The Exposure of Shipping Stock Returns to Freight Rate Risk and Other
Risk Factors
ABSTRACT
The exposures of monthly stock return of 113 shipping firms from 27 countries to 9 risk factors including the market return are estimated in OLS regression models using monthly data from 2005 through 2010. Freight rates appear to enhance the explanatory power of the model in explaining bulk and tanker shipping stock returns. Freight rates reflect much market information about shipping services, resulting in no joint residual explanatory power of traditional demand drivers steel production, oil production and world export. Firm specific variables size, book to equity market value, leverage and liquidity cannot explain the cross sectional variation of absolute risk exposure levels.
Master Thesis of Steven Roos
University of Groningen, The Netherlands
Author Steven Roos
Student number S1612514 Thesis supervisors: dr. N. Brunia
dr. P.P.M. Smid
Date: August 24th, 2012
1 1. Introduction
The objective of this study is to find relevant risk factors that explain the time series variation of marine transportation firms’ stock returns. Therefore, an Ordinary Least Squares (OLS) regression model estimating shipping firms’ stock returns to 9 risk factors including the market portfolio is estimated. The model is developed from the point of view that the risk and return characteristics of firms from three maritime subsectors are heterogeneous. This heterogeneity is confirmed by different stock return exposure coefficients to risk factors between firms operating in different subsectors of the maritime industry. The main difference between this model and comparable models from previous studies is that this model includes the freight rates for the transportation of oils, bulk goods and containers. In addition to estimating exposure coefficients, a second model estimates whether firm level variables can explain the cross sectional variation of obtained exposure coefficients.
2
the importance of marine transportation. Furthermore, they also note that equally limited attempts have been made to uncover factors, other than the market, to which stock returns in these subsectors are exposed. The bulk shipping market is closely linked to the steel industry, the tanker shipping market to the oil production industry and the container shipping market to the export of finished and semi-finished goods, it is straightforward that freight rates for the transportation of these goods develop unequal. The uncertainty whether shipping stocks are homogenous in terms of stock return risk exposure is further exploited in this study. The results of this study will answer the question to which relevant risk factors shipping firms’ stock returns from the tanker, bulk and container subsector of the maritime industry are exposed and how obtained exposure coefficients of the different subsectors compare to each other.
The model estimates the exposure to 9 risk factors of 113 firms’ stock return for every firm individually. The mean and median exposure to the risk factors are compared to the mean and median exposure of other firms to risk Factors. Bartam (2002) explains that in the case that constituencies of a portfolio exhibit heterogeneous levels of exposure, estimating individual regressions is more powerful than estimation portfolio regressions. Since heterogeneity caused by among other hedging activities is an assumption underlying the model, estimating individual firms’ stock return exposure is preferred over forming portfolios of shipping stocks based on the subsectors of the maritime industry and estimating portfolio exposures. The applied regression technique is OLS. Although this technique is not very complicated, it is widely used in studies which estimate time series exposures to risk factors.
3
The obtained exposures to risk factors are estimated in OLS regression models including the assumed determinants of exposure: size, boot to equity market value, leverage and liquidity. Among many others, He and Ng (1998), Choi and Kim(2003) and Jorion (2010) used firm specific variables to explain the cross sectional variation of stock return exposure. The only study that also applies the integrated approach of first estimating shipping firms’ stock return exposure and consecutively modeling the determinants of the obtained exposures is done by El-Masry et al. (2010). Although the methodology applied in this study shows similarities to the methodology of that study, the included variables do not match, of which the inclusion of freight rates in this study is the most notable difference.
The outline of the remainder of this paper is as follows: in section 2, relevant literature on this topic is discussed and the hypotheses are developed. In section 3, the methodology is discussed. In section 4, the empirical results are presented. In section 5, the empirical findings are discussed and in section 6 the conclusion is presented.
2. Literature
2.1. Analysis of Shipping Firms’ Stock Returns
4
Examples are: Bandhari (1988), (1994) and Hou and van Dijk (2008). This study focuses, apart from estimating time series exposure of stock returns to risk factors, on explaining the cross sectional variation of risk exposures rather than the cross sectional variation of stock returns. Studies that have successfully applied OLS regression models to explain the cross sectional variation stock return exposure by firm specific variables are done by Geczy et al (1997), He and Ng (1998) and El-Masry and Abdel-Salam (2007).
2.2. The size of future cash flows
5
Since the freight rate is the equilibrium between demand for and supply of shipping services, the freight rate contains information about both the volume shipped and the price paid per tonne per mile, which together constitute the revenues of the shipping firm. An unexpected rise of freight rates can be caused either by an unexpected increased demand for shipping services or by an unexpected decreased supply of shipping services. In the case that freight rate changes result from demand side adjustments, transporting firms’ cash flows should generally be positively exposed to freight rate risk, because a joint rise of volume and prices cause industry cash flows to rise. In the case that supply side variables change, the freight rate rises in the case that supply and thus total volume transported decreases. Since supply side adjustments cause freight rates to increase and transported volumes to decrease simultaneously, the firm level cash flow exposure depends on both the degree to which the firm level cash flows decreases compared to industry cash flows decreases, and the elasticity of freight rates to supply changes. The cash flow exposure to freight rate risk can be negative when the freight rate change is caused by supply side fluctuations and this volume effect outweighs the price effect. However, in most cases freight rate fluctuation should be caused by demand side factors, because these factors, such as commodity markets and economic activity, are generally more volatile than supply side factor such as fleet size and new building. This results in an expected positive sign of the stock return exposure coefficient to freight rate risk. All analyzed previous studies that estimate exposure of shipping firms’ stocks to risk factors did not include freight rates in their models. The expected positive relation between freight rates and stock shipping returns is formulated in Hypothesis 1.
H1: The freight rates of seaborne oil, bulk and container transportation have joint explanatory power in explaining time series variation of shipping firms` stock returns.
