Facing continued energy price volatility and tightening requirements for risk estimation and disclosure, banks need stronger analytical guidance to make their way.
Recent turbulence in the energy industries has sent shockwaves across North American banks, triggering a cascade of questions from investors, regulators and boards of directors. Stakeholders are clamoring for more information on lending exposures, compounding the pressure on senior management to justify decisions on loss reserves.
But what about decision-making and communication tools for commercial lending risk? In most cases they fall far short of what is needed. On the one hand, current risk models reflect years of expert development and consider credit implications in good and bad market scenarios. On the other hand, most have a critical weakness — they do not assign any probabilities to the wide range of possible energy pricing developments.
Unsupported by a dynamic and statistically sound view of potential outcomes, the typical energy loan risk model paints a blurry picture of loan valuation and loss exposure. Even the better renditions struggle to keep up with a changing market, given their heavy reliance on a trailing series of dry internal credit ratings.
This is an unacceptable handicap at a time of serious energy lending challenges for North American banks. For a period of years prior to the oil price collapse that began in mid-2014, energy loan underwriting was based on market prices ranging from $75 to $110 per barrel for the benchmark West Texas Intermediate crude (WTI). While prices have recovered somewhat from the $26 nadir seen earlier this year, they remain seriously underwater relative to prior market conditions and business assumptions.
To put it in perspective, based on a mid-July 2016 price of roughly $45 per barrel, all of the energy loans booked during the four years ended mid-2014 were originated when market prices were 67% to 144% higher than today (Figure 1: Collapsing Price Support for Yesterday’s Energy Loan Decisions). This is not to suggest a hopeless situation, as it is thought that most borrowers can eke by if prices stay in the $40s. But given the outlook for further global over-production and energy price volatility, it does suggest continuing tight circumstances for borrower and lender alike. Banks will need improved information to:
- Diagnose expected losses for risk management and shareholder communications;
- Optimally modify problematic loans that come up for reegotiation;
- Value loan contracts in the event of elective or forced sales; and
- Meet new accounting standards on the horizon, which will force an earlier recognition of impaired loans. These include the Current Expected Credit Losses standard (CECL), as specified by the Financial Accounting Standards Board, and the International Financial Reporting Standard 9 (IFRS 9), as promulgated by the International Accounting Standards Board.
To meet this analytic challenge, a new type of modeling framework is needed, one that takes fuller advantage of the rigorous simulation techniques now used in many areas of science and industry. In contrast with today’s energy loan risk modeling, which typically considers a handful of best- and worst-case scenarios to deliver qualitative estimates at a single point in time, simulation can continuously project the interaction of key risk variables across thousands of potential scenarios, identify major patterns, and rigorously evaluate the probability of potential outcomes, both good and bad.
To be sure, energy lenders are quite thorough in the application of their current risk models. They develop expert views on reserves, anticipate revenues by matching conservative production estimates with various price forecasts and upside/downside scenarios, and net out production costs and the impact of derivatives to forecast the earnings and free cash flow of the enterprise. Essential in initial underwriting, this work is replicated/refreshed to varying degrees when banks conduct their periodic portfolio reviews and update internal ratings of credit quality.
The problem, as exposed by the collapse in market prices, is that the conventional management dashboard provides only partial guidance (Figure 2: Questions that Most Bank Models Typically Cannot Answer). It is not designed, for example, to independently generate and evaluate various emerging risk scenarios based on the observed variability of key performance factors, or to assign probabilities to those potential outcomes. It does not provide continuously refreshed and statistically robust estimates of default risk. Nor does it provide mark-to-market insights on impaired loans, or those considered for sale.
Properly constructed, modern simulation models help to address these and other questions by sorting through hundreds or thousands of computer-generated scenarios, each representing a unique set of variations in market indices and key risk factors, to identify emerging patterns and their implications for today’s portfolio. And while this may come across as technical mumbo-jumbo to veteran energy lenders, the fact is that simulation techniques are already carrying a heavy workload elsewhere in many of their own banks, and also among their client energy companies:
- Capital markets — Energy analysts long have used simulation analysis to assess the risk and valuation of all types of energy supply contracts; derivative contracts; counterparty peak future credit exposure; and credit instruments, such as credit default swaps for energy assets and companies. Simulation also forms the bedrock for energy risk reporting in market risk and credit risk functions, as well as front office pricing and structuring decision-making.
