Backtesting and Optimizing Commodity Hedging Strategies

Transcription

1 Backtesting and Optimizing Commodity Hedging Strategies How does a firm design an effective commodity hedging programme? The key to answering this question lies in one s definition of the term effective, which can have a very different meaning even to people within the same organization, depending on their personal views and functional perspectives. For example, back-office functions rely on prevailing hedge accounting standards to make this determination. In contrast, mid-office personnel in risk management will tend to focus on the hedging program s success in reducing the volatility of key financial metrics, such as operating margins 1 or earnings. And from the senior management perspective, no matter how well a hedging programme performs on other measures, it will not be deemed effective if it generates heavy losses or results in large margin calls that damage the firm s liquidity position. Based on our experience, these are not mutually exclusive goals. An integrated approach can support a well-rounded and robust programme designed to have a positive impact on the firm s risk-adjusted performance. In this article, we introduce a set of hedge performance metrics and reveal how they can complement and reinforce one another in the context of a case study focused on a North American airline s jet fuel hedging programme. Jet Fuel Hedging by Airlines Given how sensitive their operating performance is to swings in energy prices, most large airlines do conduct active hedging operations. Although jet fuel hedging has been standard industry practice for well over a decade, there is still little consistency across airlines hedging strategies, and little agreement as to what constitutes best practice. Some programmes are highly systematic, while others rely heavily on trader discretion to determine the timing of trades and the size of hedge positions. There are also significant differences in the extent to which different carriers use options in their hedge portfolios, in the hedging proxies they employ and in their hedge horizons. Moreover, recent changes in the energy markets and regulatory environment have prompted airlines to reassess and, in many cases, redesign their hedging programmes. In Figure 1, we can see the rapid rise of oil prices to over $140/bbl in the summer of 2008, followed by their precipitous $100/bbl collapse within just six months, generated severe mark-to-market losses for a number of carriers that adversely affected operating results for several quarters thereafter. In the wake of that episode, some airlines reexamined their practice of always maintaining substantial hedge positions, regardless of prevailing market levels. And others reined in their tendency to increase their hedge positions as energy prices rose. During the same period, several carriers began to question and revise their practice of using crude oil derivatives as a more liquid and therefore less expensive alternative for hedging their jet fuel risk. The correlation breakdown between crude oil and petroleum product pricesresulted in an unacceptable 1 Operating margins are defined as operating income divided by revenues. 1 P a g e

2 level of basis risk between jet fuel and the proxies being used to hedge it (see Figure 1). In February of 2009, for example, the price of prompt month NYMEX NY Harbor heating oil declined by almost 13% despite a 7.4% increase in the price of NYMEX WTI crude oil. Figure 1: WTI Crude Oil, NY Heating Oil and LA Jet Fuel Prompt Prices WTI Crude Oil Price ($/bbl) $ $ $ $ $ $ $80.00 $60.00 $40.00 $20.00 $0.00 WTI Crude Oil, NYH Heating Oil and LA Jet Fuel Prompt Month Prices (Dec. '07 - Mar. '13) $4.50 $4.00 $3.50 $3.00 $2.50 $2.00 $1.50 $1.00 $0.50 $0.00 Heating Oil & Jet Fuel Prices ($/gal) WTI Heating Oil LA Jet Fuel Additional challenges in recent years have emerged as a result of changing regulatory and financial accounting standards. Dodd-Frank and Basel III have created uncertainty regarding the costs and funding requirements for non-cleared, over-the-counter trades, driving some corporate hedgers to reconsider their product preferences. At the same time, corporate accountants have been busy trying to ensure compliance with rapidly evolving and converging domestic and international accounting norms 2. The Subject of Our Case Study Our airline is a medium-sized carrier whose most important hub is Los Angeles International Airport (LAX). It has a hedging programme in place that utilizes NYMEX WTI crude oil futures to hedge its purchases of jet fuel in Los Angeles, which are priced off of the Platts LA jet fuel index. The airline uses approximately 10 million gallons of LA jet fuel each month, and the proportion hedged has historically been 50% of the next two quarters expected consumption, distributed equally across the months. 2 For an overview of upcoming hedge accounting changes in International Financial Reporting Standards, see Blanco et at (2012) 2 P a g e

