You have already been introduced to the primary concepts associated with credit and market risk. Credit risk concerns the potential default on an obligation, whether you are the creditor or debtor. Market risk arises from the normal course of doing business and the ups and downs of the overall market or industry.

Let’s take a closer look at the special challenges posed by each risk and risk type.

To estimate your credit risk, employ the following three measures:

To determine your expected loss due to financial risk, multiply your figures for probability of default, exposure at default, and loss given default, as follows:

To determine your unexpected loss, calculate the standard deviation of your expected loss.

To determine your overall financial risk, add up each of the individual unexpected losses as you learned in Chapter 10. Each group or type of financial risk is accompanied by correlations. Correlations exist between individuals, regions/geographies, industries, and more. In doing so, you will see that there are many opportunities for diversification.

How do you measure this type of diversification and exposure? When there are many exposures, companies often use sophisticated models to manage these properties and analyze how they combine to generate the big beta distribution. They often employ a Monte Carlo analysis (see Chapter 10). A number of companies supply these sorts of models or variations of them.

Recall that market risk can come from both traded financial instruments and structural financial risk. Structural financial risk, or nontraded market risk, comes from a business’s cash flow requirements or liquidity risk. You don’t need to be a bank or insurance company to hold such risk. Market risk can result from a mismatch in the way cash flow is structured, interest rate risk, and the way money moves in and out of your business.

Market risk distribution resembles a normal distribution more than credit risk does, particularly in the case of traded market risk (from trading financial instruments on an exchange). There is a roughly equal likelihood of achieving more or less of the target value than you expect. This also means that, unlike with credit risk, the tendency is not to have an expected average loss. Instead, the main concern associated with market risk is unexpected loss, or its market risk equivalent, value-at-risk (VAR).

There are some interesting twists to consider when calculating market risk; they involve what are called distribution tails.Sometimes, a distribution has “fat tails” or little speed bumps on its ends. The fat tails result in bigger estimates of risk than we would naturally expect as part of a normal distribution. Sometimes they can be surprisingly large. These are the sorts of issues that cause big problems for banks (and other companies that have large trading portfolios). When you hear about a bank losing a boatload of money in a trading problem, you can be certain the distribution tail was overly fat. (These kinds of tail problems are normally associated with derivatives.)

Risk managers use different types of VAR models to assess market risk, depending on the types of instruments they are examining and the types of distributions common to them. Let’s take a closer look at three models and how they can work for you.

Parametric VAR. This model assumes that the distribution is normal. Generally, the means and standard deviations of previous values will behave the same way in the future as they have in the past. Consequently, they can be applied directly to the instrument or set of instruments that you analyze. When analyzing a portfolio, historical correlations are generally used as well.

Historical simulation VAR. With historical simulation VAR, a historical rate series, such as interest rates or daily foreign currency exchange rates from the past year, is run through the financial instruments in your portfolio to see how newer instruments in different proportions in the portfolio would have behaved under past conditions. This involves saving historical rate series data, particularly that from periods of stress. Believe it or not, some market risk analysts collect rate series like some people collect baseball cards. You can imagine how valuable it would be for businesses to have access to rate series from certain periods of market behavior—the Mexican peso crisis, the Russian debt collapse, the U.S. subprime disaster in 2008, and so on.

Monte Carlo VAR. The Monte Carlo VAR is effectively a random number generator that creates a set of market rates. From that springs countless simulations. Investment banks sometimes use supercomputers to simulate all of the possibilities.

How do you decide which type of VAR model to use? It comes down to what sorts of instruments you have and whether you must adhere to any regulatory guidelines.

If you are going to trade derivatives of any kind, you need historical simulation and/ or Monte Carlo. These models can pick up the fat distribution tails and speed bumps mentioned earlier. Parametric VAR assumes that your distribution is perfectly normal and doesn’t simulate real behavior in any specific way; therefore, it doesn’t work very well for derivatives.

Most regulators want to see at least historical simulation models; some also encourage Monte Carlo simulations.

Each model has its downside. Historical simulation is rooted in past events, so unless you make up new rate series to run through these models (which can be done), you could miss out on new effects that might happen in the future. Monte Carlo can generate new effects, but it can be taxing on computer systems and sometimes makes incorrect assumptions about the distribution of the rates that it is generating.

In the end, many companies use all three types of models to test and compare assumptions and approaches.