Available online at www.sciencedirect.

com

Transportation Research Part A 42 (2008) 445–454

www.elsevier.com/locate/tra

Efficiency analysis of public transport systems:

Lessons for institutional planning
Breno Ramos Sampaio a,*, Oswaldo Lima Neto a, Yony Sampaio b
a
Dept. de Engenharia Civil, Universidade Federal de Pernambuco, Cidade Universitária, 50670-901 Recife, PE, Brazil
b
Dept. de Economia/PIMES, Universidade Federal de Pernambuco, Cidade Universitária, 50670-901 Recife, PE, Brazil

Received 21 November 2005; accepted 1 January 2008

Abstract

The transport system of the Recife Metropolitan Area is under change. As a subsidy to its change an efficiency analysis
is done with the purpose of highlighting characteristics of the efficient systems. Twelve transport system from several coun-
tries in Europe and seven from Brazil are analyzed: nine from Europe and only one from Brazil were found efficient. These
system are characterized by very different power structure and tariff structure. Efficient ones adopted a more democratic
power partition among communalities and established a more broad system of tariffs. Among other lessons it is suggested
that RMR adopts a structure that allows a more equal partition of the several municipalities comprising the Metro Area
including representatives of users groups like workers associations and syndicates. Also it should adopt a more flexible
tariff systems giving advantages to usual users at the same time that decreases costs and improves the operational efficiency.
Ó 2008 Elsevier Ltd. All rights reserved.

Keywords: Public transport; Efficiency analysis; Institutional planning

1. Introduction

Metropolitan areas have experienced in the last decades an increasing expansion bringing, as a conse-
quence, several socio-economic problems such as an unequal spatial urban development, a high pressure
on disposable infrastructure, land and housing shortages, and, with emphasis, lack of urban services. These
problems, in addition to low income and unemployment, expel poorer people to urban peripheries where
housing costs are lower. But these peripheries are diploid of public services and increase the cost of providing
urban infrastructure. Public transport, in particular, planned to operate in more density populated areas, offer
a lower frequency and quality service, due in part to larger distances and a precarious road system. Unorga-
nized urban expansion leads to an unorganized and irrational transport system in which superimposition of

*
Corresponding author. Present address: Department of Economics, University of Illinois at Urbana-Champaign, 2005 South Orchard
St., Apt. C, Urbana, IL 61801. Mobile: 1-217-819-1477. Tel.: +55 81 34581409.
E-mail addresses: brenosampaio@hotmail.com (B.R. Sampaio), oswaldolimaneto@yahoo.com.br (O.L. Neto), sampyony@yahoo.-
com.br (Y. Sampaio).

0965-8564/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tra.2008.01.006
446 B.R. Sampaio et al. / Transportation Research Part A 42 (2008) 445–454

routes is one of its characteristics. In addition, municipal system if not centrally coordinated results in super-
imposition and low coordination of routes and irrationality of the whole system.
Urban expansion, a conurbation phenomenon in which city limits loose expression bring planning difficul-
ties. Notwithstanding the difficulties, people require in each area an adequate public transport that allows easy
moves to work, shopping, educational, health, and cultural centers. Thus, a metropolitan public transport sys-
tem needs to assure mobility and accessibility through a fast, secure, regular, and trustable transport at a rea-
sonable cost. Unfortunately it is not easy to assure all these characteristics due to complex institutional
arrangements between state and several municipalities. Thus, a first step consists of working an agreement
among all political institutions involved. In particular, questions such as power division among them, admin-
istrative coordination, financing, and selection and operation of all concession to operate the several services
involved (bus, metro, vans, and so).
The main objective of this paper consists of directive propositions to a new institutional arrangement to the
Recife Metropolitan Area – RMA based on efficiency analysis of several transport systems. A data envelop-
ment analysis – DEA is adopted to select efficiency systems and their characteristics are analyzed to highlight
key propositions that may help the improvement of RMA transport system.
In the next section, questions related to quality and efficiency in public transport systems are revised. In the
third section, efficiency analysis and the DEA method are presented. In the fourth section, prior efficiency
studies of transport systems are briefly revised. In the fifth section, the systems analyzed, data basis and
selected variables are presented. In the sixth section, the Recife Metropolitan Area Transport Agency and
the Metropolitan Transport Consortium are described. The following two sections bring the results and a dis-
cussion of its consequences. Finally, in the last section, proposals for the institutional re-organization of RMA
are shown.

