The crazy thing about debt crises is that they don’t get incentivized away by an efficient market. First, prices are way too high, then they’re way too low—why don’t self-interested people profit enough on avoiding them for us all to avoid them?

Part of why I have never seen a satisfactory answer to this question is because it turns out it’s highly polycausal. I’m going to save the interesting epistemic reasons for the next essay; this one will be on the three biggest economic causes, in my understanding.

The first reason is unalterably cyclical economics. The second reason is that price forces are asymmetric to the upside because shorting assets is so much more costly than investing in them. The third reason is your bog-standard organizational malincentives exacerbated by some oligopoly-creating forces in the debt-underwriting sector.

First, the macroeconomic backdrop of cyclical credit.

Cyclical economics

Ray Dalio says that the “big debt cycle” over 70+ years tends to repeat from cause-and-effect reasons that echo the smaller debt cycles within it. Smaller debt cycles make up a larger one, because each smaller cycle tends to end with more leverage than it began, for unstated reasons.

People tend to view talk of economic or business cycles like astrology, but it actually makes sense: every instance of debt creation is like its own little cycle, where someone defers payment now in favor of payment later.

Credit creation naturally happens at similar times for many people (after a war is over, for example), and debt collection does too (when the economy suffers a setback like increased energy prices and creditors want to make sure you can pay). So the amount of outstanding “cash” in an economy is going to have a wave-shape cycle as it progresses from credit creation to credit repayment. Credit creation just “creates” “new” “cash”, so the economy can function much differently depending on how much credit is outstanding.

Dalio offers a Monopoly metaphor about how in the beginning of the game property is king, but later cash is king. Collective liquidity is just sometimes higher and sometimes lower. 

Aside from the naturally cyclical timing, I think there’s a driving force in this credit cycle that comes from the two feedback loops I mentioned late in the last post:

  1. When interest rates go down, more profit is earned by everyone who has loans, demand for cash decreases because more gets created, and everyone can leverage up and charge less interest, completing the loop
  2. When interest rates go up, more debts become unsustainable, which means everyone has to deleverage and increase their liquidity buffer, increasing demand for cash and causing interest rates to go up in turn

I can’t point to exactly the thing that causes the loop to stop. I think it’s one of those economic cases where in the short term there’s an effect, but actors adapt to it in the medium term. Maybe someone else can help me out here

So this is already a large chunk of the epistemic explanation: it’s not just that we don’t know it’s coming, it’s that we can’t really avoid the upswing-downswing nature of the cycle without trying a pretty strange strategy to prematurely pop the feedback loops! Consider how you would actually do it: once a crunch destroyed some credit, you’d try to create more to make up for it, but a lot of creditors are suddenly in dire straits, or are floundering but haven’t yet gone bankrupt. Sounds pretty risky—you could do some of this, like the interest rate reductions the government already does, but you have to respect the fundamentals. If you make a lot of new credit when it can’t be put to good use, you risk exacerbating the amount of bad debt outstanding.

This puts things in a substantially different light than popular explanations of bubbles caused by “greed” or popping from “fear”. There is something to those too: the subprime mortgages of ‘08 were in fact greedy, bad decisions; the private wealth-hoarding after recessions can in fact delay recovery. But we can have the understanding that the economy will have some cyclical nature from these feedback loops no matter what, and ask the question: how stable could it be aside from this?

Let’s consider the markets and organizations meant to control the system and avoid large deviations from reasonable prices.

Markets aren’t clean, and shorting is trash

In Inadequate Equilibria, Eliezer talks about how you shouldn’t think of whether the market is efficient in a binary sense, but what outcomes the market setup is adequate to provide. For example, markets that don’t offer the ability to short-sell are not actually adequate to provide protection against bubbles, like the mortgage market of ‘08.

But honestly, despite the seeming symmetry with longs, short-selling isn’t a magic wand that fixes the upward bias if allowed. Before I explain this, I want to make a brief conceptual note here: the pervasive “counting-down” framework still seems pretty misleading to many people who assume adequacy until shown inadequacy. Instead of thinking of market mechanisms as machines that accomplish a purpose until they break, I think it’s better to think of them as a motley set of tools with which you are trying to tame a jungle. The market is not like a Carnot engine of efficiency with some weaknesses, it’s a mess of actors doing crazy things that can have some order imposed by some legal market mechanisms. We’ve over-abstracted them from their roots in an obfuscatory manner.

