Natural, man-made, absurd

I don’t know about you but back in school I was taught there are two types of disasters: Natural and Man-Made.

Natural were the ones that occured without human error/moral failure etc. An earthquake, a flood, a drought. Man-Made were ones where humans screwed up. War, famine, terrorism. And an overlap too — A river could flood from unusually heavy rainfall or it can flood from a war criminal blowing up a dam or it can flood from a dam overflowing because the conditions were deemed to be too rare to spend money on or because the conditions were predictable but corruption kicked in and so on and so forth. And let’s not get started on global warming.

But I want to propose that some disasters don’t cleanly fit into either category even when you account for humanity’s moral failures. To be clear that isn’t always the same thing as an absence of moral failure — simply that you can’t explain it with moral failure alone.

Absurdity

Here’s a hypothetical scenario: a new type of fireproof wood flooring is invented. It is revolutionary and the initial safety tests are glowing. It starts trickling through regulatory approvals. People are absolutely loving it.

And then Bob’s home blows up. A few dozen houses down the line his cigarette now extinguished. Probably happened when the flooring suddenly blew up so violently it sucked all the oxygen in the room.

Why? Because Bob did a bad thing. He smoked indoors. He figured the wife and kids are away, how often do I get a chance? I’ll be careful not to start a fire. Blowing up was simply not on his radar. But when he decided to put it out on his fireproof floor it turns out cigarette ash causes this new flooring to not just catch fire but do so violently enough to blow the roof off his house.

It’s absurd. It is not normal and if you challenged me I could not find an example of popular flooring that blows up anywhere in the world in history.

Is Bob culpable? Yep. Can we close the book saying this was just Bob being irresponsible? Nope.

And yet I argue that such absurdity happens all the time and sometimes to disastrous ends. I like to call them complex disasters. Scenarios where the outcome was so absurd and unpredictable that nobody could realistically predict their tragic fate, sealed not by incompetence or conspiracy or stupidity or greed but the laws of mathematics and combinatorics themselves.

Chernobyl

“How does a RBMK reactor explode?”

If you’ve seen HBO’s marvelous series Chernobyl (2019) you probably remember this line, whether and who actually said it we can never be sure but the series shows you everybody asking the same question: How the hell can it explode? The senior most engineer at the site considered it so absurd in the series he refused to even entertain the idea despite first-hand testimony.

We can easily call him immoral in hindsight. And in the series he paid a dear price. But a man with his head on the line found it absurd because at the time it was absurd. There was supposed to be no way. Worst case scenario was a meltdown. The fuel got too hot and burned through the floor. How — and why — would it blow up?!

Just like Bob knew he could be starting a fire. But his cigarette going airborne was absurd.

How does a RBMK reactor explode?

Simple, silly. They didn’t realize their graphite tipped rods could cause mayhem. Their control rods went in too slow. They were reckless building a reactor with a positive void coefficient. You can judge them without even understanding what a positive void coefficient is (I certainly didn’t on my first watch). They were reckless with their testing. The series even eludes to the KGB covering up the vulnerability which is obviously cherry on top stupidity. And we can prevent it by not being absurdly greedy and stupid.

Can we really? Because my one gripe with that series is it didn’t do justice to highlighting the fact that the roof went off when they hit the emergency brakes (The other being how it implied the soviets didnt realize/factor in graphite spiking reactivity. The graphite was there on purpose. It’s job was to moderate the very reaction it ended up destabilizing) AZ-5. Drop all the control rods, risk killing the reactor, risk massive economic loss and embarrassment that might end with the scapegoat taking a trip to the gulag. And it was the emergency brake, the failsafe itself that doomed them.

This is of course reductionist. The people building a positive void coefficient reactor did not imagine in their wildest imagination that the threat isn’t just meltdown but conventional explosion. It was absurd. The people running a reactor in a way the manuals didn’t allow thought they always had the AZ-5 emergency brake to stop a meltdown. Explosion was absurd.

