MSc International Business and Management: International Financial Management University of Uppsala
MSc Business and Economics
MASTER THESIS
The Determinants of Exposure and Hedging in the Airline Industry
Karst Tuinstra Student number: 1475290 k.w.tuinstra@student.rug.nl Supervisor: Dr. N. Brunia Co-assessor: Prof. Dr. C.L.M. Hermes
The Determinants of Exposure and Hedging in the Airline Industry
Abstract
Table of Contents
I. INTRODUCTION ... 4
II. LITERATURE REVIEW ... 6
2.1 Exposure Definition and Identification ... 6
2.2 Exposure Sign and Magnitude... 8
2.3 Exposure and Hedging ... 11
2.4 Financial Hedging Incentives ... 12
2.4.1 Financial Distress Costs ... 12
2.4.2 Underinvestment Costs ... 13
2.4.3 Taxes ... 13
2.4.4 Agency Costs ... 13
III. METHODOLOGY ... 15
3.1 Estimation of Exposure and Its Determinants ... 15
3.2 Determinants of Exposure Defined ... 16
3.2.1 Gross Exposure Determinants ... 17
3.2.2 Reported Financial Hedge Levels... 18
3.2.3 Theoretical Financial Hedge Level Proxies... 19
3.3 Estimation of Determinants of Reported Financial Hedge Level ... 21
IV. DATA DESCRIPTION ... 24
4.1 Sample Selection ... 24
4.2 Data Sources of Exposure Estimation ... 25
4.3 Data Sources of the Determinants ... 27
V. RESULTS ... 31
5.1 Exposure ... 31
5.2 Determinants of Exposure ... 33
5.3 Determinants of Reported Financial Hedge Level ... 37
VI. CONCLUSION ... 40
APPENDIX I – Reported Financial Hedge Level Calculation ... 42
Jet Fuel Price ... 42
Interest Rate ... 43
Exchange Rate ... 44
I. INTRODUCTION
Changes in market prices like commodity prices, exchange rates and interest rates can change the market value of a firm. A firm’s exposure to market prices is of interest to managers because effective risk management requires in the first place a proper assessment of the risk. Up until 2008, airlines had had a ten year period of steady increases in the price of jet fuel. In 2006, fuel costs replaced personnel costs as the leading industry operating expense.1 When halfway 2008 the fuel price suddenly showed a steep drop, not all airlines were evenly relieved. In 2009 a header in The Financial Times read “Fuel hedges weigh on Air France”2. For the 2009/2010 fiscal year, pre-2009 hedging had a total negative impact of 637 million euro on the results of Air France-KLM.3 Taken together with declining passenger demand due to the economic downturn, the year ended with an operating loss of 1.28 billion euro. Air France-KLM used financial hedging through instruments like options and swaps to manage their exposure to the price of jet fuel. Whether financial hedging by firms is desirable or not is still a matter of debate. Investors can off-set exposure themselves through portfolio diversification would they desire to do so. The main argument for corporate risk management is that it can increase a firm’s value when financial instruments are used in such a way that they pay-off in times when cash flows are lower than what the firm needs to operate optimally. More specifically, hedging can increase the value of a firm in the presence of capital market imperfections. Smith and Stulz (1985) argue that bankruptcy costs are a market imperfection which allows hedging to add value to a firm since less volatile cash flows can decrease the probability of a firm going under. Another incentive for hedging is that it can match internally generated funds with investment opportunities in times when external financing is costly (Froot, Scharfstein and Stein, 1993). Graham and Rogers (2002) argue that a reduction in risk through hedging allows a firm to take on more debt, thereby increasing the tax shield benefits. A firm may also want to hedge when risk-averse stakeholders require additional compensation for bearing the undiversifiable risk of their claim on the firm (Stulz, 1996). My thesis is interested in finding out whether financial hedging practices in the airline industry conform to the predictions of the optimal hedging theories. I do this by answering my research question:
Research Question: What are the determinants of exposure and hedging in the airline industry?
The airline industry is particularly suitable because they face a multitude of market risks which they manage actively. The price of jet fuel is highly volatile and the jet fuel expenses make up a substantial part of operating expenses. The industry is characterized by a relatively high level of debt because
1 Source: Association of European Airlines.
URL: http://files.aea.be/RIG/Economics/DL/SumRep07.pdf (August 8th, 2012).
2 Source: The Financial Times. URL: http://www.ft.com/cms/s/0/b57678a0-d47e-11de-a935-00144feabdc0.html#axzz24E1vu3Ki (August 8th, 2012)
Financial Times,18-11-2009
aircraft serve as excellent collateral for debt financing. This makes airlines susceptible to changes in the interest rate. The industry is also characterized by intense competition. In such an environment, changes in exchange rates can seriously affect an airline’s performance. On top of that, the United States dollar (US$) denomination of costs such as fuel and aircraft acquisition is an important exchange rate factor to consider for non-US airlines. I therefore aim to find the determinants of exposure and hedging for the jet fuel price, interest rate and exchange rate.
To find the determinants of exposure I use a two-step regression model. First I estimate the exposure of airlines to the three market prices. I than relate the magnitude of these exposures to firm characteristics that I have identified as possible determinants. I expect the magnitude of exposure to decrease with the level of hedging. For each market price, I test two measures of the level of hedging. The first is a single measure extracted from the annual reports. The second measure is a set of firm characteristics that proxy the level of hedging based on the optimal hedging theories. Finally, I empirically test the relationship between the reported level of hedging and the theoretical hedge level proxies with a Tobit model.
My sample consists of 28 airlines from both the United States of America (USA) and European Economic Area (EAE) which provide a total of 183 firm year observations in the period 2001 to 2009. My thesis is the first study with a large sample to examine the fuel price, interest rate and exchange rate exposure of airlines. Loudon (2004) quantified the exposure of airlines to these three market prices, but for only one Australian and one New-Zealand airline. Carter, Rogers and Simkins (2006) estimate the determinants of the level of jet fuel hedging of USA airlines. My thesis aims broaden the understanding of hedging in the airline industry by including the risk management practices with regard to the interest rate and the exchange rate.
I find that airlines are negatively exposed to the price of jet fuel, where USA airlines are significantly more exposed than EEA airlines. EEA airlines are also negatively exposed to the interest rate. Cross-sectional and time-series variations in the magnitude of observed exposure is partly explained by the relative success of airline to pass-through exposure to customers. The level of financial hedging is also a significant determinant of exposure magnitude. However, bankruptcy costs, investment costs and taxes are poor explanation to account for the differences in the observed levels of financial hedging. The level of financial hedging is observed to increase with firm size and liquidity and decreases with leverage and investment opportunities.
II. LITERATURE REVIEW
To answer my research question I make use of both the exposure literature (Table I) and the hedging literature (Table II). Section 2.1 defines exposure and continues by identifying the market prices to which the airline industry is most likely exposed. These are the price of fuel, the interest rate and the exchange rate. Section 2.2 discusses a model that predicts the sign and magnitude of exposure. A recurring observation in the exposure literature is that firms are less exposed than predicted by simple models. Bartram and Bodnar (2007) call this the exposure puzzle. They propose the explanation that firms are reducing their exposure through risk management activities. In other words, the magnitude of exposure is determined by the degree of risk management undertaken by the firm. Section 2.3 further expands on this relationship and discusses the different tools available to firms for managing exposure. My thesis focuses on one subset of these tools, namely financial hedging. Finally, section 2.4 discusses the risk management theories that explain a firm’s incentive to use financial hedging.
