icto d'agostino-iannone-ferrari 2012

The Relative Economic and Technical Efficiency of Selected International

Container Terminal Operators

May 2012

Dr. Emanuele D’Agostino (PhD in Transport economics, Genoa, Italy)

Dr. Fedele Iannone (University of Genoa, DIEM, Genoa, Italy)

Prof. Claudio Ferrari (University of Genoa, DIEM, Genoa, Italy)

E-mail: [email protected]

Structure of the presentation

• The international container terminal industry

• Review of DEA applications

• Overview of the methods and data used to assess the ICTOs’ efficiency

• The DEA models used for evaluations

• The panel approaches used for the ICTOs’ DEA over time

• Descriptive statistics of models’ variables and price data

• Some analysis’ results

• Main conclusions and future research

The international container terminal industry (1)Industry group ICTOs Ownership

(majority)Country of

originGeographical spread of

terminal portfolioCMHI Public Hong Kong Regional

DP World Public UAE GlobalDragados Private Spain Multi-marketEurogate Private Germany Multi-marketGrup TCB Private Spain Multi-market

HHLA Public Germany RegionalHPH Private Hong Kong Multi-marketICTSI Private Philippines Multi-market

Modern Terminals Private Hong Kong RegionalNew World Holdings Private Hong Kong Regional

PSA Public Singapore Multi-marketSIPG Public China Multi-market

SSA Marine Private US Multi-marketChina Shipping Public China Multi-market

Evergreen Private Taiwan Multi-marketHanjin Private South Korea Multi-market

Hyundai Private South Korea Multi-marketK Line Private Japan Multi-marketMOL Private Japan Multi-marketMSC Private Switzerland Multi-marketOOCL Private Hong Kong Multi-market

Yang Ming Private Taiwan Multi-marketAPL Private Singapore Multi-market

APM Terminals Private Netherlands Multi-marketCMA CGM Private France Multi-market

Cosco Public China Multi-marketNYK Private Japan Multi-marketBBI Private Australia Regional

Euroports Private Luxembourg RegionalGoldman Sachs Private US Multi-market

Macquarie Private Australia Multi-marketPorts America Private USA Regional

RREEF (Deutsche Bank) Private Germany Multi-market

Financial holdings

Pure stevedores

Integrated carriers

Hybrid operators

Source: author’s processing

The international container terminal industry (2)

237.

018

1.1

283.

117

2.2

367.

413

9.7

443.

117

0.7

548.

513

6.2

596.

814

4.6

644.

710

7.2

688.

013

2.9

0.0

100.0

200.0

300.0

400.0

500.0

600.0

700.0

800.0

900.0

Wor

ld co

ntai

ner t

erm

inal

capa

city

(Mill

ion

TEU

)

2003 2004 2005 2006 2007 2008 2009 2010Year

Control of world container terminal capacity by ownership (Million TEU)

International container terminal operators Other private and public operators

Source: author’s processing based on data by Drewry


Source: author’s processing based on data by DrewrySource: author’s processing based on data by Drewry

175.

822

3.4

201.

624

1.5

230.

626

6.6

247.

627

8.2

226.

425

1.6

252.

529

6.0

0.0

100.0

200.0

300.0

400.0

500.0

600.0

Wor

ld co

ntai

ner p

ort t

hrou

ghpu

t (M

illio

n TE

U)

2005 2006 2007 2008 2009 2010Year

Control of world container port throughput (Million TEU)

International Container Terminal Operators (equity throughput) Other private and public operators


Breakdown of ICTOs' equity throughput

0.0

20.0

40.0

60.0

80.0

100.0

120.0

140.0

160.0

2005

2006

2007

2008

2009

2010

Year

Equi

ty th

roug

hput

(Mill

ion

TEU

)

Pure stevedores

Integrated carriers

Hybrid companies

Financial operators



Euroports

Hyundai Merchant MarineYang Ming LineMacquarieGrup TCBOOCLMitsui OSK LineK Line

Deutsche Bank (RREEF)Dragados / Noatum

NYK ICTSIAPLHanjinCMA CGM

HHLAEurogateEvergreenSSA MarinePorts America Group

MSC

Cosco

APMT

DPW

HPHPSA

0.0

20.0

40.0

60.0

80.0

100.0

120.0

140.0

0.0 10.0 20.0 30.0 40.0 50.0 60.0

Equity throughput in 2010 (million TEU)

