topleftround (1K) toprightround (1K)
earth01 (3K)
photo may07 (10K)
Arvid Kruze, B Eng, MBA
International Power Sector Consultant

Financial and Economic Analyses
Electricity Tariffs
Team Leader / Project Manager

Montreal, Quebec, Canada
+1 (514) 484-8050

Marginal Cost Pricing Studies


The concept of marginal cost pricing for electricity goes back about 40 years. From the 1970s and into the 1980s, marginal cost pricing of electricity was a relatively hot topic, with most people working in the electricity supply industry acknowledging its merits. Electricity prices based on marginal cost, specifically long-run marginal cost (LRMC), looked at the future costs of supplying the next kWh or kW, while the traditional cost of service studies that had been used by regulators and electric utilities for decades as a guide for tariff setting were quashed as backward-looking.

Well, looking to the future is really the way to live life (as opposed to living in the past) and, besides, who wants to be regarded as backward-looking? So, trying to base electricity tariffs on LRMC became a favourite topic among regulators. Even the international aid agencies jumped on board.

The subject of marginal cost pricing seemed to become less important during the 1990s and faded somewhat as new issues such as price regulation, unbundling and deregulation in general came to the forefront. At that time, some people in the industry actually figured that any kind of electricity tariff study based on cost of service, whether these were traditional embedded costs or LRMC, was now irrelevant. Of course, this is absurd.

Today, electricity pricing based on marginal cost or LRMC is still arises occasionally as a topic, judging from the number of recent references one can obtain through an internet search. I was actually involved in a tariff study as recently as 2006 in which the calculation of LRMC was included in the study terms of reference.

So, how to consider LRMC or marginal costs in electricity tariff setting? This is a good question, as electricity pricing has never been that straightforward, even when there was no such thing as marginal cost to consider. I have often heard people refer to electricity tariff setting as an art-form, as many alternatives are available to set a tariff when one considers blocked energy charges, blocked demand charges, customer charges and time-of-use differentiation of these charges, and then, different combinations thereof. However, I believe calling tariff setting an art-form is going a bit too far. There are many alternatives available in terms of charging, but their boundaries are generally quite well-defined.

In any case, there is a certain degree of nebulousness considering the different charging methodologies available combined with various ways that costs for tariff setting can be allocated for this purpose. This area can be murky enough to a novice without considering marginal costs. Introducing the concept of marginal cost into the mix has made it even more so, and quite needlessly. This is because: a) there is no single clear generally-accepted methodology for measuring/ defining LRMC and b) once marginal costs have somehow been estimated, they must be adjusted to be consistent with the financial realities faced by electric utilities, which are, after all, commercially oriented entities (even if some happen to be heavily subsidised government-owned money-losers). Given all this, it is no surprise that the whole concept of marginal cost, and, specifically, LRMC, has been referred to as a Theoretical Construct with little practical applicability.

OK, personally, I would not go that far, as I do believe that reflecting marginal costs in electricity tariffs, in a practical way, is not only possible, but desirable. I think marginal costing applied to electricity pricing has earned a bad reputation to some extent only because economist-theoreticians have spent lots and lots of time constructing not-so-useful theoretical frameworks and models that, indeed, have very little practical applicability when tried in a real-world electric utility setting.

Given the above long-winded intro, following is my preferred methodology for applying marginal costing concepts to electricity pricing.


The three basic steps in setting tariffs and the possible role of marginal cost

I like to think to think of tariff setting as comprising three basic steps, which require quite independent exercises to be undertaken:

Step 1. Determine the overall average selling price per kWh required by the electric utility to cover its costs, which is determined by building up a revenue requirement so that an adequate level of sound financial health is achieved. How to define this adequate level of sound financial health? Well, that is really quite another topic, debated over and over and over again, in public hearings, by regulatory authorities, utilities, consumers, consultants and other stakeholders in the power industry.

