THE SWEDISH MODEL SYSTEM FOR GOODS TRANSPORT - SAMGODS by dfs18652

VIEWS: 226 PAGES: 50

									THE SWEDISH MODEL SYSTEM FOR GOODS TRANSPORT – SAMGODS

A brief introductory overview

SAMPLAN Rapport 2001:1

THE SWEDISH MODEL SYSTEM FOR GOODS TRANSPORT – SAMGODS

A brief introductory overview

SAMPLAN Rapport 2001:1

3

Preface

This purpose of this report is to provide a brief but reasonably comprehensive overview of the Swedish national goods model system SAMGODS. The system in the form presented here has been used during 1999 and 2000 for policy analyses and in the strategic infrastructure planning process for the period 2002-2011. The report will hopefully increase the understanding of the pros and cons of the modelling approach, and thereby also contribute to the transparency of the modelling exercises undertaken in the processes of planning and policy analysis. The intention is also that this report, together with other more detailed background reports, will help to provide a solid basis for further development. Referring to the goods model system as ”the national system” is perhaps slightly misleading, since it might lead one to think of a highly formalised and consistent model structure, which has been conceived of as an entity at one time. The fact is, however, that the model system sooner could be looked upon as a rather loosely fitted set of system modules, which in some instances have originally been developed for quite different purposes. Of course this structure will also cause some consistency problems, which will have to be dealt with in practical applications. The reader who takes an interest in specific modules of the system is recommended to look into the detailed technical reports, which are available for all parts of the system. These reports are either in Swedish, Danish or English and will also be made available and updated on SIKA´s web site (www.sikainstitute.se). Henrik Swahn now working as an independent consultant, but who has held a leading position within SIKA for many years of development and application work related to the model, has been commissioned by SIKA to write this summary report. Stockholm, February 2001 Staffan Widlert

SAMPLAN Rapport 2001:1

4

Detailed technical reports
The following technical reports can be found as pdf-files at SIKA’s web site www.sika-institute.se. • • • • • • • • • Demand models (Swedish/Danish) Demand matrices (Swedish) Transport network models (Swedish) Evaluation models (Swedish) Calibration and validation (Swedish) Value of time and transport quality (Swedish/Danish) Analysis of STAN-instability (Swedish) Air freight transport (English) Maritime freight transport (Swedish)

SAMPLAN Rapport 2001:1

5

Contents

1

SUMMARY............................................................................................................................. 7

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
2

On the needs for national goods transport models in Sweden ................... 7 The model system ...................................................................................... 8 The institutional context for the Swedish national goods transport model 8 Major application areas .............................................................................. 9 The transport demand models .................................................................... 9 The transport market model (network model implemented in STAN) .... 11 Evaluation models and tools .................................................................... 11 Calibration and validation........................................................................ 12 Background and purpose.......................................................................... 13 The modelling approach........................................................................... 14 Organization of the report ........................................................................ 15

INTRODUCTION ................................................................................................................ 13

2.1 2.2 2.3

3 APPLICATION AREAS AND THE INSTITUTIONAL SETTING OF THE NATIONAL MODEL SYSTEM ................................................................................................. 17

3.1 3.2 3.3
4 5

The application arenas.............................................................................. 17 Actors and arenas ..................................................................................... 19 Development and administration of the national goods model system.... 20

OVERVIEW OF THE MODEL SYSTEM AND ITS COMPONENTS......................... 21 THE DEMAND MODELS................................................................................................... 23

5.1 5.2 5.3 5.4
6

Introduction.............................................................................................. 23 The multi-sectoral system of models of Swedish economy (ISMOD) .... 24 Modelling interregional transport demand within Sweden (VTI/TPR) ... 28 The foreign trade model ........................................................................... 32 Introduction and purpose.......................................................................... 35 Brief development history........................................................................ 36 Basic functions and components of the transport market/network model36

THE TRANSPORT NETWORK MODEL ........................................................................ 35

6.1 6.2 6.3
7 8

EVALUATION TOOLS AND MODELS........................................................................... 43 APPLICATION EXAMPLES ............................................................................................. 45

8.1 8.2 8.3 8.4

Demand projections, scenarios and forecast ............................................ 45 Analyses of major infrastructure projects ................................................ 46 Corridor analyses ..................................................................................... 47 Demand effects of transport policy measures including taxes/charges ... 47

SAMPLAN Rapport 2001:1

6

SAMPLAN Rapport 2001:1

7

1 Summary
1.1 On the needs for national goods transport models in Sweden
Goods transport is a vital component of the production and distribution processes of modern societies. The separation of production and consumption as well as a more or less global division of labour, act as driving forces for a seemingly ever increasing demand of goods transport. From time to time the development of the production and distribution processes as well as the logistical systems have been such as to incur a growth rate of goods transport demand higher than the growth rate of the transported quantities of goods and also at a rate higher than the general GDP growth rate. Transport demand plays a crucial role in the development of the transport sector. Total demand development and changes of the demand structure initiate capacity changes (investment as well as divestment), and is also one of the major factors influencing transport policy decisions e.g. policies aiming at influencing the modal structure emerging from market decisions, the level of demand, the spatial distribution of transport activities, transport technology and the emissions of transport operations. Both transport policy implementation and capacity alterations of the transport systems take time, which is sometimes very considerable. A process of rational decision making therefore requires some foresight of future demand changes, and also some understanding of the underlying driving forces. The needs for forecasting demand changes is by no means confined to policy makers, and decision makers involved in the development of public infrastructure. Many market actors involved in strategic decisions which have long implementation lead times should have similar requirements. Demand information is therefore an essential component in the processes of planning and decision making about transport policy and infrastructural measures. An assessment of future demand changes is a key component of evaluation methods such as Cost Benefit Analysis (CBA), Strategic Environmental Assessment (SEA) , Environmental Impact Assessment (EIA), analyses of the incidence of policy measures, and other types of policy analysis. Some kind of structured approach is necessary to meet the demand information requirements of these planning and decision processes. A quantitative modelling approach is needed. The purpose of this paper is to outline the contents and structure of the model system which was developed to provide demand analysis capabilities for the Swedish national infrastructure planning process, and for national transport policy analysis.

SAMPLAN Rapport 2001:1

8

1.2 The model system
The problem of modelling the entire goods transport system of a whole nation is very complex. It is necessary to balance many needs against each other, and many compromises and simplifications will be inevitable. One has to recognise also, that alterations and amendments of the model will be necessary over time due to changes of the driving forces of demand as well as changing requirements on the model output due to shift of focus of the policy making processes. The Swedish national model system for goods transport could most correctly be described as a set of separate models, sometimes developed for entirely different purposes than the analysis of national goods transport demand, which have been tied together to enable them to interact in a reasonably effective and consistent way in the process of modelling Swedish goods transport demand. The system comprises two major subsystems namely the transport demand models, and the transport network/transport market model. A third model package, the evaluation models and tools, are associated with these models. The demand models generate spatially disaggregated demand matrixes which are also disaggregated into separate commodities. The spatial structure covers not only Swedish territory but also the European union as well as all other areas involved in Swedish foreign trade. The demand matrices are used as input to the transport network/market model, which allocates the flow matrices to the transport modes and networks according to a system cost minimising algorithm. The model system allows projections to be made of transport flows on transport modes and networks, analyses of transport chains, modelling of transport system effects of certain transport policy measures as well as the effects of changes to the infrastructures.

1.3 The institutional context for the Swedish national goods transport model
The national goods transport model is used within a few rather complex institutional frameworks. Perhaps the most important one is the system for 10year strategic planning of transport infrastructures, which includes all transport modes. Despite focusing on long term infrastructure issues, in the planning process attention also has to be paid to many transport policy matters such as transport related taxation, charges, traffic subsidies etc. The ambition is that the model system should be applicable for all modes, all planning levels (national, regional, international connections), and for as many stages of the process as possible. Additionally, the system should also provide a truly inter-modal demand analysis and be able to deliver consistent results for different levels of detail. Undoubtedly therefore, the complex institutional setting also causes complex and sometimes conflicting demands to be put on the model system. Other complex institutional frameworks where the model is used are the transport policy process and the process to support Swedish contributions to EU level policy analyses.
SAMPLAN Rapport 2001:1

9

Since there are many participating actors in the processes where the national goods models system is used, it is mandatory that the system is as transparent as possible. The Swedish national model system is continuously developed in co-operation between the members of a consortium of Swedish state agencies involved in infrastructure planning and development as well as transport research. Members of the consortium are the Swedish Institute for Transport and Communications Analysis (SIKA), the Swedish National Rail Administration, the Swedish National Road Administration, the Swedish Maritime Administration, the Swedish Civil Aviation Administration and the Swedish Communication Research Board.

