I analyze macroeconomic issues from a fundamental perspective, and I analyze market behavior from a technical perspective. Original macroeconomic analysis can be found here and both macro analysis and commentary can be found on my Caps blog. If you like or appreciate my analysis, please add yourself to my Following List

Sunday, February 21, 2010

Of Modeling, Risk, Financial Innovation, and Liquidity Crises

I was reading the Feb 13th issue of The Economist and they have a section on risk and risk modeling for financial institutions. The one that I thought was particularly relevant was the article about liquidity risk - "When the river runs dry: The perils of a sudden evaporation of liquidity". The whole section is a very well written set of pieces, as most Economist articles are, but was not as hard-hitting as I think the topic deserves (I think a lot of punches were being pulled and a lot of the conclusions for future systemic risk were not being fully drawn).

"But binv is just a permabear, and he's always coming up with reasons to be bearish, and we just had a massive stock market rally. So why should we listen to him?"

..... You shouldn't.

I am not going to sit here and make some argument about why I am the finest macroeconomic mind around, nor am I going to try to convince and dazzle you with my charting prowess.

I am an analyst. I make observations. I examine the macroeconomic landscape and I draw conclusions.

I lay my observations out for you to follow.

So you read them and you agree with them, or you read them and you disagree, or you ignore them altogether. It is immaterial to me. Because the point of this post is not to convince you of anything. The point of this post is to share information and observations. I will draw and share my conclusions, and I offer them to you if you are interested in reading them. But your conclusions are up to you.

But more to the point, I am not bearish for the sake of being a bear. I would much rather be long. I would much rather be bullish. I am a very optimistic guy, and I want to invest in companies and the US economy for the long term. But when I honestly assess the problems that our economy faces in the near term (next couple of years) and the actions that have been taken to deal with those problems, I am utterly unconvinced that they are being solved or in some cases even taken seriously.

This is ultimately why I believe that the current environment suggests that investors should still be very defensive. That is my honest $0.02. I have no agenda other than to call things like I see them.

Yet I know something about modeling, creating mathematical representations to describe phenomena. But more importantly I know quite a lot regarding the limitations and abuses of models.


I have stated before that I am a thermal and structural analyst for the Aerospace industry. I create mathematical models to describe physical systems on a daily basis. And it is very important that these models are accurate, as there is often very expensive hardware performance dependent on the predictions from these models.

Also I have made the argument that financial systems are quasi-physical systems (one example from here: First and Second Derivatives of the SPX - http://caps.fool.com/Blogs/ViewPost.aspx?bpid=298547), that is they display properties such as inertia, capacitance, flow, diffusion, and respond to inputs such as impulses, step functions, ramps, etc. I am not making the argument that you can model financial systems and physical systems in a strict one-to-one manner, rather I am saying that there is often a corollary between the two systems and an understanding of physical systems will often given you insight into the behavior of financial systems.

Physical systems are inherently non-linear.

Even something as simple as a beam in bending, which is described by kinematic, equilibrium and constitutive relationships, has non-linear terms in them. For engineers to be able to work with these relationships in a practical manner we make simplifying assumptions. In the case of the beam equation above, we assume a small angle approximation (sin theta = theta) and this allows us to linearize many of the relationships. And this is a very good approximation for small deflections and angles. Whenever you look up a beam diagram to determine the deflection or bending moments from a particular load case, baked into those relationships are these linearizing assumptions. The point is that these assumptions allow you to build and successfully analyze very complex systems, but you must always be aware of their limitations. Deflections cannot be extrapolated infinitely because at some point the linearizing assumptions are no longer valid, and non-linear effects become very important.

The key concept here is that there is a range of conditions over which these linearized systems can be properly analyzed, and cannot be extrapolated out indefinitely with out serious error (i.e. the model gives you false predictions).

I keep talking about linearization. Why is this so important? Linearization means that effects can be added to each other to give a valid composite result. This is the principle behind superposition. An example would be a fast transient signal on top of a slower transient signal: such as the temperature response of a filter wheel on an Optical Bench that is responding to diurnal temperature swings, or the flutter response of a control surface relative to a wing which is responding to the air stream. Superposition is the key for models to be able to accurately predict the responses of several combined inputs.

Then there are some effects that are non-linear and have no linearizable simplifcation: Radiation and Convection from a high temperature exhaust plume as the vehicle moves from atmosphere to vacuum, the effective clamping constraint (degree of freedom) of card locks on a PCB as the chassis goes from low frequency to high frequency vibration, etc. When your system is subjected to these non-linear inputs then the must be tested and verified against these inputs.

