The Impact of Changes in Fuel Price, Interest Rate and Foreign

Our study starts with an analysis of the extent to which airline companies’ risk exposures have an impact on investors’ ability to forecast earnings. If investors have imperfect information about firms’ risk exposures, we expect that earnings

24 Since we cannot obtain the effective trade-weighted exchange rate for every country in our sample from Datastream, we replace it with this one.

announcements contain additional information about the impact of recent oil price, interest rate and foreign exchange rate shocks on stock performance. We also expect that the magnitude of market reaction to earnings announcements is higher when there are large shocks to oil price, interest rates and foreign exchange rates, and for firms with greater exposures to these variables.

We use the three-day window (-1, 0, 1) around the yearly earnings announcement of a company to measure the market reaction to the announcement. Following the literature, we measure the value of abnormal returns using the absolute value. To estimate abnormal returns, we use a market model in which the model parameters are estimated over a 200-day window ending 50 trading days before the yearly earnings announcement date (Mikkelson and Partch, 1986). We also divide our sample into different sub-groups based on different categories, such as US and non-US firms to see if these price shocks have different effects on each groups.

Table 3-2 shows the summary statistics of three-day earnings announcement returns and the macroeconomic control variables used in our regression analysis.

The mean (median) absolute value of announcement return for the total sample is 2.96% (1.77%), which is consistent with prior research. In the sub-groups based on the volatility of jet fuel price, we can see that the absolute value of announcement return for the stable period (1995 to 2000) is higher than that for the volatile period (2001 to 2007). Investors face more uncertainty in US rather than non-US airlines with regard to three-day announcement returns. This table also shows that three-day announcement returns of strong-governance airlines are larger than these of weak-governance airlines.

In addition to examining the current year and one-year lagged shocks of oil price, interest rate and foreign exchange rate to announcement returns, we are also interested in whether companies’ hedging activities for these risk exposures have any effect on investors’ information set. We collect hedging information about oil price, interest rate and foreign exchange rate risk exposures from the airlines’ annual reports. We then construct a dummy variable for each hedging behavior and set it equal to 1 if a airline company hedges for a given risk exposure, and 0 otherwise.

We use the following regression to analyze the relationship between absolute abnormal stock returns around earnings announcements and absolute changes in oil

price, interest rate and foreign exchange rate hedging activities:

Abs(AR)t = α + β1 Abs(ΔFUEL)t + β2 Abs(ΔFUEL)t × Hedged in FUEL + β3 Abs(ΔIR)t + β4 Abs(ΔIR)t × Hedged in IR

+ β5 Abs(ΔFX)t + β6 Abs(ΔFX)t × Hedged in FX

+ β7 Abs(ΔFUEL)t-1, t + β8 Abs(ΔFUEL)t-1, t × Hedged in FUEL + β9 Abs(ΔIR)t-1, t + β10 Abs(ΔIR)t-1, t × Hedged in IR

+ β11 Abs(ΔFX)t-1, t + β12 Abs(ΔFX)t-1, t × Hedged in FX

+ β13 Abs(Deviation in GDP)t + β14 Market Volatilityt + εt (3-1)

where Abs(AR)t is the absolute value of the abnormal returns during the three-day period around the yearly earnings announcements; Abs(ΔOIL)t is the absolute percentage change of crude oil price coming from WTI Spot Cushing for year t;

Abs(ΔIR)t is the absolute percentage change of three-month yield on Treasury bills or inter-bank offering rate coming from Datastream for year t; Abs(ΔFX)t is the absolute percentage change of foreign exchange rate for each country retrieved from Datastream for year t; Abs(ΔFUEL)t-1, t, Abs(ΔIR)t-1, t and Abs(ΔFX)t-1, t are the respective cumulative price shocks over one year prior to the earnings reported year t;

Abs(Deviation in GDP)t is the absolute value of the difference between the change in GDP in year t and the average yearly change in GDP for each country over our sample period; and Market Volatilityt is the standard deviation of daily returns for the Datastream market index value for each country in year t. We include Abs(Deviation in GDP)t and Market Volatilityt to control for general macro-level uncertainty in the economy, which is likely to affect investors’ uncertainty when they form their firm-specific expectations. Finally, εt is the error term. To alleviate potential contaminating effects of outliers on our results, we winsorize the absolute announcement return variables at the 99^th percentile of their distributions.