2. Quality and efficiency in public transport

Quality and efficiency of public transport systems may be analyzed based on several factors relating to the
quality of the service that is offered – service performance – and to the performance of the agencies and com-
panies in charge of it. As an example, Santos (2000) points several characteristics required for a good
performance:

(a) System accessibility, determined by the distance between users origin and the initial station and between
the last station and the final destination. The shorter this distance higher is route availability and, as a
consequence, geographical coverage increases, making it easy and better to people to move from one
place to the other.
(b) Travel time, determined by velocity and geometry of routes. Velocity is a function of distances, of traffic
conditions and road quality. The geometry of routes is a function of the development of a complex con-
nection of more direct and subsidiary routes.
(c) Trustworthiness, determined by uncertainty of time schedules. It can be measured by the number of trips
on time in relation to the other with delay and how much are the delay. Punctuality bring users trust and
fidelity.
(d) Frequency, determined by the time interval between each trip. Users must know the timetables, and its
changes along the day, during weekends and other special occasions.
(e) Maximum load, determined by the number of passengers in rush hours in relation to vehicle capacity.
(f) Vehicle characteristics, including age, conservation, and technology all bringing users comfort. Conser-
vation requires general maintenance, and noise and temperature control. By technology one also under-
stands door size, steps, and adoptions required by special passengers.
(g) Adequate information and support facilities, such as covered stations, schedule and timetable informa-
tion, clear indications of stations and vehicles.
(h) Mobility in accordance with necessities, that is, routes must be planned to cover the whole area and
allow flexibility in choosing an appropriate route. In addition, adaptations are required to attend pas-
sengers with motion restrictions.
 B.R. Sampaio et al. / Transportation Research Part A 42 (2008) 445–454 447

Besides quality requirements, efficiency is related to performance indicators, such as low operational cost to
users, minimum number of vehicles and personnel but without a decrease in the quality of service provided.
And efficacy is related to the number of users of public transport in relation to population, kilometers of
routes provided in relation to area, and the satisfaction level, all represent in a high quality service for the low-
est fare as possible.

3. Measuring efficiency in public transport – a DEA analysis

The efficiency of transport systems is determined by a data envelopment analysis – DEA. Urban transport
systems are considered decision making units – DMUs that relatively measured in relation to those that deter-
mine the efficiency frontier. There are two major approaches – a parametric and a non-parametric one. Para-
metric frontiers are characterized by a production function of constant parameters. This method was
originally developed by Aigner and Chu (1968). A functional form is defined and usually it is estimated by
econometric models. The specification of a functional form is the main limitation of the parametric approach,
as efficiency measures vary according to the adopted function.
The non-parametric approach does not require the a priori specification of a function. The estimation of the
frontier of the production set only requires that the production set satisfy some properties. The DEA method,
a non-parametric approach, uses mathematical programming to estimate production frontiers and calculus
efficiency scores. This method is based in the seminal paper by Farrell (1957) as later proposed by Charnes
et al. (1978).
In the DEA method, DMUs are assumed similar and differences results from input use and output
obtained. It is assumed that the production set satisfy certain properties, as mentioned, but no assumption
is made in relation to the frontier.
The production set is limited by a frontier that connects DMUs considered efficient. DMU efficiency deter-
mination results from estimation of a system of linear equation.
The model proposed by Charnes et al. (1978), assuming constant returns to scale, may be represented by N
firms or DMUs that use I input to obtain P products. Input and output quantities are represented by xi and yi
and i refers to the ith DMU. The objective is to obtain a non-parametric frontier that envelopes the data in a
manner that all united are placed on or under this frontier.
For each DMU the ratio between the weighted sum of inputs and the weighted sum of outputs is maxi-
mized, where u is a P  1 vector of weights associated to outputs and v a I  1 vector of weights associated
to inputs. The unknown vectors u and v are obtained as a result of the efficiency maximization of each DMU.
For each DMU the following problem is solved

Maxu;v ðut y i =vt xi Þ

Subject to ðut y j Þ=vt xj Þ 6 1; j ¼ 1; . . . ; N ; ð1Þ
uP0 and vP0

The model presented obtains infinite solutions. If (u*, v*) is a solution, so (au*, av*) is also a possible solu-
tion. This problem was solved by Charnes et al. (1978) imposing the condition vtxi = 1. Thus, the new pro-
gramming model is

