1. One restatement
Start with the sentence everything else rests on.
Compensation is the transfer of purchasing power through time.
Not the delivery of pay. The transfer of purchasing power, through time. Most people read that and find it unremarkable, which is the point. It is not a contrarian claim. It is a description of what compensation has always actually been doing, written in a way that makes the time dimension impossible to skip past.
The interesting part is what it costs you to accept it. If compensation is a transfer through time, then a measurement taken only at the moment of payment is a measurement of the starting point of a journey, and nothing else. Everything the profession does well, benchmarking, structure design, pay equity analysis, market pricing, happens at that starting point. Very little of it is designed to ask what condition the value is in when it arrives.
2. The Delivery Assumption
Give the gap a name and it becomes easier to argue with.
The Delivery Assumption is the inherited belief that compensation's obligations end at payroll. That what is delivered is what is received. That everything happening to the money afterwards belongs to financial planning, or investment management, or personal responsibility, but not to compensation.
It matters that this is not an idea anyone argues for. Nobody writes it down. It lives in instruments and institutions: in the fields a compensation system has and does not have, in what a total rewards statement chooses to display, in where a job description draws the boundary of the role. It is an assumption in the structural sense, which is why naming it changes anything at all.
There is a version of this claim that is too strong, and the papers are careful not to make it. Economics did examine the difference between what pay is worth at delivery and what it is worth over a lifetime. The Bulow and Ippolito debate about pension accrual is exactly that examination. The indexation research of the 1980s is exactly that examination. So the honest statement is narrower and, I think, more damning:
The Delivery Assumption is not an assumption economics failed to examine. It is the assumption the design field failed to inherit the examination of. No mainstream compensation textbook, framework, or total rewards model carries a variable for the unit value is denominated in, or an objective defined at arrival.
The nominal fallacy
The Delivery Assumption has a day-to-day corollary, which is the treatment of nominal value as economic value. Your salary went up three percent. Your bonus went up five. Your pension balance reads four hundred thousand dollars. Each of those sentences reports how many currency units a person has. None of them reports what happened to the time that was traded for them.
No competent practitioner confuses a nominal raise with a real one when the distinction is put to them directly. That is not the claim. The claim is that every system we have built produces the nominal answer by default, and the real answer only if somebody goes looking.
3. The compensation journey
If the measurement problem is that we only look at the start, the fix is to change the unit of analysis. Not the compensation event, but the compensation journey: the whole path the value of an hour of work travels, from the labour that earns it to the life that eventually spends it.
This immediately produces three registers that are usually collapsed into one word.
| Register | What it measures |
|---|---|
| Delivered | What the employer transfers. This is what every compensation system already records. |
| Arrived | The purchasing capacity that survives to the point of intended use, measured against what it has to pay for. |
| Perceived | What the employee believes they received. Related to neither of the other two in any reliable way. |
Almost every disagreement about whether compensation is generous turns out, on inspection, to be two people using different registers without saying which.
4. The Executive Exception
Here is the part that is hardest to dismiss, because the profession has already conceded it in practice.
For the five people at the top of an organisation, compensation design is exquisitely temporal. Committees spend hours on vesting schedules, performance periods, holding requirements after vesting, clawback windows, deferral elections, the tax treatment of the timing. The entire apparatus exists because everyone involved understands that when value is delivered and how long it is held changes what the award actually is.
For the five thousand people below them, we benchmark the number.
The Executive Exception is that contrast: demonstrated rigour about pay's temporal architecture for a handful of employees, and near-silence about it for everybody else. It is useful because it removes the easiest objection. Nobody can argue that thinking about compensation across time is impractical or exotic. We do it already. We do it for five people.
