Friday, July 6, 2018

Upside Down CAPM: Part 6 - Leverage before a crisis

I recently saw these graphs, from the IMF:


Source
Here is text from their Global Financial Stability Report (Chapter 2),  "The prolonged period of loose financial conditions in recent years has raised concerns that financial intermediaries and investors in search of yield may have extended too much credit to risky borrowers, potentially jeopardizing financial stability down the road. These concerns are related to recent evidence for selected countries that periods of low interest rates and easy financial conditions may lead to a decline in lending standards and increased risk taking."

It seems to me that there is a strong and common presumption that complacency or risk-taking lead to borrowing, and that this presumption really does all the work here.


Source
In the text above, there are several red flags.  Has there been a prolonged period of loose financial conditions?  I don't think so.  There is that phrase, "in search of yield".  People who want yield buy equity.  High yield bonds may be somewhat like equity on the gradient from low yield/low risk securities to high yield/high risk securities.  This horrible phrase is central to my "Upside Down CAPM" framework.  Low yields for fixed income securities aren't a signal of risk-taking, and investors complacent about risk wouldn't push yields down, even in high yield bonds.  If expected returns for equities are around 7% plus inflation, then investors funding bonds that pay 5%, nominally, aren't searching for yield or risk, because there is a better source for both of those things in equity markets.

I'm not sure we can even conclude how attitudes about risk might influence the relative level of high yield corporate debt.  And, household debt, which is mostly associated with mortgages, should grow inversely with risk-taking, as it would signal a bias toward real estate with stable cash flows over corporate investments with highly cyclical cash flows.  It seems plausible that the relative amount of risky debt securities could either rise or fall with attitudes about risk.

This seems like a simple counterfactual to consider.  What if the level of risky debt outstanding wasn't positively correlated with systemically dangerous attitudes about risk.  What would these charts look like?  Wouldn't they look just like this?  When a contraction hit, wouldn't we see disequilibrium in the market for high risk debt and a surge in low yield safe securities?  And, then, as cash flows stabilized and markets healed, wouldn't we see markets for riskier securities recovering?  In other words, the drop in risky lending and the initial recovery reflect the market dislocations that come from uncertain and volatile cash flows, not from attitudes about risk.  This is movement into and out of dislocation, not a shift in a steady equilibrium.

So, the top graph is a spurious correlation.  Of course risky debt outstanding is highest right before economic contractions.  There is no reasonable counterfactual where this wouldn't be the case.  And, the second graph shows that (in the years before a financial crisis), risky debt levels first rise as markets re-attain normalcy, and then the level of risky loans actually levels out in the years before the crisis.

There are any number of models that could explain this pattern.  The conclusion comes from the presumptions, not the evidence.

This is where turning the model upside down seems useful.  In the IMF report, low yields are associated with easy financial conditions.  I think this leads to confusion.  Low yields are associated with demand for certainty in cash flows.  I realize it takes some hubris to stand up against an entire body of research, so that having this discussion is sort of unwise of me.  I'd love to see evidence that there is more to it than this.  But, from where I stand, it looks like low interest rates are a sign of difficult financial conditions, and that fact is so counterintuitive that we have just gone barreling along with economic models that are backwards.

Part of the problem is the unfortunate habit of equating low long term interest rates with loose monetary policy.  Using the financial sector as a measure for financial conditions might, itself, be part of the problem.  When the financial sector increases in size, this is generally treated as risk-seeking behavior.  But, the financial sector is generally in the business of being an intermediary for fixed income securities.  Investors didn't need the financial sector to invest in Pets.com.

Note, that Pets.com wasn't put out of business through bankruptcy, because the ownership was completely in the form of equity.  They simply liquidated and paid any remaining cash to the shareholders.  Notice that the collapse of the internet bubble is not associated with a financial crisis.  That is because there was little debt involved.  The IMF has it correct in this regard.  But the reason it wasn't associated with debt and the reason it didn't lead to a financial crisis is because it was associated with risk-taking!  And risk-takers had equity positions.

So, this is where the standard models go off the rails, because that error flips the story on its head, and it leads the IMF and most other observers to a position where they see signs of risk-aversion and their solutions are to limit lending and cut back on monetary expansion.  Then, the collapse in cash flows that ensues gets blamed on risk takers.  The correlations between debt - even high yield debt - and subsequent financial crises are correct.  The interpretations are suspect.  What leads to a crisis is when a large portion of the set of savers demands certain cash flows and then subsequent cash flows become so volatile that those demands are not met.

Zero is also part of the problem.  Zero looms large in the way we construct these models, and it shouldn't.  Take zero out of the equation.  Now, think of an economy where savers can expect yields of minus 2%.  Or, let's take that to an extreme.  Savers can expect yields of minus 50%.  These economies exist.  Would you associate these economies with "easy financial conditions"?

13 comments:

  1. I disagree with your thinking on expected returns. Let's consider the corporate bond market. The expected return can be approximately observed by the spread to treasuries less defaults. That spread varies greatly over cycles. Sometimes expected returns are quite high relative to no-risk assets, other times they are quite low. This must be true for equities as well or else there would be a "free lunch" (if expected returns on equities were 7% real but corporate bonds only 2% real you could easily buy equities and short a basket of corporate bonds with a beta=1 to capture a risk-free spread).

