(Continued from Previous Page)
This article first appeared in the newly released book from NERA Economic Consulting, 'The Line in the Sand: The Shifting Boundary Between Markets and Regulation in Network Industries'. With a foreword by Alfred E Kahn.
D. Using Market Mechanisms to Allocate Slots
Market mechanisms can be used, in theory, to achieve an efficient allocation of airport slots. They may take the form of a primary mechanism replacing the administrative allocation with a payment scheme,11 and/or they may take the form of a secondary mechanism, where airlines trade slots with each other or with third parties once the primary allocation has been made.
1. Primary Allocations
a. Posted prices
Under a system of posted prices, the coordinator or airport would levy a charge for each slot. The charges would be fixed according to a transparent system of tariffs published some months prior to the start of the season. This would reduce the extent of excess demand and ensure that sought-after slots were not allocated to low-value services. Airlines would bear the charge for slots allocated to them, irrespective of whether the slots were actually used, thus also alleviating the problem of airlines failing to use the scarce slots that have been allocated to them.12
The system would require airport operators to forecast demand some months in advance, and in the first few seasons there would be little information about the way airlines would be likely to respond to higher prices. To reduce the risk of setting prices too high and slots remaining unsold, prices may be deliberately set on the low side. Hence prices might still fail to clear the market and there would still be excess demand for some slots.
This mechanism would be relatively straightforward to implement and would also provide incentives for airlines to use slots efficiently. But there is a risk that such charges could lead to disputes, challenges, and possible retaliation by states that have not implemented a similar system.
Under this mechanism, airlines would bid for the slots they required, and the coordinator would allocate slots to the highest bidders. The auction could be restricted to pool slots or to a proportion of total slots, with the remainder allocated using grandfather rights. Auctions are widely used in other sectors and have been considered by the Federal Aviation Administration as a possible means of slot allocation.13
In theory, large-scale auctions could achieve the most efficient allocation of slots possible and would have a relatively early impact. In practice, however, unlike telecommunications and other licences, auctions would be highly complex for both the organisers and the airlines. Airline bidding strategies would be unavoidably complicated. Airport slots at different times may not be good substitutes for each other, and there are significant demand interdependencies. At a minimum, airlines would require a departure slot a certain length of time after the arrival slot, to allow for efficient turnaround. It may wish to schedule a frequent service, with several evenly spaced flights a day. Moreover, if the destination airport is also congested, the airline would need to coordinate its bidding strategy for the two airports concerned. Thus, barriers to participation and the strong possibility of administrative error could undermine the scope for efficient allocation of slots.
Unless such auctions were restricted to pool slots, they might have a disruptive impact on airline schedules, and co-ordination problems might occur because of the need to hold several auctions (for slots at different airports) at the same time. They might also provoke challenges and retaliation by states that did not hold such auctions, and they would be strongly opposed by many airlines because they could involve the suppression of grandfather rights.
2. Secondary Trading
Secondary trading occurs following the initial allocation of slots by airport coordinators. Exchanging slots with other airlines allows carriers to adapt their schedules and make better use of their overall slot portfolio. It also safeguards them against the risk of ending up with slots they cannot use. It thus complements administrative allocations and auctions that are applied only to a proportion of slots. Trading typically consists of a financial transaction between airlines or their agents and may form part of a larger package of cooperation, often within the context of an airline alliance, which removes the constraint that the slots be of equivalent worth. It has been practiced for some years at high density airports in the US and to a limited extent in the UK.
The risk in bilateral negotiations is that potential deals might not take place, either because buyers and sellers could not identify each other or because airlines were reluctant to sell slots to their competitors. The first of these problems might be reduced by independent agents acting as facilitators and monitoring each airline’s willingness to buy or sell slots. Concern that airlines would be reluctant to sell to competitors, however, is reinforced by the US experience of secondary trading. The high-density airports where the systems operated were characterised by high slot holdings of incumbent airlines and little new entry. This outcome could have resulted either from the slot allocation regime or might simply reflect efficient slot allocation. As most potential slot sellers will be airlines wishing to reduce their presence at the concerned airport, they would not be competing directly with airlines who wished to increase theirs, and therefore should not be inhibited about releasing slots.
Secondary trading is likely to have low implementation costs and is unlikely to interfere with existing slot allocation and scheduling procedures. But because airlines are confronted only with an opportunity cost rather than a cash outflow, the response in some cases might be delayed or might not occur at all. Secondary trading might be less successful than primary trading mechanisms in promoting a more efficient use of slots.
