In 2018 Ray Dalio wrote a book called Principles of Navigating Big Debt Crises. He argues that we are at least near enough to being in a debt crisis that it is important to start thinking about how to most gracefully deal with them. After reading it, I feel similarly. And yet, I didn’t know what a debt crisis was before his book, and many people I talk with don’t understand why bad debts can not only be hard on their owners but on the whole economy. This is an intro to what a debt crisis is and why specifically it is bad, to be followed up with some posts on their causation and how they relate to the current US situation.

Disclaimer: Dalio is hard to interpret on some central points of causation. I am relaying my best guess as to the truth after looking at some other sources, but I didn't know much about macroeconomics until the last year so this is decidedly not Lindy.

Basics of a debt crisis

In short, debt crises are when a country ends up with too much bad debt, and then its economy suffers. For example, the recent government-debt crisis in Greece was a debt crisis. This was part of a larger and slower debt crisis often called the European sovereign debt crisis. The crash in 2008 was a banking crisis, a type of debt crisis where the debt is held by banks. The Great Depression was a debt crisis (and became a banking crisis). Hyperinflation in the German Weimar Republic in the 20s was a debt crisis, because they owed the Allies a lot of money in war reparations.

Ray Dalio documents ~100 debt crises in various countries from the last century in the end of his book, and I think all of them led to recession. Not every economic crisis can be thought of in this manner—sometimes there are currency crises, commodities crises, or pandemic crises—but nearly every big recession I know of was caused by bad debt (the main counterexample was probably the dot-com bubble, which was in equity and not debt and barely caused a GDP drop). To give you a sense of scale, Dalio’s ballpark for big crises was that 20% of debts got written down 40%, and with a 2x debt-to-GDP ratio this meant that an amount equivalent to 16% of GDP was lost by that country in wealth (before follow-on effects).

People often think of debt issues as stemming from public/government debt (both national and state/municipality). But this isn’t necessarily true—about half the crises listed above were about private rather than public debt, where private includes corporations. For example, if several of a nation’s companies can’t pay back their loans, they lose a lot of money in forced liquidations, and that reverberates around the economy. Private debt is often higher than public debt and is easier and more general to model, so I’ll focus on the economics of private debt crises rather than government debt crises in particular. 

I like to think of a debt crisis as consisting of two main phases.

First, there is a proliferation of hidden bad debts (bubble). Second, cash flow bottlenecks domino outward when these debts are seen to be bad (crunch). 

Hidden bad debt bubble

My understanding of the stereotypical arc of events in a debt bubble are roughly as follows:

  1. Unsustainably high real economic growth rates continue for some time due to lucky economic circumstances.
  2. Expected growth decouples from actual growth when rates drop back to normal. Debt and equity prices are forecasts, and thus don’t have tight feedback loops because their feedback is years in the future. People (somewhat reasonably) assume the downturn is a blip on the otherwise-straight trajectory.
  3. Targets: to hit the expectations of those around them, borrowers and lenders chase high growth targets and continue transacting assuming medium-term growth rates that would make these actions good deals. Out of this comes a lot of debts that aren’t fundamentally sound in producing a large margin more growth than interest. Further, companies aren’t using debt as one-time borrowings for which they pay back the principal at maturity: they just hold debt on their balance sheet, refinancing with a new loan whenever it comes due at maturity.
  4. At some point, interest rates rise, and companies can’t keep their cycle of low-rate loans: they have to either pay back the debt or refinance with a loan at higher interest rates than their growth! If they act quickly to pay off the loans they might be ok, but it can be hard if interest rates rise quickly: consider a company with debt equivalent to 50% of yearly revenue, with 5% free cash flow. If it has to pay back a loan, starting from 15% cash buffer would take it 7 years ((.5-.15)/.05=7). Realistically it will have to liquidate other assets to meet such an obligation.
  5. Debt service payments begin to exceed growth after refinancing at higher rates, but companies have no way to escape. Executives and analysts notice things are unsustainable, but don’t really make it public knowledge: executives try to turn things around, hoping they can weather this blip or get lucky with a new product or vision. The longer this continues, the longer unpayable debts will accrue, and the harder it will be to spread out the damage when the crunch comes. For the businesses destined to fail this is a charade phase; but undoubtedly some businesses do manage to make things sustainable during this period.

Now in normal circumstances, plenty of transparently risky loans go out all the time, and lenders price these in. It never reaches a crisis. So this is a stark example of market inefficiency. The question becomes, as with all bubbles: how does the asset become so mispriced? In this case, how is the badness of the debts hidden?