3 The airline wishes to evaluate two of the main components of its hedging programme. First, given recent volatility in the spread between WTI and its products, the airline is considering whether it should instead be using a different proxy, such as New York Harbor heating oil futures. Second, management is searching for ways to improve upon its current practice of hedging a constant proportion of its fuel requirements. It would like to explore an alternative strategy where, as market conditions change, hedging levels are systematically adjusted to try and maintain operating margins above a given threshold. Given the financial accounting implications, the question of whether heating oil might be a more effective proxy than crude oil requires the application of analytical methods based on hedge accounting principles. Hedge accounting tests will not be very helpful, however, on the question of when or how much to hedge. In a later section, we will introduce backtesting analysis, which involves simulating a potential hedging strategy s performance over time using actual market and operating data, is an especially powerful and intuitively accessible technique for determining optimal hedge levels. Therefore, even in a case where only a few design elements of a hedging programme are being assessed, we must apply at least two methods which are drawn from different functions within the organization.. We will now illustrate the application of these two techniques in detail and conclude by discussing additional analytical methods for simulating hedging programmes performance and optimizing their design. The Financial Accounting Perspective Most public companies with significant hedging programs prefer to use special hedge accounting, since it allows them to recognize in earnings only those hedging gains and losses whose timing corresponds to the recognition of cash flows being hedged. This avoids excessive earnings volatility due to swings in the mark-to-market value of the entire hedge portfolio. In order to qualify for hedge accounting treatment, companies must establish, analytically, that their hedges are effective, on both a prospective and retrospective basis. This is the key requirement that must be satisfied either through IAS 39, if a firm follows international accounting guidelines, or ASC 815 (formerly FAS 133), for US companies. In order to pass the prospective and retrospective tests, it must be shown that the notional price of the derivative (i.e. hedge instrument) is strongly correlated with the price underlying the hedged cash flow (i.e. hedged item). And generally accepted industry practice is that the R-squared statistic for the regression between the hedge instrument and the hedged item must be at least 80% to pass the test. This standard provides a basis for comparing the performance of heating oil versus crude oil as proxies for hedging jet fuel. Since the airline has a 6-month hedging horizon, we chose a time series of 3 rd nearby forward prices for our analysis. We expected these to be more representative of the hedge portfolio s overall behavior than shorter or longer maturity contracts might have been. 3 P a g e

4 It is also a generally accepted industry practice to use price returns, rather than absolute price levels, when calculating correlations. This provides a more normalized measure of correlation, and is the approach preferred by the major audit firms. Finally, we used three years of historical monthly forward curves, since it is a rule-of-thumb within the industry to use at least 30 observations. Figure 2 shows linear regressions between LA jet fuel and each of the potential hedging vehicles, WTI crude oil and NYH heating oil futures. Figure 2: Joint Log Returns for LA Jet Fuel and Alternative Proxy Hedges - R-Squared and Slope Statistics WTI Crude Oil vs. LA Jet Fuel Monthly Log Returns (Apr '10 - Mar '13) NYH Heating Oil vs. LA Jet Fuel Monthly Log Returns (Apr '10 - Mar '13) R-squared Slope WTI Crude Oil 76.6% 0.75 Heating Oil 96.1% 0.98 Prospective testing indicates that heating oil clears the R-squared threshold for hedge effectiveness, while WTI crude oil does not. Another statistic of interest for meeting established hedge accounting guidelines is the slope of the regression between LA jet fuel returns and each of the potential hedging proxies. And again, while WTI crude oil futures failed to fall within the acceptable range of 0.80 to 1.25, NYH heating oil futures performed very well. While a complete analysis for hedge accounting purposes would need to include retrospective testing as well, it is clear that this analysis provides very useful guidance on the relative effectiveness of different possible proxies. We now turn to a complementary analytical method to provide more insight on this issue, but to also help address the question of optimal hedging levels. Backtesting Financial Performance When senior managers evaluate the effectiveness of a hedging strategy, they tend to focus on the programme s contribution to financial results over time. There is usually little interest in achieving a high degree of certainty if it comes at the expense of overall performance. Therefore, some of the questions commonly asked by the CEO, CFO and Board of Directors before a hedging programme is formally approved include: 4 P a g e