This brings together the next key concept: When a system is subject to inputs, especially non-linear inputs, or an input will cause the system to behave in a non-linear fashion, the model must be correlated to these conditions. Using a model to extrapolate beyond correlated conditions will lead to inaccurate predictions

Together, these concepts of model correlation and the range of applicability for a model are the basis for all aerospace analysis and testing procedures. Think about it, do you design a new thruster that has a new plume shape and temperature profile and stick it on a satellite and just launch it? What if it doesn't work the way your model predicts (non-linear behavior). What if it breaks hardware? What if it ends the mission? It is impossible to retrieve a satellite and fix it when it is in orbit 22,000 miles above you. This is why models and hardware are verified before they are deployed, so that you can be sure they work and you can be sure that the model will predict all of the situations the system will encounter on orbit (sun angles, surface degradation, shadowing, thruster pulse timing, etc).

What this does it allows you to retire risk. The more you know about your system, and the better your models predict the system behavior through correlation, the more confident you are that the system will behave for conditions that are similar to the ones that you tested for. AND RETIRING RISK WILL GIVE YOU CONFIDENCE THAT YOUR MODEL WILL ACCURATELY PREDICT SYSTEM BEHAVIOR AT THE EXTREMES (i.e. IN A CRISIS)!

Risk and Financial Innovation

Good economics, just like good engineering, should be BORING!.

Models should be clean and as simple as possible, so that they can be vetted by multiple analysts that should agree on the interpretation of the results. And when new designs are introduced in (which happens all the time in engineering), the design must be tested thoroughly and the models describing the behavior of the new design very well understood before the design goes into production.

The term "financial innovation" should scare the SH** OUT OF YOU!!. For 2 reasons:

a-1) There was never any mechanism by which these instruments (CDOs, CDSs, MBSs, etc.) were truly exercised before mass proliferation within the marketplace. The first time we saw how these derivatives reacted during a crisis was when they collectively represented more than the GDP of the entire world.

It would be the same as designing a new satellite, putting it in a thermal vacuum chamber and testing it for nominal conditions only. Then doing some handwaving and saying "yeah it works fine at hot and cold conditions too". Are the radiators properly sized for the hot case? Is there any heat leak to sensitive components during hot conditions? Are the heaters properly sized for the cold case? Are there any conditions where the propellant freezes creating a catastrophic scenario? These are the questions that you want to have answers to on the ground before you launch, where you can take action if the system does not perform as expected. That last thing you want to have happen is to find out there is a major system design flaw on orbit.

Same with these financial instruments. The last thing you want to have happen is to find out that these responded to crisis conditions not at all like you expected while in the middle of a crisis. Yet that is *exactly* what happened.

a-2) It didn't really matter anyways because this instruments were fraudulent to begin with: Financial Carcinoma -- Denninger: Did You Need a PhD For That? - http://caps.fool.com/Blogs/ViewPost.aspx?bpid=322718

So why did these models perform so poorly? Why was risk so grossly under-represented?

Because the financial systems these risk models are trying to describe are also extremely non-linear

Models such as Value at Risk / VAR are fine for short term forecasts where the liquidity environment is known. But leverage greatly alters the liquidity environment and a very small change in leverage can completely change how liquid any asset is. Does that change its underlying value? YOU BET IT DOES!. An MBS is worth the fully securitized cost only if somebody is willing to pay you that amount for it. And if liquidity is only available to purchase that asset in a highly leveraged environment, then its value is a function of liquidity and leverage. Regarding Economic Debates and Opinions: The Fallacy of "Purely Objective" Analysis - http://caps.fool.com/Blogs/ViewPost.aspx?bpid=305849.

Additionally these models looked at housing risk simply from a historic default rate standpoint, without considering the possibility of a massive fall in the underlying assets (housing prices) and how that would further impact the default rate. (Effects combining in a very non-linear fashion).

There are quite a lot of financial environments that are non-linear and sometimes even binary, such as a highly leveraged liquidity environment.

The current and next generation of modeller's will benefit themselves and all of us by adhering to the Modeler's Hippocratic Oath: http://www.financialmodelingguide.com/financial-modeling-tips/tips/financial-modelers-manifesto/:

* I will remember that I didn’t make the world, and it doesn’t satisfy my equations.
* Though I will use models boldly to estimate value, I will not be overly impressed by mathematics.
* I will never sacrifice reality for elegance without explaining why I have done so.
* Nor will I give the people who use my model false comfort about its accuracy.
* Instead, I will make explicit its assumptions and oversights.
* I understand that my work may have enormous effects on society and the economy, many of them beyond my comprehension.

This is the reason why I *always* put caveats on my analysis and projections. Because any analysis, whether physical, financial, TA, EWP, etc., is an approximation and a projection, nothing more.

Liquidity, Liquidity Crisis, and why another Crisis is Extremely Likely

So we have discussed modeling and risk. And we have already had a major financial crisis and a stock market crash. Since all these effects are known, isn't the risk retired? If we are aware of some of the underlying problem, haven't we fixed them?

NO, not by a long shot.