Anyway, the above step is all about meeting financial obligations without earning what might be considered excess revenue. I have never seen or heard about marginal costing being employed to establish a total revenue requirement anywhere, so just forget about using marginal costs here. The straightforward calculation of unit marginal costs, expressed on a per kWh, per kW and on a per customer basis will yield a revenue total that is different from the financial requirement

Step 2. Once the revenue requirement is set, determine the average selling price and total revenue required for each customer category. This should reflect cost differences in providing electric service to each category. Essentially, if the revenue requirement is thought of as a pie, this exercise divides the pie amongst the various customer categories.

This step has traditionally been addressed through embedded cost of service studies, where the pie is allocated to each customer category through a relatively complex, but well thought-out process. I describe this process under another Topic.

Marginal costing proponents largely oppose this exercise, instead advocating the calculation of full LRMC, on a per kWh, per kW and per customer basis, which inevitably results in a revenue total that is different from the revenue requirement. To deal with this discrepancy, the resulting marginal cost based revenue levels by customer category are then adjusted, usually on an equiproportional basis so that they yield the total financial based revenue requirement. Effectively, this results in a slightly different allocation of the revenue pie to the various customer categories. The argument for calculating/ adjusting LRMC in this manner is that at least the revenues to be collected from each customer category are determined in a more economically optimum fashion and LRMC based revenues are at least somehow reflected by maintaining the same structure.

I guess this sounds logical to some extent, but I do not buy it. Having been through this exercise a few times, it sort of feels analogous to fitting a square peg into a round hole.

With a traditional embedded cost of service study, at least one is starting with a real, financially based revenue amount that can be directly allocated to customer categories based on their actual usage of system resources. OK, this usage is based on current consumption patterns - not the future - but really, how much will these change in the future to the extent that it will make a big difference? Well, not very much.

Step 3. Set the tariff for each category in terms of energy charges, demand charges and customer charges so that the revenue requirement for each category is realised. This revenue requirement can be determined through either the ill-defined LRMC process discussed in Step 2 or, preferably, through a regular embedded cost of service study.

Once the revenue level by customer category is set, this step is where I believe the concept of marginal cost is best applied. One merely sets the energy rates at what I call the short-run marginal cost (SRMC), with the balance of the revenue requirement for the category collected from the fixed customer charge (small customers) or the demand charge (large customers), as the case may be. For reasons that totally escape me, most LRMC methodologies define the SRMC as the LRMC, so one may end up with somewhat similar tariffs under my preferred approach as with some LRMC methodologies.

What about marginal cost based demand charges? Well, we have already set the energy charges at marginal cost. That means the LRMC of demand must bear the adjustment to get the revenues in line with the financially based revenue requirement. Actually, calling this an adjustment to LRMC is much too ingenuous to describe the setting of the demand charge, as LRMC has nothing to do with it. The demand charge is simply set so that the financially based revenue target is reached. Thus, if one has gone through the exercise of calculating a marginal cost of demand, it has been a totally useless one.

Here, an LRMC proponent may argue that marginal costing will allow one to time-differentiate demand charges (based on relative loss-of-load probabilities, for example), but that is really an independent exercise whose results can be equally applied to a financially based demand charge.


LRMC versus SRMC

It is appropriate at this point to define SRMC, briefly describe how this differs from LRMC and, provide a method for calculating it. There is really no sense in describing any LRMC methodology in detail, since, as seen in the previous section, the concept of LRMC does not seem very relevant for tariff setting purposes. However, in deference to international lending agencies such as the World Bank and the ADB, I should mention that their economists have, in the past, required LRMC calculations to be undertaken as part of technical assistance projects in the power sector, but I have never seen these calculations carried out for tariff setting purposes (except perhaps by their consultants). I can see the merit in judging the adequacy of current tariffs in terms of the LRMC of future supply (at least partly), but this LRMC is more appropriately derived as a number representing the present worth of future costs to meet incremental load growth, as contained in a system expansion program. These are also known as levelized future costs for meeting incremental demand, which I regard as the best indicators of LRMC and these are quite different from the so-called LRMCs arising from the rather convoluted calculations of other methodologies. In any case, for the setting of actual tariffs for individual customer categories, any LRMC calculation is quite secondary. Tariff levels are best determined through a financial forecast of revenue requirements. Their adequacy may be partly judged through a comparison with a levelized LRMC, but in the end, it is really the financial requirement that matters, a financial requirement that is determined through a financial forecast.

