Allocating Cost Between Customer Classes

Setting the revenue requirement answers how much the utility may collect. It does not answer from whom. That second question is worked out through a cost of service study, which divides the total among customer classes according to the costs each class causes the system to incur. The study produces a number for each class, and commissions then decline to follow that number all the way.

Functionalization, classification, allocation

The study proceeds in three steps. Functionalization assigns every cost in the revenue requirement to a function: production, transmission, distribution, or customer service. Most costs functionalize obviously from the accounts, since the prescribed accounting system already separates generating plant from distribution lines. General plant and administrative expense are functionalized by spreading them in proportion to the directly assigned costs.

Classification sorts the functionalized costs by what drives them. Demand-related costs vary with the maximum load the system must be capable of serving, and cover the capacity of generating units, transformers and lines. Energy-related costs vary with the quantity consumed, principally fuel and variable operating costs. Customer-related costs vary with the number of customers served regardless of consumption, covering meters, service drops, billing and the customer-facing portion of distribution.

Allocation distributes each classified pool to the classes using a factor that measures each class's contribution to the driver. Energy costs are allocated on kilowatt-hours, adjusted for losses at each voltage level. Customer costs are allocated on customer counts, sometimes weighted for the differing cost of serving a residential meter and a large industrial one. Demand costs are allocated on a measure of peak load, and that is where the disagreements are.

Some costs bypass the sequence entirely. Where a facility exists to serve one customer, its cost is directly assigned to that customer rather than allocated, and a large industrial line extension is the standard example. Direct assignment is the most accurate treatment available and the least often applicable, since most of a utility's plant is shared.

Why the demand allocator is contested

Generating and transmission capacity exists to meet peak load, but there is no single correct way to measure a class's responsibility for that peak. Several methods are in use. A coincident peak method allocates on each class's load at the hour the system peaks, which favors classes whose usage is flat. A non-coincident peak method uses each class's own maximum, which favors classes that peak with the system. Average and excess methods blend a class's average demand with its excess over that average.

Methods that recognize several peak hours, or that weight hours by the probability of capacity being needed, sit between the extremes. Each produces a defensible allocation and each moves substantial dollars. A shift between two accepted methods can change a residential class's indicated responsibility by several percentage points of the total revenue requirement, which is why the choice draws the most testimony in the allocation phase.

Losses complicate every allocator. Energy delivered at low voltage requires more generation than energy delivered at high voltage, because some is lost in transformation and transmission. Studies therefore gross up each class's measured consumption by a loss factor appropriate to its delivery voltage, and a change in those factors moves cost between classes without any change in the underlying method.

What the completed study shows

The output is a table of class revenue responsibility: the cost the study assigns to each class, the revenue that class currently produces, and the resulting rate of return by class. Where a class returns less than the system average, the study indicates it is being subsidized by classes returning more. Where it returns more, the reverse.

Those indications are the starting point for a discussion rather than an instruction. Every figure in the table depends on the classification and allocation choices made upstream, and reasonable studies of the same utility can indicate different directions for the same class. Commissions treat a study as evidence of the general direction and magnitude, and they discount fine distinctions that a change in one allocator would erase.

Class definitions themselves are occasionally reopened. A class whose members have diverged in usage patterns may be split, and classes that have converged may be merged. Because a class boundary determines which customers share an allocation, redefinition can accomplish more than any change in allocator, and proposals to create a new class are examined for whether the distinguishing characteristic is a genuine difference in cost causation.

Gradualism and the movement actually ordered

Even where a study clearly indicates that one class is paying below cost, commissions rarely move that class to indicated cost in a single case. The principle applied is gradualism, sometimes called rate shock mitigation: the increase to any class is capped at some multiple of the system average increase, and the remaining movement is deferred to later cases. A class indicated to need a large increase might receive one and a half times the system average, with the rest carried forward.

The justification is partly practical and partly about the limits of the study. Customers make long-lived decisions on the basis of prevailing rates, and an abrupt change imposes costs the cost study does not measure. And because the indicated figure rests on contestable allocation choices, moving all the way to it would treat a model output as a fact. Gradualism hedges against the possibility that the model is wrong.

The result is that class revenue responsibility converges on indicated cost slowly, across several proceedings, and never entirely. What follows the allocation decision is the conversion of each class's revenue target into a tariff, described in designing the rate once the revenue is set. The total being divided is assembled as described in the revenue requirement and how it is built, the proceeding in which the division is litigated is described in the rate case and who takes part in it, and costs recovered outside that proceeding follow the route in riders and trackers outside a rate case.