters
of the model. However, from a methodological point of view our analysis still remains
useful, particularly in the context of the PEIT, because it provides a simple procedure to
classify and rank different project alternatives in the 20092020 horizon, according to
the expected demand.
Results in Table 5 are changed if the reference parameters do so. For example,
if we modify the assumptions with respect to the origin of the diverted traffic –
assuming, for example, that HSR captures most of diverted traffic from air transport
(95%) and road transport shares remain the same (car, 4%; bus, 1%),
17
the total
minimum demand threshold would increase to 21.7 millions of passengers, out of
which 18.7 million would come from the airplane. This increase is due to the reduction
of the time savings: given the existence of less time savings in the airplane than in the
conventional rail it would be necessary to attract a larger number of travellers to offset
the social costs of the project.
Similarly, it is possible to make different types of sensitivity analysis on the
model parameters, some of which are summarized in the following table.
17
This would be the case of a highspeed line built on a new route where conventional rail did not
previously exist.
23
Diverted
traffic
Generated
traffic
Total
demand
Initial situation 16,702,817 1,670,282 18,373,098
Increase of train capacity ( q = 450 seats) 13,284,484 1,328,448 14,612,932
Reduction of the load factor (ì = 60%) 27,202,245 2,720,225 29,922,470
Increase of the construction (ckm = 30 mill.) 24,840,191 2,484,019 27,324,210
Reduction of the construction cost (ckm = 15
mill.)
12,634,130 1,263,413 13,897,543
Reduction of the acquisition cost of trains (c
A
=
15 €/ pas.)
10,494,766 1,049,477 11,544,243
Increase in the value of time (v = 30 €/h) 9,971,601 997,160 10,968,761
Reduction of the demand rate of growth (u=
1%)
21,873,643 2,187,364 24,061,007
Table 6. Sensitivity analysis of the model results
4.2.2. A step further: the model under uncertainty
Although the sensitivity analysis procedure is fairly standard, the usefulness of the
results in Table 6 is limited because they are built upon a ceteris paribus approach.
Only one parameter is changed in each case whereas the rest remain unchanged.
What happens when the project is simultaneously affected by several sources of
uncertainty? In this case, an alternative approach consists in identifying the parameters
and variables in Table 4 whose exact values are unknown by the analyst and model
them through explicit probability distributions (normal, uniform, triangular, etc.)
depending on the known range of values. Under this approach, the expressions of the
SNPV used in Section 3 should be now viewed as the expected value of the SNPV.
The initial demand threshold is also a random variable and, therefore, the results in
Tables 5 and 6 can be given in terms of probability distributions.
Just to illustrate these ideas, consider for example the case of the value of time
(VoT). As noted above this is a critical variable in the project appraisal literature.
However, in most cases its specific value is very difficult to determine, either because
of lack of information about the current and future users (a detailed survey could be
carried out, but this is very expensive and often unpractical) or because the
transposition of values from other (previous) projects is always imperfect. In any case,
we can now consider that, instead of setting a deterministic value (€15, according to
Table 4), the VoT is distributed between €5 and €30 according to a triangular
probability distribution. After this, and using a suitable worksheet, the calculations in
our model can be repeated several times to obtain the probability distribution of the
initial demand.
Figure 1 shows this probability distribution, obtained after 5,000 iterations in the
computations of the previous model using @RISK software. As it can be seen, the
average value approximates to the one obtained in the first row of Table 5 although it
does not exactly coincide because this figure is the result of a simulation process. The
interpretation of the figure is immediate: to guarantee a SNPV > 0 our project should
attract a minimum number of passengers during the first year of between 12.5 and 25.9
24
million. This result is valid with a probability of 89.5%. Interestingly, note that there are
no limits to introduce additional sources of uncertainty in the model and even that
correlations between those sources could be properly taken into account.
Figure 1. Probability distribution of the initial demand (total)
From a methodological point of view, the utility of this approach is selfevident:
the decision of accepting or rejecting a particular investment project is carried out
taking into account the demand uncertainty and other sources of uncertainty
simultaneously, involving a margin of error into the decision that can be set ex–ante by
the analyst. Ultimately, it is a more complex approach to CBA, but also more realistic.
5. Conclusion
Highspeed rail is widely reckoned as one of the most significant technological
breakthroughs in passenger transportation nowadays, but building, maintaining and
operating HSR lines is very expensive and their potential benefits are always subject to
great deal of uncertainty, particularly on the demand side. In this context, the aim of
this paper has been to extend received CBA methodology in order to develop some
new procedures to detect the circumstances under which such proposals might be
socially worthwhile.
Our approach departs from traditional CBA but changes its perspective: since
demand uncertainty is the critical variable to accept or reject most rail projects, we
transform the traditional SNPV > 0 acceptance criterion into a „demanddriven criteria‟
and thus obtain different values for this critical demand under alternative simulated
scenarios. What is wanted, in the words of the Spanish Strategic Infrastructures and
Transport Services Plan (PEIT) is to „select (or rank) projects according to their real
demand, finding the minimum traffic threshold (in the year of entry into service) which
25
guarantees a positive social profitability‟. To achieve this goal, and in order to also
provide some realistic examples, we first analyzed the main characteristics of the rail
corridors proposed in the rail sector plan within the PEIT. We then defined a reference
hypothetical project that involved the construction and operation of a highspeed (new
route or improvement of an existing conventional line). From here, and expanding in full
detail the traditional the expression of social NPV we derived an assessment model
that allows us to obtain that minimum demand threshold.
Under the appropriate simplifying assumptions, this model can be easily
implemented (and suitably changed, if needed) according to the analyst‟s needs. It is
basically a methodological tool to provide a quick reference regarding the size of the
corridors where HSR projects are socially viable or not. Our results show, just as an
example, that under reasonable assumptions extracted from previous experiences the
minimum demand necessary to ensure a positive social return can be placed around
18.3 million passengers (in the first year), a figure that is substantially higher than any
of the projects currently discussed in Spain. However, the relevance of this exercise
does not lie in this value, but into the methodology itself. A particularly important
additional element that is also worth noticing here is the possibility of introducing
several sources of uncertainty in the estimation of demand thresholds, which is a
novelty with respect to previous literature.
References
AENA (2006): Evaluación del impacto en las previsiones de tráfico de los aeropuertos
de Aena. Working Paper. Dirección de Planificación de Infraestructuras. Madrid.
Campos, J. (2006): “Spain: the end of an era”, in GómezIbáñez, J. and De Rus, G.
(eds.) Competition in Railways, Edward Elgar. New York.
Campos, J. and G. de Rus (2009): The cost of building and operating a new high
speed line, en G. de Rus (ed.) Economic Analysis of High Speed in Europe.
