Figure 1. The SASI model


European Developments
Here assumptions about European developments are entered that are processed by the subsequent submodels. European developments include assumptions about the future performance of the European economy as a whole and the level of immigration and outmigration across Europe's borders. They serve as constraints to ensure that the regional forecasts of economic development and population are consistent with external developments not modelled. Given the expected rapid population growth and lack of economic opportunity in many origin countries, total European immigration will be largely a function of immigration policies by national governments of the countries of the European Union. Another relevant European policy field are transfer payments by the European Union via the Structural Funds or the Common Agricultural Policy or by national governments to assist specific regions, which, because of their concentration on peripheral regions, are responsible for a sizeable part of their economic growth. The last group of assumptions are those about policy decisions on the trans-European networks. As these are of focal interest in SASI, they are modelled with considerable detail. A network scenario is a time-sequenced investment programme for addition, upgrading or closure of links of the road, rail or air networks. Besides a 'baseline' scenario several TETN scenarios will be specified.

Regional Accessibility
This submodel calculates regional accessibility indicators expressing the locational advantage of each region with respect to relevant destinations in the region and in other regions as a function of travel time or travel cost (or both) to reach these destinations by the strategic road, rail and air networks.

Regional GDP
This is the core submodel of the SASI model. It calculates a forecast of gross domestic product (GDP) per capita by industrial sector (agriculture, manufacturing, services) generated in each region as a function of endowment indicators and accessibility. Endowment indicators are indicators measuring the suitability or capacity of the region for economic activity. Endowment indicators may include traditional location factors such as availability of skilled labour and business services, capital stock (i.e. production facilities) and intraregional transport infrastructure as well as 'soft' location factors such as indicators describing the spatial organisation of the region, i.e. its settlement structure and internal transport system, or institutions of higher education, cultural facilities, good housing and a pleasant climate and environment. Accessibility indicators are derived from the Regional Accessibility submodel. In addition to endowment and accessibility indicators, monetary transfers to regions by the European Union such as assistance by the Structural Funds or the Common Agricultural Policy or national governments are considered, as these account for a sizeable portion of the economic development of peripheral regions. The results of the regional GDP per capita forecasts are adjusted such that the total of all regional forecasts multiplied by regional population meets the exogenous forecast of economic development (GDP) of Europe as a whole by the European Developments submodel.

Regional Employment
Regional employment is derived from regional GDP by exogenous forecasts of regional labour productivity by industrial sector (GDP per worker) modified by effects of changes in regional accessibility.

Regional Population
Regional population changes due to natural change and migration. Births and deaths are modelled by a cohort-survival model subject to exogenous forecasts of regional fertility and mortality rates. Interregional migration within the European Union is modelled in a simplified migration model as annual net migration as a function of regional unemployment and other indicators expressing the attractiveness of the region as a place of employment and a place to live, whereas immigration to and outmigration from the European Union are modelled separately. The migration forecasts are adjusted to comply with total European immigration and outmigration forecast by the European Developments submodel and the limits on immigration set by individual countries. In addition educational attainment, i.e. the proportion of residents with higher education, is forecast as a function of national education policy.

Regional Labour Force
Regional labour force is derived from regional population and exogenous forecasts of regional labour force participation rates modified by effects of regional unemployment.

Socio-economic Indicators
Total GDP and employment are related to population and labour force by calculating total regional GDP per capita and regional unemployment. Accessibility, besides being a factor determining regional production, is also considered a policy-relevant output of the model. In addition, equity or cohesion indicators describing the distribution of accessibility, GDP per capita and unemployment across regions are calculated.


3.2 Space and Time

The SASI model forecasts socio-economic development in the 201 regions at the NUTS-2 level defined for SASI for the fifteen EU countries. These are the 'internal' regions of the model. The 27 regions defined for the rest of Europe are the 'external' regions which are used as additional destinations when calculating accessibility indicators. The four regions representing the rest of the world are not used.

The spatial dimension of the system of regions is established by their connection via networks. In SASI road, rail and air networks are considered. The 'strategic' road and rail networks used in SASI are subsets of the pan-European road and rail networks developed by IRPUD and recently adopted for the GISCO spatial reference database of Eurostat. The 'strategic' road and rail networks contain all TETN links laid down in Decision No. 1692/96/CE of the European Parliament and the Council (European Communities, 1996) and the east European road and rail corridors identified by the Second Pan-European Transport Conference in Crete in 1994 as well as additional links selected for connectivity reasons. The SASI system of regions and the strategic networks used in SASI are also used in the concurrent DGVII projects STREAMS, EUNET and STEMM.

