Financial institutions rely on asset and liability management (ALM) to mitigate the risks associated with mismatched assets and liabilities. However, traditional ALM frameworks are narrowly focused, and typically do not consider the dynamics of a firm’s full balance sheet. For firms seeking to optimize capital allocation, a more dynamic, multi-scenario-driven approach to ALM is therefore needed.
Before discussing the challenges one must address and the steps one should take on the road to dynamic ALM, it’s helpful to provide a bit of background about this approach. The levels of assets and liabilities at each financial institution are driven by many factors, including supply/demand, cost of funding, and consumer behavior.ALMreduces the risks connected with a mismatch of assets and liabilities, which can arise due to changing circumstances.
Traditional ALM assumes deterministic levels of assets and liabilities, focusing on interest rate, currency, and liquidity risks. However, in addition to this, ALM should ideally be applied to the entire balance sheet of an organization.
In reality, most ALM frameworks today don’t simultaneously adjust for inflows and outflows of deposits or demand for loans - or at least fail to do so consistently to account for fluctuations of interest rates, credit spreads, equity markets and other critical drivers.
By imposing a mandatory haircut of assets and liabilities to stress test banks’ liquidity, liquidity regulations - like IRRBB, LCR and NSFR - are addressing this problem to some degree. But these regulatory exercises are incapable of estimating the full distribution of potential ALM outcomes, which must incorporate tail correlations that are specific to each institution.
In a crisis, some banks have seen their deposits going up (flight to safety), while others have experienced a bank run. Ultimately, the mix of assets and liabilities amid a downturn depends on the nature of the stress scenario, as well as a bank’s specific balance sheet profile.
Setting Up Dynamic ALM
Recently, ALM has been moving in the right direction, via firms employing a variety of market scenarios and balance sheet alternatives.Dynamic ALMapproaches, for example, allow users to evaluate not only the potential impact of a strategy (before it is implemented) but also its risk/return trade-offs.
Lamentably, such dynamic ALM is not very helpful when, say, a firm is potentially prevented from implementing proper management strategies because of the balance-sheet-altering behavior of its depositors and obligors. For example, an operational risk event or cyberattack can produce a reputational risk impact and cause an outflow of deposits. Similar to unexpected increases in loan defaults, this can affect ALM considerations more substantially than changes in interest rates.
One therefore has to use multiple scenarios to model asset and liability segments dynamically. Subsequently, to project respective cashflows consistently, one needs to consider scenario-dependent interest rates, respective levels of interest-bearing liabilities and interest-generating assets. Only then is it possible to maximize profitability, incorporating necessary regulatory, market and business constraints
While consumer products are driven by behavioral patterns and reactions to changing market environments by various market segments, commercial and industrial (C&I) loans might require additional adjustment. For example, as a reaction to the current market environment, a number of loans can be implicitly upgraded or downgraded. (The actual review of a loan’s status typically happens either once a year or once a quarter.)
Even with publicly-rated companies, any change in rating usually trails behind the actual trading levels of a company’s debt. But dynamic ALM must realize when the implicit rating of the loan should be changed - based on the unfolding future scenario and how it would impact the loan level.
In the case of an upgrade, there is a high possibility the obligor will prepay or refinance a loan with a lower spread. In case of a downgrade, obligors are known to grab a portion of unfunded commitments.
In both cases, loan modifications are driven by behavioral patterns specific to the bank - and will substantially impact ALM outcomes both from volume and paid-interest perspectives.
Dynamic ALM must also include asset-liability rebalancing of the scenarios, as illustrated in Figure 1:
The rebalancing details depend on the policies of an institution (including its risk appetite) and market conditions. All of this leads to a couple of important questions: (1) How much of a firm’s high-quality liquid assets (HQLA) can be sold, and, when raising short-term debt, at what funding cost? and (2) Is there a demand for new loans or an opportunity for new investments?
These questions can be answered within a fully dynamic ALM framework that includes holistic balance-sheet optimization on the full range of future scenarios.
One potential problem is that some balance sheet segments - such as operational deposits - are unobservable. The behavior of these segments is hard to predict, and the regression with regularization techniques used for other segments do not apply, so a modified approach to their projections within dynamic ALM is required.
Operational deposits have some interesting characteristics. They are stable and not sensitive to interest-rate fluctuations. Bank clients need these deposits to pay their regular scheduled expenses, and thus cannot reallocate these funds to higher-yielding instruments.
Operational deposits are also important for HQLA estimation, and liquidity regulation defines very well what operational accounts might look like. But the liquidity factor doesn’t prevent banks from keeping additional funds in these types of accounts. So, the actual level of operational deposit within an operational account is not observable. To qualify for stable funding, banks therefore have to separate excess cash from the actual level of operational deposits.
One way to estimate such unobservable levels is to train a logistic regression on two easily identifiable sets: accounts with zero levels of operational deposits and accounts that are 100% operational. These two sets can be found through their behavioral patterns. The former type has rare cash outflows, which means it can’t possibly be used for operational expenses. The latter type makes frequent payments and has a low deposit-to-payment ratio, which means there is no excess cash in these accounts.
Logistic regression trained on these two sets can be applied to all other accounts. Obtained results can then be treated as a probability of an account being 100% operational. Such probability, multiplied by the total deposit amount in the account, gives the expected level of operational deposits.
Implementing a truly dynamic ALM process enables financial institutions to inform decision-making in both strategy and risk. Moreover, modern analytical tools based on the full distribution of potential outcomes can help financial institutions determine optimal capital allocations.
While dynamic AML requires an extra effort initially, financial institutions can use it to substantially improve both profitability and their preparedness for weathering future storms.
Given today'schallenging environment, advances in ALM strategies will surely prove critical in maintaining competitiveness.
Alla Gil is co-founder and CEO ofStraterix, which provides unique scenario tools for strategic planning and risk management. Prior to forming Straterix, Gil was the global head of Strategic Advisory at Goldman Sachs, Citigroup and Nomura, where she advised financial institutions and corporations on stress testing, economic capital, ALM, long-term risk projections and optimal capital allocation.