In the previous chapter, we discussed the interest-rate risk that is caused by mismatches between the bank’s assets and liabilities. One function of the ALM unit is to measure and manage this interest-rate risk. The ALM unit is also involved in the management of the funding-liquidity risk that arises from mismatches between the assets and liabilities.
This risk arises because banks generally fund themselves with liabilities that have very short contractual maturity (e.g., demand deposits such as checking accounts). Banks take the money they receive from these liabilities, set aside a small amount in cash, and invest the rest in assets that have long maturity, e.g., commercial loans. In general, customers leave most of their money in the demand deposits for a long time, and the small amount of cash that the bank sets aside is sufficient to meet customers’ requests for withdrawals. However, if withdrawals are unusually high, there is a risk that the bank would not have enough cash to meet the demand. Such a situation could happen if there was a rumor that the bank could have a liquidity problem, which would lead customers to withdraw their funds, thereby creating a liquidity problem and increasing the rumors. This vicious cycle is called a “run on the bank.”
In such a situation, the bank’s choices can be simplified into three: borrow money from other banks, if they are willing and able to supply more cash; sell some of the loans, possibly at deeply discounted prices; or default to the customers, and go out of business. This risk of defaulting or being forced to sell at a loss is called funding-liquidity risk or cash-crisis risk.
Funding-liquidity risk is different from the liquidity risk discussed in the chapters on trading risk. The liquidity risk in trading arises from the possibility of the bank’s losing money by being locked into a position that is losing value. But in ALM, we are concerned that the bank will be unable to raise enough cash to pay its customers, or that it will be forced to sell (“cash in”) assets at an awkward moment, incurring a significant liquidation cost. There is also an important “reputation” element in funding risk because if a bank is seen scrambling for funds, other market participants will start charging the bank high interest rates on any funds it borrows. In this chapter, we discuss in more detail how funding-liquidity risk arises, how it can be measured, and how it can be managed.
Let us begin our analysis of liquidity risks by considering the bank’s uses and sources of funds. The uses of funds are the outflows of payments from the bank to customers or other banks. The sources of funds are inflows from customers and other banks.
To structure our discussion, we classify payments as scheduled, unscheduled, semidiscretionary, and discretionary. Scheduled payments are those which have previously been agreed on by the counterparties. Unscheduled payments arise from customer behavior. Semidiscretionary payments occur as part of the bank’s normal trading operations but can be quickly changed if necessary. The discretionary transactions are those carried out by the bank’s funding unit to balance the net cash flow each day. Using this classification, the typical daily outflows are as follows:
• Scheduled loan disbursements to customers
• Scheduled repayments to customers, such as maturing fixed deposits (FDs)
• Scheduled loan repayments to other banks as well as corporations
• Unscheduled repayments to customers, such as withdrawals from checking-account deposits
• Unscheduled loan disbursements to customers, such as credit cards and lines of credit
• Unscheduled contracted payments to corporations, such as contingent standby lines of credit
• Semidiscretionary payments for the purchase of securities by the bank
• Semidiscretionary payments as cash collateral for borrowing securities or for reverse repos
• Discretionary lending to other banks in the short-term interbank market
The bank typically has the following inflows and sources of cash:
• Scheduled payments being made into the bank by customers, including loan repayments
• Unscheduled payments by customers, such as checking-account deposits, and rollover of expiring fixed deposits into new fixed deposits
• Semidiscretionary payments from the sale of normal trading securities
• Semidiscretionary payments as cash collateral for lending securities or for repos
• Discretionary payments from the sale of securities from a liquidity reserve
• Discretionary borrowing from other banks in the interbank market
• Discretionary calls on backup lines of credit prearranged with other banks
• Cash on hand, stored in the vault or deposited with the government
• In great emergency, calling on the government for expensive short-term funding
This classification allows us to construct a framework to measure funding-liquidity risk. The measurement of funding requirements can be considered for three situations: expected requirements, unusual requirements, and crisis conditions.
The expected requirements are relatively easy to measure conceptually, although they require a large amount of daily data collection. The expected requirements are all the scheduled payments, plus the average levels of all the other payments. For example, if the total balance of the checking-account portfolio is expected to increase steadily over the next few months, this would create a net expected inflow. A more detailed, daily model would probably find that balances on personal checking accounts were expected to decline steadily towards the end of the month, then jump up as wages are paid. The mirror image of this is that corporate accounts would be expected to increase steadily, then drop as they pay wages. In this case, the net flow to the bank would be close to zero.
An important component of the scheduled funding requirements are new asset originations. For example, if the commercial lending unit is planning to give a $500 million loan to a corporation, this will cause an additional expected outflow of $500 million. For such a loan, the ALM and funding units should have advanced warning that they will need to source additional funds on the day that the loan is disbursed to the company.
The expected rarely happens. Instead, the unscheduled demand will usually be above or below the mean. As part of normal business, the bank should be ready to make payments easily on days when the net outflow of funds is 2 standard deviations above the daily mean. Two standard deviations for a Normal distribution corresponds to a 2% tail probability; therefore, such an event can be expected 5 times per year. This is not a crisis, just an unusual day. To meet this demand, the ALM desk will need to find funds from discretionary sources, such as interbank loans or selling liquid securities.