Table 3-3 shows the results of the regression estimates of Equation (3-1). We estimate the regressions using a pooled sample with robust standard errors, which account for the clustering sample effect of our study period. Because we use

absolute values of three-day earnings announcements and changes in price shocks in the regression analysis, we predict a positive relation between investor uncertainty and the price shocks. Following the literature, we draw inferences about the significance levels of the regression coefficients based on one-tailed t-statistics.

Furthermore, we find that there is correlation between current shocks and one-year lagged shock in the empirical tests.²⁵ Therefore, we first report two reduced form regressions of Equation (3-1), which are presented in Columns (1) and (2) of the tables. Column (1) shows the current shocks, while Column (2) shows only the one-year lagged shocks.

Panel A of Table 3-3 reports the results for the total sample. The evidence shows that investors do not fully anticipate the firm-specific effects of changes in oil price and interest rate for firms with risk exposures in the current year and one-year lagged. In contrast, when airlines hedge for these risk exposures, there is a negative relation between shocks and the absolute value of abnormal returns. Thus, it seems that hedging can reduce investor uncertainty about the effects of these risk exposures on airlines’ earnings.

Panel B of Table 3-3 shows the results for the sub-sample based on jet fuel price volatility, with the sample period from 1995 to 2000 representing a relatively stable fuel price period, while 2001 to 2007 a relatively volatile one. We can see that oil price shocks in current year and one-year lagged have a positive impact on the three-day absolute value of abnormal returns only in the stable period. This suggests that investors face more uncertainty in the stable period than in the volatile period, and this result is not consistent with our expectation. Moreover, investors in the volatile period face more uncertainty for current year and one-year lagged interest rate shocks, and airlines that hedge for shock would reduce investor information uncertainty.

Panel C of Table 3-3 reports the results for the sub-sample of US and non-US airlines. The result shows that current year shock of oil price increases investor uncertainty for US airlines, while one-year lagged shock of oil price increases investor uncertainty for non-US firms. It may be that the US market is more efficient than the non-US market, so that information can be incorporated into investors’ decisions

25 For example, the correlation coefficient between the absolute value of current period shocks to oil price and the absolute value of one-year lagged shocks to oil price is 0.6186 in our sample.

more quickly and efficiently. The coefficients of interest rate shocks also indicate this phenomenon, since for US airlines, we find that hedging for interest rate risk exposure in current year and one-year lagged shock reduces investor uncertainty.

The reason could be that information about financial disclosure of derivatives use is clear and easily observed in the US market, so that investors can gather related information on the effects of hedging on firms’ earnings more efficiently.

Panel D of Table 3-3 shows the results for the sub-sample based on corporate governance mechanisms. We construct a composite index using six corporate governance indices suggested by LLSV (1998), which are shareholder protection laws, creditor protection laws, law enforcement, efficiency of judicial system, corruption and expropriation. By adding ranked deciles scores of these six indices, we take the median value to partition our sample into two sub-groups. The sub-group with the composite index higher than the median is called the “strong governance sample”, while that with the composite index below the median is the “weak governance sample”. Overall, investors encounter more uncertainty only for current year shocks with airlines in the strong governance sample, especially for oil price and interest rate shocks. Our results are consistent with Chung et al.’s (2004) finding that better corporate governance is related to higher-quality information. Furthermore, the results also demonstrate that airlines that hedge for oil price, interest rate and foreign exchange rate shocks reduce investor information uncertainty in current years.

3.2 The Impact of Changes in Fuel Price, Interest Rate and Foreign Exchange Rate on Analysts’ Forecast Errors

The results in Table 3-3 show that investors encounter difficulties in dealing with the uncertainties caused by changes in oil price, interest rates and foreign exchange rates. In this section, we further examine the association between absolute changes in these three risk exposures and information uncertainty using analysts’ forecasts errors to replace investors’ earnings expectations as our dependent variables.

Baron et al. (1998) analytically prove that the mean forecast error, together with forecast dispersion and the number of forecasters, can be used to estimate analysts’

total uncertainty and their degree of consensus (common uncertainty relative to total uncertainty). They show that analyst forecast errors and dispersion in analysts’

forecasts increase with analyst uncertainty. We focus on the median analysts’

consensus annual earnings per share forecasts,²⁶ as reported on I/B/E/S. I/B/E/S releases monthly analyst forecasts and our data are adjusted for stock splits and dividends. We calculate the absolute value of analysts’ forecast errors as follows:

Percentage absolute forecast error = ︱(AFt-EPSt)/EPSt︳ (3-2)

where AFt is the median analysts’ forecast in the year prior to the end of the year for which earnings are reported, and EPSt is the actual reported earnings per share in year

t (as defined by I/B/E/S). We winsorize the absolute analysts’ forecast errors at the

99^th percentile values to minimize the effect of extreme observations in the empirical studies.