Capa

city

in 2

010

(Mill

ion

TEU

)




PSA

HPHDPW APMT

CoscoMSC

Ports America GroupSSA MarineEvergreenEurogateHHLACMA CGMHanjinAPLICTSI NYK Dragados / Noatum

Deutsche Bank (RREEF)K LineMitsui OSK LineOOCLGrup TCBMacquarieYang Ming Line

Hyundai Merchant MarineEuroports

0.0

10.0

20.0

30.0

40.0

50.0

60.0

0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0

Total throughput of all participated terminals in 2010 (million TEU)

Equi

ty th

roug

hput

in 2

010

(mill

ion

TEU

)


Top 5 ICTOs: container throughput and the market share on global container port throughput

0.0

50.0

100.0

150.0

200.0

250.0

300.0

350.0

2005 2006 2007 2008 2009 2010

Year

Cont

aine

r thr

ough

put (

Mill

ion

TEU

)

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

ICTO

's m

arke

t sha

re o

f glo

bal

cont

aine

r por

t thr

ough

put

Throughput of a l lterminals in whichshareholdings held

Throughput of a l lterminals in which10%+ shareholdingsheld

Equity throughput

Market share of top 5ICTOs ' throughput (foral l terminals in whichshareholdings held)

Market share of top 5ICTOs ' throughput (foral l terminals in which10%+ shareholdingsheld) Market share of top 5ICTOs ' equitythroughput


Herfindahl-Hirschman Index (HHI)

980.001000.001020.001040.001060.001080.001100.001120.001140.001160.001180.00

2005 2006 2007 2008 2009 2010

Year

HHI




HPHDPW

APMT

CoscoMSC

SSA Marine

Evergreen

EurogateHHLA

CMA CGM

Hanjin

APL

ICTSI

NYK Dragados / NoatumK LineMitsui OSK LineOOCL

Grup TCB

Yang Ming Line

Hyundai Merchant Marine

Ports America/AIG Highstar

Deutsche Bank (RREEF)Macquarie

-20%

-10%

0%

10%

20%

30%

40%

0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20%

Share of total equity throughput in the international container terminal operating industry in 2010

Aver

age

grow

th o

f equ

ity th

roug

hput

dur

ing

2007

-201

0

“Question marks” “Stars”

“Dogs”“Cash cows”


Source: author’s processing based on data by Drewry and companies’ annual reports

Average profitability in the international container terminal operating industry (real values: 2007 US$ per equity TEU) - unbalanced panel data set (14 ICTOs)

0.00

20.00

40.00

60.00

80.00

100.00

120.00

140.00

2005 2006 2007 2008 2009 2010

Year

2007

US$

per

equ

ity T

EU

Unitrevenue

Unitoperatingcost

Unit profit


Source: author’s processing based on data by Drewry and companies’ annual reports

0.00%

10.00%

20.00%

30.00%

40.00%

50.00%

60.00%

2008 2009 2010

Eurogate HPH ICTSI PSA NYK APMT DPW HHLA

EBITDA margins of selected ICTOs during 2008-2010


Review of DEA applications (1)

Cont’d

Authors Objectives Sampled DMUs and covered time period

DEA modeling approaches

Inputs Outputs

Marchese et al. (2000)

Examine the relative efficiency of terminals to evaluate the appropriateness of their mix of inputs

12 Italian container terminals, 1997-1998

Input oriented CCR and BCC contemporaneous analysis

Quay lenght, number of gantry cranes, yard capacity, manpower

Container throughput

Itoh (2002)Analyze the operational efficiency of container ports over time

8 Japanese container ports, 1990-1999

Input oriented CCR and BCC contemporaneous analysis, window analysis, intertemporal analysis, and time series analysis

Terminal area, number of berths, number of gantry cranes, manpower


Cullinane et al. (2004)

Evaluate the efficiency of container ports over time

25 world's leading container ports, 1992-1999

Output oriented BCC window analysis

Quay lenght, terminal area, number of quayside gantry cranes, number of yard gantry cranes, number of straddle carriers


Turner et al. (2004)

