Market timing and fuel hedging : U.S. Airline industry

(1)

Market Timing and Fuel Hedging – U.S. Airline Industry

Abstract:

The purpose of this research is to identify if and how CEOs of U.S. airlines perform market timing in their fuel hedging strategies. More specific, the relationship between changes in Brent oil prices and the percentage of expected fuel consumption hedged is analyzed by conducting an empirical research. Furthermore, this paper explains the main motives for hedging and relates these motives to the airline industry in specific. Using a sample of twenty U.S. airlines for the period 2011 – 2015, I found several airline specific factors to have a significant impact on the expected fuel consumption hedged. The effect of changing oil prices on hedging ratios turns out to be negative but insignificant. This thesis does find that larger airlines are more likely to hedge. When operating costs are relatively high, hedging is more effective and therefore the hedge ratio will be higher for airlines with large operating costs. This article also finds evidence for the effect of average age of the aircraft, total revenues and the number of aircrafts leased on the hedge ratio. These effects operate in a negative manner and are found to be statistically significant. Furthermore, the average hedge ratio more than doubled compared to the years before the global financial crisis.

Lara van Zalingen 10400443

Finance and Organization Field: Finance

18 ECTS (combined with internship at KPMG – Financial services)

Supervised by: J.E Ligterink Faculty Economics and Business Amsterdam, 26 – 07 – 2015

(2)

2

1. Introduction

Lately there is a discussion going on about why airlines do not reduce their ticket prices when oil prices decrease. One reason for this may be that most airlines hedge a substantial fraction of their projected fuel consumption to reduce the exposure of a firms’ cash flows to the underlying risk; in this case the fuel price. Fuel costs represent a great portion of total operating expenses for airlines. Resulting from this, airlines can suffer losses when fuel prices rise. Controlling these costs is of major concern for airlines since it stabilizes profit, hence stabilizes the share price. The premium for these stocks of airlines that make use of hedging instruments accounts for 12 to 16 percent, as is shown in Carter et al. (2006). The shortcoming of using this hedging strategy is that airlines cannot benefit from decreasing prices, since they have locked in prices far in advance. I find it very interesting to investigate the difference in fuel hedging strategies between airlines, and the degree of ‘market timing’ resulting from changing oil futures prices on the amount of fuel consumption hedged. American airlines, for example, does not hedge at all. However, Ryan Air hedges ninety percent of its fuel contracts, as can be seen in the 2014 annual reports of both companies.

Carter et al. (2014) concluded in their paper that airlines tend to increase their hedging activity when fuel prices are high and decrease when prices are low. Based on Carters’ conclusions I expect that decreasing oil prices will cause CEOs of airlines try to time the market and decrease their hedge ratio. They do this to be able to benefit from lower expected prices. When oil prices are increasing, CEOs will increase their hedging activity to lock in lower current prices for the future.

My research is in line with the procedure used by Carter et al. (2014), except for the fact that I used slightly different control variables which are in my opinion are a better fit to the airline industry in specific. This will be more thoroughly explained in the research section.

In this paper I conduct an empirical research on the effect of changing oil prices on the hedge ratios of twenty American listed airlines during the period January 1st_{2011 to January 1}st_{2015 in order}

to answer the research question if CEOs of airlines try to time the market and adopt their hedging strategy accordingly. Although focusing on the airline industry raises concerns about the generality of the results, several features of this industry make it well suited for an analysis of risk management policies. First of all, the market is exposed to common risk; the price of oil and gas is highly volatile. Secondly, various methods are available to hedge against this risk. Futures and forwards are traded on NYMEX and forwards and swaps are traded in the over-the-counter market. Third, there is a large dispersion in hedging policies between the airlines (Haushalter, 2000). Most importantly, fuel is the

(4)

4 largest cost driver for almost every airline, which makes the airlines’ cash flows highly sensitive to price fluctuations (Carter et al., 2014).

There has been done extensive research regarding the field of hedging and the rationale behind this. Less research has been done regarding the hedging strategies of the U.S airline industry in specific, which can be explained by the fact that in the past there were no disclosure requirements regarding the expected percentage of fuel hedged (FASB, 1998). Nowadays airlines are obliged to report this percentage in their annual report, which makes it possible to do a deeper investigation regarding the subject. This paper thus contributes to existing literature since it examines whether managers are trying to time the market based on changing oil prices and adopt their hedging strategy accordingly. Moreover, the paper places this in a more recent time series, after the global financial crisis.

The paper proceeds as follows: The basic motivation for this study will be introduced in section 1. In section 2 a literature review will be provided about why firms hedge, what the different hedging strategies are, the link to the current fuel market and an optimal hedging strategy. Section 3 provides a description of the data and the method. The findings will be further explained in section 4, which also provides a summary.

2. Literature

Does hedging makes economic sense?

Modigliani and Miller (1958) stated that under perfect capital markets, hedging is fruitless. This can be explained by classical investment theory, which holds that investors reward stocks based on their overall portfolio return, assuming that investors can replicate hedging themselves possibly at a lower cost than airlines by diversifying their portfolio (Crouhy and Briys, 1993). These findings are supported by the CAPM theory of William Sharpe (1964), which says that by owning a portfolio, investors only pay for reductions in market risk. Under perfect capital markets, investors can thus easily replicate or undo hedging decisions of firms themselves, which makes hedging unnecessary.

