Big Debt Crises, Bigger Debt Crises
I’ve been looking forward to reading Ray Dalio’s Big Debt Crises ever since I slowly gave up on Principles. Dalio runs Bridgewater Associates, a hedge fund notable for two things:
It’s an extraordinarily large fund with a track record that’s either solid (in terms of risk-adjusted returns) or unprecedented (in terms of dollars earned). You might recognize Dalio from his TV appearances (CNBC, January 2018: “If you’re holding cash, you’re going to feel pretty stupid” Business Insider, September 2018: It’s like 1937. To be fair, in January he said we were in the late stages of a blow-off rally, and it’s necessarily true that feeling stupid for holding cash peaks when asset prices do.)
Its culture is… intense. Bridgewater prizes brutal honesty, directness, constant ranking and evaluation, and a fair amount of internal secrecy. Of their nearly two thousand employees, only a handful actually know what strategies they use.
Obviously, this is an institution you can learn from: either you can say that they engage in a bunch of crazy behaviors and still make money, in which case the people in charge have to be staggeringly talented (you’ll rule out luck if you look at how they handled the crisis), or their superficially crazy behaviors are actually reasonable.
I found Principles frustrating, because it described both an ideal workplace for me and a set of behaviors that would get me promptly fired from nearly every actual workplace I’ve encountered. Norms and traditions encode vast amounts of accumulated wisdom; there are good reasons not to tell someone with a stupid idea “Your idea is stupid” before giving them more detailed feedback. I would prefer the direct feedback — it was years later that I realized “We want to put you to your highest and best use” meant “You suck at what you’re currently working on.” But I’m a weirdo. I found Principles wistfully unsatisfying.
Big Debt Crises was teased as part two of Principles, and it’s great. If you care about finance, go out and buy it right now. Or just download it. It’s free on Kindle Unlimited.
The structure of the book is:
A general overview of the credit-based view of the economic cycle.
A detailed look at how crises work. Something I’d never seen before: Dalio looks at the archetypal crisis by creating a blended average of historical crises. This lets you see the general process: economic growth, then asset price inflation, which is tied to growth in leverage, followed by underlying inflation, a peak soon after rates rise, followed by a massive, painful deleveraging. Here the paths diverge: Some countries go through a long and painful deleveraging (Japan in the 90s)), some go through hyperinflation (Germany in the 20s), and some bounce back (the US after the most recent crisis).
The real meat of the book is the summary of three crises: Weimar hyperinflation, the US in the Great Depression, and the Great Financial Crisis. These writeups summarize the thesis behind the boom (I didn’t know the post-WWI Germany was where all the hot money went), what led to the crack-up, and how policymakers responded. Each case study ends with a timeline of contemporary headlines; the 2008 version includes contemporary Bridgewater client memos from the crisis, which make fascinating reading. I learned a ton from this section, and relearned some things I’d forgotten. Did you know that Germany’s central bank was probably picking the best of a set of bad options when they opted for hyperinflation? Remember that time the Fed met the day after the Lehman bankruptcy and didn’t cut rates, because they were afraid of higher oil prices? Fracking may be bad for the environment, but it’s great for monetary flexibility.
The final section is a series of computer-generated summaries of other crises. I got kind of lulled to sleep by this, although if Dalio ever expands each of those examples into something as detailed as the big three case studies, I’d happily read another 2,000 pages or so on debt crises.
The book is more geared towards policy than towards trading. Before I read it, I thought a drop in asset prices was generally preceded by a big increase in asset prices, then an increase in rates and an inversion of the yield curve; I have more datapoints now, but that’s still what I think. On the policy side, Dalio is a big advocate of printing money to stave of painful deleveraging, and allowing companies to not mark their distressed assets to market. And he makes a strong case.
The Dalio Model
The basic template Dalio uses for assessing the macro cycle would be uncontroversial to a Keynesian or Modern Monetary Theorist: in the near term, economic growth depends on credit, i.e. on people making investments or consuming goods in exchange for the promise that they’ll deliver a set amount of cash at a future date. A credit-based economy has a nice benefit: the supply of money tends to rise to meet the demand. And it has a big drawback: economic downturns are self-perpetuating, since a contraction in credit produces a contraction in spending, which leads to lower returns on investments and thus lower valuations for assets, which itself leads to more credit contraction.
The solution most of the developed world has arrived at is the Bagehotian model: when credit contracts, central banks should freely lend to financial institutions, to inject liquidity into the system.
Back in 2008, when I was less informed and more opinionated about economic issues, the bailouts struck me as an accounting fiction: the government disburses dollars to buy assets; the assets go up in price, but there are more dollars, so we’re back where we started: we have more dollars that are worth less, and on net we’re just as poor. But when the Fed expanded the monetary base from $870bn in the summer of 2008 to $1.7 trillion in the summer of 2009, and the CPI didn’t double in response, I had to reconsider.