5. The Four Jobs of a paycheque
If you sort every dollar an employer transfers by what it is supposed to accomplish, you get four categories, and they are exhaustive at that level.
| Job | What it does | State of the science |
|---|---|---|
| Consumption | Funds the life being lived now. | Mature. Market pricing, structures, pay equity. |
| Protection | Absorbs shocks. Health, disability, life, income continuity. | Mature. Actuarial, insured, well modelled. |
| Alignment | Ties reward to outcomes over a defined period. | Mature. Incentive design, equity, deferral. |
| Preservation | Carries the value of today's labour to a distant future. | Orphaned. |
Three of those four have decades of theory, a professional body, a literature, and a set of instruments purpose-built for them. The fourth had one, and it was the defined benefit pension. When that institution receded, the job did not go away. Its owner did.
What replaced it was a savings account with a tax wrapper and a menu of funds, which is a fine thing but is not a preservation architecture. It hands the employee a set of decisions that the pension used to make on their behalf, without the pooling, the professional stewardship, or the outcome-denominated promise that made those decisions survivable.
The forced investor
Which produces a result worth stating carefully, because the careless version of it is wrong.
When the real return on the safest accessible claim is below zero after tax, standing still requires taking risk. The gap between zero and that real return is the preservation burden. When it is positive, an employee who wants to merely maintain the value of what they earned has to accept variance to do it.
Preservation-motivated investment is the portion of a worker's investment risk accepted to stand still rather than to get ahead. It is defined counterfactually and its size is genuinely unknown. I do not put a number on it and neither should anyone else. The concept does not need one. The observation is simply that we never asked employees to be investors. The design made them one.
6. What carries the value
If some part of an hour worked in 2026 is meant to fund a life in 2056, something has to transport it. That something is not an asset. It is an assembly.
The carrier is the instrument, wrapper and custody composition in which deferred value travels. Calling it a design variable rather than a neutral container is most of the work. Two employees can hold the identical fund and have materially different outcomes because one of them held it inside an architecture that survived a job change and the other did not.
Preservation architecture is the full stack: ten layers across six domains, covering instrument, wrapper, ownership, custody, contribution, defaults, pooling, tax, portability and conversion. Layers substitute for each other across normal ranges, which is why the stack is robust. Each one also has a failure region no other layer can compensate for, which is why the stack is worth auditing.
The target is a liability, not an asset property
This is the correction that does the most work in the second paper, and it is the one most often skipped.
Preservation is not a property an asset has. It is a relationship between a holding and a future life. The target is the worker's future liability: the stream of things they will have to pay for, tranched into a floor they cannot go below, a flexible middle, and an aspirational top.
Once the target is a liability, the question "is this asset a good store of value" stops being answerable in the abstract. A holding that is entirely wrong for the floor tranche can be defensible in the aspirational one. Most arguments about preservation assets are conducted without either party naming which tranche they mean.
The Six Tests
Any preservation architecture, existing or not yet invented, can be put to six questions.
| Test | The question it asks |
|---|---|
| Arrival | Does purchasing capacity sufficient for the liability actually show up when it is needed? |
| Assignment | Who bears each risk, and was that assignment chosen or merely inherited? |
| Ownership | Whose claim is it, and against whom? |
| Journey | Does it survive the events of a real working life, above all a change of employer? |
| Human | Can an ordinary person operate it under stress, without expertise, over decades? |
| Conversion | Does it turn back into spendable living at the far end, on acceptable terms? |
The tests are deliberately blunt and deliberately candidate-neutral. They were written before any asset was named, and the second paper ends without ranking anything, which was the discipline the argument required.
Property and promise
Every carrier sits somewhere between two poles. A promise is a claim against a named counterparty who has undertaken to deliver an outcome. Property is a holding with no counterparty at all.
They are complements, not rivals, and each does something the other cannot. Property carries; promises pool. Only a promise can spread longevity risk across a population. Only property survives the failure of the institution holding it. An architecture built entirely from one of them has a known failure mode.
Governed property is the interesting middle: worker-owned claims where the decisions are institutionally stewarded. Property at the claim layer, stewardship at the decision layer.
Two kinds of portability
When somebody changes jobs, what crosses the boundary?
