Are We Headed for a Credit Market Crash?

April 17, 2014

In a series of speeches, Federal Reserve Governor Jeremy Stein emphasized the importance of financial stability concerns in monetary policy-making. But how does one measure whether threats to financial stability are lurking?

Put differently, can we know that there is a credit bubble about to burst?

In his speeches, Stein cites the work of two Harvard Business School professors, Robin Greenwood and Samuel Hanson. Their research argues that a good indicator of credit market overheating is the share of all new corporate debt issues coming from low-grade issuers.

This measure is based on the quantity of credit issued, not just interest rates. Others focus exclusively on credit spreads, or the interest rate differentials between, say, junk and investment grade firms. Greenwood and Hanson argue that quantities of credit issued by low-grade versus high-grade firms add a lot of power when predicting credit market crashes.

So how big is the risk of a credit market crash today? Robin and Sam were nice enough to send us the updated data through 2013. This chart shows the high yield share of corporate debt issues. Or in other words, the fraction of all corporate debt issues by high yield (or junk) firms:




You can see right away that this variable predicts crashes pretty well. The high yield issue share peaks about two years before major meltdowns we’ve seen in credit markets. So when risky firms are issuing a ton of debt, bad things tend to happen.

The high yield share in 2012 and 2013 indicates elevated risk, but not an impending disaster. For example, the 2013 high yield share is still below the peaks seen prior to other credit crashes. This may be driven in part by the fact that investment grade firms are also issuing a ton of debt. So in some sense the denominator is rising so fast that the high yield bond issues cannot keep up with it.

As mentioned above, another measure people use to predict credit market overheating is the interest spread between Baa rated-debt and Aaa rated-debt. If this spread narrows, then credit markets are willing to fund riskier firms at lower interest rates. Here is the chart showing this spread:




Looking at the long history, you can see why interest spreads don’t do as good a job predicting crashes as the high yield share. The interest spread falls before every crash, but the pattern is not nearly as stark as it is with the high yield share of debt issues above.

As this chart shows, the interest spread has fallen substantially in 2012 and 2013, but remains slightly above the very low levels it reached before the 2001 recession and the Great Recession.

Greenwood and Hanson put both of these variables into a simple regression framework to predict excess returns on risky bonds over the next two years. They can use the historical data to estimate the model, and then they can show the predicted return on risky bonds over the next two years. Here is the graph of the predicted excess return on risky bonds:



This is the picture that worries many. The Greenwood-Hanson model is currently predicting negative excess returns on risky bonds. That is the definition of a bubble — people are buying risky bonds that in expectation will deliver less return than riskless bonds. As you can see, their model predicts negative excess returns prior to the Great Recession as well. This is the exact same graph that Governor Stein showed in one of his speeches (see his Figure 1 left panel).

Gene Fama is right that predicting bubbles is near impossible. The methodology of Greenwood and Hanson is based on a pretty small sample, and we cannot be sure that it will work out of sample. Further, their measure predicted low returns in 2010, but returns over the next two years were strong. At the very least, their model should lead to some worries. And it also tells us that the financial stability concerns of Governor Stein and others are based on something more than just instinct.

Bookmark and Share

5 Responses to Are We Headed for a Credit Market Crash?

  1. Matt Busigin on April 17, 2014 at 7:39 ami

    This article could be summarized as: when lending becomes uneconomical, negative returns ensue, and if there is a lot of it, it can lead to a rapid re-pricing.

    I don’t disagree that low quality credits will necessarily have a positive nominal return in the future if bought here.

    What is missing is the justification for the jump to conclusion that uneconomical lending necessarily begets crashes, the mechanics of how that happens, and a way of measuring the probabilities.

    The most important independent variable not discussed is cyclicality. In 2004, having low or even negative expected nominal return on low quality credits is just a problem for the creditor. In 2007, it’s a problem for the system.

    It all hinges on the Marginal Utility of Credit. If it’s positive, even uneconomical lending can lead to higher output, and be positive for monetary velocity, and reduce instability (for all but the creditor). If it approaches zero or negative, the risk shifts from the individual creditors to the system. The Marginal Utility of Credit is very much a function of the Marginal Efficiency of Capital (Keynes, General Theory Ch. 11). If the marginal efficiency of new real capital is high, borrowers will increase leverage for capital formation. This should raise your output assumptions, and lower your loss assumptions (something which I imagine is not considered in your loss adjustments for expected return).

    I don’t dismiss the findings, but they have to be properly contextualised in terms of cyclicality, marginal efficiency of capital, and marginal utility of credit. I have tried to expand upon how to measure the last piece, which I think is possibly the one necessary to complete this model, here:

    Matt (@mbusigin)

    • Marko on April 18, 2014 at 3:56 ami

      If you look at the marginal utility of credit for all private debt ( i.e. households plus businesses ), it fell below 1 in ~1983 , and is now about 0.7.

      Similarly , if you were to look at it from the household’s perspective , the bottom 95% of income-earning households had been experiencing a < 1 marginal utility of debt for quite a while prior to 2007. Their gains in debt were not nearly being matched dollar-for-dollar by income gains.

      Distribution – of incomes AND debts – matters. If you look at pages 3-4 of this recent BLS article , it appears that the middle three income quintiles must be depleting savings or selling assets to support their increased spending during the recovery , since their income increase has been minimal. On the other hand , the top quintile does not appear to be spending much of their income gains , so any savings that we see are coming from the top earners only.

  2. Mark Close on April 17, 2014 at 12:15 pmi

    How do they account for the issuance of debt that was incorrectly rated? We now know that prior to the financial crisis a significant percentage of RMBS were issued with ratings that did not reflect the credit quality of the underlying loans. This was not just that the models were wrong, but we were seeing deficient underwriting, out-right fraud, first payment defaults and incompetent or corrupt rating agencies. Many of these tranches carried very high ratings and were a fundamental fuel for the credit bubble.

  3. dwb on April 18, 2014 at 8:56 ami

    “You can see right away that this variable predicts crashes pretty well.”

    Uhh, no. What I see is that the high yield share peaks when the economy peaks. It peaked in 2004, and 1997. Lest we forget, most recessions were passively tolerated by the Fed to disinflate the economy. And, investment is pro-cyclical. By definition investment, and therefore high-yield borrowing, drops when we have a recession. The higher the unemployment, the deeper the drop in investment. Moreover, “high yield” is not a well defined concept over time. What looks like high yield in 2014 with unemployment still high (meaning, low demand and higher likelihood of project failure) is NOT what looks like high yield at the peak of the economy. The same tranche of debt could be AAA rated with low unemployment and high growth, or CCC rated with high unemployment. That is true whether its municipal debt, corporate debt, or a tranche of prime mortgages. Defaults are *endogenous* to the state of the economy and monetary policy. If the Fed decides to disinflate the economy, unemployment will go higher, investment will go lower, corporate issuance to fund investment will go lower, and defaults will rise. Projects that were AAA rated will default.

    I read some of your stuff, you see bubbles everywhere, you really need to check your confirmation bias at the door. When you hear hoofbeats you should think of horses, not zebras.

  4. tew on April 19, 2014 at 10:10 ami

    Re: “You can see right away that this variable predicts crashes pretty well.” Um, no it doesn’t. It seems the author assumes the reader knows when every credit crash happened and can superimpose this knowledge on the charts.