Measure seaport productivity growth over time and explore several causal relationships between infrastructure productivity and industry structure and conduct

26 North American container ports, 1984-1997

Quay lenght, terminal area, number of cranes


Cont’d

Authors ObjectivesSampled DMUs and covered time period


Inputs Outputs

Cullinane et al. (2005)

Examine the relationship between privatization and the relative efficiency of container ports

30 world's leading container ports and terminals, 1992-1999

Output oriented CCR and BCC contemporaneous analysis, and intertemporal analysis



Rios and Macada (2006)

Analyze the relative efficiency of operations in container terminals

23 container terminals of Mercosur, 2002-2004

Input oriented BCC contemporaneous analysis

Number of cranes, number of berths, number of employees, terminal area, amount of yard equipment

Container throughput; number of containers

handled per hour/ship

Lin and Tseng (2007)

Compare the efficiency measurements obtained by applying DEA and SFA to the same data set for the container port industry

27 international container ports, 1999-2002

Input oriented CCR and BCC

Quay lenght, number of gantry cranes, number of stevedoring equipment in yard


Ng and Lee (2007)Study the performance of container ports

6 container ports of Peninsular Malaysia, 2000-2005

Output oriented CCR and BCC contemporaneous analysis, time series analysis, and window analysis

Yard area, number of yard cranes, lenght of berths, number of quay cranes

Container throughput, ship

calls

Al-Eraqi et al. (2008)


22 cargo ports of the Middle East and East Africa, 2000-2005

DEA CCR, BCC, output oriented, window analysis

Berth lenght, terminal area, number of handling equipments

Ship calls, general cargo throughput


Cont’d



Inputs Outputs

Cheon (2009)

1) Estimate the impact on port efficiency due to the global operators' participation in container terminal operations 2) Understand how different characteristics of global operators can influence efficiency 3) Understand the implications on port strategies of the recent emergence of global operators and their performance levels

110 global hub ports and major national gateway ports, 2004

Tiered oputput oriented BCC

Berth lenght, terminal area, crane capacity, aggregated hinterland size


Lun and Cariou (2009)

Develop a reference for terminal operators to evaluate their operational performance

7 International Container Terminal Operators (ICTOs), 2006

Input oriented CCR Total operating costs

Container throughput (for all terminals in

which shareholding

held); total profit




Inputs Outputs

Al-Eraqi et al. (2010)


22 cargo ports of the Middle East and East Africa, 2000-2005

DEA BCC, output oriented, normal and superefficiency, window analysis

Berth lenght, terminal area, number of handling equipments

Ship calls, general cargo throughput

Cullinane and Wang (2010)

Compare the efficiency measurements obtained by applying panel DEA approaches to the same sample of container ports

25 world's leading container ports, 1992-1999

Input oriented BCC contemporaneous analysis, window analysis, and intertemporal analysis



Lun et al. (2011)

Examine whether container terminal operations with better firm performance and a larger scale of operations are prone to take more business risk

6 International Container Terminal Operators (ICTOs), 2006

Input oriented CCR Total operating costs

Container throughput (for all terminals in

which shareholding

held)


Overview of the methods and data used to assess the ICTOs’ efficiency (1)

Two groups of models based on different inputs and outputs variables have been constructed. The (a) group is summarized in the above table. The (b) group in the next table.

Sampled DMUs and covered time

periodDEA models (a)

Panel data approaches

Inputs in models (a)

Outputs in models (a)

Prices of inputs/outputs in models (a)

New Cost efficiency

Average unit operating cost

per TEU of deployed

capacity (real US$/TEU)

Input oriented CCR

Input oriented BCC

New Revenue efficiency

Average unit revenue per TEU

of equity throughput (real

US$/TEU)

Output oriented CCR

Output oriented BCC

Contemporaneous analysis; window

analysis; intertemporal

analysis; time series analysis

8 ICTOs (APMT, DP World, Eurogate, HHLA, HPH, ICTSI, NYK, PSA), 2006-

2010

Handling capacity (TEU

million)

Equity throughput

(TEU million); total profit (real US$ million)

Handling capacity (TEU million); total

operating costs (real US$

million)

Equity throughput

(TEU million)

Two groups of models based on different inputs and outputs variables have been constructed. The (b) group is summarized in the above table.