Maxu;v ut y i
subject to vt xi ¼ 1;
ð2Þ
ut y j  vt xj 6 0; j ¼ 1; . . . ; N ;
uP0 and vP0

This new model is known as a multiplicative model and presents a great number of restrictions. Using the
linear programming dual property, the problem may be present in an equivalent form but with a smaller num-
ber of restrictions (I + P < N + 1).
448 B.R. Sampaio et al. / Transportation Research Part A 42 (2008) 445–454

Minh;k h
Subject to Y k  y i P 0;
ð3Þ
hxi  X k P 0;
k P 0:

where h is the efficiency score, k is the N  1 is the constant vector, X is the input matrix (I  N) and Y is the
output matrix (P  N).
The linear programming model is solved N times, one for each DMU. The efficiency score h might satisfy
the condition h 6 1.
Adopting the constant returns to scale assumption when not all DMUs are operating with optimal scale,
may result in efficiency measures influenced by the scale efficiency. In this case, adopting variable returns to
scale allows the measurement of efficiency independent of scale efficiency.
The model for variable returns was developed by Banker et al. (1984), by addition of a convex restriction
(ztk = 1):
Minh;k h;
Subject to Y k  y i P 0;
hxi  X k P 0; ð4Þ
t
z k ¼ 1;
kP0

where z  N  1 is the unitary vector.

One of the advantages of the DEA approach for measuring efficiency is that it produces automatically ‘‘tar-
get units” when inefficient units are found. These ‘‘target units” may be ‘‘virtual” and do not really need to
correspond to a real DMU, that is, the ‘‘target unit” may be a linear combination of efficient units in relation
to an inefficient DMU. Thus, at the same time that the DEA model identifies that a certain DMU is inefficient
it also identifies the DMUs in relation to which this DMU is inefficient. It is determined a set k weights, indi-
cating a combination of efficient units and representing the output proportion that an inefficient unit could
obtain using less inputs, in relation to ‘‘target units” (Régis, 2001).

4. Background

Several studies have been carried out to analyze the efficiency of urban transport services, using non-para-
metric techniques. A brief review of some of these studies are presented.
Karlaftis (2004) presented a review of papers analyzing the performance of transport systems. Tomazinis
(1977) specified a number of parameters to measure public transport systems and defined some basic concepts
for the evaluation, such as efficiency, productivity and service quality. Fielding et al. (1978, 1985a,b) presented
an impressive number of parameters that could be used to evaluate performance, isolating three categories:
efficiency, effectiveness and overall performance, this last one including the first two.
Viton (1997) studied the efficiency of the US bus system, applying DEA to a sample of 217 public and pri-
vate companies, using the following parameters: vehicles/distance in miles and passengers transported (out-
puts) and average speed, fleet average age, miles traveled, gas used, personal employed in the transport
service, maintenance personal, administrative personal, capital, and costs (inputs).
Chu et al. (1992) and Viton (1998) used a DEA model to develop a unique measure of performance and
concluded that the US bus system improved its productivity between 1988 and 1992. Noted also that, in gen-
eral, efficiency and effectiveness are negatively correlated.
Nolan (1996) studied technical efficiency in 29 average size US bus systems using a DEA model. As input
used the number of buses, total number of employees and gas consumed and as output vehicles per mile.
Levaggi (1994) applied a DEA model to 55 urban transport companies in Italy. As output selected the num-
ber of kilometers traveled, average speed, capital represented by a proxy, the number of vehicles, and a coef-
ficient of capacity defined by the ratio between passengers by kilometer and disposable seats per kilometer and
 B.R. Sampaio et al. / Transportation Research Part A 42 (2008) 445–454 449

