You are currently following Michael Pettis
You are no longer following Michael Pettis
In the past few weeks I have been getting a lot of questions about serial sovereign defaults and how to predict which countries will or won’t suspend debt payments or otherwise get into trouble. The most common question is whether or not there is a threshold of debt (measured, say, against total GDP) above which we need to start worrying.
Perhaps because I started my career in 1987 trading defaulted and restructured bank loans during the LDC Crisis, I have spent the last 30 years as a finance history junkie, obsessively reading everything I can about the history of financial markets, banking and sovereign debt crises, and international capital flows. My book, The Volatility Machine, published in 2002, examines the past 200 years of international financial crises in order to derive a theory of debt crisis using the work of Hyman Minsky and Charles Kindleberger.
No aspect of history seems to repeat itself quite as regularly as financial history. The written history of financial crises dates back at least as far back as the reign of Tiberius, when we have very good accounts of Rome’s 33 AD real estate crisis. No one reading about that particular crisis will find any of it strange or unfamiliar – least of all the 100-million-sesterces interest-free loan the emperor had to provide (without even having read Bagehot) in order to end the panic.
So although I am not smart enough to tell you who will or won’t default (I have my suspicions however), based on my historical reading and experiences, I think there are two statements that I can make with confidence. First, we have only begun the period of sovereign default.
The major global adjustments haven’t yet taken place and until they do, we won’t have seen the full consequences of the global crisis, although already Monday’s New York Times had an article in which some commentators all but declared the European crisis yesterday’s news.
Just two months ago, Europe’s sovereign debt problems seemed grave enough to imperil the global economic recovery. Now, at least some investors are treating it as the crisis that wasn’t.
The article goes on to quote Jean-Claude Trichet sniffing over the “tendency among some investors and market participants to underestimate Europe’s ability to take bold decisions.” Of course I’d be more impressed with Trichet’s comments if pretty much the same thing hadn’t been said before nearly every previous crisis. Before the decade ends, I am pretty convinced, there will be several countries, including European, struggling with the process of debt restructuring, and some of the victims will surprise us.
The second statement I think I can make with some confidence is that there is no threshold debt level that indicates a country is in trouble. Many things matter when evaluating a country’s creditworthiness.
As a rule anything that increases the chance of a sustained mismatch between earnings and debt servicing undermines the creditworthiness of the borrower. But what really matters is not the expected outcome so much as the probability of an extreme outcome. The expected variance, in other words, is more important than the mean expectation, which is another way of saying that a country with less debt and more variance can be a lot riskier than a country with more debt and less variance.
What are the risk factors?
I would argue that there are at least five important factors in determining the likelihood that a country will suspend or renegotiate certain types of debt:
1. Of course debt levels – perhaps measured as total debt to GDP or external debt to exports – matter. As a general rule, the more debt you have, the more difficulty you are going to have servicing it.
But we shouldn’t get too caught up in nominal debt levels. Coupons matter too. So, for example, as part of the Brady restructuring of the 1990s, most loans were exchanged either for “discount bonds”, which included an explicit amount of debt forgiveness via a reduction in principle, or “par bonds” which included no explicit reduction in principle, but the coupon was reduced.
In fact par bonds and discount bonds implied the same real amount of debt forgiveness, but this debt forgiveness did not show up as a lower nominal debt level in the case of the par bonds. It showed up as a lower nominal coupon.
This Brady-bond talk may seem largely academic, but it has a very important modern-day implication. It means that financial repression also matters a lot – even though it gets little attention in discussions about sovereign credit risk. In some countries, most notably Japan and China, interest rates are set artificially low – much lower than they would be by the market. Local central banks can do this because the financial systems in these countries are heavily banked (i.e. most savings and financing occur through the banking system), there are few investment alternatives, and the financial authorities determine deposit and lending rates.
Forcing down interest rates in this way has exactly the same effect as the lowered coupons on the “par bonds” described above. It implies significant (and hidden) debt forgiveness, so when we look at Japanese and Chinese debt-to-GDP ratios we must remember that we should conceptually reduce the nominal debt levels to reflect the fact that the interest coupon is artificially low – perhaps reducing nominal debt by as much as 30-50%.