Sampled DMUs and covered time

periodDEA models (b)

Panel data approaches

Inputs in models (b)

Outputs in models (b)

Prices of inputs/outputs in models (b)

New Cost efficiency


per TEU of deployed


Input oriented CCR

Input oriented BCC

New Revenue efficiency

Average unit revenue per TEU

of equity throughput (real

US$/TEU)

Output oriented CCR

Output oriented BCC

8 ICTOs (APMT, DP World, Eurogate, HHLA, HPH, ICTSI, NYK, PSA), 2006-

2010

Contemporaneous analysis; window

analysis; intertemporal

analysis; time series analysis

Handling capacity (TEU

million)

Equity throughput

(TEU million)

Overview of the methods and data used to assess the ICTOs’ efficiency (2)

The DEA models used for evaluations (1)

1

1 1

where is the non-negative input matrix with the row vector

is the non-negative output matrix with

the row vector

( , ) , , 0

( ,..., ) ( ,..., ); ( ,..., )

m nn

s nj j mj n

P X Y

X = R = x x Y = R

x y x y

x xx y y

y

1 1and is a semipositive vector of

weights in which forms the linear combinations of the DMUs.

( ,..., ); ( ,..., )

j j sj n

n N

= y y = λ λ R

Assuming there are n DMUo (o = 1,…, n) and they use m inputs xio (i = 1,…,

m) to produce s outputs yro (r = 1,…, s) with the unit input costs cio (i = 1,…, m

and o = 1,…, n). The production possibility set P is:

Cont’d

Input oriented CCR and BCC models (Charnes, Cooper and Rodes, 1978; Banker, Charnes and Cooper, 1984)

Using the set P, we solve the following LP for each DMUo:

subject to:

(CCR, BCC)

(BCC)

where is a row vector in with all elements being equal to

Min

01

o

*o oθ ,

o o

θ

θ

me R

= θ

X Y

=

o

xy

e

one. The optimal

solution is not greater than one and it is called "Farrel", "technical",

"weak" or "radial" efficiency.

*θo

Input oriented CCR and BCC models (Charnes, Cooper and Rodes, 1978; Banker, Charnes and Cooper, 1984)


Using the set P, we solve the following LP for each DMUo:

subject to:

(CCR, BCC)

(BCC)

where is a row vector in with all elements being equal t

Max

01

o

* *o oφ ,

o

o

φ

me R

= φ

X φ Y

=

o

xy

e

o one. The optimal solution is greater

than or equal to one and it is called "Farrel", "technical", "weak" or "radial" efficiency. In addition,

such solution relates to that of the input oriented

*φo

.model via: Finally, in order to compare

output and input oriented efficiency scores and also to easily compute scale efficiency scores, it is usually

employed an output oriented technica

* * o oφ = 1/θ

.l efficiency index equal to *o1/φ

Output oriented CCR and BCC models (Charnes, Cooper and Rodes, 1978; Banker, Charnes and Cooper, 1984)



1

1 1 1

where is the non-negative input matrix with the column vector

is the non-negative unit cost

matrix with th

( , ) , , 0

( ,..., ) ( ,..., ) ; ( ,..., )

c

m nn

T m nj j j mj mj n

P X Y

X = R = c x c x C = R

x y x y

x xx c c

1 1

1

1

e row vector is the

non-negative output matrix with the column vector

and is a semipositive vector of weights in whic

( ,..., ); ( ,..., ) ( ,..., ) ;

( ,..., )

s nj j mj n

Tj j sj

nn

= c c Y = R = y y

= λ λ R

c y yy

h forms the linear

combinations of the DMUs.N


m) to produce s outputs yro (r = 1,…, s) with the unit input costs cio (i = 1,…, m

and o = 1,…, n). The cost-based production possibility set Pc is:

Cont’d

New-Cost model (Tone, 2002)

Using the set Pc, we solve the following LP for each DMUo:

,

subject to:

(constant return-to-scale, variable return-to-scale)

(variable return-to-scale

Min

01

o

*o o

o

=

X Y

=

x

o

ex ex

xy

e

)

where is a row vector in with all elements being equal to one. Let the optimal

solution of the above LP be Then the New-Cost efficiency is

defined by = and is not greater than o

. *o

** oo

me R

γ

xexex

ne.