I unfortunately don’t have the cleanest of answers to this. Because it’s so complicated, I’m going to relegate it to my next post, despite it being both at the heart of my interest and also surprisingly relevant to the whole cycle.

We’ll instead move on to how the liquidity crisis would tend to play out assuming some downturn happens after a period of some prosperity. As such, this applies to “small” debt crises as well.

So back to the plot: the longer the decoupling of the bubble continues, the more loss is piling up that isn’t being paid by anyone. Eventually, we see the big reveal and it all comes due at once. The loss itself is bad but a small enough fraction of GDP even in big cases (10-20%) that it wouldn’t be catastrophic if spread out over a decade—but it is made much worse by receiving it all at once, which sparks the cash flow crunch.

Cash flow crunch

This is basically a liquidity crunch, spread out over everyone.

Single expenses of $50k in medical bills can bankrupt many people, even if they could easily pay it over a decade (cf. student loans). They just can’t get the cash fast enough. And while lots of human hardship is sort of neutral in economic standards—e.g. a business getting fairly outcompeted by an alternative is financially net-positive in terms of the total assets of the group—a person going bankrupt isn’t neutral even by economic standards. It’s not just that all their assets get transferred to other entities, it’s that the assets become worth less too. Any synergies they have are gone; good cars get junked because buyers can’t tell them from lemons; houses get put on the market and are unlived-in for a while; items the buyer has adapted to will go to people who don’t know how to use them; etc. They had annealed into comparatively-advantageous assets through learning and selective purchase, and all that advantage is lost in an asset sell-off.

So it goes for companies. Re-allocating the goods, financial assets, and career capital from a company is a long undertaking that destroys much of the value. Specialized career capital is burned, current projects are burned, IT work is burned, etc etc. And credit squeezes don’t have to progress all the way to bankruptcy to hurt. Any partial squeeze that requires liquidation will bring partial costs.

At high leverage or thin margin, large costs must be paid even for small adverse market movements or lending conditions. How small? 

Here are some Fermis I made while surfing through 10-Ks and Statista and knowing very little about business. They don't lead to any strong conclusions and are hard to make clear, so please skip if confusing. 

Average profit margin is about 8%, and companies seem to hold cash reserves ~10% of revenue. So for the average large company, if a crisis lowers margins by 7%, nothing happens. If they’re reduced 9%, they start losing money but have 10 years to recover. If they go down 11%, now they will go bankrupt in 3 years unless they raise new cash. And stochasticity matters a lot, as The Goal will tell you. So companies can’t just ride it out perfectly, knowing that if their margin picks up in 2 years and 364 days they’ll be fine. They have to start deleveraging with productivity-hampering sell-offs significantly before the final bankruptcy date. (If they can—if not, bankruptcy passes their losses on to others: shareholders, lenders, companies that sold them products they haven’t paid for yet.)

So, in this Fermi, the average business has to initiate sell-offs once margin drops about 10%. However, revenue only drops on average about 5% during recessions (because that's how much GDP drops). And if revenue drops only 5%, profit margin probably drops only 1-2% (because of that lost revenue maybe 60-80% was going to cost of goods sold (COGS), and only 20-40% was going to profit).

This would imply the average business only loses 1-2% margin, and most don't even have to initiate sell-offs. I don't think this is exactly true. Perhaps COGS is lower for many businesses, especially in certain classes like restaurants or the service industry; or costs are sticky and so more lost revenue comes directly from profit; or something else in my Fermi is wrong. Regardless, the companies most hurt would seem to be those with low COGS, low initial margins, sellers of discretionary goods and others preferentially hurt by a smaller GDP, and businesses relying on continued financing (new small businesses, startups, growth companies, etc). 

But no matter which factors hurt the most, this is where profitability crunch and cash flow crunch blend together. It starts as a cash flow crunch in company X who can’t raise money; they then defer payments to company Y, who has a cash flow crunch that hurts profitability; then company Y reduces its net outflows, decreasing revenue and causing a profitability crunch in company Z. This cascades around the economy. If everyone has to cut back at the same time, there just can’t be good allocation of all the "excess" capital that was cut, because other people don't have uses for it yet. So the economy’s total output will shrink for a while until free cash is accumulated and new uses can be found for the excess capital.