Fundación BBVA. Bilbao. Available at www.fbbva.es
Campos, J. and G. de Rus (2009): “Some stylized facts about highspeed rail. A review
of HSR experiences around the world”. Transport Policy, 16: 1928.
De Rus, G. and C. Román (2006): “Análisis Económico de la Línea de Alta Velocidad
MadridBarcelona”, Revista de Economía Aplicada, XIV, 42: 3580.
De Rus, G. and C. Nash (2008): ¿En qué circunstancias está justificado invertir en
líneas de alta velocidad ferroviaria? Working Paper 3/209. Fundación BBVA.
Bilbao. Available at www.fbbva.es
De Rus, G. and G. Nombela (2007): “Is the Investment in HighSpeed Rail Socially
Profitable?”. Journal of Transport Economics and Policy, 41(1): 323
De Rus, G. and V. Inglada (1997): “Costbenefit analysis of the highspeed train in
España”, The Annals of Regional Science, 31, 175188.
Ministerio de Fomento (2007): Plan Sectorial del Transporte por Ferrocarril 20052012.
Available at www.mfomento.es.
SDG (2004): High Speed Rail: International Comparisons. Prepared by Steer Davies
Gleave for the Commission for Integrated Transport. London.
UIC (2005): Estimation des resources et des activités économiques liées a la grande
vitesse. Prepared by CENIT (Center for Innovation in Transport, Universitat
Politecnica de Catalunya). October 2005. Paris.
1. Introduction
The economic evaluation of public investment projects is one of the fields of applied microeconomics which has gained larger relevance in the current context of budgetary constraints. Pressures to find the most efficient allocation of resources are increasingly being translated at every level of government into the need of selecting good projects and investing public money where its social return could be more rewarding. There exist several wellestablished techniques for conducting this assessment, but most of them simply try to compare the benefits that the society as a whole accrues from different types of public intervention in the markets, with the opportunity cost of the resources used to finance such interventions. Transport markets make an especially fruitful area for all sorts of costbenefit analyses (CBA), since measuring and giving money values to traffic flows, travel time and other transport costs seldom result as controversial as, for instance, evaluating the impact of health or environmental policies. The economic evaluation of any public intervention in transport markets particularly aims at quantifying in monetary terms the change in social welfare generated by this intervention. To evaluate a transport project is to identify, assess, aggregate and compare the costs and benefits generated by the project once discounted over time. These costs and benefits can be solely valued both from a financial point of view (just considering the revenues and costs generated by the project), or from a more general, social, point of view, which is the focus followed in this paper. In the case of highspeed rail (HSR) infrastructure and services, the main social benefits are primarily gained from time savings enjoyed by existing users, by users that switch in from other modes, and from the willingness to pay of the newly generated demand. Where appropriate, these benefits should also be weighed against those experienced by users of alternative transport modes, either because congestion (in roads or airports) is reduced and capacity is freed up, or because lives are saved (in roads) due to the deviation of road traffic. On the opposite side, the main social costs of HSR investments correspond to the expensive construction and maintenance costs of the infrastructure, the operation and maintenance of services provided over it, and other external costs (landscape effects, noise, pollution, etc.) whose components are often more difficult to quantify. If the benefits outweigh the costs, investment in highspeed rail is, in principle, socially desirable. However, even a positive net present value does not preclude the possibility that there may be alternative projects that could provide a higher social benefit. When the amount of the investment is large enough – for example in the case of huge national or supranational investment programmes such as the Spanish PEIT3 or the European Structural and Cohesion Funds – the accuracy required when adopting decisions on particular projects is also larger. Although this is not a particular requirement for rail investment and our analysis could be easily extended to other infrastructure areas, the current worldwide expansion of HSR projects, jointly with the fact that most of these projects are very expensive and particularly uncertain, sometimes with dubious social returns in a 2030 years horizon, all make of this sector a very interesting case study.
3
The term PEIT (Plan Estratégico de Infraestructuras de Transporte) stands for Strategic Plan for Infrastructure and Transport represents the dorsal spine of current Spanish public investment policy. It compromises projects worth more than €200 billion over the 20052020 period, as explained in Section 2.
2
Note that. we will follow and upsidedown approach where demand uncertainty plays a central role. as in the overall amount of investments to be made in this sector (more than €108 billion). 2009). Although the Plan includes actions in many different areas. 4 The MAGLEV (magnetic levitation) technology is increasingly viewed as the next standard for high speed surface transportation. Section 4 is then devoted to the practical implementation of this model in full detail and result analysis. Section 2 describes in larger detail the rail investment projects within the PEIT. We will not simply compare (social) benefits and costs discounted (from an initial estimate of demand figures) to calculate a single value of a social net present value (SNPV). is that it internalizes and isolates most of the demand uncertainty in HSR projects. as it has been used. aimed at modernizing the country infrastructure in the 20052020 horizon. Using the characteristics provided by this description. while in Section 5 the main findings and lessons are discussed. HSR projects in the Spanish PEIT program The Strategic Plan for Infrastructure and Transport (PEIT) is an ambitious planning tool developed by the Spanish Ministry of Public Works and Transport (Ministerio de Fomento) to encompass all its medium and longterm policy actions within a single framework. the society would gain more investing its resources on alternative projects. where the project (the construction and operation of a new highspeed line) is defined and the assessment model is formulated to calculate the value of the initial demand figures that guarantee a positive expected SNPV. we can then build an „ideal HSR project‟ that comprises most of these characteristics. 3 . This ideal is a standard highspeed rail corridor whose technical and cost data can be easily extracted either from the PEIT or the existing literature (see Campos et al. The Plan itself gives the railroad a „leading role‟ and establishes as one of its specific objectives „(.) to achieve a complete restructuring of the current railway network‟. using well estimated demand and cost data. that will be illustrated with data and examples extracted from the Spanish PEIT and other international references. both in quantity (43. 2. If this minimum amount is not met. the rest of this paper is structured as follows. rail transport concentrates the most important government interventions.. making it possible to determine. in De Rus and Nombela (2007) and De Rus and Nash (2008).It is precisely the inevitable uncertainty associated with factors such as demand predictions or the future technology developments in the rail sector4 what leads us to adopt in this paper an unconventional approach into costbenefit analysis. for example.. as a decision reference. Instead of conducting a standard assessment of a single highspeed line or corridor. After this introduction. The main advantage of this methodology. our contribution adds a larger role for uncertainty and explicitly shows that the reject/accept decision cannot be made based upon a deterministic SNPV but on a complete probability distribution of SPNV values.7% of total actions). the „minimum demand threshold from which a project would be socially profitable‟. we will try to figure out what is the initial demand level that the project should have in order to provide a positive (expected) social return.. if successful. all current investments on HSR could become obsolete in 1020 years. Although this methodological approach is not completely new in the CBA literature. On the contrary. This is carried out in Section 3.
