The Actuary and the Abyss
How life assurance learned to measure uncertainty across a population, and what finance misunderstood when it carried the method into markets.
London, 1762. A new society begins offering life assurance on a principle that now appears obvious. The premium should reflect the age of the person entering the policy. It should be calculated from evidence about mortality, combined with the mathematics of compound interest, and set at a level intended to support the promise being made.
At the time, this was a substantial advance.
Earlier schemes had found ways to spread the financial consequences of death, but their pricing could be crude. Some restricted entry by age. Some collected broadly similar contributions from members whose prospects were clearly different. The arrangement might survive while claims remained manageable and enough new participants entered. Its weakness appeared when the contributions no longer matched the obligations.
The Society for Equitable Assurances on Lives and Survivorships approached the problem differently. It used age-related premiums designed to remain level over the policyholder’s life. The calculation recognised that the chance of death changes with age and that premiums collected today can earn interest before claims are paid.
The Society did not know when any particular member would die. It did not need to.
Its promise rested on a larger pattern. Across a sufficiently broad group of people, observed mortality could be organised into a table. That table could not reveal an individual’s future, but it could help estimate how many claims might arise across the pool, and when.
For the first time in this series, measurement does more than inform judgement. It becomes part of the institution’s foundations.
This is not a story about foolish mathematicians imposing an elegant fiction on the world. It is the opposite. The mathematics worked because the problem offered enough regularity for it to work.
That success changed finance.
It also created a temptation that would reach far beyond life assurance: if one form of uncertainty could be measured, priced and pooled, perhaps all uncertainty could eventually be treated the same way.
DODSON'S UNFINISHED DESIGN
James Dodson did not live to see the Equitable begin business.
In the 1750s, he sought admission to the Amicable Society, an established life assurance scheme, and was refused because he was older than its entry limit of 45. The rejection gave him a practical problem to solve. Could a society admit people at different ages if each paid a premium appropriate to the obligation being created?
Dodson believed it could.
His approach drew upon advances that had been gathering for decades. In 1693, the astronomer Edmond Halley had published a life table using records from the city of Breslau. The data was imperfect, and Halley’s work was not the last word in mortality measurement, but it demonstrated something profound. Death was unpredictable for the person and patterned across the population.
Dodson joined that insight to compound interest. A promise to pay money in the future does not require the full amount to be held in cash today. Premiums can be collected over time. Reserves can earn interest. Future claims can be translated into present values.
The task was to bring those relationships together without pretending to know the fate of any one policyholder.
Dodson calculated level premiums for entrants of different ages and worked on the principles of a society capable of supporting long-term assurances. He died in 1757. The attempts to secure a royal charter were unsuccessful, but others continued the project. In 1762, the Equitable was established through a deed of trust using the broad principles he had developed.
Its innovation was not merely charging older people more. It was building the price of a long promise from mortality evidence, interest and reserves.
That changed the nature of the business.
A premium was no longer simply a negotiated amount or a contribution shaped by custom. It became the output of a model linking present payments to future obligations.
The table revealed a pattern. It did not contain the world.
WHAT A MORTALITY TABLE CAN KNOW
A life table is easy to misunderstand.
It does not say that every person of the same age is equally healthy or faces the same future. It does not predict who will die next year. It summarises what happened across a defined population and turns those observations into estimated rates of survival and death by age.
The distinction between an individual and a pool is the heart of insurance.
Imagine 10,000 people of similar age. We cannot identify in advance which particular lives will end during the year. Yet if the population is sufficiently broad, the data relevant and the period reasonably stable, an insurer may estimate the total number of claims far more reliably than it can predict any individual claim.
That difference allows uncertainty to be pooled.
It does not remove uncertainty. Actual claims can still depart from the estimate. The insured population may differ from the one behind the table. Medical treatment, sanitation, nutrition and behaviour can alter mortality over time. War, pandemic, heat or disaster can affect many lives together.
Deaths are not perfectly independent, and mortality never stands still.
What makes life assurance workable is more modest. Across large and appropriately defined populations, mortality has often changed slowly enough for tables to be updated, premiums to be revised for new business, reserves to be recalculated and deviations to be absorbed with capital.
The model is useful because the institution is built around its limitations.
Data must be monitored. Assumptions must be reviewed. Reserves must be held. The pool must be sufficiently large and understood. Guarantees must be supported across decades, not merely at the moment a policy is sold.
This is the real achievement of actuarial science. It does not make the future certain. It converts a recurring population-level pattern into a promise that can be financed.