However, when markets are imperfect, hedging can alter a firm’s value by influencing its

investment decisions, expected costs of financial distress, or expected taxes (Haushalter, 2000). A lot of research has been done trying to prove whether hedging increases firm value, and the results vary across industries and between the various studies. As is mentioned in the section above, Carter et al. (2006) find a positive ‘hedging premium’ of 12 to 16 percent for hedging airlines due to decreased risk at the firm specific level. Their findings are supported by Morell and Swan (2006), who state that the market will respond to reduced volatility in profits, resulting in a higher stock price for the

(5)

5 corresponding airline. In addition to this, Smith & Stulz (1985) show that when a firms’ debt level is high compared to total assets, and a firm faces bankruptcy costs, hedging is used as an effective instrument to smoothen cash flows. The authors conclude that hedging is more important for firms with greater dependence on external funding since hedging increases a firms’ debt capacity and tax shield. Corresponding to this statement, the authors find that when taxes are a convex function of earnings, it will be more profitable for firms to hedge. To sum up: hedging provides a great source of risk management and therefore reduces financial distress costs, increases debt capacity and tax shields, and therefore diminishes expected taxes.

Reducing risk by hedging

If we assume that hedging increases firm value, this increase will also result from a reduction in risk. Existing literature that aims to find out if hedging reduces risk has shown contradictory results. One study of Hentschel and Kothari (2001) shows that there is little association between hedging and interest rate risk, currency risk and total risk for non-financial firms. They comment that this weak relation is a result of derivative users reducing their risk to a level comparable with non – derivative users. On the other hand, Guay (1999) shows that firms who implemented a derivatives program experience a reduction in total risk, firm specific risk, interest rate risk and exchange rate risk relative to firms who do not use derivatives. In his study Guay examines the effect of various levels of exposure on the firms’ hedging policy.

What drives hedging policies?

According to Haushalter, the variation in hedging policies is associated with several specific firm characteristics. This study examines the differences in hedging policies by questioning 97 oil and gas producers for the years 1992 – 1994. He finds that the fraction of production hedged is positively related to differences in financial leverage, measured as the level of debt to total assets. This thesis aims to find the same positive relationship between the debt ratio and hedging behavior of airlines. Haushalter also finds that the hedging activity is negatively correlated with the basis risk associated with hedging instruments and that economies of scale play a significant part (Haushalter, 2000). Hence, the motives discussed above explain the cross sectional variation in the use of hedging strategies.

(6)

6

2.1 Market Timing

The central topic in this thesis regarding the drivers for hedging is market timing. Faulkender (2005) examines the effect of market timing on making certain hedging decisions and finds this to be an important determinant. This market timing is illustrated by a paper of Beber & Fabbri (2012) where the authors link market expectations and CEO characteristics to the firms’ hedging decisions. They document that managers change the amount of currency derivatives in response to past behavior of the foreign exchange rate. This market timing can be explained by several behavioral biases. First of all, they found evidence that manager characteristics such as overconfidence, young age, short experience and an MBA degree, affect hedging decisions. According to their study these personal characteristics lead to more risk taking by managers of non-financial firms. Analyzing the specific CEO characteristics of the airlines in the sample to reach our own conclusions goes beyond the scope of this paper, but could be an interesting subject for future research.

Another behavioral bias that explains market timing is mental accounting. Managers seem to frame specific amounts of losses and profits in their mind, which can result in selective hedging. This factor is often linked to loss aversion (Beber & Fabbri 2012).

One last factor to take into account regarding market timing is compensation structure. Beber and Fabbri also argue that the degree of speculation in hedging programs is partly dependent on the firms’ remuneration policy. When CEOs are compensated in the form of options, a higher vega – measure of sensitivity to a firms’ stock price volatility - , and a higher delta – sensitivity of the CEO’s compensation to a firms’ stock price – result in a higher CEO compensation. This higher compensation increases the level of risk taking in hedging. Beber and Fabbri’s findings are supported by Tufano (1996), who as well found a significant relationship between derivatives and compensation. On the other hand, Haushalter (2006) argues that changes in the hedging strategies are not connected to the compensation of CEOs.

An important note is to be made when comparing the paper of Beber and Fabbri and this research, since they use currency derivatives as their measure for hedging. However, in this thesis I only investigate the effect on fuel hedging since fuel covers the largest part of operating costs for airlines.

(7)

7

2.2 Analysis of Fuel Hedge strategies among airlines

This section provides a brief description of the hedging strategies of four airlines taken from the sample to illustrate the differences in fuel risk management among airlines, together with a suggestion made by Nascimento & Warren (2008) for a ‘perfect hedge strategy’.

According to a company’s risk profile, the airlines adopt their fuel hedging strategy. Ryan Air, as mentioned before, hedges ninety percent of its expected fuel usage as their CFO does not want the company to bear any risk in this prospect. It’s all part of the strategy of the company. This strategy paid off during the gulf war, where oil prices went sky high. During this period Ryan Air made a lot of profit. But now, when oil prices are low, it seems that not hedging, like US Airways, would have been the right strategy. What strategy pays off in the long term though? It depends not only on the percentage hedged but also on the type of derivatives used.

Avianca Holdings reported in their annual statement that they only want to limit the losses when prices increase and still profit when prices decrease. Therefore they only use call options. This may sound like a good strategy but the premiums that have to be paid for these options are high and drive up the costs of hedging.