1.4 Major application areas
The national model system for goods transport has multiple purposes. The focus is on the strategic long term (10 year) infrastructure planning which encompasses all modes and regions. The planning exercise therefore also has a process aspect, since many state agencies at the national and regional level interact in the process. The role of the system in this planning process is to provide demand forecasts, policy and project evaluation information, analyses of effects and consequences of alternative strategies. Other important application areas are analyses of strategic corridors and investment projects as well as structural analyses of the terminal systems of transport modes. Such analyses are either an integral part of the 10 year strategic planning process or carried out as entirely separate analyses for policy evaluation purposes. A third application area is analyses of general transport policy measures as well as policy measures related to specific infrastructures e.g. road, rail etc. The advent during the last decade of some substantial transport policy initiatives at the EU level, has created still another application area for the national model system.

1.5 The transport demand models
The goods transport demand models (”generation of origin/destination matrices”) comprises five separate model modules: • • • Multi-sectoral models of the Swedish economy (ISMOD) A model for regional disaggregation of sectoral employment (EARLY) linked to ISMOD A model for modelling inter - regional transport demand within Sweden (VTI/TPR)

SAMPLAN Rapport 2001:1

10 • • Model for regional forecasting and regionalising Swedish foreign trade Models for forecasting of implicit commodity value for aggregates of commodities

The multi-sectoral model ISMOD is an input-output model. A characteristic feature of the model is the endogenous generation of the capital stock, where productivity varies with capital vintage. The model is driven exogenously by world market commodity price forecasts as well as domestic public sector demand forecasts. Additionally two exogenous balancing restrictions are given namely total employment and the balance of current account. The model provides as output production, investment, intermediate commodity demand, employment as well as wage and profit levels for all sectors of the Swedish economy. Also, private consumption, export and import are solved for commodities and services specified in the model. The model is designed to produce projections of Swedish economic development for a medium term perspective (5-10 years). Within the goods transport demand system the output of the ISMOD model for production, export, import and consumption for each sector is used as input to the Early model, to the system for modelling interregional transport demand, and to the foreign trade model. The EARLY model is linked to the ISMOD model. Based on the ISMOD demand projections per sector total national employment changes are calculated per sector. These changes are further disaggregated per region in the Early model. The disaggregation is carried out in a number of steps. One important tool is the separation of sectors into “engines” and “wagons”, with the idea that some sectors (engines) could be viewed as primary employment generators, while employment in other sectors (wagons) could be looked upon as induced by employment changes in “engine” sectors. By the combination of historical data, “engine” sector growth and decline probabilities and the model of engine/wagons, employment changes per sector and region are calculated. The model for inter - regional transport demand within Sweden (VTI/TPR-model) uses as input data the output from the ISMOD multisectoral model, output on regional employment changes from the EARLY model, output from the foreign trade model system, and output from the model for forecasting implicit prices for the commodity aggregates used in the model system. The VTI/TPR models utilises an entropy algorithm to estimate forecast demand matrixes for the relevant commodity groups. Estimates of the present domestic inter-regional transport flow matrices based on available empirical information as well as the corresponding foreign trade matrices are used as à priori matrices in the entropy algorithm. The marginal conditions for each commodity are derived from the ISMOD-model output. The purpose of the foreign trade model system is to forecast trade (export and import) between Swedish municipal areas and numerous foreign trade regions disaggregated into commodity aggregates. The model system has two main subsystems namely the bilateral trade model subsystem and the subsystem modelling lower level regional trade flows. The model system comprises some 50 bilateral trade models for suitable trade areas (countries or groups of countries)
SAMPLAN Rapport 2001:1

11 and commodity aggregates. These models are of the gravitation type using inter alia GNP, population, and distance as independent variables. In principle the model system is intended to provide forecasts of inter-area trade patterns on the most disaggregated area level used in the transport market model (demand matrix level) as well as a suitable commodity disaggregation.

1.6 The transport market model (network model implemented in STAN)
The purpose of the network model is to represent the transport markets allocation of transport demand to combinations of transport modes and also to transport routes (paths). Transport demand matrices, which are output from the model for interregional transport demand, are input to the transport network model (disaggregated per product). The allocation to modes and routes is controlled by the cost conditions of the transport services production system, which are represented as functions and parameters of the network model. The network model is implemented within the framework of the STAN-model package, which uses a system cost minimising algorithm to solve the allocation problem. It is essential in the policy context where it is used, that the national model systems produces demand projections and also allows for policy analyses related to the transport network. Only the capability of dealing with the network level makes it useful for analyses of a range of issues related to transport infrastructure development (investment, divestment, system level changes of the general properties of subsystems of the infrastructure, corridors, missing strategic links etc) as well as for the analysis of transport policy issues such as pricing/taxation).

1.7 Evaluation models and tools
The purpose of the evaluation models and tools is to make use of information already available in the national model system as well as utilising the system’s analytical capabilities for evaluation and effect analyses. The evaluation tools will support CBA, EIA and SEA analysis of major projects and programmes as well as CBA and SEA analysis of a wide range of policy measures and general changes of subsystems of the infrastructure. The effects of infrastructure measures and certain policy measures on transport cost, transport time etc. could be calculated and presented with the appropriate regional distribution. Major transport corridors could be studied with regard to time/cost characteristics. The evaluation models comprise tools for getting access to model system data. The evaluation modules are planned to interact with other components of the national model system in order to make the best use also of already available built-in tools of various software components.

SAMPLAN Rapport 2001:1

12

1.8 Calibration and validation
Generally, more efforts should be made to calibrate as well as to validate the model system as a whole and the separate modules. Considering that early versions of the system have been used for quite a few years now, there is also scope for evaluation of the model system performance against the actual development. For the VTI/TPR model an evaluation/validation report has been produced in the early 90s. For the entire model version of 95/96 an evaluation was initiated by RRV. The report was written by Anders Karlström. There are to our knowledge no more recent efforts within this field.

SAMPLAN Rapport 2001:1

13

2 Introduction
2.1 Background and purpose
This report aims to give an overview of the present Swedish national model system for goods transport and the purposes for which is has been built. Goods transport is a vital component of the production and distribution processes of modern societies. The separation of production and consumption as well as a more or less global division of labour, act as driving forces for a seemingly ever increasing demand of goods transport. Historically, the growth of goods transport demand has been strongly correlated with GDP-growth. From time to time, however, the development of the production and distribution processes as well as the logistical systems have been such as to incur a growth rate of goods transport demand higher than the growth rate of the transported quantities of goods and also at a rate higher than the general GDP growth rate. The increasing role of international trade as well as the high absolute level of foreign trade for Sweden, also acts to stimulate a growth rate of transport demand which is higher than GDPgrowth. To some extent this tendency is counteracted by a continuos shift from low to high value goods for longer distances1. Transport demand plays a crucial role in the development of the transport sector technology and resources. Total demand development and changes of the demand structure initiate capacity changes (investment as well as divestment), and is also one of the major factors influencing transport policy decisions e.g. policies aiming at influencing the modal structure emerging from market decisions, the level of demand, the spatial distribution of transport activities, transport technology and the emissions of transport operations. Both transport policy implementation and capacity alterations to the transport systems take time, which is sometimes very considerable. A process of rational decision making therefore requires some foresight of future demand changes, and therefore some understanding of the underlying driving forces. The needs to forecast demand changes is by no means confined to policy makers, and decision makers involved in the development of public infrastructure. Many market actors

1

The traditional way to measure demand in the context of transport planning is in tons or tonkilometres. The obvious rationale for these measurement scales is the seemingly direct relation to the physical properties of the infrastructure and vehicles which define transport capacity. However, it becomes more and more important to consider also to measure demand more generally as the market value of the transport services carried out in goods transport, probably broken down into sub-markets. Such a more general market orientated definition highlights that transport services are complex and often defined by a bundle of quality attributes, which go beyond the simple physical categories tons and ton-kilometres.
SAMPLAN Rapport 2001:1

14 involved in strategic decisions which have long implementation lead times should have similar requirements. Demand information is therefore an essential component in the processes of planning and decision making about transport policy and infrastructural measures. An assessment of future demand changes is a key component of evaluation methods such as Cost Benefit Analysis (CBA) and policy analysis. Some kind of structured approach is necessary to meet the demand information requirements of these planning and decision processes. A quantitative modelling approach is needed. The purpose of this paper is to outline the contents and structure of the model system, which was developed to provide demand analysis capabilities for the Swedish national infrastructure planning process, and for national transport policy analysis.

2.2 The modelling approach
The problem of modelling the entire goods transport system of a whole nation is very complex. It is necessary to balance many needs against each other, and many compromises and simplifications will be inevitable. One has to recognise also, that alterations and amendments to the model will be necessary over time due to changes of the driving forces of demand, new knowledge as well as changing requirements on the model output due to shift of focus of the policy making processes. The modelling approach chosen for Sweden is based on the following set of key assumptions: • • goods transport demand is closely related to economic activity the structure of economic activities between sectors is important for goods transport demand since it influences the composition of final and intermediate products, which in turn influences goods transport demand the domestic regional structure of production, investment and consumption is important for the regional distribution and composition of transport demand and also for the distribution of demand on the transport networks the country and commodity distribution of foreign trade as well as its domestic regional distribution is (almost) as important as the domestic distribution of production, consumption and investment because of the extreme ”foreign trade intensity” of the Swedish economy the logistic and transport systems have inherent robust structures meaning that the geographical distribution of terminals, e.g. ports, railway terminals, road haulier terminals, and logistic centres by and large is fairly stable also over long periods of time it is a reasonable approximation to assume that the transporting markets attempt to minimise generalised cost - a concept which besides operating out of pocket cost also includes the shippers/receivers direct and indirect cost e.g. cost of stocks, cost of time delays, and goods damages.
SAMPLAN Rapport 2001:1

•

•

•

•

15 • a network representation of the Swedish national goods transport system is feasible and such a network could yield a sufficiently good representation of actual transport flows on the real network for the purposes of strategic infrastructure planning and transport policy analysis

These assumptions formed the basis on which the main model components were selected and designed as well as for the structure of the Swedish model system. Of course, over the years of development of the system, numerous methodological, empirical and practical considerations and restrictions have constrained development and forced developers to (over)simplifications and compromises.