Anyway, moving on, the basic difference between LRMC and SRMC is that the LRMC assumes all cost inputs are variable, as fixed plant can be modified over the long-term in order to provide outputs more efficiently. In the short-term, certain inputs, such as capital equipment, cannot vary. Thus, the only variable in the practical application of SRMC to electricity tariffs is the cost of fuel, variable O&M and associated power system losses (or, in the case of purchased power, the price paid).

Certain LRMC methodologies use the argument that the SRMC is equal to the LRMC...

In any case, the SRMC of energy is the incremental cost of the most expensive unit currently in use. I have seen this sometimes referred to as the system lambda. The calculation of the SRMC requires an analysis of the system lambda on a time-of-use basis. The method of calculation entails an assessment of typical daily load curves in present and future years, and seasonal and monthly curves (if required), in order to estimate which resources will be meeting the load at different times and, in particular, which resources and their associated incremental costs are meeting the load at the margin.

This exercise may be carried out in a simple manner, through a cursory examination of existing load profiles and the generating units (or purchases) likely to be operating at the margin over time. Alternatively, it may entail the computer simulation of how daily load will be met by generating unit, day by day, season by season and year by year. This latter exercise can be extensive and may be carried out with considerable effort. But really, for tariff-setting purposes, it does not have to be so complicated. One only needs to have a basic understanding of how the current power system operates, associated incremental costs and, how these will basically change in the future as the system grows.

And, as seen in the last section, SRMC is useful only in the final step of tariff setting, i.e., tariff design after the revenue requirement by customer category has been defined. Quite simply, this entails setting the energy charge(s) at the SRMC. As SRMC varies by time of day, it can be a strong basis for designing a time-of-use tariff. Or, depending on the SRMC cost structure, it can be used in the design of a blocked energy tariff, with the last block set at some level of SRMC. With the energy charge(s) set at SRMC, demand and/ or customer charges may then be set accordingly so that a given revenue target (based on a cost of service study) is met.

The SRMC is a very effective cost signal to send consumers, as they can then decide whether or not to consume the extra kWh based more or less on the current incremental cost of generation. Also, customers generally tend to be more responsive to energy charges than other charges that are fixed.


Summary on using marginal costs for setting

From all of the forgoing, it can be said that LRMC can be used for tariff setting purposes only when the particular LRMC methodology happens to define the LRMC of energy as the SRMC (although I have conceptual problems with this idea). Of course, an economic theorist might also like to have an LRMC calculation on hand, however calculated, to judge the adequacy of a given tariff schedule to conform to long-run economic costs, but really, future revenue requirements are better derived through financial forecasts.

So, this is how an electricity tariff study based on marginal costs should be conducted:

1.  Set the total revenue pie through a financial forecast.

2.  Undertake a traditional embedded cost of service study to divide this pie among the customer categories.

3.  Determine the SRMC of energy supply at different voltage levels and different times of day, times-of-year, etc. (i.e., whatever differentiation is appropriate, based on cost differences).

4.  In designing tariffs for each customer category, set the energy rates at the SRMC, with the balance of the revenue requirement for the category collected from the fixed customer charge (small customers) or the demand charge (large customers), as the case may be.

5.  Depending on the extent that the power system is capacity constrained (as most power systems are), demand charges may be time differentiated based, for example, on relative loss-of-load probabilities.

While the above framework will apply to most power systems, I have come across a few situations where a little tweaking of the approach may be required; specifically, all (or mainly) hydropower based power systems, as well as situations where the SRMC is greater than the revenue requirement. However, such situations tend to be an exception. In any case, an expert should probably be hired before undertaking any such exercise from scratch.