Anyways, short-selling doesn’t just abstractly fix bubbles. They actually have a terrible risk-reward profile compared to long investing, so they’re not actually going to “correct” the price—just move it part of the way there.

Some problems with shorts: they have negative expected value (the opposite of the underlying asset), perhaps about -6% compared to 6% for equities. They have a terrible performance distribution, being capped at 2x and with unlimited downside compared to longs capped at losing your investment and with unlimited upside. The counterparty risk has to be paid for as a few percent p.a. Shorts create ownership rights totaling >100% of outstanding stock, threatening short squeezes like the GameStop debacle. If you add these up, your expected value might be around -12% + -4% + -2% + -1% = -19% per year compared to longs just from these issues alone, which are not exhaustive! This discrepancy is even higher in more bubbly assets you would most want functional shorts for! That’s crazy!

Just as prediction markets should be interpreted as indicating a WIDE spread of possible probabilities rather than a point estimate, the above shows that all market prices should be interpreted as indicating a spread of value too. Just these several mechanistic difficulties of short-selling mean there’s something like a 20% downward spread on actual value one year out. And the further out you go, the more these premiums add up, since holding it 3 years would be a 60% EV discrepancy. Shorts require you to get the timing right in a way that longs don’t. 

I want to reiterate that point. The fact that shorts have a carry cost that adds up over time is actually a core reason why bubbles are hard to pop. You’d think there’d be some way to make a more symmetric asset, but there really isn’t, not one that people are interested in. Sometimes you want a tool to form the jungle in a certain way, and there’s just no clean way to make it.

The corollary core reason: you don’t hear about market trenches because prices are supported against going too low by hard cash streams like dividends and interest payments that pay out immediately; but you do get market bubbles because the price has no immediate counterforce from rising too high, as you have to wait until people lose interest in the asset (potentially years down the line). There isn’t some analogous fundamental income stream for short assets, like a regular cost to owning a stock. That would actually make things interesting. 

All of this is to say that even when markets seem like they have a solution to something, it can be much hackier or weaker than you’d expect. I focused on short-selling because that’s the location of the weaknesses that most directly enable bubbles, but mechanisms like high-frequency trading and options also offer a mix of benefits and drawbacks to a market. They and other market technologies can help bring prices into line, but leave some spread that they can’t incentivize away, which depends on situation. Designing a market’s plumbing so that its prices avoid unreasonable fluctuations is a difficult engineering task that differs from market to market and works in different measures for different assets on a market, and is especially hard on the bubble side. 

Anyways, the market isn’t really expected to be adequate to avoid debt crises, because on top of all the inadequacies from standard shorts, it can be very hard just to obtain a contract to short a debt. You can short some large indices of bonds, but it isn’t that easy and you pay a significant carry cost. If you’re in money management, you can sometimes get credit default swaps on specific loans (like Michael Burry convincing banks to create new vehicles for him in ‘08), but it seems unfeasible to do this for an index. You can buy gold or cryptocurrency to hedge against money dilution from stimulus, but this won’t do so well in the other half of debt crises that are deflationary. There are lots of ways to “sort of” bet on a debt crisis, but it’s very hard to make low-cost competitive bets against the assets that are most likely to fail so as to reduce their number created in the first place. Another possibility is that the bubble before a big debt crisis can just be too widespread to short well. (Compare the “Everything Bubble”.)

Another issue is whether a society’s debts can be efficiently abstracted for the market to correctly price them. Equities and commodities are highly liquid and easy to price, because every share of a company’s common stock is typically the same; but every mortgage is to a separate homeowner, every credit card and auto loan is in a different circumstance, and even every debt offering a company makes has a new idiosyncratic priority in the ordering of payoff in case of bankruptcy. They get batched in statistically-similar groups, but often still contain large discrepancies. Michael Burry found some tranches of mortgages that were nearly all adjustable-rate, even when other tranches of the same rating were not, and created vehicles to short the worst first. This was far more work than any other fund manager would do.

So, shorting is a difficult way to put downward price pressure on the markets. But, if debt is briefly very overpriced in a bubble, might buyers just be disciplined enough to avoid losing money on overpriced assets?

Organizations and their internals

There are lots of competing funds trying to make money on the stock market, paid in great sums for being correct. However, the stock market is not where debt crises usually happen. Debt crises happen in debts, a lot of debt is created by investment banks, and there are far fewer investment banks in the United States than hedge funds. A nation’s economy can rise and fall with a few small groups in a few investment banks managing billions of dollars in subprime loans (or a small group of bankers determining federal monetary policy).