- Reserve estimation — Popular software tools already in use at major banks (e.g., PHDWin) offer thousands of reserve simulations, allowing bank loan analysts to evaluate energy company strategic responses to volumetric surprises. The effects of volumetric uncertainty are incorporated into modeled borrower pro formas and used to assess the likelihood of loan repayment.
Even in the face of these compelling examples, some executives still may argue that “We don’t need this level of sophistication because the worst is over, the market is coming back and risk is being reduced.” But we disagree for two main reasons:
First, many of the factors that drove the price collapse are still in play, and while a return to the depths of the cellar may not be likely, the sector is facing an extended period of market volatility, sluggish pricing and heightened risk exposure (Sidebar: Likely Turbulence Ahead). Second, tightening accounting and regulatory standards will ultimately force greater sophistication in energy risk modeling — no matter the future price environment. Why wait to build these tools when they are needed now?
A Tale of Two Companies
Banks have voluminous information about energy clients and loan contracts, but most of it is loaded into static risk models that are refreshed only occasionally. The main output of these periodic reviews is an updated internal credit rating based on a simplified version of the bond rating scales used by Moody’s and Standard and Poor’s. Some credits are confirmed in their current category (e.g., “AA, BBB, BB+”), and others are reclassified upwards or downwards. Then the bank rolls up all the credits in each ratings category for a view of portfolio risk.
While this work is rigorous in its own way, it errs on the side of being more of a static reference system than an active tool for decision-making and communication. When pressed by bank stock analysts or regulators about energy portfolio risk and loss reserve adequacy, executives necessarily must stick to a narrow script. Their models are not set up for a multi-faceted review and forecast of risk variables and probabilistically-handicapped outcomes, either for individual credits or the composite portfolio.
For a glimpse of how the progressive bank will address these questions in the future, consider a sample cash flow simulation for two energy companies (Figure 3: A Tale of Two Companies — Free Cash Flow vs. Debt). The two companies were hit by the same tidal wave when energy prices collapsed, yet their cash flow simulations paint very different pictures of resilience in the face of future market volatility.
Company A — This energy company has a track record of production efficiency, avoids heavy debt leverage, and is skilled in the use of derivatives to hedge its exposure to fluctuating energy prices. By examining the varying influences of key performance factors over period of years, and then extensively modeling how these factors might interact in combination with various price scenarios, the lender can see that Company A has a low probability of default. By no means is it immune to distress, to be clear, but it likely will be able to service its debt in a variety of scenarios.
Company B — This energy company aggressively expanded at the top of the market, incurring a heavier debt load to mobilize the higher-cost production resources available in peak market conditions. While simulation indicates that it still has a chance of staying afloat in the emerging market, it also shows much higher default risk.
Once the bank assembles this type of analysis for each borrower, the results can be rolled up for a portfolio-level view of the emerging risk environment. For senior management, the benefit is a rolling five-year forecast that includes a month-by-month summary that plugs into formal accounting/regulatory classifications, including probability of default (PD); loss given default (LGD); exposure at default (EAD); and expected loss (EL). This type of modeling also provides mark-to-market insights — increasingly needed but unavailable at many banks today.
Building the Model
The key to simulation modeling is extracting detailed risk insights from a statistical study of historical performance drivers (and their variability) and projecting these factors into future market scenarios. In energy borrower risk modeling, much of the work lies in preparation. This includes building a proper foundation of historical and current information, establishing the right variability assumptions for each performance factor, and linking market data and projections for dynamic modeling.
The model drivers should include borrower characteristics, loan terms including covenants, reserve economics, dynamic energy prices, energy volatility and energy price probability distributions. They also should consider hedging programs and the knock-on effects of changing energy prices, such as costs of energy services and transport (since the cost of these services also varies with energy prices).
The model outputs should include expected losses, loan valuations, and sensitivities of losses and valuation to changes in inputs. Finally, the models should also be able to include covenant valuations in order to guide the lender in modification, restructuring and sales decisions.