A DIFFERENT KIND OF UNCERTAINTY
The difficulty begins when the same confidence is carried into a system that behaves differently.
Human mortality and financial markets are both uncertain, but uncertainty alone does not make them equivalent.
A mortality table describes outcomes across people. The death of one policyholder does not ordinarily cause thousands of others to revise their expectations and die because the first death occurred. There are common shocks, but the everyday process is not dominated by participants responding strategically to the table itself.
Markets are reflexive.
Participants observe prices, estimates and one another. A model can influence the behaviour it is trying to describe. If investors come to believe an asset is safe, they may buy more of it, use more leverage and compress the return available for accepting its risk. The resulting stability can appear to validate the original belief. When conditions change, the same positioning can amplify the reversal.
The measurement enters the system.
That is the crucial difference.
An estimate of market behaviour drawn from the past may remain useful. But the people, instruments, regulations, leverage and sources of liquidity can change faster than a mortality population. Dependence that appeared weak can strengthen under pressure. Positions that looked diversified can become linked through funding, collateral or common ownership.
The problem is not that markets sometimes move a great distance. Large movement is not, by itself, risk. Volatility describes the scale or dispersion of returns. The danger lies in the exposure attached to that movement, the leverage supporting it, the liquidity available during it and whether the portfolio can remain intact long enough to continue.
A stable estimate of dispersion can help size a position. It cannot guarantee that the next regime will resemble the one from which the estimate was drawn.
Actuarial science did not fail by being mathematical. Trouble begins when the reliability achieved in one domain is treated as evidence that every uncertain domain will offer the same kind of regularity.
THE LONG SHADOW OF THE TABLE
Harry Markowitz changed portfolio construction in 1952 by insisting that investments should not be assessed in isolation.
The question was no longer simply whether an asset looked attractive. It was what the asset contributed to the portfolio as a whole.
Markowitz represented each investment through an expected return and the variance of its returns, then considered the covariance between assets. Two volatile holdings could sometimes create a less volatile portfolio if their movements differed sufficiently. From those relationships came the efficient frontier, the set of portfolios offering the highest expected return for a given level of variance, or the lowest variance for a given expected return.
This was a genuine breakthrough. It gave diversification a mathematical form.
It did not prove that variance was risk in every meaningful sense. Variance measures dispersion around an average, treating movement above and below that average within the same calculation. Investors may dislike losses more than equivalent gains, care about drawdown and ruin, or depend upon liquidity that disappears precisely when diversification is needed.
Nor does the basic construction of the efficient frontier require the claim that all returns are normally distributed. Normality makes mean-variance reasoning especially convenient because mean and variance then describe the distribution completely. Other assumptions, including particular forms of investor preference, can also support the framework.
The vulnerability lies elsewhere.
Expected returns, variances and covariances must be estimated. The optimiser treats those estimates as inputs, but it cannot know how durable they are. Small changes in an expected return can produce large changes in the selected portfolio. Correlations measured during ordinary periods may offer little guidance when participants are forced to reduce exposure together.
The mathematics can solve the problem it has been given perfectly while the problem itself has been specified badly.
William Sharpe’s capital asset pricing model extended mean-variance thinking into an equilibrium account of expected return. In its familiar form, an asset’s expected excess return depends on its beta, its covariance with the market relative to the market’s variance.
Again, the achievement was real. CAPM provided a common language for thinking about systematic exposure. Its simplified world was not offered as a complete description of every market event.
Normal returns are one sufficient route into mean-variance analysis, but it is too crude to say that CAPM simply assumes normality and therefore fails whenever returns are not normal. Its limitations involve a wider set of issues: the market portfolio, investor expectations, borrowing and lending conditions, transaction costs, changing betas and the gap between a one-period model and a market unfolding through time.
By the time Fischer Black and Myron Scholes published their option-pricing work in 1973, finance had acquired a powerful habit. Define the uncertainty. Specify the process. Translate future possibilities into a present price.
That habit had produced remarkable advances.
The next article will examine what happens when the assumed process meets a market capable of changing faster than the model.
THE EQUITABLE'S RECKONING
There is a grim symmetry in the later history of the Equitable, although it must be understood correctly.
The Society’s crisis was not caused by the failure of a mortality table. Nor was it proof that actuarial science had been useless.
From the 1950s through the 1980s, Equitable sold some pension policies containing guaranteed annuity rates. At retirement, the policyholder could convert accumulated benefits into income using either the guaranteed rate or the prevailing rate, whichever was more favourable under the policy terms.
