Testing for market linkages between Hawaii and Japan's tuna markets

Testing for market linkages between Hawaii and Japan's tuna markets

Fisheries Research 109 (2011) 351–359 Contents lists available at ScienceDirect Fisheries Research journal homepage: www.elsevier.com/locate/fishres...

314KB Sizes 0 Downloads 1 Views

Fisheries Research 109 (2011) 351–359

Contents lists available at ScienceDirect

Fisheries Research journal homepage: www.elsevier.com/locate/fishres

Testing for market linkages between Hawaii and Japan’s tuna markets Hui Huang ∗ , PingSun Leung University of Hawaii at Manoa, 3050 Maile Way, Gilmore 111, Honolulu, HI 96822, USA

a r t i c l e

i n f o

Article history: Received 2 November 2010 Received in revised form 10 March 2011 Accepted 10 March 2011 Keywords: Market linkage Hawaii Japan Tuna price

a b s t r a c t Even though international and regional fish markets may be geographically dispersed, the prices across different markets may exhibit long-run spatial linkages, suggesting that certain locations are integrated and that prices provide relevant market signals. In other words, it is possible that fish prices in different markets may vary in the short run but maintain the equilibrium in the long run. This paper tests whether the markets for yellowfin tuna and bigeye tuna in Hawaii and Japan are integrated. To this end, market integration test procedures are applied to determine whether there exists market linkage among the fish markets using semi-monthly average wholesale price data recorded in Honolulu’s and Tokyo’s wholesale markets. The market integration tests also provide a study of tuna pricing in Hawaii from the perspective of market price linkage, to understand whether markets outside Hawaii influence local pricing. © 2011 Elsevier B.V. All rights reserved.

1. Introduction Tuna species in various product forms are produced and traded extensively on an international scale. Hawaii, an important producer and consumer of tuna, is regarded as a part of a larger network of inter-connected worldwide markets. The advent of reliable packing materials and air shipment has allowed fresh tuna to enter the Hawaii market from around the world, mostly from the Indo-Pacific region that includes Fiji, Tonga, Marshall Islands, and the Philippines (NMFS, 1987–2008). Meanwhile, even though the local consumption of fresh tuna provides the main market for the tuna harvested in Hawaii, a substantial amount of tuna is also exported to Japan, the mainland US, and, to some extent, Canada and Europe (NMFS, 1987–2008). Hawaiian exporters and fishers target the Japan’s tuna markets because it is profitable to sell highquality tuna in these markets. If these two markets are linked to each other, that is, if two international or regional markets for tuna are well integrated, then one would expect that the prices of tuna in two regions would follow a similar pattern as prices change due to profit maximization and commodities arbitrage between these markets. This implies that the price of tuna in one market cannot diverge significantly from the price in the other market before market forces operate to restore a balanced price relationship across both markets, even though there may be differences in absolute prices due to transportation differences, supply characteristics, and demand. To the extent that this is true, such a balanced price relationship will represent an integrated system in a time-series framework. Short-run

∗ Corresponding author. Tel.: +1 808 9568562; fax: +1 808 956 9269. E-mail address: [email protected] (H. Huang). 0165-7836/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.fishres.2011.03.004

price deviations are possible, but in the long-run, an equilibrium relationship will exist for the price of tuna in different markets. The issue of international tuna market integration is an interesting empirical question with respect to Hawaii’s tuna market in particular. Previous studies of the factors affecting tuna price in Hawaii’s fish market offer an analysis of tuna pricing in the short run. For example, McConnell and Strand (2000) utilized a dataset on tuna sold in Hawaii’s auction market to estimate a hedonic price model. The model provided empirical estimates of price increments based on species, quality indicators (such as size or fat content), method of handling, and market conditions. Pan and Pooley (2004) also analyzed the main factors that affected the seasonal variation of the price of fresh tuna using a statistical approach. Factors such as substitution effect, sea surface temperature, holidays and the number of tourists were studied. While previous studies have focused on the local economic factors and fish qualities affecting the price of tuna in the short run, pricing factors beyond the local market have not been explored. Hence, it is useful to study tuna pricing in Hawaii from the perspective of market price linkages in order to understand whether markets outside Hawaii influence local pricing. To study tuna market integration, the data on tuna species, particularly bigeye tuna and yellowfin tuna were obtained for the Honolulu market in Hawaii and the Tokyo market in Japan. This investigation is of interest not only for statistical modeling but also for policy analysis. If it is observed that tuna prices are not integrated internationally, demand modeling should focus on the market of interest and ignore, at least in a statistical sense, the characteristics of regional demand for the species. In this way, timeseries techniques can be viewed as a pretest procedure used to specify the price variables in demand regressions (Gordon et al., 1993; Gordon and Hannesson, 1996). Furthermore, this indicates