One airline that does hedge effectively in times when Oil prices are not highly volatile, is Airfrance – KLM. This company mainly uses collars; a long position in a call option and a short position in a put option with a different exercise price, to hedge fuel price risk. In this way they raise the premium with the written put to pay for the call, and are insured against major spikes and drops in fuel price. An important note to this is that the hedging companies always run basis risk since the kerosene fuel price is only 97 % correlated with Brent oil prices (Hull, 2000). Airfrance -KLM does not use swaps, but since Transavia does, it is still incorporated in the annual report of Airfrance – KLM. The company makes extensive use of four-ways: a combination of a long call option spread and a short put option spread. The airline purchases one call option while at the same time it sells a call that is more out of the money, and does the same for the put. This strategy provides a partial hedge against rising fuel prices while limiting the exposure to declining prices and does not require an upfront cost. Under the current scenario of relatively low oil prices, Airfrance - KLM is adjusting their lower bandwidth accordingly. The company had already lowered the maturity of their derivative contracts from four years to two years, to be able to better react on the market. Carter et al. (2014) mention in their paper that airlines may sometimes increase their hedging in response to a spike in oil prices. They believe that management thinks their exposure to fuel prices had increased and that because of that, investors will favor hedging. When they adopt their hedge ratios accordingly, they act ‘too late’ (Carter et al, 2014). Hedging in a

(8)

8 lower exposure environment would have had positive results on hedging instruments after an oil spike, which might as well be the case for Airfrance – KLM.

An extreme form of hedging is performed by Delta Airlines who bought an oil refinery in 2012 to be able to hedge against volatile oil prices. According to many specialists, this was a mistake. It had cost 420 million dollars to build the refinery and it is not generating any profits. Instead, in 2014 they suffered a loss of 2.0 billion due to fuel hedges as is mentioned in their annual statement.

Another way to hedge against price fluctuation is the use of future contracts, which are standardized and exchange traded. It gives the owner the right to buy the underlying at a fixed price at a specific time in the future (Morell and Swan, 2006). According to Nascimento & Warren (2008) the most commonly used hedging strategy considering these future contracts is the dynamic minimum variance optimal hedge ratio (OHR).

(1)

In formula (1), the numerator stands for the covariance between the future used (oil) and the spot price of fuel. Since there are no exchange-traded options for jet fuel traded in the US, airlines buy futures commodities with a high correlation with fuel prices, such as crude oil in this case (97%). The denominator is the variance in the price of the future (Hull, 2000). This method minimizes the deviation between jet fuel and the hedge and it does not take into account other factors such as the expected price or the volatility of jet fuel. Due to this fact the assumption is made that this hedge ratio stays the same during the period, this method is also called ‘hedge and forget’ (Nascimento &Warren, 2008). Since managers are still unable to predict oil prices, this method seems logical. However, as is explained in earlier sections, managers try to time the market and during the year, they adopt their hedge ratios according to market developments.

(9)

9

2.3 Market developments

In 2014, the oil energy market has declined significantly and primarily in the last 6 months of 2014. This decline is related to several economic and geopolitical aspects, like the decline of economic growth of China, Japan and Europe resulting in a decrease of demand; the United States of America import less oil due to their freckling production. If the supply and demand are not consistent with each other then the OPEC normally decreases the production of crude oil to stabilize the price developments.

However, Saudi Arabia (the biggest supplier) did not change their production and has reduced their prices for Asian countries. Lower prices have a positive effect on Saudi Arabia’s market share, as freckling in Texas has high production expenses and lead to high losses for those companies involved (when Crude oil prices are low). The effect of this decline in oil prices on the airlines’ hedging strategies will be discussed later in the paper. The graphic on next page shows the trend in Brent Oil prices compared to 2013.

(10)

10

2.4 Hypothesis

Based on the theoretical framework derived above, I derive the following hypothesis:

H0: There is no relation between the change in Brent Oil prices and the hedge ratio of Airlines.

H1: There is a significant relationship between the change in Brent Oil prices and hedge ratio of Airlines.

3. Research Design

To answer the research question if managers of U.S. Airlines perform market timing, an empirical analysis is constructed using firm specific data found on Compustat, historical Brent oil prices from Bloomberg and the hedging percentages of the airlines. I found most of these hedging percentages by conducting a keyword search in the 10-K reports filed for fiscal years between 2010 and 2014, since the firms report the expected percentage of fuel hedged for the coming year in the annual report of their current year at December, 31. I also hand collected total assets for the years 2011 until 2015 using the same method. In this paper I investigated twenty different airlines, of which I used 5 years, all together giving me 100 observations. The choice to the specific sample period responds to the aim of having homogenous accounting standards of hedging disclosure and the need to rely on hand collected data. While a longer sample is always preferable, the 5-year panel used can be seen as a good choice since it eliminates the effect of the financial crisis on the results. Although in a recent survey, Bodnar, Graham Harvey and Marston (2011) document that the financial crisis of 2008 had no significant impact on firm derivative usage, it was most certainly a driver in the volatility in oil prices, which is why I exclude these years from the panel.

I investigate whether the change in Brent oil price has a significant effect on the percentage of fuel consumption hedged using the multiple regression function in STATA. I tried to mimic the results of Carter et al (2014), by using the same independent variables to measure exposure, except for one. In their article they created one additional exposure variable: the average of the volatility and direction of fuel price changes. In my belief this extra variable can lead to collinearity and will result in a bias, which is why I am not including it. In addition to this, Carter et al. (2014) used control variables as capital expenditures and Tobin’s Q, but I chose to include more airline specific control variables to better reflect the specific airline characteristics. These variables will be further explained in next section.