2.1 Exposure Definition and Identification
Exposure is the sensitivity of the future value of the firm to the future value of market prices like commodity prices, interest rates and exchange rates. For each market price, a firm’s exposure can be measured as the regression coefficient of the change in the firm’s stock return on the change in the market price. The relationship expressed by this coefficient is not necessarily causal as stock prices and market prices can have the same exogenous determinant. Jorion (1990) controls for macroeconomic shocks, as these can simultaneously move stock prices and market prices. This measures exposure relative to the average exposure of firms in a market index. To be able to compare my results I conform to the majority of the exposure literature I predict, measure and interpret residual exposure, which is exposure after controlling for market movements (see table I). When a firm has no residual exposure, this does not mean that the firm is not exposed. This only means that it is not more exposed to the market price than the firms in the market index.
to USA customers and competitors. The Association of European Airlines4 report that for members in
2005 North-American revenues (17.5%) where the largest after European (33.6% excluding domestic) revenues. I therefore expect that the most important currency risk for EEA airlines comes from the euro to US$ exchange rate. Taking the euro to US$ exchange rate even for EEA airlines with a domestic currency that is not the euro seems appropriate because domestic revenues only account for 15.9% compared to 33.6% from European operations. Since the euro is the dominating currency in Europe I assume that its exchange rate effects are more important than the domestic currency to US$ exchange rate. Airlines for America5 report that for its members in 2005 domestic revenues make up
4 The Association of European Airlines is a non-profit industry organization. Source of passenger revenue data: 2006 Operating Economy of AEA. The total revenue of included airlines was 52,144 million euro (this translates to 64,814 US$ with the average 2005 exchange rate).
URL: http://files.aea.be/RIG/Economics/DL/SumRep06.pdf (July 6th, 2012).
5 Airlines for America (formerly known as the Air Transport Association) is the only trade organization of the principle United States airlines – Source of passenger revenue data: 2006 Economic report of the US airline industry. The total revenue of included airlines was 93,449 million US$.
URL: http://www.airlines.org/Documents/economicreports/2006.pdf (July 6th, 2012).
Table I | Literature Summary of Exposure and Determinants
The table gives an overview of the exposure literature. The second column displays the determinants with their sign (+/-) and the significance level (1/5/10%). Determinants that were not found to be significant but which are referred to in this thesis are denoted with NS. The last column lists the characteristics of the study. These include the market risk studied, sample size, sample period, the used model, data frequency of stocks, the predominant sign of the exposures results and the explained variable (EV).
Author (year)
Exposure Determinants Characteristics Tufano
(1998)
Gold price level
Size as equity(MV)+debt
Production hedged % of assets in mining
Leverage as (LTdebt+current debt)/equity(MV)
-/1 +/10 -/1 +/1 +/1
Risk [Gold price]
Sample[48 gold mining firms] Period [1990-1994]
Model [Dimson adjusted 2-step OLS] Stock data frequency [Daily]
Exposure sign found [+]
EV [Quarterly residual exposure] Bartram
(2002)
Liquidity as Operating cash flow / Total Assets Leverage as Debt-to-Equity(MV)
-/1 +/NS
Risk [3-month interest rate] Sample [490 nonfinancial firms] Period [1987-1995]
Model [2-step OLS]
Stock data frequency [Monthly] Exposure sign found [+/-] EV [abs. residual exposure|] He and Ng
(1998)
Export ratio as Exports over total sales Size (+/- hedging) as log[equity(MV)] Liquidity (- hedging) as measured by
(A low) Dividends payout ratio (A high) Quick ratio
Leverage (+ hedging) as LTdebt/equity(MV) Growth (- hedging) as Book/equity(MV)
+/1 +/1 -/1 +/5 -/1 +/NS
Risk [Exchange rate]
Sample [171 Multinational firms] Period [1979-1993]
Model [2-step OLS]
76.2% of total revenues. The highest foreign revenues of USA airlines come from the Transatlantic/European segment (10.6%). USA airlines face competition from European airlines that through code sharing agreements have easy access to the domestic USA market. The competition from European airlines is likely only to increase as international rules and regulations are increasingly liberalized6. I therefore expect the US$ to euro exchange rate to be an important factor in the price
competitiveness of USA airlines.
2.2 Exposure Sign and Magnitude
Exposure is empirically measured as the sensitivity of a firm’s stock returns to unexpected changes in the market prices. To predict the effect of changes in the market prices on the firm value of airlines I use a model developed by Bodnar and Marston (2002). While this model was developed for exchange rate exposure, it can also be applied to commodity prices as is demonstrated by Valkova (2009). To serve the purpose of my thesis I build the model up in a generalized form so that it is applicable to exchange rate, fuel price and interest rate exposure. The model by Bodnar and Marston (2002) does not require the use of stock data as it expresses the value of a firm as the present value of its future cash flows. When net investments are assumed to be zero, cash flow is the after-tax profit (), which is calculated by subtracting total costs () from total revenues () and correcting it for the corporate tax rate (). In this context, a firm is said to experience economic exposure when cash flows (tax profits) change due to market price fluctuations. With a constant tax rate, a change in after-tax profits can come from varying revenues, varying costs or both. If all revenues and costs are constant in period t, except those affected by the market price, the amount that profit changes in period t due to the a change in the market price can be expressed as
= (1 − )(ℎ − ℎ) (1)
The left hand side of this equation is an exposure coefficient,, multiplied by the after-tax profits. On the right hand side, ℎ is the ratio of market price affected revenues to total revenues and ℎ is the ratio of market price affected costs to total costs. When we further define the firm’s profitability as
= ( − )/ and use the definition of after-tax profit = (1 − )( − ), equation (1)
can be rewritten to=ℎ − ℎ 1 − 1 = ℎ + (ℎ − ℎ)(1 − 1) (2)
6 10-K Continental Airlines, Inc. 2007, Item 1A: “The recent "open skies" agreement between the U.S. and the
which, under the assumption of perfect market, gives the exposure elasticity of a profit maximizing firm. With this equation, the percentage change in firm value due to a 1% change in the market price is measured as a function of only three inputs: ℎ , ℎ and r. Using this equation I will first predict the signs of exposure of airlines to fuel price, interest rate and exchange rate. I do this by defining and approximating the ℎ , the ℎ and their proportional interrelatedness for each of the three market prices. This is followed by a discussion of why and how the third input of the model, (profit margin), affects exposure. Later, I will empirically test proxies of ℎ ’s ℎ’s and as determinants of the exposures.
Equation (2) shows that when a firm’s profit margin () is positive, the sign of exposure can become negative when the market price affected costs (ℎ) are large enough. With respect to fuel price this is most certainly the case for airlines. ℎ is here the costs of purchasing fuel as a percentage of total costs. Since airlines do not have fuel price revenues, ℎ is zero. This makes the fuel price exposure coefficient,, negative, provided that the airlines have a positive profit margin.
Hypothesis 1a: Airlines are negatively residually exposed to the price of jet fuel.
