Network pricing - Part 2

Part 1 explained that network prices should mimic, as far as possible, competitive market outcomes. It also explained why, in competitive markets, prices reflect both short and long run marginal cost - at least on average, over time. Putting these things together, it follows that the price a customer should be charged for consuming an ‘additional unit’ of a regulated network service should reflect the marginal cost of supplying it.

Although this principle is relatively straightforward, it can be challenging to put into practice. Careful consideration needs to be given to the principal drivers of network costs and the type of marginal cost pricing to employ. There's also the question of how to go about recovering costs that aren't marginal (e.g., the sunk costs of existing assets) – an issue that doesn't arise to the same degree in competitive markets.   

What's driving costs?

To develop marginal cost-based usage prices that promote efficient use of and investment in network infrastructure, it's necessary to understand the principal drivers of network costs. Only then can tariffs be developed that signal to customers when they're engaging in behaviour that's contributing the most to future costs. Those signals can then give customers an opportunity to change their conduct in efficient ways. For most network businesses, future costs are typically driven by factors like:

• the number of customers that are using the network;
• the maximum demand on the network;
• when customers are using the network; and
• the locations within the network that services are sought. 

Understanding the relationship between each of these characteristics of the business (and any others that are relevant) and its costs is consequently essential to efficient pricing. In principle, there will be a unique marginal cost associated with changes to each of these different cost drivers, i.e., the marginal cost of a new customer using the network, the marginal cost of changes in system demand, the marginal cost based on the time of infrastructure use and the marginal cost based on the location of that usage.

A network tariff structure could therefore contain a series of quite ‘granular’ usage charges based on these different marginal costs, which could be levied on customers depending upon which categories applied, e.g., peak vs. off-peak charges; location-specific elements, etc. But of course, more complex pricing structures also give rise to other costs. They're typically more expensive to design and administer and, perhaps even more importantly, they can be harder for customers to comprehend and respond to – especially if the different components fluctuate over time.

For that reason, regulated businesses don't usually try and isolate and price every single significant price driver. Instead, they're more likely to focus on the primary cost drivers and maintain a simpler tariff structure that's easier to understand and cheaper to administer. Indeed, because the regulated business’ customers will often have to make their own long-term investment decisions, simpler, more certain price structures often have broad appeal. Ultimately, there's no ‘right or wrong’ way to go about striking this balance between specificity and simplicity. It all depends on the circumstances.  

SRMC or LRMC?

In competitive markets, prices reflect both expected SRMC and LRMC (see Part 1 if you want to know why). However, it's sometimes more practical, in a regulatory setting, for prices to entail elements of both. Consider a simple example. Imagine that, several times a day, a train travels along a segment of railway line. Most of the time, it's the only train that wants to use that path, i.e., it's not preventing any other operator from using it. Let's suppose also that's unlikely to change in the longer-term. But, at other times of day, there's several other operators wanting to run trains along that path as well.

In this simple scenario, at the times of day the train is using the line when there's no other demand (i.e., when nobody else wants to run trains), the marginal incremental infrastructure costs of that usage are likely to be trivial. A SRMC-based tariff for use of the rail link during those times might therefore be a very small sum, e.g., a usage price equal to the additional impact upon operating and maintenance costs (probably a modest total). Conversely, at times when the line is in high demand, and that operator’s usage is depriving others from using the link, the SRMC would include those costs imposed on others, i.e., it would include short-term constraint costs.

A SRMC-based usage price that included the 'externality effects' of those additional costs would therefore provide signals that might alleviate the congestion. Specifically, if the operator wasn't prepared to pay those higher SRMC-based prices (inclusive of congestion costs), this would reveal that it derived fewer benefits from using the line at that time than other prospective users. If it then decided to relinquish the path to another user, that would constitute more efficient use of the network, i.e., the line would be being used at that time by a customer who valued it more.

In a similar vein, a LRMC-based price levied on consumption during the periods in which there was excess demand would reflect the incremental cost of alleviating any such constraints by expanding the capacity of the line, e.g., through double-tracking. Once again, if the operator wasn't willing to pay that LRMC-based price, it would reveal the operator derived fewer benefits from using the line at that time than other prospective users. Here again, it would presumably surrender the path and allow someone else to use it instead, promoting more efficient usage.

It follows that, in this simple example, given that the cost of expansion is only caused by users during certain periods of ‘peak demand’, efficient utilisation and investment in this fictional rail network could, theoretically, be achieved through a tariff structure that:  

1. Set usage charges during ‘off-peak’ periods equal to the SRMC of providing the service during those periods, i.e., equal to operating and maintenance costs; and
2. Set usage charges during ‘peak’ periods equal to either:

• the SRMC of providing the service at those times, i.e., equal to operating and maintenance costs plus constraint costs; or
• the LRMC of providing the service during those periods, i.e., equal to the future costs of capacity expansions.

It'll often be easier for firms in these circumstances to use LRMC pricing rather than SRMC-based congestion pricing (for the ‘peak’ usage component). Because of the fluctuating nature of SRMC, it doesn't provide enduring investment signals and could result in volatile prices. Moreover, as a practical matter, network businesses probably aren't going to be able to accurately measure congestion costs and feed them into prices. That would require them to know how much - in quantitative terms - congestion and delay was affecting other customers, which would be very challenging.

It's usually much simpler for a regulated firm to ascertain the impact of congestion on its own future investment costs (i.e., to estimate LRMC) than to measure the externality costs of that excess demand, i.e., the costs imposed on others. These practical limitations mean that network businesses usually use LRMC as the basis for signalling future network costs, and limit SRMC-based pricing to situations where capacity is ‘unconstrained’, i.e., the ‘off-peak’ periods in the example above.