Fermi: Specifically, you can imagine that a nation starts with GDP of X/yr and wealth of 10X. They like to have 10% of wealth as cash-on-hand, so they have X cash. They lose .2X of their cash at once from bad debts, taking wealth to 9.8X and cash to .8X. They need .18X of cash to be re-stockpiled, which will take 4 years if their businesses have 5% free cash flow = .05X/yr. In the meantime, everyone cuts 5% of their outflow so they can stay solvent or accumulate cash quicker, and GDP drops to .95X/yr. 

[Caveat: I'm not sure if these numbers are reasonable, as it's hard to figure out how corporate wealth is counted and whether the sum of free cash flows can be directly estimated from GDP, etc.]

One way to think about this liquidity shortage is in terms of monetary velocity and nationwide cash availability, which may be familiar from crypto utility tokens or from The Dark Lord’s Answer or NGDP-level targeting within MMT. Everyone finds out at once that a small segment of debts is bad and they won’t receive cash back from them, and they need to quickly raise cash some other way. Some people sell assets to raise cash, but demand for money is high—there’s not actually enough cash available in M1/M2/M3 that people are willing to part with to be exchanged in the ripple of transactions that needs to occur. Velocity is not high enough, so cash gets bid up in a deflationary manner, which makes demand even higher. (Note that this means that you are providing a valuable service by re-investing in things with your cash when cash has dried up toward the bottom of a crash.) (Interestingly, this means that the faster economies get in sending payments to each other via internet or blockchain or what-have-you, the less deadweight loss you’d theoretically have from liquidity crises.)

Another way to think about it is again from the point of view of personal finance. Consider someone who is always in a bit of debt, late on utilities or credit card payments. Things can continue on like that for a long time and the person can be fine, as long as they make a little more than what they spend. But if they suddenly get foreclosed, or have their utilities shut off, or some other demand for money is called upon, things really spiral quickly—they can’t go to their job and get money, they quickly have their other loans foreclosed, and they have to declare bankruptcy or something. Then they can’t earn and contribute to the economy, even though they were fairly productive beforehand. This is what happens to many companies at the edge of profitability when interest rates go up and they suddenly have to pay rates they can’t afford over the long-term, or pay off the entirety of loans they can’t afford at this moment. A sudden small shock that they can’t spread out over time can destroy their entire productivity.

The way the economy surfs the wave right at the edge of profitability helps explain the mystery of why crunches can happen so quickly. If there was more marginal profit to be gotten from lending, lending would occur, which pushes rates down a little bit, which actually increases marginal profit to be gotten (if rates were at 1.5% but drop to 1%, now anyone who could make 1.2% should get a new loan), so more lending occurs and pushes rates down in a feedback cycle (and all the formerly-profitable loans become even more profitable when they rollover at lower rates). But when just a tiny bit too much has happened, then not only are those last marginally-profitable loans proved to be no longer profitable—rates go up, and suddenly the batch of loans before that becomes unprofitable, which forces rates further up, and more loans become unprofitable in the reverse, nastier feedback loop. (Technically, the loans that were made are still profitable, but if they can’t directly repay then the rollover will be unprofitable.) Suddenly there are a bunch of people in tight spots trying to pay off loans, with not enough cash flow to go around and save everyone. 

If you’re still wondering the age-old question of whether this can all just be averted by collectively deciding not to be stingy with our cash, this is a little true—I’m sure there’s some hoarding psychology that exacerbates the crash at the bottom. But the general pattern is unavoidable, because the cash has to come from somewhere and there just isn’t enough of it. You can nationalize corporations to keep them productive and pay back the taxpayers with their future profit, or you can make government emergency loans to big players that are profitable for taxpayers like the 2008 bank bailout. But the bad loans have to be paid for by someone; if the lender can’t shoulder the loss, they can’t pay their lenders; if the lenders can’t shoulder that loss, they can’t pay their lenders in turn; so something has to drop in value equivalent to the value of the loss. The only question is where it comes from.


As I type these things I’m hit by some deja vu from reading Ray Dalio, where I felt like he kept dodging a key question about *exactly what was happening* to make the debt crisis occur. I felt this way reading most explanations of bubbles over the last few years. Why is there a positive feedback loop rather than a negative feedback loop? Why don’t entities adequately prepare for this? Why isn’t every “reason” for a crash countered by an equal and opposite reason for self-preservation? Why does it seem like there are so many accounting terms thrown when there should be a simple core with fewer terms?