6
Grammenos and Arkoulis (2002) and Drobetz et al. (2010) who use extensive multifactor models to estimate exposure of shipping stocks to risk factors. Both studies identify the oil price and industrial production as important return drivers for shipping stocks. Westgaard et al. (2007) estimated the exposure of tanker shipping stocks to seven variables including industrial production risk and oil price risk. Drobetz et al. (2010) argue that industrial production is a main parameter affecting seaborne trade volume. Although industrial production is intuitively closely linked to seaborne trade, produced goods do not need necessarily to be exported. Therefore, total export volume would be a better variable helping explain shipping stock return exposure fluctuations than industrial production. Therefore, the world total export of goods is included as proxy for the demand for container transportation. For the estimation of tanker demand, the production volume of oil products will be included, and for the estimation of bulk carrier demand the volume of steel production will be included. The choice for steel production is supported by the United Nations Review of Maritime Transport (2010), which explains that steel production is the main driver for bulk shipments. Although for oil and steel markets also holds that export figures are intuitively better proxies for transportation demand, these figures are not available on a monthly basis.
Hypothesis 2: Inclusion of traditional demand drivers for shipping services in a model that explains time series variation of shipping stock returns that also includes freight rates does not increase explanatory power of that model.
It is expected that when the freight rate is being regressed as explanatory variable of stock returns, the expected regression coefficient is positively signed. For the more traditional explanatory variables oil production, steel production and world export, no significant exposure is expected in the same model, since theory predicts that all information on demand for shipping services is included in the freight rate. Among other risks, freight rate risk can be hedged by the use of financial derivatives. The incentives for and effects of hedging risk exposure will be discussed in section 2.5.
2.3. The Present Value of Future Cash Flows and the Cost of Debt Financing
7
markets such as the shipping industry. However, risk free debt does not represent the complete cost of debt financing, since shipping firms are not able to borrow at the same interest rates as governments, but also face credit spreads. Nevertheless, increasing risk free interest rates are expected to lead to higher debt financing costs and consequential lower net cash inflows for shipping firms. This holds that both from a discount rate perspective and from a debt financing cost perspective, shipping firms’ stock returns are expected to be negatively exposed to changes in risk free interest rates. It is possible that the effect of interest rate fluctuations is priced into the market portfolio, because changing interest rates affect the discount rate for all stocks on the market. In that case, regressions are not able to detect significant residual stock return exposure to interest rate risk when the return on the market portfolio is also included in the regression model. Of considerable importance is the decision which debt instrument and which time to maturity to include, when estimating risk free interest rate exposure.
8
9
exposures obtained by the studies using portfolio analyses of Prasad and Rajan (1995) and Drobetz et al (2010). The observation that generally a few firms show significant exposure to interest rate risk might indicate that that the effects of interest exposure are included in the market portfolio. Another explanation of the small number of stocks exhibiting exposure to interest rate risk might be that many firms use successful hedging strategies to reduce the exposure of stock returns to interest rate risk. However, the firm level analysis of El-Masry et al (2010) show that a not negligible part of shipping firms actually exhibits significant negative exposure to long term interest rate risk. This is in line with the theoretical expectation presented earlier in this section. In line with previous studies, the 10 year United States Treasury Note is used in the models of this study to serve as proxy for the long term interest rate. The short term interest rate is not included in the model, because a rate discounting cash flows 10 years from now is preferred over a discount rate of for example 3 months, because most cash flows will be received after many years. Also the maturity of most debt obligations for financing purposes will be years rather than months.
2.4. The Dollar Value of Future Cash Flows
10
11
magnitudes of exposures in developing countries appear to exceed the magnitudes of exchange rate exposures in developed countries. A significant negative exposure of return to the value of the domestic currency is detected in 5.0% of the cases, while a positive signed exposure is estimated in 6.4% of the cases, both at the 5% level of confidence. These results are inconclusive about the sign of exposure coefficients. El-Masry et al (2010) measured exposure of shipping stock returns to exchange rate risk. They define the exchange rate as USD value expressed in local currency. They find that 9% of the firms exhibit negative exposure at the 99% level of significance. 10% of the firms exhibit positive exposure at the 99% level of significance.
Although theory predicts a positive residual exposure coefficient of shipping stock returns to USD value risk for non-U.S. shipping firms, previous studies found mixed results. An explanation for low numbers of significant residual exposures is offered by Choi and Jiang (2009), who study the difference in exchange rate exposure between U.S. multinational companies and U.S. non-multinationals between 1983 and 2006. They define multinational firms as firms with over 500 million USD foreign sales and at least three country representations in the Compustat database. Their main finding is that non-multinationals are exposed to exchange rate risk while multinational companies are not. The explanation they present is that non-multinational firms are indirectly exposed to exchange rate risk, because of the changing competitiveness compared to foreign competitors as a result of changes of the exchange rate. Furthermore, multinational firms appear to employ more operational and financial hedging strategies to mitigate exchange rate risk than non-financial firms do. Based on the findings of previous literature, only a small number of significant coefficients of non-U.S. stock return exposure to value changes of the U.S. dollar should be expected, and if any, most of them are expected to be positive because of the mismatch between dollar denoted cash inflows and outflows that is typical for the shipping industry.
2.5. Explaining the Cross Sectional Variation of Risk Exposure
12
who use these instruments aim at is a lower absolute exposure to exchange rate risk, interest rate risk or freight rate risk. Samitas and Tsakalos (2010) find empirical evidence that ship owners successfully use freight rate derivatives as hedging instrument and in this way manage their freight rate risk exposure. As opposed to dollar value risk, interest rate risk and freight price risk, the risk associated with volatility of the demand for seaborne transport cannot be hedged by the use of financial instruments. Therefore, if shipping firms appear to exhibit residual exposure to demand volume risk, this should intuitively not be explained by firm specific factors. Many studies have suggested that firm characteristics help explain the cross sectional difference in incentive to use hedging instruments. Most studies that make an attempt to explain cross sectional variation of stock return exposure among firms, explain exchange rate risk. The expectation based on previous studies is that the absolute level of exchange rate risk of firms should be negatively related to size, growth opportunities and leverage. The relation between the absolute level of exchange rate risk and liquidity is expected to be positive. Larger firms are expected to have more resources and manpower to effectively manage risks, while highly leveraged firms, illiquid firms and firms with more growth opportunities are expected to have greater incentives to hedge exposure to risk factors. However, leverage in itself increases the absolute level of exposure by construction. Leveraged firms are thus expected to have a greater incentive to hedge, resulting in lower absolute exposure levels, but have at the same time higher exposure levels due to their leveraged capital structure resulting in higher absolute levels of exposure. The sign of the regression coefficient between the absolute level of interest rate risk exposure and freight rate risk exposure and the explanatory variables size, leverage, liquidity and growth opportunities are expected to be identical to the expectations about the relation between absolute dollar value risk exposure and the explanatory variables, because theory about firms’ hedging incentives is equivalent applicable to these sources of risk, and similar hedging instruments, such as options and futures, are available.
