Lessons of 2007-2008

Portfolio theory, liquidity and risk management

Originally published in the May 2010 issue

The harsh lessons of 2007-2008 continue to unfold. Traditional ideas about correlation in a portfolio failed many managers. The risk assessment for a series of varied asset class investments proved to be difficult as volatility underlying the entire portfolio exploded and even worse, the correlation of diverse trades merged towards 1. Quantitative models fall apart when macro-risk isn’t considered. What made 2007-2008 so difficult was that those indicators also failed as many proved to be coincidental rather than leading.

The biggest concern going forward remains the confusion surrounding liquidity and volatility. Markets need to understand the new roles and premium put on transparency, volatility and leverage. The simple cause of the 2007-08 crisis was liquidity as access to credit collapsed when banks had to recognise losses due to bad loans and any leveraged position began to unravel towards a debt-deflation dynamic. Forced selling led to deflation and fears of a global depression. But as global policies have worked their magic in righting the ship of assets with excess money from loose fiscal and easy monetary policy – now is the time to ask what will be the next set of indicators to warn investors and portfolio managers away from the next bubble risk.

The traditional search for a better value-at-risk model inevitably leads to measures of volatility both implied and actual and the spread between them. The limits of Modern Portfolio Theory continue to be an issue as risk is not the same as volatility (see www.kamny.com/load/publications/ p03_eng.pdf). As Michael Keppler noted in the journal Die Bank: An accurate measure of risk must factor in the probability of loss and its potential magnitude. The expectation of loss – measured over a long period – meets this requirement. Another indication of investment risk is the maximum drawdown from a previous high – peak to trough. But there are other tools – particularly those that watch liquidity in a portfolio and its value to the larger system.


In search of the macro risk indicator
Quantitative models work best when regulated by global risk indicators. The search for better global macro risk indicators starts by separating out systemic portfolio risk factors from specific ones. For many years, volatility has been the favourite choice as it’s a universal indicator underlying each individual asset. Using the equity market as an example, implied volatility as a leading indicator for systemic risk, the VIX and S&P500 – has a long history of academic study. The VIX and the S&P 500 move in an opposite direction on 76% of all trading days. The median daily move of the VIX has been built into many models at -4 the daily move of the SPX – making the contract a “hedge” for many managers when they see global macro risks rising.
The issues surrounding the VIX as a proxy for global risk remain:

1) Returns are not normally distributed in most portfolios leaving many rushing to new models explaining underlying non-normal returns, further returns across time may prove more chaotic than trending.
2) Optimal risk management misses other factors beyond “profits” such as taxes, strategic needs, social restrictions and environmental concerns.
3) Risk-free rates don’t exist – as interest rates move towards zero other factors become larger impediments – such as inflation expectations, future volatility and credit concerns. The concern of 2010 is that sovereigns default.

In search of new global risk indicators
There are a number of indicators other than volatility used in models for risk.

1) Credit spreads: The difference between AAA-BBB paper has been used across many decades as a signal for the tightness or easiness of credit. Credit and liquidity have a strong correlation and this has become a target for central bankers to prove their operations to support markets are working. The TED spread (US Treasury Bills – Eurodollar rates) is another favourite example for more short term risk indicators – but again this indicator has been compromised by being targeted by governments globally.
2) Bid-Ask Spreads: The bid-ask spread of a portfolio divided by its volatility remains a useful tool for understanding the liquidation risk. If you have a lack of transparency in price, the actual ability to generate return may be overstated or misunderstood. This was clearly the case in some mortgage derivatives.
3) Debt-Service Ratios: Measuring the cost of capital against its expected investment return has been a key tool for understanding the 2007-08 subprime mortgage crisis. High ratings and low yields can be a toxic combination for risk management. As money becomes too easy, the leverage required to generate reasonable return targets becomes unsustainable.
4) Trade Volumes: Market movements that happen in low volume mean less than those that occur in high volume. This technical rule has some fundamental underpinning and models that aggregate total interest in a market have been successful in avoiding some erratic moves.
5) Correlation Matrices: The sum of the parts of risk – each individual asset investment’s volatility should be compared to the experience of the whole. The problems of 2007-08 were ones of rapid correlation shift. Understanding how that drift occurred remains a fruitful place to understand future dangers.
6) Curves: Understanding how illiquid assets shift in value across duration may be a useful tool to risk managers. The front-month contract of a low volume commodity can be compared to the furthest out contract to understand future expectations, more than just a net present value calculation, forward curves capture some information about market outlooks.
7) Positions: The total position size of any asset in a portfolio needs to be relative to the total market as well as the total portfolio. At some point, the total size of risk outstanding may also be a limit – consider the case of a CDS outstanding going beyond the total debt underlying it. Technical indicators frequently use speculative positioning as a trigger against adding to risk.