The Dalio view is roughly what I settled on: economic activity rises and falls with the supply of credit, not the quantity of dollars, and when total credit has been a large multiple of total dollars, a small drop in the amount of credit per dollar of currency can only be offset by an enormous increase in the quantity of currency. The simplest way to explain why the CPI didn’t double is that the cash offset what would otherwise have been a decline. (A fancier argument: since a currency’s value is a function of demand for that currency, whether it’s driven by trade or investment, monetary stimulus raises the currency’s value by raising the expected growth rate, and lowering the volatility of that growth, both of which promote outside investment.)
The argument against mark-to-market is similar: if a company’s mark-to-market value shows that it’s insolvent, it will collapse, liquidate, and depress assets further. If we think the market value is temporary, then we’d expect it to recover.
Knowing what I know now, if I were Fed chair or Treasury Secretary during a 2008-style meltdown, I’d probably do something similar to what the Fed did: guarantee lots of assets, print lots of money, buy a ton of safe assets (to drive yields to zero so banks are forced into more distressed assets), and buy some distressed assets (just so there’s a bid).
But I’d have doubts.
The Model’s Assumptions
In a closed system where “consumption,” “investment,” “cash,” and “credit” are all heterogeneous, the model works just fine. Credit contracts, cash offsets the contraction, consumption continues, so investments don’t lose value; meanwhile, the stability of investments’ cash flows reduces their risk premium, so a successfully-managed recession actually leaves us better off.
But we don’t have a closed system, and these things aren’t heterogeneous. There are different populations in the US with different marginal propensities to spend, and the financial world is global: monetary stimulus in the US has effects worldwide, and in a global recession, other countries’ stimulus programs affect us, too.
I’d like to walk through a couple of the details that make the pure offsetting-credit-contractions model work suboptimally, and explore what world we’d expect to live in if these details made the model break down:
Because of agglomeration effects and economies of scale, simultaneous stimulus spending in different countries can accelerate industries’ migration to new countries: The Chinese stimulus program in 2008 was around 12% of GDP; the American Recovery and Reinvestment Act spent about 5% of the US’s 2009 GDP, spread out over ten years. Over half of China’s spending was on infrastructure and affordable housing, both of which encouraged urbanization. And in China, urbanization means transitioning workers from farms to factories — and that means increasing the amount of manufactured goods China exports, especially to the US. If China was directly stimulating their export economy at the same time that US exporters were economically stressed, it would accelerate the shift of manufacturing from the US to China. This is not bad for the US as a whole; China had a comparative advantage in manufacturing, and although their labor costs have still risen, many supply chains have migrated there, so the cost advantage persists. However, while this is a benefit to Americans as consumers (we get cheaper stuff), it’s a harm to Americans as low-skill workers. If you lose your job in a recession and get hired back in six months, that’s going to put a hole in your savings account; if you lose your job and your job leaves the country, that can wreck your life.
We don’t necessarily redistribute money to people who will spend it. In a levered economy, the temptation is to spend money to plug holes: you want to minimize the number of insolvent institutions, because insolvent institutions liquidate, and liquidation drives prices lower. But if you make a list of people who are likely to spend a cash windfall, “Person who just lost 90% of their wealth in a levered bet and isn’t sure if they’ll ever make it back” is surely at the bottom. Unless consumption is dominated by Jim Beam and therapy, allocating money to those people just doesn’t juice the consumer economy.
It’s forced to make assumptions about what the real trend in economic growth is. The underlying assumption of any counter-cyclical policy is that there’s a cyclical, not secular, trend. A counter-secular policy has to make deeper changes: not getting people back to work, but getting them to change jobs; not keeping companies solvent, but making sure the right companies go bankrupt. You can draw a trendline through US GDP growth and get a remarkably stable number, but that’s a what without a why. If productivity growth has been slowing, and social capital is eroding, then we should expect our growth rate to decline — and we should see bubbles and busts as a general symptom of excess credit availability, not as deviations around a trend. Remember, if you measure at the right points, you can get an exponential function to have a reasonable fit on a sigmoid curve. Or a sine wave.
Absence of mark-to-market makes all levered borrowers suspect, forcing more money to be injected into the system or privileging borrowers who borrow against liquid collateral. Marking assets to cost rather than market doesn’t change reality, it just changes what we’re willing to admit. And if you permit some fudging, you force people to assume that everybody’s doing some fudging. And the people who, post-fudging, are most eager for your money are probably the ones who most egregiously benefited from lax accounting. The net result of letting banks not mark their assets to market is that fewer lenders trust banks, unless they’re guaranteed in some other way. As Big Debt Crises notes, the US government ended up guaranteeing a majority of all debt in the US, which gives you a sense of the scope required.
The people who get money may spend it in ways that don’t produce much wealth. This is the big risk I see, and it’s an effect of the three other issues I raised. Suppose your stimulus policy is most effective at bailing out investors, i.e. the middle to upper-middle class. (It also helps pension recipients, but they seem to treat their pensions as guaranteed until they go bust, so there’s no wealth effect from making a pension solvent.) These people generally don’t need more stuff; if they spend, they’ll spend on non-tradable goods, and positional goods with finite supply. Housing, for example. Education (a positional good insofar as the sheepskin effect argues against school producing much human capital, and Ivy League class sizes over the last fifty years argue against schools expanding supply to meet demand). Even healthcare counts, since it’s the one thing people are always willing to consume more of on the margin. When we look at a chart of what products are responsible for the most inflation in the last twenty years, that’s what we see. Alternatively, they might save, or, if rates are unacceptably low, buy other income-producing assets.