2.3 Organization of the report
Before going into the details of the model and its components there is a discussion in the following chapter 3 on the intended uses of the model and the institutional settings of its application. We think that a reasonably elaborate explanation of these issues will help the reader to get a better understanding of what the model sets out to accomplish and under what circumstances and therefore also will facilitate the understanding of the model as such. Of equal importance is the fact that the discussion in chapter 3 hopefully in the first place will provoke curiosity about the model presented in the following chapters but also provoke questions on the relevance and quality of the model for its alleged purposes. We hope that such questions could contribute to ideas for amendments to the models and also to more profound long term re-thinking of the model system. Chapter 4 gives an overview of the national goods model system and its components. The sequential nature of the relation between the different subsystems is illustrated diagrammatically. System modules are also related to major input and output data sets which interconnect the subsystems. Chapters 5 and 6 provide a more detailed description of the two blocks of the national model system. In chapter 5 the sub-models of the demand model block are described and chapter 6 explains briefly the function, principles and structure of the transport demand/network model. The evaluation tools interacting with the national model system at various levels are described in chapter 7. The report is concluded in chapter 8 with some application examples from the fields of transport projections/forecasts, infrastructure related CBA and traffic/transport analyses, and policy analysis.

SAMPLAN Rapport 2001:1

16

SAMPLAN Rapport 2001:1

17

3 Application areas and the institutional setting of the national model system
3.1 The application arenas
The national models systems for goods and passenger transport are primarily developed and used to support decision processes of the state as for the allocation of state funds for transport sector purposes such as infrastructure investment, operation, and maintenance as well as safety and environment related measures, taxation of the transport sector, the use of transport policy measures such as vehicle standards, environmental issues etc. These decision processes take place in four major arenas namely: 1. National and regional transport policy analysis 2. The national and regional infrastructure planning process 3. Swedish contributions to EU transport policy development as well as formation of Swedish policy positions related to the EU policy level 4. In depth analysis of major infrastructure projects The requirements of the decision processes and actors on these arenas provide the primary basis for the demands on the national transport models. It is of course not possible to go into any detail on these demands in this overview. However, a brief comment on each one of the four points mentioned above, will indicate the nature of the processes on each arena and therefore also hint at the kind of decision support expected from the national transport model systems. Each of the four points above is briefly commented below. (1) The national transport policy is shaped in a continuos process where decisions are codified by government and parliament in the regular legislative processes. Major revisions of the policy, however, have been made approximately every 10 years. The policy now in force was decided by the parliament in 1998, based on a government bill, after extensive studies and analyses had been carried out by a parliamentary commission in the years 1995-1997. (The present policy was preceded by policies of 1988, 1979, 1963 etc). (2) The principles of the national infrastructure planning processes have developed over the years. Therefore it seems fair to say that there is no single unified piece of legislation which regulates these processes, but rather a bundle of regulations and principles of different dates and origin, and many important
SAMPLAN Rapport 2001:1

18 aspects are not legally codified at all. The state involvement in transport infrastructure planning is subdivided into two separate parts. The first is planning for infrastructure and other measures financed over the state budget, which covers road, rail and regional infrastructure. The second is the maritime and aviation infrastructures, which are financed by user charges and planned in a decentralised way by two separate state agencies. The parts of infrastructure planning which are (wholly or partly) financed over the state budget are subject to decisions within the framework of the yearly and the 3year budget mechanisms of the state. Many infrastructure investments, implementation of maintenance policies etc are complex projects which require long construction or implementation time, and have huge resource requirements. Transport infrastructure is also of a long lasting nature. It has therefore been considered necessary to supplement the 3-year and yearly budget procedures by a 10-year planning process, which is carried out every four years. The formal responsibility for this 10-year planning process lies with the national state agencies for road and rail respectively, and with each one of the county administrations, which are also state agencies at the regional level. The central agencies for rail and road have to provide the counties with study and analysis material (e.g. CBA, SEA) as a basis for the regional planning process. There are at least two inherent problems in the 10-year planning process. First, there is justified demand for appropriate political influence on the direction and outcome of the 10-year planning process. Second, there is a need for co-ordination in at least three dimensions namely between the national and the regional level, between the regions themselves, and between the transport modes – rail, road, air, maritime – of which the last two are not directly involved in the budget allocation process. It is not a simple task to come to grips with these co-ordination issues, but the need for some kind of co-ordinating body has been recognised over the years. The present mechanism is really a mix of mechanisms, where a separate state agency (SIKA) has one role, a common steering group including SIKA, the national ”modal” sector agencies, and representatives of the counties have another role. The government steers the process at various stages by means of formal directives. Many demands on the national model systems originate within the requirements of this 10-year planning process. The processes are demanding, since both strategic overview and sufficient detail are required. The modal co-ordination perspective requires an inter-modal approach while retaining relevance for the modal 10-year planning process of each separate agency. The inclusion of the regional level calls for sufficient detail for the study of regional transport problems. (3) Quite a few years have lapsed since Sweden joined the EU. Gradually it has become apparent to policy actors at all levels that essential elements of Swedish transport policy is in fact formulated in EU fora. Therefore there is demand for Swedish contributions to EU transport policy development as well as the formation of Swedish policy positions related to the EU policy formation processes. The Swedish analytical tools, including demand models, should

SAMPLAN Rapport 2001:1

19 preferably be sufficiently versatile to be able to contribute to decision processes related to Swedish –EU interaction in the formation of transport policy. (4) For major strategic infrastructure projects in-depth analyses are required. Examples are the Öresund bridge, Fehmarn bridge, the Botnia railway, the Stockholm ring roads. Of course, the many aspects and dimensions of such large infrastructure projects call for many different types and methods of analysis. However, it is a great advantage if the national model system can be used for demand analyses also for such projects, whereby the demand analyses of the major project will become comparable with other projects and bundles of project. Such an approach will also ensure that the often nation-wide system effects of a large projects will be treated in a defendable way.

3.2 Actors and arenas
Many actors perform at more than one of the arenas mentioned above. The following table summarises some major state actors and their roles at the different arenas (1-4).
Table 3.1. Major state actors in transport policy and planning areas in Sweden.
Arena/Actor Ministry, SIKA, Commissions Policy analysis National Rail Administration Rail sector policy analysis National Road Administration Swedish Maritime Administration Swedish Civil Aviation Administration Aviation sector policy analysis Profitability analysis, strategic analyses Regions/ County Municipal level Spatial planning, corridors 10 year strategic regional plan, regional/ land use, environ mental issues, regional projects Informal, project particip Regional/ land use, environ mental issues

Transport policy Infrastructure planning

Road sector Maritime policy sector analysis policy analysis 10-year strategic infraplan, project/mea sure planning Strategic planning, project evaluation

Co-ordination, 10-year common strategic principles infraplan, project/mea sure planning

Influencing EU-policy

Decision/ negotiation support

Decision/ negotiation support Basic analysis and evaluation

Decision/ negotiation support Basic analysis and evaluation

Decision/ negotiation support Basic analysis and evaluation

Decision/ negotiation support Basic analysis and evaluation

Major strategic Ad hoc projects evaluation Reviews

SAMPLAN Rapport 2001:1

20

3.3 Development and administration of the national goods model system
The Swedish national model system is continuously developed in co-operation between the members of a consortium of Swedish state agencies involved in infrastructure planning and development as well as transport research. Members of the consortium are the Swedish Institute for Transport and Communications Analysis (SIKA), the Swedish National Rail Administration, the Swedish National Road Administration, the Swedish Maritime Administration, the Swedish Civil Aviation Administration and the Swedish Communication Research Board. Since SIKA is formally endowed with the responsibility to carry out method development of common interest to the sector, and to ensure the compatibility of methods and information used in the planning processes, SIKA is also chairing the steering group for development and administration of the national goods model system. (The group is called the SAMGODS-group).