13
implies that for these samples the effect that leveraged firms have greater incentives to hedge outweighs the effect that leveraged financing increases absolute exposure levels. Furthermore, Choi and Kim (2003) also find that firms with higher growth opportunities, which they measured as Tobins Q and R&D investments, have more incentive to hedge which reduces currency exposure. Geczy et al (1997) estimate growth opportunities as the book-to-market value and find that growth opportunities are an incentive to hedge for managers, resulting in a negative relation between book to equity market value and exchange rate risk exposure. This negative relationship between growth opportunities and currency risk exposure is also found by El-Masry and Abdel-Salam (2007). Geczy et al. (1997) explain this negative relationship by the underinvestment problem that arises when firms are not able to exploit their entire growth opportunities due to currency risk exposure. The studies of He and Ng (1998) and Choi and Kim (2003) also find that the degree of foreign operations, measured as foreign sales or foreign assets are positively related to currency risk exposure. Since all firms analyzed in this study are internationally operating shipping firms, they can be considered to have a more or less homogenous degree of foreign operations. Therefore, this variable will not be included in the models of this study.
To my knowledge, only one previous study estimates exposures of shipping firms’ stock returns to various risk factors and consecutively estimates the explanatory power of firm specific variables. This study, done by El-Masry et al. (2010), first estimates the time series exposure of stock returns to the variables: market return, oil price, dollar exchange rate, short term interest rate and long term interest rate. They explain the obtained exposure coefficients by the variables: total assets, foreign sales to total sales ratio, foreign assets to total assets ratio, long term dent tot total equity and reserves ratio, market-to-book value of equity, dividend payout ratio and current assets to current liabilities ratio. The risk factors to which time series exposures of stock returns have been estimated by El-Masry et al (2010) that match those to which exposure is estimated in this study are the USD exchange rate and long term interest rate. They find that none of the firm specific factors that were expected to explain exposures can actually explain the exposure of shipping stock returns to exchange rate risk. The variable firm size measured as total assets appears to be negatively related to the long term interest rate exposure coefficient at the 5% level of confidence. This confirms their hypothesis that larger firms are less exposed to interest rate risk. The other factors they included in their estimation were not able to explain the interest rate exposure coefficient. Hypothesis 3 tests whether this study confirms the significant finding of El-Masry et al. (2010).
14 Table 1
Expected exposure coefficient signs and determinants of absolute exposure
Expected sign of stock return exposure of to risk factor, per subsector of the
maritime industry
Expected coefficient of determinant of absolute level of exposure
Sector Bulk Container Tanker Size BtM
value
Leverage Liquidity
Market Return + + +
Interest Rate - - - - - - +
Dollar Value + + + - - - +
Bulk Freight Rate + - - - +
Tanker Freight Rate + - - - +
Container Freight Rate + - - - +
World Steel export +
World Oil Export +
Adjusted World Export +
+ Denotes positive expected coefficient, - denotes a negative expected coefficient
3. Methodolgy
3.1. Identification and Classification of Stocks
15
3.2. Time Series Regression
Time series regression model 1, displayed in equation 1, is employed to estimate the exposure of monthly stock returns of all 113 firms to the independent variables market risk, interest rate risk and dollar value risk1. The regressions of model 2, presented in equation 2, are the same regressions augmented by the independent variables: bulk freight rate risk, tanker freight rate risk and container freight rate risk. The regressions of model 3, presented in equation 3, are the same regressions as those of model 2 augmented by the independent variables: world oil production risk, world steel production risk and world export risk. OLS regressions are used to determine the exposure of individual stock returns to the concerning risk factors. Every time series regression includes a constant, and the market return as independent variable. The specification of this model is in line with, among many others, Jorion (1990), Prasad and Rajan (1995), Sadorsky (2008), Bartram and Bodnar (2009) and El-Masry et al. (2010) who all use OLS regression including a constant to estimate individual firms’ exposure to one or more risk factors. The betas associated with these risk factors only estimate unexpected changes of risk factors. As explained by Jorion (1990), the effects of expected changes of the value of any given risk factor on the stock price are reflected in the value of the constant or intercept. The error term captures the noise in stock price volatility, which cannot be captured by the other variables of the model. Prasad and Rajan (1995) and El-Masry et al (2010) test for significance of each coefficient at the 1%, 5% and 10% level of confidence. Jorion (1990) tests the significance of each coefficient at the 1% and 5% level of confidence. Because of the limited statistical and economic meaning of coefficients that are statistically significant only at the 10% level of confidence, this study tests for the significance of coefficients at the 5% level of confidence. OLS is only applicable if some assumptions are met about the distribution of variables. One assumption is that the series do not suffer from multicollinearity. Another assumption is that independent variables are exogenous, i.e. that they are independent from the dependent variable. This model assumes that independent firm stock returns do not explain any of the independent variable. This can be tested by the use of instrumental variables. However, it is very difficult, if not impossible, to find instrumental variables for all independent variables. This is possibly the reason that the discussed previous studies do also not address the problem of endogeneity.
1
16
The use of monthly data instead of higher frequency data is necessary because some of the independent variables are not available in a higher frequency. The large impact on regression outputs caused by a single shock in the stock return of a single stock, i.e. outliers, may cause incorrect inferences. This problem is reduced by the application of Winsorization at the 10% level to every time series regression variable. Each value in the segment of the highest 5% of the values of each series is set equal to the highest value in the segment of the lowest 95% of values. Also, each value in the lowest 5% segment of values of each series is set equal to the lowest value in the segment of 95% highest values of that series. Wisorization in an OLS regression model in a stock return estimation context has also been applied by Bartram and Bodnar (2009). They Winsorize only 0,1% of their sample, but the size of their sample is over 100,000 observations, which is considerable larger than the sample size in this study. In the dataset of this study, Winsorizing of at least 3.4% of observations is necessary to correct at least the highest and the lowest value of each series. Another procedure that is commonly used to deal with the effect of outliers is data trimming. That method removes the highest and lowest values of each series. Lusk, Halperin and Heilig (2011) show that Winsorizing is an equally effective method of dealing with the possible inference problems associated with outliers compared to data trimming. In section 5, the mean and median coefficients, probabilities, mean R-Squared and mean Adjusted R-Squared values of every coefficient of every subsector of the maritime industry are presented. The R-Squared value indicates what percentage of stock return volatility is explained by the model. Since this value will increase by the addition of every explanatory variable, also the Adjusted Squared will be presented. This variable adjusts the R-Squared value for the number of explanatory variables in the regression model.