Furthermore. This situation had created notable differences in the performance levels across different lines of the network. the exact investment estimate for the projects included in the 20052020 horizon amounts to €108.494 km without electricity.The reasons that explain this particular interest can be found in the initial diagnosis made in the PEIT of the poor conditions of the Spanish railways at the beginning of the century. that is. gradually extending the HSR network across the country and improving the interoperability with the French network. The railway restructuring projects included in the PEIT were supposedly designed to upgrade the Spanish network in order to progressively transform this mode into „a central element for the articulation of intermodal transportation‟. These lines clearly make the future pattern of the rail network rigid.) actions from 2013 onwards will be set out in accordance with the results obtained in the previous period in terms of evolution of the modal share of railways.600 km of new tracks. HSR for passengers or improved speed for cargo). have focused on completing highspeed corridors that were already under construction (most notably the MadridBarcelona one). Actions to be taken after 2009 are mainly designed to improve services. Madrid to the provinces) by developing peripheral routes and getting all the provincial capitals connected to the high performance network.000 km of the network.628 km of electrified singletrack and still 5. The projects affect more than 9. In the case of rail. The network. The guidelines and broad lines of action included in the PEIT have been developed through sectorspecific plans. only had 1.000 sq. using parameters which do not allow mixed traffic.500 km in 2005 (in a country of about 500.. with the risk of a possible underutilization of the two networks and of increased administrative costs for the infrastructure this involves. and on consolidating the new institutional framework. with greater timesaving to the destinations served.031 km with international (UIC) width. 2007): 1. which are revised by various government agencies (at least every 8 years) to redefine specific actions in each mode of transport. so that 90% of the population is situated within 50 km of a rail station. almost completed at the time of writing this paper. so that the conventional network has to be used for goods. but many others are yet to be defined. the Plan also aims to modify the current radial structure of the Spanish transport system (from the city capital. had also produced difficulties in their integration within the European network and led to the existence of corridors and lines with very low utilization and unlikely survival. ensuring the greatest possible social returns and territorial impact. In particular. on investment in the conventional network to improve the operating conditions of rail freight services and to facilitate exchange with the road and maritime transport. in terms of the real demand‟.095 km of electrified double track. in order to improve accessibility. 3.760 million. with a total length close to 13. it is especially relevant for the purposes of this paper the fact that the PEIT states that „(. These refer essentially to sections or lines which exceed a given traffic threshold in the year they come into service.. Actions in the 20052008 period. km). After the PEIT the network will cover the entire country thanks to the construction of over 5. These ambitious objectives are to be met in 15 years (20052020). although this figure varies across three different project types (Ministerio de Fomento. but divided into two different stages. 2. which will be improved to support high performance traffic (that is. At this moment many of the projects included at this stage have started. New trunk lines and corridors exclusively for passenger services. 4 .
to prepare for the gauge changeover. structural transversal axes and crossborder links.2. the prolongation of trunk lines. plus the addition of a future option to close the conventional line. 3. since the potential of the corridors they serve does not seem at present to allow them to specialize in passengers. if the traffic. This concerns sections or lines usually at the termination of routes. Terminal sections. but there also are actions designed for the improvement of the conventional network in order to equip it with high performance features. with less passenger traffic and some goods traffic. The systematic introduction is therefore proposed of the multilateral sleeper on Iberiangauge lines. where the traffic is significantly less than in the other two cases. As shown in Table 1. or to operate the two lines on a specialized basis. except where the existing line is being used substantially. with the resulting duplication of the network. The aim is to renew these sections entirely. functionality and service conditions allow. designed in principle for mixed traffic. seeking full interoperability in the medium term with the rest of the European network. for mixed traffic. such as the Mediterranean corridor or ValladolidBurgosVitoria. In such cases. or even the construction of new infrastructure in areas where the conventional grid did not exist or did not meet the adequate technical conditions. they all include the three types of actions described above. this development of the highspeed rail network in Spain has been carried out through five major radial corridors covering the Iberian Peninsula and connecting the national network with French and Portuguese ones. In all cases. parallel specialized lines could be created on part or all of these axes. the mixed traffic design provided for in the PEIT has to be understood as applying to the entire ambit of the corridor and not just of the specific line so that. In any event. Lines and sections on routes varying substantially from existing ones. among others. safety and service quality. the most relevant projects involve the extension of existing HSR lines. Such actions would come in a second phase of programming. 5 . This second case relates to lines with medium traffic. the final design will depend on the balance between the increased cost of construction of the mixed system and the advantages of a concentration of traffic and enhanced occupancy. Although the characteristics of each corridor are very different. for mixed traffic (passengers and freight). to increase speed. and particularly on lines with greater traffic.
Valencia (2010) [350 km] Madrid . 6 .Cadiz (+ 2012) [660 km] Madrid .Málaga (2007) [540 km] Madrid .Granada (+ 2010) [430 km] Madrid . between two major cities). This line should have an average length of between 300 and 600 km. then connecting with a transversal axis from there to France. in accordance with the above discussion and the information provided by the PEIT.Corridor Andalusia Description Will connect Madrid with the most relevant Andalusian capitals through the MadridCórdobaSeville line.Valladolid (2007) [190 km] Madrid . This line would operate in a context of great uncertainty about demand and costs associated with it. as it happened in the CórdobaAntequera (2006) line or in LleidaTarragona (2006). in April 1992. The MadridSegoviaValladolid axis will be extended to the north coast and Galicia. In most cases there are lines that start operations before the full completion of the corridor.Caceres (+ 2012) [300 km] Cáceres . while actions in subsequent phases of the Plan will materialize – as mentioned above – only if the minimum demand criterion is met. Operation starting year [distance]5 Madrid .es) Table 1. MadridMalaga).Galicia (+ 2012) [600 km] Madrid .Badajoz (+2012) [90 km] Badajoz .Barcelona (2008) [620 km] France is almost completed and will Barcelona . Will connect Madrid with Lisbon..France (2009) [150 km] become the most relevant one in the country. It is on this type of future action where the rest of the work done in this paper mainly focuses. most of the actions undertaken in the 20052009 period have consisted on the conclusion of some lines that had been in planning and construction for several years (MadridBarcelona. It should include the building of new HSR infrastructure (either by designing a new line or through a substantial improvement into a existing conventional one) and its operation within a pointtopoint corridor (i. a period of planning and construction of about five years. Madrid .Portugal (+2012) [50 km] NorthNorthwest Extremadura Portugal Source: PEIT (www.mfom. The MadridZaragoza route was inaugurated in 2003. we can now define a „representative project‟ that could serve as the unit of analysis for the following sections of this paper. Northeast Corridor The MadridSaragossaBarcelonaMadrid . MadridSeville.e.Alicante (2012) [420 km] Mediterranean Will connect Madrid with the Axis Mediterranean coast (Valencia and Alicante). 5 These are estimated dates for the entire corridor. with a total estimated project life of 40 years. MadridValladolid. although the latter can be mitigated in part by using information extracted from other projects. HSR corridors in Spain as designed in the PEIT Indeed. and the pioneering one. Therefore. as we immediately will see. Connections with provincial cities are also planned.Bilbao (+ 2012) [395 km] Madrid .
