Every return is a ratio, and every ratio has a denominator.
To say that an asset returned seven percent is to say that it gained against a particular measuring stick. That is not a complaint about finance, which has never been naive here. The distinction between nominal and real returns is as old as the field, and every practitioner knows it.
The deeper problem is which deflator, and it becomes visible the moment you accept last week's proposition that the target of preservation is the worker's future liability rather than an index.
A CPI-adjusted real return reports performance against a generalized consumer basket. For the consumption portion of a worker's liability, that is exactly the right question. But the liability is not a generalized basket. It is a particular future life: a house in a particular market, care costs with their own price trajectories, an income stream whose cost depends on interest rates and lifespans that nobody can pin down in advance.
So three concepts need separating.
The nominal return: how many more units of currency the worker holds.
The general real return: how much more of a standard consumption basket those holdings command.
And the liability-relative return: how much more of the worker's actual future liability the accumulated capital can now fund.
A portfolio can produce a positive nominal return, a positive real return, and still make barely any progress against a liability whose price is rising faster than the basket. Anyone who spent the 2010s saving for a house in Toronto or Vancouver or Austin lived inside that arithmetic without needing the vocabulary for it.
The third concept is the one preservation cares about. It is also the one that no statement reports. Not the plan statement, not the total rewards statement, not the annual benefits summary. Nowhere.
Even Safety Is Relative
The point generalizes further than it first appears, and this is where it started to change how I read plan menus.
A treasury bill is riskless in units of its own currency over its own term. It is risky in units of retirement income, because the price of retirement income moves while the bill's value stands still. When rates fall, every future dollar of income gets more expensive, and the holder of the bill has lost ground against the liability without a single statement showing a loss.
Follow that logic and the genuinely low-risk asset for the retirement-income portion of a liability is not cash. It is a deferred, inflation-indexed income stream, which is an instrument most workers have never been offered. That is Merton's point restated in this framework's vocabulary, and it is uncomfortable, because "we put the conservative money in cash and short bonds" is a sentence I have heard from thoughtful plan sponsors who were doing their honest best.
A Diagnostic, Handled Carefully
One exercise makes the denominator problem hard to unsee, and it has to be handled carefully because it is a diagnostic and nothing more.
Change the measuring stick and watch the story change.
Price a salary in houses in the worker's own city, and the last two decades read differently than they do in currency. Price a portfolio in years of median wages, and it changes again. Price it in the annuity income it could purchase, and it changes a third time, because interest rates move the price of income independently of what the portfolio is doing.
Each re-denomination asks the same question against a different portion of the liability: what can this claim actually command?
Gold belongs on that list as one diagnostic among several, useful precisely because it is nobody's liability and no committee resets it. Its own record disciplines any romantic use of it. Jastram found gold's purchasing power roughly constant only across very long spans, with deviations lasting decades and poor hedging in several inflationary episodes, and the scholarship on gold-anchored monetary regimes records instabilities and failures of their own.
The conclusion here is deliberately modest and deliberately double-edged. Perceptions of investment performance depend partly on the unit of account in which performance is measured. No traditional numeraire, whether currency, commodity or index, has an unimpeachable claim to be the measuring stick. And the economically relevant benchmark for preservation is therefore the liability itself.
That destination, it should be said plainly, is standard. Liability-driven institutional practice has benchmarked against liabilities for decades and would find none of this surprising. What is new is the route, walked for a readership that has never taken it, because compensation scholarship has not previously asked what its own deferred instruments are denominated in.
The Steelman, at Full Strength
Against that benchmark, the conventional evidence deserves its full weight, and it is genuinely strong. I want to state it properly rather than gesture at it, because the version of this argument that skips this paragraph is the version I do not want to be associated with.
The most comprehensive long-run data assembled, covering the major asset classes across sixteen countries since 1870, together with the global equity record, supports a clear conclusion. Diversified claims on productive assets have delivered substantial positive real returns over long horizons, across most times and places measured. Any preservation architecture that ignores productive assets is ignoring the most powerful accumulation engine in the historical record. That is not in dispute here.
What the framework adds is a precision about what that evidence establishes.
A long-run average real return describes a representative holder who stayed invested across the whole span. It does not by itself tell us about a particular worker's liability-relative outcome, which depends on the sequence of returns across their particular career, the fees and behavior of their particular stack, and the trajectory of their particular liability.
Put it this way: the rate of return on everything is not yet the rate of preservation of anyone.
And the gap is measured rather than rhetorical. Dollar-weighted returns, meaning the returns investors actually experienced given the timing of their flows, run persistently below the buy-and-hold returns the indices report. That gap between the published return and the experienced return is the empirical trace of everything I am calling architecture. Connecting the two is the work that architecture, and only architecture, performs.
Why This Is a Compensation Matter
The obvious objection is that all of this belongs to investment management, and that compensation professionals have no business in it.
I would have agreed a few years ago. Here is what changed my mind.
The liability-relative return is the number that no total rewards statement and no employer plan statement currently reports to the worker whose hour it is. We report contribution amounts, employer match, account balance, and one-year and five-year fund performance against a market index. Every one of those numbers answers the departure question. None of them answers the arrival question.
The employer did not choose the market. The employer did choose the menu, the default, the wrapper and the fee schedule through which that money travels, and those choices are compensation decisions made by compensation people. Which means the denominator problem reaches the worker through a door that our profession is standing in.
You do not have to solve the measurement problem to take it seriously. The first move is smaller than that: stop treating the index as the neutral background against which performance is read, and start asking what the number would look like if the denominator were the life it exists to fund.
Next week: three things changed at the same time, and the interaction is the part nobody has been assigned to notice.