Cash is king
As markets moved in 2007, the cash component and cash equivalent proportion of a portfolio became more important as credit extensions proved impossible. The process of deleveraging overwhelmed other forward-looking indicators. So the focus on liquidity then has brought an unintended consequence to trading now. The effect of holding more cash to prevent future squeezes on positions and the push to keep positions liquid enough has made the investment horizon shorter. Duration of risk and its liquidity have become factors that inevitably lead to a short-term market where future implied volatility is always higher. If you look at the current actual volatility in equities – or the volatility in almost any market – the implied rates remain far too high suggesting in the medium term further drops in things like the VIX. This also calls into question their accuracy in measuring real risks. Low volatility alone is insufficient to justify portfolio leverage. The spread between actual and implied volatility should also have an eye to the shape of the implied future curve. After crisis events position liquidation leads to significant disruption in trading activity that inevitably drives volatility lower, but actual potential volatility remains closer to longer-term averages. This dynamic puts longer-term risk takers at a disadvantage. So as liquidity pressures mount on markets the push to hold risk back to short-term positioning takes over – making risk curves steeper than they otherwise optimally should be. The confusion between liquidity and risk doesn’t stop there.

Collateral limits liquidity
Central Bank liquidity operations have a limit – as the private sector in times of crisis shifts its demand for liquidity, central banks have to extend credit based on private sector assets. The present environment limits central bank credibility by the assets swaps from private to public sector balance sheets. But some of those assets may not meet the usual criteria for a central bank balance sheet. The moves by the Bank of England, the Federal Reserve and the European Central Bank to widen the eligibility criteria have changed what markets use for liquid instruments. The blending of risks will in the long-term have an unintended consequence. Using credit spreads or futures curves as indicators of risk will no longer work. In effect, the “illiquidity of some bank debts” was transformed into cash by this government action changing the risk equations for risk free rates. As governments intervened to support markets in 2008 and 2009, many mangers may be ignoring the Goodhart law at their peril. As the Bank of England economist in 1975 wrote: “Any observed statistical regularity will tend to collapse once pressure is place on upon it for control purposes.“ Just as monetary targets failed central bankers in the late 1970’s, so collateral targets will fail in the 2010’s. This study remains a work in progress. The recent paper by FED Vice Chairman Kohn and others (see http://federalreserve.gov/pubs/feds/2010/201020/index.html) underscores the effort of central bankers to understand what statistics worked during the crisis and which failed them.

As risk managers look for better tools to control their leverage and positioning in large portfolios, the process of central banks and governments to reform the present system may seem at odds. The uncertainty of rule changes in already fragmented markets adds to volatility. The future margins required for risk taking will remain a drag on the present trading environment until they are settled.

The present push for bank regulatory reform has some important implications for risk management. Not all of them are negative. The biggest improvement for model trading systems comes from a shift towards transparency. The relationship between liquidity and transparent markets comes through via a number of risk indicators including positions, bid-ask spreads, trade volumes and future curves. Better data about assets and how they trade will inevitably lead to better trading models. The search for a perfect global indicator requires looking beyond volatility. The inclusion of econometric models into a risk system seems to add value but which one matters may require a more fuzzy logic approach.

The uncertainty of future regulation may be dwarfed by the uncertainty of understanding the data inherent to any portfolio or market. Call this the Heisenberg uncertainty principle applied to risk management. The more precisely one property is known, the less precisely the other can be known. The future indicators for portfolio risk management will use more than volatility and correlation. The role of liquidity defined by the collateral inherent in the underlying assets and their price dynamics will need to become a core driver for new models of risk.