Firstly. During each of the years of operation of the line (between t =T’+1 and t=T). which includes both the costs of planning and construction. the society makes a gross investment equal to GIt euros. Project definition According to Section 2. Figure 1.1. showing different results under different scenarios. Thus. valued at t = 0. 3. in order to simplify the model. the initial investment net of its residual value (RVIT):6 6 Although it is not the most appropriate approach. the project contributes annually to the society a change in the social net benefit (with respect to the situation without project) equal to SBt. by using – for example – the social net present value (SNPV) as the decision criterion. Project structure Once the flows associated to the project are discounted with the interest rate i. step by step. we have considered that the residual value is just a percentage of the initial investment. t t T ' 1 (1 i ) T (1) where I0 summarizes. our aim is to calculate „the minimum level of traffic in the year of entry into service the new line that would guarantee a positive social return‟. the calculations to be done to produce the expected result. In this paper we will follow an inverse procedure: since the critical (uncertainty) variable is the demand. a sensitivity analysis is usually carried out ex post. the formal expression of the social net present value at t = 0 is traditionally given by: SNPV I 0 SBt . during the construction period (between t = 0 and t = T ‟). our objective will be to determine the threshold value of the initial demand that equals to zero the (expected) SNPV. 7 . our representative project consists in the construction and operation of a highspeed line of L km long. we will begin with the technical definition of the project considering some reasonable assumptions about it in order to facilitate its assessment and secondly we will illustrate. As shown in Figure 1. which may be positive or negative. All these flows are valued at 31th December each year.3. Project definition and assessment model The economic appraisal of a rail investment project traditionally consists in comparing its discounted benefits and costs. If there is uncertainty about the variables or parameters of the model.
Assuming that this value does not change with the project. Diverted traffic In the initial situation without the project (denoted by superscript 0) the users‟ generalized cost of travel for each of the existing modes (m = a. 8 . changes in taxpayers‟ surplus (GSt) and changes in the indirect effects (IEt). Thus. 0 = total travel time in mode m (without project). m 7 8 From now on. we will proceed step by step. Computing the (change in) consumers’ surplus7 All the traffic in the new high speed line in period t = 1 comes from two possible sources of demand: the diverted traffic from existing modes. subscript 1 is dropped. by users who did not previously on this route. and m 0 = monetary valuation of the travel quality in mode m (without project). or conventional rail (r). let first assume that the project has a total length of one period: all investment takes place at t = 0 and the social costs and benefits associated to the operating period are valued at t = 1. Under these conditions. we will simply denote it as vm . r) is defined as: 0 0 0 gm pm vm 0 0 m m (5) where: 0 pm = monetary component of the cost in mode m (without project). to simplify the notation. to simplify our analysis we will only consider that the first two effects are relevant in our case.8 or the newly generated traffic. An initial benchmark: the analysis of a static model To illustrate the calculations associated to the assessment model.I0 t 0 T' GI t RVIT t (1 i) (1 i)T (2) On the other hand.e. the annual change in the net social benefit (SBt) can be decomposed into changes in consumers‟ surplus (CSt). it is immediate to verify that expression (1) becomes: SNPV I 0 CS1 PS1 (1 i ) (4) 3.2.2. car (c). However. i. from a simple (static) framework to a dynamic one. i.1. Conventional rail can be omitted when the project consists on the construction of a new line. 0 vm = value of time in mode m (without project).e. which will be generically denoted as: airplane (a). (T’. 1). GSt = IEt = 0. where this mode did not previously exist. changes in operators‟ (or producer‟s) surplus (PSt). bus (b). c. b. T) = (0. so that our society only includes transport users and transport producers: SBt CSt PSt (3) 3.
It is conventionally determined by: 2 (g m 1 0 m I 0 g m )(q I qm ) . that we will simply estimate as a proportion () of the diverted traffic. and εIm represents a monetary valuation of the HSR service quality. (6) where: 0 qm = HSR trips made by travellers from other m modes.In the situation with project. 0 Q0 m qm . pI = average fare of a HSR trip. 3.2. 10 Alternatively. and the generated traffic. I is the average duration of that trip. the change in the consumers‟ surplus of users diverted from other existing modes can be expressed as:9 (g m 0 m I 0 g m )qm . so that: q I (1 )Q0 . which is assumed close to zero. (9) This expression. and the operation and maintenance costs of the HSR line. (7) where q I are total highspeed trips per year. These trips consist of the sum of all diverted demand from other modes m. conventionally denominated „rule of one half‟ in CBA. I gm pI vm I I = generalized cost of travel at HSR for travellers form other m modes. implicitly considers that the change in the social welfare of each traveller can be identified with the change that is produced in the last traveller.10 Generated traffic The surplus associated with new trips (made by both old and new users) must be added to the previous expression. 0 q I Q0 Q0 m qm . 9 . plus avoidable costs associated with the reduction of service in each mode m (because of the traffic that is deviated to the high speed line): ( p q p q I I m 9 0 0 m m 0 Cm CI ). after the investment (denoted by superscript I). Computing the (change in) the operators’ surplus (8) The change in producers‟ surplus is determined by the difference between the operators‟ income with and without project.2. this valuation could be also incorporated to the value of travel time (v).
using expression (8): N[q I ] I I qI (1 )Q0 N[Q0 ] . The number of trains in operation (N) is a crucial variable in order to determine the supply of transport services. 10 . avoidable costs depend on the characteristics of each mode. we consider q = 350. cA = unit cost (per traveller) of acquisition of the rolling stock. operating and maintenance costs of a highspeed line could be expressed as: CI I M cAqI (cO c M )N . the difficulty to obtain real values of the acquisition price (which are directly negotiated with the producer and usually includes part of the maintenance costs) suggests the possibility of calculating through the cost per passenger what simplifies the calculus and could be interpreted as a sort of leasing. the number of trains needed to serve the annual demand of travellers qI is given by I F . F H Thus. 0 In general. cM = unit maintenance cost (per train) of the rolling stock. the capacity of trains (assumed to be constant) and the daily frequency of services (assumed to be homogeneous throughout the year) are required.I where Cm are the avoidable costs in mode m and C the operating and maintenance costs of the highspeed infrastructure. It is therefore possible to express them as: 0 0 Cm cm qm (10) where cm is the unit cost (per trip) saved in mode m. if q is the effective average occupation of each train (where is the load factor q e and q is the number of seats) and H are the hours of operation per day. we also assume that all operators in alternative transport modes obtain normal profits and all their costs are recoverable.11 cO = unit operation cost (per train) of the rolling stock. since it changes as the demand does. which we assume to remain unchanged after the project (compared to the situation without the project). the level of competition and the volume of diverted traffic. as an important simplification. allowing us to omit the calculation of surpluses in other modes different from m. 365Hqe 365Hqe (12) 11 Although this cost should be formulated as a cost per train (equivalent to the acquisition price of each train). so that the frequency (F) is: ((q I / 365) / qe ) . where: (11) IM = infrastructure maintenance costs (usually fixed in terms of total traffic). For example. 12 In this case. net of its residual value.12 then the number of daily services during a year in each direction is obtained from the ratio (q I / 365) / qe . Therefore. To compute N. = 80% and H = 18. and N = number of trains needed to satisfy the demand each year. See Table 4 below. Moreover. that is.