The temporal dimension of the model is established by dividing time into discrete time intervals or periods of one year duration. By modelling relatively short time periods both short- and long-term lagged impacts can be taken into account. The base year of the simulations is 1981 in order to demonstrate that the model is able to reproduce the main trends of spatial development in Europe over a significant time period of the past with satisfactory accuracy. The forecasting horizon of the model is 2016.


3.3 Model Data

Two major groups of data can be distinguished: data required for running the model and data needed for the calibration and validation of the model:


Simulation Data

Simulation data are the data required to perform a typical simulation run. They can be grouped into base year data and time series data. Base year data describe the state of the regions and the strategic rail and road networks in the base year 1981. Base year data are either regional or network data. Time series data describe exogenous developments or policies defined to control or constrain the simulation. They are either collected or estimated from actual events for the time between the base year and the present, or are assumptions or policies for the future. Time series data must be defined for each simulation period, but in practice may be entered only for specific (not necessarily equidistant) years, with the simulation model interpolating between them.

Base year data
   Regional data (201 EU regions)
       Regional GDP per capita by industrial sector in 1981
       Regional labour productivity (GDP per worker) by industrial sector in 1981
       Regional population by five-year age group and sex in 1981
       Regional educational attainment in 1981
       Regional labour force participation rate by sex in 1981
   Network data (pan-Europe)
       Link data of strategic road network in 1981
       Link data of strategic rail network in 1981
       Link data of air network in 1981

Time series data
   European data (EU)
       Total European GDP by industrial sector, 1981-2016
       Total European immigration and outmigration, 1981-2016
       National data (15 EU countries)
       National GDP per worker by industrial sector, 1981-2016
       National fertility rates by five-year age group and sex, 1981-2016
       National mortality rates by five-year age group and sex, 1981-2016
       National immigration limits, 1981-2016
       National educational attainment, 1981-2016
       National labour force participation by sex, 1981-2016
   National data (23 non-EU countries)
       National population, 1981-2016
       National GDP, 1981-2016
       Regional data (201 EU regions)
       Regional endowment factors, 1981-2016
       Regional transfers, 1981-2016
       Network data (pan-Europe)
       Changes of node and link data of strategic road network, 1981-2016
       Changes of node and link data of strategic rail network, 1981-2016
       Changes of node and link data of air network, 1981-2016


Calibration/Validation Data

The regional production function in the GDP submodel is the only model function calibrated using statistical estimation techniques. All other model functions are validated by comparing the output of the whole model with observed values for the period between the base year and the present. The following data for calibration/validation are required:

Calibration data
   Regional data (201 EU regions)
       Regional GDP per capita by industrial sector in 1981, 1986, 1991
       Regional endowment factors in 1981, 1986, 1991
       Regional labour force in 1981, 1986, 1991
       Regional transfers in 1981, 1986, 1991
       Regional net migration in 1981, 1986, 1991
       Regional unemployment rates in 1981, 1986, 1991
   Network data (pan-Europe)
       Node and link data of strategic road network in 1981, 1986, 1991
       Node and link data of strategic rail network in 1981, 1986, 1991
       Node and link data of air network in 1981, 1986, 1991

Validation data
   Regional data (201 EU regions)
       Regional population (by age and sex) in 1981, 1986, 1991, 1996
       Regional GDP (by industrial sector) in 1981, 1986, 1991, 1996
       Regional labour force (by sex) in 1981, 1986, 1991, 1996
       Regional employment (by industrial sector) in 1981, 1986, 1991, 1996
       Regional unemployment rate in 1981, 1986, 1991, 1996


Data Sources

The Eurostat data base REGIO has been identified as the primary data input to the project as a whole, as it is the main official source of regional data that is provided on a regular basis and in a harmonised framework (Masser et al., 1997).

Data problems identified were large differences in the size of regions, changes in region boundaries and the creation of new regions all resulting in outliers and gaps in the data set. Data coverage was found to be very poor for the new member states Austria, Finland and Sweden and the new German Länder.

Missing data, in particular for the base year 1981, are estimated or derived from other data sources such as national statistical offices. Although REGIO covers a considerable amount of the data required, calculation of regional endowment factors requires other data sources, as does the information needed for the European Developments submodel.

Network data used for SASI are the 'strategic' road, rail and air networks described in Section 3.2. For past years they contain information on the historical development of transport infrastructure, whereas for future years they represent the transport network investments and transport system improvements to be investigated. Travel cost is presently represented by travel time only; in future applications also generalised travel cost consisting of a combination of travel time, travel cost and mode-specific inconvenience will be used.