This is why Japan was able to raise its nominal debt level to what seemed unimaginably high (and why if it is ever forced to raise interest rates to a more reasonable level, it will face real difficulty), and why although I believe China has a debt problem, I do not believe this problem will show up in the form of a banking or sovereign debt crisis (instead it will show up as lower consumption, as I explain in my July 4 post).
2. The structure of the balance sheet matters, and this may be much more important than the actual level of debt. In my book I distinguished between “inverted” debt and hedged debt. With inverted debt, the value of liabilities is positively correlated with the value of assets, so that the debt burden and servicing costs decline in good times (when asset prices and earnings rise) and rise in bad times. With hedged debt, they are negatively correlated.
Foreign currency and short-term borrowings are examples of inverted debt, because the servicing costs decline when confidence and asset prices rise, and rise when confidence and asset prices decline. This makes the good times better, and the bad times worse. Long-term fixed-rate local-currency borrowing is an example of hedged debt. During an inflation or currency crisis, the cost of servicing the debt actually declines in real terms, providing the borrower with some automatic relief, and this relief increases the worse conditions become.
Inverted debt structures leave a country extremely vulnerable to debt crises, while hedged debt helps dissipate external shocks. Highly inverted debt structures are very dangerous because they reinforce negative shocks and can cause events to spiral out of control, but unfortunately they are very popular because in good times, when debt levels typically rise, they magnify positive shocks. I discuss this a little more below when I talk about virtuous and vicious cycles.
3. The economy’s underlying volatility matters. Less volatile economies can safely bear more debt because their earnings are less subject to violent fluctuations, especially if the performance of the economy is correlated with financing ability. This is especially a problem for countries whose economies are highly dependent on commodities. Not only are commodity prices volatile, there is a long history suggesting that global liquidity dries up at the same time that commodity prices collapse.
This is a deadly combination for highly indebted economies with big commodity sectors. Commodity importers, however, benefit because their volatility is negatively correlated to market conditions (unless of course they have stockpiled commodity prices in a misguided decision to “hedge” themselves – effectively reinforcing inversion in their balance sheet).
It is possible to create a measure that adjusts debt levels according to underlying economic volatility. The first academic piece I ever published, in 1993 I think, looked at 1975-80 external-debt-to-export ratios for a number of developing countries and found no predictive ability. In other words if you had used these ratios back then to predict which countries would have defaulted on their external debt in the 1980s and which didn’t, you would have done no better than if you simply tossed a coin.
But when I used an option formula to adjust the ratios to incorporate the volatility of their export earnings, suddenly the predictive ability of the adjusted ratios became extremely good. The more volatile the country’s export earnings, in other words, the more likely it was to default for any given amount of external debt.
4. The structure of the investor base matters. In my opinion contagion is caused not so much by “fear”, as most people assume, but by large amounts of highly leveraged positions (including leverage through forwards, options, and leveraged notes), which force investors into various forms of “delta hedging” – i.e. buy when prices rise, and sell when they drop.
This kind of trading strategy automatically reinforces price movements both up and down and spreads them across asset classes. Highly leveraged markets are highly susceptible to contagion, whereas markets with little imbedded leverage almost never are.
5. The composition of the investor base also matters. A sovereign default is always a political decision, and it is easier to default if the creditors have little domestic political power or influence. Unless foreign investors have old-fashioned gunboats, or a monopoly of new financing, for example, it is generally safer to default on foreigners than on locals. It is also easier to “default” on households via financial repression than it is to default on wealthy and powerful locals.
One corollary, by the way, is that the total value of assets owned by a government does not matter in determining likelihood of sovereign default as much as many might assume. Governments are not subject to corporate or bankruptcy law. In any individual country you will often hear optimists say that in spite of high debt levels the country will not default because the government owns more assets than it has liabilities.
You should ignore this argument. This is muddled thinking on many counts (for example how easily can you sell assets in a liquidity crisis?), but rather than go into detail, let me just point out that throughout history defaulting governments have almost always had significantly more assets than the value of their liabilities (in fact I cannot think of any exception).
There is usually, however, a significant political cost to relinquishing those assets – that is usually why the government owns them in the first place. If that cost is greater than the cost of default, the government will default.
