Start with the plainest fact about work. Someone gives an hour of their life to an employer, and the hour does not come back. Money can be earned again. Portfolios recover. Careers restart. The hour does not.
The first paper in this series argued that compensation is the transfer of purchasing power through time, and that the profession which designs compensation has historically measured that transfer only at its starting point. It named the Delivery Assumption, moved the unit of analysis from the compensation event to the compensation journey, and identified Preservation as the fourth job of compensation: the one that carries the value of today's labor into the distant future, and the only one of the four without a mature science of its own inside compensation practice.
This second paper takes up the question that follows, and it turned out to be harder than it sounded when I wrote it down.
If part of an hour worked in 2026 is meant to fund a life in 2056, something has to transport the economic value of that hour across thirty years. The transport does not happen by itself. Value has to be held in something, owned somehow, kept somewhere, moved when careers move, defended against a long list of hazards, and finally converted back into the living that was deferred. The whole apparatus that does this, and not any single asset inside it, is what I have called preservation architecture.
The Question I Am Not Asking
The question is not which investment is best. That question has a large and capable literature, and I have nothing to add to it.
The prior question is stranger once you sit with it: what properties should an architecture possess if we expect it to carry the economic value of finite human labor into an uncertain future? Retirement planning assumes an answer to this. Asset allocation assumes an answer to it. Every default fund in every plan you have ever administered assumes an answer to it. The second paper is an attempt to earn one rather than assume it.
It helps to see the whole sequence at once. Human time and human capital together produce labor. Labor produces compensation. Some compensation is consumed and some is saved. Saving becomes capital. And preservation architecture determines what that capital is still worth when the future arrives to spend it.
Every link in that chain is the subject of a mature discipline except one: the system-level design problem connecting compensation delivery to durable future consumption. Human capital formation has been studied since Becker. The lifecycle of earning and consuming has been modeled since Modigliani and Friedman. The interaction of human capital with portfolio choice was formalized by Bodie, Merton and Samuelson. Converting a stock into late-life income is the home terrain of actuarial science. The pieces are all there. Nobody owns the assembly.
Compensation scholarship has been precise about this boundary without meaning to be. The field's research syntheses examine pay level, structure and basis with real rigor. The standard texts organize around attraction, retention and motivation. The benefits literature studies retirement plans as objects of employee satisfaction, valuation and communication. What that literature does not study is whether the plans preserve. The question is not answered badly. It is not on the agenda.
What an Endowment Knows
There is one institution whose entire purpose is preservation, and it is worth asking what it knows.
A university endowment exists to support present spending while preserving purchasing power for beneficiaries who do not yet exist. Endowment practice has spent decades on exactly the questions this paper asks: spending rules that balance today against tomorrow, portfolio construction for generational horizons, governance that outlives any individual steward. If preservation has an institutional science, it lives here and in the pension funds that share its structure.
So try the comparison. An employee can be understood, cautiously, as a finite-duration endowment funded by human capital. The working years throw off income the way a corpus throws off returns. Some must be spent and some preserved. The preservation must survive changing conditions.
The analogy is useful right up to the point where it breaks, and the breaks are the whole point.
The endowment is perpetual. The employee is not, and does not know their own horizon, which may be cruelly short or expensively long. That single difference carries a corollary practitioners will recognize: an endowment's celebrated tolerance for illiquid assets is a function of its perpetual horizon and its governance, which are precisely the two things the worker lacks. The popular advice to invest like an endowment imports the strategy while omitting the conditions that make it safe.
The endowment employs professionals. The employee is an amateur at the one task on which their later life depends.
The endowment has governance, committees and institutional memory. The employee has evenings. Read this way, the individual's documented failures during market drawdowns look less like a character defect and more like the absence of a committee. No quorum stands between the panicked holder and the sell order.
The endowment pools nothing, because its horizon is indefinite. The employee faces longevity risk that no individual balance sheet can efficiently self-insure.
The endowment's beneficiaries are abstractions. The employee's beneficiary is their own future self, whom the behavioral research tells us is treated, in decision after decision, like a stranger.
And the endowment never changes sponsors, while the modern employee moves through many employers while remaining one economic unit.
The Disanalogies Are the Specification
That list is usually presented as a reason the comparison fails. I think it is the most useful output the comparison produces, because the differences are not weaknesses of the analogy. They are the design specification, and they were written by the structure of the problem rather than by anyone's theory of assets.
Stated carefully: institutional preservation systems possess capabilities that individual workers structurally lack. Governance and delegated expertise, documented in the endowment literature. Pooling, documented in the actuarial literature since Yaari. The enforcement of present discipline on behalf of a future self, documented in the behavioral literature whose founding models treat the saver as two selves in systematic conflict.
Preservation architecture exists to supply institutionally what the individual cannot self-supply: the governance the employee lacks, the pooling the individual balance sheet cannot manufacture, the continuity the career no longer provides, and the discipline the future self cannot enforce on the present one.
Whatever the architecture turns out to be, that is the job description.
Where This Goes
Over the next seven weeks I want to walk through the framework the second paper builds. What preservation is actually preserving, which is less obvious than it sounds. What return means once you ask "relative to what." The three changes that made this problem harder at the same time. Who is carrying which risk, and whether anyone chose. Why the stack matters more than the instrument. Why an instrument that carries value well becomes harder to leave, and what that does to retention design. And six tests that can be applied to any preservation architecture, including ones that do not exist yet.
One thing to say at the outset, because it governs everything after it. This is not a paper arguing that the inherited system has failed. The opposite is closer to the truth, and week five will make that case as strongly as I know how. The argument is narrower: the architecture we inherited was assembled under employment, institutional and monetary conditions that no longer exist in the same form, and nobody in compensation has been assigned the job of noticing.
Underneath the formal question sits a human one, and it is the one I keep returning to. At the start of a career, "what is your time worth" sounds like a question about wages, and the labor market answers it every payday. But that answer is only the opening entry in a longer ledger. The real question is how much of the value of an hour you can never recover will still be there when your future self needs it.