Dynamics of the co-movement between stock and maritime markets

International Review of Economics and Finance 25 (2013) 282–290

Contents lists available at SciVerse ScienceDirect

International Review of Economics and Finance journal homepage: www.elsevier.com/locate/iref

Dynamics of the co-movement between stock and maritime markets Oral Erdogan a,⁎, Kenan Tata b, B. Can Karahasan c, M. Hakan Sengoz a a b c

Istanbul Bilgi University, Turkey Turkon Holding, Turkey Okan University, Turkey

a r t i c l e

i n f o

Article history: Received 27 December 2010 Received in revised form 20 May 2012 Accepted 13 July 2012 Available online 20 July 2012 JEL classifications: C32 G11 G15

a b s t r a c t This study demonstrates the existence of economically significant information spillovers between stock markets and markets for shipping freight by sea. Using multivariate correlation models on the returns of the Dow Jones Industrial Average (DJIA) and the Baltic Dry Index (BDI), we find mutual feedback between the two markets, which becomes stronger during the periods of financial turmoil. Results also suggest that the extent of information spillover between the markets varies over time, depending on market-specific conditions. We conclude that, being an indispensable factor for price discovery, such a relationship provides a link between two markets that are otherwise rather distinct with respect to the assessment of available information and real activity. © 2012 Elsevier Inc. All rights reserved.

Keywords: Stock markets Maritime markets Financial crisis Multivariate volatility modeling

1. Introduction It has been already validated that agents in financial markets take the linkage between financial and economic activity as given and form their investment decisions accordingly. However, contemporary developments in financial markets, especially after the most recent global financial turmoil, make it necessary to ask updated questions which may add valuable information for investors. Basically, investors' decision function compares the expected and realized returns of the assets they own. Among numerous endogenous factors affecting investors' decision function, interaction between different markets and asset classes is an emerging and important issue both on macro and micro levels. Every signal that contains information about the possible interdependencies and spillovers between the real economy and the capital markets will be valuable. In that regard, interaction of stock markets and markets for shipping freight by sea (i.e. maritime markets), respectively as a benchmark for asset prices and a proxy for assessing the path of real economic activity enters the realm of this study. Intuition regarding the importance of the proposed interdependency originates from the strong connection between financial development and economic activity. The insight of McKinnon (1973) and Shaw (1973) hypotheses and contemporary studies such as Bencivenga, Smith, and Starr (1996) and Levine and Zervos (1998), all formulated why finance matters for the real economy. On contrary, Ross (1976), later Roll and Ross (1980) and Chen, Roll, and Ross (1986) explained that although financial development affects the real economy at the macro level, there are also factors coming from the real economy that is influencing the asset prices directly and the capital markets indirectly. Yet, the common property of both approaches is observing the link from a macro perspective without focusing on the dynamics of real economy. Therefore, the general link between finance and ⁎ Corresponding author at: Santral Kampusu, K. Karabekir C. No:2/13 Eyup, Istanbul 34060, Turkey. Tel.: +90 2123115000; fax: +90 2123117776. E-mail address: [email protected] (O. Erdogan). 1059-0560/$ – see front matter © 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.iref.2012.07.007