(11)

11

3.1 Variables:

A. Dependent Variable: Fraction of Expected Next Year’s Fuel Requirements Hedged

This percentage will be a fraction of (expected) total volumes consumed for the coming year, calculated for twenty American listed companies over a period of five years (2011 – 2015), displayed as a percentage. I do not build in long-term hedging in this analysis, since most airlines are not including any hedging beyond a one-year time horizon. Stata will omit 20 observations since the data necessary for the control variables is not yet publicly available for the year 2015.

The years chosen for the sample are relevant because of the latest drop in oil and gas prices, which is the main reason to conduct this research. When the airline is not conclusive about this percentage in their annual report, I eliminated the airline and took another one. I understand that using this method may result in selection bias and I tried to avoid this. When the airline does not report this percentage but does report number of barrels hedged, it is straightforward to calculate this ratio. For consumption considered to be hedged, either the price must be fixed or is insured from dropping below a fixed level.

B. Independent variables: Exposure proxies

The change in Brent oil prices as reported in the Bloomberg database from January 1st_{, 2011 until April}

1st, 2015 is used as an independent variable. By taking only the last years we exclude the major effect of the financial crisis. The average oil price per year is calculated using excel. From this average yearly oil price I calculated the percentage change per year, compared to the year before. I used this factor as an independent variable, since my expectations are that the change in oil price is underlying in what drives the amount of fuel consumption hedged. I only took the yearly average oil price instead of daily prices, since the hedge ratios are also calculated on a yearly basis. Using daily oil prices would result in a bias. The change in oil prices in the year 2011 is calculated using the available data from 2010 to compare with. I included another exposure variable; Change Q3, to measure the difference between the average price of the first two quarters and the average price of the third quarter of that year to check if this percentage has a significant impact on a managers decision to hedge. This variable is included because we expect managers to anticipate on next years’ fuel hedging strategy during the period around September, before the annual report has been established. For this reason I included the variable ChangeQ3 in the second regression to compare with. Again, for the year 2015 this factor is not yet available.

(12)

12 As a control variable the volatility of Brent oil is used. This variable is calculated by taking the standard deviation of the daily percentage change in Brent oil. My expectations are that this coefficient will be positive, since a higher volatility induces more risk. Haushalter finds a positive relationship between financial leverage measured as total debt to total assets and hedging. For this reason, I also included the factor Debt/Assets (DA) in the model as a control variable.

Other airline specific control variables included are (the logarithm of) the age of the aircraft, fuel consumed in gallons, total aircrafts in service, operating expenses per Available Seat Mile (ASM), passenger revenue, number of aircrafts leased and total revenue per ASM. My expectation is that the older an airplane is, the more fuel it consumes. Carter et al. (2006) show a positive but insignificant relation between capital expenditures and hedging, therefore this factor is not included in the model. In their paper they do find a positive relation between size and hedging and according to them this is related to economies of scale influencing hedge decisions. Based on their findings I included this factor in my model as ln_Total Assets. Larger firms are expected to be better diversified and have more cash available to handle potential losses but on the other hand are also large enough to have a risk management department. These variables are tested by doing a multiple regression in STATA.

Table I provides an overview of the variables used, followed by a correlation matrix in table II.

Table I; Summary statistics

Variable Obs Mean Std. Dev. Min Max

DA 75 13.4669 79.2464 0.0000 508.8714

Hedge 96 0.3077 0.2939 0.0000 1.3650

Total Aircrafts in Service 73 436.0137 361.7784 37.0000 1285.0000

Aircraft Leased 65 186.8000 197.7903 0.0000 879.0000

Total Rev. p ASM 74 14.7550 3.6352 8.1100 26.1800

Volatility Oil price 100 0.0165 0.0062 0.0111 0.0279

Change in Av. Oil Price 100 -0.0230 0.2558 -0.3998 0.3976

ChangeQ3 80 -0.0624 0.1056 -0.2429 0.0172 Dummy 100 0.2000 0.4020 0.0000 1.0000 ln_Total Assets 79 22.3663 2.0415 14.5087 24.7145 ln_Age Aircraft 78 2.2200 0.4320 1.3350 3.1303 ln_Fuelcons (gln) 58 13.2401 1.2692 10.8935 15.2815 ln_Operating Expenses 58 16.2816 0.2495 15.7324 16.6777 ln_Pass. Revenue 74 22.5958 1.2129 20.0597 24.3375

(13)

13 Table II: Correlation Coefficients

Hedge ln Total Assets ln Age Aircraft DA Fuel Consumption (gln) total Aircrafts in Service Operating Expenses Passenger Revenue Aircraft Leased Total Revenue Change

Oil Price ChangeQ3

Volatity Oil Price Hedge 1 ln_Total Assets 0.2814 1 sig level 0.0126 ln_AgeAircraft -0.1485 0.2526 1 sig level 0.2005 0.0267 DA -0.1050 -0.6001 -0.0489 1 sig level 0.3731 0.0000 0.6788 Fuel Cons. 0.4143 0.5864 0.4080 -0.0485 1 sig level 0.0014 0.0000 0.0016 0.7176

