The Death Bet: How Wall Street Puts a Price on Human Life
Section 1: The Trade Nobody Talks About
Every year, Americans walk away from more than nine million life insurance policies, over $725 billion in death benefit, by letting them surrender. Nearly 88% of universal life policies never result in a death benefit claim and 76% of universal life policies sold to seniors at age 65 never pay out.
The fix has existed for over 100 years and almost no one knows about it. In 1911, a decision of Justice Oliver Wendell Holmes established that a life insurance policy is considered an asset, and therefore a client has the right to sell it. A patient named John Burchard, in need of money for surgery, sold his life insurance policy to his physician for $100. When Burchard died a year later, the doctor's claim to the death benefit was challenged and eventually reached the United States Supreme Court. Justice Holmes ruled in favour of the doctor, making a $100 transaction the legal foundation of a multi-billion dollar industry.
Today, the market's own trade body, the Life Insurance Settlement Association, reports that its licensed buyers complete roughly 3,000 transactions a year covering about $3.7 billion in face value, though only around $627 million actually reaches the sellers. Pension funds, asset managers and hedge funds are the primary buyers, and the spread between what they pay and what they collect, net of every premium in between, is the trade. The market has two layers: the secondary market, where policies are bought directly from seniors, and the tertiary market, where institutions trade already-purchased policies amongst themselves. According to AIR Asset Management, a life settlement fund manager, the tertiary market now substantially exceeds the secondary. Institutions are not just buying these policies, they are trading them like any other asset. That is no longer a niche curiosity, it is a functioning market.
But it is also a market built on a single number, and almost nobody involved has a strong incentive to get that number right. Right now, there's a real reason to think that number could be wrong for everyone in this market at the same time, not just one policy here or there. The opportunity is real. So is the fragility.
Section 2: Pricing the Worth of a Human Life
When a hedge fund buys a portfolio of various life settlements, they are placing a mathematical bet on the exact date a human dies. The bet must factor in how much the policy pays out when the insured dies, the price to keep paying premiums until then, and how much longer the person is expected to live.
In order to do this, they hire third party medical underwriting firms. A handful dominate the market, among them AVS Underwriting, TwentyFirst, and Fasano Associates. The last two take opposite approaches. Both start from a baseline mortality table (how long an average person of that age and sex is expected to live) then adjust it for the individual, assigning debits for conditions that shorten life and credits for factors that extend it. The difference is who makes that call. Fasano puts physicians at the centre of the assessment, reading the medical file and applying clinical judgment. TwentyFirst is deliberately rules-based, running the same inputs through a fixed system so the answer does not depend on which underwriter happens to open the file.
Either way the debits and credits get combined into a score called a mortality multiplier (A multiplier of 200% means this person is twice as likely to die in a given year as an average healthy person their age, which pulls their entire life expectancy estimate shorter). Underwriters feed medical history, family history, body mass index, and lifestyle factors into that calculation, and what comes back to the buyer is a single mean or median life expectancy (LE). That number is the midpoint of a distribution, not a prediction about any one person. Pricing and leverage both treat it like one.
The buyer then runs a simple calculation. If I pay X today, keep paying premiums of Y every year, and collect Z when this person dies, what return does that generate, and at what death date does that return become unacceptable?
Think of it like a bond with an unknown maturity date. With a traditional bond, you know exactly when you get your money back: you buy the bond, collect coupons along the way, and receive the face value on a fixed date. When dealing in life insurance policies, there is no fixed date. The death benefit is the face value, the premiums are the negative coupons you pay instead of receive, and the maturity date is whenever this person dies. Every month the insured lives longer than expected is another premium payment out of pocket and another month before the payout arrives. The longer they live, the worse the return gets.
TwentyFirst and Fasano are not competitors. Both sit under Longevity Holdings (the rebranded parent of what was formerly ITM TwentyFirst, now a Stone Point Capital portfolio company) which also owns LexServ, by its own description the largest portfolio servicer in the secondary insurance market, along with Longevity Trading & Analytics, which sells independent valuation and pricing. The company says it has produced over 450,000 underwritings on 143,000 unique lives and services more than half of all outstanding life settlement policies. Underwriting sets the LE estimate, valuation prices the policy off it, and servicing later reports whether the person actually died on schedule, the one real feedback loop that would catch a bad estimate. At Longevity Holdings, all three sit under one roof, so that check never comes from outside the company.