352

H. Huang, P. Leung / Fisheries Research 109 (2011) 351–359

that changes in the price of tuna in Hawaii may not be the result of regional or international trends. However, if price linkages are observed, an outside price shock will have short-run effects on the local market, but in the long-run, all prices from linked international markets must again be balanced out. Thus, an integrated price system ensures that the price shock will be short-lived. Also, since Hawaii’s tuna market is dynamic and intrinsically tied to a larger market, trade policy or fishery management changes in Japan as well as the demand and supply outside Hawaii will to a large extent shape the local tuna market. It is also noteworthy that the availability of wholesale market data makes this study feasible. In both Hawaii and Japan, producers usually land their products into competitive wholesale auction markets, where the price is likely to be determined by the interaction between demand and supply. The data from wholesale auction markets are thus usually well documented. This auction system offers an opportunity to study the link between the quality attributes of fish and the prices as well as price linkages. To investigate the price linkage of the tuna markets, this study first reviews some price causality and co-integration models that can be used to investigate market integration. The models include the law of one price, the stationary data procedure, and nonstationary time series techniques developed by Engle–Granger (1987) and Johansen (1988, 1991). The choice of model depends on the time-series characteristics of the underlying data series. The remainder of this paper proceeds as follows. Section 2 discusses some price causality and cointegration models that can be used to investigate market integration, as well as addresses data sources. Section 3 presents and discusses the empirical results, and Section 4 provides further discussion.

2. Materials and methods 2.1. Hawaii and world tuna trade Tuna is the largest component of the pelagic catch and is a primary source of fresh fish to the local fish market in Hawaii. In 2006, commercial tuna landings totaled 7 million kilograms and represented 59% of total commercial landings. The landings, valued at US $44.6 million, were equivalent to 71% of total ex-vessel revenue in Hawaii (WPRFMC, 2008). Bigeye tuna represented the largest component (40%) of the total landings, followed by yellowfin tuna (12%). Among the major tuna catches, bigeye tuna was the largest component since 1996 and averaged about 44% during 1987–2006, while the composition of yellowfin tuna averaged 28% over this 20-year period (WPRFMC, 2008). As the major component of the pelagic catch, tuna has also shown an increasing trend, rising from 4.3 million kilograms in 1992 to 7.8 million kilograms in 2005 (WPRFMC, 2008). In recent years, this trend is largely due to a significant change involving a shift of longline effort away from swordfish and toward tunas. Since tuna is a key economic resource of the Pacific islands (Gillett et al., 2006) and an increasingly important component of Hawaii’s commercial fishery landings due to the prohibition of swordfish fishing, it is of particular importance to investigate the pricing behavior of the tuna market in Hawaii. The price impact from markets outside Hawaii is possible due to the existence of trade across international borders. During the period 1991–2008, the annual amount of tuna imported into the Honolulu Customs District from foreign countries ranges from 0.54 million kilograms to 1.23 million kilograms, with an average of 0.83 million kilograms. While imports tend to be from a large variety of origins and in the form of lower-quality fish sold at discount prices, the export markets are also important for Hawaii’s tuna market. For the period 1991–2008, the annual export to foreign countries, excluding the

Table 1 Import and export of bigeye tuna, 2001–2008 (in 1000s kilograms). Year

Landings

Import

Export

% export destined to Japan

2001 2002 2003 2004 2005 2006 2007 2008 Average

2773 4976 3858 4788 5298 4772 6249 6128 4855

51 140 199 226 187 271 316 239 204

20 20 99 177 320 150 79 68 117

0 0 0 48.5 79.3 61.7 61.9 53.9 38.2

Source: Pelagic Fisheries of the Western Pacific Region 2008 Annual Report, compiled by Western Pacific Regional Fishery Management Council, 2010; Annual trade data through Honolulu customs, NMFS Fishery Statistics and Economics Division.

mainland US export market, ranged from 0.11 million kilograms to 0.98 million kilograms, with an average of 0.42 million kilograms. In other words, around 8.6% of the tuna catch is exported internationally. If the mainland US export market is included, it is estimated that around 15.6% of the tuna catch was exported in 1997 alone (Peterson, 2002). Tables 1 and 2 report the import and export of bigeye tuna and yellowfin tuna, respectively. Around 3.2% of bigeye tuna was exported from 2002 to 2006, as was 14% of yellowfin from 1997 to 2006. Among the export destinations, Japan is the largest international market for tuna produced in Hawaii. It was almost the sole foreign market for Hawaii’s exported tuna prior to 1997 and has remained a major export market in recent years, even though exports to other countries such as Canada have gradually increased. As reported by NMFS, 48.5–79.3% of the total value of internationally exported bigeye from Hawaii was sent to Japan from 2004 to 2008 (Table 1). As for yellowfin tuna, Japan accounts for an even larger percentage. Around 77% of internationally exported yellowfin from Hawaii was destined for the Japan’s market during the period 1989–2008 (Table 2). Given the magnitude of tuna trade between Hawaii and Japan, Japan was therefore selected to study linkages between its tuna market with the Hawaii tuna market. Another reason for selecting Japan is that the country is traditionally the most important Table 2 Import and export of yellowfin tuna, 1989–2008 (in 1000s kilograms). Year