New-Cost model (Tone, 2002)


m) to produce s outputs yro (r = 1,…, s) with the unit output prices pro (r = 1,…,

s and o = 1,…, n). The price-based production possibility set Pp is:

Cont’d

1

1 1 1

where is the non-negative output matrix with the column

vector is the non-negative

unit price matrix with

( , ) , , 0

( ,..., ) ( ,..., ) ; ( ,..., )

p

s nn

T s nj j j sj sj n

P X Y

Y = R = p y p y P = R

x y x y

y yy p p

1 1

1

1

the row vector

is the non-negative input matrix with the column vector and

is a semipositive vector in which forms t

( ,..., ); ( ,..., ) ( ,..., ) ;

( ,..., )

m nj j sj n

Tj j mj

nn

= p p X = R = x x

= λ λ R

p x xx

he linear combinations of

the DMUs.N


New-Revenue model (Tone, 2002)

Using the set Pc, we solve the following LP for each DMUo:

,

subject to:

(constant return-to-scale, variable return-to-scale)

(variable return-to-scale

Max

01

o

*o o

o

o

=

X Y

=

yey ey

xy

e

)

where is a row vector in with all elements being equal to one. Let the optimal

solution of the above LP be Then the New-Revenue efficiency is

defined by = and it is not greater

. *o

* oo *

o

me R

ρ

yeyey

than one.


New-Revenue model (Tone, 2002)

The panel approaches used for the ICTOs’ DEA over time Contemporaneous analysis involves the estimation of T frontiers, one for each year. This approach allows for technical progress and regress. It also allows for intersecting frontiers, which would signal local progress in a region of output-input space and local regress in another region.

Window analysis involves the estimation of a sequence of overlapping pooled panels, each consisting of a few time periods of arbitrary lenght. This approach allows for tracking efficiency trends through successive overlapping windows. It also alleviate volatility in efficiency estimates.

Intertemporal analysis involves the pooling of data to estimate a single grand frontier. This approach assumes an un-varying best practice technology, which may be tenable in short panels. It generates T efficiency estimates for each DMU, alla against the same standard, and trends in efficiency estimates of individual DMU may be of interest.

Time series analysis involves the estimation of separate single frontiers comparing each single DMU with its own efficiency across time. When there is a change of efficiency, it is difficult to determine whether the cause is an external or internal factor.

Descriptive statistics of models’ variables and price data

Equity throughput

(TEU million)

Total profit (real US$ million)

Handling capacity

(TEU million)

Total operating costs (real

US$ million)


per TEU of deployed


Average unit revenue per

TEU of equity throughput

(real US$/TEU)

Mean 20.29 655.50 46.80 1408.32 42.68 141.91Standard deviation 16.48 552.96 37.55 945.18 31.37 88.44Range 49.25 1795.73 110.63 3160.41 135.01 336.28Minimum 2.06 33.96 4.40 176.30 17.36 55.48Maximum 51.31 1829.69 115.03 3336.71 152.37 391.76Sum 811.69 26220.03 1871.91 56332.81 1707.40 5676.42Count 40 40 40 40 40 40

Outputs Inputs Price of inputs/outputs

Some analysis’ results (1)

Average results from contemporaneous analysis, window analysis, intertemporal analysis, and time series analysis

Models (a) 2006 (average)

2007 (average)

2008 (average)

2009 (average)

2010 (average)

Average

New Revenue efficiency (CRS) 0.836 0.871 0.845 0.811 0.844 0.841Output oriented technical efficiency (CRS) 0.776 0.783 0.773 0.671 0.724 0.745Output oriented scale efficiency 0.939 0.941 0.945 0.927 0.923 0.935New Cost efficiency (CRS) 0.566 0.596 0.648 0.618 0.661 0.618Input oriented technical efficiency (CRS) 0.709 0.725 0.708 0.614 0.651 0.681Input oriented scale efficiency 0.831 0.850 0.854 0.837 0.840 0.842New Revenue efficiency (VRS) 0.910 0.932 0.912 0.885 0.929 0.913Output oriented technical efficiency (VRS) 0.830 0.831 0.816 0.721 0.777 0.795Output oriented scale efficiency 0.939 0.941 0.945 0.927 0.923 0.935New Cost efficiency (VRS) 0.688 0.692 0.715 0.685 0.735 0.703Input oriented technical efficiency (VRS) 0.861 0.857 0.833 0.740 0.783 0.815Input oriented scale efficiency 0.831 0.850 0.854 0.837 0.840 0.842