The test designers, different from the people who ran the reactor different from the people who designed the reactor different from the people who built the reactor different from the people who hid embarrassing facts about the reactor, might have thought about safety culture not being taken seriously. The whole thing was literally a safety test and if the soviets understood one thing it was psychology. But the worst that can go wrong is a meltdown and if you can’t even assume your operators to hit AZ-5 what can you do?

And we can prevent it by not being absurdly greedy and stupid

So here’s my question to you: As a local actor, without knowing what was going on with all the other people also skirting the rules, how on earth would you have prevented something that was already impossible?

Bob could have avoided smoking indoors and the soviet engineers could have avoided pulling out so many control rods. But the reactor blew up when they did the right thing and the graphite tipped control rods happened to touch the bottom first but the poisoning and rod withdrawal caused the reaction to be bottom-heavy which caused the reactor to suddenly release 100x heat and flash boil the coolant into steam causing a massive, conventional explosion.

Nobody thought of this. Even the KGB, if they suppressed the fact about the graphite and even if they consulted any leading experts before suppressing it, the expert would have said yeah this can cause a meltdown and let’s patch it silently. Even the test engineers weren’t worried about the test specifically restricting water pumping because AZ-5 was always on the table.

Nobody expected that the reactor would be poisoned, that the bottom was gonna be more reactive than the top or the graphite rods would increase reactivity at the bottom that causes a temperature spike that makes the reactor care less about being poisoned, making the rest of the reactor fire up in a chain reaction and that instead of the reactor melting down the water blew up.

Are the operators culpable? Yep. Can we close the book saying this was just them being irresponsible? Nope.

PS: I do want to hedge that fundamentally we are talking about human beings that successfully designed nuclear power plants. There is no way they didn’t understand positive void coefficient risks or the fact that super hot stuff can turn water into a bomb. They had layers of power backups to keep the pumps running. But the leap from theoretical yeah the reactor can get super hot to reactor gets hot, the water circulation is hobbled, hitting the brakes actually causes causes an explosion instead of a meltdown?

How does a financial system explode?

In the Chernobyl example I highlighted how certain people made some perfectly reasonable assumptions:

• RBMK designers assumed the operators wouldn’t be so incompetent as to pull out more rods than the minimum limit. After all if you can’t even trust basic rule following you’re paralyzed.
• Operators assumed they can pull out more rods than the pesky designers because AZ-5 was the universal failsafe. Nothing can go wrong if you kill the reactor itself, right?
• Test designers and safety engineers had painfully modeled bad outcomes and reasoned that the risk from temporarily shutting down a safety mechanism in a controlled way was lower than the risk of not knowing and something going very, very wrong.
• Even if the water were to flash boil into steam the pressure tubes were called pressure tubes because they could handle the pressure. It’s not like that particular material when hit by very high heat suddenly becomes brittle enough to release the same steam it can otherwise contain just fine, right?
• Everybody assumed physics gave them a very long rope. And it did right up to the point it didn’t.

I now make the claim that complex disasters are not limited to what my textbook covered under the usual natural and manmade categories but everywhere, getting more and more prevalent the more and more we do stuff.

Imagine a banker that’s about to light up a cigarette and come up with a moneymaker:

• Americans rarely default on their mortgage. And those mortgages can be sold on the market if I call them Mortgage Based Securities (MBS).
• Pension funds, retirement funds, sovereign wealth funds and what not have high accountability. They know american housing bonds are a safe and profitable way to park money even if a few default. But if too many homeowners default the pension fund manager is screwed and so the manager will not touch it.
• What if I bundled them up? Statistically some of them will default, that’s guaranteed. If the economy turns down more will default. I am smart enough to model just how many defaults it’ll take for this to be profitable, how many to break even and how many to lose money. And the math is in my favor. I’ll call them Collateralized Debt Obligations (CDOs) instead of housing bonds.
• But conservative investors still don’t want my CDOs. They are too afraid of losing money. They want safety. What if I could get my bonds guaranteed by somebody they do trust? What if I could get big insurance companies, the most respected players that live and die by their perception and math to back my CDOs?
• I’ll go to the insurer and say I will pay you this much money. In return you will promise me that if the homeowner defaults you’ll cover the loss. I’ll call that a credit default swap (CDS).