All of this preparation comes together in the mark-to-market energy loan model, which differs from today’s basic pro forma models in three major ways:
Knock-on effects. Most pro forma spreadsheets are based on static analysis, where a few variations are introduced (e.g., energy price and foreign exchange scenarios) while all other factors are held constant. Potential offsetting factors and consequent developments are not considered. The first step is to assure that changes in energy prices and foreign exchange rates ripple correctly through the pro forma; e.g., lower energy prices and output would imply reductions in capital expenditures at some point.
Price scenarios. Robust scenarios must be generated for major capital markets inputs, including forward curves, implied volatilities and correlations. This must be done in a way that ensures consistency with the current market pricing environment while considering the full range of possible future settings.
Simulation analysis. Once the model is equipped to handle dynamic linkages, then it can be put to work in generating a cascade of probability-weighted upside and downside scenarios, typically 1,000+ iterations. Each scenario processes a unique combination of input variables to assess occurrence and timing of default, loan balance and asset revaluation at default, and recovery at default.
Once a full library of 1,000+ cases is populated, the weighted results are compiled to determine overall default frequency by month, EAD by month, LGD by month and EL by month. These metrics can be further aggregated to determine annual averages for regulatory reporting purposes. The model also evaluates loan cash flow, which is discounted to determine the value of the loan at inception under each scenario. The weighted average of the 1,000+ cases implies the value of the loan contract, and the probability distribution of the present value can be taken as a risk measure, similar to value-at-risk.
Benefits of Simulation Modeling
Why should banks use a mark-to-market style model instead of their current static models? The first benefit is improved scenario analysis. The bank can use a comprehensive set of scenarios based on projections of emerging market prices, and assign probabilities to these scenarios. It can also assess whether conservative scenarios are conservative enough. These insights rely on the incorporation of market probabilities implied by forward curves and option prices.
Second, periodic internal credit ratings do not work well for companies whose cash flow distributions hinge on rapidly changing oil prices and their attendant volatilities. Key metrics, such as the probability of default and loss given default, should be based on probability distributions of changing oil prices over time.
Third, the mark-to-market approach can accommodate changing forward curves and implied volatility curves. While a conventional bank model can accommodate changing oil price forecasts, it usually does not reflect price volatility — a critical consideration for banks, given that the benefit of a price surge is capped at full repayment of principal and interest, while downside exposure can be devastating. Higher volatility means higher expected losses, and few bank models capture this.
Fourth, the mark-to-market approach provides real-time loan contract valuation, along with dynamic risk assessments. Over the life of the credit, this is far more useful than conventional bank models, which are not built to value loans past inception.
Fifth, the mark-to-market approach provides a methodology for evaluating covenants. This is useful at origination, where alternative loan covenants can be evaluated for effectiveness in risk reduction. Further, upon renegotiation of terms, new covenants can be evaluated to determine how bank protections can be maximized.
Sixth, production and capital expenditure (CAPEX) optionality can be built into a simulator and valued explicitly.
Call to Action
Though energy lenders may have caught their breath as prices retraced from their lowest levels, volatility shows no signs of permanent abatement in the wake of Brexit; increasing political challenges in the Middle East; and rising U.S. production strength. Falling oil prices are not only a geopolitical exigency, but the obvious market response to a stronger dollar caused by investors seeking safe havens. Oil, like most dollar based commodities, shrinks as the dollar swells.
The oil markets and currency markets have jointly exposed a widening gap in the analytic models that banks critically rely upon for answers to their loan portfolio questions. Just as a general should not go into battle unequipped, bank executives should not face their constituents without a foundation of cogent analysis.
Today, simple and reasonable shareholder questions like “What is the liquidation value of the bank’s energy loan portfolio?” or “What would a $10 drop in oil prices mean for expected losses?” or “How has the bank managed the risk of its energy loan portfolio?” largely go unanswered. Wall Street analysts and investors are tempted to assume the worst, and share value suffers.
It is not enough to try and make do with today’s limited energy credit risk models and hope that oil prices rise (or at least stabilize) so that constituents stop paying attention and the bank can return to business as usual. Facing continued energy price volatility and tightening requirements for risk estimation and disclosure, banks need stronger analytical guidance to make their way. Simulation modeling should be part of the answer.
David Shimko and Brett Friedman are Directors in the New York office of Novantas. They can be reached at firstname.lastname@example.org and email@example.com, respectively.