For years, the guarantee appeared relatively unimportant. Then market annuity rates fell. The contractual option became valuable because eligible policyholders could secure more retirement income than current market terms would otherwise provide.
That value created a liability.
Equitable attempted to manage the cost by paying lower final bonuses to policyholders who exercised the guarantee. Policyholder Christopher Hyman challenged that approach. In 2000, the House of Lords ruled that the Society could not use its bonus discretion to neutralise the guaranteed benefit in that way.
The judgment exposed a financial weakness large enough to threaten the Society’s future. Equitable tried to find a buyer, failed, and closed to new business in December 2000. Existing policies continued, but the institution that had pioneered age-based life assurance could no longer continue as it had before.
It is tempting to turn this into a simple morality tale: interest rates departed from history, the model failed and the abyss opened.
The reality is more instructive.
A long-dated guarantee had become more valuable as the financial environment changed. The Society lacked sufficient financial strength to absorb the full consequence after losing the legal dispute. Bonus policy, disclosure, governance, reserving and regulation all formed part of the failure.
The lesson is not that guarantees are foolish or that actuaries cannot estimate liabilities.
It is that a promise can outlive the conditions under which it first appeared affordable.
Long-term institutions must survive not only the expected path, but also the interaction between contracts, markets, behaviour, law and capital when the path changes.
No mortality table can answer that problem on its own.
THE TABLE AND THE WORLD
The great contribution of actuarial science was not perfect prediction. It was disciplined approximation joined to institutional design.
Observe a population. Build a table. Calculate premiums and present values. Hold reserves. Review experience. Maintain enough capital for the outcome to differ from the estimate.
The method works when the data remain relevant, the population is understood and the institution respects the uncertainty around its calculations.
Markets demand an additional humility because the participants can respond to the model itself. The relationship being measured may change as capital moves, strategies spread and constraints tighten.
For an Outlier Hunter, the practical conclusion is not to reject measurement. ATR helps compare movement across markets and scale positions. Historical testing helps reveal how rules behaved across earlier regimes. Portfolio statistics help identify concentrations that intuition may miss.
But none of those tools turns dispersion into risk or converts the future into a known distribution.
The structure must carry what the estimate cannot.
That means beginning with small exposure, diversifying broadly, using predefined trailing exits and preserving the ability to continue after being wrong. It means allowing profitable positions room to develop because the rare outlier cannot be known in advance or reliably reduced to an average outcome.
The table remains useful.
It tells us what happened across a defined population under observed conditions. It can expose patterns that judgement alone would miss. It can support decisions, comparisons and promises.
It cannot certify that the domain will remain unchanged.
That boundary does not diminish mathematics. It tells us where architecture must take over.
The table revealed a pattern.
It did not contain the world.
Next: Article 4, The Formula That Ate the Market
Chicago, 1973.
The Chicago Board Options Exchange opens its doors. In the same year, Fischer Black and Myron Scholes publish a formula that will transform global finance. Its great insight is that, under specified conditions, an option’s payoff can be replicated through a dynamically adjusted position in the underlying asset and risk-free borrowing or lending. That makes it possible to derive the option’s price without forecasting the asset’s expected return.
The formula is elegant, practical and enormously influential.
Its assumptions are also about to meet the market.
Richard Brennan writes on systematic trading, complex adaptive markets, and the philosophical foundations of trend following at atstradingsolutions.com. His books include The Fractals of Finance, Complex Adaptive Markets, Carved by Impossibility and The Aussie Turtles Trend Following Guide.
Want to explore why structure exists at all?
Carved by Impossibility: What Remains When Everything Else Is Eliminated
The book explores the architecture of constraint, emergence, and reality itself, and what it means for how we understand markets, life, and the universe.
Available now on Amazon in paperback, hardcover, and Kindle.
Want the theoretical foundation for why trend following works?
The Fractals of Finance: Determinism, Adaptation and the Geometry of Markets
The book explores the full architecture of feedback, fat tails, and fractal structure in financial markets, and what it means for how we trade, invest, and understand risk.
Available now on Amazon in paperback, hardcover, and Kindle.
Want a practical field manual for trading trends and capturing outliers?
The Aussie Turtles Trend Following Guide: A Field Manual for Hunting Outliers adapts the timeless principles of the original Turtle traders into a systematic, rules-based approach for modern markets. Co-authored with Adam Havryliv.
Available now on Amazon in paperback, hardcover, and Kindle.