Landings

Import

Export

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Average

1843 2023 1661 1335 1756 1862 2241 1747 2099 1767 1820 2285 1957 1208 1575 1429 1459 1414 1606 1578 1733

167 533 392 454 595 478 384 418 303 186 192 213 466 198 87 50 64 190 123 88 279

464 408 364 160 273 416 928 487 298 154 401 119 90 83 17 23 37 12 284 175 259

% export destined to Japan 92.7 93.7 98.3 99.4 98.7 99.4 99.9 100.0 99.5 99.2 99.3 94.6 88.0 60.8 100.0 51.9 36.8 26.1 0 0 76.9

Source: Pelagic Fisheries of the Western Pacific Region 2008 Annual Report, compiled by Western Pacific Regional Fishery Management Council, 2010; Annual trade data through Honolulu custom, NMFS Fishery Statistics and Economics Division.

H. Huang, P. Leung / Fisheries Research 109 (2011) 351–359

Import quantity of bigeye Japan

Thailand

United States

a sizable amount of yellowfin tuna for its large canned tuna industry.

World

million kg

200

2.2. Models

150 100 50 0

1990

1992

1994

1996

1998

2000

2002

2004

2006

2004

2006

2004

2006

Import quantity of yellowfin World

Japan

Spain

Thailand

million kg

600 500 400 300 200 100 0

1990

1992

1994

1996

1998

2000

2002

Import value of bigeye World

Japan

United States

Thailand

million USD

1,200 1,000 800 600 400 200 0

1990

1992

1994

1996

1998

2000

2002

Import value of yellowfin World

Japan

Spain

Thailand

1,400

million USD

353

1,200 1,000 800 600 400 200 0

1990

1992

1994

1996

1998

2000

2002

2004

2.2.1. A brief literature review Earlier empirical measures of market integration focus on the relationship among prices to test for causality and to test for the Law of One Price (LOP). When the LOP holds, the market integration is complete. To account for the existence of non-perfect substitutability of products, price adjustments occurring over time and testing a long-run LOP relationship are considered (Goodwin et al., 1990). To account for transportation costs which may impede market adjustment, causality tests using stationary price series prove useful in study market relationships (Ravallion, 1986; Slade, 1986). With the development of cointegration methods, price series which are nonstationary can also be empirically tested to for market relationship. To study tuna fisheries, the cointegration and causality tests are typically framed within the context of horizontal integration between final product markets or vertical market integration between raw material and final products. Most recent papers focused on the two main tuna species, skipjack and yellowfin. Jeon et al. (2008) found the markets of the cannery-grade skipjack tuna throughout the globe are spatially integrated by price. Market leadership can be identified in either the Americas or Thailand, over other prices throughout Europe, the south west Pacific islands, the Indian Ocean islands and Africa (Jeon et al., 2008; Squires et al., 2006). For yellowfin tuna, the results are less clear and this species cannot be found to fully connect to the world tuna market (Squires et al., 2006) or such markets for yellowfin tuna are spatially independent (Jeon et al., 2008). Jiménez-Toribio et al. (2010) further evaluated the degree of integration between the major European markets and the world market of frozen and canned tuna through both vertical and horizontal price relationships. They found the high level of market integration at the ex-vessel stage and the price leadership of yellowfin tuna over skipjack tuna. They also found, at the ex-factory level, the European market for final goods appears to be segmented. The present research will use the cointegration and causality tests framed within the context of horizontal integration to test the market integration for the Japan and Hawaii markets of the two tuna species, bigeye tuna and yellowfin tuna. Detailed model reviews are to be presented in next sections.

2006

Source: Fishstat Plus, Globefish , FAO Fig. 1. World imports of tuna into the principal importing countries. Japan was almost the sole import market before 2000 and remained a significant player after 2000 for bigeye. For yellowfin tuna, Japan and several other countries shared substantial world import quantities; however, Japan accounted for the major import values. Source: Fishstat Plus, Globefish, FAO.

international market for tuna. As for bigeye tuna, most of the world’s catch is marketed in Japan. It is no exaggeration to say that the high value of sashimi-quality bigeye sold on the Japan’s market drives the international longlining fleet of all oceans (Bayliff, 2001). Fig. 1 illustrates world imports by quantity and by value of bigeye and yellowfin tuna among the principal importing countries. For bigeye tuna, Japan was almost the sole import market before 2000 and remained a significant player after 2000. For yellowfin tuna, Japan and several other countries shared substantial world import quantities; however, Japan accounted for the major import values, which indicate that Japan imported highquality yellowfin tuna while other import markets were targets for lower-quality yellowfin tuna. For example, Thailand imported