Now imagine being in the insurer’s shoes:

• This guy here wants to pay me money for this bond and in return I cover the loss on homeowners that default.
• I am not comfortable with this.
• But when am I ever comfortable? Look at all these life insurance policies. For all I know a mass-casualty event could happen. Worldwide famine, drought, a pandemic. If too many people die I go out of business.
• You know what, I will give myself a long, long rope with the premiums and trigger thresholds. Something that would only happen if civilization is so doomed I’m going out of business anyway (They say nothing’s safer than US treasuries because if the US government can’t pay you have bigger problems than your treasuries).
• After all this banker doesn’t want a good deal, he wants any deal. His customers just want to know I will step in if things ever go so sideways.

And now a pension fund manager:

• I know this is a good deal.
• How can it not be? Better interest than treasuries and this banker’s math checks out. I only lose money if home loan defaults are absolutely humongous which can only happen if some other big thing has completely wiped out Americans’ ability to pay for their homes that the government couldn’t stop.
• But if I buy it questions will be raised. It’s a new thing. I could get fired.
• Wait. AIG, a 100 year old behemoth that already holds my own life insurance says they have my back? So it’s a win-win?
• It sounds too good to be true but I can’t poke any holes in it? I mean the housing market can get screwed but AIG? If the biggest insurance companies are out of business CDOs are the least of my concern.

And now an investment bank:

• Pension funds & insurers want a huge amount of bonds rolled up into a CDO to be on the safe side.
• Which means only people that are both really smart & really rich can actually buy so many bonds as to be able to repackage them. Even I don’t have that kind of money.
• But.. not everybody hates MBS. Only the geezers do. What if I bought a MBS and asked my friend for some money and gave him the MBS as collateral?
• And then I can use that money to buy another MBS. And then pledge that to get the money back with a tiny haircut. Rinse and repeat and I can buy a lot of MBS and make really attractive packages.
• The worst thing that can happen is that the defaults are higher than expected in which case the stuff I’m holding becomes less valuable and the debtors make margin calls (I want extra money or I’m selling your collateral). But there’s an obvious floor. Maybe the market loses it’s appetite for these bonds. Okay, breathe, I know it’s risky but taking risk is the job description. And what’s the worst case scenario? The homeowner defaults and the insurance company makes me whole?

And now the friend lending money:

• Huh, my friend wants more money again.
• You know what, what if I gave this MBS that he gave me to another friend and he lends me money which I lend to my friend at the investment bank.
• Now to be clear none of these guys were objectively stupid. Every time you pledged an MBS the other party only gave you most of the money and reserved a little, called a haircut. This was actually good for both parties because if the value of the bond fluctuated the other party will be forced to make a margin call which stresses the first party’s cash flow.

And while the party was going on a few smart people that would later be prominently features in film realized the home loan fundamentals didn’t add up and that the math said there’s some CDS deals that have a good chance of triggering, which would be a massive payout because instead of waiting for the interest to be paid over time you get paid the full amount upfront. That’s a LOT of money.

This text shortcuts a few things like synthetic CDOs (CDOs that don’t directly own the underlying bonds) and people buying credit default swaps without owning the bonds the swaps are meant to protect against. Like buying life insurance on your neighbor.

Again, none of that is illegal. And up until that point nobody thought it could blow up the house. Worst case scenario they go bankrupt. And that’s probably not happening because insurance companies have backed the bonds. It’s not like calling in the insurance will blow the roof off, right? That’s absurd.

And to a pension fund sitting on the sidelines watching gamblers make money and going bankrupt speculating on these weird options, it was not their problem. They knew some idiots were gambling with this stuff but for the most part it was a good arrangement. It’s not like the guy selling me this CDO made it by buying more bonds than the money he’s got, right? That’s absurd. Sure he can get loans but he’d have to put up collateral. If he goes bankrupt with his loans putting these together that’s not my problem, it’s between him and whoever was stupid enough to lend him money. I’m paying a premium because he went through the trouble & the risk & talked to the ratings agencies (cough cough) & the insurance companies to put this deal together. I won’t end up like him because I’m not absurdly greedy and stupid.