And that is not the only problem with how these estimates get produced. There is a documented conflict of interest baked into the chain. On the seller's side, the broker represents the individual, often a senior citizen, selling the policy and earns a commission on the sale price. A shorter LE estimate means higher bids from buyers, a better outcome for the seller, and a bigger payday for the broker. Even life settlement companies themselves aren't immune: aggressive LE estimates yield higher offer prices, which can win more deals and grow assets under management. This is a bias against the fund's own long-term investors that only shows up years later, once the policies have been bought and the premiums start running long.
In practice, the only real check on this system is competitors watching each other, and once in a while it catches something. One provider, Coventry, ran a study on a rival firm, Lapetus Solutions, and found that Lapetus was telling institutional buyers people would die sooner than other firms predicted, in about 85% of cases, by an average of two and a half years. Independent academics then re-ran the analysis themselves and got the same result. A shorter estimate like that makes a policy look more attractive to buy, since it implies a faster payout, which is exactly the kind of bias this market is supposed to guard against. This is what the system looks like when it works: a competitor with something to gain from finding the problem found it, and outsiders checked the math before anyone relied on it. What the market doesn't have is a version of that check that runs on its own, without needing a rival to go looking first. That gap caught up with the market in 2007.
In late 2007, two leading LE firms revised their own mortality tables, quietly admitting that the people in their portfolios were going to live longer than they'd estimated. In February 2008 the Valuation Basic Tables were updated as well (the industry-standard mortality tables, built from pooled insurer death-claims data, that the entire life insurance business uses to price and reserve for policies). Two firms quietly changing their minds was one thing. An industry-wide table update, affecting every insurer and every policy priced off it, was another.
Policies tied to those tables repriced downward overnight. For funds that had borrowed money against them, the math turned ugly. Consider an example: assets that were worth $100 on paper were suddenly worth $70, but the loan against them was still $80. Lenders called their money back. Funds that couldn't pay had to dump portfolios at whatever price the market would bear. That is distressed selling. It is what happens when a fund holding a leveraged, illiquid asset finds out its single most important input was wrong.
Section 3: The Machinery
A single life settlement is a bet on one person's lifespan. A fund holding ten thousand of them is a business. Getting to that scale takes the same machinery Wall Street built for mortgages and corporate debt.
Longevity Holdings isn't the only place this consolidation is happening. It's happening at the fund level too. Abacus Global Management is one of the largest publicly traded life settlement companies, buying policies and managing portfolios. In 2023, Abacus went public and started acquiring immediately. In 2024 it bought Carlisle Management, a Luxembourg-based life settlement fund manager overseeing $2 billion of client money, for $200 million, a deal that closed that December.
Here's why that price is strange. Companies are usually valued as a multiple of their yearly profit, pay 12 times what a business earns in a year, say, and that's a normal deal. You only pay far more than that when a business is growing fast, because you're betting on much bigger profits down the road. Carlisle's profits were falling. There are ordinary reasons to still pay up for a deal like this, more assets under management, a foothold in Europe, but none of them usually justify tripling the standard multiple for a business getting less profitable, not more. Abacus still paid, by one short seller's calculation, roughly 36 times earnings, about triple the usual rate. That only makes sense if Abacus wasn't buying Carlisle's profits. It was buying control: $2 billion of policies, and one more piece of the pipeline brought in-house, regardless of what the earnings said it was worth. In an industry where the entire trade depends on one unverifiable number, controlling more of the chain that produces that number is worth more than the profit Carlisle happened to be generating that year. With that scale, Abacus started tapping the capital markets directly. But first, it needed financing.
Before you can bundle these policies into something sellable, you need money upfront to buy enough of them to make the bundle worthwhile. That money usually comes from a bank, in the form of a loan against the policies the fund already owns. The bank that arranges this loan is first in line to get paid a fee, the opening charge in a chain of fees that runs all the way through the deal.
Once the pool of purchased policies is big enough, it can be securitized. In its fullest form, that means bundling the policies together, having a rating agency assess the risk, and dividing the pool into tranches (slices of varying risk and return), which are then packaged into notes (essentially bonds) and sold to institutional investors. That is the same chain used in every other asset-backed securities market. In practice, though, a rated, tranched deal is the exception rather than the rule. In October 2025, Abacus Global Management took a first step down that road, selling $50 million of securitized life insurance assets to banks, pension funds and insurers at a mid-single-digit yield, but as a single collateralized note rather than a tranched structure, so every investor in the deal holds the same slice of risk with no senior layer absorbing losses first. Abacus describes the note as above investment grade. It does not name the rating agency.
That's not an isolated deal. It's the same instinct behind Longevity Holdings owning the underwriter, the servicer, and the valuation firm all at once. Different company, same logic: whoever controls the most steps in this chain controls the most valuable thing in it, which is why buyers keep paying up for control rather than for profit. And every step that gets folded into one company is a step where the outside check, a second firm, with its own incentives, looking over the first firm's shoulder, disappears. That's the real cost of this kind of consolidation. It isn't just higher fees. It's fewer and fewer people checking whether the number this entire market runs on (how long someone has left to live) is honest.