Empirical evidence in support of this hypothesis comes from Loudon (2004) who finds that at the 0.05 significance level both the Australian and New Zealand airline in his sample are negatively exposed to the price of fuel. Carter et. al. (2006) find that their equally weighted portfolio of the US airline industry returns has a fuel price exposure of -.11, significant at the 0.05 level.
rate in Air Berlin, FS 2009, p179). The second reason is that the theoretical effect of changes in the long-term rate on the value of a firm is less distinct than for the short-term rate. An increase in the long-term rate makes new long-term rate debt relatively more expensive, so negative exposure is predicted. However, this effect is mitigated when at the same time the market value of old long-term rate debt decreases. Contrarily, changes in the term rate do not affect the market value of short-term debt, so an increase in short-short-term rate will only negatively affect firm value. With approximately half of the borrowing costs affected by the short term rate and a theoretically larger net result on the value of a firm I prefer to measure short-term interest rate exposure over the long-term interest rate exposure. In terms of equation (2), I expect the interest rate costs as a percentage of total costs (ℎ) to outweigh the effects of interest rate revenues as a percentage of total revenues (ℎ ).
Hypothesis: 1b airlines are negatively residually exposed to the short-term interest rate.
Bartram (2002) finds that in his sample of 490 nonfinancial firms the percentage with significant (0.05 level) short-term interest rate exposure varies between 5.4 and 9.3, depending on the observed period. These are not predominantly negative as was expected.
With respect to the exchange rate, measured as the home currency cost of the foreign currency, equation (2) estimates a firm’s exposure as a function of foreign currency costs as a percentage of total costs (h2) and the foreign currency revenues as a percentage of total revenues (h1). For a pure exporter (h2=0), exposure is positive. A depreciation of the home currency makes its products cheaper in terms of the foreign currency. Along the same line of reasoning, a pure importer (h1=0) benefits from an appreciation of the home currency. An EEA airline with operations in the USA is both an exporter through (among other things) the sale of tickets and an importer due to the US$-denominated costs like fuel. The net effect of exposure is therefore less distinct than with the pure export or pure import firm. I expect that for EEA airlines the costs are larger than the revenues and therefore predict a negative exposure to the euro-to-US$ exchange rate.
Hypothesis 1c: EEA airlines are negatively residually exposed to the euro-to-US$ exchange rate.
This means that EEA airliners are net negatively affected by a depreciation of the euro with respect to the US$. It is possible that the negative effect of a depreciating euro on EEA airlines gives USA airlines a competitive pricing advantage. For example, a purely domestic operating USA airline with no cash flows in euros is not affected directly by the change in the euro-to-US$ exchange rate. However, indirectly it may gain a competitive pricing advantage over EEA airlines operating in the USA market because these have relatively expensive euro denominated costs7. For USA airlines with foreign operations, a depreciating euro also decreases euro revenues in terms of US$s, so the effect is
not one directional. I therefore determine the sign of the US$-to-euro exchange rate exposure empirically. In their sample of 171 Japanese firms, He and Ng (1998) about 25% experience significant (0.05 level) positive exposure and 2% negative.
In equation (2), the profit margin, , should be interpreted as a measure of industry competitiveness. One strategy of firms to deal with exposure is to pass-through market price changes to the customer by changing the price of their products. An example of price pass-through in the airline industry is the fuel surcharge on ticket prices. This strategy will be harder to apply when price competition in an industry is high (Bodnar, Dumas and Marston; 2002). The airline industry is a highly competitive industry and price-pass through is not always possible8. The alternative is to absorb the market price change and accept a lower profit margin. Therefore, profitability measures the relative success of firm to pass-through market price changes. One can imagine, for example, that airlines with effective customer loyalty programs or a larger market share are more successful in passing-though exposure. Equation (2) therefore predicts that, with ℎ and ℎ unchanged, the magnitude of exposure decreases with profitability.
A limitation of the model in equation (2) is the assumption of perfect markets. In the presence of real world market imperfections, financial hedging can increase the value of a firm. Financial hedging is aimed at reducing exposure and therefore its role as a determinant of exposure needs to be examined.
2.3 Exposure and Hedging
A phenomenon in exposure studies is that the measured exposure is not as large or significant as expected. Bartram and Bodnar (2007) call this the exposure puzzle. Models like the one discussed in section 2.2 do not take into account the risk management activities that a firm undertakes as a reaction to their exposure. Bartram and Bodnar (2007) suggest that the exposure puzzle be explained by the fact that exposure is empirically measured net of corporate hedging. A firm with high gross exposure that hedges intensively can show a similarly low net exposure to a firm that had a low gross exposure to begin with.
With corporate hedging a distinction can be made between operational and financial hedging. Operational hedging is the practice of matching foreign currency costs with foreign currency revenues to reduce exchange rate exposure. For example, when operating activities leave a firm with a surplus in a certain currency, this firm can initiate investment opportunities denominated in that currency to
8 10-K AirTran Airways 2007, Item 1: “Subject to market conditions, we may implement fare increases to offset
reduce its exchange rate exposure. To a degree, a firm can also operational hedge itself against commodity prices and interest rate changes. For example, airlines can try to reduce their consumption of fuel and their dependency on debt financing. Financial hedging uses financial contracts to offset substantial gains or losses arising from changes in the market prices. Financial instruments used towards this goal include forwards, futures, options, swaps and collars. Jin and Jorion (2006) empirically confirm the negative relationship between financial hedging and exposure in the oil and gas industry. It would be interesting to see if financial hedging reduces jet fuel, interest rate and exchange rate exposure in the airline industry.
Hypothesis 2: The magnitude of exposure is negatively related to the level of financial hedging in the airline industry.
2.4 Financial Hedging Incentives
A firm can manage its exposure through financial hedging. Whether this is desirable over shareholders managing the risk through portfolio diversification has been a matter of debate. Carter et. al. (2006) find that an USA airline that financially hedges the industry average of 29.4% of next year fuel requirements has a market value (as measured by Tobin’s Q) 10.2% higher than an airline that does not hedge at all. For European airlines the hedge premium increases to 20.6% according to results by Kvello and Stenvik (2009). However, Jin and Jorion (2006) find no evidence that hedging increases market value for oil and gas producing firms. Modigliani and Miller (1958) show that in perfect capital market the financing policy of an investment cannot add value to a firm. Financial hedging can be classified as a financing policy, so for it to add value would require one or more of the perfect market conditions to not hold in practice. The remainder of this section will discuss how certain market imperfections can make a volatile cash flow or firm value costly. When these costs outweigh the costs of hedging it makes economic sense to hedge.
Hypothesis 3: The levels of financial hedging in the airline industry conform to the predictions of the optimal hedging theories.
2.4.1 Financial Distress Costs
2.4.2 Underinvestment Costs
In perfect capital markets the choice of internal versus external financing of investment projects is irrelevant (Modigliani and Miller, 1958). A shortfall in internally generated funds due to changing market prices can then easily be supplemented with external funds. In the real world, however, external financing can become expensive when the marginal costs increase with the amount of externally raised funds (Froot et. al., 1993). When internally generated funds are not sufficient and external funds are expensive a firm may underinvest. Hedging can benefit the firm by matching internally generated funds with available investment projects. Carter et.al (2006) find that the jet fuel hedge premium of airlines is best explained by the underinvestment theory, where airlines with both a high level of hedging and a high level of investment opportunities are valued the highest.