(Also, a lot of debts are created by other organizations like companies selling corporate bonds, smaller banks creating loans and mortgages, and pension funds and low-risk actors buying these objects. The problem with the low-risk market is there isn’t a lot of selection pressure making sure it’s working well. I’m not going to talk further about these other than to say that it seems to me they’re really not in the business of predicting macroeconomics so much as pencil-pushing.)

Back to investment banks as the debt-production engine of the economy.

Some of the failures hopefully need little exposition. Lots of banks appear to have bonuses for amount made, but no anti-bonuses for amount lost. On the scale of years, teams that make more money will expand lots more than those that make less, with imperfect correction with risk. Moral mazes pervert incentives internally. Transaction-based fees incentivize dealmaking at the expense of performance (and the 2-and-20 fee structure of funds that hold the debt doesn’t help).

But actually, these are just pretty standard organizational issues in all industries. Most products you can buy are the result of vastly imperfect processes; incentives for dealmaking are just as common in any sales branch. The only reason it matters so much here is that the whole economy rests on them.

(I suppose there’s also the question of why it’s such an oligopoly. My guess is that it’s a similar reason as consulting, law, and universities end up with such fat tails: when we’re selecting for prestige rather than results, there’s no hard cap on how much we’re willing to pay, so winners can really scale with the amount of unconstrained capital available in the society in a way that isn’t true in material goods, for example.)

And remember, it isn’t like the CEO is necessarily cackling behind her desk as she watches her company get rich while dealmaking under incentive problems she didn’t fix. As Zvi’s moral mazes sequence showed, orgs with too many layers just have fundamental problems with transferring information up and down the chains. Amongst the many layers of managers are many projects trying to justify themselves with a lot of “bullshit”, and I doubt the CEO could tell bullshit-covering-bad from bullshit-covering-good even if they tried hard (which I’m not claiming they do).

Anyways, one answer you already have as to why debt crises happen: look at Goldman Sachs and ask “is this entity, with its employees incentivized as they are and socialized as they are, going to well-protect the nation from a debt crisis caused by creating too many bad loans that are too abstruse to see through?” I think the answer for most people is a resounding no. There are tons of factors that you know from your own organizations, and you can sub in your own answers. But my summary would be that we’ve tried to tame a jungle, and done a necessarily imperfect job with what we could get.

(It isn’t like debt crises are a modern problem caused only by our modern institutions. It was always a jungle.)

New to LessWrong?

New Comment


11 comments, sorted by Click to highlight new comments since:

What's your take on the claim that NGDP targeting basically ends recessions?  

No longer true with Covid, but Australia went an impressively long time without a recession basically by adjusting their FX rate to do NGDP targeting.  China too seems to have adopted the policy "the government will decide on the rate of growth and take whatever means necessary to achieve it".

The best counterargument I'm aware of is that China's exploding debt will cause a problem in the future.  But it seems that this is much more likely to result in Japanification than a traditional debt crisis.

You talk about the "half of debt crises that are deflationary" and at the very least it seems that the central bank could guarantee that 100% of debt crises are inflationary by promising to mint however much money is needed to keep creditors solvent.  Perhaps this would make it easier for agents to "trade the cycle" by always buying hard assets (gold, land, stocks..) when the debt is "too high".  This should push inflation earlier in the cycle, allowing the Central bank to respond by raising interest rates and popping bubbles sooner.

First, your idea at the end is amazing and I hadn't noticed it. It seems totally possible that more monetary stimulus earlier would make it easier to trade the cycle, moving inflation up, raising interest rates earlier. I like it. However, I'm not sure about a promise to keep creditors solvent, that seems like enough moral hazard to increase debt crises; perhaps you just mean giving them loans, which seems fine. Also, I'm not convinced that this would work as great as you'd expect, just because you still have to have people bet a ton of money in the right direction (the epistemic issues I'll talk about in the next post).

Regarding NGDP targeting, I do think it would go a long ways toward eliminating "recessions" by reducing the magnitude and duration of disruptions. But, not knowing much about the details of NGDP-level targeting and where the stimulus canonically comes from, I am concerned that some sources of continued monetary injection could be bad. For example, if all of the cash creation was going into QE, I'd expect to see not enough trickle-down. And I think if the money came in ways that created much additional national debt, it would indeed cause Japanification or worse.