3.4 Model Output

Output of the model are indicators measuring socio-economic and spatial impacts of the simulated policies. Three groups of output indicators were defined:

-

Gross domestic product (GDP) was chosen to represent the economic performance of a region. GDP per capita allows to draw conclusions on regional income levels. Despite its well-known theoretical and methodological drawbacks this indicator continues to be the most commonly used indicator of regional economic efficiency.

-

The unemployment rate is used to indicate the social condition of a region. This indicator, too, presents measurement problems because there exist large differences in the definition of unemployment in European countries. Nevertheless unemployment remains the most widely used social indicator and is easily related to policy goals.

-

In addition to the above regional indicators macro indicators expressing the distribution of GDP and unemployment across regions are used as indicators of cohesion between the regions of the European Union. Cohesion indicators inform about the degree of spatial concentration or dispersion of GDP or unemployment and if applied to modelled policies show whether the implementation of a policy would contribute to the political goal of reducing socio-economic disparities or not.

Using these indicators it can be shown that cohesion and integration policies of the European Union have not always been successful. In fact there is no evidence that regional income differences in Europe have been reduced during the 1980s. In terms of regional unemployment, the gap between successful and declining regions even seems to have widened (Bökemann et al., 1997).


4. Accessibility

The important role of transport infrastructure for regional development is one of the fundamental principles of regional economics. In its most simplified form it implies that regions with better access to the locations of input materials and markets will, ceteris paribus, be more productive, more competitive and hence more successful than more remote and isolated regions (see Linneker, 1997).

However, the impact of transport infrastructure on regional development has been difficult to verify empirically. There seems to be a clear positive correlation between transport infrastructure endowment or the location in interregional networks and the levels of economic indicators such as GDP per capita (e.g. Biehl, 1986; 1991; Keeble et al., 1982, 1988). However, this correlation may merely reflect historical agglomeration processes rather than causal relationships effective today (cf. Bröcker and Peschel, 1988). Attempts to explain changes in economic indicators, i.e. economic growth and decline, by transport investment have been much less successful. The reason for this failure may be that in countries with an already highly developed transport infrastructure further transport network improvements bring only marginal benefits. The conclusion is that transport improvements have strong impacts on regional development only where they result in removing a bottleneck (Blum, 1982; Biehl, 1986; 1991).

While there is uncertainty about the magnitude of the impact of transport infrastructure on regional development, there is even less agreement on its direction. It is debated whether transport infrastructure contributes to regional polarisation or decentralisation. Some analysts argue that regional development policies based on the creation of infrastructure in lagging regions have not succeeded in reducing regional disparities in Europe (Vickerman, 1991a), whereas others point out that it has yet to be ascertained that the reduction of barriers between regions has disadvantaged peripheral regions (Bröcker and Peschel, 1988). From a theoretical point of view, both effects can occur. A new motorway or high-speed rail connection between a peripheral and a central region, for instance, makes it easier for producers in the peripheral region to market their products in the large cities, however, it may also expose the region to the competition of more advanced products from the centre and so endanger formerly secure regional monopolies (Vickerman, 1991b).

While these two effects may partly cancel each other out, one factor unambiguously increases existing differences in accessibility. New transport infrastructure tends to be built not between core and periphery but within and between core regions, because this is where transport demand is highest (Vickerman, 1991a). It can therefore be assumed that the trans-European networks will largely benefit the core regions of Europe.

These developments have to be seen in the light of changes in the field of transport and communications which will fundamentally change the way transport infrastructure influences spatial development (Masser et al., 1992). Several trends combine to reinforce the tendency to reduce the impacts of transport infrastructure on regional development:

-	An increased proportion of international freight comprises high-value goods for which transport cost is much less than for low-value bulk products. For modern industries the quality of transport services has replaced transport cost as the most important factor.

-

Transport infrastructure improvements which reduce the variability of travel times, increase travel speeds or allow flexibility in scheduling are becoming more important for improving the competitiveness of service and manufacturing industries and are therefore valued more highly in locational decisions than changes resulting only in cost reductions.

-	Telecommunications have reduced the need for some goods transports and person trips, however, they may also increase transport by their ability to create new markets.

-

With the shift from heavy-industry manufacturing to high-tech industries and services other less tangible location factors have come to the fore and have at least partly displaced traditional ones. These new location factors include factors related to leisure, culture, image and environment, i.e. quality of life, and factors related to access to information and specialised high-level services and to the institutional and political environment.