    The challenge is that anytime expected returns are low relative to safe assets it takes very little fear to send investors rushing out of risky assets and into safe assets. In other words, a crisis. In fact, i would argue that low expected returns relative to safe assets are a key risk factor for crises. The expected returns on equities is not always 7% real...it probably varies between something like 0% - 15%. Yes, it's harder to observe than in corporate bonds but it must be true nonetheless.

    And this problem occurs even before we consider the irrationality of "7% real in an environment of 1% labor force growth, 1-2% productivity, and 2% inflation."(corporate profits would eventually overtake the entire economy).

    ReplyDelete
    Replies
    1. As a follow-up consider the chart i've linked to below.

      You may be familiar with the "hussman model" of equity returns which has something like a 0.90 R-squared with 10-year forward equity returns. It turns out that the residuals to the model closely follow consumer confidence. In other words, when consumers are optimistic, they push equity returns very high relative to what a non-biased model would suggest (and thus push expected returns down). Valuations are pro-cyclical and this is a big part of financial crises...

      https://twitter.com/NoBServations/status/944273318756175872

      Delete
    2. The issue with equities is that in the short run, they are highly vulnerable to changing cash flows and growth expectations. Shorting bonds is basically borrowing. That is a winning position on average, but eventually with catastrophic outcomes, because equities are vulnerable to real shocks.

      The reason their prices fluctuate so much is because their expected returns remain relatively stable while cash flows and growth expectations fluctuate.

      Regarding your last paragraph, keep in mind that most of those returns come from dividends and buybacks, which have generally run about 5% annually for many decades. That's not an assertion. It's just historical fact. r>g and all that. They don't overtake the economy because some of that income is consumed.

      Delete
    3. So you're suggesting the expected returns on risky corporate bonds vary over the cycle but but not the expected returns for equities? What about for hybrids like convertible bonds? I find this hard to believe...all assets are linked.

      Delete
    4. You are right. Some bonds have equity-like qualities, so their returns involve various spreads that are affected by different factors which may cause them to act more like equities. The basic relationship I am talking about is the basic CAPM equation where ERP is added to the risk free rate to get total expected returns to equity. I am saying ERP is negatively correlated with the risk free rate. There are a lot of securities that exist between risk free bonds and a diversified basket of equities. And the returns on all of those securities depend on an array of complications.

      Delete
  2. Very thoughtful blogging.

    Anyway, when consumers bought housing that was, until 2008, a very low risk proposition.

    One reason the market accepted securitized pools of mortgages was that housing prices had been steady and rising for the previous 50 years.

    Moreover most middle-class families probably should diversify assets with some in the financial sector and some in real property, the latter being, of course, a house.

    In many other nations house prices did not collapse. The Fed made house buying and financing a risky activity, but only retroactively.

    It is a monument to the imperviousness of independent public agencies that there was no housecleaning at the Fed after 2008.

    ReplyDelete
    Replies
    1. It is galling to see policymakers talking about how executives needed to take the fall because of their recklessness while they were sucking the lifeblood out of the economy. But, paradoxically, they feel as certain of their positions as I feel that they are mistaken, so I suppose I shouldn't feel that way.

      Delete
  3. Here is a great observation: "So, the top graph is a spurious correlation. Of course risky debt outstanding is highest right before economic contractions. There is no reasonable counterfactual where this wouldn't be the case."
    I see the same thing whenever commentators talk about equity inflows and outflows. They note that investors always buy and sell "at the wrong time" as if the prices of stocks are moving independently of supply and demand. By definition, the majority of investors can't zig when the market zags.

    ReplyDelete
    Replies
    1. Yes. Complaints about buybacks are another example. People complain that corporate managers have bad timing and tend to overbuy at market tops. But that would be the case if markets were purely efficient but were sensitive to unpredictable real shocks. People love the behavioral fear/greed explanation so much they accept it wherever it fits.

      Delete
    2. Great example. People complain about that without remembering that the other likely alternative was investing in the business. And those investments, at that time, probably would have had similar (disappointing) results. The third alternative, cash, is just market timing and likely to be worse if undertaken on a regular basis.

      Delete
  4. The Fed tightened because inflation was FAR over target, not because of the housing market. 10y inflation expectations did not dip below 2% until september 2008, at which point i agree the Fed should have acted a little quicker (although nothing would have changed).

    Any counterfactual needs to consider the risk that failing to tighten as they did would have led to a true inflation crisis. I dont see how you can rule this out.

    ReplyDelete
    Replies
    1. Core CPI inflation when the Fed initially began to lower rates in September 2007 was 2.1%. Breaking that out between shelter inflation (which is mostly imputed rents of homeowners) and non-shelter components, Shelter inflation was 3.4% and non-shelter inflation was 1.1%. Inflation in noisier non-core categories was high. That is true. But, it is also true that core inflation was well below the 2% target by then, and core inflation only seemed like it was near their target because they had already made the housing shortage worse by raising rates too high.

      Sometimes I judge the Fed in hindsight. In this case, I wouldn't judge the Fed so much, in terms of their discretionary decisions. I would judge the proxies they are using, because they were getting false signals from them. It is a good example of how NGDP targeting would be more stabilizing, because NGDP targeting would have led them to accommodate earlier here, since falling residential investment was the cause of the rent inflation.

      Delete