SAMPLAN Rapport 2001:1

21

4

Overview of the model system and its components

The Swedish national model system for goods transport could most correctly be described as a set of separate models, which have been tied together to enable them to interact in a reasonably effective and consistent way in the process of modelling Swedish goods transport demand. These models were sometimes developed for entirely different purposes than the analysis of national goods transport demand. The system comprises two major subsystems namely the transport demand models, and the transport network/transport market model. A third model package, the evaluation models and tools, are associated with these models. The demand models generate spatially disaggregated demand matrixes which are also disaggregated into separate commodities. The spatial structure covers not only Swedish territory but also the European union as well as all other areas involved in Swedish foreign trade. The demand matrices are used as input to the transport network/market model, which allocates the flow matrices to the transport modes and networks according to a system cost minimizing algorithm. The model system allows projections to be made of transport flows on transport modes and networks, analyses of transport chains, modelling of transport system effects of certain transport policy measures as well as the effects of changes to the infrastructures. A diagrammatic representation of the system and its functional components is given in figure 4.1. below. The figure also illustrates the major data sets connecting the modules as well as specific input data sets for the various system modules.

SAMPLAN Rapport 2001:1

22

-World market prices -Public sector demand

Multi Sectoral Model of the Swedish Economy (ISMOD)

Exogenous balance conditions: -Current account -Employment level

Production, export, import etc per sector Model for subcountry level region to region Swedish foreign trade

Regional level and distribution of employment (EARLY)

-Flow matrices for base year -Domestic regional distribution of production, consumption -Conversion sector - product matrices -Input/-output tables

Model for interregional domestic transport demand (+export/import demand) VTI/TPR

Model for calculation of implicit price/ton for commodity aggregates

Demand matrices domesic, export, import per product Transport costs on links and nodes; capacities

Networks for all modes

Transport Network Model (Transport Market Model) STAN

Evaluation models and tools (CBA, effect analyses)

Figure 4.1. Overview of system components.

SAMPLAN Rapport 2001:1

23

5 The demand models
5.1 Introduction
The goods transport demand models (”matrix-generation”) comprises five separate models: • • • • • Multi-sectoral system of models of the Swedish economy (ISMOD) A model for regional disaggregation of sectoral employment (EARLY) linked to ISMOD A model for modelling inter - regional transport demand within Sweden (VTI/TPR) Model for regional forecasting of Swedish foreign trade Models for forecasting of implicit commodity value for aggregates of commodities

An overview of the functions of the separate models and their interaction was given in chapter 4 above. In the following sections each of the subsystems is described briefly as for function, input and output. The multi sectoral model ISMOD is an input-output model. A characteristic feature of the model is the endogenous generation of the capital stock, with a productivity which varies with capital vintage. The model is driven exogenously by world market commodity price forecasts as well as domestic public sector demand forecasts. Additionally two exogenous balancing restrictions are given namely employment and the balance of current account. The model provides as output production, investment, intermediate commodity demand, employment as well as wage and profit levels for each of 31 sectors. Also, private consumption, export and import are solved for commodities and services specified in the model. The model is designed to produce projections of Swedish economic development for a medium term perspective (5-10 years). The output of the ISMOD model for production, export, import and consumption for each sector is used as input to the EARLY model, and to the system for modelling interregional transport demand (VTI/TPR). The EARLY model is linked to the ISMOD model. Based on the ISMOD demand projections per sector total national employment changes are calculated per sector. These changes are further disaggregated per region in the Early model

SAMPLAN Rapport 2001:1

24 The purpose of the foreign trade model system is to forecast trade (export and import) between Swedish municipal areas and numerous foreign trade regions disaggregated into commodity aggregates. The model system has two main subsystems namely the bilateral trade model subsystem and the subsystem modelling lower level regional trade flows. The models system comprises some 50 bilateral trade models for suitable trade areas and commodity aggregates. These models are of the gravitation type using inter alia GDP and distance as independent variables. In principle the model system is intended to provide forecasts of inter-area trade patterns on the most disaggregated area level used in the transport market model (demand matrix level) as well as a suitable commodity disaggregation. The model for inter - regional transport demand within Sweden (VTI/TPR) uses as input data the output from the ISMOD multisectoral model, output on regional employment changes from the EARLY model , output from the foreign trade model system, and output from the model for forecasting implicit prices for the commodity aggregates used in the model system. The VTI/TPR models utilizes an entropy algorithm to estimate forecast demand matrixes for the relevant commodity groups. Estimates of the present domestic inter-regional transport flow matrices based on available empirical information as well as the corresponding foreign trade matrices are used as à priori matrices in the entropy algorithm. The marginal conditions for each commodity is derived from the ISMOD-model output.

5.2 The multi-sectoral system of models of Swedish economy (ISMOD)
Background on some problems of planning and uncertainty The general economic and structural development of the Swedish economy are issues which are primarily dealt with by institutions outside the communications/ transport sector. In the Swedish context the long term aspects (10-20 years) of the development of the economy are studied and discussed by the ad hoc long-term study group within the ministry of finance. Official reports are released every 2-4 years. The earlier tradition of “indicative planning” with detailed development paths for the economy as a whole and for different sectors was abandoned in the 1990 Long term study of the Swedish economy. (“From planning tool to a basis for discussion and debate”). The dominating time perspective in these analyses is the middle term 10 year perspective for the analyses of the main structural issues for the Swedish economy. Issues of stability of the economy are analysed with a five-year horizon. Some exercises are carried out with a longer time perspective – 25–35 years, but it seems fair to say that the “long term studies” for Sweden have their focus on the 5-10 year horizon. Rather than indicating “most likely development paths”, projections conditioned on certain sets of assumptions are presented and discussed. Though the co-ordination of the long-term studies are in the hands

SAMPLAN Rapport 2001:1

25 of the ministry of finance, the study work draws on contributions from many sector ministries and agencies. It is well known that transport sector investments ideally have to be evaluated on a long term time basis. The technical life of rail track investment or road investment are in the range of 25-100 years. However, it is widely recognized today that the uncertainties associated with e.g. composition and level of demand for such long horizons are paramount. Despite this fact, for the present infrastructure planning paradigm, procedures are based on the idea that a “most likely” very long term scenario (up to 20 years) could be developed for the key factors determining transport demand, and that a reasonable extrapolation could be made from the “most likely” situation 20 years from the present time, onwards into the future for another 40 years. This will yield the transport demand development path over time and subsequently the basis for the CBA covering the whole life span of proposed investments for the next 10 years. As was said above, such very long term development paths, are beyond the scope of the national long term studies, partly due to the fact that the study teams have found it questionable, whether such long term scenarios/projections are meaningful. Also, in the background material from different sectors outside transport, the time perspectives are normally very much shorter than 20-60 years probably with the exception of national aggregate demographic forecasts, which have a comparable time horizon. Therefore, to sustain the planning paradigm, it has turned out to be necessary for the transport sector to take specific action to generate adequate long term scenarios for the long term development of the economy at large, as well as its sectoral production structure, and the Swedish demography. The transport sector agencies have chosen to make use of the ISMOD system of multi-sectoral models to generate the required long term structural scenarios. Co-operation links have been established with the ISMOD expertise at the governmental agency level as well as the research level.

A brief description of the model The ISMOD system generates overall and sectoral scenarios for production and import on one side and intermediate goods deliveries private consumption, government consumption, investment, export, and trade/transport on the other. The model system’s point of departure is the current state of the above mentioned variables. Through a stepwise iteration process, future states of the same set of variables are calculated, subject to a set of consistency conditions. For each specific “forecast” year values for export/import prices and public sector consumption have to be given exogenously. The model system also requires total employment and a condition for the external balance of current account to be given exogenously. For each sector the total supply (domestic production + import) is to be exactly equal to total demand (private and public sector consumption + export +

SAMPLAN Rapport 2001:1

26 intermediate goods deliveries + investment). The current account condition has to be satisfied as well as the employment condition. The latter condition is satisfied through the adjustment of private consumption until the pre-determined level of employment is reached. Total investment as well as investment per sector are determined within the model. A set of parameters are given as input data, which define the productivity of old (there is a vintage structure for the old capital) and newly invested capital in each sector. New investment capital is assumed to have the properties of the 25 % of the capital stock with the lowest specific labour requirement. Private consumption is described by a linear consumption system for ten commodities where the arguments are price index of each commodity and the total level of outlay. The consumption system used is simultaneously estimated and has also been used within the national long term studies. There is a transformation matrix to translate the 10 commodities of consumption into the 28 commodities of the production system. Export and import are described by similar functions. Export (import) is assumed to be determined by a) the ratio between the world market price and the domestic price (inverted for import) b) a price independent time trend The models for export and import have been estimated on time series data. Sectors and products are not identical. Most sectors produce more than one product. Therefore, an output matrix relating sector output to products is used within the model system. This matrix is based on the regular input/output statistics produced regularly by Statistics Sweden.