Finally. for the sake of simplicity. m m 2 m (14) showing that this change has got three elements: time savings in the diverted traffic.2. (7) and (9): (g m 0 m 1 0 I 0 I 0 0 0 0 g m )qm (g m g m )(q I qm ) ( p I q I pmqm Cm C I ) 2 m m (13) Expanding the first term of the previous expression yields: ( p m 0 m 0 pI ) vm(0 I )qm. after incorporating the expressions of costs (10) and (11). we finally obtain as a benchmark: SB CS PS 0 1 vm (0 I )qm m 2 m 0 0 ( pm p I )qm 2 m 0 cmqm m (15) I M A (cO c M ) I 0 c (1 ) qm 365Hqe m 11 .e. i. the sum of (6). together with (12). m m m Substituting the previous result in expression (13) and simplifying the outcome.e. the previous expression leads to: v m 1 0 0 0 0 0 (0 I )qm ( pm p I ) p I qm vm (0 I )qm Cm C I . that the project does not significantly alter the monetary value of the quality associated to travel on different modes. the value of the new generated traffic and resource savings in the modes that lose traffic.3. the change in the social benefit is determined by the sum of changes in the surpluses of consumers and producers. Computing the (change in) social surplus According to (3). m where it is assumed. the change in the net social benefit can be rewritten as: v m m 1 0 0 I 0 0 (0 I )qm (g m g m ) p I (q I qm ) Cm C I . i. m m m 2 m m 2 m where. (0 1 ) 0 .3. given that m 0 0 (q I qm ) m qm and developing the intermediate term in (14).
4. For a better understanding of the procedure. m = a). for example that the diverted traffic can be expressed in relation to one of the modes.13 Among the different alternatives we might consider. the airplane. Thus. In this case. we easily obtain: 0 qa I0 (1 i) I M 0 (cO c M ) I (1 )va (0 I ) ( pa p I ) ca (1 ) c A a 2 2 365Hqe . it is necessary m to introduce some additional assumptions about the relation between different values of 0 the diverted traffic. (16) 0 where this value of qa determines the minimum value of the highspeed diverted traffic 0 from the airplane that makes this project socially profitable.2. we first consider a simpler case (where the diverted traffic comes only from a mode) and then we later generalize this result. In the general case. qm .. from qa we can immediately recover the volume of generated traffic (using the parameter) and the total demand. the change in social welfare can be finally obtained as: 13 Alternatively. expression (15) simplifies to: SB (1 0 0 0 )va (0 I )qa ( pa p I )qa a 2 2 (cO c M ) I 0 (1 )qa 365Hqe 0 ca qa I M c A 0 According to (4). So.g. 12 . Therefore. Note that. taking again the airplane (a) as the reference mode. we would have: 0 0 0 0 0 0 0 0 Q0 qa qb qc q0 qa bqa cqa f qa (1 b c f )qa f (17) where the parameters (that reflect the relation between different diverted traffics) can be determined endogenously in terms of the relative market share of each mode in relation to the highspeed rail mode. the problem is to 0 determine the value of qm in (15) so that I0 (1 i) SB . Using this procedure. so the m values are reduced into just one. isolating qa from the previous expression.3. the problem would be unsolvable. let us consider that there is only an alternative mode (e. with m alternative transport modes. Finding the minimum demand threshold: SNPV=0 The last step of the calculations is to solve the equation SPNV = 0 using expressions (4) and (15). SNPV = 0 requires that I0 (1 i) SB . since the expression SNPV = 0 would have infinite solutions. To get it.
3. To repeat the same steps. 3. The general model of T periods After this simplified introduction. we first introduce a time subscript t from expression (15) so that the change in social welfare in year t is given by: SBt 0 1 vm (0 I )qm t m 2 m 0 0 ( pm p I )qm t 2 m 0 cmqm t m (cO c M ) I 0 I M cA (1 ) qm t 365Hqe m 13 .SB 0 1 va (0 I ) vb (0 I ) b vc (0 I ) c v f (0f I ) f qa a b c 2 0 0 0 0 ( pa p I ) ( pb p I ) b ( pc p I ) c ( p0 p I ) f qa f 2 0 ca cb b cc c c f f qa (cO c M ) I 0 I M (1 ) c A (1 b c f )qa 365Hqe and solving again I0 (1 i) SB we get: 0 qa I 0 (1 i ) I M 0 I 0 I 0 I 0 I (1 ) va (a ) vb (b )b vc (c )c v f ( f ) f 2 0 I 0 I 0 I 0 I ( pa p ) ( pb p )b ( pc p )c ( p f p ) f 2 (c O c M ) I A ca cbb ccc c f f (1 )(c 365Hq (1 b c f ) e whose interpretation is similar to that of expression (16). we will now generalise the procedure assuming that the change social benefit in (1) is distributed over T years.
. Thus. 14 . which 0 is assumed to grow at a constant rate (). it is immediate to obtain the following: t T '1 T SBt (1 i)t 0 T qm t 0 I 1 vm ( m ) t 2 m t T '1 (1 i) 0 T qm t 0 ( pm p I ) t 2 m t T '1 (1 i) 0 T qm t cm t m t T '1 (1 i) (19) IM t t T '1 (1 i) T 0 T qm t A (cO c M ) I (1 ) c 365Hqe m t T '1 (1 i)t where 2 T ( T 1) 0 qmT '1 1 1 1 1 (1 i)t (1 i)T '1 1 i 1 i . qmt represents diverted traffic from mode m towards HSR in year t.The main simplifying assumption underlying this expression is that all the parameters of the model remain constant over time with the exception of traffic. 0 0 qmt 1 qmt (1 ) . (18) Therefore.. 1 i t T '1 T 0 qm t Note that this final expression is just a sum of terms of a geometric progression with common ratio (1+)/(1+i) so that we can rewrite it as: 1 T T '1 1 0 0 T T ' 1 T qmt q (1 i)T T '1 0 0 1 i mT '1 (1 ) qmT '1 qmT ' 1 1 (1 i)t 1 ( i)(1 i)T 1 (1 i )T '1 t T ' 1 i 1 and then replace it in (19).