5 What is the programme s expected impact on operating margins and earnings? What are the initial and maximum margin requirements, on an expected and worst-case basis? What are the other costs of the hedging programme (e.g., option premia, transaction fees)? As we shall discuss later, forward-looking simulation techniques can provide valuable insights into these and other concerns. But a more intuitive and straightforward approach is to evaluate how a strategy would have performed under actual historical conditions. It is difficult to make the case that a hedging programme that would have failed to protect a company from the most adverse price movements of the past decade will be successful in countering unfavorable price spikes in the future. In the same vein, a program that would have historically reduced the variability of operating margins will not generate much enthusiasm among senior management if that stability came at the cost of substantially depressed performance. This historical analysis approach is known as backtesting, and is commonly used by traders to evaluate the performance of alternative trading strategies. Actual historical data has dynamic relationships and complex behaviors embedded within it that are present within financial markets and the broader economy, but may not be captured by other modeling approaches. This important strength is one of the main reasons why historical simulation, a close cousin of backtesting, is a very popular method for estimating Value-at-Risk (VaR) among leading financial institutions. While past performance may not guarantee future results, it is often the best predictor that we have. Returning to the subject of our case study, we are going to compare the historical performance of the current hedging strategy to a more dynamic approach. If you recall, our airline s current practice is to hedge a constant proportion of expected future jet fuel consumption and to simply roll the hedges forward. Let s assume that the risk manager has proposed to systematically adjust the size of the hedge in response to changes in contemporaneous operating margins. More precisely, when prevailing operating margins are relatively high, accounting for seasonality and historical levels, hedging levels are increased. And in those periods when operating margins are relatively weak or negative, the airline s hedge positions are reduced. The central objective of this margin-dependent hedging strategy is to identify periods when margins are exceptionally robust and to try and preserve them into the future. Conversely, when margins are well below historical averages, minimal hedging is done, since large hedges would essentially lock in the poor margins and limit their ability to recover. This approach is based on the empirical evidence that operating margins in many industries have a tendency to mean revert, or return to a long-term equilibrium level, over time. To approximate real-world conditions as closely as possible, we used publicly available data on operating margins for a major airline 3 from June 2004 until September 2012 as the basis for our airline s historical performance(see Figure 3) in our backtesting analysis. We then applied a scaling formula that determined the percentage of jet fuel exposure that would have been hedged at the start of each 3 Since other elements of the case study may not reflect the practices of this airline, we have chosen to keep the company name confidential. 5 P a g e

6 quarter under the dynamic hedging strategy, based on prevailing operating margins. The resulting hedging levels are also shown in Figure 3. Figure 3. Quaterly Operating Margins and Margin-Dependent Hedging Levels Operating Margins Quarterly Operating Margins and Margin-Dependent Hedge Levels (%) (June '04 - Sept' 12) Hedge increased to above average level Hedge decreased to minimum level 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% % Hedge Margin-Dependent Hedge % Operating Margin (pre-hedging) Rolling-Avg. Margin Once we had determined historical hedging levels for the margin-dependent strategy, we were able to backtest and compare it to the existing approach ( constant-proportion roll ). We did this for a five year period starting in September 2007 for both WTI crude oil and NY Harbor heating oil futures as possible hedging proxies. The operating margins shown in Figure 4 were calculated by adding the P&L realized each quarter under the different hedging strategies to the pre-hedging margins shown in Figure 3. Figure 4. Backtested Operating Margins for Constant Proportion and Margin-Dependent Hedging Strategies Using Crude Oil (left) and Heating Oil (right) Proxies Operating Margins for Alternative Hedging Strategies with WTI Crude Oil Futures (Sept'07 - Sept '12) Operating Margin with Margin-Dependent Hedge Operating Margins for Alternative Hedging Strategies with NYH Heating Oil Futures (Sept'07 - Sept '12) Operating Margin with Margin-Dependent Hedge Operating Margin with Constant-Proportion Roll Hedge Operating Margin with Constant-Proportion Roll Hedge 6 P a g e