I think national debt mechanics and potential catastrophes require a much longer discussion than this comment, but the short version is that 1) you can only reasonably print much money if you have a reserve currency or you end up with crazy inflation because your foreign debts require more of your currency to pay them off 2) even if you have a reserve currency, there's an important question about whether adding 5% to NGDP causes more or less than 5% of national debt growth that year. If it requires more, your debt:GDP ratio will just keep going up in an unsustainable fashion. And whether you can succeed at keeping this ratio from growing obviously depends on a lot of factors, like your background GDP, whether you make cash by lowering inter-bank interest rates by fiat vs lowering interest rates by buying bonds in QE (which adds debt), whether interest rates are even high enough to lower, etc etc. While I agree with most people that we sometimes have a shortage of cash in the economy to make transactions and this causes huge deleterious effects, I'm also concerned that we have such low interest rates that we're "pushing on a string" and going to be unable to do anything if a debt crisis does hit. I don't know how to raise interest rates to give us more slack without crashing the economy, or how to get out of this mess, but one idea is to increase monetary velocity by improving the SWIFT system or other parts of the settlement layer (crypto might come in handy here, though 99% of the time when that's uttered it won't pan out).

Anyways, I'd certainly want to try NGDP-level targeting and see how practical you can make it, but I think there are a lot of other parameters that you have to also get right. especially regarding national debt. Regarding Australia, I don't understand how FX can target NGDP and everything I can find just says that Australia is considering doing NGDP targeting after 25 years of inflation-targeting; also, their population growth and mining boom both seem like the kind of economic boons that would make things go well under a variety of macro policies they could enact, so I'm a a bit hesitant to take that as a real example without further evidence. Maybe you can explain more?

I am concerned that some sources of continued monetary injection could be bad. For example, if all of the cash creation was going into QE, I'd expect to see not enough trickle-down.

This is absolutely a valid concern.  My preferred version of NGDP targeting would pair a VAT with a UBI.  Any time the central bank wanted to raise NGDP, it would increase the UBI or decrease the VAT and anytime they wanted to lower NGDP they would decrease the UBI or increase the VAT.  I actually do think that you can do NGDP targeting using interest rates/QE alone, but this requires a much stronger "signaling" component from the central bank where they commit not to raise interest rates/pull back on QE until NGDP is back above target.

 

1) you can only reasonably print much money if you have a reserve currency or you end up with crazy inflation because your foreign debts require more of your currency to pay them off 

You can't actually get out of control inflation simply by doing NGDP targeting.  Reserve currency or not, sustained (above target) inflation occurs when the government spends more than it taxes and overrides central bank independence in order to monetize the debt.  

2) even if you have a reserve currency, there's an important question about whether adding 5% to NGDP causes more or less than 5% of national debt growth that year. If it requires more, your debt:GDP ratio will just keep going up in an unsustainable fashion.

The debt:GDP ratio is determined by the savings rate and productivity growth rate of the underlying economy.  Unfortunately in the developed world, we've seen the savings rate grow and productivity growth stagnate due to a combination of wealth inequality and an aging population.  These are both serious concerns, but raising interest rates and driving the economy into recession is definitely not going to help.  Looking at the 2008 recession, the result was to drive down fertility and wage growth for the bottom 20% didn't really happen until we reached near-full employment in 2019.

 

Regarding Australia, I don't understand how FX can target NGDP and everything I can find just says that Australia is considering doing NGDP targeting after 25 years of inflation-targeting

I think I was just misremembering here.  However one way you could target NGDP (if you aren't a reserve currency) is just to raise/lower your exchange rate target when you want to adjust NGDP.  This solves the "pushing on a string" problem with low interest rates, since it is always possible to lower your exchange rate by selling domestic currency and buying the reserve currency.  

Oops, you're obviously right about not getting runaway inflation from NGDP targeting. Not sure if there's still an issue there with printing when not a reserve currency, but maybe not.

Oops, I actually meant public_debt:GDP. I think growth there continues to be a concern, because it's not "determined" in the way you said afaik, and e.g. COVID?

Ah, the FX idea is good! I still don't actually understand some of the core of NGDP targeting—my reasoning runs into a contradiction because it seems like if you lower your FX, your nominal GDP should go up because your produced goods are now worth more of your own dollars, but that would mean that you've actually increased the amount of cash needed to make all the transactions go through the economy, such that you've created a cash bottleneck where there wasn't one, which is the opposite of what is supposed to happen. Perhaps the lowering of your FX requires increasing the currency in circulation to a greater extent than the drop in its value, such that you end up with more currency x value at the end?