1 2 3 ij i i j i j i j ij R C MRK USD INR (1) 1 2 3 4 5 6 ij i i j i j i j i j i j i j ij R C MRK USD INR FRT FRB FRC (2) 1 2 3 4 5 6 7 8 9 ij i i j i j i j i j i j i j i j i j i j ij
17
Where the elements of the models are denoted as:
ij
R - The monthly log return of stock i in period j
i
C
- The intercept of stock ij
MRK - The log return of the market portfolio in period j
j
USD - The log change of USD value expressed in local currency in period j
j
INR - The log change of the long term interest rate in period j
j
FRT - The log change of the tanker transport freight in period j
j
FRB - The log change of the bulk transport freight rate in period j
j
FRC - The log change of the container transport freight rate in period j
j
WPO - The log change of the world oil production in period j
j
WPS - The log change of the world steel production in period j
j
WEX - The log change of the world export in period j
ij
- The error term of stock i in period jNi
- The coefficient of stock return exposure of stock i to unexpected changes in the Nth risk factor3.3. Analysis of coefficient distribution
18
3.4 Cross-sectional regression
The variables in model 4 have not been Winsorized, because no other analyzed study has done this for cross sectional regression variables. A t-test is used to determine whether the mean value of every independent variable from model 4 within a subsector differs from the mean value of the same variable of the firms that are not in that subsector. The significance of the difference of medians is determined by a Wilcoxon rank sum test. of Many studies have made attempts to explain the cross sectional variation of exposure to risk factors among firms. Most of the studies discussed in section 2 that do so, employ OLS regressions. Among others, He and Ng(1998) and El-Masry et al. (2010) have done so. The absolute level of each exposure obtained from equation 3 is regressed on the factors size, leverage, liquidity and growth opportunities. Equation 4 displays this model used to estimate the cross section of stock return risk exposures. In section 5, the coefficients, probabilities, R-Squared and Adjusted R-Squared values will be presented.
1 3 4 5
ki Cki kSIZEi kBTMi kLEVi kLIQi ki
(4) Where the elements of the model are denoted as:ki
- The exposure of firm i to time series independent variable kki
C
- The constanti
SIZE - The log USD market value of firm i
LEV
i - Total book value of long term debt divided by the total market value of totalshareholder equity of firm i
i
LIQ
- Net current assets divided by current liabilities of firm ii
BTM
- Total book value of common equity divided by the total market value of common equity of firm iki
- The error term in the regression of firm i to time series independent variable kNk
19 4. Data and descriptive statistics
4.1. Description of dependent variables in time series regressions
The empirical analysis in this study consists of 2 stages. First, the time series exposures of stock returns to risk factors are estimated. Second, the cross sectional variation of the absolute levels of exposure is explained by firm specific factors. The dependent variable in every time series regressions is the monthly stock return of one of the 113 firms for a period of 60 months. The stock returns are monthly log returns of the stock prices retrieved from Thomson Datastream. The stock price data are USD denoted values, adjusted for stock splits, stock repurchases and dividends. The distributions of these returns are displayed in table 2. Container shipping stocks appear to exhibit the highest mean and median return, while tanker firms exhibit on average negative stock returns during the analysed period. This also holds for the shipping firms that are not in the bulk, container or tanker subsector.
Table 2
Distributions of stock returns after Winsorization
60 log stock return observations per firm
4.2. Description of independent variables in time series regressions
The independent variable in the time series regression model are market risk, interest rate risk, dollar value risk, tanker freight rate risk, bulk freight rate risk, container freight rate risk, world oil production risk, world steel production risk and world export risk. The proxy for the market return is the log return of Morgan Stanley Capital International (MSCI) World Index. This index is a US dollar denoted composite index of stocks traded on worldwide markets. The MSCI World Index has been elected as proxy for the market return of this model because the maritime transportation is a true international industry. The MSCI world index matches this international character, because it a free float-adjusted market capitalization weighted index composed of stocks from of 24 developed countries’ stock markets. Westgaard et al (2007a) explain that this is a commonly used proxy for the market return in the estimation of shipping stock returns risk exposure. The historical values of this
Sector Nr. of firms Min Max Mean Median Skewness Kurtosis JB
Tanker 15 -0.286 0.244 -0.003 0.003 -0.24 -0.03 23.54
Bulk 13 -0.612 0.377 0.004 0.006 -0.56 1.95 5.93
Container 11 -0.397 0.335 0.008 0.007 -0.14 0.42 16.81
Other 74 -0.557 0.528 -0.002 0.003 -0.26 1.68 5.02
20
21 Table 3
Distributions of time series regressions independent variables after Winsorization
Variable Min Max Mean Median Skewness Kurtosis JB
Market return -0.094 0.073 0.002 0.005 -0.457 -0.406 31.09
Interest Rate -0.005 0.005 -0.000 0.000 -0.031 -0.500 30.63
Freight Rate Tanker -0.313 0.296 -0.002 0.012 -0.055 -0.772 35.61
Freight Rate Bulk -0.413 0.324 0.007 0.039 -0.416 -0.376 30.23
Freight Rate Container -0.128 0.061 -0.014 -0.009 -0.568 -0.420 32.47
Oil Production -0.012 0.012 0.001 0.000 -0.038 -0.192 25.48
Steel Production -0.057 0.108 0.005 -0.003 0.641 0.092 25.25
World Export -0.046 0.044 0.008 0.010 -0.711 0.210 24.52