Table 3-4 is the summary statistics of analysts’ forecast errors and other control variables used in the regression analysis, and we also report the sub-sample’s summary statistics. In Panel A of Table 3-4, the results show that the average absolute value of analysts’ forecast errors is 0.6609, indicating that on average the median forecast errors is about 2/3 of reported earnings. The distribution of analysts’ forecast error is right-skewed in our sample, and the median forecast error is about 22.5% of the reported earnings. The average number of analysts following an airline company is 15.17, and the median is 13. Panel B of Table 3-4 shows that average absolute value of analysts’ forecast errors is 0.6968 in the stable period, slightly greater than that in volatile period, 0.6320, and the number of analysts following an airline company is almost the same for these sub-samples. Panel C of Table 3-4 shows that the average absolute value of analysts’ forecast errors is greater for non-US than that for US airlines, and that the number of analysts who make forecasts for the former is also greater than that for the latter. Panel D of Table 3-4 depicts the descriptive statistics for strong and weak governance sub-groups. The average absolute value of analysts’ forecast errors is higher for airlines from weak governance regions than that for these from strong governance ones.

26 The empirical results of using mean values of analysts’ consensus annual earnings per share forecasts are similar to those obtained using median values. The summary statistics in Table 4 show that our sample and sub-sample are right-skewed, therefore, we use the absolute error in the median analyst forecast to alleviate the influence of large errors caused by individual analysts, as documented by most previous studies.

We examine the relationships between changes in oil price, interest rate and foreign exchange rate shocks and analysts’ forecast errors as follows:

Abs(AFE)t = α + β1 Abs(ΔFUEL)t + β2 Abs(ΔFUEL)t × Hedged in FUEL + β3 Abs(ΔIR)t + β4 Abs(ΔIR)t × Hedged in IR

+ β5 Abs(ΔFX)t + β6 Abs(ΔFX)t × Hedged in FX

+ β7 Abs(ΔFUEL)t-1, t + β8 Abs(ΔFUEL)t-1, t × Hedged in FUEL + β9 Abs(ΔIR)t-1, t + β10 Abs(ΔIR)t-1, t × Hedged in IR

+ β11 Abs(ΔFX)t-1, t + β12 Abs(ΔFX)t-1, t × Hedged in FX + β13 Abs(Deviation in GDP)t + β14 Market Volatilityt

+ β15 Log(No. of Analysts)t + εt (3-3)

where Abs(AFE)t is the absolute change in analysts’ forecast errors described in Equation (3-2), and Log(No. of Analysts)t is the logarithm of the number of analysts making forecasts in the year prior to the end of the year for which earnings are reported. The accuracy of consensus forecasts is likely to increase along with the number of analysts making forecasts, and there is a negative relationship between number of analysts making forecasts following a given company and both the noisy estimates of consensus and volatile analysts’ forecast errors. The rest of variables are the same as in Equation (3-1).

Table 3-5 reports the estimate results of Equation (3-3). Column (1) gives the result of the impact of current year shocks to oil price, interest rate and foreign exchange rate on analysts’ forecast errors, while Column (2) shows the influence of one-year lagged shocks to fuel, interest rate and foreign exchange rate on analysts’

forecast errors. Finally, Column (3) presents the estimation for the full specification.

Consistent with previous research, we expect that the absolute changes of these shocks have a positive relation with analysts’ forecast errors. From our regression estimates, we show that only interest rate shocks are consistent with this expectation for total sample, as shown in Panel A of Table 3-5. There is no positive or significant relationship in oil price and foreign exchange rate shocks. It is worth

noting that for airlines with hedging for oil price, current year and one-year lagged oil price shocks with have a positive impact on forecast errors, while hedging for interest rate and foreign exchange rate shocks have a negative impact on forecast errors. We suggest several reasons to explain this phenomenon. First, airline companies are very concerned with the volatility of fuel price because such costs account for large amount of their operating expenses. Therefore, predicting fuel prices is important for airlines and analysts when making earnings forecasts. The positive relation between oil price shocks coupled with fuel-hedging and forecast errors may imply that analysts can not evaluate the influence of fuel price on fuel-hedged airlines’

earnings, as the hedging may add to analysts’ information uncertainty during an oil price rise, and thus increase forecast errors. Second, financial reports only disclose certain information about the results of fuel hedging, and analysts can not observe the exact processes and policies adapted. Therefore, analysts encounter difficulties when making earnings forecasts, because they lack information on how changes in hedging policy may affect reported earnings during the fiscal year. Third, the efficiency of fuel hedging may not be as great as analysts predict. Financial accountants’ hedging skill, financial derivatives’ accessibility and top managers’

attitude to hedging are all factors that can affect airlines’ hedging efficiency, and thus influence analysts’ forecasting ability.