Section 4: The Pitch
The pitch in favour of life settlements to institutional investors is simple: returns that don't move with the stock market. Whether markets boom or bust, people die at roughly predictable rates. A recession does not change when someone passes away. A rate hike does not accelerate your date of death. A portfolio of life settlements should in theory be completely unaffected by whatever is happening in the market.
For institutions perpetually hunting yield without adding correlated risk, the numbers are attractive. Life settlement funds historically target IRRs of 8 to 12%, equity-like returns with cash flows driven by actuarial tables rather than companies' earnings cycles. The long duration of the asset matches what pension funds actually need: predictable future cash flows to pay obligations to their members years from now. 11,200 Americans turn 65 every single day through 2027, and by 2030, every baby boomer will be 65 or older, according to the U.S. Census Bureau. Baby boomers own life insurance at higher rates than any other generation, and institutional buyers' strongest offers typically come for insureds between 70 and 85, an age bracket this generation is now entering at scale. Supply should grow. It hasn't yet. Transactions fell from 3,218 in 2023 to 2,699 in 2024 before recovering to 2,955 in 2025, and the dollars reaching sellers are still below their 2023 peak. Even Conning (the research firm the industry itself relies on for market forecasts) called the 2024 decline a "pause," insisting the long-term growth is still coming. Maybe they're right. But "pause" is doing a lot of work in that sentence, and it's coming from the same firm that has been promising an unstoppable demographic wave for years.
In 2025, 2,955 transactions were completed against more than nine million policies that lapse or surrender annually across all ages, a penetration rate of roughly three hundredths of one percent by policy count. Either this is one of the most overlooked opportunities in alternative assets, or there are structural reasons the market has stayed this small despite decades of existence.
Whether this market grows is one question. Whether the portfolios already inside it are priced correctly is a separate one. The second question is more urgent, because if the model is wrong, it's wrong right now, on every policy that already exists.
Section 5: Where the Bet Goes Wrong
Risk 1: The Model Can Be Wrong For Everyone At Once
The bull case for life settlements rests on one idea, human mortality is random. This means that a large enough pool of policies diversifies the risk away. If one person lives longer than expected, another dies sooner, and it averages out. That logic is sound up to a point. It breaks when everyone starts living longer at the same time.
Buying thousands of policies protects you against individual variation. It does nothing to protect you against a shift in the mortality curve. A change that moves every policy in every portfolio in the same direction at the same time. And that is exactly what may be happening right now.
Take Ozempic. The SELECT trial, a clinical study on semaglutide, the drug behind Ozempic, found that it reduced major cardiovascular events like heart attacks and strokes by 20% in overweight patients with existing heart disease. That matters because the typical life settlement insured is in their 70s, and cardiovascular disease is one of the primary conditions life settlement underwriters analyze when estimating how long someone will live, it is the impairment that makes a policy worth buying. A drug that meaningfully reduces heart disease in that age group is a hit to the core assumption on which an entire life settlement portfolio is priced.
Here's the catch. The strongest evidence for this drug's benefits comes from people who don't look much like the people in these portfolios. The SELECT trial (the main study behind the 20% cardiovascular improvement) tested patients whose average age was 61.6. The typical life settlement seller is closer to 75. That's not a small gap. It's most of a generation. The SELECT trial only enrolled patients with a BMI of 27 or higher, the medical definition of overweight. That matters because a higher BMI has traditionally been treated as a factor that shortens, not lengthens, a LE estimate. The trial's dramatic result came from exactly the group underwriters already treat as higher-risk on weight, which says nothing about whether the same benefit shows up in someone who isn't overweight to begin with.
The pattern holds beyond this one study. Munich Re (one of the world's largest reinsurers, whose entire business depends on getting mortality right) found the same drug produces a smaller benefit as people get older, with the weakest effect in the group over 60, and even their data didn't go past age 80.
Then there's the question of who's actually taking the drug. Among adults 50 to 64, about 22% currently use a GLP-1 drug. Among adults 65 and older, that drops to 9%, largely because Medicare has historically refused to cover the drug for weight loss, and many older Americans are on Medicare. And of the people who do start taking it, a large share quit within the first year.
Even the strongest recent test aimed squarely at older adults came back negative. In November 2025, Novo Nordisk ran trials specifically on adults aged 55 to 85 (essentially the exact age group life settlements are built on) to see whether the drug slowed Alzheimer's disease. It didn't work.