E. Impacts of Market Mechanisms
NERA considered the potential benefits of moving towards an efficient system for allocating slots. Given data constraints, NERA measured efficiency gains by estimating the increase in passengers using the airport that may result from the introduction of market mechanisms.
The estimate was derived by considering five European airports in detail and by segmenting demand for slots into eight categories (for example, hub carrier short haul, hub carrier long haul, low-cost carrier, charter). Demand was forecast for each of these categories, and NERA derived an elasticity of demand with respect to the charge for a slot for each of the eight categories by taking into account, inter alia, airport charges as proportion of total flight costs. The slot price was then increased until demand equalled supply, and the change in the mix of flight categories was used to determine average passenger loadings and hence the number of passengers using the airport. We also used cross-elasticities to determine the extent to which flights might be rescheduled to use lower-cost slots at off-peak times.
Overall, we estimated that the number of passengers using major airports in Europe would increase by 7.3 percent as a result of using an ideal market mechanism to allocate the slots. At highly congested airports, the increase would be 6.3 percent – slots are already intensively used throughout the day at such airports, so the main change would come from the mix of aircraft using the slots. At airports that were congested at peak times only, the number of passengers would increase by 8.5 percent – resulting from a change in the mix of aircraft using slots, from more use at off-peak times, and also from fewer slots being allocated but then remaining unused.
We then estimated how these changes might differ for each of the market mechanisms. As shown in Table 2, secondary trading may be less effective than an ideal market mechanism because airlines may not be inclined to sell slots even though they would profit from doing so. Posted prices may not be fully efficient because they would be prone to forecasting errors. Auctions may also fall short because of the difficulty of bidding for the full set of slots required to fit particular schedules. In the low case for auctions, the efficiency gains may be marginal because of the significant risk of airlines making major bidding errors or being unwilling to participate.
We concluded that these mechanisms increase the concentration of airlines using the airport as a hub. This reflects both the hub airline’s tendency to value the slots more than other airlines – at least in part reflecting an efficient allocation – and the fact that it would be able to use the mechanisms more effectively. For example, potential trading partners may approach the hub airline first. Overall, we concluded that each of the market mechanisms would deliver material benefits, though auctions may be difficult to adopt in practice, and mechanisms that transferred wealth from airlines to other parties would be subject to legal dispute.
Summary of Main Properties of Market Mechanisms for Airport Slots
IV. THE ALLOCATION AND PRICING OF RAIL INFRASTRUCTURE
A. Introduction: Rail Infrastructure Allocation
The problems of how to allocate capacity on the rail network among competing users, and how to charge for the use of the network, have assumed greater importance in recent times. For many years, such problems rarely if ever arose because most rail services were provided by vertically-integrated organisations that were responsible for providing both the rail network and the trains that ran on it. Where trains did run on other companies’ networks (for example, in the US and Canada), the commercial arrangements were often negotiated between the parties with relatively little regulatory control over tariffs or the terms and conditions of access.
More recently, vertical separation between infrastructure provision and train operation has become more common. This has occurred in two main situations. Some countries (including Australia, Sweden, and the UK) have voluntarily implemented vertical separation, usually as part of a more general restructuring of the rail industry, which often also involved some private sector participation. Vertical separation can facilitate competition in downstream markets, either competition within the market where different train operators compete directly with each other, or more usually, competition for the market, where different organisations bid for time-limited franchise contracts or similar rights to operate certain train services. Other countries have introduced vertical separation, in some cases somewhat reluctantly, in order to comply with European Union directives that aim to make it easier for rail services (especially freight services) to operate across national boundaries and to liberalise rail freight markets.
Vertical separation raises important questions about (1) the allocation and pricing of access to the rail network and (2) the pricing and capacity allocation principles that might be applied where an infrastructure manager is supplying network access to a number of independent train operators. The manager must develop a common charging and capacity allocation framework that can be applied to all train services using the network.
This is somewhat different from the situation that occurs where a vertically integrated rail organisation allows other train operators to use its network. If such access is mandated, difficult issues arise in relation to the pricing of such access (since the vertically integrated supplier may be providing network access to train operators that compete with it in downstream markets). Indeed, the "efficient component pricing rule," which has been proposed and to some extent used to determine interconnection charges between telecom operators, was originally developed in the context of rail infrastructure access.