O. Erdogan et al. / International Review of Economics and Finance 25 (2013) 282–290

283

economic growth needs to be expanded by focusing on more specific areas that are expected to contain the information set that we regard as factors affecting the investors' decision function. Co-movements without a restriction on the exogeneity have been investigated in numerous influential studies. The primary focus is on understanding and estimating the linkage of two or more financial markets while recent studies on co-movement analysis have focused on the combinations of stock prices on one hand and commodity prices, exchange rates or equity markets on the other (see Antoniou, Pescetto, & Violaris, 2003; Bartram, Taylor, & Wang, 2007; Bhar & Hammoudeh, 2011; Kim, Moshirian, & Wu, 2006; Lin, 2012; Martens & Poon, 2001; Savva, 2009; Wahab, 2012). Specifically, studies with methodologies similar to our study; Erdogan and Schmidbauer (2005) for currency and stock markets, Chiang, Jeon, and Li (2007) for a number of Asian stock markets, Li and Zou (2008) for Chinese capital markets (bond vs. stock markets), Savva, Osborn, and Gill (2009) for US and European stock markets, Lin, Menkveld, and Yang (2009) for Chinese and Western capital markets, Aslanidis, Osborn, and Sensier (2010) for US and UK capital markets, and Syllignakis and Kouretas (2011) for Central and Eastern Europe (CEE), US, German and Russian stock markets, have all remarked that the markets in question experienced a process of co-movement in varying degrees. In our belief, all findings are crucial, yet should be evaluated carefully. The presence of high level of co-movement mainly suggests a lesser degree of efficiency in portfolio diversification while the efficiency of portfolio construction is a clustered motivation for most of the aforementioned studies. Furthermore, findings of high level of co-movements should signal the contagion effects evident for the bad as well as good events spreading over different markets. Therefore, the possible high correlation between the markets under investigation will contain the previously mentioned concern about how a finance or real economy related event can spread over the financial markets and real economy. In both cases, the result will be the evolution of the information set that should be valuable for investors' decision function in an area which is relatively less explored in the literature. In order to observe the interconnection between financial markets and real economy, international trade will be our benchmark to assess the developments in the real side of the economy. Financial development and international trade are endogenously correlated with economic growth from separate channels (for international trade and growth issue, see Barro, Mankiw, & Sala-I-Martin, 1995; Coe & Helpman, 1995; Grossman & Helpman, 1994; Rivera-Batiz & Romer, 1991; for finance and economic growth linkages, revisit Bencivenga et al., 1996; Chen et al., 1986; Levine & Zervos, 1998; Roll & Ross, 1980). One way to talk about the path of the international trade is to go over the trade volumes. Moreover, as already discussed by Krugman (1991), focusing on the transportation costs will yield more accurate information about the international trade developments. Among different channels, by concentrating on the seaborne trade we discuss how transportation costs in maritime markets act as a representative measure for the international trade 1 and how the uncertainty in financial side of the economy is correlated with real side of the economy through maritime markets. Within this approach, markets for shipping freight by sea entered the agenda of numerous studies where only the risk structure of the market is investigated across different transportation markets and time spans (Erdogan, 1996, 1997; Goulielmos, 2008; Kavussanos, 1996; Kavussanos & Nomikos, 1999). On the other hand, it seems more meaningful to use a general representative index to control for the asset prices. Eventually, we focus on the contagion effect between these two markets. For this purpose, DJIA represents the stock markets while BDI stands for the representation of maritime markets. Given such a framework regarding the interdependency between these two markets; any hints about the structure of the relationship between financial markets and real economy should be providing information for investors. Therefore, this study applies a multivariate framework without making a restriction on the exogeneity of real and financial side of the economy. It is noteworthy to remark that such an objective is highly influenced as well as inspired by the recent concerns of Tong and Wei (2008), and Reinhart and Rogoff (2009). These studies regard financial side of economy as an exogenous factor of the process and explain the spillover of the financial crisis from financial markets towards the real economy through employment and output growth. Although we prefer to approach the subject from a different perspective, allowing financial markets to be endogenous, an emphasis on the effects of the most recent global financial turmoil is an additional motivation to examine the co-movements between these two markets. Then, we evaluate the endogenous interaction between stock markets and maritime markets to obtain valuable hints about not only the univariate behavior but also the dynamic behavior lying in the core of certain events. The responsiveness to internal and external shocks leads to complexity of the dynamics of the relationship between the two markets. Thus, a deeper understanding requires an analysis far beyond a simple univariate approach. Interactions on return level, such as a vector autoregressive setup, are far from being able to reveal the required risk dynamics. While modeling volatility has been subject to increasing attention by using univariate autoregressive conditional heteroscedasticity family models, the co-movements among markets started to gain popularity in recent empirical studies (see Bauwens, Laurent, & Rombouts, 2006). Moreover, Bollerslev, Engle, and Nelson (1994) underline the efficiency gains in using multivariate models to observe volatility and co-movements. Although the multivariate generalized autoregressive conditional heteroscedasticity (M-GARCH) models are fashionable and informative, the number of parameters increases along with the dimension of the model as the specification and estimation procedures of M-GARCH models become more complex. Our special interest is on those models allowing for a dynamic setup so that the risk spillover effects can be observed in a timely manner. Accordingly, this study aims to unfold a dynamic relationship among two markets which is beneficial for both markets' participants through realizing an informative feedback. First, a meaningful co-movement pattern with a term structure provides that maritime 1 As discussed by Hummels (1999), the seaborne trade represents more than eighty percent of the world trade transportation in volume and more than half in value. Hence, among different modes of transportation, we believe that focusing on the maritime transportation will yield more information with respect to other modes of transportation.

