What is RISK ?

It’s a word we hear a lot these days...

CEO, Investment bank:

• 2006: “We’re able to spread the risk by creating a security made up of uncorrelated assets”
• 2008: “Oh dear, all those uncorrelated assets have just become correlated! There’s no way our risk models could have predicted this”
• 2009: “Whoops, I didn’t understand the risk of these toxic assets but I’m not giving my massive pension back!”

... but what does risk mean ?

A lot of people think they understand risk, but they actually don’t.

Risk is a combination of 3 things, but it’s essentially a measure of how potentially bad a situation is.

Those 3 things are:

1. Preferences – What’s bad for me might not be bad for you, and my pain threshold might be different.
2. Distribution – The spread of the different outcomes
3. Sensitivity - An idea of by how much we’re affected by a small change from status quo

My favourite analogy for explaining risk is this:

a) A man is sitting on the edge of a cliff that has been completely sturdy for as long as he has know, it’s a beautiful day, and there’s no wind.... is that risky ?

- No.

b) The man is now standing, and the wind is blowing a light breeze.... is that risky ?

- Yes, it sounds pretty risky... he cld fall off and die.

c) Suppose he has paragliding equipment on

- Ah, It’s not risky now, looks like he’s going paragliding

d) Now suppose the wind picks up... is it risky now ?

- Y, it’s riskier now, he could be blown off before he’s ready.

e) Now lets say he straps his feet to the ground and takes off his paragliding equipment

- Well, it’s not risky any more no matter how hard the wind blows

f) Now suppose he’s standing away from the cliff, and the wind is blowing hard.

- Well, it cld be very risky if the wind is blowing at such a strength that he’s unable to move further away from the cliff

g) Now, we thought it wld never happen but the whole cliff gives way... and he dies...

– it turns out is was risky to sit on the cliff !

Lets now have a look at how this analogy works in the financial world of investment banking:

a) A CEO is sitting on a massive portfolio of CDO’s and property prices have gone up for as long as he has know.... is that risky ?

– No... Or at least that’s what his risk models will be telling him.

b) Property prices go down for the first time in ages.... is that risky ?

- Yes, it sounds pretty risky... there’s a chance he cld be blown closer to the financial cliff

c) Suppose he has had enough of investment banking and is looking for a way to retire

- Ah, It’s not risky now, looks like he doesn’t really care

d) Now suppose the pace of property price falls picks up... is it risky now ?

- Y, it’s riskier now, he could be blown off before he can get out of his positions.

e) He sells all the positions to the market

- Well, it’s not risky any more no matter how far prices go down

f) He has re-entered the market albeit with smaller positions, but the market is now going down fast.

- Well, it cld be very risky if the liquidity in the market dries up and he’s unable to sell them

g) Now, all assets become completely illiquid, and the market tanks.

– it turn out is was risky to sit on the financial cliff !

- Our investment banks assumed that the market would never go down based on only a few observations !

In this example, the preferences are the easiest to explain. Here, his preferences for paragliding immediately converted the situation from being a risky one to barely being risky at all. Falling off the cliff was what his wanted to do. Relating this to our current banking crisis, preferences relate to money, and you’ll be hard pushed to find someone who actually sets out to lose money (the guy who wants to jump off the financial cliff)... So as a rule, losing money is a something you don’t want to do, and making money is something you aim to do. The edge of the cliff represents bankruptcy for the firm. However, every company has a different point at which it goes bankrupt, so that’s where preferences come in. Decisions are likely to be made based on an extreme aversion to bankruptcy.

i.e. we can handle a little risk, but not so much that it puts our organisation in jeopardy every year. Here’s a screen shot of how that works:

Here, the job of a “risk management” team is to add up all the positions the organisation takes, and come up with an Annual Probability of bankruptcy.

The CEO’s job is then to say whether it’s too big or not.  The more risk an organisation takes, the smaller the life expectancy.

In our example, company 2 is sailing close to the wind, and has a 50:50 chance of existing in 7 years time.

Download sheet to have a look at survival probabilities based on the amount of risk taken each year.

Going back to our analogy... the amount I can be blown around by the wind is the equivalent to the size of a position in the financial world, and the cliff (or the terrain of where we are standing currently) is the distribution. A cliff is a physical thing so we can observe it. The distribution of asset prices is not so we can only guess what it is based on making observations on the past. Even with our physical cliff example, the cliff was sturdy for his entire lifetime, but that didn’t mean that it would never collapse. I have often heard the phrase: “But this is a 10 sigma event, how were our models supposed to predict that ?!!” Well, according to your incorrect model assumption that the world is one big normal distribution, you’re right, but the problem is that you’ve guessed what the entire distribution is based on very few observations.

Suppose house prices go 10% up for 10 years in a row with a variance of 1%... What’s the probability that the next move is 10% down ? Well, if we guess the distribution based on these observations, we’ll come up with an extremely low probability. Well, a lot of models in banks made the assumption that the next 10 years would also go up at 10% per year.

In investment banks, risk management systems are OK at measuring SENSITIVITIES (the amount you can make or lose if the market moves) , but are less good at measuring Risk (which includes the probability of that move). The reason is because they make a guess of the distribution.

There are several other problems with risk management systems at investment banks:

1. A lot of them also incorrectly net risk numbers. For example, you may have a $25,000 per basis point position in a Brazilian Government bond, and a -$25,000 per basis point position in a US Treasury bond. You might be tempted to add these numbers together to come up with a net “risk” calculation of zero, but this is simply wrong. It’s fine if the 2 assets are 100% correlated AND that the correlation will never change, but clearly the US Treasury far more risky than the Brazilian Government bond ; - )

2. The other problem with adding up sensitivities is one of liquidity. If we are attempting to hedge the sensitivity of one position with another, the hedge is only effective if the there is good liquidity in both the assets.

I don’t honestly believe that CEOs with 20 years of experience at these investment banks ignored warnings from their risk management systems. Most likely, they simply didn’t have enough red lights flashing because the models used were simply inadequate.