A pattern runs through every preservation institution I looked at, and it is worth stating as a principle. It also needs stating carefully, because the careless version invites attack.
When an architecture is chosen, risk is rarely eliminated. Most preservation design reallocates risk, transforms it, pools it, hedges it, or concentrates it. And the deepest consequence of an architectural choice is the resulting assignment: who ends up carrying what.
The principle beneath the pattern is this. Risk-bearing is an architectural choice even when nobody consciously makes it.
The Ledger of Transfers
Run the inventory through that lens and the pattern is consistent enough to be uncomfortable.
The defined benefit plan did not eliminate market and longevity risk. It moved them onto a sponsor's balance sheet and manufactured a new exposure, promise risk, in the process.
The defined contribution plan did not eliminate promise risk. It handed market, longevity and sequence risk to the individual, and manufactured behavioral risk, meaning the possibility that the person will not or cannot execute, at scale.
The annuity does not eliminate longevity risk for the system. It pools it, and manufactures counterparty dependence and illiquidity.
Indexed bonds hand inflation risk to a treasury and accept the treasury's own risk in exchange.
Even holding cash is an assignment. It trades market risk for the full, unhedged weight of whatever the monetary era does to idle balances.
Each of those moves is defensible. Several of them are excellent. The point is not that any of them was wrong. It is that each one is an assignment, and the assignment is the thing nobody writes down.
What Genuinely Reduces Risk
The principle is only honest if it names the exceptions and their limits, because "risk is only ever moved" is the kind of claim that sounds sophisticated and is false.
Pooling across independent exposures is the most important genuine reducer. Mortality pooling turns a thousand uncertain lifespans into one nearly certain liability, and no individual holding of any asset can replicate it.
A precision the actuarial literature insists on belongs right next to that. Pooling reduces idiosyncratic longevity risk, meaning the uncertainty of one person's lifespan around the average. It does not reduce systematic longevity risk, meaning the possibility that the whole cohort lives longer than the tables predicted. That has to be carried by a guarantor, transferred through reinsurance, or shared by adjusting benefits. Somebody still holds it.
Pooling is also not free. Insurance markets carry adverse selection, loading costs, basis risk, and the possibility that supposedly independent exposures turn out to be correlated exactly when it matters.
Diversification reduces idiosyncratic asset risk while the common risk remains.
Hedging transforms exposures rather than removing them, at a price.
And prevention exists, which gets forgotten in every conversation about portfolios. Working longer, protecting health, and maintaining the human capital that funds the whole chain reduce risk at its source. An architecture can support those or undermine them, and most benefits programs contain instruments on both sides of that line without anyone having mapped it.
The refined principle, then: risk reduction is real but scarce, purchased mainly through pooling, diversification and prevention, each with costs and conditions. Everything else in the inventory is a transfer with a transformation attached.
The Most Practical Question in the Paper
What follows is the same either way, and it may be the single most useful thing the framework produces.
For any architecture: name every risk, ask who bears it, and ask whether anyone chose that assignment.
Grouping the roster keeps it usable. Four groups.
Risks to value: inflation, deflation, market decline, sequence-of-returns.
Risks to the claim: counterparty, issuer, governance, custody, regulatory change, confiscation and capital controls.
Risks of fit: longevity in both directions, liquidity need, portability failure, concentration, behavioral abandonment.
And systemic risks: institutional insolvency and technological change, which can strike any layer.
Most workers today carry an assignment of those risks that nobody designed. It was inherited from the accidents of plan history, and it is invisible precisely because no instrument reports it.
The first paper argued that the profession measures deliveries rather than journeys. This sharpens the point. It also fails to measure who is carrying what.
Two Places the Assignment Question Bites Hardest
The first is sequence risk, because it is the assignment that is least visible and most consequential.
Two workers with identical average returns across identical careers can retire into materially different lives depending on the order in which the returns arrived. Nothing in a contribution rate, a match formula, or a fund's ten-year performance number carries any information about it. It was assigned to the individual by the shift to individually directed accumulation, and it was assigned silently.
The second is the sovereign layer, and it is the one I expect to be most argued with.
The instruments conventionally treated as the low-risk foundation of household portfolios, meaning government bonds and the deposits and funds built on them, are claims on sovereign balance sheets. The historical record on those is extensive and sobering without being apocalyptic. The most instructive channel is what Reinhart and Sbrancia called the liquidation of government debt: extended periods, with the postwar decades prominent among them, in which moderate inflation, regulated interest rates, negative real yields and captive domestic savings steadily reduced real public debt burdens.
No default occurred. Every contractual nominal promise was honored. And the holders of government obligations, disproportionately ordinary savers through their banks, insurers and pension funds, experienced a sustained real transfer all the same. Reinhart and Sbrancia estimate the annual effect at three to four percent of GDP for the United States and United Kingdom across the postwar decades they measure.
I want to be exact about what that does and does not establish. It was achieved gradually, under institutional preconditions, including rate caps, captive financial systems and capital controls, that no advanced economy currently operates in full. It is a demonstrated mechanism with documented boundaries, not a standing condition. Nothing here says a crisis is coming, and the paper makes no such claim.
What it demonstrates is the distinction the whole framework has been building toward. A sovereign bond can satisfy its contractual nominal promise completely while failing the preservation objective for which the worker held it.
The defensible proposition is narrow and sufficient: high sovereign indebtedness introduces additional failure modes into long-duration preservation architecture, and those failure modes attach precisely to the instruments that conventional design treats as the safe foundation.
That is not an argument against holding government bonds, which remain among the best instruments ever devised for several preservation sub-tasks. It is an argument that the risk they carry should be named, assigned and priced like every other risk in the taxonomy, rather than defined out of existence by the label "risk-free."
Why It Lands Inside Compensation
The assignment question could stay in macroeconomics if it were not for one thing.
The default funds, target-date glide paths and stable-value options through which employer plans deliver the deferred increment are built on precisely these instruments. The sovereign balance sheet's risks reach the worker by way of choices that the employer's plan design made.
That is what makes this a compensation matter and not only a macroeconomic one. Somebody selected the default. The risks that default carries were assigned to a workforce that was never shown the assignment, by a process that was never asked to document it.
Naming the assignment costs nothing. It is a one-page exercise using information the plan already has. And I have yet to see it done.
Next week: why the instrument matters less than the ten layers around it, and why the default menu has become one of the most consequential compensation decisions an employer makes.