284

O. Erdogan et al. / International Review of Economics and Finance 25 (2013) 282–290

markets provide information for stock markets. Second, stock markets contain information to clear the maritime freight markets which in return may provide a better freight rate discovery as well. Finally, the study outlines that maritime market indicators would be employed as control variables in asset pricing models for further studies. Following this section, Section 2 introduces the data and the methodological aspects. Then the results of the empirical analysis will be given while the last section of the study concludes. 2. Data and methodology The co-movement between the stock markets and maritime markets is investigated by using DJIA and BDI, respectively. Baltic Dry Index, the traditional leading indicator of maritime markets, comprises the indexes for four sub-categories of vessels, classified in terms of the vessel size, of dry cargo. These size-related components of the index have equal weights and it is calculated according to the daily rate of chartering a vessel (i.e. time charter rates). To simply put, a shipper who is in need of transportation service of certain types of goods may inquire about the cost and delivery time of shipping those goods and this inquiry may be through one-to-one business networks, online systems or intermediation, which is a more common method, of brokerage and/or forwarder systems. These chartering rates are obtained through a collection process and do not necessarily represent the cost/revenue of moving a certain type of good from locations A to B. In fact, it is the daily price for a vessel to be hired and, as the technicality of vessels bound them to operate in markets of certain products with certain departures and destinations, these rates represent a necessary input for the shippers and a valuable significant input for the dynamics of real economic activity. The rates are announced through fixtures as the vessels are fixed between a shipper and a ship-owner itself or a charterer, who hires the ship to operate. These fixtures are the main source of collecting the data for the freight rates but if no fixture is established for a certain route, the panelists of the intermediation agencies, who essentially quote the fixture data in the Baltic Exchange, submit their estimations for a daily rate of chartering vessels of the relevant routes. Eventually, these figures are calculated and the sub-indexes are formed. As these sub-indexes are calculated, BDI is also set as an arithmetic average of those sub-indexes. Some dynamics of BDI bear unique properties that are not available in other markets. For instance, fixtures for the transportation of dry cargo essentially represent the demand for raw materials 2 rather than direct consumption as it is mainly attributed to containerized seaborne trade. In addition, the economic activity reflecting the need for allocation of resources directly impacts the price of the contracts nominating the shipping cost for the near future. Hence, the demand for raw materials due to private and/or governmental policies serves as an important leading indicator. Moreover, the selection of a number of routes worldwide enhances the usage of such indexes globally, compared to other local based alternatives. 3 And last but not least, the supply of vessels has a unique structure. Supplying ships is not easy as demanding them. Over or under supply may have some long term effects on the indexes (Erdogan, 2008). Especially, during and after the periods of financial turmoil, the idle fleet due to contraction in global demand may have persistent negative effects on shipping rates. This is related with slow utilization of the fleet during the revival of the global demand and continuation of shipbuilding activities assisted by governments. Also, the competition of shipping firms over routes and efforts to sustain market shares at any cost drive rates down even more. For BDI data, the daily and weekly series are obtained from Clarksons Shipping Intelligence Network for the period between November 1999 and January 2012. This makes weekly and monthly closing prices up to 618 and 147 observations for each frequency, respectively for both BDI and DJIA. In addition, the monthly classification of crisis periods; business cycles (i.e. contractions) announced by National Bureau of Economic Research (NBER), is used as a control variable in an attempt to explain the possible clustered structure of the estimated dynamic conditional correlation (DCC). In order to exploit the nature of the co-movement between the two markets, a simple methodology accounting for non-linear effects is employed. The procedure involves main three steps; modeling the first and second moments and specification of the correlation. In the scope of M-GARCH modeling, we apply the procedure suggested by Engle (2002). The general representation can be expressed as; yt ¼ μ t þ ε t

ð1Þ

where yt = (BDIt, DJIAt)′, μt is the VAR (vector auto-regressive) specification including a constant and the residuals, and εt ~ N(0, Ht) are error terms conditional on the information available at time t − 1. The diagonals of Ht, hii,t, may be referred as the univariate specification of each dependent variable and the off-diagonal element h12,t as the covariance. Moreover, following pffiffiffiffiffiffiffiffi Engle (2002), Ht can be expressed as a decomposition of the form DtRtDt, where Dt is a diagonal matrix consisting of hii;t 's and Rt is the correlation matrix. Engle's (2002) suggestion for the correlation is of the form;  ð1−α−βÞ þ αε1t−1 ε2t−1 þ βqij;t−1 qij;t ¼ ρ

ð2Þ

2 BDI is an equally weighted index for four sub-categories of vessels (in terms of size) carrying dry cargo and dry cargo refers mainly to commodity raw materials (i.e. iron ore, coal, grains, bauxite/aluminum, and phosphate rock). The largest vessel size included in the BDI calculation is called as “Capesize” which has 100,000–180,000 tons deadweight (DWT), and navigates only on deep waters and serves on deep water ports/terminals. 3 A reference for such an index is ISTFIX to represent time charter rates for small size vessels.

O. Erdogan et al. / International Review of Economics and Finance 25 (2013) 282–290

285

and qij;t ρij;t ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi : qii;t qjj;t

ð3Þ

After calculating the conditional correlation, tests suggested by Bollerslev (1990), Tse (2000) and Bera and Kim (2002) are carried out in order to check the constancy of the conditional correlation. pffiffiffiffiffiffiffiffi Bollerslev (1990) compares the Ljung–Box statistic of the cross-product ε1t*ε2t*. These are the standardized residuals (εit = hii;t ) obtained from a univariate GARCH(1,1) estimations for each independent variable. On the other hand, Tse (2000) suggests an LM statistic to test the null hypothesis H0 : δ = 0, where δ is the parameter in;  þ δε1t−1 ε2t−1 : ρij;t ¼ ρ

ð4Þ

Finally, Bera and Kim (2002) developed an Information Matrix (IM) statistic computed as; " IM ¼

#  2 T  P ðv1t v2t Þ2 −1−2ρ2

i¼1

  4T 1 þ 4ρ2 þ ρ2

;

ð5Þ

Weekly Data 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0

Monthly Data 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0

Fig. 1. Level series for DJIA and BDI. (Dashed lines represent BDI; the shaded region on monthly figure is NBER Business Cycles (i.e. contractions)).