Total Aircrafts in Service 0.1640 0.5773 0.3722 -0.0826 0.8498 1

sig level 0.1656 0.0000 0.0013 0.4873 0.0000 Operating Expenses -0.1536 -0.1895 0.0428 0.0985 0.2411 -0.0771 1 sig level 0.1795 0.0944 0.7100 0.4007 0.0683 0.5167 Passenger Rev. 0.3615 0.5237 0.4064 -0.0703 0.9954 0.8383 0.0137 1 sig level 0.0011 0.0000 0.0002 0.5488 0.0000 0.0000 0.9042 Aircraft Leased -0.0281 0.3886 0.2727 -0.0214 0.5628 0.8324 0.0072 0.5478 1 sig level 0.8244 0.0014 0.0293 0.8658 0.0000 0.0000 0.9546 0.0000

Total Rev. ASM 0.2217 0.2352 0.2715 -0.0154 0.3530 0.1997 -0.2532 0.5361 0.0212 1

sig level 0.0595 0.0437 0.0202 0.8963 0.0066 0.0904 0.0295 0.0000 0.8671

Change Oil price -0.0176 0.1064 -0.0835 0.1218 -0.0375 0.0036 0.0604 0.0717 0.0004 0.0578 1

sig level 0.8786 0.3508 0.4672 0.2980 0.7798 0.9756 0.5943 0.5276 0.9973 0.6249

ChangeQ3 -0.0212 0.1469 -0.0757 0.0716 -0.0379 0.0053 0.0237 0.0666 -0.0138 0.0642 0.9656 1

sig level 0.8537 0.1965 0.5102 0.5413 0.7778 0.9645 0.8344 0.5573 0.9133 0.5869 0.0000

Volatility Oil Price 0.0150

-0.1565 0.0728 -0.0648 0.0364 -0.0097 0.0027 -0.0562 0.0153 -0.0615 -0.8944 -0.9725 1

(14)

14

3.2 Models:

To test the effect of changing oil prices on an airlines percentage of next year’s fuel consumption hedged (hedge ratio), I estimated the following model using Ordinary Least Squares in STATA with robust standard errors; 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑖𝑖,𝑡𝑡= 𝛽𝛽0 + 𝛽𝛽1 ∗ ln�𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟𝑇𝑇 𝐴𝐴𝐴𝐴𝐴𝐴𝐻𝐻𝑟𝑟𝐴𝐴𝑖𝑖,𝑡𝑡� + 𝛽𝛽2 ∗ ln�𝐴𝐴𝐻𝐻𝐻𝐻 𝐴𝐴𝑟𝑟𝑟𝑟𝐴𝐴𝑟𝑟𝑟𝑟𝐴𝐴𝑟𝑟𝑖𝑖,𝑡𝑡� + 𝛽𝛽3 ∗_{𝐴𝐴𝐴𝐴𝐴𝐴𝐻𝐻𝑟𝑟𝐴𝐴}𝐷𝐷𝐻𝐻𝐷𝐷𝑟𝑟𝑖𝑖,𝑡𝑡 𝑖𝑖,𝑡𝑡+ 𝛽𝛽4 ∗ ln (𝐹𝐹𝐹𝐹𝐻𝐻𝑇𝑇𝐹𝐹𝑟𝑟𝐹𝐹𝐴𝐴𝑖𝑖,𝑡𝑡) + 𝛽𝛽5 ∗ 𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟𝑇𝑇 𝐴𝐴𝑟𝑟𝑟𝑟𝐴𝐴𝑟𝑟𝑟𝑟𝐴𝐴𝑟𝑟𝐴𝐴 𝑟𝑟𝐹𝐹 𝑆𝑆𝐻𝐻𝑟𝑟𝑆𝑆𝑟𝑟𝐴𝐴𝐻𝐻𝑖𝑖,𝑡𝑡+ 𝛽𝛽6 ∗ ln (𝑂𝑂𝑂𝑂𝐻𝐻𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝐹𝐹𝐻𝐻 𝐸𝐸𝐸𝐸𝑂𝑂𝑖𝑖,𝑡𝑡) + 𝛽𝛽7 ∗ ln (𝑃𝑃𝑟𝑟𝐴𝐴𝐴𝐴𝐻𝐻𝐹𝐹𝐻𝐻𝐻𝐻𝑟𝑟 𝑅𝑅𝐻𝐻𝑆𝑆𝑖𝑖,𝑡𝑡) + 𝛽𝛽8 ∗ 𝐴𝐴𝑟𝑟𝑟𝑟𝐴𝐴𝑟𝑟𝑟𝑟𝐴𝐴𝑟𝑟𝐴𝐴 𝐿𝐿𝐻𝐻𝑟𝑟𝐴𝐴𝐻𝐻𝐻𝐻𝑖𝑖,𝑡𝑡+ 𝛽𝛽9 ∗ 𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟𝑇𝑇 𝑅𝑅𝐻𝐻𝑆𝑆. 𝐴𝐴𝑆𝑆𝑀𝑀𝑖𝑖,𝑡𝑡+ 𝛽𝛽10 ∗ Δ𝑂𝑂𝑟𝑟𝑇𝑇 𝑃𝑃𝑟𝑟𝑟𝑟𝐴𝐴𝐻𝐻𝑡𝑡+ 𝛽𝛽11 ∗ (𝜎𝜎𝑂𝑂𝑟𝑟𝑇𝑇 𝑃𝑃𝑟𝑟𝐴𝐴𝐻𝐻𝑡𝑡) + 𝜀𝜀

