Railway (De-)Regulation – A European Comparison
Guido Friebel, Marc Ivaldi, Catherine Vibes
November 2003Railroad Conference, Toulouse
Subject of the study Intensive reform discussion: how to increase
European railroad efficiency? Three types of reforms (Directive 91/440):
Separation infrastructure/operations Independent regulatory body Third-party access
Large variation across time and countries in adopting these reforms
Main question: What does the experience in EU countries teach us about the effect of reforms on railway efficiency?
Approach Production function Measure of efficiency: closeness of a given railway firm to
“production frontier”? Efficiency = residual that is not explained by:
Technological elements Reforms in the law book
For instance: quality of management, implementation of reforms...
Our specific interest: what is the impact of reforms on productivity trends?
Approach allows to look at Effects of reforms Efficiency of railways: over time and across countries
Main results Reforms increase output More reforms are not necessarily better
than less reforms It depends on sequencing:
packages of reforms are neutral or even bad sequential reforms improve efficiency
More favorable efficiency development for smaller countries than for larger countries,
except for Sweden and Germany.
Data Worldbank: information about physical
inputs and outputs Inputs: route kilometers, staff Outputs: total kilometers (freight and
passenger), passenger kilometers Reforms: date of adoption of three reforms:
Separation Regulatory institution Third-party access
Data strengths and weaknessesStrengths: Physical data: most comprehensive data set
available Institutional data: variation over time and
across countriesWeaknesses: Lacking data of UK during reform period:
clearer (better) results without UK Institutional data:
Problem of compatibility across countries Very different types of reform implementation
Deregulation events, three main aspects
separation infra-structure, operations
third party access
independent regulatory entity
Austria 1997 1995 Belgium 1998 Denmark 1997 1999 Finland 1995 1999 France 1997 1997 Germany 1994 1994 Italy 1998 1999 The Netherlands 1995 1995 Portugal 1997 1997 Spain 1996 1995 Sweden 1988 1989 United Kingdom 1993 1993 1993
The model Cobb-Douglas production function
After log-linearization
Country fixed effects and time trend
y=output, K=Capital, L=Labor OLS Estimation, robustness check: LISREL
LK LAKy
LKAy LK lnlnln
0( )it i it itA Deregulation t
Result 1: Reforms increase productivity
Variables Parameter estimate t-value Intercept -1.327*** -4.92 Logarithm (Capital) 0.526*** 9.77 Logarithm (Labor) 0.737*** 15.53 Deregulation Productivity trend 0.004* 1.76 Productivity trend Austria 0.009** 2.03 Productivity trend Belgium 0.020*** 4.40 Productivity trend Denmark 0.038*** 7.75 Productivity trend Finland -0.002 -0.39 Productivity trend France 0.049*** 9.62 Productivity trend Germany 0.024*** 4.42 Productivity trend Italy 0.050*** 11.57 Productivity trend The Netherlands 0.081*** 16.47 Productivity trend Portugal 0.032*** 6.28 Productivity trend Spain 0.039*** 7.15 Productivity trend Sweden 0.026*** 3.85 Productivity trend United Kingdom - -
Distinguishing reforms Result 1 does not take into account:
Intensity of reforms Type of reform
Regression on quantity of reforms only: more than one reform does not improve efficiency
Distinguish sequencing of reforms: Partial Sequential Package
Result 2: Sequencing matters
Variables Parameter estimate t-value Intercept -1.133*** -4.23 Logarithm (Capital) 0.518*** 9.53 Logarithm (Labor) 0.711*** 14.93 DeregulationPartial Productivity trend 0.008* 1.59 DeregulationSequential Productivity trend 0.011** 3.05 DeregulationPackage Productivity trend -0.005 -1.28 Productivity trend Austria 0.004 0.89 Productivity trend Belgium 0.018*** 3.63 Productivity trend Denmark 0.035*** 6.98 Productivity trend Finland -0.009 -1.48 Productivity trend France 0.054*** 10.15 Productivity trend Germany 0.031*** 5.31 Productivity trend Italy 0.049*** 11.04 Productivity trend The Netherlands 0.085*** 16.98 Productivity trend Portugal 0.035*** 6.87 Productivity trend Spain 0.034*** 5.80 Productivity trend Sweden 0.017** 2.25