2.4.3 Taxes
Taxes can be an incentive to hedge in two ways. First, when taxes are convex it means that the amount of tax paid increases disproportionate with the amount of taxable income. A firm with a convex tax function would benefit from a smoothed taxable income, which can be achieved with hedging. For example, a firm that is equally likely to earn 200,000 as it is to lose 200,000 euro has an expected tax expense of 40,000 euro when earnings are taxed at 40 percent. When the earnings volatility is reduced through hedging, the expected earnings and taxes become zero. The relationship between tax convexity and hedging is empirically observed by Smith and Stulz (1985). However, there are also characteristics of the tax code that reduce the convexity implied by the previous example. For instance, firms can carry forward losses to shield future earnings against taxations. Graham and Rogers (2002) do not find evidence that convex tax functions are an important determinant of hedging. They argue that the benefits from tax convexity hedging are not as large as the second way in which taxes can be an incentive to hedge; an increased debt capacity. The interest tax reduction, or tax shield, is directly related to the debt capacity. The reduction in income volatility and financial distress probability, which can be achieved with hedging, increases debt capacity (Leland, 1998). This allows firms to capitalize on the extra interest tax reduction.
2.4.4 Agency Costs
Table II gives an overview of the main studies referred to in this thesis with respect to the determinants of hedging.
Table II | Literature Summary of Hedge Level Determinants
The table gives an overview of the determinants identified in the literature for commodity price (CP), interest rate (IR) and exchange rate (ER) hedge levels. The second column displays the significant determinants with their sign (+/-) and the significance level (1/5/10%). The last column lists the characteristics of the study. These include the market risk studied, explained variable (EV), sample period and sample size, data frequency and the model.
Author (year) Determinants Characteristics Bartram, Brown
and Fehle (2009)
CP IR ER ALL Risk [CP, IR, ER]
EV [Derivative usage extent as measured by key-word ranking index] Period [2000-2001]
Sample [739 nonfinancial firms] Frequency [Yearly]
Model [second stage Tobit] Leverage Debt Maturity Dividend paid Quick Ratio Net ER exposure +/1 +/1 +/1 +/1 -/10 +/1 -/1 +/1 +/1 +/1 +/1 -/1 Berkman and Bradbury (1996) Ln[size] Interest Cover Leverage Tax Loss E/P ratio Liquidity Dividend Payout +/1 -/1 +/5 +/5 -/10 -/5 +/10 Risk [CP,IR,ER]
EV [Derivatives fair value/MV Equity]
Period [1994]
Sample [116 New Zealand firms] Frequency[Quarterly]
Model [Tobit] Tufano (1996) Leverage
Managerial stock ownership Managerial option ownership Large outside block percentage Quick Ratio +/1 +/5 -/1 -/1 -/5
Risk [Gold price]
EV [% of 3 year production sold forward]
Period [1991-1993]
Sample [36 NA gold mining firms] Frequency[Quarterly]
Model [One sided Tobit model] Haushalter
(2000)
Debt Ratio Bond rating Marginal tax rate Production costs Tax Indicator Inside Ownership Production location +/1 -/5 +/5 +/5 +/1 -/1 +/5
Risk [Gas and Oil price] EV [% of production hedged] Period [1992-1994]
Sample[100 US oil/gas producers] Frequency [Quarterly Model [Tobit] Kvello and Stenvik (2006) Ln[size] Dividends Debt Ratio
Capital expenditures / Sales
+/1 +/10 -/1 +/10
Risk [Jet fuel price] EV[% of next years’ fuel requirements hedged] Period [2001-2008]
Sample[18 EU and US airlines] Frequency [Yearly] Model [OLS] Carter, Rogers and Simkins (2006) Tobin’s Q Size as Ln[Assets] Leverage as LTdebt-to-assets Credit Rating Fuel pass-through
Executive Shareholdings / Shares Outst. +/1 +/5 -/5 -/1 -/1 +/1
Risk [Jet fuel price] EV[% of next year’s fuel requirements hedged] Period [1992-2003]
Sample [28 North American airlines] Frequency [Yearly]
III. METHODOLOGY
3.1 Estimation of Exposure and Its Determinants
I test the determinants of exposure with a two-step model as is common in exposure studies (See table I). In the first step a four-factor regression, like that of Loudon (2004) and Sadorsky (2001), simultaneously estimates the yearly residual exposure to the fuel price, interest rate and exchange rate.
,= + + + + + , (3) = !(") − ! ("$ ) (4) , is the total weekly return on the stock of firm i in period t. is the total weekly return on the market index for period t. Both are a total return, which means they include dividends and are adjusted for stock splits. The market index variable is included to measure residual exposure. That is exposure after adjusting for market movements. When fuel price, interest rates and exchange rates move with the market index, the estimated exposures would be over or underestimated without the inclusion of a market variable (Jorion, 1990). is the intercept and is the error term. The three gammas, , are the estimated exposure to the jet fuel price (%"), the local short-term interbank interest rate (&) and the home to foreign currency exchange rate (') respectively. All the returns (R) are calculated from their price (P) with equation (4). The exchange rate P for the EEA airlines is the euro cost of 1 US$ and for the USA airliners it is the US$ cost of 1 euro. A negative (positive) exchange rate gamma, , means that the firm is negatively (positively) exposed to a devaluation (appreciation) of the home currency. The stock price and fuel price are both denoted in the airline’s local currency.
The market price returns are assumed to be predominantly unexpected. The expected future spot price may be better approximated by the futures price instead of the current spot price when there is seasonality in the market price, like with agricultural commodity prices (French, 1986). However, this does not apply to the market prices used in equation (3), so the spot price seems appropriate.
I use weekly returns in equation (3) to estimate yearly exposures. The choice of weekly data is a trade-off between having enough observation to estimate yearly exposures and having to deal with biases resulting from high-frequency data (Scholes and Williams, 1977). Dominguez and Tesar (2006) find that the magnitude of exposure may be underestimated when a weekly return horizon is used, but in their study the qualitative results did not change with longer return horizons. I estimate yearly exposures because it allows an airline’s exposure to change over time. That a firm’s exposure is not constant over time is an observation made by Tufano (1998).
Verschoor (2007) find that in their estimations of exchange rate exposure 96% of error term are heteroskedatic. Heteroskedasticity invalidates the test statistic of OLS estimations. I therefore add a GARCH (1,1) specification to equation (3). This model is also used by Muller and Verschoor (2007) and it does not require homoscedasticity of the residuals.
,= (,∗ *ℎ,+ ⁄ -!. ℎ, = /+ ∈,$ + 1ℎ,$ (5) The conditional variance of the residuals is denoted by ℎ, and the white noise error term by (,. Hypothesis 1a, 1b and 1c are tested with a Wilcoxon signed-rank test. When the medians of the measured exposures are negative and statistically different from zero at the 0.05 level, the hypotheses are confirmed. To test whether the median exposures of EEA airlines are significantly different from the median exposures of USA airlines, a Wilcoxon rank-sum test is used. Both the Wilcoxon signed-rank and the Wilcoxon signed-rank-sum test are non-parametric test. I prefer the Wilcoxon tests over a (paired) Student’s t-test because they do not require a normal distribution of the observations.