In a second model I replaced the variable Δ𝑂𝑂𝑟𝑟𝑇𝑇𝑃𝑃𝑟𝑟𝑟𝑟𝐴𝐴𝐻𝐻_𝑡𝑡 for the variable Change Q3_𝑡𝑡. I did the same for the models thereafter since the rationale behind the variable ChangeQ3 seems more ligit than Δ𝑂𝑂𝑟𝑟𝑇𝑇 𝑃𝑃𝑟𝑟𝑟𝑟𝐴𝐴𝐻𝐻_𝑡𝑡, because it better incorporates the effect of market timing by taking into account the approximate date when managers ‘time the market’. The variable aircrafts owned is omitted because of multicollinearity with the factor total aircrafts in service since this last variable sums up aircrafts leased and aircrafts owned.

In models (3) and (4) I omitted the variable 𝐹𝐹𝐹𝐹𝐻𝐻𝑇𝑇𝐹𝐹𝑟𝑟𝐹𝐹𝐴𝐴_{𝑖𝑖,𝑡𝑡}, in order to solve a multicollinearity problem and to increase the sample size. As is shown in the correlation matrix in table IV the correlation between Fuel Consumption and the variables Total Aircrafts in Service and Passenger Revenue is rather high. Apart from increasing the samples accuracy by eliminating the variable with the least observations, this third model thus also reduces the effect of multicollinearity by eliminating the highly correlated variable Fuel Consumption (gln). (table II).

To be able to better predict the degree of market timing, I added a dummy in model 4. This dummy equals one if the average oil price for year t is above the average of year t-1. The results of the regressions are presented in table III, followed by an analysis of the results in part 3.3

(15)

15 Table III. Regression Results – Analysis of Hedging Policy

(1) (2) (3) (4) Independent Variables Intercept -5.4580 -5.8020 -6.681* -6.594* (3.4060) (3.7360) (3.5940) (3.6900) DA 0.0001 0.0002 0.0004 0.0004 (0.0005) (0.0005) (0.0004) (0.0004)

Total Aircraft in Service 0.0004 0.0004 0.0003 0.0003

(0.0004) (0.0004) (0.0003) (0.0003)

Aircraft Leased -0.0005 -0.0005 -0.0009** 0.0008**

(0.0006) (0.0006) (0.0004) (0.0004)

Total Revenue pASM -0.0200 -0.0211 -0.0530** -0.0532**

(0.0241) (0.0246) (0.0242) (0.0244)

Volatility Oil price -15.8400 1.7650 1.9250 -6.1980

(28.0100) (11.1300) (11.7800) (25.7200)

Change in av. Oil price 0.2510

(0.3710) Change Q3 -0.0965 -0.1960 -0.1690 (0.5100) (0.4820) (0.4970) ln_Total Assets 0.0255 0.0302 0.0477* 0.0465* (0.0257) (0.0313) (0.0253) (0.0258) ln_Age Aircraft -0.173* -0.176** -0.223** -0.223** (0.0860) (0.0833) (0.0833) (0.0849) ln_Fuel cons(gln) 0.0839 0.0826 (0.1210) (0.1260) ln_OperatingExpenses 0.433** 0.439** 0.371* 0.370* (0.1660) (0.1790) (0.2100) (0.2150) ln_Pass. Revenue -0.0995 -0.1020 0.0476 0.0500 (0.1080) (0.1100) (0.0690) (0.0711) Dummy 0.0565 (0.1480) Observations 51 51 64 64 R-squared 0.2860 0.2810 0.3560 0.3570

Dependent var: Fraction of Consumption Hedged Robust standard errors in parentheses

(16)

16

3.3 Analysis

The results of the regressions in Table III provide some interesting findings. First of all, in table I we find that the average hedge ratio of the sample is now 30.77 percent. This is more than twice as high as the average hedge ratio found by Carter: 14 percent. (Carter et al., 2014). Moreover, the first model in table II shows a positive but insignificant relationship between the Change in av. Oil Price and the dependent variable; the hedge ratio. Apart from the insignificance, the direction of this coefficient is in line with the central hypothesis in this paper that when oil prices are decreasing, CEOs of airlines will try to time the market and decrease their hedge ratio to be able to benefit from lower expected prices and when oil prices are increasing, they will increase their hedging activity to lock in lower current prices for the future. The direction of this coefficient is in line with the results found by Carter et al. (2006), who concluded in their paper that airlines tend to increase their hedging activity when fuel prices are high and decrease when low. Unfortunately the interpretation of this coefficient is rather difficult, since it is not significant at the 5 or 10 percent confidence level.

The same positive but insignificant relationship holds for the debt ratio (DA), which is in line with

the results found by Haushalter (2000), as described in earlier sections. In addition to this, the same positive effect on hedging behavior is found for Total Aircrafts in Service, ln_Total Assets and ln_Fuel Consumed. Other variables show a negative, insignificant relationship with hedging; Volatility of Oil prices, ln_Passenger Revenue, Aircrafts Leased and Total Revenue. Surprisingly though, these last two variables become significant at the 5 percent level in model 3 and 4, just like the variable Total Assets. This sudden significance can be explained by the elimination of one highly correlated variable in model 3 and 4: Fuel Cons (gln.), and a higher accuracy of the results by an increased number of observations (64).