On the other hand, there are also tendencies that increase the importance of transport infrastructure:

-	The introduction of totally new, superior levels of transport such as the high-speed rail system may create new locational advantages, but also disadvantages for regions not served by the new networks.

-	Another factor adding to the importance of transport is the general increase in the volume of goods movements (due to changes in logistics such as just-in-time delivery) and travel (due to growing affluence and leisure time).

Both above tendencies are being accelerated by the increasing integration of national economies by the Single European Market, the ongoing process of normalisation between western and eastern Europe and the globalisation of the world economy.

The conclusion is that the relationship between transport infrastructure and economic development has become more complex than ever. There are successful regions in the European core confirming the theoretical expectation that location matters. However, there are also centrally located regions suffering from industrial decline and high unemployment. On the other side of the spectrum the poorest regions, as theory would predict, are at the periphery, but there are also prosperous peripheral regions such as the Scandinavian countries. To make things even more difficult, some of the economically fastest growing regions are among the most peripheral ones.

The central task of SASI is therefore to identify the way transport infrastructure contributes to regional economic development in different regional contexts. This means to develop indicators measuring not infrastructure investments as such but the benefit they bring to firms and households in the regions by more capacity, higher speeds, better quality and more reliable transport. These indicators are called accessibility.


4.1 Basic Accessibility Indicators

Accessibility is the main 'product' of a transport system. It determines the locational advantage of a region relative to all regions (including itself). Indicators of accessibility measure the benefits households and firms in a region enjoy from the existence and use of the transport infrastructure relevant for their region.

Accessibility indicators can be defined to reflect both within-region transport infrastructure and infrastructure outside the region which affect the region.

Simple accessibility indicators consider only intraregional transport infrastructure expressed by such measures as total length of motorways, number of railway stations (e.g. Biehl, 1986; 1991) or travel time to the nearest nodes of interregional networks (e.g. Lutter et al., 1993). While this kind of indicator may contain valuable information about the region itself, they fail to recognise the network character of transport infrastructure linking parts of the region with each other and the region with other regions.

More complex accessibility indicators take account of the connectivity of transport networks by distinguishing between the network itself, i.e. its nodes and links, and the activities or opportunities that can be reached by it (cf. Bökemann, 1982). In general terms, accessibility then is a construct of two functions, one representing the activities or opportunities to be reached and one representing the effort, time, distance or cost needed to reach them:




where A_i is the accessibility of region i, W_j is the activity W to be reached in region j, and c_ij is the generalised cost of reaching region j from region i. The functions g(W_j and f(c_ij) are called activity functions and impedance functions, respectively. They are associated multiplicatively, i.e. are weights to each other. That is, both are necessary elements of accessibility. A_i is the accumulated total of the activities reachable at j weighted by the ease of getting from i to j.

It is easily seen that this is a general form of potential, a concept dating back to Newton's law of gravitation and introduced into regional science by Stewart (1947). According to the law of gravitation the attraction of a distant body is equal to its mass weighted by a decreasing function of its distance. Here the attractors are the activities or opportunities in regions j (including region i itself), and the distance term is the impedance c_ij.

The interpretation here is that the greater the number of attractive destinations in regions j is and the more accessible regions j are from region i, the greater is the accessibility of region i. This definition of accessibility is referred to as destination-oriented accessibility. In a similar way an origin-oriented accessibility can be defined: The more people live in regions j and the more easily they can visit region i, the greater is the accessibility of region i. Because of the symmetry of most transport connections, destination-oriented and origin-oriented accessibility tend to be highly correlated.

Different types of accessibility indicators can be constructed by specifying different forms of functions g(W_j) and f(c_ij). Table 1 shows the three most frequently applied combinations of g(W_j) and f(c_ij), where W_min and c_max are constants and a and b are parameters.


Table 1. Typology of accessibility indices

Type of accessibility	Activity function	Impedance function
Travel cost: accumulated travel cost to a set of activities
Daily accessibility: accumulated activities in a given travel time
Potential: accumulated activities weighted by a function of travel cost

4.2 Disaggregate Accessibility Indicators

Virtually all accessibility indicators used so far have concentrated on network nodes or centroids representing cities or regions and so have ignored the internal spatial organisation within regions.