On any day, the net inflow of discretionary funds to be raised (ID) equals the scheduled, unscheduled, and semidiscretionary outflows minus the scheduled, unscheduled, and semidiscretionary inflows:
ID = (OS + OU + OSD) – (IS + IU + ISD)
The scheduled flows are known, but the unscheduled and semidiscretionary flows evolve randomly according to the behavior of customers and the bank’s normal operations. Let us group these random terms into a single variable, R:
R = (OU + OSD) – (IU + I SD)
We can now say that the amount of discretionary funds to be raised (ID) is equal to the scheduled flows plus the random term:
ID = OS – IS + R
The funding requirement on unusual days is then the scheduled requirement, plus the average for R, plus an additional two standard deviations of R:
and σR can be estimated by collecting historical data for OU, OSD, IS, and IU to calculate R for each past day.
The analysis above gives us the amount of funds that should be available to make payments on one unusual day. If we wish to be protected for an unusual period, we can make the bold assumption that cash flows from one day to the next are not correlated, and therefore can use the “square-root-of-T” approximation to scale up the standard deviation from one day to multiple days:
The results can be given as in Figure 14-1, which shows the cumulative scheduled funding and the scheduled funding plus two standard deviations of the uncertain requirements.
This analysis of the normal flows of funds gives a good measure for the amount of discretionary sources of funds that the ALM unit must keep available to continue conducting normal business on those unusual days that happen a few times each year.
It is tempting to use this analysis with a higher multiple of the standard deviation to get to a higher level of confidence. For example, if we had enough discretionary funding sources to cover 3 standard deviations of σR then, based on the shape of the Normal distribution, we might hope that the probability of meeting the payments would rise from 98% to 99.8%. However, this assumes that the underlying mechanisms causing an “unusual day” are the same as in a crisis. Unfortunately, this is not true, so we need to develop a different kind of model to cover crises.
FIGURE 14-1 Typical Results for the Required Payments
By their nature, crises are rare, and there is little direct data to study. One possibility is to look at international crises and get a high-level indication of the behavior of customers and markets in such situations. However, such data is hard to find because banks naturally do not publish details about the occasions in which they almost defaulted.
An alternative approach to measuring the risk in a cash crisis is to use two steps. First, estimate the possible funding requirement by modifying the model that we used above. Then go on to estimate how much value would be lost in such a crisis. This gives us an indication of the economic capital required to be held against funding-liquidity risks.
The times of greatest liquidity risk are those in which there is a general market crisis. In such a crisis, customers lose confidence in the bank, and other banks are not able or willing to lend. In this situation, the bank may disrupt its normal business to minimize semidiscretionary outflows of cash, such as buying new securities. The bank will generally do everything it can to maximize its net cash available. It may also maximize the discretionary and semidiscretionary inflows, such as calling on standby lines of credit.
For a crisis, it would be reasonable to assume that the bank will make all its scheduled payments, and most of the scheduled inflows will be received, but there will be some defaults. We could also assume that customers made no unscheduled payments into the bank, and that the unscheduled payments to customers were a multiple of the usual standard deviation of unscheduled customer demands.
We can apply these assumptions to modify the requirement for funds on an “unusual” day, and thereby get an estimate of the amount of cash that would need to be generated in a crisis.
The next step is to estimate how much value the bank could lose in such a crisis. The loss would arise from selling assets that are usually illiquid but have been discounted to “fire-sale” prices to get cash.
The estimate of the loss can be achieved by making a list of the bank’s assets in the order in which they would be sold to generate cash. Then, for each asset, estimate the amount of “fire-sale” discount that would be required to sell it immediately. Table 14-1 illustrates such a compilation. The columns list the amount of each asset class held by the bank, the discount that could be expected if they were sold in a crisis, and the consequent loss. If our analysis had shown that $13 billion of additional funds would be required in a crisis, then we would expect that we would need to sell all the cash, money-market instruments, and government bonds. This would give the bank the required $13 billion in cash to meet customers’ demands.
However, in making these sales, the money-market instruments would be sold at $20 million below their value and the government bonds at $200 million below, making a total loss of $220 million. This would indicate that $220 million of economic capital should be maintained to absorb this potential loss.
Once management has a picture of its current liquidity position and the associated probabilities of default, they can then begin to modify the position. This can be done in the following ways:
TABLE 14-1 Losses Due to Selling Assets in a Liquidity Crisis
• Borrow long-term funds in the interbank market or issue bonds. Then use the proceeds to buy liquid assets, such as government bonds to be sold or pledged in times of crisis.
• Establish contingent standby lines of credit with other banks, whereby the bank providing the line of credit guarantees to give funds in a time of crisis. Establishing such a line can be expensive, especially if the bank providing the line sees that it is likely to be called when it is also in trouble.
• Limit the amount of funds that are lent for long maturities in the interbank market. The ultimate case would be to use only the overnight market for lending the proceeds from demand deposit accounts (DDAs) such as checking. This would perfectly match the contractual cash flows and eliminate liquidity risk, but it would give a relatively low yield.
• Reduce the liquidity of the bank’s liabilities. For example, the bank could promote fixed deposits instead of savings accounts, and it could encourage customers with short-term FDs to move to longer-term FDs. This can be done by offering higher rates for more illiquid products. It can also reduce the value of the customer’s option for early withdrawal by adding early-withdrawal penalties, and thereby increasing the exercise price.
It is also important that management should plan the optimal liquidation of the balance sheet for times of crisis. The funding desk should have a crisis-response plan prepared in advance so they know all the possible places in the bank that can either reduce their requirements for cash, get cash back if it has been lent or pledged, or increase cash inflow. The plan for generating cash inflow should list the order in which securities will be sold to minimize the amount of discount required.
In this chapter, we discussed how ALM liquidity risk can be measured and managed. Next, we explain how to calculate funds-transfer pricing, which is the framework for establishing prices and moving interest-rate risk between business units.