But such situations have not generally arisen in rail industries. Where the infrastructure manager is vertically integrated (i.e., it also operates trains), the requirements to provide network access to potential competitors are often weak and sometimes voluntary (as in the US and Canada). In such cases, there is often considerable flexibility about charging principles (for example, the application of stand-alone cost ceilings, which in practice permit a wide range of charges to be applied below the ceiling). And where the infrastructure manager is subject to stronger requirements to provide network access to its potential competitors, it is often required also to deal with its "own" train operator on an arms-length basis and subject to the same terms and conditions as it applies to independent train operators. So all train operators pay access charges.
B. The Problem of Allocating Rail Network Capacity
Especially where the infrastructure manager is either independent or required to treat all train operators equally,14 and when several train operators are seeking access to the network and not all their requirements can be met, it may be necessary to establish rules or procedures about how to allocate capacity. However, deciding between competing demands for network capacity can be a difficult exercise, not least because of the complexities arising where operators want to use the same network facilities to run different types of service.
One problem is that rail network capacity itself is difficult to define and measure. The number of trains that can use a particular part of the network depends on technical constraints such as the minimum separation (or "headway") that must be maintained between trains, which in turn depends on factors such as the signalling equipment installed on the route. Then, within these constraints the number of trains that can be accommodated will depend on the precise nature of the traffic seeking to use the route and the order in which the trains are scheduled.
Figures 3 to 6 illustrate the problems with simple "train graphs." The vertical axis refers to physical locations along the route, the horizontal axis refers to time, and each line shows the planned path of an individual train. These graphs show one of the simplest examples possible, a single uni-directional route with uniform line speed and a single intermediate station stop.
The maximum use of capacity is generally achieved when all train services are identical. They travel at the same speed, use the same type of rolling stock (so they share the same braking and acceleration characteristics), stop at the same stations for the same amount of time, and take exactly the same route. Figures IV.1 and IV.2 show simple cases where identical services use the route. These are either fast non-stop services (Figure 3) or slower stopping services (Figure 4). In both cases, however, the uniformity of the traffic allows a relatively large number of trains to use the route.
Train Graph with Fast Services
Train Graph with Slow Services
Train Graph with Alternating Fast and Slow Services
Train Graph with Flighted Fast and Slow Services
Figures 5 and 6 then illustrate both fast and slow trains using the route. In Figure 5 these services are alternated, whereas in Figure 6 the services are "flighted" so that several fast trains run in quick succession, followed by several slow trains. The situation shown in Figure 5 often arises in practice, because users of both the fast and slow services generally prefer a service spaced at regular intervals (rather than, for example, three trains in quick succession followed by a long gap with no services). But alternating services carries a penalty in terms of the lower number of trains that can be accommodated.
In practice, the situation facing real-life train schedulers is very much more complex than these simple examples. Among other complications:
- Routes diverge, converge, overlap, and cross each other. On any particular route, services may enter the route at particular junctions and/or leave at others. The timetables on different routes around the network need to be coordinated to allow for such interactions.
- At many locations, trains using junctions need to cut across the path of trains travelling in the opposite direction. Space needs to be left in the timetable to accommodate such conflicting movements.
- There are many different types of trains, not just the fast/slow versions shown above. They may have quite different running speeds, stopping patterns, and acceleration/deceleration capabilities (especially, for example, heavy freight trains).
- As well as the headway necessary to maintain a safe separation between trains, additional allowances must be inserted to allow for "resilience," so that the timetable is not unduly disrupted by individual trains running late and the day-to-day perturbations that are normal on a complex and congested rail network.
- On more complex routes, operators will typically look to maintain connection opportunities for passengers to change trains part-way through a journey.
- Some train operators (especially freight) may require access rights to run particular services but may not be able to specify exactly when the train will run.
Thus train timetabling is a complex task, often undertaken manually by timetabling experts and very often involving only incremental changes from an established timetable that has been tried and tested over a number of years. It relies on specialist knowledge and non-market based processes to decide, for example, which trains should be "first on the graph" on the basis that the services (often long distance or with complex routes) would be difficult to fit into a timetable that was already partially complete.
C. Could Prices or Auctions Be Used to Allocate Rail Network Capacity?
In theory, a variety of market mechanisms can be considered potential candidates for allocating capacity where there is excess demand for train paths. In broad terms, these approaches involve either (1) pricing, that is, some form of pre-determined tariff so that train operators can calculate how much it would cost them to run a particular service or (2) auctions, a process whereby train operators can indicate the value that they attach to particular train paths, thereby revealing the allocation that generates the highest value.
