SIXTH FRAMEWORK PROGRAMME PRIORITY [Sustainable surface transport] Call identified: FP TREN-2

Transcription

1 SIXTH FRAMEWORK PROGRAMME PRIORITY [Sustanable surface transport] Call dentfed: FP TREN-2 Generalsaton of Research on Accounts and Cost Estmaton Case study 1.2E: Track Mantenance Costs n Swtzerland Annex to Delverable D 3 Margnal cost case studes for road and ral transport Verson October 2006 Authors: Mchael Mart, René Neuenschwander Contract: FP Project Co-ordnator: ITS, Unversty of Leeds Funded by the European Commsson Sxth Framework Programme GRACE Partner Organsatons Unversty of Leeds; VTI; Unversty of Antwerp; DIW; ISIS; Katholeke Unversty of Leuven; adpc; Arstotle Unversty of Thessalonka; BUTE; Chrstan-Albrechts Unversty; Ecoplan; IER Unversty of Stuttgart; TNO Inro, EIT Unversty of Las Palmas; Unversty of Gdansk 1

2 GRACE FP Generalsaton of Research on Accounts and Cost Estmaton Margnal cost case studes for road and ral transport Ths document should be referenced as: Mart Mchael, Neuenschwander René (2006), Case study 1.2E: Track Mantenance Costs n Swtzerland, Annex to Delverable D 3 Margnal cost case studes for road and ral transport Delverable D 1, Informaton Requrements for Montorng Implementaton of Socal Margnal Cost Prcng. Funded by Sxth Framework Programme. ITS, Unversty of Leeds, Leeds, March October 2006 Verson No: 1.0 Authors: as above. 2

3 Track Mantenance Costs n Swtzerland Fnal report, Verson No October 2006 ECOPLAN Economc Research and Polcy Consultancy CH Berne, Thunstrasse 22 CH Altdorf, PO box nfo@ecoplan.ch

4 Report Informaton Recommended ctaton Author: Mchael Mart, Neuenschwander René (Ecoplan) Ttle: Track Mantenance Costs n Swtzerland, Case study 1.2E: Track Mantenance Costs n Swtzerland, Annex to Delverable D 3 Margnal cost case studes for road and ral transport Delverable D 1, Informaton Requrements for Montorng Implementaton of Socal Margnal Cost Prcng. Funded by Sxth Framework Programme. Ecoplan, Berne, Commssoned by: European Unon Place: Berne Year: 2006 Project Team Ecoplan Mchael Mart René Neuenschwander The responsblty for the content and the opnons expressed n ths report les solely wth the authors. Ecoplan Economc Research and Polcy Consultancy Thunstrasse 22 CH Berne Tel Fax bern@ecoplan.ch PO box CH Altdorf Tel Fax altdorf@ecoplan.ch

5 Table of Contents Abstract Introducton Descrpton of data Infrastructure data Traffc data Cost data Model specfcaton and estmaton results General model specfcaton Detaled model specfcaton Estmaton results Margnal costs Conclusons Appendx: Model wth alternatve specfcaton...17 References

6 Abstract To mprove the effcency of ral transport s an mportant objectve of both the Swss and the European transport polcy. Track access and nfrastructure chargng are mportant nstruments of such a polcy. Wth respect to these nstruments, Swtzerland pursues a polcy that follows closely the European approach of lberalzaton. In ths paper, we provde a bass for a more margnal cost orented ralway nfrastructure chargng scheme. The objectve s to estmate margnal costs of ralway mantenance for Swtzerland. The hypothess s that margnal costs of ralway mantenance are a functon of dfferent ndependent explanatory varables. Such explanatory varables are output measures lke gross-ton-klometers as well as techncal and spatal features of the ralway system. The methodology used to estmate margnal costs of ralway mantenance refers to former European studes. We estmate cost functons usng econometrc methods and calculate margnal costs. A database coverng 371 track sectons from 2003 to 2005 (1113 observatons) has been avalable for ths study. Acknowledgements: Data provsons by SBB and the dscussons wth ts experts are gratefully acknowledged. Ths paper has benefted from comments from Mats Andersson and Leopold Sögner. 2