Finally. we know that I0 SNPV =0. Model implementation and results analysis Implementing the expressions just developed in the previous section and the formal calculation of the minimum demand thresholds that guarantee the social profitability of this HSR investment project requires an exogenous definition of the set of parameters that complete the model. These are mainly technical parameters related to producers‟ and consumers‟ costs and the specific characteristics of each project. once this initial value is found. the calculation of total diverted demand. the value of the initial (first period of operation) demand (diverted from the airplane) that equals the social NPV to zero is given by: T IM 0 qaT '1 I 0 t t T ' 1 (1 i ) 0 0 (1 ) va (0 I ) vb (b I )b vc (c I )c v f ( 0f I ) f a 2 0 I 0 I 0 I 0 I ( pa p ) ( pb p )b ( pc p )c ( p f p ) f 2 O M I A (c c ) ca cbb ccc c f f (1 ) c (1 b c f ) 365Hqe As discussed below. Although the 15 . solving for qaT '1 .: T t T '1 T SBt (1 i)t must hold to guarantee that I0 t T '1 SBt (1 i)t 0 1 va (0 I ) vb (0 I ) b vc (0 I ) c v f (0f I ) f qa T '1 a b c 2 0 0 0 0 ( pa p I ) ( pb p I ) b ( pc p I ) c ( p0 p I ) f qa T '1 f 2 0 ca cb b cc c c f f qa T '1 IM t t T '1 (1 i) T A (cO c M ) I 0 (1 ) c (1 b c f )qa T '1 365Hqe 0 Thus.e. 4. from (1). generated demand and all other variables of the model is immediate. i.
In the case of the acquisition costs of the rolling stock (cA) raised in expression (11).1. for example) of the gross investment. annual maintenance costs of infrastructure are denoted by IM which are considered independent of the traffic volume.5 million euros. whose life is estimated in 40 years (T). We assume that the same number of kilometres is built every year. the relevance of our simulation results will be highly conditioned by the plausibility of the reference values that we use to perform our simulations. the provision of highspeed rail services involves two types of costs: those related to the maintenance of the infrastructure itself and those associated to the provision of the services (i. the expression (2) for the initial investment would be: I0 GI t T ' GI t T ' GI (1 i)t 1 (1 i)T (1 it )t . (2008) use a database of over 160 highspeed rail projects around the world and obtain average values for the construction costs per country. According to Campos et al. these values range between 9 and 17. In the case of Spain. (2008). Campos et al. so that the annual gross investment (GIt) between t = 0 and t = T‟ (valued at the end of each year) is given by: GI t L ckm . acquisition. 2005). 16 . With respect to the residual value of the investment in infrastructure (RVIT). therefore. 2009). out of which the first five ones (T’) correspond to the planning and construction period. I cmk L . it is not easy to obtain appropriate references on its value.000 euros per kilometre and year. our reference project is a highspeed line between two cities (nonstop) with a total length of 500 kilometres (L). According to (11). Thus. operation and maintenance costs of rolling stock). Parameters 4. a reasonable value of this parameter for Spain would be around 35. making the annual cost for a line of 500 kilometres line around 17. T' where ckm is the construction cost per kilometre in euros. Most of these come from other projects (see Campos and de Rus. Modelling the producers’ cost In accordance to previous sections.5 millions of euros per kilometre if costs of planning are included.. In France. 2009) and the authors‟ experience in analysing HSR markets (Campos. where cmkI is the maintenance cost per kilometre. it must be taken into account – as discussed in footnote 11 – that they are being considered the unit cost per traveller and not by train.1. according to UIC data (UIC.1.main point of this paper is a methodological one. we can simply consider that this is adequately represented by a proportion ( =10%. 4.e. The value of this unit cost of construction varies across projects. t (1 i )T t 0 t 0 (1 i ) t 0 T' Once the infrastructure has been built. because it is directly affected by terrain circumstances. because they only depend on M I the line length.
2. 2006) 14 This value primarily includes labour costs and energy. TGV Dúplex). However. technical specifications and other characteristics may change the price dramatically. however. 15 Recall that.there is reliable information about total acquisition values for different models of train (TGV Réseau. which will not be considered in this case.000 passenger per train and year. what implies that a train of 350 seats costs a minimum of 17 to 18 millions euros.14 The equivalent figure for the maintenance costs (cI) would be 1 million euros per train and year. Thus.000 euros per seat. all operators in alternative modes have normal profits in the corridor and all their costs are recoverable).000 kilometres per year. In principle. Table 2 shows reasonable estimates for pm and pI (HSR). Modelling the users’ cost As discussed above. assuming an average use intensity of 500. since our analysis is based on a hypothetical PEIT corridor it is difficult to estimate a value for the generalized cost. using as a reference information from MadridBarcelona corridor (620 km) and adjusting it to 500 km line.000 and 500. we consider that pm = cm..55 million euros per year. these values are around 50. as well as the proportion of the overall cost of administration and service management. i. Campos et al. a 100% degree of cost coverage in all the alternative modes.1. can be between 25 and 40 euros per passenger and year for an intensive use of between 300. according to the avoidable costs assumption in (10) (i. Table 2. (2008) shows that the average operative cost per seat is around 53.e. Price estimates per mode (one way trip. taking into account differences in vehicle types. total travel time () valued by each passenger and a set of additional factors () globally related to the „quality‟ of each mode. Thus.5 € 48 – 80.5 € 56 – 121 € Source: AENA (2006) and own elaboration. factors such as the inside design. the users‟ generalized cost (g) includes the price of each trip (p). The estimation of unit costs for operation and maintenance of rolling stock is less difficult due to the existence of better information.15 Mode Vehicle (Company) Price (620 km) Estimated price (500 km) Airplane (a) Bus (b) Car (c) Conventional rail (r) High speed rail Boeing 757 (IBERIA) Eurobus (ALSA) Mediumsized car Alvia Train (RENFE) AVE Train (RENFE) 30 – 160 € 25 – 45 € 85 € (*) 60 – 100 € 70 – 150 € 24 – 130 € 20 – 36 € 68. and although the range of variation can be high.000 euros. (*) Includes tolls and fuel consumption.e. 17 . so that the cost per a train of 350 seats (cO) would be 18. 4. with little dispersion between countries. an initial estimation of the unit cost cA.