In practice, except in extremely simple cases, it is highly unlikely that either approach could be used to efficiently allocate scarce rail network capacity. Whereas the allocation of scarce airport take-off and landing slots discussed in Section III involves just the node at each end of a journey, the problem of allocating scarce rail network capacity can apply at every location along a route, with the many complexities noted above.
To use a pricing mechanism to allocate scarce capacity between competing users would require a pricing system that could generate a price for any train path that an operator might require. The infrastructure manager would need sufficient information about train operators’ likely demands for these paths at different times of day even to create a market-clearing schedule of prices for a "standard" train path. This in itself would be extremely difficult. But also, there is no such thing as a standard train path. In addition to differences in speed, stopping patterns, etc., even minor retimings of identical train paths could either create or remove conflicts with other services. We see only one exception to the unsuitability of using pricing to allocate capacity: Levying rail infrastructure charges on the basis of planned services (for which the infrastructure manager has allocated capacity) rather than on the services that actually run, would discourage train operators from reserving scarce capacity that they may in fact not use.
The use of auction mechanisms also raises major potential problems, but in theory at least it could provide a mechanism through which train operators could express their requirements and their willingness to pay for these, and the auctioneer could compute the set of feasible allocations that generates the maximum value. Borndörfer et al. show the hypothetical application of a simultaneous ascending combinatorial auction to a small part of the German rail network.15 Bidders can specify flexible time requirements (for example, frequencies rather than specific departure times), require connections between services, and specify rolling stock diagrams. At each round of the auction, the highest value feasible allocation is calculated.
However, we believe that applying this approach to real-world situations would be extremely difficult and probably impossible. Train operators would be concerned with many more variables than those considered by bidders in Borndörfer’s paper. The process of drawing up bids, addressing interactions between different services and deciding how much to bid for each would be complex and possibly intractable. Moreover, this would need to be repeated at each round of the auction. It is also likely to be computationally infeasible. We note, for example, that the authors carried out quite simple tests involving bids for 946 pre-defined candidate train paths but with varying degrees of freedom around the precise departure time. When the time window for each slot was increased to five minutes, it took three days to compute the results of a single round. Though interesting from a theoretical point of view, and despite recent advances in auction design, we believe the prospects for using auctions to allocate scarce rail network capacity are very remote.
Instead, therefore, where capacity allocation problems have arisen in practice, they have sometimes been addressed through analysis focused on the particular location and the specific services seeking access, rather than through more generic approaches that could be applied to all conflicts. This has included
- Government decisions, backed up by varying degrees of economic analysis, about the services that should be allowed to run (and sometimes funded by the government);
- Specific regulatory decisions that examine the merits of the different services that operators wish to run but cannot all be accommodated; and
- Detailed studies in the UK resulting in "Route Utilisation Strategies" that seek to establish the best use of the available infrastructure along particular routes. These studies are unusual in that they can recommend changes (or even removal) of existing services and assess the relative merits of possible new services.
In other cases, however, potential capacity allocation problems have been resolved through the application of administrative rules rather than analysis. These include operational decisions about which trains should be "first on the graph" and political decisions about particular types of services that should have priority over others. However, an implicit system of grandfather rights is also common, such that potential new entrants will simply be told that there is no room on the network for their service rather than having any opportunity to argue the case that their proposed service should displace others that are currently using the network (for example, because it would generate higher net benefits).
D. Pricing of Rail Infrastructure
Even if the pricing mechanism cannot be used to allocate scarce capacity on the rail network, rail infrastructure charges can still play a valuable role in encouraging the efficient use of less congested parts of the network. But the prices need to be set correctly. If infrastructure charges are too high, potentially viable traffic (i.e., traffic that could afford to pay the additional costs that it creates) could be priced off the network. Equally, infrastructure charges that are too low could result in inefficient use of the network – either by services that generate no additional value, or because there are too many services in total and therefore the level of congestion is inefficiently high.
The efficient infrastructure charge for a train operator seeking to run an additional service is the short-run marginal cost (SRMC) of that service.16 SRMC covers costs such as:
- the marginal wear and tear (hence increased maintenance costs or advancement of renewals) imposed on the track and other assets by the additional train;
- for an electric train, the traction current consumed;
- any additional operations costs, reflecting for example increased work for signallers and train planning staff; and
- any costs imposed on other users of the network.