7 1 Introducton To mprove the effcency of ral transport s an mportant objectve of both the Swss and the European transport polcy. Track access and nfrastructure chargng are mportant nstruments of such a polcy. Wth respect to these nstruments, Swtzerland pursues a polcy that follows closely the European approach of lberalzaton. 1 The Swss ralways reform ntroduced n 1999 the separaton of nfrastructure and transport sectors n terms of accountng and organsaton. Access to the ralway network was adapted to the EU-drectve 91/440. There s open access for Swss ralway undertakngs n freght traffc, for passenger traffc a concesson for the conveyance of passengers s requred. In freght traffc also foregn companes have open access to the Swss ralway system on the bass of recprocty. As there are no dscounts on quantty or other non-lnear tarffs the reform ncreased competton substantally, especally n the market of transalpne freght traffc where several new ralway companes entered the market. Accordng to Art. 9b para. 3 of the Swss Ralway Act (Esenbahngesetz) the nfrastructure manager has the basc rght to charge the access. The charge has to be non-dscrmnatory and t can take nto account dfferent nfrastructure costs (e.g. caused by topography), the envronmental mpact of vehcles as well as the characterstcs of demand. The Swss Federal Councl determnes the basc prncples of chargng and defnes the rules of publcaton. These detals are subject of the Netzzugangsverordnung (Swss Order of Network Access). Bascally, for regular and lcensed passenger traffc the charge conssts of the standardzed margnal costs and a part of the revenues of traffc servces (contrbuton margn). Today the tran path prce n Swtzerland conssts of several components. 2 Mnmum prce : Mantenance ( CHF/Gtkm 3 ), tran operaton servce (0.4 CHF/tran- km), purchase of energy ( CHF/Gtkm) and supplements for nodes (bg nodes: 5 CHF, small nodes: 3 CHF). Contrbuton margn: Long-dstance passenger traffc (4% share of revenue), regonal passenger traffc (14% share of revenue), goods traffc ( CHF/Nettkm) 4 We can conclude that tran path prces n Swtzerland correspond to some knd of calculated average costs respectvely a standardzed level of margnal costs ( Normgrenzkosten ). The equvalence to margnal costs s qute rudmental as there s no reflecton of the specfc costs of axle load, qualty of rollng stock ( track frendlness ) or speed. There are also no scarcty charges for congested lnes, no peak load charges and the qualty of a tran path s not taken The European ralway polcy has been descrbed n many places, see e.g. Commsson of the European Communtes (2001), Whte Paper: European Transport Polcy for 2010: Tme to Decde, COM (2001)370, Brussels. Due to subsdes (wth the poltcal am of modal shft) the charges for freght traffc are temporarly reduced. Gtkm means gross-ton-klometers. Ths s for the ralway network of SBB (Schwezersche Bundesbahnen), on the one of BLS t s CHF/Gtkm. 3

8 nto account. Overall, the nfrastructure chargng scheme offers only small ncentves for a more effcent use of the ralway system. In ths paper, we try to provde a bass for a more margnal cost orented ralway nfrastructure chargng scheme. The objectve s to estmate margnal costs of ralway mantenance for Swtzerland. The hypothess s that margnal costs of ralway mantenance (as well as those for operaton and renewal) are a functon of dfferent ndependent explanatory varables. Such explanatory varables are output measures lke Gtkm or the number of trans or axles as well as techncal and spatal features of the ralway system. The methodology used to estmate margnal costs of ralway mantenance refers to former European studes, whch focus on margnal costs and use mcro-level data such as Johansson and Nlsson (2004), Tervonen and Idström (2004), Gaudry and Qunet (2003), Munduch et al. (2005), and Andersson (2006). The outlne s as follows. In secton 2, we descrbe the collected data used for our econometrc estmatons. Secton 3 summarses the model specfcaton and presents the estmaton results. In secton 4 we derve results wth respect to margnal costs whle secton 5 covers our conclusons. 2 Descrpton of data The frst project step concerned the data base. To begn wth, data avalablty was dscussed wth experts from SBB (Schwezersche Bundesbahnen), the natonal ralway company of Swtzerland. Responsble persons from SBB have assured ther nterest n the study and ther wllngness to provde the data needed. However, the dscusson also showed that t would be a demandng task to generate a consstent data set because the data had to be taken from dfferent sources. In collaboraton wth SBB we managed to provde a unque data set for Swtzerland by mergng three dfferent data sources 5 nto one data set. The data used s based on the whole ralway network of Swtzerland ncludng all man lnes. Ths network can be dvded n almost 500 sectons. Most of these sectons are mantaned by SBB, some by other lcensed ralway companes. For every secton a record of data was gathered for the years 2003, 2004 and Ths record contans: Infrastructure data Traffc data Cost data The data set we got from SBB ncludes a vast amount of varables, especally for nfrastructure and cost data. In the followng sectons we descrbe the dfferent types of data. 5 Data sources provded by SBB: DfA (Datenbank der festen Anlagen) for nfrastructure data, PANDA for traffc data, and cost data. 4