The second-step regression tests the determinants of the exposures estimated with equation (3). Since I am interested in what factors determine the magnitude of exposure, the absolute value of the estimated exposures as the dependent variable is used.
23,42 = - + 5 ∅7%7,,4 …
79
+ ,4 (6)
where 3,4 is the exposure gamma (3, 3 or 3 ) estimated with equation (3) for firm i in year x. Equation (6) is estimated for each of the three market prices. The next section (3.2) will discuss the factors, %7,,4, that are tested as determinants of exposure. A positive coefficient, ∅7, means that factor : increases the magnitude of exposure.
With annual observations of multiple airlines, the data has a panel structure. The coefficients of the determinants are estimated with a cross-sectional random effects model. This is preferred over a fixed effects model because it produces a more efficient estimation (Brooks, 2008, p500).
3.2 Determinants of Exposure Defined
is extracted directly from risk management section in the annual report of airlines. These are defined in section 3.2.2. The reported hedge level may however be a poor indication of the actual level of hedging. For example, due to data availability constraint they do not distinguish between the instruments (i.e. options, collars, swaps) used for financial hedging. The second approach is approximating the level of hedging by a set of determinants based on the risk management theories. These are discussed in section 3.2.3. This approach is also used by He and Ng (1998) and hinges on the conjuncture that hedging reduces exposure. When a factor should theoretically increase the level of hedging we expect to measure a reduction in exposure. Equation (6) is estimated with both approaches to measuring the level of financial hedging and both with and without the determinants controlling for gross exposure differences.
3.2.1 Gross Exposure Determinants
Recall that the model of equation (2) expresses exposure as a function the market price affected revenues to total revenues (ℎ ), market price affected costs to total costs (ℎ) and profitability (). These determinants do not account for financial hedging activities. I therefore refer to them as the gross exposure determinants throughout the rest of my thesis.
To control for differences in gross fuel price exposure I test the fuel expenses as a percentage of operating expense as a determinant of net fuel price exposure. In accordance with model of equation (2), I expect the magnitude of fuel price exposure to increase with this measure of ℎ.
To control for difference in gross interest rate exposure I test net interest rate expense as a percentage of operating expense as a determinant of net interest rate exposure. Since a lot of interest rate exposure arises from the liability side Bartram (2002) proposes a firm’s leverage as a measure of ℎ. Bartram (2002) did not find a significant relationship between leverage, measured as debt to the market value of equity, and interest rate exposure. He offers the explanation that the measure of leverage may neglect the effect of assets on a firm’s interest rate exposure. In terms of model (2), this means that ℎ is not zero because assets generate interest rate revenues. Using net interest rate expense instead of leverage alleviates this problem because the measure can take a negative value when interest rate revenues out way interest rate expenses.
currency costs (i.e. imports) are hardly ever reported by airlines, while geographical segment reports of revenues do reveal the export ratio.
The final determinant from the model of equation (2) is profitability (). Bartram, Brown and Minton (2010) use the gross profit margin, a measure of operating profitability. In the airline industry the EBITDAR margin is common measure of operating profitability. The EBITDAR margin is the earnings before interest, taxes, depreciation, amortization and rent to total revenues. EBITDAR is preferred over the perhaps more familiar EBITDA margin because airlines can choose to rent aircraft instead of buying them and the exclusion of depreciations only controls for aircraft acquisitions. I expect airlines with a higher EBITDAR margin to exhibit smaller fuel price, interest rate and exchange rate exposure because they are more successful in passing-through exposure to the customer.
3.2.2 Reported Financial Hedge Levels
For jet fuel hedging in the airline industry, Carter et. al. (2006) and Kvello and Stenvik (2006) use the percentage of next year’s estimated fuel requirements hedged. For comparability purposes I will not deviate from this measure. Most annual reports of airlines give this percentage or at least provide data that allow it to be calculated. It is a measure comparable to proxies used in studies of other industries. For example, as a proxy for the gold price hedge level, Tufano (1996) scales the ounces of gold sold forward against the estimated production of the next three years. Appendix I illustrates with annual report extracts how the proxy is calculated when reporting styles differ.
I measure the level of exchange rate hedging as the net notional amount of foreign currency derivatives to operating expense. Recall that I estimate the exposure of EEA and USA airlines to the euro-to-US$ and US$-to-euro exchange rate respectively. Foreign currency derivatives are therefore US$-derivatives for EEA airlines and euro-derivatives for USA airlines. I use the net amount of derivative outstanding. That is for either buying or selling the foreign currency. In the most common case a net buyer of a foreign currency that wishes to hedge this exposure will acquire more contracts to buy the foreign currency forward than sell. With the assumption that airlines do not use currency contracts for speculating the hedge level value will not be negative. Berkman and Bradbury (1996) scale the notional amount of commodity price, interest rate and exchange rate derivatives to firm size. I scale to operating expense because a lot of the US$ dollar denominated expenses like fuel, maintenance, aircraft rent are operational.
3.2.3 Theoretical Financial Hedge Level Proxies
Section 2.4 has discussed in what circumstances it can be attractive for a firm to financially hedge their exposure to market prices. Based on these risk management theories, certain firm characteristics can approximate a firm’s incentive to financially hedge. I use a set comparable to the one used by He and Ng (1998). These firm characteristics are leverage, firm size, dividend payout ratio, quick ratio and the book-to-market value of assets.
As a substitute for financial hedging, a firm may try to reduce the expected costs of financial distress and agency costs by maintaining a larger short-term liquidity (Nance, Smith and Smithson, 1993). By investing more in liquid assets and by paying fewer dividends, stakeholders are assured that funds are available to pay fixed claims. I therefore expect that hedge levels increase with a firm’s quick ratio and reduce with its dividend payout ratio. These measures of liquidity are found to be significant determinants in the study of He and Ng (1998), where exchange rate exposure increases with a firm’s liquidity. That hedging is negatively related to liquidity is empirically observed by Berkman and Bradbury (1996) and Tufano (1996). Opposing evidence comes from Bartram (2002), who finds that the magnitude of interest rate exposure decreases with liquidity, as measured by operating cash flow over total assets. Bartram (2002) also argues that a high liquidity reduces expected financial distress costs. The difference in interpretation is that Bartram (2002) argues that this in itself reduces the magnitude of exposure. He and Ng (1998) argue that it reduces financial hedging and therefore increases exposure. Bartram (2002) also tests other measure of liquidity, including the quick ratio. These measures show the same relationship with respect to exposure, only the results are not statistically significant.