Beware virtuous cycles
What does all this tell us about the probability of a country’s being forced into default or restructuring? Perhaps not much except that tables that rank countries according to their debt ratios are almost useless in measuring the likelihood of default. This would be true even if those rankings were accurate, but not surprisingly countries hide a lot of their real obligations, and the riskier they are the more likely they are to hide them, so the inaccuracy is always biased in the wrong direction.
It also suggests that investors really need to look very carefully into each country’s underlying economic volatility and, most importantly, the country’s debt structure, since the structure of the balance sheet, and the correlation between asset values and liability values, may actually be more important than the outstanding amount of debt. Countries with a lot of short-term debt, external debt, and asset-lending-based banks, especially large amounts of real estate lending, are far more vulnerable than they might at first seem because the debt burden is likely to soar at the worst time possible – just when everything else is going wrong.
Lots of hidden and off-balance sheet debt is also a very bright red flag, because these structures nearly always implode just when economic conditions sour. One of the main points of the IADB’s Living with Debt (2006) is that nominal debt levels just before a crisis often seem reasonable, but suddenly surge because of an unexpected (but easily predictable in retrospect) explosion in contingent liabilities.
In fact some of the recent “star” sovereign performers may very well be the biggest risks, since their great performance may have been caused in part by highly inverted balance sheets. These kinds of debt structures ensure that good times are magnified, but they also ensure that bad times are exacerbated.
Remember this when someone argues that Country X is doing very well and has even locked itself into a virtuous cycle, in which a good event causes other good events that are self-reinforcing. There are few things as risky as highly virtuous cycles, which are almost always caused by inverted balance sheets. Many of my Brazilian friends, for example, wince whenever they hear about virtuous cycles, because they know first hand how virtuous cycles can quickly collapse into vicious cycles.
Until 1997, for example, Brazil’s biggest credit problem was its huge fiscal deficit, more than 100% of which was explained by interest payments on short-term debt. As global conditions improved during the middle of the decade, Brazil was caught up in a powerful virtuous cycle. The improving external position caused local interest rates to decline, which dramatically reduced the projected fiscal deficit, and so boosted confidence, causing interest rates to decline even more.
Inverted structures are toxic
It was wonderful – and happening very quickly – with real interest rates dropping from the 30-40% range to the 20-25% range in a matter of two or three years. But the 1998 crisis set off a devastating reversal of that process.
A global flight to quality caused Brazilian interest rates to rise. Rising rates dramatically pushed up the government deficit (the financial authorities had not bothered to lock in the low rates, believing that the game would go on until domestic interest rates were at an “acceptable” rate), which caused confidence to drop. Declining confidence forced interest rates higher, and so on with the result that interest rates spiraled out of control as each event reinforced the other. Brazil was forced into a currency crisis in January 1999.
It was a similar process for the countries participating in the Asian crisis of 1997. During the early and mid 1990s it seemed obviously clever to borrow in dollars to fund local operations since dollar interest rates were much lower than local currency rates, and moreover the dollar was depreciating in real terms. The more locals borrowed dollars and converted into local currency, the more local asset markets boomed and the lower the real cost of the financing (compared to borrowing in local currency).
It seemed like such an easy way to make money, until it stopped. At some point the risk caused by the massive currency mismatch (a highly inverted structure) became unbearable and the market went into reverse. Suddenly, and just as local asset markets were collapsing because of capital flight, so did the value of the local currency.
With the collapse of local currency values, all the once-cheap dollar debt went toxic, soaring in relative terms until one company after another faced bankruptcy. Of course each company made overall conditions worse by trying to hedge its dollar debt – buying dollars simply pushed local currency even lower, and increased the cost of the dollar debt.
The Asian wreck was magnified by another inverted debt structure: asset-based loans in the banking sector. When the economy is doing well, rising asset prices make existing loans seem less risky and encourage riskier debt structures (i.e. loans whose servicing cannot be covered out of minimum expected cash flows) because creditworthiness seems constantly to rise.
But once the crunch comes, asset values and creditworthiness chase each other in a downward spiral. The fact that this has happened a million times before, most spectacularly in Japan in the 1980s, never seemed to affect anyone’s evaluation of the risks.
The extent of the carnage in Asia shocked everyone, but it shouldn’t have. We were lulled into overconfidence precisely because balance sheets were so inverted, and made good times so much better, but the very fact of the inversion determined the speed and violence of the balance sheet contraction.
So who is at risk?