9 2.1 Infrastructure data The Swss natonal ral network comprses approxmately 4,900 klometers of track, of whch 56 percent s double track. The SBB defnes a lst of track sectons (almost 500 sectons n operaton) that we wll make use of. A track secton s a part of the network, whch vares n length from 0.39 to 81.6 klometers. A secton s not strctly homogeneous, that s between ts endponts, t can vary n terms of ral and sleeper types, ballast, curvature, slope etc. For ths study, a vast amount of nfrastructure data was avalable per secton. Not all of the track sectons n the data base are possble to be used. Some defned track sectons are mantaned by other countres (16 sectons n border areas, mantaned by foregn ralway companes as DB, ÖBB, FS and SNCF) others by other ralway companes (58 sectons). 36 track sectons are marshallng yards where ether no traffc data or no cost data are avalable. Fnally, 18 track sectons have to be dropped because they have been redefned n the perod or have only been n operaton snce Ths results n 371 observatons (track sectons) per year to analyze wth complete nformaton. The nfrastructure data was provded by SBB through ther data system DfA (Datenbank der festen Anlagen). The DfA shows the current status of the network and contans all exstng physcal nformaton about the ralway network n Swtzerland. The DfA was bult up n recent years. Therefore, experts from SBB strongly recommend usng the 2005 state of the DfA for all observatons. As n our sample we dd not nclude sectons wth new tracks constructed between 2003 and 2005 ths recommendatons makes sense and we follow t. Table 1: Infrastructure data used (371 track sectons, data for 2003, 2004 and 2005) Varable Measure Observatons Mean Standard devaton Track length m Track dstance m Swtches m Brdges m Tunnels m Level crossngs No Radus 500 m Slope 1 m Slope 2 m Nose / fre protecton m Platform edge m Rals aged > 25 % Sleepers aged > 25 % Maxmum speed km Swtches 1 No Swtches 2 No Swtches 3 No Shafts No Mn Max 5

10 Table 1 shows mean values for those nfrastructure varables of the years that wll be used n the econometrc estmaton. The varables are all expected to affect mantenance and renewal costs. An analyss of correlaton coeffcents between varables shows that there are only lttle nterdependences. We assume to have no serous multcollnearty problems n the model estmaton. However, we added dummy varables to the data to cover for dfferent years and spatal dfferences, ncludng dummy varables for dfferent dstrcts. Fgure 1 shows the locatons of the headquarters of the 23 dstrcts of SBB. The south of Swtzerland s less densely populated than the rest of Swtzerland; therefore, as can be seen n fgure 1, the dstrcts n the south are larger than n the rest of the country. The south-east of Swtzerland s not run by the natonal ralway company SBB, but by another ralway company ( Rhätsche Bahn ). Fgure 1: Locatons of the 23 dstrcts (regons) of the SBB Legend: The red dots show the locatons of the headquarters of the 23 dstrcts of SBB. A specal remark wth respect to tran statons: Larger tran statons such as Zurch, Geneva or Berne and statons whch are ralway junctons such as Brg are lsted as sectons. Snce the ralway nfrastructure s very complex n large statons we take these sectons nto consderaton as ordnary sectons. However, for estmaton purposes we wll defne a dummy varable for sectons whch nclude only statons. An addtonal varable the platform edge n meters takes nto account that small statons are ncluded n other sectons. 6

11 2.2 Traffc data Traffc data were provded by SBB through ther traffc data system PANDA. The system gves average daly data on number of trans, axle load and gross-tons per track, as well as yearly data on tran klometers, axle load klometers and gross-ton-klometers per track for the man lnes. Analogous to the nfrastructure data, table 2 gves the key values for traffc varables of the years Table 2 contans gross ton-klometers (Gtkm) and tran klometers (tran km) per secton. Table 2: Traffc data used (371 track sectons, data for 2003, 2004 and 2005) Varable Measure Obs. Mean Std. Dev. Mn Max Total tran km km ' '720 3'308 2'522'007 Total tran km km ' '480 4'306 2'794'162 Total tran km km ' '415 3'646 2'654'653 Total tran km km ' '691 3'308 2'794'162 Total axle km km '155'267 19'282'200 87' '417'983 Total axle km km '457'683 21'956'408 68' '837'933 Total axle km km '406'014 20'912'474 74' '182'425 Total axle km km '006'321 20'754'971 68' '837'933 Total Gtkm km ' '871 1'037 2'040'742 Total Gtkm-km km ' ' '236'565 Total Gtkm km ' ' '118'660 Total Gtkm km ' ' '236' Cost data SBB provded us wth very detaled cost data per secton for whch SBB s responsble for mantenance. The cost data contan nformaton such as costs on: Operaton mantenance (e.g. cleanng, snow and ce removal) Track mantenance Forestry Engneerng Sgnal tower mantenance Wre mantenance Electronc nstallaton Moreover, wthn these dfferent cost categores SBB separates between short-run mantenance costs ( Contractng A ) that arse yearly and long-run costs whch arse perodcally and have the characterstcs of renewal costs ( Contractng B ). Due to the fact that the excellent data base s only avalable snce 2003, the estmaton of renewal costs s based on a relatvely short tme perod of three years. Therefore, we do not estmate renewal costs by them- 7