If regulated usage prices depart significantly from SRMC or LRMC (as the case may be), this can compromise both 'static' and 'dynamic efficiency'. The efficiency of the usage of the existing network – i.e., static efficiency – can be harmed, because:

• when usage prices are above SMRC/LRMC, customers will be unduly discouraged from using the network, even though they'd obtain more value from consuming the service than the incremental costs imposed on the operator; and
• when usage prices are below SRMC/LRMC, customers will use the network even though the benefits they derive from doing so are less than the additional (marginal) costs that their consumption imposes on the operator. 

Inefficient network usage (i.e., static inefficiency) can then result in inefficient investment in the longer-term (i.e., dynamic inefficiency). If usage prices are above SRMC/LRMC, this may cause customers to inefficiently curtail their demand (i.e., if those prices exceed the benefits customers derive), inappropriately pushing-back future capacity expansions. Conversely, if prices are constantly below SRMC/LRMC (e.g., during periods of peak demand), then this may inefficiently bring forward those expansions.

What about sunk costs?

So far, I've focused on the importance of setting the price for an additional ‘unit’ of a regulated service equal to its forward-looking marginal cost. But what about ‘non-marginal’ costs, e.g., sunk network costs? How should those be recovered? In competitive markets, these are recouped during periods of short-term excess demand, e.g., during New Year’s Eve, Harbour-front hotel prices skyrocket as demand greatly outstrips supply. There's no limit to how high prices can rise at these times in such markets.

In contrast, in regulated industries, there tends to be restrictions on how far prices can increase when constraints emerge that limit the extent to which firms can make contributions to their fixed costs. This is because most regulatory regimes don't allow businesses to recover more than the ‘total cost’ of delivering the service during the relevant period. For example, regulated businesses generally aren't permitted to ‘over-recover’ their total costs in one period via SRMC and/or LRMC charges, with the expectation that they'll ‘under-recover’ by an equivalent amount in some later period (e.g., after an investment has been made and congestion has disappeared). Rather, the 'up-side' is capped. 

This means that marginal cost-based usage charges will almost never yield sufficient revenue to cover a regulated firm’s total (i.e., marginal and non-marginal) costs, i.e., 100% of its revenue requirement. There's therefore invariably a need to recover the remainder of the business' revenue requirement through additional charges. Recall that the key objective of the marginal cost-based usage charge is to signal to customers future costs to encourage efficient changes in behaviour. In stark contrast, the primary objective of any additional residual charges is to discourage inefficient changes in customer conduct.

Specifically, the idea is to recoup the residual revenue in a way that minimises changes in the use of the existing network, relative to what would have happened if consumers had paid only the marginal cost of supply. In other words, the marginal usage charge incentivises efficient conduct and the main job of any additional charges is to recover the residual costs without undermining those beneficial signals. Broadly speaking, there's two options that can satisfy this 'non-distortionary' principle:

• levying fixed charges on customers that don't vary with their respective levels of usage; and/or
• levying 'mark-ups' on customers in ways that reflect their willingness to pay, i.e., to engage in so-called 'Ramsey pricing'.

Of the two approaches, the former tends to be employed more often due to the additional informational challenges associated with Ramsey pricing, e.g., the need to have data on customers’ elasticities of demand. But, once again, there's no 'right or wrong' answer to the question of how best to set residual charges. There's inevitable trade-offs to be made between complexity and simplicity and it all depends on the circumstances.

What about customers’ expectations?

Regulated businesses must frequently make lumpy, irreversible investments that give rise to considerable 'sunk costs'. As such, they're constantly relying on regulators to resist the urge to engage in any opportunistic conduct that compromises their ability to recoup those costs once they've been incurred, since there is no realistic option of reselling the assets in second-hand markets. For this reason, changes to regulatory frameworks that give rise to such 'asset stranding' risks must be extremely well-justified before they're countenanced.

But regulated businesses aren't the only parties that face these 'hold-up' risks. The customers of regulated businesses will often also need to make critical investment decisions that may be influenced to a significant extent by the existing regulatory frameworks. For example, a rail network customer might decide to invest in a certain type of train (e.g., electric rather than diesel) to make the most of the existing regulatory pricing methodology, i.e., to minimise its exposure to such charges, based on the current methodology.

Regulated businesses – and, in turn, regulators – should consequently be mindful of those past investment decisions when contemplating making changes to regulatory pricing structures (e.g., suddenly changing the prices applied to, say, diesel vs. electric trains). Of course, that's not to say that such structures can never change. Rather, it's simply to say that such modifications should only be made when there's a clear and compelling case for doing so, i.e., when an alternative is available that would be materially better at promoting efficient outcomes than the status quo.

Furthermore, even when a superior pricing methodology exists, it will often be worthwhile seeking to manage the financial ramifications of the reform by transitioning to the new prices over time, e.g., phasing them in over several years to allow firms time to restructure their businesses. It may even be appropriate, in some circumstances, to put in place 'grandfathering' arrangements. These mechanisms usually mean that the new pricing arrangements apply only to new entrants and the existing frameworks continue to apply to those existing customers that have invested under them.  

Summary and implications

Broadly speaking, an efficient network pricing structure can be expected to exhibit the following basic characteristics:

• a price for 'network usage' that reflects that marginal cost of that consumption, to provide efficient signals to customers to change their behaviour in efficient ways when this would save costs; and
• a residual charge (or charges) to recover the remainder of the firm's total costs whilst giving rise to the fewest possible changes to consumption, i.e., discouraging inefficient changes in conduct.  

Changes to that basic structure should occur only when there is clear evidence that an alternative is available that would be materially better at promoting efficient outcomes, and when transition mechanisms have been considered.