286

O. Erdogan et al. / International Review of Economics and Finance 25 (2013) 282–290

Table 1 Summary statistics (log returns). Monthly

Number of obs. Average Maximum Minimum Standard dev. Skewness Kurtosis Jarque–Bera Uncon. corr.

Weekly

DJIA

BDI

DJIA

BDI

147 0.0012 0.101 −0.152 0.045 −0.60 3.84 13.15*** 0.158

147 −0.0048 0.71 −1.297 0.23 −1.58 11.27 480*** –

618 0.00017 0.131 −0.125 0.0269 −0.106 6.2 265*** 0.0034

618 −0.00024 0.502 −0.415 0.0749 −0.252 9.7 1178*** –

Notes: *, ** and *** indicate significance levels of 10%, 5% and 1%, respectively.

  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi where vit ¼ εit −ρ εjt = 1−ρ2 , ρ* = (ε1t*ε2t*)/T and εit* are the standardized residuals from Bollerslev (1990) constant conditional correlation (CCC) model. The statistic is distributed as χ 2 with 1° of freedom. Following the constancy tests, we obtain the model depending on Eqs. (2) and (3). Furthermore, to check the effect of recession times, a simple regression model is estimated with respect to Aydemir (2008) and Cai, Yeutien Chou, and Li (2009), who observed an increase in the interdependence among markets during recession times. As explanatory variables, following Syllignakis and Kouretas (2011), the estimated conditional variances of DJIA and BDI and the categorical “crisis” variable are regressed with a modification. When the correlation deviates from its long time average, the fluctuations may go both positive and negative with similar amplitudes. The conditional variances and the “crisis” variable on the other hand can approach zero at most on the downside. This may lead to an underestimation of the impact of recession periods on the variability of the correlation coefficient. As a result, the dependent variable used in the analysis is the absolute value of the deviation from the mean. 3. Empirical findings and implications The monthly and weekly log returns of DJIA and BDI series are available in Fig. 1 and the summary statistics for series are presented in Table 1. Our first remark is that weekly and monthly BDI returns experience significantly larger fluctuations compared to DJIA returns. Additionally, both series exhibit negative skewness and excess kurtosis both on weekly and monthly observations. Indeed, Jarque–Bera test results are in line with the aforementioned deviations from the normality and the unconditional correlation is larger on the monthly sample whereas it is close to zero on the weekly sample. Instead of proceeding with univariate models separately, VAR specification on the mean of the model is preferred. As mentioned before, it allows for a more “loose” approach to constraints on variables' endogeneity and a broader information feedback from previous observations of both series. VAR models are estimated up to 15 lags and based on Akaike Information Criteria (AIC), Schwarz Information Criteria (SIC) and parsimony in case of ambiguity, the relevant lag lengths of the mean equations are determined. 4 Results indicate that the lag lengths of the models should be 3 for the monthly series and 2 for the weekly series, respectively. Both monthly and weekly series reveal different feedback results regarding the effects of the lagged variables. Table 2 clearly indicates that the previous changes in BDI help to explain the changes in DJIA in the longer term, while a similar effect is not evident in the short term. Just the opposite holds for the effects of DJIA on BDI in the weekly series. Besides, BDI is more helpful in terms of self-explanatory power compared to that of DJIA. To sum up, the feedback between the markets follows different paths in the short and long terms. The short term fluctuations in the financial markets tend to add to the information set of maritime market participants, whereas such a feedback is not evident in the longer term. BDI on the other hand, seems to add explanation to the fluctuations in the financial markets in the longer term. This may be attributed to the relatively slower update of macroeconomic phenomena, which are presumably reflected in the freight rates as discussed before. An important check before proceeding with DCC estimations is a couple of tests against the constancy of the correlation coefficient. Results in Table 3 suggest that the weekly data does not reveal a lot in this respect and the results other than the Bollerslev test are weak whereas monthly results seem promising. While Bollerslev test allows rejecting the null hypothesis at 1%, Tse test and Bera and Kim test achieve this at 10% and 5%, respectively. It is also worth pointing that the test procedure by Tse (2000) bears some weakness with samples less than 500 observations, which is the case for our monthly sample. In spite of weak test results in the weekly sample, series with both frequencies are used in the DCC estimation procedure. Monthly and weekly samples exhibit different results regarding the behavior of the correlation coefficient. The results summarized in Panel C of Table 4 indicate a strong dependence of the correlation on the cross product of the residuals from the previous period (measured by the estimate of α) while both cases fail to show any dependence on the previous realizations of the correlations (measured by the estimate of β). Weekly observations, on the other hand do not reveal a similar behavior as this finding also supports the test outcomes available in Table 3. 4 Prior to the VAR estimation, Johansen (1988) cointegration tests on the price level are also performed. However, these results do not reveal any evidence in that direction, so VAR model for a linear interdependence is considered adequate.