18 . year. access time and waiting time – it can be estimated using a similar approach. we must include access time to the airport and hence. Table 3.77 hours Plane 2:37 2:24 2:24 2:14 2:22 2:30 2:22 2:29 2:20 170 km /h 2. AVE 3:25 2:40 3:30 1:40 2:00 2:45 2:25 3:55 3:00 180 km /h 2. ** An hour and a half is added to bus (6. although in this case we have in Table 3. because it depends on specific factors such as type of user.Cadiz [660 km] (Jerez) Madrid .Valencia [350 km] Madrid . the average commercial speed is much lower because of intermediate stops and traffic restrictions. common to all transport modes and users and it has an (initial) constant value of €15 per hour.A Coruña [600 km] Madrid . frequency. In the case of the airplane.5 hours). etc. and in order to simplify our analysis we initially consider that vm is an average value.94 hours Road* 4:18 5:24 6:36 3:30 4:12 6:12 3:57 6:00 3:54 100 km /h 5 hours * In road transport (car) is assumed a constant speed of 100 km/h. Although the value of time is one of the critical parameters in the appraisal of any transport project. travel purpose (work.Barcelona [620 km] Madrid .Malaga [540 km] Madrid . *** Total time in AVE is multiplied by 1. more detailed information on several Spanish corridors and their exact travel time for different modes.Badajoz [390 km] Average Speed (km/h) Estimated time for 500 km Source: AENA (2006).16 Line [length] Madrid . This simplification is removed later in order to introduce uncertainty in our model.Alicante [420 km] Madrid . For this reason.…). leisure.Bilbao [395 km] Madrid . This allows us to calculate in the last two rows. the average speed of each mode and extrapolate it to a line of 500 kilometres.5 times for conventional rail. 16 Although it is generally assumed that HSR services run at speeds exceeding 250 km/h. the average speed is further reduced. it cannot be obtained straightforwardly. Total travel time by mode: different corridors Once the travel time for each mode has been estimated it is finally necessary to define the value of time (vm) that will be used in the valuation of time savings from diverted traffic to HSR.Granada [430 km] Madrid . period of the day.With respect to total travel time – including time spent in the vehicle.
. i. while the effects on the road transport have been less significant. The following table summarizes the values assigned to the main parameters of the model. that were among the rail operator (RENFE) main lines before 1990. decreasing the load factor and flight frequencies. the annual growth rate of demand () is assumed constant throughout the life of the project. Using these shares (sm). For example in the MadridSeville corridor.4. Today. MadridMalaga and MadridCordoba). Globally speaking. the parameters in expression (17). m parameters are determined by the ratio: m sm . the introduction of the highspeed rail in 1992 reduced the demand of air transport by 3040%.e. In particular. and selecting the airplane as the reference mode. we consider a reasonable =0. Demand parameters The final set of parameters that are needed to solve model are those related to the demand characteristics. and taking a conservative approach using the data from other existing highspeed lines (MadridSaragossa or MadridBarcelona). 20% (from airplane). 4% (from car users) and 1% (from bus passengers). The conventional rail transport was also affected by the introduction of new services (MadridSeville.3. these lines have lost most of its traffic.10. We must first determine what percentages of diverted traffic to HSR come from each alternative mode. and equal to 3%. this is estimated – according to (8) – as a proportion of diverted traffic.1. 19 . it could be considered that diverted traffic ratios percentages may be safely given by 75% (from conventional rail to HSR). sa With respect to the generated traffic. Finally.
000.25 Year Year Seats % Hours Unit of measure Kilometres H ckm ckmI cA cO cI pI pa pb pc pf ca cb cc cf € /km % € / km € / passenger € / train € / train € € € € € € € € € Table 4.5 64.000 10 35.5 77 28 68. Summary of the values of the main parameters of the model 20 .500.000 1.000 88.Parameter Line length Last year of the construction of the project Last year of the project Average capacity of each train Load factor Hours of operation Unit construction cost per kilometre Proportion of the residual value of the infrastructure Unit maintenance cost of the infrastructure Unit acquisition cost of the acquisition of rolling stock (per passenger) Unit operative cost of rolling stock (per train) Unit maintenance cost of rolling stock (per train) Average price of a high speed trip (one way) Average price of a plane trip (one way) Average price of a bus trip (one way) Average price (operative cost) of a car trip (one way) Average price of a conventional train trip (one way) Avoidable average costs (plane) Avoidable average costs (bus) Avoidable average costs (car) Avoidable average costs (conventional train) Notation L T’ T q Value 500 5 40 350 80% 18 20.000.000 30 18.25 77 28 68.5 64.
15 15 15 15 15 15 20 1 4 75 10 3 Hours Hours Hours Hours Hours € / hour € / hour € / hour € / hour € / hour % % % % % % Unit of measure Table 4.) 21 .77 2.0 4.Parameter Total travel time on high speed rail Total travel time by plane Total travel time by bus Total travel time by car Total travel time on conventional rail Value of time (high speed rail) Value of time (plane) Value of time (bus) Value of time (car) Value of time (conventional train) Diverted traffic from plane Diverted traffic from bus Diverted traffic from car Diverted traffic from conventional rail Ratio of generated traffic with respect to diverted traffic Annual rate of growth of demand Notation I a b c f vI va vb vc vf sa sb sc sf Value 2.5 5.94 6. Summary of the values of the main parameters of the model (cont.
2. Results in Table 5 are changed if the reference parameters do so. particularly in the context of the PEIT.7 millions of passengers.2.670.817 passengers 3. 22 .17 the total minimum demand threshold would increase to 21. In particular. the project will not be socially profitable. Diverted traffic: from the airplane: from the bus: from the car: from the conventional train: 16.702. Calculating the initial demand thresholds From the values of the parameters in Table 4 it is straightforward to compute the expressions derived in Section 3 using a standard worksheet. However.1. 17 This would be the case of a highspeed line built on a new route where conventional rail did not previously exist. It is obvious that this is just an approximate figure conditioned by the assumptions about the parameters of the model.563 passengers 167. This increase is due to the reduction of the time savings: given the existence of less time savings in the airplane than in the conventional rail it would be necessary to attract a larger number of travellers to offset the social costs of the project. Evaluation results 4.113 passengers 12. with the proposed values and a discount rate of 5% we get the following results with respect to the initial demand (when it starts the operation period t = 6) that at least guarantees a SNPV equal to zero. from a methodological point of view our analysis still remains useful.282 passengers 18.340. 1%). according to the expected demand. First year demand requirement for a socially profitable HSR line These figures show that if the total annual demand in the first year is less than 18. for example.4. Similarly. if we modify the assumptions with respect to the origin of the diverted traffic – assuming. it is possible to make different types of sensitivity analysis on the model parameters. because it provides a simple procedure to classify and rank different project alternatives in the 20092020 horizon. out of which 18.373.3 millions of passengers.112 passengers 1. For example.7 million would come from the airplane. bus. 4%.527.028 passengers 668. some of which are summarized in the following table.098 passengers Generated traffic: Total required demand: Table 5. that HSR captures most of diverted traffic from air transport (95%) and road transport shares remain the same (car.