Models (b) 2006 (average)

2007 (average)

2008 (average)

2009 (average)

2010 (average)

Average

New Revenue efficiency (CRS) 0.502 0.515 0.488 0.453 0.493 0.490Output oriented technical efficiency (CRS) 0.709 0.719 0.699 0.598 0.644 0.674Output oriented scale efficiency 0.857 0.866 0.868 0.862 0.842 0.859New Cost efficiency (CRS) 0.492 0.498 0.533 0.490 0.524 0.507Input oriented technical efficiency (CRS) 0.709 0.719 0.699 0.598 0.644 0.674Input oriented scale efficiency 0.858 0.878 0.884 0.878 0.883 0.876New Revenue efficiency (VRS) 0.771 0.774 0.753 0.726 0.788 0.762Output oriented technical efficiency (VRS) 0.829 0.826 0.804 0.695 0.763 0.783Output oriented scale efficiency 0.857 0.866 0.868 0.862 0.842 0.859New Cost efficiency (VRS) 0.620 0.599 0.624 0.583 0.627 0.611Input oriented technical efficiency (VRS) 0.828 0.818 0.790 0.679 0.727 0.768Input oriented scale efficiency 0.858 0.878 0.884 0.878 0.883 0.876

Average efficiency score of 8 selected ICTOs 2006 2007 2008 2009 2010 AverageRevenue efficiency (VRS) 0.930 0.922 0.924 0.952 0.954 0.936Output oriented technical efficiency (VRS) 0.811 0.812 0.791 0.780 0.752 0.789Output oriented scale efficiency 0.915 0.928 0.956 0.925 0.938 0.932Cost efficiency (VRS) 0.778 0.699 0.781 0.743 0.732 0.747Input oriented technical efficiency (VRS) 0.834 0.842 0.841 0.815 0.792 0.825Input oriented scale efficiency 0.805 0.832 0.835 0.919 0.834 0.845


Variable return to scale (VRS) contemporaneous analysis: average results from models (a)

0.600

0.650

0.700

0.750

0.800

0.850

0.900

0.950

1.000

2006 2007 2008 2009 2010

Year

Efficie

ncy

scor

e

Revenue efficiency (VRS)

Output oriented technical effi ciency(VRS)

Output oriented scale effi ciency

Cost effi ciency (VRS)

Input oriented technical efficiency(VRS)

Input oriented scale effi ciency



Variable return to scale (VRS) intertemporal analysis: average results from models (a)

0.400

0.500

0.600

0.700

0.800

0.900

1.000

2006 2007 2008 2009 2010

Year

Efficie

ncy

scor

e









Variable return to scale (VRS) contemporaneous analysis: average results from models (b)

0.5000.5500.6000.6500.7000.7500.8000.8500.9000.9501.000

2006 2007 2008 2009 2010

Year

Efficie

ncy

scor

e









Variable return to scale (VRS) intertemporal analysis: average results from models (b)

0.400

0.500

0.600

0.700

0.800

0.900

1.000

2006 2007 2008 2009 2010

Year

Efficie

ncy

scor

e







VRS New-Cost models (a)2006 2007 2008 2009 2010 Average

APMT Contemporaneous (WL1) 0.681 0.455 0.565 0.347 0.349 0.479Window Analysis WL2* 0.517 0.449 0.472 0.330 0.349 0.423Window Analysis WL3** 0.512 0.446 0.441 0.295 0.342 0.407Window Analysis WL4*** 0.512 0.446 0.423 0.278 0.303 0.392Intertemporal (WL5) 0.512 0.446 0.423 0.278 0.303 0.392Time series 1.000 0.965 1.000 0.850 1.000 0.963

DPW Contemporaneous (WL1) 0.772 0.589 1.000 0.920 0.912 0.839Window Analysis WL2* 0.543 0.583 0.809 0.863 0.912 0.742Window Analysis WL3** 0.533 0.577 0.724 0.685 0.901 0.684Window Analysis WL4*** 0.533 0.577 0.622 0.561 0.610 0.581Intertemporal (WL5) 0.533 0.577 0.622 0.561 0.610 0.581Time series 1.000 0.985 1.000 0.988 1.000 0.994