O. Erdogan et al. / International Review of Economics and Finance 25 (2013) 282–290

287

Table 2 Estimation results for the weekly and monthly VAR specifications.  The model estimated is of the form

DJIAt BDI t



 ¼

  



ε c1 þ ½A DJI t−i;i¼1;2;3 þ ½B BDI t−i;i¼1;2;3 þ 1t , where A and B are 2 × 3 matrices for monthly and 2 × 2 c2 ε2t

matrices for weekly data. Constants Monthly   DJIAt BDI t Weekly   DJIAt BDI t

2

3 0:002 6 ð0:004Þ 7 4 −0:005 5 ð0:018Þ 2

3 0:0002 6 ð0:001Þ 7 4 −0:004 5 ð0:003Þ

[A] 2

0:038 6 ð0:084Þ 6 4 0:125 ð0:428Þ

[B]

−0:175 ð0:082Þ 0:729 ð0:421Þ

2

−0:11 6 ð0:04Þ 6 4 0:267 ð0:101Þ

3 0:064 ð0:083Þ 7 7 −0:263 5 ð0:423Þ

3 −0:002 ð0:041Þ 7 7 0:189 5 ð0:102Þ

2

0:051 6 ð0:018Þ 6 4 0:358 ð0:092Þ 2

−0:005 ð0:019Þ −0:181 ð0:095Þ

−0:003 6 ð0:016Þ 6 4 0:365 ð0:04Þ

3 0:042 ð0:018Þ 7 7 −0:005 5 ð0:093Þ

3 0:015 ð0:016Þ 7 7 0:112 5 ð0:04Þ

Notes: *, ** and *** indicate significance levels of 10%, 5% and 1%, respectively. The standard errors of the estimates are in brackets. The decision on the lag length is given upon information criteria like AIC, SC and HQ.

Fig. 2 shows the plots for conditional correlation series of monthly and weekly estimations. Both correlation series follow a bi-directional pattern and monthly correlations take values between approximately − 24% and 74% while weekly correlations differ between approximately − 18% and 29%. In addition, the plots of the conditional correlation suggest an additional check against the cluster of periods with high level of correlation. Due to the nature of the categorical “crisis” variable and the relative constancy of the weekly correlation, we investigated only monthly series. Two models are employed in line with our purpose of explaining the clusters with “crisis” related variables. The first model (Model 1 in Table 5) takes conditional variances to explain the absolute deviations from the long run average of conditional correlation. The second model (Model 2 in Table 5) takes NBER “crisis” dummy variable into account with the same purpose. According to the estimation results in Table 5, the conditional variances and the “crisis” dummy variable tend to add to the explanation of the deviations from the long run average of conditional correlation. Furthermore, we attempt to model all variables (conditional variances and NBER “dummy” variable) together in a single equation. Since there is an overlap for the cluster of periods (also evident from observing Fig. 1) with high volatility in series (i.e. conditional variances) and the categorical variable of “crisis” (denoted by shaded region in Fig. 2), such specified model is omitted. Several implications for the participants of both markets can be derived from the outcomes discussed above as several mechanisms are already suggested in the literature regarding the effects of real economy on the stock markets. As a further discussion, we present a practical note on the effect of findings on both markets. Our methodological view contributes in such a way that there is a relationship between two markets with a nonsynchronous structure. In other words, both markets provide informative feedbacks with different lag structure. For instance, VAR findings on monthly sample suggested that the first and third lags of BDI significantly help to explain DJIA while the findings on weekly sample suggested that the first and second lags of DJIA significantly help to explain BDI. Additionally, for financial markets, we argue that trade patterns and freight rates yield information to market participants on a longer term (i.e. one or three months). Moreover, we suggest that freight rates would be counted as a factor for pricing financial assets. On the other hand, for maritime markets, comparatively short-term relationship (i.e. one to two weeks) may be beneficial for better predictions. It should be noted that the inner dynamics of the maritime markets play a significant role for a complete assertion of the benefits of such relationship. This may also be due to the maturity of contracts, and the type of cargo or both. We also attempt to unravel the contribution and possible explanations of the methodology which would provide additional practical insight. The most common attribution of dynamic co-movement analysis is seeking significant increase in pair-wise correlations of samples in question with a unidirectional characteristic. On contrary, in our study the conditional correlation between DJIA and BDI exhibits a significant bi-directional pattern. As there would be numerous factors that may explain this

Table 3 Tests against the constancy of the conditional correlation.

