Balance Sheet Optimization
Balance sheet optimization has been the Holy Grail of companies’ management for many years now. But to quote Fintekminds on the topic, “A lot has been tried and written on the subject, but very few have been able to achieve meaningful results.”
Capital planning, Capital optimization, balance sheet optimization, forward-looking scenarios, risk appetite, strategic planning, scenario generation
16832
post-template-default,single,single-post,postid-16832,single-format-standard,bridge-core-1.0.4,cookies-not-set,ajax_fade,page_not_loaded,,footer_responsive_adv,qode-child-theme-ver-1.0.0,qode-theme-ver-18.2.1,qode-theme-bridge,qode_header_in_grid,wpb-js-composer js-comp-ver-5.7,vc_responsive
 

Balance Sheet Optimization

Balance Sheet Optimization

Balance sheet optimization has been the Holy Grail of companies’ management for many years now. But to quote Fintekminds on the topic, “A lot has been tried and written on the subject, but very few have been able to achieve
meaningful results.”

There were attempts to treat institutions’ (and corporate) balance sheets as portfolios of assets, with balance sheet optimization reduced to traditional investment portfolio optimization in a Markowitz mean-variance framework. It works by finding the best expected return (mean) on portfolio assets for each level of risk (defined as portfolio variance). The difficulty in applying this framework, however, is that banks’ balance sheet optimization, while looking at capital allocation from a risk-reward perspective, has to consider also the full capital structure of the institution (e.g. loans, securities, deposits and wholesale funding).

The main assumption in the mean-variance portfolio optimization framework is that portfolio variance-covariance is constant over the optimization horizon. Since this assumption doesn’t hold over any reasonable time horizon (a week, a month or at maximum a quarter), it implies that the portfolio must be rebalanced periodically; that is, it should be liquid enough. Such rebalancing is infeasible when it comes to managing the balance sheet and making strategic decisions like M&A, divestitures, capital raising, determining risk appetite and establishing long-term goals for the institution.

First, due to longer time horizons of strategic decisions, changes in market conditions can cause dramatic shifts in correlations between the asset classes. Second, if one decides to extend certain loans or merge with another institution, it is not easy (and often not possible) to reverse these decisions. Finally, there are numerous regulatory constraints from capital and liquidity angles that trigger respective changes on the asset and liability sides (e.g., maintaining certain levels of ALLL, high-quality liquid assets and capital ratios).

In addition to the above challenges to balance sheet optimization in the Markowitz framework, regulatory and investment communities expect institutions to incorporate stress considerations into their baseline projections of capital, profit and allowance. It’s not enough to find a balance sheet composition that is optimal from a risk-return perspective. One must also explain why assets or liabilities would behave in a certain way and what macroeconomic and market environment factors caused this behavior.

Even the definition of risk-return trade-off is very different when it comes to balance sheets. It is not clear what is being optimized – capital, return on equity or profits – or what the relevant measures of risk should be, with the variance of the above Key Performance Indicators (KPIs) only one of the candidates. And the constraints of balance sheet optimization are numerous and dynamic – capital and liquidity ratios should never fall below a certain level over the optimization time horizon, earnings should never decline by more than X in any single quarter, worst-case net income should exceed Y, etc. These are difficult to incorporate into a Markowitz framework.

From the regulatory perspective, it is crucial to predict how institutions may respond to some adverse macro-economic scenarios because the practical balance sheet optimization depends on an evolving macro-market scenario and on management actions that are scenario-dependent.

Given these challenges, most optimization approaches simplify the problem by either dealing with narrow silos of an institution and/or by using very restrictive assumptions (e.g., a single-time step) to maintain computational transparency and still use the static mean-variance approach. As stated in a recent European Central Bank research paper, “Financial institutions need a new approach from the very senior strategic level on how to optimally manage their balance sheets and make them more resilient to potential new stress, both from capital and liquidity perspectives.”

Multiple forward-looking scenarios are the only way to incorporate longer-term risks into analysis. An institution’s balance sheet can be projected on these scenarios, but manually building a few dozen scenarios is not enough. What’s needed is the full distribution that incorporates potential extreme events and their consequences, captures dynamic correlations between PDs, LGDs and EADs, and allows the integration of macroeconomic and market risk drivers and their impact on various deposit segments, non-interest income (e.g., fees), demand for new loans, etc. Such scenarios include all components necessary to calculate high-level KPIs: capital and liquidity ratios, ALLL and ECL, funding costs, net income and earnings. An institution can then overlay its specific covenants on the scenarios (e.g., when the unfunded commitment will be closed, or additional collateral required).

Once all the details of critical outcomes are calculated on all of the generated scenarios at every monthly or quarterly point over the time horizon, comprehensive balance sheet performance can be assessed (e.g., expected net income in Q1, 95th percentile worst case net income in Q1, same for capital and liquidity ratios, same for Q2 and others). Then one can back-track (reverse) scenarios leading to particular outcomes, identify specific risk drivers’ values producing these outcomes and design potential balance sheet alternatives that might mitigate adverse outcomes and/or strengthen the good ones.

Multitudes of such balance sheet alternatives or managerial actions will create a cloud of choices from which an efficient frontier can be formed. Each balance sheet alternative (represented by the weights of balance sheet and P&L segments) has many outcomes (represented by statistical expectations and the worst cases of all key indicators). Select two dimensions – one of which will represent risk (e.g., by how much capital can drop over the time horizon) and another return (e.g., expected profitability over the horizon) – and calculate all necessary regulatory and internal constraints (e.g., capital and liquidity ratios, maximum drop in earnings in a single quarter). Now the frontier can be constructed – for each level of risk, what is the balance sheet alternative with the maximum return that satisfies all constraints?

This approach to balance sheet optimization requires only high-level data, helps one discover hidden pockets of risk through the generation of unprecedented scenarios that capture increases in correlation, and makes it possible to overlay dynamic decisions on these scenarios (e.g., reinvestment, refinancing, risk appetite, capacity and limits). Not only is it fully transparent, it incorporates all types of risk considerations and truly informs an institution’s strategic decisions.