This essay builds a model of the human connectivity offered by Spain’s railways, revealing the patterns between Spaniards and the democractic tension therein, with income analysis that explores the import of “Obligación de Servicio Público”. “Understanding Obligación” is the fourth essay in the sequence “Café Para Todos“, an exploration of the contemporary relationship between the railways and the people of Spain. The first essay, “Saving Ferroviarias“, reviews the broad policy context of Spain’s passenger railways, highlighting the residual tension between pre and post-democratic eras, the financial impetus to make the high speed network more viable, and the evolving policy paradigm of rationalisation. “Disassembling Trenes“, the second essay in the sequence, deconstructs Spain’s current passenger railways to expose the deceptions of AVE and nation therein. “Deconstructing Estaciones” provides a demographic analysis of Spain’s railway stations, that explores the unserved areas and probes the differences between regions. “Reanimating Regional” outlines the regional biases of Spanish railway connectivity, reassesses the role of Castilla in the national railway, and ponders the balance between actuality and perception inherent in Adolfo Suárez’s doctrine of “café para todos”.
Transport models have acquired a reputation for becoming impenetrably complicated, their results rendered as factual knowledge however internally flawed their logics actually are. Spanish policymaking has its own form of complexity, that in the relationships between people, and thus complex modelling risks being distilled down into simple statements in support of a pre-determined policy position. Instead this analysis tries to place greater emphasis on understanding, using only commonly agreed tokens (people, trains, distance), and making only practical assumptions that hopefully reflect “common sense”. To that end, a model of connectivity across Spain’s passenger railway network has been built in simple stages:
- Basic Connectivity – who is connected by train to who: A matrix of routes between municipalities with stations on which at least one train per day links the pair directly. Each pairing is multiplied by the municipal population of the destination, the result for all pairs from the origin then summed and attributed to the origin. The population of the origin municipality is added to the result, which is assumed to have perfect connectivity to itself – an assumption that only tangibly affects the overall connectivity score of the largest, and avoids cities attaining worse connectivity scores than the suburbs that connect to them (because those suburbs would gain the connectivity of the city’s population, while the city would not).
- Service Connectivity – who is connected to who by what frequency of train service: As basic connectivity above, except each route pairing is additionally multiplied by a factor representing service frequency, ( 1 – ( 1 / daily trains ) ), where daily trains is the total of both directions. This formula gives no value to the first train (which logically supposes no possibility of return), but thereafter values of each additional pair marginally, as half the value of the previous pair. Such weighting places emphasis on attaining the most basic level of service, as befits the non-urban regional networks that are the focus of this sequence of essays, while weighting high frequency metro services very marginally indeed.
- Hinterland Connectivity – who is connected to who by what frequency of train service, but where people use the station with the best ratio of connectivity to proximity, not necessarily the nearest station: For every municipality (both with and without stations), calculate the straight line distance from the centroid of the origin municipality to all municipalities with stations within 150 kilometres, and then find the municipality with the highest ( Service Connectivity of municipality * ( 1 / distance to municipality in kilometres ) ), assigning that calculated value to the initial origin municipality. This gravity model reflects the tendency of municipalities with many more trains to attract passengers from more distant markets. The assumed distance tapper is approximate, but generally succeeds in both re-assigning relatively poorly served municipalities that are close to a much better served neighbour (for example, a municipality 10 kilometres away would need to offer at least 10 times better Service Connectivity than a local station), and assigning people in municipalities without a station to the most attractive station in their proximity (the best served relative to distance). Every place in Peninsula Spain is within 100 kilometres of at least one station, and the 150 km buffer ensures a range of stations are considered, including provincial capitals.
- Connectivity Index – how does this connectivity compare to that of the average Spaniard: Hinterland Connectivity is expressed as a percentage of the average for all the municipalities scored (in the base case, those within 150 kilometres of at least one station, almost the entire population of Spain), with that average weighted by population: For example, if Madrid represented 7% of the total population of scored municipalities, Madrid’s score would count towards 7% of the overall average. This population weighting serves only to distribute the resulting indices around a meaningful average, where a connectivity index of 100 is what the average Spaniard (with a station within 150 kilometres) would obtain. The indices are thus entirely relative to other members of the population, reflecting policy themes of balance and equality.