Bollerslev test Tse test Bera–Kim test

Monthly (147 obs)

Weekly (618 obs)

Q statistic (4 lags) = 15.55 (0.004)⁎⁎⁎ T-stat = 1.697 (0.092)⁎

Q statistic (4 lags) = 8.12 (0.087)⁎ T-stat = 0.022 (0.98) Chi square = 0.26 (0.74)

Chi square = 6.06 (0.013)⁎⁎

Notes: The values in parentheses are the relevant p-values. *, ** and *** indicate significance levels of 10%, 5% and 1%, respectively.

288

O. Erdogan et al. / International Review of Economics and Finance 25 (2013) 282–290

Table 4 Estimation results for the bivariate GARCH models. 

    



ε DJIAt c ¼ 1 þ ½A DJIt−i;i¼1;2;3 þ ½B BDI t−i;i¼1;2;3 þ 1t , where A and B are 2 × 3 for monthly and 2 × 2 matrices for BDI t c2 ε2t 2 + biihii,t−1 for each variable. The conditional correlation is estimated weekly data. The conditional variance on the other hand is of the form hii,t = cii + aiieii,t−1 qij;t  p ffiffiffiffiffiffiffiffiffiffi ffi via qij;t ¼ ρ ð1−α−β Þ þ αε 1t−1 ε2t−1 þ βqij;t−1 and ρij;t ¼ q q : The mean equation is of the form

ii;t jj;t

Panel A: mean equations [A]

Constants 2

Monthly

Weekly

3

0:003 6 ð0:006Þ 7 4 0:004 5 ð0:028Þ 2 3 0:002 6 ð0:0008Þ 7 6 7 4 0:001 5 ð0:001Þ

Weekly

Monthly

Weekly

−0:038 −0:187 0:068 ð0:096Þ ð0:095Þ 7 6 ð0:083Þ 4 0:058 0:697 −0:321 5 ð0:256Þ ð0:487Þ ð0:740Þ 2 3 −0:056 −0:067 ð0:047Þ 7 6 ð0:040Þ 4 0:070 0:032 5 ð0:059Þ ð0:060Þ

0:0012 6 ð5:4e−05Þ 7 6 7 4 0:0419 5 ð0:0064Þ 2 3 3:8e−06 6 ð1:5e−05Þ 7 6 7 4 8:1e−06 5 ð8:4e−06Þ

α

Panel C: conditional correlation parameters β

2



0:177 ð0:0238Þ 0:031 ð0:045Þ

3

[B] 3

Panel B: variance equation [a] 2 3 0:0229 6 ð0:042Þ 7 6 7 4 0:061 5 ð0:0162Þ 2 3 0:203 6 ð0:043Þ 7 6 7 4 0:181 5 ð0:034Þ

Constants

Monthly

2

2

3 0:017 0:009 0:026 6 ð0:015Þ ð0:019Þ ð0:008Þ 7 6 7 4 0:085 −0:027 −0:041 5 ð0:066Þ ð0:119Þ ð0:090Þ 2 3 0:0007 −0:0006 6 ð0:013Þ 7 ð 0:012 Þ 6 7 4 0:433 −0:049 5 ð0:04Þ ð0:04Þ

[b] 2

3 0:143 6 ð0:076Þ 7 6 7 4 0:353 5 ð0:046Þ 2 3 0:747 6 ð0:052Þ 7 6 7 4 0:845 5 ð0:025Þ

0:0 ð0:26Þ 0:328 ð0:378Þ

Notes: *, ** and *** indicate significance levels of 10%, 5% and 1% respectively. The standard errors of the estimates are in brackets.

pattern, the first and foremost factor is the combined effect of a shock on dynamics of the co-movement. In line with this assertion, results for monthly series, available in Table 4 imply that the co-movement of both markets depends on the combined effects of the previous period shocks of both markets. The estimated coefficient (α) is positive which suggests that shocks affecting both markets in the same direction strengthen the expected correlation whereas the shocks with opposite effects weaken the expected correlation (or strengthen it, but with a negative sign). Such cases are easily observable during the recent turmoil. In the second half of 2008, when the crash in financial markets becomes apparent, combined with the weakness in global demand, financial turmoil led to a huge decline in both markets. Afterwards, some emergency measures were taken, like renowned quantitative easing by the Federal Reserve and such measures resulted in a revival in the stock markets, meanwhile suppressed demand and deliveries of vessels (which were already ordered back in previous growth periods) caused a decrease in the freight rates. This is one of the few key points that explain the divergence of bounce-backs occurring in stock markets and real economy. However, this does not necessarily imply that markets exhibit simultaneous and opposite in direction developments. For instance, China's decision to replenish iron ore inventories during the turmoil caused a temporary surge in the dry bulk freight rates in line with the surge in stock markets. Therefore, the co-movement dynamics of financial markets and real economic activity also empirically suggest existence of the combined effect of shocks resulting from both markets.