60 observations per variable
Table 4
Correlation between time series regression independent variables Market Return Interest Rate Tanker Freight Bulk Freight Container Freight Oil Production Steel Production Interest Rate 0.18 Tanker Freight 0.11 0.11 Bulk Freight 0.03 0.07 0.11 Container Freight 0.26 0.08 0.12 0.20 Oil Production -0.01 0.00 0.13 0.00 0.14 Steel Production 0.11 0.33 0.24 0.20 0.27 -0.03 World Export 0.18 0.04 0.20 0.08 0.40 0.18 0.19
4.3. Description of independent variables in cross section
22 Table 5
Characteristics of cross-sectional independent variables
Mean Median
Nr of
firms Min Max
M ea n su b se cto r (M ea n su b se cto r) – (me an o th er ) M ed ia n su b se cto r (M ed ia n su b se cto r) – (me d ia n o th er ) Tanker 15 Size -1.74 11.30 5.96 0.59 5.50 0.12 Book to Market 0.00 3.50 1.05 -0.41 0.77 -0.21 Leverage 0.00 2.93 0.81 0.01 0.59 -0.02 Liquidity 0.29 7.71 2.35 -0.32 1.70 0.14 Bulk 13 Size 4.24 10.42 6.77 1.46 * 6.83 1.55 Book to Market 0.45 3.65 1.43 0.03 1.00 0.03 Leverage 0.04 2.79 0.82 0.03 0.62 0.03 Liquidity 0.37 7.96 2.97 0.39 2.04 0.50 Container 11 Size -0.12 6.89 3.71 -1.95 4.26 -1.53 * Book to Market 0.24 8.09 1.76 0.40 0.94 -0.04 Leverage 0.00 1.81 0.81 0.01 0.88 0.29 Liquidity 0.27 5.70 1.41 -1.34 1.07 -0.57 * All 114 Size -2.71 13.50 Book to Market 0.00 17.68 Leverage 0.00 3.29 Liquidity 0.27 29.24
Minimum, maximum, mean and median value of each variable within each subsector. A Levene’s test has been applied to determine whether the values of a subsector and all other firms have equal variances. If so, a Student t-test has been applied to determine whether the values of that subsector and all other firms have equal means. If not, a Welch t-test has been applied to test whether the values of that subsector and all other firms have equal means. A Wilcoxon rank sum test has been applied to determine whether the values of a subsector and all other firms have equal medians. * and ** denote significant different means or medians at the 5% and 1% level of confidence respectively.
5. Results
5.1. Exposure of stock returns to risk factors
23
of model 1, implies that addition of freight rates of tanker, bulk and container shipping to the time series model improves the model. Even when adjusted for the increased number of variables, the explanatory power increases. This is confirming hypothesis 1, which expects an increase of explanatory power of the model resulting from inclusion of freight rates for bulk, container and tanker transportation. This finding supports the view of Beenstock and Vergottis (1989, 1993) that freight rates reflect relevant market information. Once the freight rates are included in the model, joint addition of proxies for oil production, steel production and total exports does not improve the model. This follows from the almost equal average Adjusted R-Squared values of the models 2 and 3. This also supports the view of Beenstock and Vergottis (1989, 1993), and hypothesis 2, that all relevant information is included in the freight rate, and no explanatory power is added by inclusion of proxies of demand drivers for shipping services. The R-Squared and Adjusted R-Squared values for models 1, 2 and 3 are presented in table 6. El-Masry et al. (2010) also estimated individual risk exposure of shipping firms. They estimate exposure to the market return, exchange rate risk, long and short term interest rate risk and oil price risk. Unfortunately, they do not report R Squared values or Adjusted R-Squared values of the regressions. They also estimate a panel to determine the exposure of all firms’ returns to the same risk factors. The Adjusted R-squared value of this panel is 0.0011. compared to this value, the average Adjusted R-Squared value of 0.236 that is found by estimation of model 3, is very high. It should be noted that comparing Adjusted R-Squared values of panel and average values from individual regressions is not completely fair.
Table 6
Explanatory power of different time series models
Model number: 1 2 3
Included risk factors:
Market risk Interest rate risk USD value risk
Market risk Interest rate risk USD value risk Tanker freight risk Bulk freight risk Container freight risk
24
The minimum, maximum, mean and median value of all estimated regression coefficients from model 3 are presented in table 7. The presentation is based on subsector of the maritime industry. The difference between the mean coefficient of a subsector is compared to the mean of the same coefficient of all other firms. T-tests determine whether the difference between both means is significant. The mean of every coefficient of every subsector is compared to the mean of the coefficients of all other firms. Wilcoxon rank sum tests are applied to test whether differences in median coefficient are present. The results of these comparisons of means and medians are also presented in table 7. The significant higher mean and median exposure coefficient of tanker transporting firms to tanker freight rate risk compared to the mean and median exposure of other firms’ exposure to tanker freight rate risk confirms the finding that the tanker freight rate is an important explanatory factor of tanker stock returns. The same logic applies to bulk shipping stocks; the mean and median exposure coefficient of bulk shipping firms’ stock returns to bulk freight rate risk is significantly higher than the exposure of other firms’ stock returns to this risk factor. In contradiction to bulk and tanker transportation firms, none of the container shipping stocks exhibits significant exposure to container freight rate risk. This may be caused by effective hedging practices, or by the fact that the HRCI container index does not proxy the price of seaborne container transportation as good as the BDI proxies the bulk transportation rate and the BDWI the oil transportation rate.