The latter can be important in cases (which are common) where capacity is theoretically available but the network is actually already congested. The costs include both the costs that the "new" service might impose on existing services if it runs late and therefore disrupts them, and also the increase in expected delays that will result simply because the network is even more congested and therefore has less resilience to recover from any disruption that does occur (even if it is not caused by the new operator). The UK is unusual (though not unique) in that it has introduced a component of its track access charge that attempts to capture such costs, but in practice it has proved difficult to estimate these costs at the level of detail originally envisaged.
A more general difficulty with SRMC pricing is that it generally leads to a very low level of cost recovery. In Europe, SRMC is typically estimated at around 10 to 20 percent of average cost. The conflict between the objectives of efficient pricing and of cost recovery is one of the main challenges that infrastructure managers and governments have had to address when introducing track access charges. A few governments, mostly in Scandinavia, have been willing to fund the difference between marginal and average costs, though even in some of these cases this is being reviewed. Elsewhere, even if some government subsidies are available, the infrastructure manager needs to set access charges higher than marginal cost for at least some train operators in order to meet its cost recovery target.
Where prices (in any industry) have to be raised above marginal cost, the two standard approaches used to minimise the adverse impact on economic efficiency are two-part tariffs and variable mark-ups over marginal cost. In a two-part tariff, each user pays a fixed charge (usually the same for each user or class of user) plus a variable charge based on their consumption. The variable charge remains at marginal cost and therefore promotes efficient consumption decisions. But this approach can only be applied where the fixed charge will not lead to a significant number of potential consumers reducing their demand to zero. When applied to intermediate goods such as rail infrastructure, it can also distort competition between firms in downstream markets (e.g., large and small train operators). Where it is possible to use two-part tariffs without pricing some train operators off the network or distorting competition, this approach may avoid the efficiency loss from pricing above marginal cost. But such circumstances are rare – the main examples are France, where SNCF faces little or no meaningful competition, and in the UK, where the fixed charge is paid by franchised passenger operators who know what the fixed charge will be in advance and therefore simply factor it into their franchise bids.
Variable mark-ups over marginal cost are based on Ramsey Pricing principles, which state that the mark-up for each consumer should be inversely proportional to that consumer’s elasticity of demand. The main way the principles have been implemented in practice is through mark-ups that vary among different market segments or broad types of customers. In some countries, access charges do vary by factors that may be related to the underlying demand elasticities – these include the type of service (e.g., freight, local passenger, long-distance passenger, high-speed passenger) and route. But it is not clear that these charges are based on differential mark-ups over marginal cost rather than simply an adjustment to the average charge per train required to meet the cost recovery target, or that the mark-ups fully reflect the price elasticities of different types of traffic.
V. PRICING AND INVESTMENT
Our theme in this chapter has been the importance of considering the contribution of pricing and other techniques of economic allocation to improving efficiency within the transport sector by improving use of existing capacity. Nevertheless, decision makers cannot therefore neglect investment decisions. While different allocation mechanisms will alter investment priorities over the coming years, they will not avoid the underlying need to consider how far capacity should be altered to cope with changing transport demands.
11 While the payment would be made to the coordinator, the coordinator would probably not retain the revenue, but use it, for example, to fund improvements to airport infrastructure.
12 Dusseldorf airport has explored the concept of levying a slot reservation fee.
13 Federal Aviation Administration (2001), Notice of Alternative Policy Options for Managing Capacity at LaGuardia Airport and Proposed Extension of the Lottery Allocation. Published in the Federal Register of 12 June 2001.
14 In the European Union, EC Directive 2001/14 requires that, where the infrastructure manager is not independent of all train operators, responsibility for ensuring that infrastructure capacity is allocated on a fair and nondiscriminatory basis shall be assigned to an independent body.
15 R. Borndörfer, M. Grötschel, S. Lukac, K. Mitusch, T. Schlechte, S. Schultz, and A. Tanner , "An Auctioning Approach to Railway Slot Allocation," Competition and Regulation in Network Industries, Special Issue on Recent Trends in the Privatization and the Regulation of Network Industries, I:2 (2006),163-196.
16 For a discussion of how SRMC might be applied in practice, see NERA (1998), An Examination of Rail Infrastructure Charges. This is a report for the European Commission that provided the basis for many of the charging principles included in EC Directive 2001/14.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.