Panel B of Table 3-5 shows the regression results for the sub-sample based on volatility of fuel price. The absolute changes of these shocks have a significant influence on analysts’ forecast errors in the volatile period. We can see that the regression coefficients of oil price and foreign exchange rates are significant in current year shocks, no matter whether they engage in hedging activities or not.

Moreover, the influence of shocks to oil price on forecast errors is similar for the total sample in this period. The possibilities of making inefficient hedging decisions and the difficulties in predicting fuel prices reduce the precision of analysts’ forecasts, and thus fuel hedging seems to lead to more inaccurate forecasts in the volatile fuel price period.

Panel C of Table 3-5 presents the results for US and non-US airlines.

Compared with non-US airlines, analysts face more information uncertainty when making earnings forecasts for US airlines with current year and one-year lagged oil price shocks. Conversely, analysts’ forecast errors are significantly greater for

non-US airlines than for US airlines with jet fuel hedging for current year and one-year lagged shocks. The coefficients of absolute change in interest rate are all significant for non-US airlines. Contrary to oil price shocks, there is a positive relation between absolute changes in interest rate (current year and one-year lagged) and forecast errors, and a negative relation between absolute changes in interest rate with interest rate hedging (current year and one-year lagged) and forecast errors.

This means that shocks to interest rates can increase analysts’ forecast errors, but airlines with interest rate-hedged can reduce this uncertainty for the non-US airlines sub-sample. Furthermore, we see that the number of analysts making forecasts for a given firm has a negative effect on analysts’ forecast errors for US airlines, and the coefficients are all significant in the three regression estimates. This is consistent with our expectation that more analysts can reduce the noisy estimates of consensus and volatile forecasting errors. Overall, the US market seems more efficient than non-US markets as the related analysts’ forecast errors are less affected by such shocks and show a greater consensus.

Panel D of Table 3-5 provides the regression results for sub-samples based on corporate governance mechanisms. The coefficients of regressions in strong governance airlines are mostly insignificant, which implies that shocks to oil price, interest and foreign exchange rates have no impact on analysts’ forecasts errors.

This result is consistent with Chung et al.’s (2004) findings that better corporate governance is associated with higher-quality public information and generates greater consensus among analysts. On the other hand, we find that shocks to oil price and interest rates have significant effects on forecast errors in weak governance airlines.

The current and one-year lagged oil price shocks reduce forecasting errors, while hedging for jet fuel price increases analysts’ uncertainty. In addition, our results show that interest rate shocks have a positive effect on analysts’ forecast errors, both in current year and one-year lagged shocks, but airlines with interest rate hedging can reduce this uncertainty. Finally, we find that there is a positive relation between the number of analysts and forecast errors in weak governance airlines. This relationship contradicts our previous findings, and our explanation is that when airlines have weak governance analysts will have more diverse opinions about earnings forecasts, because of inefficient information transmission or low-quality public information.

3.3 The Impact of Changes in Fuel Price, Interest Rate and Foreign Exchange Rate on Revisions in Analysts’ Forecasts

A number of recent papers suggest that market participants are concerned with

analysts’ forecast accuracy. This implies that the intensity of the market’s response to earnings forecast revisions should increase along with analyst forecasting ability.

Using a broad sample of over 15,000 analyst-firm-year observations from I/B/E/S, Park and Stice (2000) find that there is a positive relation between past forecast usefulness and the market’s response to individual analysts’ forecast revisions.

Francis and Soffer (1997) also find that the market responds more strongly to earnings forecast revisions accompanied by buy (versus hold or sell) recommendations, using 556 analyst research reports available in the Investext database from 1989 to 1991.

In this section, we examine whether and to what extent analysts revise their

In this section, we examine whether and to what extent analysts revise their

在文檔中 Two essays on corporate hedging policy and firm value of global airline industry (頁 66-0)