25 Table 7 Properties of risk exposure coefficients from model
Significant Exposures Mean Median
Nr of firms Positive coefficients Negative coefficients min Max M ea n su b se cto r (M ea n su b se cto r) – ( me an o th er ) M ed ia n su b se cto r (M ed ia n su b se cto r) – (me d ia n o th er ) Nr % Nr % Tanker 15 Market return 8 53% 0 0% -0.03 0.02 -0.01 -0.01 * -0.01 -0.01 Interest rate 2 13% 0 0% 0.29 10.79 5.56 1.90 6.08 2.31 USD value 9 60% 3 20% -0.23 4.04 1.46 2.58 1.04 -0.22 Tanker freight 3 20% 0 0% -0.20 0.24 0.05 0.08 ** 0.05 0.07 * * Bulk freight 4 27% 0 0% 0.00 0.24 0.07 -0.04 0.03 -0.06 Container freight 0 0% 0 0% -0.39 0.33 0.02 -0.05 0.03 -0.01 Oil production 0 0% 0 0% -2.90 2.74 -0.68 -0.40 -1.28 -1.02 Steel production 0 0% 0 0% -0.42 0.19 -0.12 -0.02 -0.15 -0.07 World export 1 7% 0 0% -0.53 1.16 0.39 0.19 0.35 0.24 Bulk 13 Market return 9 69% 0 0% -0.03 0.03 0.00 0.00 * 0.00 0.00 * Interest rate 2 15% 0 0% -7.89 19.04 3.45 -0.52 4.31 0.33 USD/LC value 9 69% 4 31% -0.67 18.54 4.09 2.96 1.52 0.45 Tanker freight 0 0% 2 15% -0.24 0.14 -0.05 -0.04 -0.07 -0.07 Bulk freight 10 77% 0 0% 0.08 0.32 0.19 0.10 ** 0.19 0.12 * * Container freight 0 0% 0 0% -0.18 0.63 0.13 0.08 0.01 -0.03 Oil production 0 0% 0 0% -6.26 2.23 -0.22 0.13 -0.42 -0.09 Steel production 0 0% 0 0% -0.72 0.17 -0.23 -0.13 -0.17 -0.10 World export 1 8% 0 0% -0.69 2.76 0.69 0.52 ** 0.53 0.37 * Container 11 Market return 4 36% 0 0% -0.02 0.05 0.01 0.01 0.01 0.01 Interest rate 1 9% 0 0% -3.98 8.96 3.48 -0.48 4.11 0.13 USD value 6 55% 2 18% -4.92 4.17 0.58 -0.91 1.25 0.15 Tanker freight 0 0% 1 9% -0.28 0.14 -0.04 -0.03 -0.05 -0.06 Bulk freight 4 36% 0 0% -0.04 0.37 0.12 0.01 0.06 -0.02 Container freight 0 0% 0 0% -0.29 0.47 0.12 0.06 0.21 0.17 Oil production 1 9% 0 0% -2.31 5.20 0.81 1.26 * 0.99 1.36 Steel production 0 0% 1 9% -0.64 0.44 -0.17 -0.07 -0.19 -0.12 World export 0 0% 0 0% -1.05 0.98 -0.07 -0.33 -0.03 -0.24 All 114 Market return 51 45% 0 0% -0.07 0.05 0.00 0.00 Interest rate 13 12% 0 0% -20.3 25.53 3.91 4.00 USD value 63 56% 14 12% -4.92 18.54 1.40 1.11 Tanker freight 6 5% 6 5% -0.28 0.24 -0.02 0.00 Bulk freight 38 34% 2 2% -0.20 0.37 0.10 0.08 Container freight 3 3% 0 0% -0.69 0.94 0.07 0.04 Oil production 1 1% 1 1% -6.26 5.20 -0.33 -0.36 Steel production 0 0% 3 3% -0.92 0.44 -0.11 -0.08 World export 4 4% 0 0% -1.67 2.76 0.22 0.21
26
5.2. Explanation of exposures by firm-specific variables
The firm specific factors size, book to equity market value, leverage and liquidity are modeled to explain the cross sectional variation of stock return exposure coefficients to time series risk factors. These coefficients have been obtained from estimations of model 3. The result of this analysis is summarized in table 8. From the low R-squared and adjusted R-squared values can be concluded that the explanatory power of this model is very low, compared to for example the time series models. Hypothesis 3 expects a significant negative relationship between firms size and the exposure to interest rate risk. This is not confirmed by the results of the cross sectional modeling presented in table 8.
Table 8
Explanation of absolute levels of exposure by firm-specific variables
Independent: Constant Size Book to equity
market value
Leverage Liquidity
Dependent: Coeff P-val Coeff
. P-val. Coeff. P-val. Coeff. P-val. Coeff. P-val. 2
R
AR
2 Market return 0.591 0.000** 0.009 0.521 0.021 0.365 -0.090 0.132 0.009 0.438 0.04 0.01 Interest rate 6.519 0.000** 0.011 0.933 -0.083 0.705 -0.586 0.303 -0.130 0.245 0.02 -0.02 USD value 2.103 0.004** 0.033 0.687 -0.168 0.190 -0.529 0.114 -0.059 0.366 0.05 0.01 Tanker freight 0.071 0.000** 0.001 0.782 -0.001 0.831 0.002 0.822 0.001 0.383 0.01 -0.00 Bulk freight 0.124 0.000** -0.002 0.431 0.003 0.469 -0.006 0.632 0.001 0.798 0.02 -0.02 Container freight 0.222 0.000** -0.010 0.048* 0.001 0.869 0.043 0.060 0.006 0.186 0.08 0.05 Oil production 1.526 0.000** -0.002 0.949 -0.047 0.452 0.088 0.585 -0.019 0.551 0.01 -0.02 Steel production 0.220 0.000** 0.000 0.949 0.002 0.826 0.015 0.523 0.000 0.933 0.00 -0.03 World export 0.632 0.000** -0.010 0.482 -0.011 0.645 -0.052 0.384 0.009 0.463 0.02 -0.0227 6. Conclusion
28
exposure coefficient of stock returns to USD value expressed in local currency cannot be explained by existing theory.