One thing to point out is that there are actually just a lot of things going on in an economy-wide liquidity crunch. A business going bankrupt has to sell many different kinds of capital: some capital is easily accounted for as wealth (accounts-receivable, or machines), but other capital is very different (human capital, or organizational capital). We try to use simple categories like "wealth", but the disbursement of different wealths shows up in very different ways on the balance sheet and leads to very different outcomes. The edge case of a debt crisis is enough to make lots of concepts start to fray across the very complex web of transactions and accumulations in different entities across the economy. 

A second point is that feedback loops are very difficult to quantify. Because of the nature of a nationwide domino-ing liquidity crunch where reduced wealth causes reduced demand causes reduced wealth, this falls into that category. Plus, the vast differences in how it affects businesses with low COGS or safe financing or what-have-you, compared to other businesses, makes it somewhat hard to aggregate into a single fixed-point equation. However, much like fractional reserve banking leads to a feedback loop that still results in a finite and estimable amount of new cash, there should be a succinct series of equations relating reduced wealth to reduced aggregate demand to reduced wealth. Or perhaps the present elasticity curves take that into account since they're already empirical. Anyways, I don't entirely grasp how all this fits together, but I imagine there's some simple explanation that will be laid out eventually.

If anyone has a pithier conceptual explanation for the reasons behind the bubble or the crash, I would love to hear them; for now, I hope that this at least helped characterize debt crises, and perhaps my future posts will convey more of the causation.

Thanks to John Steidley for many good discussions slowly clarifying this topic.

New Comment
5 comments, sorted by Click to highlight new comments since: Today at 9:21 AM

An alternate perspective that I sometimes use is to view the crash as a short squeeze.

Going into debt is equivalent to selling short a currency. It's usually safer than shorting a stock, because currencies are designed so that their value is stable, rather than to usually go up.

Just like someone who owes GameStop shares needs to have the ability to buy back those shares, a player in debt needs the ability to regain dollars.

There are important limits to how many dollars one can safely owe, and it depends a fair amount on how other traders of dollars behave.

If one big player, or many little players, misjudge their risks, they need to acquire more dollars. Sometimes everything works out fine because lenders are optimistic enough to lend more. Sometimes big players expect things to work out poorly, and that causes them to compete for drive up the price of dollars (as they would drive up the price of GameStop) in order to cut their risks. Guesses about what others will do can play a key role here, which is part of why the result is hard to predict.

One somewhat unique feature of modern currencies is that central banks are at least nominally able to supply unlimited quantities of them, and there are widespread expert opinions that they out to do roughly that when the value of the currency is being bid up (i.e. when there's deflation). There's some sign of a trend towards central banks doing better at that. But it's still not obvious what's preventing them from doing better. (This is mostly relevant to standard recessions, which are caused by an unexpected decline in the inflation rate; some debts can't be handled by stabilizing the currency, so I'm a bit hesitant to accept your broad framing of the term debt crises).

This is a great frame, thanks.

In the ending parenthetical, it sounds like you're saying that I'm overapplying the term "debt crisis" compared to "standard recession" because what-you-would-call-a-standard-recession is caused by a quick decline in inflation, whereas what-you-would-call-a-debt-crisis is merely sparked by or correlated with a quick decline in inflation and is sometimes inflationary. Is that a correct paraphrase?

I will think about that claim more, but for now your frame seems very compatible with mine to me. My current reconciliation would say that your frame is more elegant for describing how the liquidity crunch works (just like a short squeeze on cash once people start bidding it up as they trade in debt, with an emphasis on the primacy of other actors' reactions), but that it doesn't really capture how you get to that vulnerable point, which is usually not by huge inflation moves but via the debt/credit cycle and some bad debts that take a bunch of cash out of the economy at once.

That paraphrase is mostly good. I'm trying to separate monetary phenomena, which are the main problem in recessions, from reckless debt levels, which are the main contributor to government debt crises.

Yes, my explanation is mostly compatible with yours.

I didn't try to explain how a system becomes vulnerable. I think that happens via recency bias causing misjudgments, plus competitive pressures that Romeo mentions.

My impression is that competition pushes fragility in good times. If two firms are basically the same but one takes out bigger loans and levers up their investments more they will have more cash to play with to try to take some market share from their competitors.

Without a shred of support I'll throw out there that perhaps we're dealing with a scale-free network [1]. There would be no "typical size" for economic disruptions; and it would invalidate all the 20/20 hindsight "why did it happen" stories (it makes reading the news a lot more relaxing, I find). I don't know how to evaluate this hypothesis, though.