population density. As input used employees cost, gas used, miles traveled, population density and number of
vehicles. They concluded that companies operated with excess capital, excess bus capacity and that salaries
represent a high percentage of total costs.
Karlaftis and McCarthy found that systems scoring highly on one attribute of performance (such as effi-
ciency, effectiveness or overall performance) generally performed well on the remaining attributes, a finding
that contradicts Chu et al. (1992).
Karlaftis (2004) applied a DEA model to 259 US systems, defining three different models according to
parameters selected. The model differ on outputs: vehicles/distance in miles; passengers per mile or passengers
transported; or both, jointly. As inputs, used: total number of employees, gas used and the number of vehicles.
Concluded that efficiency is positively related to effectiveness and that the magnitude of the economies of scale
depends on the output chosen.
Husain et al. (2000) also estimated a DEA to evaluate the efficiency of the public transportation service of
Malaysia, a sample of 46 service units. As inputs used the number of employees and total labor costs. As out-
put selected total service and companies gross revenue. They concluded that more efficient companies achieved
higher revenues.
Pina and Torres (2001) used a DEA model to analyze the efficiency of public and private transport services
in Spain. They choose as inputs: gas/distance in kilometers; cost/km or cost/passenger; and subsidy/passenger.
As output selected: bus number per kilometer per employee (bus-km/employee) a variable that provides infor-
mation on urban transport performance in respect to total employees; number of buses per kilometer per year –
a variable that shows average productivity and the bus utilization level; number of buses per kilometer per
inhabitant, representing public transport offer in each city; and also as indicators of quality: the accident rate
and frequency and agility of service provided.

5. Selected transport services, variables, and data

Nineteen public metropolitan transport systems were analyzed (Table 1): seven Brazilian, five Spanish, two
English, one French, one German, one Dutch, one Greek, and one Lithuanian. The selection of different coun-
tries services is justified, even if public transport policies are different, because of proposals similarity, that is,
all present the goal of decreasing inputs and increasing outputs, assuring the highest quality as possible. Data
analyses were applied showing consistency of the selected services; others systems were dropped out according
to this analysis.
It is known that selecting key indicators to evaluate performance of transport systems and comparing sys-
tems with use of mathematical models is a goal pursued by all authors, as a great number of indicators may
turn almost impossible a comparison and makes it difficult to generalize results (Benjamin and Obeng, 1990).
Table 1
Selected metropolitan services
Brazila European systemsb
Cities Cities Country
Recife Seville Spain
Belo Horizonte Madri Spain
Fortaleza Barcelona Spain
João Pessoa Bilbao Spain
Salvador Valencia Spain
São Paulo Londres England
Teresina Manchester England
Lyon France
Frankfurt Germany
Amsterdam Holland
Athens Greece
Vilnius Lithuania
a
Data for Brazilian services were obtained from ANTP (2001), American Public Transportation Association (2005) and European
Metropolitan Transport Authorities (2005).
b
For other services, data are available in Internet sites and additional data were provided by the services upon request by the authors.
450 B.R. Sampaio et al. / Transportation Research Part A 42 (2008) 445–454

Table 2
Variables
Inputs Outputs
X1 – operational cost Y1 – number of passengers transported
X2 – number of equivalent vehicles
X3 – number of employees

Transport service efficiency analysis are based in three basic inputs: labor, gas, and capital. Labor may be
represented by total number of employees (operational, maintenance, and administrative) or by the total
labor cost. Gas used is obtained directly in gallons or litters per year. Capital may be represented by the
number of vehicles. But several alternatives can be chosen, as shown in Section 4. Defining output is more
complex, but generally is based in efficiency and effectiveness indicators. Several authors have suggested (see
Fielding, 1987) the use of vehicles/distance traveled as a measure of efficiency and the number of passengers/
distance traveled as a measure of effectiveness, and a combination of these two as a measure of overall
performance.
Taking in consideration the data available for the selected systems, three inputs and one output were
selected (Table 2). The three inputs were total operational cost; total number of equivalent vehicles; and num-
ber of employees. These three inputs represent the basic inputs: labor, capital, and operational costs, including
gas and other expenses. The number of equivalent vehicles corrects for different transport modalities – metro
and bus; equivalence considers the number of seats of each vehicle. The number of employees considers only
those employed by the central administrative unit of the system. Thus, the emphasis is in the administration of
the metropolitan system and not in the companies that actually deliver the transport service. As output the
total numbers of passengers transported was selected, a variable that represents the efficiency of the service.
The output oriented model was selected to maximize input use, that is, labor, vehicles and cost determine
the maximum output, be it the numbers of trips, the number of kilometers traveled or the number of passen-
gers transported, among other alternatives. Also, given the different sizes and heterogeneity of the systems, a
DEA with variables returns to scale was adopted.
A comparative analysis based on efficiency scores considers those on the frontier, that is the ones efficient,
and those under the frontier. Based on the results, the comparative analysis highlights characteristics of the
efficient services that may help the re-organization of the Recife Metropolitan Area Agency. A brief review
of the RMA Agency and Metropolitan Transport Consortium is presented in the next section.