The variable ln_Age Aircraft does show a significant coefficient on hedging at the 5 percent level. This variable is operating in a negative manner, meaning that when the average fleet of an airline gets older, the percentage of fuel hedged is declining. This finding is supported by analyzing the existing literature: as stated by Morell and Swan (2006), airlines that exist for a longer time period, are better able to hedge their future fuel needs since most newer carriers need their credit to finance high growth rates. An additive assumption regarding this matter has to be made that an older average fleet age means an older company existence.

Another significant relation is found in the variable Total Revenue per ASM, which shows a negative significant relationship at the 10 percent confidence level. Again, this finding is only applicable for models 3 and 4. This result indicates that the higher the revenues of an airline, the less it hedges. This

(17)

17 finding is supported by Zea (2004), who explains in his paper that airlines with higher cash flows are better capable of managing the risk of fluctuating oil prices, hence reducing the need to hedge.

On the other hand, the variables I hoped to find evidence for; Volatility Oil Price, Change Q3 and Change in av. oil Price did not turn out to be significant at any near level which can be explained by several potential biases.

Potential Biases:

First of all, the number of observations in the regressions is limited to 51. I hand collected 100 observations for the hedging ratios, but since I cannot use last years’ results in the sample I am therefore forced to eliminate the last 20 observations. The data available on Compustat was very limited as well. This is why STATA removed missing observations from the sample, reducing it to a number of 54.

In order to increase the accuracy of my results, I ran a third and fourth regression eliminating the variable with the least observations; Fuel Consumption gln. (Table I). This model is resulting in a higher number of observations (64), a higher R squared and three additional significant variables (Table III). These results can be explained by the elimination of a multicollinearity problem as well as an increased sample size.

Nevertheless, there can still be a bias in the form of multicollinearity since Aircrafts in Service is highly correlated with Passenger Revenue and Aircraft leased. This makes sense because the more aircrafts an airline leases, obviously the more it has in service and the higher the number of aircrafts in service, the higher the revenues of the airline will most likely be.

Another factor to take into account when analyzing these models is that there might be an omitted variable bias. A variable about the tax structure of airlines could have helped to better explain the model. This rationale is supported by the prevailing literature. Firms with lower marginal tax rates in the current period are more likely to face non-constant tax rates in the future and, hence, are expected to hedge more extensively (Haushalter, 2000). As mentioned in the literature review, the paper from Smith and Stulz (1985) show that if taxes are a convex function of earnings, it will be more profitable for firms to hedge. Analyzing the effect of tax structure on a firms’ hedge ratio might be an interesting suggestion for a follow-up research. Credit rating is another possible omitted variable to take into account, but since this specific data was not available for the selected sample, I took the debt to assets ratio instead.

The external validity of the selected sample is to be questioned. This research is limited to a small number of U.S. airlines that could be found on Compustat, which may result in selection bias. Including more airlines from different continents will increase the generality of the results.

Another important consideration to be made is about internal validity. As mentioned in the article of Beber and Fabbri (2012), there are many important social factors that have a causal relationship with

(18)

18 hedging. In their paper they investigate the effect of CEO characteristics as young age, an MBA degree, overconfidence and short experience on a companies’ hedging strategy. Other factors may determine a companies’ risk profile as well. For example investing climate, company morale, expectations about a companies’ future etc. These factors are hard to measure but can play an important role in determining the risk profile and the degree of market timing in their hedge ratio. This requires a deeper investigation that can be interesting for a follow – up research as well.

Some airlines are flying in behalf of other airlines or group companies and they make use of a ‘fuel pass through agreement’. Carter et al.(2006) already described this in their paper where they assumed that a fuel pass through agreement provides an alternative fuel price risk strategy resulting in a negative relation with hedging. This factor is not included in this model and could therefore harm internal validity. The most important pitfall in this paper is the fact that the hedge ratios are reported on a yearly basis. Airlines do not report when and how exactly they change their derivative portfolio. During the year, airlines adapt their risk profile according to the market. At Airfrance-KLM, The first quarter is hedged for a large part, and the last quarter only to a relatively small extent as can be seen in table 2 of the Appendix. This percentage provides us only with an indication and during the year the portfolio changes. If we could obtain these exact data on a monthly or even daily basis and could match them with the daily oil prices, it would be more realizable to find a significant relation. This can only be done with access to highly protected databases, which is out of the scope of this thesis.

Finally, some airlines used WTI (West Texas Intermediate) as a cross hedge for their fuel consumption. WTI moves roughly the same as Brent, but they do still slightly differ. This difference in basis risk can also result in a bias.

4

Conclusions

Carter et al. (2014) found that the U.S. airlines hedged 14 percent of their next year’s fuel requirements between 1994 and 2008. As shown in table I, this paper finds an average hedge ratio of 30,77 percent of (almost) the same sample of U.S. airlines in the time after the global financial crisis (2011-2015). This last year is included in this percentage since the projected 2015 fuel hedging percentages were given in the 2014 financial statements. We can conclude that airlines currently hedge more of their expected fuel requirements compared to the area before the global financial crisis. Carter et al. also concluded that 30.4 percent of their sample had a hedge percentage greater than zero. Our sample shows that on average, 75 percent of the airlines have a hedge ratio of higher than zero. A significant increase when compared to the years before 2008.