And we can prevent it by not being absurdly greedy and stupid

So what actually happened? Well Michael Burry was right and housing took a plunge shortly before Burry himself would have gone bankrupt. The worst case scenario models kicked in.

And that’s when the investment banks got margin calls faster than they could sell stuff to pay their lenders.

And then the lenders got paranoid since they had also borrowed money to lend money and their lenders were doing margin calls but the bank wasn’t giving them money fast enough.

And the news was breaking that this is a housing crisis and so nobody wanted to buy those super-safe MBS anymore. So their value started to drop which means the CDS started to get called.

For an insurer this is fine. Not good but you planned for this. You took money from people for exactly this thing. You made excruciatingly painful models and did surveys and everything else to figure out the worst case scenario of how many people could default and how much you’d have to pay out even in the absolute worst case.

Except the bonds were no longer crashing because of defaults, they were now crashing because of panic. And every minute the cash didn’t fly out fast enough was every minute the next guy in the chain didn’t get paid fast enough and every minute the market wanted to pay less and less for MBS. Which means more margin calls which dropped their value more which made the insurance companies pay more CDS which dropped the MBS value more which meant more margin calls which meant their values dropped even more..

And then people started wondering how long the insurance companies can keep this up. And when they realized the answer is not “forever” the floor fell through.

The market started hitting AZ-5. And they were hammering it. Even if you held a perfectly good bond backed by a taxpayer that had a great job and was paying every installment on time you had no idea what the actual value of that bond is which means you wanted out. And the insurance companies took money from you for exactly this thing.

Except to the insurance companies they didn’t take money from people to make them whole when the market panics and starts treating the good and bad housing bonds the same way, they took money to guarantee the bad bonds that did default. But now suddenly they’re on the hook for every single bond (That’s a bit of an exaggeration but not far from it. CDS had thresholds like mark-to-market, credit ratings [Remember that scene in the Big Short where they’re trying to get the credit ratings dropped?] etc)

Sounds a lot like a runaway reaction, doesn’t it? Like the control rods that were meant to slow down a reaction that ended up accelerating it?

The thing that couldn’t blow up that blew up? And that no matter how smart and non-greedy you were your pension was still screwed even though you did the right thing your whole life because of something that you had no connection to.

But then again, who exactly did? The most obvious answer would be the greedy bankers. And yes, they were indeed smoking indoors.

But I don’t think anybody expected the financial industry to suddenly run out of money or the fact that the guy who lent you money was borrowing money to lend money and is hence more desperate than you expected. Or the fact that the middleman lender thought that he’s making money either way — if the guy who borrowed the money doesn’t pay back the lender just asks his friend to sell it and pays a tiny difference. If the bond drops too much the insurer puts up the collateral anyways.

Are the bankers culpable? Yep. Can we close the book saying this was just them being irresponsible? Nope.

Recap

At Chernobyl what happened was a poisoned reactor that blew up when the operators hit the emergency brakes. In 2008 the system blew up when the lenders hit the margin calls.

Graphite tipped rods -> reactivity spike -> heat -> bubbles -> higher reactivity -> more heat -> more bubbles -> the pressure tubes are no longer pressure tubes -> water inside the pressure tubes blows up

And in 2008?

Housing defaults -> Falling bond prices -> Margin calls -> Forced Selling -> Even lower bond prices -> Margin calls -> Forced Selling -> Repo lenders hike the haircut percentage -> More margin calls -> Crashing bond prices -> Repo lenders want even more security -> More margin calls -> Nobody even knows what the price is -> People owe each other more money than all the money in the world.