6. RMA Agency and Metropolitan Transport Consortium

The administrative agency for the Recife Metropolitan Area, known as EMTU/Recife, is composed of 29
members representing the state government, 14 municipalities that compose the metropolitan area, state con-
gress, municipal council of all 14 municipalities, public and private operators representatives, and representa-
tives of employees, of users, and of producers and service providers. But EMTU/Recife was not empowered to
administer and control all municipal transport systems. Thus, superimposition persists and a rational route
integration net is still to be implemented.
The public transport system is actually being re-organized. The objective is to approve an administrative
agency in which state and municipalities establish a real partnership. The new Metropolitan Transport Con-
sortium shall substitute EMTU/Recife. It is committed to plan, manage and control all transport services;
manage all finance subjects; and adopt as a goal the expansion and improvement of the metropolitan transport
service. It shall be composed of several different administrative and deliberative councils: (a) a higher deliber-
ative council in charge of approving all policies and directives relating to the transport system projects and
models; (b) an administrative council to implement decisions taken by the deliberative council; (c) a fiscal
council to monitor all financial aspects of the consortium; (d) a consultant council to discuss the proposed
policies, plans and project; and (e) an executive office to manage all transport services. The consortium shall
be a state company; the state owns 51% of shares, the municipality of Recife 30%, and the remaining munic-
ipalities own the other shares.
 B.R. Sampaio et al. / Transportation Research Part A 42 (2008) 445–454 451

7. Results

The efficiency scores obtained from the DEA model are presented in Table 3. The following systems were
considered efficient: Seville, Madrid, Barcelona, Bilbao, Valencia, Manchester, Amsterdam, Athens, Vilnius,
and Sao Paulo. The others are inefficient in different degree: London, Lyon, Frankfurt, Recife, Belo Horo-
zonte, Fortaleza, Joao Pessoa, Salvador, and Teresina.
Of all Brazilian systems, only one was efficient. This represents 14% of the analyzed systems. European sys-
tems scored much better: 75% of the services were considered efficient.
Table 4 contains a statistical summary of minimum, average, and maximum scores obtained in the DEA-V
model.
It is observed that 12 systems present scores above the average, 88,89%. Only three Brazilian systems scored
above the average.
The number that each system appears as a parameter for the inefficient systems are shown in Table 5.
Amsterdam is a reference for seven inefficient systems, followed by Seville that is a reference for six. Vilnius
is a reference for four, Madrid and Sao Paulo for two and Bilbao, Valencia and Manchester each for one and
Barcelona and Athens are not a reference. The importance of being a reference is due to the contribution of
their practices as a reference to increase the efficient of the inefficient systems.

8. Analysis

The efficiency scores were analyzed according to the following criteria: power partition among the compo-
nents of the administrative agency of the transport system and tariff structure.

Table 3
Efficiency scores of public transport services
System Score (%)
Seville 100.00
Madrid 100.00
Barcelona 100.00
Bilbao 100.00
Valencia 100.00
Londres 78.61
Manchester 100.00
Lyon 62.05
Frankfurt 71.65
Amsterdam 100.00
Athens 100.00
Vilnius 100.00
Recife 61.86
Belo Horizonte 69.88
Fortaleza 75.90
João Pessoa 91.27
Salvador 95.34
São Paulo 100.00
Teresina 82.50
Source: EMS software results.

Table 4
Statistical summary
Statistics DEA-V (%)
Minimum 61.86
Average 88.89
Maximum 100
Source: EMS software results.
452 B.R. Sampaio et al. / Transportation Research Part A 42 (2008) 445–454

Table 5
Number of references of the efficient system
System Number of references
Amsterdam 7
Seville 6
Vilnius 4
Madrid 2
São Paulo 2
Bilbao 1
Valencia 1
Manchester 1
Barcelona –
Athens –
Source: EMS software results.