12 selves but n combnaton wth mantenance costs. Smlar to Andersson (2006), we observe that renewal costs per track secton can vary sgnfcantly between years. Accordng to cost experts of SBB there are three reasonable cost categores to estmate: Model type 1: Yearly arsng mantenance costs that consder all expendtures for Contractng A (man model) Model type 2: Yearly arsng mantenance costs only for operaton and track mantenance Model type 3: Mantenance and renewal costs (all expendtures for Contractng A and Contractng B The descrptve statstcs of the cost varables used n the three model structures are shown n table 3. Cost data are avalable n Swss Francs (CHF), so the econometrc analyss was performed wth the orgnal data. The man results concernng margnal and average costs are presented n EUR for better comparson. Table 3: Cost data used, n CHF (371 track sectons, data for 2003, 2004 and 2005) Varable Obs. Mean Std.Dev. Mn Max Mantenance costs (Contractng A) ' ' '025 17'929 5'193'119 Mantenance costs (Contractng A) ' ' '118 20'759 5'027'611 Mantenance costs (Contractng A) ' ' '829 15'345 4'619'208 Mantenance costs (Contractng A) ' ' '930 15'345 5'193'119 Only operaton and track mantenance costs Only operaton and track mantenance costs Only operaton and track mantenance costs Only operaton and track mantenance costs Mantenance and renewal costs (Contractng A+B) Mantenance and renewal costs (Contractng A+B) Mantenance and renewal costs (Contractng A+B) Mantenance and renewal costs (Contractng A+B) ' ' '139 3'724 2'225'127 1' ' '076 3'449 1'810'884 1' ' '071 3'344 1'959'106 1' ' '368 3'344 2'225'128 1' ' '668 17'929 7'489'656 1' ' '387 20'759 8'361'440 1' ' '163 15'345 7'876'671 1' ' '176 15'345 8'361'440 8

13 3 Model specfcaton and estmaton results 3.1 General model specfcaton In ths secton we present the model selected to estmate margnal costs of nfrastructure mantenance. We refer to former emprcal work such as Johansson and Nlsson (2004), Tervonen and Idström (2004), Munduch et al. (2005) and Andersson (2006). An mportant aspect of former emprcal work s the applcaton of a pooled ordnary least squares technque. Pooled OLS treats all observatons as a cross secton sample assumng that there s no systematc varaton over the three years. But n order to control for a tme effect we ntroduce dummy varables for dfferent years. 6 Our general model specfcaton s: C = α + β + ε t x t t Whle ndcates track secton and t tme perod, C t denotes the costs, and x t the vector of explanatory varables. The letters α and β are the coeffcents to estmate, whle ε t s the error term of each secton n tme t. Each model contans a specfed dependent cost varable, and a number of explanatory varables (nfrastructure varables and a traffc varable) that measure the output from the track secton. Our next step concerns the functonal form of the cost functon. Most common functonal forms n lterature are lnear, log-lnear and Translog specfcatons. Whle Johansson and Nlsson (2004) as well as Tervonen and Idström (2004) used a Translog specfcaton, Munduch et al. (2005) and Andersson (2006) appled a classcal log-lnear specfcaton. In order to compare our results to most recent emprcal work we are usng a log-lnear specfcaton as well Detaled model specfcaton The detaled specfcaton of the model depends on the expectaton about cost drvers. Ralway experts assume that mantenance costs are more lkely to be drven by gross tons than by other varables. In accordance wth Johansson and Nlsson we splt the varable gross-tonklometers nto two separate varables: gross tons per track klometer and track length. Ths separaton solates the costs drven by gross tons from traffc length effects. For the other explanatory varables we rely on two sources: 6 7 Applyng dummy varables for dfferent years, we correct for the tme effect. Regardng to the tme dmenson the model s smlar to a fxed-effects model. A good presentaton of a log-lnear specfcaton s gven by Munduch et al. (2005). 9