(19)

19 The results of Beber and Fabbri (2012) mention an ‘active view’ on derivative holdings due to manager characteristics as overconfidence, young age, short experience and an MBA degree. These personal characteristics lead to more risk taking by managers of non-financial firms according to their study.

In the airline industry the CEO of Ryanair; Michael O’Leary is notorious for his attitude, according to The Guardian:

:’Ryanair boss Michael O'Leary appears to be the antithesis of the well-rounded and well-loved image that most corporate chiefs are desperate to cultivate. He is fiery, belligerent and antagonistic to competitors, unions, staff and even customers. He is seriously foul-mouthed, outspoken, and appears not to give a second thought about damaging his profile, or the company's.’ (The guardian, Rene Carayol, 7 November 2014.)

Despite this seemingly overconfident CEO, Ryanair is the most conservative company in this survey when it comes to hedging. Beber & Fabbri also mention loss aversion as a reason why managers adopt their hedging policy. If derivatives generate a loss that offsets the profits on the underlying asset (in this case Brent oil), the manager may feel a sense of regret and adopt their strategy accordingly for the following year. Again, this does not affect O’Leary’s policy as the company is pursuing its hedging strategy, even though they are still making huge losses resulting from it. This does not take away the fact that personal characteristics may affect the risk management decisions of CEOs of other companies, since this is only one extreme example to illustrate. Another factor to take into account when comparing these papers is that Beber and Fabbri use currency derivatives, which is used for the same purpose; to hedge risk. However, since fuel expenses account for more than 35 percent of operating costs for airlines (Carter et al. 2006), the importance for airlines to hedge fuel risk is higher than the importance to hedge currency risk, since this is a larger portion of operating expenses.

Unfortunately I cannot draw a conclusion with respect to the proposed hypothesis that the change in oil price has a significant effect on airlines’ hedge ratios. This would require a deeper investigation. Nonetheless, I can conclude that the average age of an airline’s fleet has a negative impact on an airline’s hedge ratio, as well as the airlines revenues. On the other hand Aircrafts leased, operating expenses and size operate in a positive manner with regard to hedging.

I do believe that finding the results originally hoped for can contribute to the effectiveness of hedging strategies in the airline industry. Adding additional insights on psychology could help explain cross sectional variation in market timing better.

(20)

20

References:

Beber, A., & Fabbri, D.(2012) Who times the Foreign Exchange Market? Corporate speculation and CEO Characteristics. Cass Business School.

Bodnar, J., Gordon, M., Graham, J., Campbell, R., Marston, H., & Marston, R,. (1998). Wharton Survey of Financial Risk management by U.S. Non-Financial Firms, Wharton School, University of Pennsylvania.

Carter, D., Rogers, D. & Simkins, B. (2006). ‘Does hedging affect firm value? evidence from the us airline industry’, Financial Management 35(1), 53–86. 1

Carter, D., Treanor S., Rogers, D., Simkins, B. (2014). ‘Exposure, Hedging, and Value: New Evidence from the US. Airline Industry’. International Review of Financial Analysis 34, pp. 200-211.

Crouhy, M., & Briys, E. (1993). Playing against (with?) the devil: managing risks for better corporate return. European Management Journal, 11(3), pp. 304-312

Haushalter, D. (2000). Financing Policy, Basis Risk, and Corporate Hedging; Evidence from Oil and Gas Producers. The Journal of Finance

Hausman, D. Does Operational and Financial Hedging Reduce Exposure? Evidence from the U.S. Airline Industry. The Financial Review.

Hentschel, L. & Kothari, S.P. (2001). Are corporations reducing or taking risks with derivatives? Journal of Financial and Quantitative Analysis, 36, pp. 93- 118.

Hull, J. (2000). Options, Futures, and Other Derivatives, Prentice Hall. 4 pp. 45 – 73

Faulkender, M., (2005) Hedging or Market Timing? Selecting the Interest Rate Exposure of Corporate Debt. Journal of Finance. Vol 60, No. 2.

Financial Accounting Standards Board. (1998). Financial Interpretation No. 44: Accounting for certain transactions involving stock compensation: an interpretation of APB Opinion No. 25.

Lee, M. C. & Hung, J. C. (2007). Hedging for multi-period downside risk in the presence of jump dynamics and conditional heteroscedasticity. Applied Economics 1(1), 1–10. 5

Modigliani & Miller the cost of capital, corporation finance, and the theory of investment. The

American economic review. (1965)

Morrell, P. & Swan, W. (2006), ‘Airline jet fuel hedging: Theory and practice’, Transport Reviews 26(6), 713–730. 1

Nascimento, J. M. & Powell, W. B. (2008), An Optimal Solution to a General Dynamic Jet Fuel Hedging Problem. Princeton University

Smith, C. & Stulz, R. (1985). The determinants of a firms’ hedging policies’, Journal of Financial and Quantitative Analysis 20, pp. 391-405

(21)

21 Sharpe, W. F. (1964). Capital Asset Prices: a theory of market equilibrium under conditions of risk.

Journal of Finance, 19(3), pp. 425 – 442

Tufano, P. (1996); Who manages risk? An empirical examination of Risk management Practices in the Gold Mining Industry’, Journal of Finance 51, 1097 – 1137.

Zea, M. (2004). Is airline industry risk unmanageable?, Technical report, Mercer on Travel and Transport. 1 pp. 312- 333

(22)

22

Appendix

Disclosure example US Airways:

Market timing and fuel hedging : U.S. Airline industry