Eurogate Contemporaneous (WL1) 0.565 0.478 0.531 0.533 0.602 0.542Window Analysis WL2* 0.485 0.444 0.464 0.483 0.599 0.495Window Analysis WL3** 0.438 0.415 0.444 0.425 0.503 0.445Window Analysis WL4*** 0.438 0.415 0.436 0.417 0.483 0.438Intertemporal (WL5) 0.438 0.415 0.436 0.417 0.483 0.438Time series 0.943 1.000 1.000 0.872 1.000 0.963

HHLA Contemporaneous (WL1) 1.000 0.880 1.000 0.935 0.795 0.922Window Analysis WL2* 0.788 0.837 0.925 0.766 0.795 0.822Window Analysis WL3** 0.681 0.794 0.900 0.688 0.725 0.758Window Analysis WL4*** 0.681 0.794 0.850 0.652 0.650 0.725Intertemporal (WL5) 0.681 0.794 0.850 0.652 0.650 0.725Time series 0.991 0.967 1.000 0.979 1.000 0.987

Notes: *average value for column by two-window lenght;** average value for column by three-window lenght;*** average value for column by four-window lenght.

Cont’d


VRS New-Cost models (a)

Cont’d

2006 2007 2008 2009 2010 AverageHPH Contemporaneous (WL1) 1.000 1.000 1.000 1.000 1.000 1.000

Window Analysis WL2* 0.785 1.000 0.706 1.000 1.000 0.898Window Analysis WL3** 0.785 0.811 0.640 1.000 1.000 0.847Window Analysis WL4*** 0.566 0.601 0.516 0.954 1.000 0.728Intertemporal (WL5) 0.542 0.580 0.506 0.908 1.000 0.707Time series 0.791 0.768 0.798 1.000 1.000 0.871

ICTSI Contemporaneous (WL1) 1.000 1.000 1.000 1.000 1.000 1.000Window Analysis WL2* 1.000 0.930 1.000 0.902 1.000 0.966Window Analysis WL3** 1.000 0.849 1.000 0.803 0.938 0.918Window Analysis WL4*** 1.000 0.849 1.000 0.800 0.892 0.908Intertemporal (WL5) 1.000 0.839 1.000 0.796 0.892 0.905Time series 1.000 0.839 1.000 0.796 1.000 0.927

NYK Contemporaneous (WL1) 0.206 0.189 0.150 0.213 0.196 0.191Window Analysis WL2* 0.194 0.163 0.150 0.192 0.182 0.176Window Analysis WL3** 0.171 0.151 0.149 0.171 0.145 0.158Window Analysis WL4*** 0.171 0.151 0.149 0.170 0.145 0.157Intertemporal (WL5) 0.171 0.151 0.149 0.168 0.144 0.157Time series 1.000 1.000 0.973 0.987 0.879 0.968

PSA Contemporaneous (WL1) 1.000 1.000 1.000 1.000 1.000 1.000Window Analysis WL2* 0.752 1.000 1.000 0.957 1.000 0.942Window Analysis WL3** 0.749 1.000 0.945 0.856 1.000 0.910Window Analysis WL4*** 0.749 1.000 0.907 0.802 1.000 0.892Intertemporal (WL5) 0.749 1.000 0.814 0.802 1.000 0.873Time series 0.855 1.000 0.814 0.843 1.000 0.902



VRS New-Revenue models (a)

Cont’d

2006 2007 2008 2009 2010 AverageAPMT Contemporaneous (WL1) 0.738 0.685 0.772 0.983 0.876 0.811







VRS New-Revenue models (a)2006 2007 2008 2009 2010 Average

HPH Contemporaneous (WL1) 1.000 1.000 1.000 1.000 1.000 1.000Window Analysis WL2* 1.000 1.000 0.993 1.000 1.000 0.999Window Analysis WL3** 1.000 1.000 0.991 1.000 1.000 0.998Window Analysis WL4*** 1.000 1.000 0.983 0.995 1.000 0.996Intertemporal (WL5) 1.000 1.000 0.982 0.989 1.000 0.994Time series 1.000 1.000 0.982 1.000 1.000 0.996