Put all of that together, and the honest version of this risk is narrower than it first sounds. These drugs may well be changing how long people live. Whether they're changing it for a 78-year-old with heart failure, in an age group that's barely taking the drug in the first place, is a much weaker, much less certain claim.
But "less certain" doesn't mean "not worth checking." If people in that age group are living even somewhat longer because of a drug that barely existed five years ago, the LE estimates those portfolios are built on may be miscalibrated. The industry's answer is that improvement is already priced in: TwentyFirst (one of the biggest firms producing these estimates) already builds in an assumption that people will keep living a little longer each year, adding roughly 0% to 1.5% onto its LE estimates annually, and it says its past predictions have been accurate, tracking close to actual outcomes over the last thirteen years. That built-in cushion is roughly the same size as the GLP-1 effect reinsurers are now forecasting. So in theory, the model may already be covering this but that cushion was sized by looking backward at how LE improved over the twentieth century, before this drug existed. Nobody outside a small number of firms can check whether that same number is actually big enough for the specific kind of person these portfolios hold: someone in their late 70s with heart disease.
With a stock, if your price guess is wrong, the market corrects you within a day, thousands of trades tell you so. There's no equivalent for these LE numbers. Nobody trades this exact policy on any public exchange, so there's nothing out there to catch it if the estimate is wrong. So the real question isn't whether the model accounts for medical progress at all. It does. The model already assumes people will keep living a bit longer each year. The question is whether "a bit longer" is still the right amount now that a new drug exists.
Only a handful of firms in the country do this kind of work at all (they're the ones with the medical files and the historical death data) so they're the only ones equipped to even attempt an answer. And nobody checks it. There's no regulator whose job this is, no required audit, and as we just saw, no public market price that would expose it if they got the number wrong. It's worth noting how uncertain this whole field currently is, even in its most institutionalized corner. LE firms use their own proprietary mortality tables, but the pension industry relies on a shared, public benchmark from the Society of Actuaries (SOA), and that benchmark is frozen. The SOA updated it every year from 2015 through 2021, then stopped, judging pandemic era mortality data too distorted. As of its October 2025 update, it still hasn't issued a new scale, and it says it doesn't expect to produce one in 2026 either. In the meantime, it has declined to publish a COVID-adjusted scale at all, telling individual actuaries to make their own adjustments using their own judgment. If the best-resourced, most scrutinized corner of mortality projection currently can't say what the trend line looks like, it's worth asking how much confidence to place in a corner of the market where the same question gets answered privately, by a handful of firms, with nobody checking the answer.
So what does the industry itself think the actual effect size is? Three of the world's biggest reinsurers have each tried to answer that question, and they don't agree. Swiss Re projects GLP-1 drugs could reduce US mortality by up to 6.4% by 2045, though that's their optimistic scenario, against a 4.0% baseline and a 2.3% pessimistic case. Munich Re puts the number at 0.2% to 0.5% annual improvement over twenty years, and argues the drugs will simply confirm improvement the industry already expected rather than add anything new. RGA lands in the middle: a central estimate of 3.5%, with a range of 1.0% to 8.8% depending on how many people actually take the drug and how well it works.
It is worth noting that these three companies sell insurance and reinsurance for a living, and part of that business is convincing clients they have this under control. Their numbers aren't neutral facts handed down from nowhere, they are best guesses from companies with a reason to sound calm rather than alarmed.
Even the industry's own numbers admit the effect isn't the same for everyone. RGA's headline figure is 3.5%, a single number that gets repeated as if it applies evenly across every age. It doesn't. The same research shows the benefit strongest for people aged 45 - 59 and shrinking with age, weakest of all in the 85-plus bracket (the exact age range these portfolios are built on). It doesn't disappear. But it's meaningfully smaller exactly where the money is. If even the industry's own study can't produce one clean number that actually applies to a 78-year-old, that's reason enough to be skeptical of any single tidy figure claiming to capture what this drug means for this market.
There's also a simpler reason this might not fully show up in the models yet: it's too new. Mortality tables that predict future improvement are built by looking at what's happened in the past, and GLP-1 drugs have only been used at large scale for the last two or three years. There hasn't been enough time for their effect to show up in the historical death data these models depend on. That's not a flaw, It's just how the math works. You can't measure a trend using data that doesn't exist yet. If the reinsurers underestimated how this plays out for older, more impaired insureds, some portfolios in this market may already be worth less than the model currently says, and there hasn't been time for that loss to show up in anyone's numbers.