There is a big difference between captital risk (and so called Value At Risk / VAR models) and liquidity risk. And as you will see in some of my conclusions, the latter is the one that gets significantly less attention and is significantly more dangerous. And why I ultimately believe that all of the actions of the Treasury and the Fed have almost guaranteed another liquidity crisis in the future, and has done nothing to ameliorate the underlying problems.

First, lets start with the Great Deleveraging Event of 2008. From above, I talked about how all these derivatives reacted to the collapse of liquidity and how leverage across the board began to dry up.

But why did conditions abate? Did the market just "have enough" when it came to deleveraging and the equity market find a compelling valuation bottom back in March 2009? NO

b-1) If you want to know why liquidity returned to the market, then read this fantastic post by Kristjan Velbri: Dollar Liquidity Swaps & The Financial Crisis. The reason why the Dollar rallied and then peaked while the market bottomed in March is because of massive Dollar Swaps pumped into the market by the Fed. That's all. There is no fundamental strength in the US Dollar. There was a crisis reaction into dollars by the Deleveraging Crisis in 2008 (in order to buy Treasuries / the standard "safe haven" move) and then the Fed flooded the market with Dollar Swap agreements to force liquidity in 2008/2009.

b-2) There was no meaningful valuation bottom found, as I have stated many times. Is the Market Fairly Valued? Did the Market Achieve Any Meaningful Bottom Back in March? - http://caps.fool.com/Blogs/ViewPost.aspx?bpid=320237 and The Long View - http://caps.fool.com/Blogs/ViewPost.aspx?bpid=314202

Okay? So maybe this is how it went down. But why am I stating that there will most likely be another liquidity crisis?

Because nothing was fixed!! All the old problems were simply wallpapered over and the same instruments with the same models are back en vogue!!

Here is a list of reasons why the underlying causes of the Liquidity and Deleveraging Crisis of 2008 are still around and worse than ever!

c-1) These assets were never allowed to be properly valued by the market. Congress extorted the FASB to allow for "mark to imagination" and "extend and pretend" valuation, so that financial institutions could keep these toxic assets on their balance sheets unimpaired. Former Regulator Talks Fraud and the Big Bank Getaway - http://caps.fool.com/Blogs/ViewPost.aspx?bpid=344209

c-2) Despite the collapse of several derivatives markets in 2008, the disproportionate size of these "assets" to any meaningful metric (such as GDP of the world), and the systemic risk of the size of these assets .... they are still growing. Financial Crash Risk Increasing Exponentially as Derivatives Monster Continues to Grow - http://www.marketoracle.co.uk/Article15437.html

c-3) The Fed is backstopping the CDS market to allow for, and in fact encourage, the more rapid proliferation of derivatives. PSW: The Fed / CDS Development - http://caps.fool.com/Blogs/ViewPost.aspx?bpid=340700

c-4) The recent rule changes regarding Money Market Redemptions has an even larger moral hazard then first observed. Connect the Dots Before Financial Depressurization - http://caps.fool.com/Blogs/ViewPost.aspx?bpid=335085 and my response - http://marketthoughtsandanalysis.blogspot.com/2010/02/flag-day.html#comment-33841524

There is One Last Issue That Puts this Whole Discussion Into Perspective

I and many others talk about a lot of terms that are important to the conversation, and many that I discussed above, but there is one term that is important beyond all others:

maturity transformation

What this means is to borrow short in order to lend long. It is the basis behind behind our entire financial system. And financial failures are not simply a random and rare by-product, but is actually a "feature" of this system.

I have linked to these two articles before, but I have never featured a post around them. But they are extremely relevant:

Maturity transformation considered harmful: an unauthorized biography of the bank crisis -

The Misesian explanation of the bank crisis - http://unqualified-reservations.blogspot.com/2008/10/misesian-explanation-of-bank-crisis.html

These post were written by Mencius Moldbug of http://unqualified-reservations.blogspot.com. His blog is extremely interesting (although it is a bit out there sometimes / fairly often). He is a Computer Scientist who likes to look at issues from an almost diagnostic point of view. I first started reading him in September 2008 (when he wrote these posts) with his discussion of Fractional Reserve Banking and Mises Banking.

The point of these linking to these posts is to not advocate for an Austrian Banking System (not because I don't think it is a good idea, but because it won't happen), but rather it is to show you from a very logical and easy to follow reasoning why liquidity crises happen to begin with.

Once you read these posts, and then go back up and re-read points c-1 through c-4 again, you will see that the Federal Reserve's / Treasury's / Congress's absolutely irresponsible actions of not allowing financials to fail, by backstopping derivatives to further encourage the taking of risk (not only for hedge fund but for Money Market Managers as well!!!) are setting conditions up for an Epic Fail even worse than the first Deleveraging Crisis.

It is not a foregone conclusion, but I think a reasonable person looking at all of these fact can say that it is a possible, if not likely, outcome.
blog comments powered by Disqus