14 Frst, we screened former emprcal work to fnd approprate varables. Other studes ncluded varables such as tunnel meters, brdge metres, swtch meters, curvature, slopes, the age of rals and sleepers etc. Second, we made use of our large set of varables n nfrastructure. We checked for dfferent varables such as level crossngs, nose/fre protecton, retanng walls, dfferent sgnals, swtches, shafts etc. Not all examned varables entered the fnal estmatons. The varables whch were appled n the fnal estmatons are shown n table 3. Table 4: Informaton per track secton, explanatory varables Name of varable Descrpton Track length Dstance of tracks Gross tons Number of gross tons (logarthm) Swtches Length of swtches n meters (logarthm) Brdges Length of brdges n meters (logarthm) Tunnels Tunnel meters (as percentage of track length) Level crossngs Number of level crossngs Radus 500 Bends wth radus < 500 n meters (as percentage of track length) Slope 1 Slopes of 10 to 20 n meters (as percentage of track length) Slope 2 Slopes steeper than 20 n meters (as percentage of track length) Nose / fre protecton Nose / fre protecton n meters (as percentage of track length) Platform edge platform edge n meters (as percentage of track length) Rals_age_25 Rals older than 25 years (as percentage of track length) Sleepers_age_25 Sleepers older than 25 years (as percentage of track length) Maxmum speed Maxmum speed per track (logarthm) Swtches 1 Number of mechanc swtch machnes, type 1 Swtches 2 Number of mechanc swtch machnes, type 2 Swtches 3 Number of electro-mechanc swtch machnes Shafts Number of shafts D_staton Dummy for statons (1 for staton, 0 for other sectons) D_03 Dummy for the year 2003 (1 for 2003, 0 for other years) D_04 Dummy for the year 2004 (1 for 2004, 0 for other years) Dummes per regon Dfferent dummes for 23 dstrcts 3.3 Estmaton results The estmaton results of the three model types are presented n table 5. The estmatons are carred out wth Intercooled STATA 9.1. In order to avod neffcent estmatons due to heteroskedastcty, we apply a robust estmator. For each model two estmatons were calcu- 10

15 lated, one wth regonal dummes and one wthout regonal dummes. As mentoned n chapter 2, the estmatons have been run over 1113 observatons, whch are 371 observatons per year for three years. Generally, the explanaton power of all three models s relatvely hgh. Adjusted R 2 -values are between 0.61 and Smlarly to the experence of Andersson (2006) for Sweden we do not face any serous problems addng renewal costs to mantenance costs (model type 3). We fnd the smallest explanatory power n model type 2. Addng more varables or replacng varables such as gross tons by the number of trans as suggested by Andersson (2006) does not lead to a hgher adjusted R 2 -value. The ncluson of regonal dummes (dstrcts) has a relatvely small effect on the explanatory power of the estmated models. We consder ths as a sgn that the dstrcts use comparable technologes for track mantenance. An mportant fndng s that the algebrac sgns of the varables ncluded do not change between the dfferent model types. Takng a closer look at our man model type 1 (estmatons (1) and (2)) we see that most coeffcents (not the regonal dummes) are sgnfcant at the 1 percent level and mostly have expected sgns. Not surprsngly, the estmatons (1) and (2) show hghly sgnfcant and postve values for track length, gross tons, swtch meters, brdge meters, level crossngs, curvature, platform edge meters, steep slopes, nose/fre protecton, and for the staton dummy varable. Addtonal nfrastructure varables such as the number of swtches and shafts are also hghly sgnfcant (mostly sgnfcant at the 5% level). At frst glance, the negatve sgn for tunnel meters s rrtatng (sgnfcant at the 5% level). However, the natural protecton of tunnels from snow and ran mght reduce wear and tear and therefore the amount of mantenance costs. Qute surprsng s the negatve sgn for maxmum speed (whch however s narrowly not sgnfcant at the 10% level). We assume that a maxmum speed can be acheved on the track sectons wth less curvature and less slope. A smple correlaton analyss confrms ths assumpton: the maxmum speed per track s negatvely correlated wth curvature (-0.52) and wth steep slopes (-0.36). As already mentoned, the ncluson of regonal dummes has lttle mpact on the explanatory power of the model. However, t has nterestng effects for two varables: estmatng the regonal dummes leads to a sgnfcant negatve varable for tunnel meters, whle moderate slopes become nsgnfcant. Fnally, the dummy varable for 2003 s sgnfcant at 5% level. Therefore, costs were sgnfcantly hgher n 2003 than n the followng years. Lookng at the results of model type 2 whch are gven by estmatons (3) and (4) we see that the explanatory power of the model s clearly less hgh than for model type 1. Usng other model specfcatons does not mprove the explanatory power of the model sgnfcantly. Compared to the results of model type 1 the most nterestng changes n varables are to be 11

16 found n curvature. Curvature seems to be less mportant for operaton and track mantenance costs. Consderng the results of model type 3 we fnd good values for the adjusted R 2. The explanatory power of the model s almost as good as for model type 1. Analogous to the results of model type 1, the estmatons (5) and (6) show hghly sgnfcant and postve values for track length, gross tons, swtch meters, brdge meters, level crossngs, curvature, platform edge meters, steep slopes, nose/fre protecton, and for the staton dummy varable. In contrast to the results of model type 1 we fnd negatve values for the dummy varable for the years 2003 and 2004, respectvely. The dummy varable for 2004 s sgnfcant at the 5% level. Combned wth the results of the estmatons (5) and (6) ths observaton leads to the concluson that n 2005 more money was spent for renewal, whle n 2003 more money was spent for mantenance. In summary, the most relable model type s model type 1. We fnd here the hghest values for the adjusted R 2. Less explanatory power can be found for model type 2 estmatons. Ths result s nvarant when we change the model specfcaton (addtonal or other explanatory varables). Fnally, the results of model type 3 estmatons are qute good. However, a longer tme span than three years would be preferable for analysng renewal costs. 12