Smith and Stulz (1985) argue that smaller firm have more incentive to hedge because bankruptcy costs as a percentage of market value tend to decrease with firm size (see Warner, 1977). Additionally, Graham and Smith (1998) show that small firms are more likely to have convex tax function. This means that they would have more incentive to hedge. In line with these predictions both the exchange rate exposure of He and Ng (1998) and gold price exposure of Tufano (1998) are found to increase with firm size. Tufano (1998) supplements the explanation that due to the higher trade volume of the stocks of larger firms; their stock prices may incorporate changes in market prices faster. This would also explain the larger exposures. Opposing evidence comes from Nance et. al. (1993) and Berkman and Bradbury (1996) who find evidence that hedging increases with firm size. Berkman and Bradbury (1996) offer the explanation that the indirect costs of financial distress (like damage to brand image) are probably higher than the direct costs of bankruptcy. The impact of disproportionate bankrupt costs becomes more subtle when the indirect financial distress costs are not disproportionate to size. Moreover, larger firms are more likely to have sophisticated risk management departments which increase the intensity of hedging. Also, economies of scale involved with the transaction costs in the derivative market can make hedging more attractive for larger firms. The level of jet fuel hedging in the airline industry seems to conform to these explanations as it is found to be positively related to firm size in the study from Carter et. al. (2006). Like Nance et. al. (1993) I measure a firm’s size as the natural logarithm of the market value of equity plus the book value of debt.
potential as the book value of assets to the market value of assets. Firms with a low book-to-market are expected to have more investment opportunities and thus more incentive to hedge. Supporting evidence comes from He and Ng (1998) who find that exposure increase with the book-to-market value of equity. Kvello and Stenvik (2006) find that the capital expenditures of airlines are positively related to the level of jet fuel hedging, but only at the 0.10 significance level. Empirical evidence opposing the underinvestment theory comes from Tufano (1996) who finds that gold mining firms with more exploration opportunities hedge less.
3.3 Estimation of Determinants of Reported Financial Hedge Level
The determinants of hedge levels are commonly estimated with a Tobit model (See table II). A Tobit model is preferred when the dependent variable is unobservable and censored. The dependent variables have unobservable characteristics because firms can have different approaches to financial hedging. The approaches can differ in the amount of risk covered, the instrument used (i.e. options, collars, swaps), but also in the terms of the instruments. For example, the cap and floor rate can be different for each collar agreement. This means that the true level of financial hedging is unobservable. The percentage of next year’s estimated fuel requirement hedged is zero when the airline does not hedge. The variable can also not be negative, because that would suggest financial instruments are in place to increase risk, which is considered speculating rather than hedging and I assume that none of the airlines held financial derivatives for trading purposes. Because many observations are for this reason exactly zero and none are negative the variable is considered left censored. The nominal amount of exchange rate derivatives to operating expense is also a left censored variable, so for these two measures a one-sided Tobit model is appropriate. The percentage of debt at a fixed rate is my measure of the level of interest rate hedging. For an airline with only floating rate debt the value is zero and for an airline with only fixed rate debt the value is one. I therefore use a two-sided Tobit model for estimating the determinants of the interest rate hedge level. The Tobit model is expressed as
%;<,4∗= - + 5 ∅=%;<=,,4 … =9 + >,4 One-sided: ;,4= %;<,4∗?@ %;<,4∗> 0 -!. Two-sided: ;,4= %;<,4∗?@ 1 > %;<,4∗> 0 and %;<,4= 0 CDℎEF?GE (7)
where %;<,4 is the reported financial hedge level of fuel price, %;<,H, interest rate, %;< ,H, or exchange rate, %;< ,H, for firm i in year x. %;<=,,4 is the set of theoretical financial hedge level proxies as identified in section 2.2.3 and ∅= is the estimated coefficients of %;< proxy . The %;<
(negative) coefficient for the determinants that are supposed to increase (decrease) the level of hedging. - is the intercept and >7 is the error term.
Recall that hypothesis 2 expects a negative relationship between the level of hedging and the magnitude of exposure. This hypothesis 2 can be tested with the combined results of equation (6) and equation (7). The theoretical hedge level proxies are tested as determinants of exposure with equation (6) and as determinants of the reported hedge level with equation (7). An observation will be in line
9 The annual reports of some of the EEA airlines do not specify domestic sales. They only specify Europe as a geographic sales segment (e.g. Aer Lingus Group PLC, AR 2008, p.65). Foreign sales for EEA airlines are therefore calculated as 1 - European sales. For the USA airlines foreign sales means sales outside the USA. 10 Foreign currency is defined as US$ for EEA airlines and as euro for USA airlines.
11 The market value of equity and book value of debt are converted to the US$ with that year’s average weekly home currency-to-US$ exchange rate.
Table III | Definition of Determinants and Theoretical Prediction of Their Relation to the Magnitude of Exposure and Level of Financial Hedging
The table displays the definition of the factors outlined in section 3.2. The columns under |Exposure| give the predicted sign of the determinant’s coefficient in the estimation of 23,42 = - + ∑…79 ∅7%7,,4+ ,4. The theoretical financial hedge level (TFHL) proxies are also tested as the determinants of the reported financial hedge level (RFHL) with %;<,4∗= - + ∑ ∅…=9 =%;<=,,4+ >,4. The predicted signs in these estimations are displayed in the last column.
Determinant Definition |Exposure| RFHL
Gross exposure FP IR ER
Fuel expense Fuel expense / Operating expense + Interest expense Interest / Operating expense + Export ratio Foreign revenues9 / total revenues +/- Profitability EBITDAR/ total revenues - - -
Reported FHL
Fuel price RFHL Fuel hedged / next year’s estimated fuel requirements
- Interest rate RFHL Fixed interest rate debt / Total debt - Exchange rate RFHL Net notional amount of foreign currency
derivatives10 / Operating expense
-
Theoretical FHL
IV. DATA DESCRIPTION
4.1 Sample Selection
A list of firms was extracted from the Amadeus database, which selected all the listed airlines in the EEA and the USA. I used Amadeus for the initial sample selection over DataStream because it allowed the inclusion of formerly listed airlines, thus reducing a survivorship bias. Table IV reports the step-by-step method of sample construction. Of the initial list, the largest 25 EEA airlines and the largest 25 US airlines were selected, based on the last known revenues. Schulz and Mosler (2011) find evidence that less frequently traded stocks distort test statistics. Stocks of smaller firms are traded less, which means that changes in market prices may not be incorporated in the price efficiently. Tufano (1998) gives the same argument for the observation that larger firms have larger exposures. A selection based on size should reduce the distortion to some extent. The sample period is from 2001 to 2009 because very few EEA airlines have online archives of annual reports that extend further back than 2001. Airlines that only operated outside the sample period are excluded. 4 EEA airlines do not have annual reports available due to either discontinuation or a non-English reporting language. The central storage of SEC filings with the Edgar database ensured no US airlines need to be excluded. Furthermore, airlines for which no stock price data is available on DataStream are excluded. Finally, a total of 3 firms are excluded because they had more than 20% of their revenues accruing from operating activities like helicopters services, cargo and road distributions. Due to the different nature of risks involved with these activities, exclusion will increase the homogeneity of the sample. The final sample consists of 15 EEA airlines and 13 US airlines. The sample size is comparable to other airline industry studies. Carter et. al. (2006) have a sample of 28 North American airlines. Kvello and Stenvik (2009) have 12 EEA and 6 US airlines and Loudon (2004) has one Australian and one New-Zealand airline.
Table IV | Sample Construction
The table displays the step-by-step sample selection procedure. The first column states the criterion and the second column shows the number of airlines that are excluded from the sample due to this criterion. The final sample consists of 15 EEA and 13 US airlines.
Criterion Excluded Resulting
Initial AMADEUS result - 76
Largest 25 of each subsample 26 50
Operated within the sample period 2 48
Annual report available 4 44
With the restriction that a firm year has the annual report available from which at least one of the three reported financial hedge levels can be extracted, the final sample consists of 97 EEA and 86 US firm year observations, totaling to 183 firm years.
