For a variety of reasons, it is common practice for investment managers to treat market risk separately from credit risk and both of them separately from liquidity risk. Market risk is considered to be carried by securities sensitive to factors such as equity or commodity markets and treasury, Libor and foreign exchange rates. Similarly, credit risk is deemed carried by corporate bonds, mortgages, commercial and leveraged loans. While both of these groups are subject to liquidity risk, it is often added as an afterthought. In fact, most market and credit risk approaches such as value-at-risk and the Merton credit model assume high or even perfect liquidity.

But reality does not often care about models. Tang & Yan1 showed in 2007 that credit spreads decrease with GDP growth rate while increasing with the volatility of the growth rate. The Basel Committee on Banking Supervision has shown that market risk and credit risk interact with each other non-linearly. Default-induced credit losses can be driven by market price changes as well as market factors such as interest rates, forex rates and commodity prices. At the same time, the changes in prices depend on the rating migration or increase in the default perception of the firm. The senior supervisors’ group of the Financial Stability Board found that those banks that were more severely affected in the global financial crisis measured their market and credit risk separately, whereas banks that used an integrated approach to market and credit risk measurement were much less severely affected.

Current practice

Market risk and credit risk are usually measured separately. Transactions that are perceived to be driven by market risk alone ignore credit risk, even in assets that can get triggered with increasing market prices, which increases the possibility of default. Similarly, credit risk is often measured without taking into account the change in the asset value as market rates move. This results in the two risks simply being added together, implying perfect correlation. But that is far from what actually happens.

Take adjustable-rate loans with coupons that change as interest rates change. Since the borrower is exposed to the volatility of the coupon rate, such a loan is perceived to be without market risk. But these loans still contain credit risk, which is then computed with interest rates fixed at current market levels. However, this calculation misses the impact of interaction between market and credit risks. For example, if the probability of default increases, with a rise in interest rates as the borrower’s obligation increases, the actual probability of default will be underestimated. Rather, both market and credit risks should be estimated together to avoid an underestimation of the total risk.

Consider the classic carry trade. Here the loan in the higher interest rate country carries both market and credit risk. If the lender calculates the market and credit risk separately, the market risk is computed by assuming the borrower is not going to default, leaving only the market risks of foreign exchange and interest rates. And when estimating credit risk, interest rates and the forex rates are assumed to be static. However, the probability of default on the loan depends upon the profitability of the carry trade. For example, suppose the currency in which the loan is invested weakens or the interest rate in the foreign country drops; either puts pressure on the profit margin and increases the probability of the default. This separate treatment of market and credit risk can lead to an underestimation of total risk.

Finally, let’s examine matching of long and short positions in over-the-counter derivatives. Suppose a bank buys an OTC derivative from one counterparty and sells the same OTC derivative to another counterparty. Assuming that the counterparty does not default, the bank is market-risk neutral because the gains in one position will offset losses in the other. Further, let’s suppose the OTC derivatives market value does not change because of daily mark-to-market and daily adjustment of the collateral. Then in case of default by a counterparty, its deliverable can be purchased at the same price. Hence, there is no credit risk. However, if the value of the derivative changes and one counterparty defaults at the same time, the change in market value and default together generates a loss for the bank. In August 1998, adverse credit events and market moves occurred simultaneously. Russian counterparties defaulted and at the same time the ruble floated and its value dropped dramatically. The dollars deliverable to the western customers had to be purchased on the market and the rubles banks received in return had lost much value.

In contrast to the situations above, the coupling of market and credit risk can have two different effects: a compounding effect or a diversifying effect. Consider lending in a foreign currency to domestic borrowers. If the domestic economy slows, the probability of default for the domestic borrowers increases and if the domestic currency depreciates, the value of the loan in domestic currency increases as it is denominated in foreign currency. It may seem that those two effects offset one another, but that view ignores the relationship between changes in the exchange rate and the default risk. Depreciation in home currency increases the probability of default of the foreign currency loan by an unhedged domestic borrower. It has been estimated that the increase in probability of default becomes severe with lower-rated obligors. The integrated risk measurement can be anywhere from 1.5 to 7.5 times the risk estimated by simple addition of the market and credit risk when measured separately.

Liquidity is another matter. Investors are generally concerned with two types of liquidity: market liquidity of securities and funding liquidity. Market liquidity conditions can impact asset prices because of buyers’ and sellers’ time requirements to unload the asset. An investor’s permissible time horizon is very often dictated by market liquidity conditions. However, market liquidity also depends upon the size of the asset. Deteriorating market liquidity often forces investors to lengthen the horizon over which they can execute their risk management strategies. As this time horizon lengthens, overall risk exposures generally increase, as does the contribution of credit risk relative to market risk.

In normal conditions over very short horizons – unlike the subprime crises – defaults tend to be largely idiosyncratic. As a result, losses due to unexpected defaults in well-diversified portfolios are negligible over short horizons. Over short horizons, the dominant credit risk in portfolios appears because of mark-to-market price changes, typically measured by VAR, which increases through time roughly following the square root rule – quadruple the time to double the VAR.

Unanticipated shocks to market liquidity conditions can lengthen an investor’s liquidity horizon and alter the blend of market and credit risk in its portfolio. Over longer horizons, defaults are driven by changes in macroeconomic conditions that are not diversifiable. Hence, the risk from unexpected defaults becomes more important. The credit risk also increases linearly with the increase in the horizon period. Depending upon the horizon period, the increase in risk behaves differently from short horizon to long horizon.

The Basel Committee on Banking Supervision established a working group to study the interaction of market and credit risk that simulated the impact of both changes in the liquidity horizon and changes in credit risk on the overall risk of a portfolio. According to this research, “a drying-up of liquidity associated with an increase in the liquidity horizon from two weeks to six months would have the same effect on VAR of a portfolio of A3-rated assets as a downgrade of these assets by two notches from A3 to Baa2 over a two-week liquidity horizon. In an alternative scenario, in which the lengthening of the liquidity horizon and the rating downgrade both occur at the same time, the combined impact on VAR is two to three times stronger than the sum of both effects measured separately, showing that nonlinearities can also have strong effects in the interaction of credit and liquidity risk.”

A partial solution

Let’s consider the treatment of both credit and market risk in a credit portfolio. Knowing that the portfolio carries both types of risk, we can simultaneously simulate correlated market and credit moves, deriving the correlation from historical data of market rates and credit spreads. Alternately, we can use the appropriate historical market driver as well as the corresponding credit driver to evaluate the combined risk of the portfolio.

This approach will take into account the systematic credit risk captured by the credit spreads as well as the market risk. However, it will not take into account the idiosyncratic credit risk. To do so, we must carry out a daily mark-to-market of the asset. Each day, we check if the asset defaults. If it defaults, we change the value of the asset by its defaulted value and carry out the process for the next day, repeating the process for rest of the portfolio until the horizon is reached. The process is cumbersome, but could account for the idiosyncratic risk. However, while adding the risk for different portfolios with different liquidation horizons, we are assuming the correlation between the risks of different portfolios to be one. This assumption about correlation ignores the diversification effect and overstates the risk.