7 The key feature evident in the backtesting charts is the margin-dependent strategy s success in sidestepping the large hedge book losses that the constant-proportion strategy incurred in the second half of 2008, when oil prices plunged to below $40/bbl. Unfortunately, this airline did suffer very significant losses during that period, as did many others, as a result of the long position it was carrying. The margin-dependent strategy would have avoided these losses because operating margins during the first half of 2008 were severely depressed by record-high oil prices, and this would have prompted the airline to systematically wind down its hedge position to a minimal level before the price collapse. Table 1. Key Return and Risk Metrics for Alternative Hedging Strategies (September 2007-September 2013) WTI Crude Oil Futures NYH Heating Oil Futures Operating Margin Constant-Proportion Margin-Dependent Constant-Proportion Margin-Dependent Mean -1.4% 1.1% -1.2% 1.7% Standard Deviation 10.1% 6.9% 11.4% 7.2% Table 1 reveals that the margin-dependent hedging strategy outperformed constant-proportion hedging for the period studied. Margin-dependent hedging would have generated substantially higher average operating margins, with much lower variability, than the airline s current approach. The backtesting results also confirmed that heating oil futures may be a more effective proxy than the one currently in use (WTI crude oil), given the 60 basis point improvement in expected margins, and only 30 basis point increase in variability. But clearly, this is only a supplementary indication to the more compelling results generated by the hedge accounting tests, revealing again how the different methods complement and reinforce one another. The Risk Management Function s Perspective As we move to the mid-office, simulation emerges as a tool of choice. From the perspective of the risk management function, a hedging program s effectiveness is best evaluated by quantifying its impact on the distribution of potential outcomes by using certain key risk indicators such as operating margins at risk, potential hedging losses, or potential funding liquidity risk, since the funding requirements associated with initiating and maintaining a hedge portfolio introduce costs and capital requirements that can undermine an otherwise sound hedging strategy. Conducting a comprehensive risk analysis of a hedging programme requires understanding the long-run dynamics of rule-based hedging strategies rather than static portfolios. This is the main reason why traditional market risk metrics such as value-at-risk are not widely used to evaluate hedging programmes as they provide limited insights The right simulation-based metrics may provide invaluable, forward-looking guidance on the risks involved in a hedging programme, but based on our experience, we have found that most firms see 7 P a g e

8 them as too complex to implement and too difficult to explain to others that lack the financial risk management expertise. Optimization of hedging strategies So far, we have discussed backtesting analysis applied to pre-defined rule-based strategies. Another application of backtesting analysis is to identify optimal hedging strategies based on certain objective functions as well as risk and return constraints. An alternative is to use simulation-based analysis with Monte Carlo. Figure 5 shows a flow diagram with the steps involved in the optimization of a hedging strategy Figure 5: Optimization of Hedge Strategies The process of optimizing a hedging strategy begins with the definition of its parameters and the specification of the range of choices available for each element of the programme. For example, liquidity, credit or other constraints may dictate a floor or ceiling on the possible size of the hedge position. Other parameters that must be fully defined include available instruments and the hedge horizon. Next, the objective function to be optimized must be clearly specified. This is comprised of key risk and return indicators that reflect the firm s strategic goals for its hedging programme. The optimization can be performed based on historical market and operating financial metrics or, alternatively, via an integrated Monte Carlo simulation. Monte Carlo simulation offers the flexibility needed to fine-tune or calibrate hedge programme parameters with greater precision than that afforded by backtesting, since the universe of market data and financial metrics is not restricted to historical operating and market data. Ultimately, optimization is a methodical process for aligning the expected performance of the hedging programme with corporate objectives. 8 P a g e

9 Summary and Conclusions Successful commodity hedging programmes are those that manage to strike the right balance between a series of tradeoffs such as risk management objectives, accounting standards and funding liquidity considerations. In contrast, hedging programs that lack commonly accepted benchmarks and metrics to track their performance in an integrated fashion are usually perceived to be a failure and often discontinued should there be significant losses. For firms that rely on hedge performance metrics, backtesting analysis can be an important tool for evaluating how a set of key risk indicators (KRIs) would have performed in the past for a particular hedging strategy and gain valuable insights from the analysis. In addition, once a series of KRIs have been identified and optimal targets have been set, backtesting and simulation analysis can be used to identify optimal hedging strategies based on a series of risk and return objectives and constraints. Our analysis shows the value of backtesting in designing key features of a jet fuel hedging programme by confirming that a margin-dependent system for determining hedging levels at the start of each quarter would significantly improve risk-adjusted performance over a more static approach. It also shows how backtesting and other methods complement reinforce one another by providing targeted insights into the various elements comprising an effective hedging program. References Blanco, C. Pierce, M. and Aragonés, J.R. (2012) IFRS 9, hedge effectiveness and optimal hedge ratios. Energy Risk. June Authors Carlos Blanco is managing director of NQuantX Tamir Druz is director of Capra Energy Group (NQuantX s alliance partner). 9 P a g e