8.1. Efficiency and power partition

Power partition among components of the administrative agency varies much for the analyzed systems. It var-
ies in the number and representation of components and also in the percentage of votes that each component
possess. According to Sampaio and Lima Neto (2005), the more power is distributed among components the eas-
iest the decision process and the acceptance of decisions by the components of the agency. On the other hand, the
predominance of one agent (more than 50% of the votes) weakens the partnership as it can decide practically
alone. If the predominance makes it easy to take decisions (does not depend on other partners) and may lead
to higher efficiency it weakens the partnership and as a consequence a division of responsibilities and costs. A
partnership among governments and user associations is another important aspect to validate the system.
To comparatively analyze the systems in relation to power partition, a index is proposed

Y ¼ N  ½1  ðK 1  K 2 Þ=K 1 
Xh X i X h X i
K 1 ¼ ð100=N Þ  Nj þ ðN iþ1 Þ and K2 ¼ Pj Nj þ ðN iþ1 Þ ð5Þ
subject to P i < P iþ1

where N is the number of components of the administrative agency and P is the percentage of votes (places)
corresponding to each component. Making Ni = 1 and defining NN+1 = 0, the value of the expression
[1  (K1  K2)/K1] varies from 0, when power partition is unequal to the maximum, to 1, when it is equal
to the maximum. This expression is similar to the Gini coefficient used to analyze the income inequality
(Sen, 1973). The expression is multiplied by N to express that the higher is the number of participants (com-

7
6 Inefficient
Number of Systems

5 Efficient

4
3
2
1
0
0,0 - 1,0 - 2,0 - 3,0 - 4,0 - 5,0 -
1,0 2,0 3,0 4,0 5,0 6,0
Y

Fig. 1. Power partition.

 B.R. Sampaio et al. / Transportation Research Part A 42 (2008) 445–454 453

ponents) more efficient the system is expected to be. Taking the value for each system the Y index was calcu-
lated. Results are shown in Fig. 1.1
Efficient systems, the only exception is Sao Paulo, present a higher Y than inefficient systems, showing that
the higher the number of participants and more equal the power partition more efficient the system tends to
be. That is, a more equal power partition is positively correlated to efficiency.
The system that presents higher Y is Madrid, with seven participants: the Community of Madrid with more
power, 25%, and the User Association with the least, 5%, show a more equal power partition than all other
systems. In addition to the two already mentioned, the municipality of Madrid and the other municipalities of
the metropolitan area, the central government, the association of companies in charge of the service and work-
ers association are components of the Agency. As also shown in Sampaio and Lima Neto (2005) a partnership
among government and beneficiaries is an important combination to improve service and efficiency.
The recently proposed (Recife) Metropolitan Transport Consortium achieves a Y equal to 2.58, an
improvement in relation to actual Brazilian systems, in relation to power partition. But it does not include
the participation of associations representing society.

8.2. Efficiency and tariff structure

All efficient systems, exception only of Sao Paulo, tariff structure is composed of three products: (a) unitary
tickets; (b) multiple trip tickets; and (c) weekly, monthly or longer periods cards for multiple trips. Inefficient sys-
tems, with exception of London, Lyon, and Frankfurt, offer one or two types, most frequently only unitary tickets.
Tariff value varies according to concentric zones. But in efficient systems, again with exception to Sao
Paulo, tariff decreases with the number of credits or the length of period of the card. Tariff final value is a
function of product type and the number of zones crossed in a trip.
Thus, tariff structure in public transport systems shall be well defined offering users several ticket types that
meet passengers interests and assure a better service and a expansion of demand. At the same time, multiple
trip tickets and cards for longer periods are provided at lower costs in the benefit of both users and transport
companies and agencies.

9. Proposals for the institutional re-organization of RMA

Efficiency analysis comparing Brazilian and European transport systems may be very important to high-
light aspects for service improvement. The DEA variable returns model used showed that only 14.3% of
the analyzed Brazilian systems are efficient or 5.3% of the total analyzed systems. In contrast, only 25% of
European systems were inefficient. A comparison of efficient and inefficient systems determines differential
characteristics. Two characteristics are emphasized in this paper: the number of participants and power par-
tition among components of administrative agencies and tariff structure.
In relation to power partition, an index was developed taking in consideration the number of participants
and the power distribution among them. It is concluded that efficient systems present a higher number of par-
ticipants, including central and local governments and associations representing communities, and that power
is more equally distributed. As a consequence, a more democratic and broad partition system shall be pursued
because it corresponds to a more developed society and may bring a higher efficient level.
Tariff structure analysis showed that efficient systems offer a great number of products, such as unitary and
multiple trip tickets and cards for longer periods, improving the quality of the service provided at the same
time that reduces costs and raises efficiency.