4. Conclusion This study has demonstrated the existence of economically significant information spillovers between stock markets and markets for shipping freight by sea. We find mutual feedback between the two markets using multivariate correlation models on the returns of the Dow Jones Industrial Average (DJIA) and the Baltic Dry Index (BDI). Both monthly and weekly series reveal different feedback results regarding the effects of the lagged variables. It is clearly indicated that the previous changes in BDI help to explain the changes in DJIA in the longer term, while a similar effect is not evident in the short term. Just the opposite holds for the effects of DJIA on BDI in the weekly series. Results also suggest that the extent of information spillover between the markets varies over time. VAR findings on monthly sample suggested that the first and third lags of BDI significantly help to explain DJIA while the findings on weekly sample suggested that the first and second lags of DJIA significantly help to explain BDI. Furthermore, the mutual informative feedback becomes more evident during the periods of financial turmoil. Finally, we suggest

O. Erdogan et al. / International Review of Economics and Finance 25 (2013) 282–290

289

Weekly Sample 30%

20%

10%

0%

-10%

-20%

Monthly Sample 80% 60% 40% 20% 0% -20% -40%

Fig. 2. Dynamic conditional correlations.

that time charter rates would be considered as a factor for pricing financial assets while stock markets indices would be counted as a variable for better price discovery in chartering contracts. Acknowledgement The authors thank the participants at the “Third Lloyd's Shipping Economist Ship Finance Conference” (Hong Kong, 2006), the “Sixth Annual International Conference on Business” (Hawaii, 2006), the “Eleventh Annual International Conference of Global Business and Technology Association” (Prague, 2007), the “Twelfth World Conference on Transport Research” (Lisbon, 2010), and especially Paul Bennett (Former Senior Vice President of New York Federal Reserve and Former Chief Economist of NYSE

Table 5 Crisis effects on the conditional correlation.   Model 1 is of the form abs ρ12;t −ρ12 ¼ c þ ah11;t þ bh22;t þ νt , where the independent variables are the conditional variances. On the other hand, Model 2   which involves the NBER dummy variable is similarly expressed as abs ρ12;t −ρ12 ¼ c þ aDNBER þ νt :

Model 1 Model 2

c

a

b

−1.31*** (0.105) 0.063*** (0.008)

734.3*** (70.14) 0.077*** (0.0198)

3.85*** (0.529)

Notes:*, ** and *** indicate significance levels of 10%, 5% and 1% respectively. The standard errors of the estimates are in brackets.