4.2 Data Sources of Exposure Estimation
Table V displays the descriptive statistics of the independent variables of equation (3), which estimates the yearly residual exposures. The dependent variable in equation (3) is an airline’s weekly stock return. The stock returns are based on the Friday closing prices adjusted for stock splits and dividends, which are collected from DataStream. The prices of the independent variables are also collected from DataStream. This remainder of this section will justify the choice of the proxies for the independent variables.
Jet kerosene is produced around the globe and the price varies depending on the location of purchase.
The most important jet fuel price for each airline’s risk analysis is the one that is sold geographically near the airline’s area of operations since that is where the airplanes are most likely to refuel. Carter et al. (2006) use the Gulf Coast spot jet fuel price for their USA airline sample. My study uses the same for the USA airlines and for the EEA airlines the Amsterdam-Rotterdam-Antwerp jet fuel price is used. With the estimation of the exposures, the price of jet fuel is expressed in the same currency as the stock price. This is done to exclude exchange rate effects from the jet fuel price exposure estimates. For the short-term interest rate I use the 3 month local rate, which is the European Interbank Offered Rate (Euribor) for EEA airlines and the US interbank offered rate for the USA airlines. The exchange rate exposure of USA and EEA airlines is estimated for the US$-to-euro and the euro-to-US$ exchange rate respectively. The majority of exchange rate exposure studies estimate the exposure to a trade-weighted exchange rate (See table 1 from Bartram and Bodnar, 2007). Williamson (2001) points out that when firms are exposed to only a few currencies, the results can lack power when a trade-weighted exchange rate is used. In my sample, the US$ costs of jet fuel, maintenance and aircraft rent alone averages at 30,5% of operating expense for the EEA airlines. This makes them net buyers of US$s. The substantial effect of the US$ on the costs of EEA airlines would be underestimated with a trade-weighted currency index. See section 2.1 for a full discussion of the exchange rate exposure variable.
picking up domestic market effects. An alternative to using a global index is to use an industry index. The problem here is that the market index may explain too much, incorporating the exposure to changes in market prices. For an industry study this may provide very powerful comparative results within the industry. However, since industry specific exposure is partly filtered out, the comparability with studies of different industries is reduced. My study will therefore use market indices that are a compromise between global and industrial. For the US airlines the Morgan Stanley Capital International (MSCI) USA is used. This is a weighted-average of the largest traded equities in the USA. For the EEA airlines the MSCI Europe index is used. This is an index composed of 16 European country market indices. The returns on both indices are adjusted for stock splits and dividends. My choice of market index is supported by Valkova (2009) who finds that the estimations of exposure with the MSCI world and the MSCI sector index have a lower fit, as measured by the R-squared, than with the MSCI country index.
Table V | Independent Variables Used for Estimating Exposure
The table displays the mean and standard deviation of the weekly returns used in ,= + + +
+ + , from 2001 to 2009. This equation estimates the yearly exposures of EEA and USA airlines to the jet fuel price, interest rate and exchange rate. The market index return, , is a variable that controls for market movements. The EEA jet fuel price is in the estimations denominated in the same currency as the airline’s stock price but in this table only the descriptive statistics are given for the euro per metric tonne of jet kerosene. Panel B displays the Pearson’s correlation of the returns. The p-value, P, of a two-tailed t-test is displayed next to the correlation value. The correlations in bold typeface are significant at the 0.05 level.
Panel A – Descriptive statistics of the returns used in estimating yearly exposures
EEA airlines Mean StDev.
MSCI Europe -0.07% 3.00%
Jet Kerosene Barges FOB ARA €/metric tonne 0.11% 4.73%
3 month Euro interbank offered rate -0.41% 1.94%
Euro to 1 US$ -0.09% 1.38%
USA airlines
MSCI USA -0.03% 2.75%
Jet Kerosene US Gulf Pipeline US$cent/gallon 0.21% 6.19%
3 month US interbank offered rate -0.69% 4.07%
US$ to 1 Euro 0.09% 1.38%
Panel B – Pearson’s correlation of the returns
– Market index – Fuel price – Interest rate – Exchange rate EEA USA EEA USA EEA USA EEA USA Cor. P Cor. P Cor. P Cor. P Cor. P Cor. P Cor. P Cor. P
EEA 1.00 USA 0.85 0.00 - 1.00 -
EEA 0.11 USA 0.13 0.02 0.01 0.060.10 0.20 0.03 1.000.71 0.00 - 1.00 -
USA 0.02 EEA 0.09 0.07 0.63 0.030.09 0.59 0.04 0.080.13 0.01 0.07 -0.010.10 0.03 0.89 1.000.48 0.00 - 1.00 -
Panel A in table V shows that, on average, fuel price returns have been positive from 2001 to 2009, with a mean return of 0.11% for the EEA jet fuel price and 0.21% for the USA jet fuel price. Of the three market price risks, the standard deviations of the jet fuel returns were the highest (4.73% and 6.19%). It is also interesting to note that the standard deviations of the USA interest rate (4.07%) and jet fuel price return (6.19%) are larger than their EEA counterparts (1.94% and 4.73% respectively). Panel B in table V shows the correlation values of the independent variables in the 2001 to 2009 period. Of the variables that will be estimated simultaneous the largest significant (0.01 level) correlation value, 0.20, is between the USA fuel price and the US$-to-euro exchange rate. This means that an increase in the jet fuel price tended to be accompanied by a depreciation of the US$ with respect to the euro. Jet fuel price returns were positively related to the returns on the market indices. The significant (0.05 level) correlation values are 0.11 for the EEA and 0.10 for the USA variables. A macroeconomic explanation is that demand for energy products rises with economic prosperity. The returns on the euro-to-US$ exchange rate are positively related to the US market returns (0.13), which means that growth in the US market is accompanied by a depreciation of the US$ with respect to the euro. Significant correlations can result in multicollinearity with the estimation of the exposures. A consequence of multicollinearity is larger standard errors and lower t-statistics of the explanatory variables (Williamson, 2001). This would result in less significant exposure estimates. Cooper and Schindler (2006) suggest that independent variables should be excluded when their correlation is above 0.80. Since the highest observed correlation between independent variables (0.20) is far below this threshold I assume that multicollinearity is not a problem in the estimations.
4.3 Data Sources of the Determinants
Most of the firm data for the determinants is obtained from DataStream. The exceptions are fuel expense, foreign revenues, the rent for EBITDAR margin and the hedge levels which are extracted from the annual reports. The annual reports of EEA airlines are published and archived on their respective websites. The 10-K filings of the US airlines are collected from the Edgar database12. Data from annual reports with a financial year not ending on the 31st of December are corrected with weighted averages. For example, 2007 data from financial years ending on the 31st of March will be one-fourth times the 2006/2007 data plus three-fourth times the 2007/2008 data.
Panel A in table VI displays the descriptive statistics of the gross exposure determinants (see section 2.3.1), the reported financial hedge levels (section 3.2.2) and the theoretical hedge level proxies (section 3.2.3) for the whole sample. All these firm characteristics are tested as determinants of exposure with equation (6). The reported financial hedge levels are further the dependent variable in the estimation of equation (7) with the theoretical hedge level proxies as determinants. Panel B in table VI gives the descriptive statistics of the determinants by subsample.