Issues of transport sector application As was partly touched upon above the demand for very long term scenarios for the Swedish economic development and for the development of the industrial structure, seems to be a specific characteristic for the transport sector planning approach. This fact has turned out to cause certain problems of credibility of the very long term scenario in a world where such long term perspectives in the particular sense of a “most likely future”, is looked upon with suspicion by influential actors. For the ministry of finance and the actors of industry and trade, the long term planning horizon seems to be rather 5–10 years than 20-60 years. In this context of inconsistent time perspectives consistency is still in demand. One would like to see consistency with (other) more or less official aggregate projections for the Swedish economy as well as consistency with the industry sectors’ own assessment of each sectors national production, export/import and international price trends. Due to diverging time horizons such consistency is

SAMPLAN Rapport 2001:1

27 inherently problematic. In practice, what could be done, is to ensure that obvious inconsistencies/contradictions do not exist. Trade and transport are treated in a simplistic way in the existing operative version of the model. The output of these sectors is assumed to amount to the aggregate trade margins of the other sectors. However, development of the transport sector representation of the ISMOD system is currently under way, and could hopefully become operative within the next few years. Experience from the application of the ISMOD model system in real world infrastructure investment planning, has made it plain that the sectoral structure directly influences what kind of goods to transport as well as goods volumes/quantities. The contents of the ISMOD scenarios, therefore, eventually will have an impact on the relative costs and benefits for e.g. rail and road investment and other measures and consequently on subsequent strategic investment choices.

The regional structure of production, employment, consumption and foreign trade. The EARLY model The ISMOD model system uses no geography; the multi-sectoral production/ consumption system defined by ISMOD encompasses the whole of Sweden as an entity. From the viewpoint of analysing transport – policy, investment and other infrastructure measures – it is obvious that the underlying geographical structure is of vital importance, since the very purpose of transport is to bridge distance. However, the need for geographical disaggregation of the structural development of industry and trade has been felt also by actors outside the transport sector. Regional and municipal authorities involved in municipal and regional development planning, building, health care, and education make use of forecast data on population and employment. Over the years different approaches have been tested, and for many years local/ regional projections were aggregated to the national level and fed back to the regional/local level after national consistency adjustments. This system was used to provide projections on regional employment changes per sector in earlier applications of the national goods model system. A totally new comprehensive such a forecasting/planning tool, the so called RAPS system, has been commissioned by a state agency to replace existing routines, but the system is still only in an early implementation phase. For the last infrastructure planning application of the national goods model, a model called EARLY, which is linked to the ISMOD-model, was used to provide projections of regional and sectoral employment changes. The purpose of the EARLY model is to provide regional and sectoral disaggregation of employment (for ISMOD sectors) as well as to provide data and projections on manpower demand for educational categories.

SAMPLAN Rapport 2001:1

28 The disaggregation process of the EARLY model is carried out in a number of steps. One important tool is the separation of sectors into “engines” and “wagons”, with the idea that some sectors (engines) could be viewed as primary employment generators, while employment in other sectors (wagons) could be looked upon as induced by employment changes in “engine” sectors. By the combination of historical data, “engine” sector growth and decline probabilities and the model of engine/wagons, employment changes per sector and region are calculated. The EARLY model is at present implemented in the Excel software. These issues will be dealt with more in the following sections. Suffice it here to finally point out the important links between regional demography and migration on one side and the (changing) structure of regional consumption, employment and production on the other. From these interdependencies it is clear that in an ideal transport planning world the framework scenarios used for analysing personal travel should be made consistent with those used for the analyses of goods transport.

5.3 Modelling interregional transport demand within Sweden (VTI/TPR)
Statistics for domestic interregional transport flows – some background remarks Official statistics are very much arranged to provide information to support the decision processes of the state. Traditionally statistics tied to the income flows of the state (taxes, fees and customs duties) as well as statistics capable of providing some guidance on the allocation of the expenditure of the state e.g. on transport infrastructure, have played an important role for the development of the national official statistical systems. The nation states created integrated domestic markets for goods, services, labour, and capital with free and unrestricted movement within the national borders. Therefore no real need has been perceived for official statistics related to domestic flows. The contrast is striking with the amount of statistics related to foreign trade and travel. For the planning of national infrastructure flow, statistics related to the existing infrastructures have been regarded as sufficient. For the purposes of regional policy and redistribution of incomes between regions, it has been felt adequate with official statistics on variables related to geographical sub-areas of a country (population, employment, regional incomes etc.). Understanding and modelling the broad economic interaction between domestic sub-regions, as well as the modelling of domestic interregional transport flows on an origin-destination (O/D) basis have had to rely basically on area data. In transport, the growing infrastructure planning ambitions of the state, aiming at the best use of public funds within the transport sector, have stimulated the development of sophisticated models for the estimation of O/D matrices. Given that such matrices exist, it becomes possible to simulate more realistically the

SAMPLAN Rapport 2001:1

29 effects of system-wide development of the infrastructures, simultaneously for all modes of transport. For the movements of people the combination of traffic counts and national travel survey data has formed the basis for the model calculation of flow matrices making use also of area data on population, income structure, employment level and structure. For goods transport, the need for a collateral to the travel surveys for the purpose, among other things, of O/D estimation for transport modelling, has only recently been recognised. (c.f. Commodity Flow survey in the US, Pilot Commodity Flow Surveys in Sweden, recent Eurostat discussions). Given the lack of genuine O/D data for goods, the Swedish modelling efforts for interregional goods flows have been based on imputations of O/D matrices from observations of points (areas) of loading/unloading of goods. For road transport the yearly road transport surveys, which are part of Swedish official statistics, have been used as a basis for these imputations. Observations have been aggregated for a number of years to form a “synthetic” road transport O/D matrix. For domestic sea transport the official statistics have been used in a similar way. For rail transport, no such statistics are public, though the information per se is available with the rail operators, but treated confidentially. Therefore, the “synthetic” rail matrix has to be estimated on scattered information aggregated for geographical areas and commodities. By “synthetic” we refer here to the fact that the matrices based on transport statistics (loading/unloading) do not capture only the true end-points (genuine O/D) of a shipment of goods, but also the intermediate points, where trans-shipment of goods take place.

A brief description of the model (VTI/TPR transport model) The main purpose of the model, as it is used presently2 by the Swedish planning authorities, is to generate projections of O/D matrices for goods flows, disaggregated into commodities and geographical zones so as to be compatible with the matrix structure required by the transport market model (see below) which is implemented in the STAN network software. The projections are based on input from the multi-sectoral model system (ISMOD) as described in section 4.2 above. The matrix projections used by the STAN-model are not mode specific. The modal distribution and computation of ton kilometres etc. are based on the multi-modal network assignment of the STAN-model (see below).

2

The development of the model was initially commissioned by the Swedish agency ”Transportrådet” to aid in forecasting long term goods transport development for Sweden for commodity groups and transport modes. The model retains the capability to produce such forecasts. However, the present use of the model is basically to generate projections of O/D matrices for goods flows, which are compatible with a) certain scenarios for the structural development of the Swedish economy as generated by the ISMOD-model (see previous sections) b) The O/D matrix structure and level of disaggregation used in the transport market model STAN, see below.
SAMPLAN Rapport 2001:1

30 In general terms the model is based on the reasonable assumption that the interregional transport flows are related to the absolute and relative regional economic activity. Specifically it is assumed that transport volume from a region for each commodity group is directly related to the total regional supply of this commodity. Analogously transport to a region is assumed to be directly related to total demand of the region3. Just like the ISMOD model the VTI/TPR model is based also on the concept of a base year and a forecast year and the model is basically focused on the change ratios between those two years. The model has two main modules namely the economic model and the transport related O/D model. Both these models operate on a regional level for Sweden based on eight regions. The first computational step of the economic module of the model is to determine the regional economic development in such a way, that it is consistent with national economic development and the structural development for sectors according to the ISMOD model. To maintain the relation to transport activity to and from each region, as mentioned above, it is essential to keep track separately of the components of supply and demand. Supply components are production and import, which are assumed to relate to transport from a region. Demand components are consumption, investment, intermediate goods/services and export which are assumed to relate to transport to a region. For both base year and forecast year equality between supply and demand is assumed to hold for each commodity aggregate. Changes between the base and the forecast years of export and import are given from the ISMOD model and the change rates are assumed to be equal for all regions (per commodity). Aggregate production levels for the base and the forecast years for the commodities are also given as by the ISMOD model. However, contrary to the treatment of foreign trade, the distribution of overall production level change over regions is based on the regional change of employment in each sector. Therefore the model requires as input data also exogenous forecasts on the change of employment per region and sector between the base year and the forecast year4. Intermediate goods volume is determined via an input/output table, common for all regions. Consumption/investment are residuals. Given that the model has determined supply and demand components in value terms for each region for the base year and the forecast year, a quotient between forecast value and base year value could be calculated separately for the supply and demand components for each region, which gives the relative value change for each component.