Risk 2: This Failure Model Has A Precedent
This has happened before, though not for the same reason. In 2008, banks weren't just guessing how many mortgages would default, they were also guessing how connected those defaults would be to each other, the idea that a wave of defaults in one city wouldn't spread to another. That assumption turned out to be wrong. When housing prices fell everywhere at once, defaults rose everywhere as well. And because so many different mortgage bonds had all been priced using that same flawed assumption, they all got marked down on the same day, not because each bond had suddenly gotten worse, but because the formula used to price all of them had just been proven wrong.
Life settlements run on the same kind of shared assumption. One person dying doesn't make the next person more likely to die, deaths aren't connected to each other the way defaults were. But every portfolio in this market is priced using the same handful of LE models. If that shared assumption is wrong, it isn't wrong for one policy. It's wrong for every policy that uses it, all at once.
There is even a precedent specific to this market. In the 1980s, AIDS was a near-certain death sentence within a few years, and funds began buying life insurance policies directly from patients, paying cash up front in exchange for the death benefit once the person died, betting on how short that wait would be. Half of those diagnosed died within the first year, and 85% within three, so investors were reasonably confident they'd collect within a short window.
Then, in 1996, new AIDS treatments arrived, and patients started living years, sometimes decades, longer than anyone had modeled. Most of the companies in that business didn't survive it. Funds that expected a payout in two years ended up paying premiums for a decade instead. Medicine changed faster than the models could.
That collapse is also where the modern life settlement industry comes from. The original business only worked for the terminally ill, and once treatment improved, there weren't enough of those cases left. So it shifted to a new customer: seniors with policies they no longer needed or could afford. That shift is what "life settlements" is. This isn't a new risk. The exact same bet, pay someone now based on a guess about how long they'll live, already failed once, under a different name, for the same reason: a drug nobody saw coming.
Risk 3: Nobody Can Verify The Price
This third problem is what makes the first two actually dangerous, not just theoretical. It's not only that one specific number from TwentyFirst is hard to check. Nothing in this market can be checked. These policies don't trade on any public exchange the way stocks do. There's no daily price to look up, and nothing independent to compare a portfolio's value against. A portfolio is worth whatever the model says it's worth, that's it, there's nothing else to measure it by. So if GLP-1 drugs are quietly making these portfolios worth less than they look, nothing forces anyone to admit it. Not the market, because there isn't one. Not regulators, who license this industry but don't check its math. And definitely not the firms getting paid to produce the numbers in the first place. The problem stays hidden until a fund runs out of money to pay premiums and is forced to sell.
There's been one exception recently, and it's worth walking through because it proves the point. In June 2025, a short-seller called Morpheus Research published a report claiming that Abacus had been underestimating how long its insureds would live. Abacus pushed back by hiring an outside actuarial firm, Lewis and Ellis, to independently check its entire portfolio, over 700 policies. That review came back within 1% of Abacus's own number. Nobody knows for certain who was right. But notice how this check happened at all: not through a regulator doing routine oversight, and not through some industry watchdog. It happened because a short-seller stood to make money if Abacus turned out to be wrong, and went looking for a problem on that basis alone. That's not the system working. That's one interested party checking another interested party, purely because there was a profit in it. There is still no routine, independent mechanism that checks these numbers as a matter of course, this is simply the one time the incentives happened to line up so that someone looked.
Section 6: A Bet Placed Twice
Every year, Americans walk away from more than nine million life insurance policies worth over $725 billion, most of it for nothing. Selling instead of surrendering is a real option that helps real people, and the industry built to do that at scale is not a scam. The infrastructure works. The demographic argument, a generation of boomers aging into the bracket where this makes sense, is real too, even if the transaction data hasn't caught up to it yet.
But when these companies pitch investors, they make a specific claim: that mortality risk is idiosyncratic. One person lives longer than expected, another dies sooner, and across a large enough portfolio it averages out to something close to the model. That claim is what makes life settlements attractive as an uncorrelated asset in the first place. It's also the entire reason a pension fund can hold this alongside equities and call it diversification.
The claim is only true if mortality moves randomly, person by person. If it moves for everyone at once (because of a drug taken by millions of people the same way) there is nothing left to average. Every policy in the portfolio moves in the same direction at the same time. And the conditions that would let anyone catch that early don't exist here. A handful of firms produce the number, no outside party checks it, no market price can contradict it, and this market has already collapsed once before when a drug outran the model.
None of this proves the number is wrong. GLP-1 drugs may turn out to matter slightly for an impaired 78-year-old, and the cushion already built into these models may absorb it. The problem is that nobody knows. Not regulators, not investors, and not the firms selling the product. Every market prices things it can't be sure about. The problem here is that this one is sold as if the uncertainty didn't exist.