References
Adler, M. and Dumas B., 1984, Exposure to Currency Risk: Definition and Measurement, Financial Management 13, 41 – 50
Akatsuka, K. and Leggate, K., 2001, Perceptions of foreign exchange rate risk in the shipping industry, Maritime Policy & Management 28, 235 – 249
Bhandari, L.C., 1988, Debt/Equity Ratio and Expected Common Stock Returns: Empirical Evidence, The Journal of Finance 43, 507 – 528
Bartram, S.M., 2002, The Interest Rate Exposure of Nonfinancial Corporations, Working Paper, Maastricht University
Bartram, S.M. and Bodnar, G.M., 2009, Crossing the Lines: The Conditional Relation between Exchange Rate Exposure and Stock Returns in Emerging and Developed Markets, Working Paper, Lancaster University and John Hopkins University
Beenstock, M. and Vergottis, A., 1989, An econometric model of the world market for dry cargo freight and shipping, Applied Economics, 21, 339-356
Choi, J.J. and Jiang, C., 2009, Does multinationality matter? Implications of operational hedging for the exchange risk exposure, Journal of Banking & Finance 33, 1973 – 1982
Davis, L., 1994, The Cross-Section of Realized Stock Returns: The Pre-COMPUSTAT Evidence, The Journal of Finance 49, 1579 – 1593
Drobetz, W., Schilling, D. and Tegtmeier., L, 2010, Common risk factors in the returns of shipping stocks, Maritime Policy & Management 37, 93 – 120
El-Masry, A. and Abdel-Salam, O., 2007, Emeral Article: Exchange rate exposure: do size and foreign operations matter?, Managerial Finance, 33, 741 - 765
29
Faff, R.W. and Brailsford, T.J., 1999, Oil price risk and the Australian stock market, Journal of Energy Finance and Development 4, 69 – 87
Forrest, M., Kavussanos M.G. and Skielse, A.J., 2003, International comparison of market risks across shipping-related industries, Maritime Policy & Management, 30, 107 – 122
Géczy, C., Minton, B.A., Schrand, C., 1997, Why Firms Use Currency Derivatives, The Journal of Finance, 52, 1323 - 1354
Grammenos, C.T. and Arkoulis, A.G., 2002, Macroeconomic Factors and International Shipping Stock Returns, International Journal of Maritime Economics 4, 81 – 99
Guay, W., 1999, The impact of derivatives on firm risk: an empirical examination of new derivatives users, Journal of Accounting & Economics 26, 319 – 351
He, J. and Ng, L.K., 1998, The foreign exchange exposure of Japanese multinational corporations, The Journal of Finance, 53, 733 – 753
Hou, K. and van Dijk, M.A., 2008, Resurrecting the Size Effect: Firm Size, Profitability Shocks, and Expected Stock Returns, Working Paper, Ohio State University and Erasmus University
Jorion, P., 1990, The Exchange-Rate Exposure of U.S. Multinationals, The Journal of Business 63, 331 – 345
Kilian, L., 2009, Not All Oil Price Shocks Are Alike: Disentangling Demand and Supply Shocks in the Crude Oil Market, American Economic Review 99, 1053 – 1069
Lintner, J., 1965, The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets, The Review of Economics and Statistics 47, 13 – 57
Lusk, E.J., Halperin, M. and Heilig, F., 2011, A Note on Power Differentials in Data Preparation between Trimming and Winsorizing, Business Management Dynamics 1, 23 – 31
Kavussanos, M.G. and Marcoulis, S.N., 2005, Transportation and other industries, Research in Transportation Economics 12, 107 – 142
Miller, J.I. and Ratti, R.A., 2009, Crude Oil and Stock Markets: Stability, Instability, and Bubbles, Working Paper, University of Missouri
30
Prasad, A.M. and Rajan, M., 1995, The Role of Exchange and Interest Risk in Equity Valuation: A Comparative Study of International Stock Markets, Journal of Economics and Business 47, 457 – 472
Randers, J. and Göluke, U., 2007, Forecasting turning points in shipping freight rates: lessons from 30 years of practical efforst, System Dynamics Review, 23, 253 – 284
Sadorsky, P., 2001, Risk factors in stock returns of Canadian oil and gas companies, Energy Economics 23, 17 – 28
Sadorsky, P., 2008, The oil price exposure of global oil companies, Applied Financial Economic Letters 4, 93 – 96
Samitas, A. and Tsakalos, I., 2010, Hedging effectiveness in shipping industry during financial crises, International Journal of Financial Markets and Derivatives 1, 196 – 212
Sharpe, W.F., 1964, Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk, The Journal of Finance 19, 425 – 442
31 Appendices
Appendix A - List of firms whose stock returns are subject to analysis in this study
nr. Firm Name Subsector Domestic
Currency
nr. Firm Name Subsector Domestic
Currency
1 Diana Shipping Inc Bulk USD 58 Chinese Maritime Transport Ltd Other TWD
2 DryShips Inc Bulk USD 59 Chu Kong Shipping Development Co
Ltd
Other CNY
3 Eagle Bulk Shipping Inc Bulk USD 60 Chuan Hup Holdings Ltd Other SGD
4 Excel Maritime Carriers Ltd Bulk USD 61 Cia Sud Americana de Vapores SA Other CLP
5 Algoma Central Corp Bulk CAD 62 Clarkson PLC Other GBP
6 Chowgule Steamships Ltd Bulk INR 63 Cosco International Holdings Ltd Other CNY
7 Courage Marine Group Ltd Bulk SGD 64 COSCO Shipping Co Ltd Other CNY
8 Cie Maritime Belge SA Bulk EUR 65 Daito Koun Co Ltd Other JPY
9 Daiichi Chuo KK Bulk JPY 66 Deep Sea Supply PLC Other NOK
10 Pacific Basin Shipping Ltd Bulk HKD 67 Dfds A/S Other DKK