The underlying dataset used is that described in the earlier essay, “Disassembling Trenes” – essentially a frequency-based matrix of all non-tourist rail passenger services within Spain on Friday 20 July 2018, alongside the municipal Padrón from the start of 2017. In the interest of transparency and the benefit of any other interested researchers, the raw network analysed is available in Geojson format – as is, neither supported nor maintained, and obviously without the associated computation described above. The analysed network can also be explored visually using Aquius. Frequency-based connectivity models are far faster to compute than those that process detailed schedules, and also far easier to edit – allowing the impact of a change in service pattern to be tested conceptually, without providing the kind of detailed schedule operational planners only produce after deciding to implement a network change. That flexibility to use connectivity models for network design was unfortunately lost during the development of these techniques in Britain in the early 2000s, ultimately because central government’s desire to understand connectivity surpassed their desire to assist those who might improve it, a rationale subsequently perpetuated in academia. Yet basic connectivity models remain powerful tools for both grand strategy and network tinkering, and in an environment with little or no interchangeable electronic schedule data (such as Spain) their deployment can add insight where otherwise there is none: Spanish railway interests produce plenty of good technical information, but remarkably little relates services to people, and much of what does is pre-occupied with appeasing the god of high speed.
The aim of this analysis is to understand the broad patterns by focusing on the key relationships, not to attempt to model every conceivable detail: Journey distance is ignored, but in practice the pattern of direct routes will tend to constrain destinations, while the tendency to lower frequencies on longer distance journeys renders remote destinations with poorer Service Connectivity. The availability of realistic return journeys is also ignored, but the probability of such return journeys is inherent in the overall service frequency. Interchange between trains is ignored, since as discussed in the essay “Disassembling Trenes“, interchange is not a dominant behaviour in most of Spain’s non-urban regional networks. Local interchange, especially between suburbs and better-served city centres, is factored into Hinterland Connectivity – the reduced connectivity with distance may be assumed a crude proxy for the reduced attractiveness of interchange. Hinterland Connectivity similarly manages the few branchlines (such as FEVE‘s Collanzo line in the Asturias) whose services require interchange to reach any major destination. Hinterland Connectivity takes no specific account of the availability of alternative modes of transport to reach the railway network, although its tendency is to link groups of people in relatively close proximity, groups who tend to establish transport links between one another. The factors used in Service and Hinterland Connectivity calculations may seem rather arbitrary – and would be for detailed microsimulation – but their use here is in the production of strategic aggregated comparators, where broad consistency of approach is more important than precise local calibration.
The connectivity of the islands and north-African autonomous cities – Balears, Canarias, Ceuta and Melilla – cannot be adequately reflected in a railway model of Spain, since even islands with railways can provide no direct connections beyond their own island. Overall Connectivity Indices include island and autonomous city municipalities within 150 kilometres of a municipality with a station, so can affect the overall average score and thus the index, but in practical terms the results for these municipalities are spurious and cannot be compared to Peninsula Spain. Analysis of the connectivity of (only) non-Renfe operators has a similar weakness because the networks of these operators do not generally connect to one another – for example, however well FGV serves Valenciana, it cannot be fairly compared to a national network that links Valenciana to other parts of Spain. Non-Renfe operators can be important to specific local municipalities, and are thus important within the most local analysis, but add only marginally to the overall connectivity of regions: Even in the provinces best served by non-Renfe operators, Madrid and Barcelona, such operators only add about 10% to the overall Connectivity Index. Lleida’s extremely high connectivity poses a particular challenge to the Pobla de Segur route, which offers a relatively infrequent service whose only major destination is Lleida, and thus provides far less direct connectivity than Lleida herself. While the route is modelled, the connectivity it offers is usurped by Hinterland Connectivity at many place close to Lleida, and even at Pobla de Segur the railway offers only a marginal connectivity advantage, hence is almost invisible in the Connectivity Indices for local municipalities. Analysis of Renfe’s “commercial” non-OSP products ignores local OSP journeys delivered as shared seats on commercial services, leaving those commercial services only to stop for the benefit of longer-distance passengers. This is an accurate reflection on current operations, but produces local quirks such as removing one of the links between Badajoz and Cáceres while retaining that between Badajoz and Madrid – with the net result of reducing slightly the overall (commercial) Connectivity Index of Badajoz. Such reduced connectivity is, however, a reasonable reflection on the marginal nature of the commercial service provided.