Efficiency measure
Global efficiency:
Passenger traffic efficiency:
))max(exp(/
)max(lnln
lnln
*
*
iiiii
iii
iii
uuyyE
uxy
uxy
exp[( max( )) (ln max(ln )]i i i i iPE tonkm tonkm
Efficiency development over time, total traffic, larger countries
Larger countries
0.2
0.4
0.6
0.8
1
1.2
Periods
Eff
icie
ncy
Lev
els
France Germany Italy Spain Sw eden
Efficiency development over time, total traffic, smaller countries
Smaller countries
0.2
0.4
0.6
0.8
1
1.2
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
Periods
Eff
icie
ncy
Lev
els
Austria Belgium Denmark Finland Netherlands Portugal
Efficiency development over time, passenger traffic, larger countries
Larger countries
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Periods
Eff
icie
ncy
Lev
els
France Germany Italy Spain Sw eden
Efficiency development over time, passenger traffic, smaller countries
Smaller countries
0.2
0.4
0.6
0.8
1
1.2
Periods
Eff
icie
ncy
Lev
els
Austria Belgium Denmark Finland Netherlands Portugal
Relative efficiency among larger countries, five-year periods, total traffic
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1980-1984 1985-1989 1990-1994 1995-2000
Periods
Eff
icie
ncy
Lev
els
France
Germany
Italy
Spain
Sweden
Relative efficiency, larger countries, five-year periods, passenger traffic
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1980-1984 1985-1989 1990-1994 1995-2000
Periods
Eff
icie
ncy
Lev
els France
Germany
Italy
Spain
Sw eden
Summary and implications First detailed test of effect of reforms on
railroad efficiency Reforms help increase efficiency More reforms are not necessarily better:
sequencing seems to matter Additional result: Institutional/full
separation of infrastructure do not score better than organizational reforms (when including UK)
Much need to dig deeper into differences in implementation
Backup slides
The mean-and-covariance structure analysis: the LISREL model
z
B
Advantage of the method: it allows to solve the potential problem of correlations between input quantities and individual effects.
The theoretical model:
η= latent variablesz=observed variablesEstimation of the model entails choosing values for the parameters so that the predicted covariance matrix fits the empirical one.
Efficiency comparison, total traffic
a) Smaller countries
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
Periods
Eff
icie
nc
y le
ve
l
Austria Belgium
Denmark Finland Netherlands Portugal
Efficiency comparison, total trafficb) Larger countries
0.40
0.50
0.60
0.70
0.80
0.90
1.00
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
Periods
Eff
icie
nc
y le
ve
l
France Germany Italy Spain Sweden
Intensity of reforms
Variables Parameter estimate t-value Intercept -1.328*** -4.94 Logarithm (Capital) 0.510*** 9.49 Logarithm (Labor) 0.750*** 15.86 DeregulationOneAspect Productivity trend 0.008** 2.35 DeregulationTwoAspect Productivity trend 0.002 0.83 Productivity trend Austria 0.01** 2.11 Productivity trend Belgium 0.019*** 4.00 Productivity trend Denmark 0.036*** 7.48 Productivity trend Finland -0.002 -0.34 Productivity trend France 0.051*** 9.81 Productivity trend Germany 0.025*** 4.50 Productivity trend Italy 0.051*** 11.75 Productivity trend The Netherlands 0.082*** 16.73 Productivity trend Portugal 0.033*** 6.58 Productivity trend Spain 0.040*** 7.46 Productivity trend Sweden 0.029*** 4.25
Separation of infrastructure from operations
Organisational Institutional or Full Austria From 1997
Belgium From 1998
Denmark From 1997
Finland From 1995
France From 1997
Germany 1994-1998
Italy From 1998
The Netherlands From 1995
Portugal From 1997
Spain From 1996
Sweden From 1988
United Kingdom From 1993