17 Table 5: Estmatons results Model Model type 1 Model type 2 Model type 3 Dependent varable All mantenance costs (Contractng A) Only operaton and track mantenance costs Mantenance and renewal costs (Contractng A+B) Estmatons (1) (2) (3) (4) (5) (6) Observatons Explanatory varables Constant Track length Gross tons Swtches Brdges Tunnels Level crossngs Radus 500 Slope 1 Slope 2 Nose / fre protecton Platform edge Rals_age_25 Sleepers_age_25 Maxmum speed Swtch machnes 1 Swtch machnes 2 Swtch machnes 3 Shafts D_tran staton D_03 D_ *** (16.43) 0.605*** (23.10) 0.200*** (9.14) 0.034*** (8.98) 0.039*** (5.46) (-0.99) 0.007*** (3.60) 0.334*** (3.80) 0.097* (1.81) 0.248*** (3.09) ** (-2.28) 0.458*** (4.91) (1.15) (-0.14) (-1.57) 0.001** (2.54) 0.002*** (5.22) 0.001*** (4.30) 0.000*** (3.89) 0.290*** (6.25) 0.068** (2.32) (0.41) 8.158*** (16.26) 0.632*** (23.09) 0.195*** (8.62) 0.034*** (8.79) 0.036*** (4.68) ** (-2.03) 0.008*** (3.50) 0.325*** (3.48) (1.33) 0.231*** (2.76) ** (-1.98) 0.531*** (5.51) (1.41) (-0.76) (-1.58) 0.002*** (3.40) 0.002*** (4.97) 0.001*** (2.97) 0.000*** (3.19) 0.329*** (7.08) 0.067** (2.37) (0.42) 4.666*** (7.77) 0.586*** (14.63) 0.285*** (5.64) 0.047*** (7.11) 0.048*** (4.50) * (-1.78) 0.007*** (3.14) (0.75) 0.235** (1.97) 0.379*** (3.42) (-0.77) 0.548*** (4.09) (1.36) (-0.97) (0.84) 0.003*** (3.10) 0.003*** (5.76) 0.002*** (4.32) 0.000** (2.04) 0.286* (1.86) (0.49) (1.26) 4.695*** (7.17) 0.622*** (13.69) 0.270*** (5.91) 0.046*** (7.34) 0.045*** (4.40) ** (-2.43) 0.007*** (3.00) (1.31) 0.250** (2.05) 0.260* (1.79) (-0.49) 0.631*** (4.29) (1.19) (-1.02) (0.63) 0.002** (2.51) 0.003*** (4.91) 0.001*** (3.39) 0.000*** (3.30) 0.320** (2.31) (0.48) (1.32) 7.243*** (11.98) 0.638*** (20.67) 0.265*** (9.73) 0.038*** (7.56) 0.048*** (5.38) (0.85) 0.009*** (4.15) 0.489*** (4.35) (1.49) 0.212** (2.19) *** (-3.09) 0.434*** (4.57) (0.25) (0.55) (-0.95) (1.02) 0.003*** (4.65) 0.001** (2.24) 0.000** (2.47) 0.243*** (4.36) (-1.54) ** (-2.20) 7.055*** (11.61) 0.682*** (20.85) 0.267*** (9.66) 0.035*** (7.00) 0.045*** (4.48) (-0.04) 0.010*** (4.48) 0.476*** (4.01) (0.91) 0.215** (2.10) ** (-2.03) 0.512*** (5.37) (0.88) (-0.20) (-0.83) 0.002** (2.53) 0.002*** (4.52) (1.38) 0.000* (1.77) 0.308*** (5.62) (-1.58) ** (-2.26) 13