the calculations in our model can be repeated several times to obtain the probability distribution of the initial demand.470 27.Diverted traffic Initial situation Increase of train capacity ( q = 450 seats) Reduction of the load factor ( = 60%) Increase of the construction (ckm = 30 mill. The initial demand threshold is also a random variable and. according to Table 4).202. we can now consider that.670.191 12. the usefulness of the results in Table 6 is limited because they are built upon a ceteris paribus approach.484.544. Figure 1 shows this probability distribution. Sensitivity analysis of the model results 4.612.373.019 1.245 24.000 iterations in the computations of the previous model using @RISK software.328.601 21.968.) Reduction of the acquisition cost of trains (cA = 15 €/ pas.971.634.484 27. but this is very expensive and often unpractical) or because the transposition of values from other (previous) projects is always imperfect.761 24.243 10.720.897.873. Only one parameter is changed in each case whereas the rest remain unchanged.) Reduction of the construction cost (ckm = 15 mill.702. either because of lack of information about the current and future users (a detailed survey could be carried out. As noted above this is a critical variable in the project appraisal literature.) Increase in the value of time (v = 30 €/h) Reduction of the demand rate of growth (= 1%) 16. the expressions of the SNPV used in Section 3 should be now viewed as the expected value of the SNPV.364 Total demand 18. an alternative approach consists in identifying the parameters and variables in Table 4 whose exact values are unknown by the analyst and model them through explicit probability distributions (normal.817 13.187. the results in Tables 5 and 6 can be given in terms of probability distributions.494. A step further: the model under uncertainty Although the sensitivity analysis procedure is fairly standard.766 9.007 Table 6. the VoT is distributed between €5 and €30 according to a triangular probability distribution.284. in most cases its specific value is very difficult to determine.448 2.098 14.922.5 and 25. However. the average value approximates to the one obtained in the first row of Table 5 although it does not exactly coincide because this figure is the result of a simulation process.840.) depending on the known range of values.225 2.263. triangular. therefore.160 2. As it can be seen. etc.9 23 .543 11.210 13.061.477 997. Under this approach. In any case. obtained after 5. Just to illustrate these ideas. What happens when the project is simultaneously affected by several sources of uncertainty? In this case.324.932 29.2.2. instead of setting a deterministic value (€15.643 Generated traffic 1. uniform. The interpretation of the figure is immediate: to guarantee a SNPV > 0 our project should attract a minimum number of passengers during the first year of between 12.049.413 1.282 1.130 10. consider for example the case of the value of time (VoT). After this. and using a suitable worksheet.
in the words of the Spanish Strategic Infrastructures and Transport Services Plan (PEIT) is to „select (or rank) projects according to their real demand. the aim of this paper has been to extend received CBA methodology in order to develop some new procedures to detect the circumstances under which such proposals might be socially worthwhile. Interestingly. finding the minimum traffic threshold (in the year of entry into service) which 24 . Conclusion Highspeed rail is widely reckoned as one of the most significant technological breakthroughs in passenger transportation nowadays. the utility of this approach is selfevident: the decision of accepting or rejecting a particular investment project is carried out taking into account the demand uncertainty and other sources of uncertainty simultaneously. it is a more complex approach to CBA. 5.million. What is wanted. but building. Our approach departs from traditional CBA but changes its perspective: since demand uncertainty is the critical variable to accept or reject most rail projects. but also more realistic. note that there are no limits to introduce additional sources of uncertainty in the model and even that correlations between those sources could be properly taken into account.5%. Ultimately. particularly on the demand side. involving a margin of error into the decision that can be set ex–ante by the analyst. In this context. Probability distribution of the initial demand (total) From a methodological point of view. Figure 1. we transform the traditional SNPV > 0 acceptance criterion into a „demanddriven criteria‟ and thus obtain different values for this critical demand under alternative simulated scenarios. maintaining and operating HSR lines is very expensive and their potential benefits are always subject to great deal of uncertainty. This result is valid with a probability of 89.
Dirección de Planificación de Infraestructuras. G. Ministerio de Fomento (2007): Plan Sectorial del Transporte por Ferrocarril 20052012. New York. Fundación BBVA. 16: 1928. G. Román (2006): “Análisis Económico de la Línea de Alta Velocidad MadridBarcelona”. and G. 175188. but into the methodology itself. de Rus (2009): “Some stylized facts about highspeed rail. and V. Available at www. De Rus. Working Paper. 41(1): 323 De Rus. SDG (2004): High Speed Rail: International Comparisons. Madrid. A particularly important additional element that is also worth noticing here is the possibility of introducing several sources of uncertainty in the estimation of demand thresholds. and C. J. Available at www. Under the appropriate simplifying assumptions. De Rus. October 2005. 25 . in GómezIbáñez. 31. and expanding in full detail the traditional the expression of social NPV we derived an assessment model that allows us to obtain that minimum demand threshold. London. Edward Elgar. References AENA (2006): Evaluación del impacto en las previsiones de tráfico de los aeropuertos de Aena. we first analyzed the main characteristics of the rail corridors proposed in the rail sector plan within the PEIT. Prepared by Steer Davies Gleave for the Commission for Integrated Transport.) Economic Analysis of High Speed in Europe. and G. Campos.) Competition in Railways. and De Rus. Fundación BBVA. Bilbao. Paris. de Rus (ed. en G. To achieve this goal. G. G. It is basically a methodological tool to provide a quick reference regarding the size of the corridors where HSR projects are socially viable or not. just as an example. Prepared by CENIT (Center for Innovation in Transport. However. de Rus (2009): The cost of building and operating a new high speed line. and in order to also provide some realistic examples.fbbva.guarantees a positive social profitability‟. Nash (2008): ¿En qué circunstancias está justificado invertir en líneas de alta velocidad ferroviaria? Working Paper 3/209.es De Rus. Our results show. We then defined a reference hypothetical project that involved the construction and operation of a highspeed (new route or improvement of an existing conventional line). From here. UIC (2005): Estimation des resources et des activités économiques liées a la grande vitesse. that under reasonable assumptions extracted from previous experiences the minimum demand necessary to ensure a positive social return can be placed around 18. if needed) according to the analyst‟s needs. a figure that is substantially higher than any of the projects currently discussed in Spain. and G. and C. Nombela (2007): “Is the Investment in HighSpeed Rail Socially Profitable?”. Inglada (1997): “Costbenefit analysis of the highspeed train in España”. A review of HSR experiences around the world”.3 million passengers (in the first year). J. XIV. Bilbao. The Annals of Regional Science. G. Universitat Politecnica de Catalunya). (2006): “Spain: the end of an era”. 42: 3580.es Campos.mfomento.fbbva. the relevance of this exercise does not lie in this value. Transport Policy. which is a novelty with respect to previous literature.es. J. this model can be easily implemented (and suitably changed. Revista de Economía Aplicada. J. (eds. Available at www. Campos. Journal of Transport Economics and Policy.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.