3 4

Though this assumption is reasonable it is nevertheless open to objections. This mechanism has been much discussed . It could be argued that the correlation per sector and region is not necessarily strong between sector output change and employment change. It could well be that the employment changes mostly occur in companies classified as belonging to a sector for which output volume increases only slowly. On the other hand increasing labour productivity e.g. due to e.g. increased capital intensity leading to relatively small employment changes might go alongside with considerable output volume growth.
SAMPLAN Rapport 2001:1

31 However, transport flows, according to the present practice, are defined in quantity terms. The ratios for value changes of supply and demand for the region therefore could not be used directly to compute the corresponding quantity changes. Changes over the forecast period of the value/quantity relation – the implicit price - for each commodity aggregate, have to be taken into consideration. Therefore, a forecast of the change ratio of this implicit aggregate price has to be introduced. Multiplying the change ratio of the value of supply and demand respectively by the corresponding (inverted) change ratio of the implicit prices yields factors of quantity change separately for supply and demand for each commodity and region. Provided base year transport quantities to and from each region is known for each commodity the corresponding quantities for the forecast year could now be calculated. The base year volumes is simply multiplied by the change factors according to the last paragraph to determine the corresponding forecast year transport volumes to and from each region for the respective commodity. How are then the transport quantities determined for the base year related to each region? The current method is to estimate an O/D matrix (aggregated for all transport modes) - for each commodity - based on transport statistics according to the above discussion. When such an O/D matrix is aggregated to the eight regions of the economic model discussed above, the row sums of the resulting 8x8 matrix correspond to quantities from the region and the column sums correspond to the quantities to the region. These sums represent the estimated values of base year quantities related to each region and commodity. These estimates are then multiplied by the change factors for economic activity and implicit price as discussed above to render the forecast transport quantities to and from each region (per commodity). There are two remaining tasks of the model. The first one is to find the forecast pattern of interregional transport flows for the 8x8 regional matrix and the second one is to disaggregate this pattern back to the level used by the STAN-based transport model. These two issues will be dealt with below. To calculate the forecast pattern an entropy model is used. The forecast quantities to and from each region gives marginal condition for the wanted forecast transport pattern5. The known matrix for the base year is used as an à priori matrix for the entropy model calculation and the output is the wanted forecast O/D matrix on the eight-region level. Disaggregation of the 8x8 matrix is accomplished in a simple way. Dividing each element of the forecast matrix by the corresponding element in the base year matrix yields a matrix with a change rate for each element. The forecast disaggregate matrix is computed through multiplying each underlying element by this change rate.

5

The marginal conditions computed in the way described above are not necessarily consistent. Therefore the model recalculates the column sums to values which ensure consistency of the column and row sum conditions. The reason for adjusting the columns is that the values of the row sum are assumed to be more reliable.
SAMPLAN Rapport 2001:1

32

Some remarks on the data requirements of the model Some important aspect on input data relating to transport O/D matrix estimation was already discussed above. The economic module which regionalises the national figures requires input data. The model uses regional distribution proportions for export, import, consumption/ investment and production for the base year. The proportion figures sum to unity for each category, and provide the basis for allocation of the overall national volumes to the regional level. For export, import and consumption/investment the proportions per region are assumed to remain unchanged between the base year and the forecast year. Moreover, the actual numerical proportions which are available at present and have been used in the practical applications of the model, date back to the 1980s. It would certainly be justified to update these figures for coming operative applications of the model, and to consider forecasting mechanisms for the future regional distribution of e.g. export and import. Production proportions are changed by the model between the base year and the forecast year. The changes are based on regional forecasts or employment change per sector as was discussed above. Long term forecasts of regional employment changes are unfortunately not readily available from sources outside the transport sector. In practical application, therefore, forecasts on regional employment have proven to be hard to come by6, which might cause problems in the application of the model. The model system VTI/TPR also uses an input/output table to generate demand for intermediate goods. Ideally, for the model as it is set up, regional i/o tables should be used. Up to now, official regional i/o tables have not been available. Until this is changed it seems reasonable to consider further the impact on the transport forecast of this deficiency.

5.4 The foreign trade model
The foreign trade model is a new component of the overall goods transport model system. It was commissioned and subsequently developed for the planning and forecasting work during 1999 and 2000 and was scheduled to provide background planning and forecast material for the long term Swedish planning cycle for the period 2002-2011. In the above discussions on the multi-sectoral model system (ISMOD) and the discussions on the VTI/TPR model, we have already touched upon the foreign
6

One option which was tried in preparing the national goods transport forecasts in 1999-2000 was to use the ”Early” model which is linked to the ISMOD model as discussed above. The development of the RAPS model system, initiated by NUTEK, aims at improving access to regional forecast data for planning purposes. The RAPS model development was recently completed and therefore practical experience is still scarce.
SAMPLAN Rapport 2001:1

33 trade component of the Swedish economy. In the ISMOD model system foreign trade forecasts are produced disaggregated into commodities, and these forecast figures are then input to the VTI/TPR model. However, there is no geographical disaggregation of these trade forecasts, neither on foreign trade partners nor on the distribution of foreign trade between Swedish regions. One would be inclined to think, that such geographically disaggregated forecasts of foreign trade would be available from sources outside the transport sector. But unfortunately this is not the case, probably because such forecast would have limited use outside the transport sector. However, the increasing international trade, is an important driving force behind transport demand. In the transport context, therefore, the geographical structure becomes very important - if not decisive - and has to be tackled. The new foreign trade model is not yet fully integrated into the demand modelling framework described above, but work is going on to accomplish this. The model in its present form is implemented with a user-friendly interface, which allows the user either to use default values for parameters and sets of independent variables, or to introduce and select customised values and procedures, which include the re-estimation of trade equations. The model comprises two main modules. The first one is the bilateral trade forecast model, which generates trade forecasts for the total Swedish foreign trade per trade partner (country or groups of countries for distant areas). The second module distributes each bilateral trade flow on Swedish as well as trade partner sub regions. The purpose of the latter is to take one step towards generating subregion to sub-region O/D matrices for the Swedish foreign trade. The bilateral trade forecast module comprises a set of trade equations, which were estimated for four commodity aggregates and three country groups (West Europe, East Europe and Rest of the World). From these equations a number of about 50 separate trade flow equations were deduced. Important independent variables are GNP and population in exporting as well as importing countries. In the estimation process also the (historical) real currency exchange rates have been included. Based on the user’s choice of change rates for the independent variables, the rates of change for export and import are computed for aggregates. Thus calculated growth rates for aggregates are applied by the model to more detailed commodity levels included in each aggregate. The regionalising module distributes the bilateral trade for commodity groups on Swedish and trade partner regions. Export and import of goods classified as intermediate or investment type goods are distributed according to industrial employment. Goods classified as being of consumption type are distributed according to disposable income, population or gross regional product per capita. It is possible to combine the regional distributions for Sweden and the trade partners to generate a complete O/D matrix, which would relate trade flows for each Swedish region to each trade partner region. The empirical basis of such an O/D matrix must be regarded as very weak, however.

SAMPLAN Rapport 2001:1

34

SAMPLAN Rapport 2001:1

35

6 The transport network model
6.1 Introduction and purpose
The purpose of the network model is to represent the transport markets' allocation of transport demand to the combined networks of all transport modes. This allocation also reflects the market’s selection of transport routes (paths). Transport demand matrices, which are output from the model for interregional transport demand (VTI/TPR), are input into the transport network model for each individually defined commodity group. The allocation of demand to the modes and routes in the combined multi-modal transport network is controlled by the cost conditions of the transport services production system, which are represented as parameters (cost functions) in the network model. The network model is implemented within the framework of the STAN package, which uses a “system cost minimising algorithm” to solve the allocation problem. In the policy context where it is used, it is essential that the national model system produces network related demand projections and also allows transport network policy analyses. Only the capability of explicitly working at transport network level allows the model system to be directly useful for analyses of a wide range of issues related to transport infrastructure development (investment, divestment, system-level changes of the general properties of subsystems in infrastructure, corridors, missing strategic links etc), as well as for the analyses of many transport policy issues such as taxation/charges related to certain areas or infrastructures as well as infrastructure access regulation. Today’s transport policy discussions focus largely on issues such as intermodality, international transport chains, level playing field of competition between transport companies with different modal focus and different domiciles (ship-owners, railway operators, road haulier companies, air freight companies). It is, therefore, mandatory that a transport market subsystem is capable of dealing effectively with inter-modality – modal co-operation and competition for all relevant modes within the context of Swedish domestic and foreign trade. It is also essential that the transport market subsystem allows for international networks to be included for all relevant transport modes. It is also obvious that there are logistical differences between categories of products and that these have to be dealt with in a reasonable way in the model. Therefore, it is a requirement that the transport market system allow a reasonable product separation, which allows the introduction of product-specific cost parameters.

SAMPLAN Rapport 2001:1

36

6.2 Brief development history
As previously mentioned in section 5.3, the earlier versions of the regional demand model (VTI/TPR) also allowed the production of transport forecasts disaggregated by mode, product, and main areas. However, there was no systematic link to a transport market/network module. It very soon became obvious, however, that the requirements of policy analysis and infrastructure planning processes made it necessary to expand the national model with a module to do so. In the course of time the gradual development of the national system was developed and the STAN package was selected in 1994/1995 as the basis for the transport market model module. The STAN package seemed to meet all the mandatory requirements, however, not completely or totally satisfactorily. Two most obvious system shortcomings at the time were network size (total number of zones and links) and the total number of products and modes. The first STAN-implementation of the Swedish national model system was operable in 1995 (see Vägverket, Banverket, Sjöfartsverket, SIKA, Modell för simulering av godsflöden, November 1995). This model was used to support the infrastructure planning process for the period 1998-2007 which was carried out between 1995-1997, and also to support the analysis of specific major infrastructure projects such as the Bothnia Railway. A thorough revision of the 1997 STAN-implementation was carried out during 1998 and 2000. This included many implementation amendments and improvements as well as a thorough re-examination of the cost functions and parameters that are essential for STAN. At the same time the introduction of functionally improved, expanded, faster, and more reliable versions of the basic STAN software were also implemented by INRO (supplier of the STAN software).