18 Regonal dummes (Dstrcts) Gubasco Brugg Bel Bern Brg Basel Bülach Delémont Frbourg Arth-Goldau Genf Lausanne Luzern Neuenburg Olten Spretenbach Rapperswl Sargans St.Gallen St.Maurce Wnterthur Zürch (1.60) (0.10) (0.96) (-0.13) (-0.34) (0.95) (0.72) (1.25) (-0.02) (-1.38) (1.05) (1.48) (-0.92) (0.32) (-0.69) 0.233** (2.11) (-0.58) (-1.15) (0.15) (1.23) (0.45) 0.239** (2.28) (0.83) (0.88) 0.402*** (2.59) 0.390*** (2.63) (0.39) (-1.29) (1.15) (0.41) (0.14) (-1.09) 0.341* (1.80) 0.383** (2.51) (0.52) (0.72) (1.39) 0.449** (2.57) (0.27) (0.43) (0.31) (0.51) (0.07) 0.437*** (2.75) (1.20) (-0.29) (1.63) (-0.48) (-1.06) (1.12) (1.07) 0.217* (1.68) (0.54) (-0.63) (1.04) 0.227* (1.86) (-0.97) (0.71) (-1.55) (1.57) (0.35) (-0.95) (0.42) (0.79) (1.49) 0.356*** (2.81) Adjusted R SE Remarks: ***/**/* denote sgnfcance at the 1-/5-/10-percent-level. One regonal dummy (the dummy for the dstrct Ascona) has been dropped; otherwse the model would have been over specfed. T-values are gven n parentheses. The sgn ndcates that the varable has not been ncluded n the model estmated. 14

19 4 Margnal costs Accordng to Johansson and Nlsson (2004) estmated elastctes can be used to derve estmates of margnal costs for each track unt. 8 Snce we are nterested n the cost effects of an addtonal gross-ton-klometer (Gtkm) per secton, we calculate margnal costs (MC) wth respect to Gtkm. The formula s gven by: C MC = Gtkm C 1 = Gton km We assume that condtonal to the specfc data base the dstance represents a constant n track margnal costs. For track, ths leads to MC Cˆ = Gtkm Cˆ = Gton 1 km Cˆ = Gton Gton Cˆ Cˆ Gtkm Ĉ denotes the ftted value of C and s calculated by C ˆ = exp( ˆ α + ˆ β ˆ σ 2 xt ).The expresson n parenthess represents the elastcty of the mantenance costs wth respect to gross-tons. To calculate the average margnal cost, we weght the margnal costs per secton, usng the number of gross-ton-klometers on each secton unt as a weght. Table 6 shows the margnal costs (MC) and the average costs (AC) for the whole network for our sx estmated models. Average costs are computed by usng the ftted values dvded per Gtkm. The margnal mantenance costs wth respect to Gtkm for our man model type 1 are and EUR respectvely. Generally the value s lower when we nclude regonal dummes. Average mantenance costs per Gtkm are approxmately EUR per Gtkm. The cost recovery for model type 1 s approxmately 20% of total mantenance costs. Ths confrms that prcng at margnal costs does not recover total costs, but s motvated by effcency thoughts. Smlar results have been found for Sweden (Andersson, 2006). The lowest values for margnal and average costs are clearly found for model type 2 where only operaton mantenance and track mantenance costs have been consdered. The hghest cost data are provded by model type 3 where renewal and mantenance costs have been ncluded. In model types 2 and 3 the cost recovery les between 26 and 29%. 8 For a good explanaton see also Munduch et al. (2005). 15

20 Table 6: Margnal costs and average costs per Gtkm, n EUR Model type Model type 1 Model type 2 Model type 3 Estmaton (1) (2) (3) (4) (5) (6) Margnal costs (MC) Average costs (AC) Cost recovery 20.0% 19.5% 28.5% 27.0% 26.5% 26.7% An nternatonal comparson shows that our results for margnal costs are n lne wth the estmatons n other countres. Accordng to the Austran study of Munduch et al. (2005) the margnal costs n Austra are EUR for all lnes and slghtly lower for man lnes ( EUR). The margnal costs for Sweden, found by Andersson (2006), are approxmately between and EUR, whch corresponds to our results of model type 1. To verfy our results, we used an alternatve model specfcaton, whch was appled by Munduch et al. (2005). The specfcaton ncludes varables whch have a drect mpact on the coeffcent of gross tons. The applcaton of ths specfcaton for Swtzerland leads to slghtly hgher margnal costs (see the estmaton results n the appendx). 5 Conclusons The good data base provded by the natonal ralway company of Swtzerland, SBB, gave us the chance to estmate margnal costs for Swtzerland. The number of track sectons and the avalablty of data for the tme span led to a reasonable number of observatons, at least for yearly arsng mantenance costs. Our results confrm the fndngs of natonal studes for Sweden, Austra and Fnland and may serve to mplement a margnal cost orented charge per gross-ton-klometer for yearly arsng mantenance costs. Lookng at renewal costs, addtonal years for estmaton are needed n order to get relable results. Our estmatons suggest that margnal renewal costs are approxmately of the same sze as margnal mantenance costs and therefore sgnfcantly rase nfrastructure charges that are based on margnal costs. But ths s only a provsonal result whch has to be confrmed by usng data of a broader tme span. There s an ongong mprovement of the SBB data base, such that n further research we should be able to extend the data base over more years. What do the above results sgnfy for Swss track nfrastructure charges? Our estmates show that mean margnal mantenance costs are about EUR ( CHF respectvely) per Gtkm. If renewal costs are addtonally ncluded, mean margnal costs would be around 16