Claim portability means the balance crosses. This we have, more or less. Architectural portability means the whole stack crosses: the contribution rate, the defaults, the pooling, the automation, the behaviour that was doing the real work. This we almost entirely do not have, and it is mostly a name for an absence. Australia's stapling reform is the first partial existence proof I am aware of.
This matters more than it sounds. Recent evidence has cut the estimated long-run effect of automatic enrolment considerably, and a large share of what is lost goes at job boundaries. The leakage is the Journey Test failing. That makes architectural portability the headline problem rather than a footnote to the defaults literature.
7. The monetary denominator
The third paper takes the thing the first two treated as background and makes it a variable.
Compensation is paid in money and measured in money. The profession pays in a unit and measures in the same unit, and therefore measures the payment with the payment. That would be fine if the unit held still.
The ruler is also the payment, and the ruler moves.
The denominator problem is what follows: nominal comparisons across long periods conceal the movement of the unit, because the measurement is being conducted in units of the vessel that is leaking.
The paper does not treat this as an indictment of elastic money. It defends the existing arrangement at full strength first, including the strongest available preservation instrument, inflation-linked government bonds, and only then assigns the residual: a decade of negative real yields, the taxation of phantom income, wide dispersion between the published index and the inflation an individual household actually experiences, the precedent of an index being redefined against holders, and hard capacity limits. Canada stopped issuing real return bonds in 2022.
Six desiderata for a monetary denominator are derived from the Six Tests before any candidate is named. Then one candidate is examined, and the finding is deliberately unflattering to the author's own commercial interest: it reassigns risk rather than reducing it, fails Arrival for floor liabilities, fails the Human Test under self-custody, and cannot pool. It is not a pension. Its bounded contribution is as a form of portable monetary property, which is a category, not a recommendation.
8. Where the employer's responsibility ends
The obvious objection arrives about here. If compensation is accountable for a forty-year journey, has the employer just been made responsible for four decades of markets and personal decisions?
No. And the boundary needs to be stated precisely enough to be usable.
The authorship principle. The employer is accountable for the architecture of its own increment. It is never accountable for the outcome of the journey.
You are answerable for what you designed and offered, for whether the thing you built survives a job change, for whether an ordinary person can operate it, for whether you told them the truth about what it does. You are not answerable for returns, for what someone chooses to do with their own property, or for a life you do not get to live on their behalf.
Two related commitments follow. Stewardship rather than retention: the employer as outfitter, equipping a traveller for a journey it will not accompany them on. And temporal agency rather than paternalism: employers owe people genuine agency over the timing and use of their economic claims, not temporal choices made for them.
There is also an ordering rule, which exists to stop this becoming an excuse to sell long-dated products to people who need cash. The adequacy gate: preservation instruments become offerable only behind funded present-pay adequacy. Emergency liquidity first, high-cost debt addressed, any available match captured. Preservation is what you build after the present is solvent, never instead of solving it.
9. What the theory does not claim
Worth stating plainly, because each of these has been read into the argument at some point.
- It does not claim inflation is a crisis. Two percent is the target, not a failure. The point is that at two percent, the value of a person's work halves in about thirty-five years, and a working life is longer than that.
- It does not claim any asset preserves value. It claims architectures do or do not deliver against a specified liability, and that assets are one layer of ten.
- It does not claim employers should make investment decisions for employees. The authorship principle exists specifically to rule that out.
- It does not offer a score. Individual-level scoring of temporal outcomes is prohibited vocabulary in this work, because it implies precision about preferences that cannot be measured. Plan-level audits are the honest version.
- It does not claim novelty for the observation that inflation erodes money. That is not the contribution. The contribution is the denomination variable, and the assembly of jobs, stack, tests and desiderata into one auditable system.
10. Where this goes
The theory's purpose is not descriptive. Time returned is the point of it: accumulated economic agency that reduces the number of future hours a person has to sell. Measured cautiously, as years to optionality, and never promised.
Because the most valuable thing an employee gives an organisation is not their productivity, or their expertise, or even their labour. It is their time. Money can be earned again. Careers can change. Companies can be rebuilt. Nobody gets Tuesday afternoon back.