6.3 Basic functions and components of the transport market/network model
The basic structure of the network model package, STAN, is similar to other network models used for transport analyses and planning purposes. Prominent features of the STAN package are the system optimising allocation algorithm and reasonably well developed mechanisms for the transfer of goods between modes. Many of the practical application problems are similar to those encountered with other software packages within the same field. It is not intended to address such discussions here. The basic building-blocks of the STAN-based network model are the following: • • Zones (O/D matrix structure and connection to the networks) Modes (a description of the various modes, together with mode-specific parameters)
SAMPLAN Rapport 2001:1

37 • • • • Products (each commodity has its own O/D-matrix and product-specific parameters) Vehicles (the possibility to define vehicle-specific parameters for each commodity AND mode combination) Networks (covering all relevant modes, as well as the possibility to transfer between them, relevant for the type of transport problems to be studied) Cost functions (describing the how to calculate the costs based on all of the above)

The STAN software contains certain limitations for these categories.

The zone structure In Sweden the zones are defined at municipal level (kommun), which means that the number of domestic zones is 288. The zoning system outside of Sweden, which is important in order to adequately handle foreign trade, the strategy has been to have a detailed zone-structure for nearby countries, but to operate with less detail for more distant countries and continents. The total number of zones is today 462. Of these there are 52 zones within the rest of Scandanavia (Norway, Denmark and Finland) and 97 zones in the rest of Europe. Outside Europe there are 25 zones. The largest version of the STAN software today can handle up to 1 600 zones. This is not much compared with the number of zones used in the SAMPERS model system where there are 668 zones in the National model and approximately 6 000 zones in each of the regional models. When defining the zone structure it is also necessary to consider the availability of background data for each zone, not only within Sweden but also for every country/region outside Sweden. Moreover, the structure has to be compatible with available transport or goods flow statistics, which can be used to support the estimation of O/D matrices. For goods transport there is also the problem of keeping data for individual companies secret. The smaller the zone-definition is the poorer the information is and the more sensitive the data becomes. For detailed information on the zone structure in the model please refer to the background report concerning “Demand matrices (Efterfrågamatriser)”.

Modes in the present model version The existing version of the network module comprises of the following modes: • • • Road haulage Rail (separate for heavy system trains, standard trains and combined road/light rail) Maritime (separate for Sweden, Europe, Overseas and Inland waterways)
SAMPLAN Rapport 2001:1

38 • • • Rail ferry Road ferry Air freight (separate for pure freight and pax-belly)

The STAN package allows up to 30 individually defined modes/submodes. In the existing goods model 13 modes are used.

Products The purpose of the product separation is to allow different transport properties of commodity categories to be reflected in the cost functions and O/D-matrix structures. The VTI/TPR model, in its present form, is capable of handling 15 or more products. The technical upper limit is set by the level of product specification of the underlying transport statistics for the matrix computations at product level. It should be emphasised however, that the reliability of detailed product matrices tend to become highly questionable due to sparse data observations in the underlying transport statistics at such detailed levels. For use within the STAN model package, product aggregates are formed based on the commodity structure available from the VTI/TPR and foreign trade models. At present the following 12 commodity groups are defined:
Table 6.1. List of existing commodity groups Commodity group 1 2 3 4 5 6 7 8 9 10 11 12 Description Dry and heavy bulk goods Liquid and voluminous bulk goods Investment goods, durable consumer goods Heavy intermediate and consumption goods Light weight consumer goods with high value Light weight intermediate and consumer goods Containerised bulk goods Iron ore from Northern Sweden (Lapplandsmalm) High value container goods Low value container goods Transit goods Air freight

Note! That a new commodity definition based on industrial sectors is now being used for the new Swedish Commodity Flow Survey (VFU). This new commodity definition will also be used for the next generation of SAMGODS models and is

SAMPLAN Rapport 2001:1

39 also more consistent with the SAMGODS demand models. The new commodities are:
Table 6.2. List of commodity groups used in VFU and STAN2002. Commodity group 1 2 3 4 5 6 7 8 9 10 11 12 13 Description Agricultural products Unprocessed lumber Processed wood products Foodstuffs Crude petroleum Petroleum products Iron ore and metal waste Metal products Paper and pulp Earth, stone and building material Chemicals Manufactured industrial products Transit products through Sweden

The number of products is at present limited to a maximum of 15 within the STAN software. The current implementation makes use of 12 products (and the coming model will make use of between 13 and 15), which limits the scope of product separation with a view to transport properties of goods. On the other hand it has turned out to be rather difficult from a practical point of view to work with too many products due to ensured matrix data quality and its links to official statistics.

Vehicles In the STAN package specific parameters can be defined for each mode and commodity group combination. This means that e.g. a particular train-type can be defined that have the characteristics of a system-train that is loaded with paper and pulp.

Networks The restrictions on the total network size (the largest STAN licence allows up to 32 000 links) (number of links, nodes and modes) today limit the scope of defining e.g. the road network for Sweden at detailed level compatible with the level used in the SAMPERS model system for the transport of persons (The SAMPERS model allows up to 120 000 links). This has partly to do with the fact
SAMPLAN Rapport 2001:1

40 that the goods model networks have to be extended to cover also the rest of Europe and (with less detail, other parts of the world, but also that the STAN system does not allow the size of networks that EMME/2 does.

Cost functions The cost functions aim to define the cost of using the different transport modes and transfers and how these costs vary according to the specific product transported. The general approach has been to separate the cost functions into two basic elements namely, operative costs and quality related costs (including frequency). The purpose of the model is to represent market interaction and the resulting market demand for various transport solutions. Shippers are assumed to make a rational choice between transport alternatives with different attributes, one of which is the pure price of the transport. The idea behind the structure of the cost functions, therefore, is to model the total perceived cost by the shipper. Of course, the shipper has to pay the invoice of the transporting company which is one of the cost elements considered from the shippers point of view. The shipper is also expected to consider his own capital cost of goods in transport which includes pure transport time as well as waiting time. Moreover the shipper has to consider the risks of damage and delay and the cost penalties associated with such events. All these cost items are included in the total cost function. The cost function approach that has been chosen requires many types of data which are not always easy to come-by. Price information, for example, is in most cases a guarded secret between buyer and seller resulting in negotiated contract prices for differentiated transport services. Quality aspects of transport are equally difficult to capture and generalise in such a model. It has been observed in the development process of the most recent version of the system that the challenge of cost function data collection should in the future be met by systematic surveys/analysis. In the present version of the model, data on the value of time, goods damage risk and the risk for delays have been deduced based on the value of time study for goods, which was commissioned by SAMGODS in 1998. There have been problems from time to time to expand the set of cost functions due to restrictions on the total size of the data-set defining those functions. The existing model has been rationalised to optimise the use the cost functions. This has led to a significant reduction in the total size of the data-set.

Some concluding comments on the network transport market model From an infrastructure planning perspective, the use of separate models for person and goods transport to allocate flows to networks, which in most cases are used for both persons and goods, is a problem. The problem is caused by the fact that infrastructure-level person and goods transport interact, not only in relation to congestion effects but also, albeit to a lesser extent, in situations with no or small
SAMPLAN Rapport 2001:1

41 congestion problems. To some extent these problems could be mitigated through sequential exchange of flow data between the models, but this approach is not ideal and does not occur today between the SAMGODS and SAMPERS models. In the present version of the model the local goods volumes transported within each zone are NOT assigned to the networks, i.e. the volumes exist at matrix level but not at network level. For broad strategic analyses this problem may be small, but for analyses of congested networks with mixed person and goods traffic the effects of this exclusion should also be considered. The same is true for analyses of transport chains where the connecting infrastructure and transfer costs between the different networks may be a poor proxy if the connecting network is congested.

SAMPLAN Rapport 2001:1

42

SAMPLAN Rapport 2001:1

43

7 Evaluation tools and models
This part of the model package is not yet implemented, and is currently under development. The purpose of the evaluation models and tools is to make use of information already available in the national model system as well as utilising the system’s analytical capabilities for evaluation and effect analyses. The evaluation tools will support CBA, EIA and SEA analysis of major projects and programmes as well as CBA and SEA analysis of a wide range of policy measures and general changes of subsystems of the infrastructure. The effects of infrastructure measures and certain policy measures on transport cost, transport time etc. could be calculated and presented with the appropriate regional distribution. Major transport corridors could be studied with regard to time/cost characteristics. The evaluation models comprise tools for getting access to model system data. The evaluation modules are planned to interact with other components of the national model system including the national system for passenger travel SAMPERS. The evaluation module will as much as possible draw on already existing tools. Within the framework of the planning process it is of particular importance from a practical evaluation point of view that a consistent (if not common) framework for evaluation could be devised to combine passenger and goods transport evaluations.