Governments around the world spent heavily to counteract the Great Financial Crisis. They succeeded in stopping it, at least in terms of asset prices. But the net result within the US was a transfer of economic activity away from tradable sectors and towards sectors with lower productivity growth and less certain results. While this might be a worthy price to pay, it’s an important cost to calculate, and I don’t see it in the math of people who lionize the government’s response to the crisis.
And even if you do buy into the stimulus approach, you should be worried: the government’s ability to print money to counteract shrinking credit is ultimately predicated on money being backed by real economic activity. To the extent that you dilute that real economic activity — to the extent that you make a country’s economy more dependent on providing non-tradable services, rather than building products other countries want to import — you reduce the government’s flexibility.
Dalio talks about this a bit when he discusses debt crises: countries that owe debts denominated in other currencies, or denominated in gold, have little to no flexibility during a crisis. They can’t inflate away a debt, so their only choices are either a) to take a massive hit to consumption by continuing to pay debts, leading to a recession, or b) to continuously devalue their currency to keep the economy humming.
But a country that can’t export much and still depends on imports is increasingly in that position. Whether our overseas obligations is denominated in gold bars or the consumers’ expectation of cheap TVs and cars, the effect is the same.
This means that money-printing interventions are fundamentally a martingale bet. In the short term, they reduce volatility by shrinking the impact of recessions, but the cost is raising the odds of a big recession that’s simply too big to contain through such tools — a recession that leads to a step-function decrease in GDP.
Suppose we’re overusing money-printing, and de facto money-printing, as a means of dealing with financial crises. What are our other options?
A few come to mind:
Much stricter regulation on leverage, and more contractionary policies during expansions. This is hard, because the economy is never perfectly aligned. You can have a rip-roaring stock market coupled with high unemployment (as we had in the early stages of the recovery). It also takes a lot of political willpower to “to take away the punch bowl just as the party gets going.”
Use tax credits or Basic Income as an avenue for stimulus. A funny thing about low rates in the US is that, due to housing finance policy decisions that can be charitably described as incompetent and accurately described as insane, US homeowners actually have a massive interest rate derivatives position that allows them to profit from lower rates through refinancing. This means that the benefits of rate cuts are disproportionately distributed specifically to the middle class. Asset purchases benefit asset owners (who are rich people, pension beneficiaries, and foreign sovereign wealth funds), but given the tight bid/ask spread on treasuries, are perhaps the single most efficient way to inject money into the economy. But simply cutting a check to every taxpayer, or giving them a giant refund, is pretty efficient, and has the benefit of also sending money to people with a high marginal propensity to consume.
Taking the pain: we can’t do it now, but we could eventually get to a system where the US government says that the FDIC represents the limit of the government’s willingness to directly bail out the creditors of insolvent institutions. This would contract credit greatly, but if expanding cash has distorting effects, that might be a worthy trade. It would be a strange and painful adjustment, but a simplifying one. If nothing else, it would give us a whole new set of topics to argue about after the next financial crisis. Ideally, the trade we’d be making is that we’d have more 19th century-style recessions — a short spike in unemployment, mass bankruptcies, and a rapid recovery — in exchange for lowering our risk of a Lost Decade.
Finding long-term ways to raise productivity growth: maybe our current economy is overextended, and has been for years; maybe our baseline GDP expectations should be reset lower, and we should treat increasing macroeconomic volatility as a necessary result of increasingly desperate measures to keep the consumption trend steady as it continuously compounds faster than the production trend. But if that’s the case, we’ll eventually run out of options, and we’ll need some kind of Manhattan Project-style effort to raise productivity growth. To the extent that malinvestment crowds out productivity growth, that means low or negative returns for malinvestors. The optimistic view is that we’ve merely screwed up our priorities for a generation or two, regulating our way into superficially low-risk economic stasis. The pessimistic view is that we’ve picked all the economic low-hanging fruit — but in that case, we should expect a future that’s like the present but with rising natural resource costs. And we should expect to feel the pinch worse than other countries; they can achieve economic growth by copying the US, but we can’t grow by doing that, so over time there will be more people expecting a US standard of living, with all the resource use that implies. Depleting resources plus a growing base of consumers would make natural resources just about the only asset class that could be expected to have positive real returns.
It would be convenient to live in a world where economic fluctuations could be dealt with by printing money or changing numbers in a database and then waiting for the status quo to return. But the assumption that “normal” is the norm allows weirdness to accumulate at the margins, and leads to an increasingly brittle system. Take a look at the US, culturally, financially, politically, and ask: is it more likely that people in authority are overestimating how weird the world is, or underestimating it? And if we’re underestimating the magnitude and nature of the future’s weirdness, do we really want our main approach to crises to be a big bet on mean reversion?