VRS New-Cost models (b)

Cont’d








VRS New-Cost models (b)

Cont’d

2006 2007 2008 2009 2010 AverageHPH Contemporaneous (WL1) 0.405 0.314 0.431 0.500 0.466 0.423







VRS New-Revenue models (b)

Cont’d








VRS New-Revenue models (b)2006 2007 2008 2009 2010 Average

HPH Contemporaneous (WL1) 1.000 1.000 1.000 1.000 1.000 1.000Window Analysis WL2* 1.000 1.000 0.987 0.916 1.000 0.981Window Analysis WL3** 1.000 1.000 0.982 0.859 0.939 0.956Window Analysis WL4*** 1.000 1.000 0.973 0.847 0.914 0.947Intertemporal (WL5) 1.000 1.000 0.973 0.847 0.914 0.947Time series 1.000 1.000 0.973 0.847 0.914 0.947






Correlations among the efficiency measures from input oriented models (a)

VRS New Cost efficiency scores

VRS New Revenue efficiency scores

Input oriented BCC efficiency scores

Technical scale efficiency scores

1.000 0.522** 0.801** 0.050

(1.000) (0.577)** (0.775)** (0.011)

0.522** 1.000 0.735** 0.199**

(0.577)** (1.000) (0.709)** (0.045)

0.801** 0.735** 1.000 0.179**

(0.775)** (0.709)** (1.000) (0.047)

0.050 0.199** 0.179** 1.000

(0.011) (0.045) (0.047) (1.000)

Upper rows show Pearson’s correlation, lower rows (between parenthesis) are the Spearman ranking´s

correlation coefficients. In order to facilitate the whole understanding of the table the sign for some of

these coefficients was changed. The values with ** are statistically significant at the 0.01 level (2-tailed).






Correlations among the efficiency measures from output oriented models (a)VRS New Revenue efficiency scores


Output oriented BCC efficiency scores


1.000 0.522** 0.567** 0.007

(1.000) (0.577)** (0.564)** (0.074)

0.522** 1.000 0.819** 0.047

(0.577)** (1.000) (0.791)** (0.124)

0.567** 0.819** 1.000 0.120

(0.564)** (0.791)** (1.000) (0.324)**

0.007 0.047 0.120 1.000

(0.074) (0.124) (0.324)** (1.000)


correlation coefficients. The values with ** are statistically significant at the 0.01 level (2-tailed).






Correlations among the efficiency measures from input oriented models (b)VRS New Cost

efficiency scoresVRS New Revenue efficiency scores



1.000 0.003 0.729** 0.047

(1.000) (0.011) (0.753)** (0.115)

0.003 1.000 0.238** 0.058

(0.011) (1.000) (0.299)** (0.061)

0.729** 0.238** 1.000 0.028

(0.753)** (0.299)** (1.000) (0.125)

0.047 0.058 0.028 1.000

(0.115) (0.061) (0.125) (1.000)



these coefficients was changed. The values with ** are statistically significant at the 0.01 level (2-tailed).






Correlations among the efficiency measures from output oriented models (b)





1.000 0.003 0.242** 0.070

(1.000) (0.011) (0.311)** (0.030)

0.003 1.000 0.791** 0.149*

(0.011) (1.000) (0.798)** (0.125)

0.242** 0.791** 1.000 0.079

(0.311)** (0.798)** (1.000) (0.058)

0.070 0.149* 0.079 1.000

(0.030) (0.125) (0.058) (1.000)



these coefficients was changed. The values with * are statistically insignificant at the 0.05 level (2-tailed).

The values with ** are statistically significant at the 0.01 level (2-tailed).






Main conclusions and future research

More attention to be paid to the cost side than to the revenue one

Build flexibility in costs

Customer involvement in business strategies

Peer ICTOs and the performance targets, including most productive scale size (MPSS) targets

Profit efficiency measurement issues

Comparison of non parametric and parametric frontier estimation techniques

Decomposition of the ICTOs performances into seaparate measures of technological change and technical efficiency change

Comparison of the efficiency levels of ICTOs and shipping lines

icto d'agostino-iannone-ferrari 2012

Data & Analytics