SAMPLAN Rapport 2001:1

44

SAMPLAN Rapport 2001:1

45

8 Application examples
This chapter, as the title says, gives examples rather than a comprehensive overview or an in depth discussion of applications of the national goods model system. The idea is to give the reader an idea of the type of analyses the system up to now has been thought suitable for. The examples below are grouped into five groups, which try to catch the emphasis of each application example, though in many instances the applications slide over into each other in the practical analysing processes. The general public availability of the model system is still unclear due to the existence of confidential information in some parts of the system. This somewhat unclear state of access has hitherto limited the use of the models by consultants.

8.1 Demand projections, scenarios and forecast
Looking back, transport growth has been a prominent feature of the development of western societies including Sweden and its neighbours during the last few decades. Transport has grown to become a phenomenon, which has profound impacts on many aspects of the modern societies. From general political and economic viewpoints, but also for the more specific problems of anticipating future transport sector challenges, the need to look ahead is felt as pressing as ever within transport and policy. The task of the forecasters has gradually become more complicated, however. Transport forecasters have been accused of using mechanistic approaches to produce forecasts of traffic growth, which themselves, as the allegations indicate, would play an important role to stimulate transport growth by providing arguments for new capacity which stimulates transport growth. This criticism has drawn more attention to the need for understanding of the interactions within the entire transport system. Attention has also been drawn to importance of the growth conditions surrounding transport, which include many transport policy issues including taxation and financing. Another important issue influencing the forecasters’ job today, is the increasing rate of change of many aspects of society - including factors thought to have a robust inherent inertia such as individual values. The times of rapid change are expected to continue at an increasing rate. These processes draw more attention to the inherent uncertainty of forecasts, and in the next step provoke questions

SAMPLAN Rapport 2001:1

46 related to all kinds of long term planning, whether such processes take place within the realms of politics or business. To my mind, these changing conditions and challenges for forecasters do only to a limited extent challenge formal modelling approaches as such. What is more and more required, however, is a very conscious mind in the use of the models. The model system (in different versions but both times including the STAN network module) has been used twice to produce reference goods transport forecast for the 10-year strategic infrastructure planning process. The first occasion was in 1995–1996 for the planning period 1998–2007 and the second in 1998–2000 for the 10-year planning period 2002–2011. This model (STAN99) was used to produce new national forecasts for 2010.7 The role of the national goods model in these applications has primarily been the following: a) to illuminate how different scenarios for the general economic development would influence transport development for modes, regions and key corridors and links b) to provide the planning processes at the state infrastructure agencies with as precise information as possible on the projected growth rates of goods transport volumes on various main categories of the networks of each mode. The precision of the model has generally not yet been considered sufficient for use for demand analysis related to specific links (for such an application, see below) c) to analyse the impact of various policy measures including infrastructure measures on goods transport volume and patterns d) to analyse how transport conditions outside Sweden influence goods transport in Sweden

8.2 Analyses of major infrastructure projects
Generally, the precision of the network model has not been considered quite sufficient for the analysis of specific corridors and links. There are though a few examples of such applications. The Bothnia railway defines a new running of the railway in the North of Sweden to complement the inland railway. The new railway would run through or close to major coastal cities along the northern part of the Swedish Bothnian coastline.8 The goods transport demand effects were analysed with the network model in the 95/96 version. However, additional calibration was necessary to obtain a reasonable precision. One of the main means used to improve the model fit was to
7

Godstransporter – Efterfrågan och utbud, Underlagsrapport till SAMPLAN, Slutrapport från arbetsgruppen för det strategiska området Det svenska och internationella godstransportsystemet ”GODIS”, november 1999 och Prognos för godstransporter 2010, SIKA-rapport 2000:7. Botniabanan – en samhällsekonomisk bedömning, SIKA 1996:1.
SAMPLAN Rapport 2001:1

8

47 subdivide the O/D matrices locally in the direct influence area of the railway. The analyses provided data on over all goods transport growth for the subsequent analyses of the costs and benefits (CBA). A particular issue, which was studied in many scenarios, was the competition between maritime transport and the new railway, and the effects of the models capacity restrictions of ports was also highlighted.

8.3 Corridor analyses
In 1995 SIKA was commissioned by the government to analyse the future needs for international transport. The study was to comprise both person and goods transport. The study was reported in August 1996.9 The national system for goods transport was used in the study to provide general information on transport growth related to Swedish foreign trade, and also to forecast the development of transport quantities per mode in Swedish foreign trade. Further the model was used to illustrate the share of total goods transport in Sweden per mode related to international transport chains, and the distribution between Swedish and foreign territories of ton-kilometres related to Swedish import/export. Another idea was to illustrate also the development for transport corridors in Swedish foreign trade, but the model, at that time, was judged to lack sufficient precision to allow such an application - however interesting this would have been because of the advent of new potential corridors making use of the Öresund and the Belt bridges. An updated version of the STAN99 model was also used for the identification and analysis of important transport corridors. This work was carried out on behalf of Godstransportdelegationen, an official delegation with the commission to develop a national goods transport strategy.10

8.4 Demand effects of transport policy measures including taxes/charges
The network model of the national goods model system allows the analyses of effects of many types of taxation/charging measures and subsidies. This is due to the fact that cost functions constitute primary building blocks of the network model. Within SIKA:s work in 1997 for the government commission on communication, the modal distribution effects of different levels of internalising external effects within the different transport modes were studied. Scenarios were constructed, which also included alternative assumptions on the charging policies in Germany.

9

Det framtida behovet av internationella transporter, SIKA rapport 1997:3. Stråkanalyser för godstransporter, SIKA Rapport 2001:1.
SAMPLAN Rapport 2001:1

10

48 The simulations illustrated the STAN network model projections of redistribution of transport flows due to the changes in taxation/charges.11 One observation is probably appropriate here. Within the framework of the total national goods model system, transport costs and capacities do not influence transport volumes and the O/D-structure of flows. These volumes and structures are assumed to be determined by other factors (see above) and not influenced by the transport system properties as such. Another example of policy orientated analysis is the Eurovinjette study12 carried out by SIKA in 2000. The network model was used to illustrate how different road network charging regimes influenced the distribution of the heavy goods traffic on different parts of the road network.

11 12

Ny kurs i trafikpolitiken, slutbetänkande av Kommunikationskommittén, SOU 1997:35. Effekter av alternativ till Eurovinjettsystemet, SIKA Rapport 2000:4.
SAMPLAN Rapport 2001:1

49

References
Anderstig, C, Lundqvist, Lena, Scheele, Siv, Prognoser över befolkning och sysselsättning, Rapport på uppdrag av SIKA, Inregia March 1999 COWI/Inregia, Udenrigshandelsmodel for Sverige, 1999-03-18 Eriksson, Jan R., VTI:s prognossystem för godstransporter, 1994-04-05 Eriksson, Jan R., Matstoms, Ylva, Metodbeskrivning av VTI:s odmatrisestimeringar 1995 avseende godstransportflöden TOS PM 1996-02-05 (uppdrag av Vägverket) Eriksson, Jan R., VTI/TPR-modellen. Ett prognossystem för godstransporter. KFB och VTI forskning och research, 31, 2000 Inregia AB, Modell för regionala prognoser av utrikeshandel. Användarhandledning Kvarnström, Christina, PM. Regionala prognoser för år 2010 och förändringar i regional efterfrågan av olika utbildningskategorier (En modellbeskrivning), Nutek Analys 1997-11-01. Lundin, Matts, Infratructure planning and Transport Policy, User perspective on STAN as an instrument for analyses, Report commissioned by SIKA from seminar in Montreal, November 25-25, Temaplan AB, December 1997 Ma Yun, McDaniel, John, Vierth, Inge, European Freight Transport Demand Model, VTI notat 56 A 1998 SIKA, Botniabanan – en samhällsekonomisk bedömning, SIKA 1996:1 SIKA, Det framtida behovet av internationella transporter, SIKA report 1997:3 SIKA, Omvärldsförutsättningar. Strategisk analys. Underlagsrapport till Samplan, November 1999 SIKA, Godstransporter – Efterfrågan och utbud, Underlagsrapport till SAMPLAN, Slutrapport från arbetsgruppen för det strategiska området Det svenska och internationella godstransportsystemet ”GODIS”, november 1999 SIKA, Effekter av alternativ till Eurovinjettsystemet, SIKA Rapport 2000:4. SIKA, Prognos för godstransporter 2010, SIKA-rapport 2000:7 SIKA Rapport, Cunningham, A , Lindkvist, A, Pettersson, A, Näringslivets transporter i Stockholms län 1998. SIKA rapport 2000:9

SAMPLAN Rapport 2001:1

50

SIKA, Stråkanalyser för godstransporter, SIKA Rapport 2001:1. SIND, ISMOD-systemets flersektormodeller av den svenska ekonomin, SIND PM 1987:32 SOU 1996:26 Inriktningen av infrastrukturplaneringen för perioden 1998-2007 SOU 1997:35 Ny kurs i trafikpoltiken. Slutbetänkande av Kommunikationskommittén. Temaplan AB, Estimering av varuvärden, Slutrapport 1999-06-23

SAMPLAN Rapport 2001:1


								
To top