11 U-Ming Marine Transport Corp Bulk HKD 68 DOF ASA Other NOK
12 U-SEA Bulk Shipping A/S Bulk DKK 69 Ezra Holdings Ltd Other SGD
13 Wilson ASA Bulk GBP 70 Farstad Shipping ASA Other NOK
14 Abetrans Ltd Container ILS 71 Norwegian Car Carriers ASA Other NOK
15 Alexander & Baldwin Inc Container USD 72 NS United Kaiun Kaisha Ltd Other JPY
16 Seaspan Corp Container USD 73 Ocean Wilsons Holdings Ltd Other BRL
17 China Shipping Container Lines Co Ltd
Container CNY 74 Pakistan National Shipping Corp Other PKR
18 Empresas Navieras S.A. Container CLP 75 Pelayaran Tempuran Emas Tbk PT Other IDR
19 Evergreen Marine Corp Taiwan Ltd
Container TWD 76 Precious Shipping PCL Other THB
20 Orient Overseas International Ltd
Container CNY 77 Premuda SpA Other EUR
21 PDZ Holdings Bhd Container MYR 78 Primorsk Shipping Corp Other RUB
22 Regional Container Lines PCL Container THB 79 Rederi AB Transatlantic Other SEK
23 Tianjin Marine Shipping Co Ltd Container CNY 80 Rig Tenders Indonesia Tbk PT Other IDR
24 Wan Hai Lines Ltd Container TWD 81 Rinko Corp Other JPY
25 Euronav NV Tanker USD 82 Sado Steam Ship Co Ltd Other JPY
26 Overseas Shipholding Group Inc Tanker USD 83 Salam International Transport & Trading
Other JOD
27 Teekay Corp Tanker USD 84 Samudera Shipping Line Ltd Other IDR
28 Berlian Laju Tanker Tbk PT Tanker SGD 85 Sanbumi Holdings BHD Other MYR
29 CLH SA Tanker EUR 86 Scomi Marine Bhd Other MYR
30 Concordia Maritime AB Tanker SEK 87 Sea Star Capital PLC Other CYP
31 CSC Nanjing Tanker Corp Tanker CNY 88 Shinwa Naiko Kaiun Kaisha Other JPY
32 DHT Holdings Inc Tanker NOK 89 Shipping Corp of India Ltd Other INR
33 Exmar NV Tanker EUR 90 Singapore Shipping Corp Ltd Other SGD
34 Nordic Tankers A/S Tanker DKK 91 Sloman Neptun Schiffahrts AG Other EUR
35 Odfjell SE Tanker NOK 92 SOC Comercial Orey Antunes SA Other EUR
36 Qatar Navigation Tanker QAR 93 Solstad Offshore ASA Other NOK
37 Solvang ASA Tanker NOK 94 SRAB Holding AB Other SEK
38 Stolt-Nielsen Ltd Tanker NOK 95 Star Reefers Inc Other GBP
39 Varun Shipping Co Ltd Tanker INR 96 STX Pan Ocean Co Ltd Other KRW
40 B+H Ocean Carriers Ltd Other USD 97 Sumatec Resources Bhd Other MYR
41 Rand Logistics Inc Other USD 98 Taiwan Navigation Co Ltd Other TWD
42 Ship Finance International Ltd Other USD 99 Takase Corp Other JPY
43 TBS International PLC Other USD 100 Tallink Group PLC Other EEK
44 Tidewater Inc Other USD 101 Tamai Steamship Co Ltd Other JPY
45 Anek Lines SA Other EUR 102 Thoresen Thai Agencies PCL Other THB
46 AP Moller - Maersk A/S Other DKK 103 Toei Reefer Line Ltd Other JPY
47 Attica Holdings SA Other EUR 104 Tokai Kisen Co Ltd Other JPY
48 Belships ASA Other NOK 105 Torm A/S Other DKK
49 Bongshin Co Ltd Other KRW 106 Touax SA Other EUR
50 Bonheur ASA Other NOK 107 Trencor Ltd Other ZAR
51 Borgestad Other NOK 108 Tsakos Energy Navigation Ltd Other EUR
52 Camillo Eitzen & Co ASA Other NOK 109 United Arab Shipping Other KWD
53 Canal Shipping Agencies Co Other EGP 110 Vallianz Holdings Ltd Other SGD
54 Chang Jiang Shipping Group Phoenix Co Ltd
Other CNY 111 Viking Line Abp Other EUR
55 China COSCO Holdings Co Ltd Other CNY 112 Wilh Wilhelmsen ASA Other NOK
56 China Shipping Development Co Ltd
Other CNY 113 Yang Ming Marine Transport Corp Other TWD
32 Appendix B - OECD member countries
Australia France Japan Portugal
Austria Germany Korea Slovak Republic
Belgium Greece Luxembourg Slovenia
Canada Hungary Mexico Spain
Chile Iceland Netherlands Sweden
Czech Republic Ireland New Zealand Switzerland
Denmark Israel Norway Turkey
Estonia Italy Poland United Kingdom
Finland
Appendix C - Correlation coefficients exchange rate time series with other time series independent variables
USD expressed in: Market
Return
Interest Rate Oil Price Oil Production Steel Production World Export
Indian Rupee 0,411 0,025 0,360 0,051 0,169 0,235
Singapore Dollar 0,404 -0,076 0,436 0,107 0,123 0,501
Hong Kong Dollar -0,035 0,026 0,018 0,057 -0,106 -0,118
Taiwan Dollar 0,327 0,083 0,485 0,059 0,124 0,424
British Pound -0,118 0,203 0,607 0,020 0,325 0,541
Israeli New Shekel 0,299 0,008 0,357 -0,105 0,266 0,469
Canadian Dollar 0,417 0,209 0,620 -0,135 0,203 0,333
Norwegian Kroner 0,339 0,213 0,701 -0,002 0,324 0,408
Chinese Yuan Renminbi -0,004 -0,030 0,192 0,133 -0,027 0,210
Chilean Peso 0,215 -0,158 0,399 -0,011 0,196 0,206 Euro 0,337 -0,203 0,456 -0,074 0,141 0,393 Japanese Yen -0,152 -0,461 -0,356 0,031 -0,141 -0,121 Swedish Krona 0,465 0,014 0,588 0,039 0,249 0,478 Thai Baht 0,225 -0,017 0,236 -0,049 0,072 0,080 Danish Krone 0,339 -0,200 0,454 -0,054 0,140 0,393 Indonesian Rupiah -0,127 0,063 0,121 0,122 0,018 0,068 Egyptian Pound 0,202 -0,212 0,264 0,196 -0,063 0,167 Pakistan Rupee 0,070 -0,214 0,148 -0,024 0,038 -0,012 Malaysian Ringgit 0,395 -0,070 0,300 0,126 0,200 0,372 Russian Rouble 0,276 -0,044 0,560 0,249 0,178 0,596 Qatari Rial -0,020 -0,106 0,019 0,034 0,022 -0,078 Jordanian Dinar 0,184 0,131 0,053 0,079 0,060 0,043 Cyprus Pound 0,324 -0,209 0,450 -0,084 0,139 0,413 Estonian Kroon 0,349 -0,194 0,456 -0,086 0,148 0,350
South African Rand 0,540 0,097 0,454 0,060 0,030 0,272
South-Korean Won 0,595 0,030 0,332 0,044 0,194 0,471
Brazilian Real 0,416 0,298 0,656 0,249 0,197 0,379