Table VI | Descriptive Statistics of Determinants
Panel A displays the descriptive statistics of the determinants of exposure for the total sample in the 2001 to 2009 period. The variables are defined in Table III. N is the number of observations per proxy for which the data was available. N(=0) is the number of observation where the value was zero. N(NA), gives the number of years where the hedge level could not be extracted from the annual reports. Panel B gives the descriptive statistic by subsample.
Panel A – Total sample
Mean StDev. Min. Median Max. N N(=0) N(NA)
Gross exposure Fuel expense 0.221 0.085 0.076 0.211 0.453 183 0 0 Interest expense 0.013 0.022 -0.088 0.012 0.096 183 0 0 Export ratio 0.181 0.169 0.000 0.181 0.550 183 66 0 Profitability 0.130 0.090 -0.297 0.136 0.404 183 0 0 Reported FHL Fuel price RFHL 0.463 0.296 0.000 0.450 1.000 172 13 11 Interest rate RFHL 0.481 0.263 0.000 0.506 1.000 179 14 4 Exchange rate RFHL 0.123 0.210 0.000 0.002 0.910 175 85 8 Theoretical FHL ln[Size] 7.934 1.479 3.589 8.302 10.062 183 0 0 leverage 1.479 1.978 0.000 0.889 14.731 183 2 0 Dividend payout 0.042 0.549 -2.685 0.000 6.320 183 134 0 Quick ratio 1.043 0.531 0.311 0.869 3.174 183 0 0 BM assets 1.489 0.561 0.063 1.506 4.569 183 0 0
Panel B – Grouped by subsample
EEA airlines USA airlines
Mean StDev. N N(=0) Mean StDev. N N(=0)
Gross exposure Fuel expense 0.200 0.080 97 0 0.244 0.085 86 0 Interest expense 0.007 0.019 97 0 0.019 0.023 86 0 Export ratio 0.220 0.165 97 21 0.137 0.163 86 45 Profitability 0.130 0.082 97 0 0.131 0.099 86 0 Reported FHL Fuel price RFHL 0.456 0.271 86 10 0.469 0.318 86 3 Interest rate RFHL 0.428 0.278 93 13 0.539 0.232 86 1 Exchange rate RFHL 0.238 0.241 90 7 0.000 0.001 85 78 Theoretical FHL ln[Size] 7.802 1.564 97 0 8.082 1.363 86 0 leverage 0.867 0.814 97 2 2.170 2.584 86 0 Dividend payout 0.073 0.752 97 63 0.006 0.021 86 71 Quick ratio 1.088 0.619 97 0 0.992 0.405 86 0 BM assets 1.550 0.471 97 0 1.421 0.642 86 0
USA. Profitability as measured by the EBITDAR margin is on average about 13% for both EEA and USA airlines.
Reporting on hedging varies across the annual reports. Appendix I contains annual report extracts and the methods of calculating hedge levels for the different reporting styles. When the value is zero, the airline is said to be not hedged for that firm year. The value is marked not available (N.NA) when the annual report mentions the risk but does not offer enough data to calculate the measure. The average level of jet fuel hedging is 46.3% of next year’s fuel consumption. No jet fuel hedging instruments were in place at the end of 13 out of 183 firm years and 11 did not have enough data available. The ratio of fixed interest rate to total debt is 0.481, which means that on average the airlines have more floating rate debt. Of the USA airlines, only 36 out of 86 firm years mention foreign exchange rate exposure in their 10-K reports, 23 with respect to the euro. 11 out of 88 firm years had exchange rate derivatives in place to fix the price to which they could sell the currencies for US$, 7 with respect to the euro. This is the reason that 78 out of 86 USA exchange rate level observations with the value zero. None of the airlines purchased financial derivatives to increase exposure. For example, a net buyer of US$s does not have more contracts to sell US$s than to buy US$s.
Panel B of table IV shows that USA airlines are on average larger (8.082) than EEA airlines (7.802).The average value of leverage is 2.170 for USA airlines. This is more than twice that of EEA airlines (0.867). A manual check of the observations shows that 4 USA airlines have at least one firm year with a leverage value higher than 6. This means that the large difference between the leverage of EEA and USA airlines is not caused by one airline. For the dividend payout ratio, 134 out of 183 observations have the value zero. This means that very few airlines pay dividends. The average quick ratios of 1.088 for EEA airlines and 0.992 for USA airlines suggest that EEA airlines have a higher liquidity. The higher book-to-market value of assets of EEA airlines (1.550) compared to the USA airlines (1.421) suggests that USA airlines have more investment opportunities.
Table VII | Correlation of Independent Variables
This table displays the pairwise Pearson’s correlation of the determinants of exposure. The p-value of a two-tailed t-test is displayed underneath the correlation value. Correlations that are significant at the 0.05 level are displayed in bold typeface. The first four columns are the gross exposure determinant, followed by the reported financial hedge levels (RFHL) and the theoretical financial hedge level proxies.
V. RESULTS
5.1 Exposure
Panel A of table XIII shows the characteristics of the yearly exposure gammas, as estimated with equation (3) for all the airlines in the sample. The average fuel price exposure is -0.204, which means that the value of an airline decreases with approximately 0.204% for every 1% increase in fuel price. Due to the volatility of fuel prices this can have a substantial economic impact. Table V reports an average percentage change of the weekly Gulf Coast jet fuel price of 0.21% over the 2001-2009 period. This change had a standard deviation of 6.19%. A one standard deviation change in fuel price would than change the value of the airline with 1.18%. With a Wilcoxon signed-rank test value of 8.033 the actual median fuel price exposure is significantly different from zero. This confirms hypothesis 1a, which states that airlines are negatively exposed to fuel price, at the 0.001 level. Carter et. al. (2006) whom measured the fuel price exposure of an equally-weighted airline industry index with monthly returns found a slightly smaller exposure coefficient of -0.110. Panel B reports the results separately for the EEA and USA airlines. The median USA fuel price exposure is -0.222, compared to -0.118 for the EEA airlines. The last column in panel B reports the probability value that the median of the two subsamples is the same according to the Wilcoxon rank-sum test. The value of 0.003 indicates that the USA airlines are significantly more exposed to the fuel price than the EEA airlines.
In line with hypothesis 1b, panel B shows that the EEA airlines are negatively exposed to the short-term interest rate, with an average exposure of -0.171 (P-value: 0.031). In contrast, the USA airlines seem to experience positive interest rate exposure with an average of 0.302. The median of 0.036 is significantly different from zero, but only at the 0.10 level. The difference between the two subsamples is highly significant (P-value 0.002). The opposing results from the subsample explain why the mean of the total sample is not significantly positive or negative as the exposures cancel each other out (P-value:0.854). Panel A reports that 30.6% of the short-term interest rate exposure estimates are significant at the 0.05 level. This percentage is higher than the 8,4% observed by Bartram (2002). A possible explanation lies in the GARCH (1,1) specification, which is used less frequently in older exposure studies13. About half of the significant exposures are negative. This is also observed in the literature. Of the two airlines examined by Loudon (2004), one has negative and one has positive interest rate exposure. For Bartram (2002), 5.6% out of the 8.4% were negative.