I actually meant public_debt:GDP. I think growth there continues to be a concern, because it's not "determined" in the way you said afaik, and e.g. COVID?

 

NGDP targeting theoretically doesn't care one-way or the other about the amount of public debt (assuming an independent central bank).  If the government spends too much money (due to a crisis like Covid), the interest rates it has to pay will rise and it will face a sovereign debt crisis.  Two outcomes are possible:  1) the government declares bankruptcy, can no longer borrow on the public markets, and is forced to raise taxes or cut spending 2) the government mandates that the central bank buy public debt, monetizing the debt and producing above-target inflation.  Case 2) (which is by far the more common one) is no longer NGDP targeting since the central bank ceases to independently follow its NGDP target.

I actually can't think of a real-world example of case 1).  When governments face a sovereign debt crisis and respond with austerity, NGDP generally falls well below trend.  See, for example, Greece.  I'm not sure why this is, but probably because austerity is forced on the country by an outside institution (usually the IMF or in Greece's case the ECB) which cares more about getting its debts repaid than about making sure NGDP stays on-trend.

 

if you lower your FX, your nominal GDP should go up because your produced goods are now worth more of your own dollars

 

If the central bank sells FX and buys domestic currency, this will cause the value of you domestic currency to rise, meaning the price (in domestic dollars) of goods produced in your country will fall.  

Obviously the central bank can only do this if it has FX reserves to sell.  If not, then it has to raise interest rates, which should similarly cause the value of the domestic currency to rise (and nominal GDP to proportionately fall).

You’d think there’d be some way to make a more symmetric asset, but there really isn’t, not one that people are interested in.

Today's Money Stuff talks about "cash-settled swaps": "just a bet in which you pay me $1 for every dollar that Tesla stock goes up, and I pay you $1 for every dollar that Tesla stock goes down."

Do you have thoughts on those? They sound like they'd be more symmetric, but I could well believe I'm missing something. (Maybe them being exactly symmetric makes them not super useful? But I don't see why they'd have to be; I assume you could make it $0.80 for every dollar that Tesla stock goes up, or one party could pay the other for entering into the trade.)

Oh cool! These definitely seem like they're on the Pareto frontier for efficiency of asset pairs returns given collateral risk. However, I think this has isomorphic risk-reward as actually owning the assets, just minus the collateral. (If the real asset makes a dollar, this makes a dollar; if the real asset loses a dollar, this loses a dollar; QED.) So, it doesn't actually fix the fact that the short side has a max return, while the long side doesn't, or the fact that the median movement is upward, etc. The main benefit here is just the potential for very increased leverage, since there is no minimum capital needed.

(The downside is just in social cost, since you're adding 0 EV to the economy but still increasing correlated volatility significantly. Normal assets have positive EV)

This puts things in a substantially different light than popular explanations of bubbles caused by “greed” or popping from “fear”. There is something to those too: the subprime mortgages of ‘08 were in fact greedy, bad decisions; the private wealth-hoarding after recessions can in fact delay recovery. But we can have the understanding that the economy will have some cyclical nature from these feedback loops no matter what, and ask the question: how stable could it be aside from this?

Treating greed as something that grows and shrinks over time and has a causative effect is a really bad model. It's much more useful to think as greed as a renewable resource that takes advantage of opportunities when the incentives are strong and feedback loops are weak. The subprime mortgage crisis had examples of both. The CRA and congressional interference gave banks incentives to issue more loans with weaker underwriting standards. The "greedy" borrowers were taking advantage of a situation where they could more easily get a loan than at other times. Borrowers always have a higher estimate of their ability to repay than the banks do; it is the banks responsibility to draw a line somewhere.

I definitely agree with this. I think the wider populace may be considering greed as almost a circular explanation, where getting money in normal ways is fine but in ways that cause bad things is greedy, and so of course after the fact of a debt crisis their actions will be labeled as "greed."

Holders of prepayable loans don't really benefit much when rates drop, so I'll assume you mean bond-like instruments (or ones that aren't likely to be refinanced out of, or that pay some bonus in that event).

I meant that borrowers throughout the economy can refinance at lower rates, which is better for them, which means it's easier for the economy to build new stuff.