Begin with the most ordinary event in compensation.
An employee earns one hundred thousand dollars. The following year, the employer awards a four percent increase. The system records it with the precision these systems are built for. Previous salary, one hundred thousand. New salary, one hundred and four thousand. Increase, four percent. A letter goes out, payroll updates, the total rewards statement regenerates, and everyone involved understands what happened.
The employee is better paid.
Is the employee better paid?
The nominal answer is obvious, and it is the one every system in this profession is built to produce. The economic answer requires a denominator. If the life that salary purchases, the rent or the mortgage, the groceries, the childcare, the commute, the insurance, the eventual retirement, became more expensive over the same year by more than four percent, the employee now receives more currency units and commands less.
Compensation professionals know this. Real wages are a standard concept. Cost-of-living adjustments exist. Inflation assumptions sit inside every actuarial valuation, and no serious practitioner confuses a nominal raise with a real one when the distinction is put to them directly.
Nothing in that observation is new. The structure underneath it is what I want to look at.
Two Jobs, One Unit
Compensation uses money in two roles at once.
Money is the medium in which compensation is delivered. Salary, bonus, employer contribution, pension promise, all in dollars.
And money is the unit in which compensation is measured. Salaries compared, raises computed, retirement targets set, replacement ratios expressed as dollars over dollars.
So the profession pays in a unit and measures in the same unit. It measures the payment with the payment.
That would be unproblematic if the unit held still. It does not. The purchasing capacity of a dollar changes through time, which means compensation is measured with a ruler whose length changes between the moment the measurement is taken and the moment it matters.
The metaphor is worth spending a moment on, because it does work for the rest of this series. Imagine measuring a room with a tape whose markings drift, so that a wall recorded as ten meters in one year reads as eleven a decade later, though the wall has not moved. Each reading is accurate on the day it is taken. A builder comparing readings across years, not knowing the tape had changed, would draw wrong conclusions about the building.
Compensation practice compares readings across years constantly. Salary histories. Pension accruals. Retirement projections. The forty-year arc of a career. Some of those comparisons adjust for the ruler and many do not, and even the adjusted ones use a general index that may or may not describe the wall in question.
The point is not that the ruler is a fraud. It is that the ruler is a variable, and a profession that measures everything with it has, with few exceptions, treated it as a constant of the problem rather than as one of the things being designed.
Stating the Claim Carefully
I want to be exact about what is new here, because the careless version of this argument would insult a profession that has thought about inflation for a century.
Compensation practice has instruments for the ruler's movement. Cost-of-living adjustments. Salary structure movement. Indexed pensions. Inflation-linked benefits. Real-return assumptions in plan valuations. All of that exists and all of it works.
What compensation theory has not generally done is treat the monetary unit as a first-order variable of compensation design itself. It treats the unit as an exogenous condition inside which compensation is designed, and then the design proceeds inside it.
This paper asks what changes when the unit comes inside the design problem. And it arrives at that question not from monetary economics but from the trajectory of the two papers before it.
The first paper argued that compensation is the transfer of purchasing power through time, and that the profession has measured that transfer only at its starting point. The second asked what carries the deferred portion across the decades between earning and need, and answered that a preservation architecture does: a stack of instrument, wrapper, ownership, custody, contribution, default, pooling, tax, portability and conversion, judged against the worker's own future liability by six tests.
Both papers deliberately held the money constant. The first defined compensation in terms of purchasing power without asking what it was denominated in. The second examined the monetary environment as terrain the architecture has to cross, and treated the money itself as a given.
If compensation is purchasing power transferred through time, and the transfer is denominated in a unit whose purchasing capacity changes, then the unit is not the background of the problem. It is one of the variables the compensation has to travel through.
Three Denominators, Not Two
The second paper introduced a layer that this one builds on.
A general price index deflates compensation by a representative basket, and no representative basket is the basket of a particular worker. Households at different incomes, regions, life stages, housing tenures and health exposures experience measurably different inflation rates, and the differences are systematic, persistent, and sometimes large enough to change the sign of a real-wage comparison.
A worker renting in a fast-appreciating city, with children approaching university and parents approaching care, faces a price trajectory no national index describes.
So there are three concepts, not two. Nominal compensation: the count of currency units delivered. Real compensation: how much of a representative basket those units command. And what I will call liability-relative compensation: how much of the worker's own future consumption, in the life the worker will actually live, the compensation will eventually purchase.
One actuarial precision belongs here, because it keeps the argument from overreaching. A worker's future obligations are not all of one kind. Some are effectively nominal, and a fixed-rate mortgage becomes easier to service as the unit decays, which makes a nominal claim its natural match. Some are real: housing, food, healthcare, care, retirement consumption, each on its own price trajectory. Some are indexed by design and some are of uncertain indexation.
The monetary denominator therefore matters differently depending on which kind of obligation is in view. The burden this series describes applies to the real and uncertain portions, not the whole.
Three things have to be said plainly, because each is easy to overstate. The general price index is not fraudulent and not meaningless; for a large fraction of any worker's consumption it is close enough. Liability-relative compensation is not fully measurable, and the point of naming it is to keep the true target visible while the proxies do their work. And the whole progression takes place inside a single monetary unit. Nominal, real and liability-relative compensation are all denominated in the same money. The corrections adjust for how much the unit buys. None of them asks what the unit is.
Where This Series Goes
The question becomes concrete the moment you ask what the compensation is for.
A salary paid this month and spent this month barely encounters the denominator. A salary paid this month and intended to fund consumption in thirty years encounters it completely, because whatever the worker holds across those years is measured, at the far end, against prices set in a unit that will have changed.
The first paper's insight was that a growing share of compensation is deferred. The second paper's was that the deferred portion needs an architecture. This one's is that the architecture is denominated, and the denomination is a design choice made, in current practice, by default.
Two notes before we start.
The first is about voice. I have spent roughly two decades inside pensions, benefits, savings plans and compensation, and I came to monetary questions late and from that direction. I am not a monetary economist telling the compensation profession it has misunderstood money. I am a compensation practitioner who ran into a question my profession did not seem equipped to answer, and followed it until it arrived at the money.
The second is a disclosure, and it belongs at the front rather than buried at the point it becomes relevant. This series ends by examining a particular monetary experiment, and I am the founder of a company that offers an employer-sponsored savings benefit denominated in it. You should weigh that. The paper is written under the second paper's own rule, which requires any application of the framework to disclose its weights, justify them against a specified liability, and evaluate complete architectures rather than isolated assets. Where the analysis finds that asset's role small, conditional or unsuitable, it says so, and by the end you will have read several pages of it doing exactly that.
The structure of the series reflects the order in which I actually worked the problem. It begins in compensation, develops a monetary question out of compensation's own materials, and only in its later articles examines the experiment that the argument makes hard to avoid.
That experiment is not the starting premise. Whether it belongs anywhere in a serious theory of compensation cannot be answered until the monetary question has been built without it.
Next week: why the worker became an investor, and the part of the answer that does not appear in the standard list.