Now here’s the uncomfortable part. Nicholas Nassim Taleb did fantastic work on the idea of the black swan event. And in turn I have seen other people call 2008 the perfect black swan event. But if you read through this article you’re probably saying that’s not a black swan event, that’s just stupidity. They didn’t realize they’ll have to pay the people they owe if the bonds went down? The insurance companies took money for something they couldn’t honor? Really?

I think that you fundamentally can’t just lump in the observation that black swans exist making you fundamentally rethink what you know about swans is the exact same bucket as a problem anybody with all the pieces visible can see but happens anyway because nobody saw all the pieces.

And to be clear, that’s not because they were incompetent. It’s just that how on earth would you know this is the specific thing to look for out of millions/billions/perhaps trillions of possible combinations that can go wrong? This was just the specific one that came true.

And the uncomfortable implication is that you can’t stop the next one by being smart and good hearted.

I really find recursion beautiful. And I think it has appeared time and again in history from the serpent eating it’s own tail to the negative space in yin & yang and that is because there might be something fundamental to it.

I have tried to identify many complex disasters and a striking element is how many of them were enabled precisely by the mechanisms failed to prevent them. Not that the mechanisms were incomplete but rather what we thought of as solid ground turned into lava and digested us.

Responsibility

I need to acknowledge that this article has been an attempt at threading a needle I can’t see or understand. Responsibility, guilt, culpability all play a role in our world.

My argument is not that complex disasters form a third source of responsibility but rather that they form a third source of causation.

Natural disasters implies something we’re completely helpless against. Man-made implies something we could have prevented. Complex disasters sit in that weird area where we can understand them but usually only in hindsight say that was so obvious I can’t believe they were stupid enough to fall for that. And then those same people fall for the next one wearing new clothes.

Not all disasters are complex though. Blowing up a dam to cause a flood is well recognized enough to be a war crime international watchdogs look out for. And if the deccan traps were to somehow go off today all the morality in the world will not save us. And quite honestly I am not sure where complex disasters sit.

CDS & Control Rods

It’s hard to capture too many details in a single article so I’d recommend looking up keywords themselves. A popular misconception is that CDS required insurers to buy the bonds themselves or to pay out the full amount when the mark-to-market rates fell etc. What they had to do was put up collateral to prove they can pay and then actually pay when the bonds defaulted.

But when the safety mechanism — asking them to put up collateral — didn’t move money fast enough the lid went off. At Chernobyl when you pull out too many rods, throw them all back in at once with higher reactivity at the bottom your graphite tips go in first and your boron follows. But the tiny time window where the graphite is in and the boron is not is enough to threaten Eastern Europe.

Conclusion

Hopefully this is the first of several articles on complex disasters. The overwhelming theme however is that complex disasters come not just from physics but mathematics & combinatorics. Both Natural and Man-Made disasters are disasters where you can reasonably attribute cause. With complex disasters you will find yourself reaching for somebody or something to blame because the alternative implication is grim.

The latest buzzword is AI and I think the same mechanism can be laid out quite beautifully in terms of gradient descent — a concept that’s deceptively simple and underlies most ML training. Imagine somebody probing around a 3D gradient descent map except there’s a cliff you literally can’t predict.

Imagine a number line where you are trying to really push the boundaries. An agent that is slowly getting to the left of the number line, trying very small numbers, when it accidentally crosses zero and ends up at a negative number.

The same problems appear over and over again in human history and I think that many of them are genuinely unknown unknown variants or black swan events. But there’s also stuff that we didn’t see because we modeled for one thing and didn’t/couldn’t account for all the edge cases where the model breaks down.

And then the disaster does happen, the field learns about it, and it happens anyway in a different field. The signature is always the same: Stuff interacting in ways people didn’t expect and the particular dangerous interaction is just one out of thousands/millions/billions of possible interactions that could have gone wrong making it impossible to plan for it in advance.

And for all we know the next disaster comes from the thing we built to prevent the last disaster and we’ll call it a black swan event again.

Do you see something I missed? Or a technical detail I got wrong? Please let me know!

You can also connect with me on X: Follow @AgentAayush