References

Aigner, D.J., Chu, S.F., 1968. On estimating the industry production function. American Economic Review 5 (4), 826–839.
ANTP, 2001. Anuário ANTP dos Transportes Públicos. Associacßão Nacional dos Transportes Públicos. National Public Transport
Association, São Paulo, SP, <http://www.antp.org.br>.

1
No data were available for London, Manchester, Frankfurt, Amsterdam and Vilnius, on power partition.
454 B.R. Sampaio et al. / Transportation Research Part A 42 (2008) 445–454

American Public Transportation Association, 2005. <http://www.apta.com>.

Banker, R.D., Charnes, A., Cooper, W.W., 1984. Some models for estimating technical and scale inefficiencies. Management Science 30,
1078–1092.
Benjamin, J., Obeng, K., 1990. The effect of policy and background variables on total factor productivity for public transit.
Transportation Research Part B 24 (1), 1–14.
Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring the efficiency of decision making units. European Journal of Operational
Research 2, 429–444.
Chu, X., Fielding, G.J., Lamar, B.W., 1992. Measuring transit performance using data envelopment analysis. Transportation Research
Part A 26 (3), 223–230.
Efficiency Measurement System – EMS. Software developed by Holger Scheel, 2005. <www.wiso.uni-dortmund.de/lsfg/or/scheel/ems/>.
Empresa Metropolitana de Transportes Urbanos – EMTU/Recife, 2005. <http://www.emtu.pe.gov.br>.
European Metropolitan Transport Authorities, 2005. <http://www.emta.com>.
Farrell, M.J., 1957. The measurement of productive efficiency. Journal of the Royal Statistical Society 120, 449–460.
Fielding, G.J., 1987. Managing Transit Strategically. Jossey-Bass Publishers, San Fransisco, CA.
Fielding, G.J., Glauthier, R.E., Lave, C.A., 1978. Performance indicators for transit management. Transportation 7, 365–379.
Fielding, G.J., Babitsky, T.T., Brenner, M.E., 1985a. Performance evaluation for bus transit. Transportation Research Part A 19 (1), 73–
82.
Fielding, G.J., Brenner, M.E., Faust, K., 1985b. Typology for bus transit. Transportation Research Part A 19 (3), 269–278.
Husain, N., Abdullah, M., Kuman, S., 2000. Evaluating public sector efficiency with data envelopment analysis (DEA): a case study in
Road Transport Department, Selangor, Malaysia. Total Quality Management 11 (4–6), 830–836.
Karlaftis, M.G., 2004. A DEA approach for evaluating the efficiency and effectiveness of urban transit systems. European Journal of
Operational Research 152, 354–364.
Karlaftis, M.G., McCarthy, P.S., 1997. Subsidy and public transit performance: a factor analytic approach. Transportation 24, 253–270.
Levaggi, R., 1994. Parametric and nonparametric approach to efficiency: the case of Urban Transport in Italy. Studi-Economici 49 (53),
67–88.
Nolan, J.F., 1996. Determinants of productive efficiency in urban transit. Logistics and Transportation Review 32 (3), 319–342.
Pina, V., Torres, L., 2001. Analysis of the efficiency of local government services delivery. An application to urban public transport.
Transportation Research Part A 35, 929–944.
Régis, F.A.P., 2001. Eficiência de Custo no Setor Bancário Brasileiro. M.Sc. – Universidade Federal de Pernambuco, Brazil.
Sampaio, B.R., Lima Neto, O., 2005. Transporte público de passageiros em regiões metropolitanas – estudo comparativo visando a
formulacßão de um novo quadro institucional para a RMR. In: Proceedings of the 15 Brazilian Transit and Transport Congress. ANTP –
Associacßão Nacional de Transporte Público, National Public Transport Association, Goiânia, Brazil.
Santos, B.J., 2000. A qualidade no servicßo de transporte público urbano. Arquivo PDF.
Sen, A., 1973. On Economic Inequality. Clarendon Press, Oxford.
Tomazinis, A.R., 1977. A study of efficiency indicators of urban public transportation systems. Final Report, DOT-TST-77-47, USDOT,
Washington, DC.
Viton, P.A., 1997. Technical efficiency in multi-mode bus transit: a production frontier analysis. Transportation Research Part B 31, 23–
39.
Viton, P.A., 1998. Changes in multi-mode bus transit efficiency, 1998–1992. Transportation 25, 1–21.