21 EUR ( CHF respectvely) per Gtkm. Ths s sgnfcantly lower than today s mantenance charge n Swtzerland of CHF/Gtkm. Furthermore, our results show that mean average mantenance costs are EUR/Gtkm ( CHF/Gtkm) whch s around fve tmes hgher than mean margnal costs. As expected, margnal cost prcng would mply drastcally lower charges compared to charges based average costs. Today s mantenance charge n Swtzerland les n between mean margnal costs and mean average costs. It has to be added that n Swtzerland mantenance charges for freght traffc are temporarly reduced, due to subsdes (wth the poltcal am of modal shft). For combned traffc mantenance charges add up to CHF/Gtkm ( EUR/Gtkm). Ths s stll more than estmated mean margnal mantenance costs, but less than estmated mean margnal mantenance and renewal costs together. The estmated mean margnal costs can be spatally dfferentated accordng to dfferent crtera, e.g. alpne regons, flat or urban areas. Our estmates show no sgnfcant dfferences of margnal mantenance costs per Gtkm between flat areas and alpne regons. 9 But for Swss urban areas margnal mantenance costs are estmated up to EUR/Gtkm whch s almost twce as much as mean margnal mantenance costs. Ths gves a frst estmate for spatally dfferentated nfrastructure charges. 6 Appendx: Model wth alternatve specfcaton In ths appendx we present the results of an alternatve model specfcaton followng the approach of Munduch et al. (2005). Table 7 shows the correspondng estmaton results. In ths specfcaton we use dfferent explanatory varables whch partly are a multplcaton of a certan nfrastructure varable e.g. such as slope metres wth gross tons. The hypothess behnd ths specfcaton s that the effect of topography (slopes, tunnels) or densty of ralway statons (platform edge) on margnal mantenance costs also depends on how the track s used e.g. the number and the weght of trans (gross tons) per year. In ths case, some of the explanatory power gven to slope, tunnel or platform edge metres n the model specfcaton used before (see table 5) s actually due to the number of gross tons. By recalculatng the margnal mantenance costs wth respect to Gtkm we can fgure out whether ths alternatve model specfcaton has an mpact on margnal and average costs. Comparng the results of table 7 wth estmaton (2) n table 5 we fnd that applyng the Mun- 9 If the calculaton s done on the bass of tran km nstead of Gtkm margnal mantenance costs are sgnfcantly hgher n alpne regons than n flat areas by a factor of 1.43, because average tran weght s hgher n alpne regons. 17

22 duch et al. specfcaton for Swtzerland leads to only slghtly hgher margnal mantenance costs. Therefore, we can conclude that n the case of Swtzerland the model specfcaton n the man report reflects almost fully the margnal mantenance costs of an addtonal gross ton. Table 7: Estmatons results (ncludng varables wth a drect mpact on gross tons) Dependent varable All mantenance costs (Contractng A) Observatons 1113 Explanatory varables Constant Track length Gross tons Radus 500 Swtches Platform edge * Gross tons Slope 2 * Gross tons Tunnels Tunnels * Gross tons 7.639*** (26.19) 0.702*** (32.52) 0.196*** (11.27) 0.751*** (7.36) 0.048*** (10.35) 0.049*** (6.61) 0.034*** (5.49) *** (-4.39) Adjusted R SE 0.45 Margnal costs (MC) Average costs (AC) Cost recovery 20.1% Remarks: ***/**/* denote sgnfcance at the 1-/5-/10-percent-level. T-values are gven n parentheses. 18

23 References Andersson Mats (2006) Margnal Cost Prcng of Ralway Infrastructure Operaton, Mantenance and Renewal n Sweden: From Polcy to Practce Through Exstng Data. Transportaton Research Record: Journal of the Transportaton Research Board, 1943, Commsson of the European Communtes (2001) Whte Paper: European Transport Polcy for 2010: Tme to Decde. COM (2001)370. Brussels. Gaudry M., Qunet E. (2003) Ral track wear-and-tear costs by traffc class n France. Paper presented at Frst conference on ralroad ndustry structure, competton and nvestment, Toulouse, France. Johansson P., Nlsson, J.-E. (2004) An economc analyss of track mantenance costs. Transport Polcy, Vol. 11, 2004, pp Munduch G., Pfster A., Sögner L. and Stassny A. (2005) An Econometrc Analyss of Mantenance Costs as a Bass for Infrastructure Charges for the Austran Ralway System. Mmeo. Venna. Tervonen J., Idström T. (2004) Margnal Ral Infrastructure Costs n Fnland Fnnsh Ral Admnstraton, Publcaton A 6/2004, Helsnk, Fnland. 19