290

O. Erdogan et al. / International Review of Economics and Finance 25 (2013) 282–290

Euronext), Metin Kalkavan (Chairman of Turkish Chamber of Shipping) and Manoulis Kavussanos (Professor of Financial Management, Athens University of Economics and Business) for helpful comments. References Antoniou, A., Pescetto, G., & Violaris, A. (2003). Modelling international price relationships and interdependencies between the stock index and stock index futures markets of three EU countries: A multivariate analysis. Journal of Business Finance & Accounting, 30(5–6), 645–667. Aslanidis, N., Osborn, D. R., & Sensier, M. (2010). Co-movements between US and UK stock prices: The role of time-varying conditional correlations. International Journal of Finance & Economics, 15(4), 366–380. Aydemir, A. C. (2008). Risk sharing and counter-cyclical variation in market correlations. Journal of Economic Dynamics and Control, 32, 3084–3112. Barro, R. J., Mankiw, N. G., & Sala-I-Martin, X. (1995). Capital mobility in neoclassical models of growth. The American Economic Review, 85(1), 103–115. Bartram, S. M., Taylor, S. J., & Wang, Y. -H. (2007). The Euro and European financial market dependence. Journal of Banking & Finance, 31(5), 1461–1481. Bauwens, L., Laurent, S., & Rombouts, J. V. K. (2006). Multivariate GARCH models: A survey. Journal of Applied Econometrics, 21(1), 79–109. Bencivenga, V. R., Smith, B. D., & Starr, R. M. (1996). Equity markets, transactions costs, and capital accumulation: An illustration. The World Bank Economic Review, 10(2), 241–265. Bera, A. K., & Kim, S. (2002). Testing constancy of correlation and other specifications of the BGARCH model with an application to international equity returns. Journal of Empirical Finance, 9(2), 171–195. Bhar, R., & Hammoudeh, S. (2011). Commodities and financial variables: Analyzing relationships in a changing regime environment. International Review of Economics and Finance, 20(4), 469–484. Bollerslev, T. (1990). Modeling the coherence in short-run nominal exchange rates: A multivariate generalized arch model. The Review of Economics and Statistics, 72(3), 498–505. Bollerslev, T., Engle, R. F., & Nelson, D. B. (1994). ARCH models. In R. F. Engle, & D. L. McFadden (Eds.), Handbook of econometrics, Vol. 4. (pp. 2959–3038): North-Holland. Cai, Y., Yeutien Chou, R., & Li, D. (2009). Explaining international stock correlations with CPI fluctuations and market volatility. Journal of Banking and Finance, 33, 2026–2035. Chen, N. -F., Roll, R., & Ross, S. A. (1986). Economic forces and the stock market. Journal of Business, 59(3), 383–403. Chiang, T. C., Jeon, B. N., & Li, H. (2007). Dynamic correlation analysis of financial contagion: Evidence from Asian markets. Journal of International Money and Finance, 26(7), 1206–1228. Coe, D. T., & Helpman, E. (1995). International R&D spillovers. European Economic Review, 39(5), 859–887. Engle, R. (2002). Dynamic conditional correlation. Journal of Business & Economic Statistics, 20(3), 339–350. Erdogan, O. (1996). Comparable approach to the theory of efficient markets: A modified capital asset pricing model for maritime firms. Ankara, Turkey: Capital Markets Board (LOC: 720187). Erdogan, O. (1997). Review of the book titled: Comparable Approach to the Theory of Efficient Markets: A Modified Capital Asset Pricing Model for Maritime Firms. Journal of Economic Literature, 35(3), 1467–1472. Erdogan, O. (2008). An investigation of the volatility of time charter markets. Marine money, 23–27. Erdogan, O., & Schmidbauer, H. (2005). Investors' selection between two financial markets: A conditional correlation approach. ISE Review, 8(30), 1–18. Goulielmos, A. M. (2008). Risk analysis of the Aframax freight market and of its new building and second hand prices, 1976–2008 and 1984–2008. International Journal of Shipping and Transport Logistics, 1, 74–97. Grossman, G. M., & Helpman, E. (1994). Endogenous innovation in the theory of growth. The Journal of Economic Perspectives, 8(1), 23–44. Hummels, D. (1999). Have International Transportation Costs Declined? Mimeo: University of Chicago, Graduate School of Business. Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of Economic Dynamics and Control, 12(2–3), 231–254. Kavussanos, M. G. (1996). Comparisons of volatility in the dry-cargo ship sector: Spot versus time charters, and smaller versus larger vessels. Journal of Transport Economics and Policy, 30(1), 67–82. Kavussanos, M. G., & Nomikos, N. K. (1999). The forward pricing function of the shipping freight futures market. Journal of Futures Markets, 19(3), 353–376. Kim, S. -J., Moshirian, F., & Wu, E. (2006). Evolution of international stock and bond market integration: Influence of the European Monetary Union. Journal of Banking & Finance, 30(5), 1507–1534. Krugman, P. R. (1991). Geography and trade. Cambridge, Mass: MIT Press. Levine, R., & Zervos, S. (1998). Stock markets, banks, and economic growth. The American Economic Review, 88(3), 537–558. Li, X. -M., & Zou, L. -P. (2008). How do policy and information shocks impact co-movements of China's T-bond and stock markets? Journal of Banking & Finance, 32(3), 347–359. Lin, C. -H. (2012). The co-movement between exchange rates and stock prices in the Asian emerging markets. International Review of Economics and Finance, 22(1), 161–172. Lin, K. -P., Menkveld, A. J., & Yang, Z. (2009). Chinese and world equity markets: A review of the volatilities and correlations in the first fifteen years. China Economic Review, 20(1), 29–45. Martens, M., & Poon, S. -H. (2001). Returns synchronization and daily correlation dynamics between international stock markets. Journal of Banking & Finance, 25(10), 1805–1827. McKinnon, R. I. (1973). Money and capital in economic development. : Brookings Institution Press. Reinhart, C. M., & Rogoff, K. S. (2009). The aftermath of financial crises. National Bureau of Economic Research Working Paper Series, No. 14656, (Retrieved from http://www.nber.org/papers/w14656) Rivera-Batiz, L. A., & Romer, P. M. (1991). Economic integration and endogenous growth. Quarterly Journal of Economics, 106(2), 531–555. Roll, R., & Ross, S. A. (1980). An empirical investigation of the arbitrage pricing theory. Journal of Finance, 35(5), 1073–1103. Ross, S. A. (1976). The arbitrage theory of capital asset pricing. Journal of Economic Theory, 13(3), 341–360. Savva, C. S. (2009). International stock markets interactions and conditional correlations. Journal of International Financial Markets, Institutions and Money, 19(4), 645–661. Savva, C. S., Osborn, D. R., & Gill, L. (2009). Spillovers and correlations between US and major European stock markets: The role of the euro. Applied Financial Economics, 19(19), 1595–1604. Shaw, E. S. (1973). Financial deepening in economic development. : Oxford University Press. Syllignakis, M. N., & Kouretas, G. P. (2011). Dynamic correlation analysis of financial contagion: Evidence from the central and eastern European markets. International Review of Economics and Finance, 20(4), 717–732. Tong, H., & Wei, S. -J. (2008). Real effects of the subprime mortgage crisis: Is it a demand or a finance shock? National Bureau of Economic Research Working Paper Series, No. 14205, (Retrieved from http://www.nber.org/papers/w14205) Tse, Y. (2000). A test for constant correlations in a multivariate GARCH model. Journal of Econometrics, 98(1), 107–127. Wahab, M. (2012). Asymmetric effects of U.S. stock returns on European equities. International Review of Economics and Finance, 21(1), 156–172.