To overcome this problem, Spiekermann and Wegener developed spatially disaggregate accessibility indicators using raster-based GIS technology (Spiekermann and Wegener, 1994; 1996, Vickerman et al., 1997). By this method the raster structure is applied to represent a quasi-continuous activity surface of Europe. As no raster data for Europe are available, synthetic raster data are generated using microsimulation in combination with a raster-based GIS. For that purpose the European territory is disaggregated to some 70,000 raster cells of 10 kilometres width. Accessibility is calculated by using each raster cell both as origin and destination, i.e. by generating a 70,000 by 70,000 origin-destination matrix. The results are accessibility values for all raster cells, which are then aggregated to regions. In this respect the method follows the suggestion by Newman and Vickerman (1993) that accessibility models should be more disaggregate in spatial resolution, economic activities and transport modes. The method is described in Schürmann et al. (1997).

One effective way of representing accessibility indicators is to display them as three-dimensional accessibility surfaces. The elevation of the surface at each point indicates the magnitude of accessibility at that point. To allow comparisons between different surfaces, surfaces to be compared are drawn to the same vertical scale. Accessibility surfaces are presented here for daily accessibility by rail to population.

Figure 2 shows daily accessibility without high-level networks. As indicated above, in this case an average speed for air-line distances of 30 km/h is assumed. This means that in the given maximum travel time of five hours destinations within a radius of 150 km are included. The no-network alternative can be considered as local or regional potential which has to be distinguished from self-potential. The destination activity is population. Consequently high-density regions, e.g. regions in south-east England, Belgium and the Netherlands, the western parts of Germany and the northern parts of Italy have the highest local potential. Remarkably not London or Paris but Belgium and the Rhine-Ruhr region seem to have the highest daily accessibility. But also the local potentials of spatially isolated but large agglomerations such as Madrid, St. Peterburg or Moscow and their hinterlands seem to be substantial.

Figure 2. Population reachable in five hours without interregional networks


Figure 3 presents daily accessibility by rail for 1996. Now the combined effects of high density and interregional infrastructure become visible. Significant disparities in accessibility appear. The highest daily accessibility values are found in France, southern England, Belgium and the Netherlands, Germany, Switzerland, Austria and northern Italy. Again not London or Paris but Belgium and north-western Germany seem to have the highest daily accessibility. There is a sharp decline from these areas towards Scandinavia, eastern and south-eastern Europe, southern Italy, the Iberian peninsula and Ireland. However, even in the high-accessibility regions there are large differences in daily accessibility between city centres (expressed as 'peaks' in the accessibility surface) and their hinterlands (expressed as 'valleys') as accessibility decreases from the nodes in the high-speed rail network to the more remote locations at the fringe of their catchment areas.

Figure 3. Population reachable in five hours by rail in 1996



5. Accessibility Indicators and Regional Economic Development

As a preparatory exercise for the calibration of the multivariate regional production function to be used in the SASI model, a tentative assessment of the relationship between accessibility and regional economic development was performed by correlating selected accessibility indicators with GDP per capita for 1981 and 1991.

Not surprisingly, the analysis results in a rather low correlation between accessibility and GDP per capita reflecting the well-known fact that location is only one factor of regional economic development. However, by temporarily eliminating characteristic groups of outlier regions and nonlinear transformation of accessibility significant increases in the coefficient of determination can be obtained. Table 2 summarises coefficients of determination between the accessibility indicators of Table 1 and GDP per capita.


Table 2. Correlation (r²) between accessibility indicators and GDP per capita

		GDP per capita, 1981	GDP per capita, 1991
Average travel cost to cities > 250,000	Road 1996 Rail 1981 Rail 1996	0.51 0.60 0.58	0.48 0.57 0.57
Average travel cost to cities > 1,000,000	Road 1996 Rail 1981 Rail 1996	0.41 0.52 0.52	0.44 0.54 0.54
Weighted average travel cost to cities > 250,000	Road 1996 Rail 1981 Rail 1996	0.49 0.59 0.57	0.47 0.56 0.57
Weighted average travel cost to cities > 1,000,000	Road 1996 Rail 1981 Rail 1996	0.46 0.56 0.55	0.45 0.54 0.55
Daily accessibility to population **0.4	No network Road 1996 Rail 1981 Rail 1996	0.30 0.36 0.43 0.47	0.21 0.25 0.29 0.33
Change in daily accessibility to population **0.3	1981 - no network 1996-1981	0.52	0.40
Daily accessibility to GDP **0.4	No network Road 1996 Rail 1996	0.37 0.40 0.50	0.32 0.32 0.32
Potential accessibility to population **0.4	No network Road 1996 Rail 1981 Rail 1996	0.34 0.39 0.47 0.50	0.23 0.27 0.32 0.35
Change in potential accessibility to population **0.3	1981 - no network 1996-1981	0.57	0.48
Potential accessibility to GDP **0.4	No network Road 1996 Rail 1981	0.43 0.45 0.55	0.36 0.35 0.43