2.2.2. Integration models for stationary data Two spatially dispersed markets are considered integrated if, in the presence of trade between them, the price in the importing market (PtA ) is equal to the price in the exporting market (PtB ) plus transport and other transfer costs (Tt ) at time t. This occurs because of the spatial arbitrage condition given by PtA = PtB + Tt . The equation is the strict version of the law of one price (LOP). The law of one price is one of the approaches that have been previously used for testing market integration. According to the strict version of LOP, the prices of one commodity in two different markets are equal, and their co-movement is perfect. Also, market integration means that price changes in the exporting market will be transmitted to the importing market on a one-to-one basis. These assumptions are, however, rarely satisfied in the real world. For this reason, a weak version of the LOP has often been used for testing market integration. This version assumes that prices have a proportional relationship and that their levels differ due to factors such as transportation and other transfer costs. Before developing the cointegration technique, the validity of the LOP is examined by estimating the regression ln PtA = a + b ln PtB + εt .

(1)

354

H. Huang, P. Leung / Fisheries Research 109 (2011) 351–359

The strict version of the LOP holds when the restrictions a = 0 and b = 1 are satisfied, and the weak version of the LOP holds when restrictions a = / 0 and b = 1 are satisfied. However, an adjustment toward integration can be delayed by certain factors (Ravallion, 1986; Slade, 1986; Goodwin et al., 1990). This delay can be modeled when investigating relationships between prices by specifying a dynamic model. The market integration test is performed by first estimating the following model: pAt

=a+

m 

bj pAt−j

j=1

+

n 

ci pBt−i

+ et ,

(2)

i=1

here p is the logarithm of P, et is white noise, and both j and i are time lags. If the joint test that all ci parameters are zero is rejected, this test supports the hypothesis that there is a relationship wherein pBt determines pAt . A test of the opposite relationship (i.e., that pAt determines pBt ) is possible by interchanging the price variables in Eq. (2). The econometric interpretations of the tests based on interchanging the dependent variables are as follows. If causality is observed in one direction with no evidence in the opposite direction, this indicates price leadership; if causality is not observed, then the markets are independent. To test for LOP relationship,  a long-run  the hypothesis that the restriction bj + ci = 1 holds is tested. 2.2.3. The concept of cointegration Many economic time series tend to change over time. If the series changes in an unstable way, the mean and variance will probably vary over time; in other words, the series is non-stationary. For two variables of the time series that are non-stationary, a linear transformation of the variables may be stationary, in which case these variables are said to be cointegrated (Engle and Granger, 1987). Cointegration implies that there exists a linear long-run relationship between the non-stationary variables in question. If this relationship exists, then the markets are said to be integrated. Engle and Granger (1987) suggested a straightforward two-stage regression procedure to test for cointegration. First, test that each variable series is stationary after differencing d times. Second, form the cointegrating vector, and test whether the errors are integrated with order zero. Therefore, for a set of variables X, a cointegrating vector  can be defined such that Zt = Xt ∼I(0),

(3)

where I(0) is called an integrated series of order zero; series that are non-stationary can often be made stable or stationary if differenced one or more times and are called integrated series of order d. Z, the error term, is white noise representing random disturbances. Despite its popularity, this method has the same normalization problem in selecting the dependent variable with non-stationary data as with stationary data. Besides, it cannot be used to test the significance of the parameters of the static long-run model, as the standard t-statistic inferences are not relevant, and this method is, therefore, also not valid to test for LOP (see Asche et al. (2004) for more detailed discussion). Fortunately, the cointegration test developed by Johansen (1988, 1991) can solve the problems with the Engle and Granger test by modeling price relationships using a VAR (Vector Autoregression) format. The Johansen cointegration test has the following feature. First, a system of equations is used to avoid the simultaneous equation bias that may be introduced in Eq. (1) if both price series are endogenous. Second, using a systems equation format, the normalization of variables in selecting the dependent variable with non-stationary data is no longer necessary. Likelihood ratio tests can be used to investigate hypotheses on the parameters and therefore test for the LOP. Third, the dynamic VAR model makes it feasible to test hypotheses based on the adjustment process and, therefore, to test for price leadership. Following Johansen and Juselius (1994), the

Table 3 Augmented Dickey–Fuller tests. Variables

Test statistics

Test statistics with trend

No. of lags

Bigeye

Hawaii Japan

−3.94 −3.67

−4.05 −3.83

5 6

Yellowfin

Hawaii −2.83 Japan −2.79 First differences Hawaii −7.18 Japan −7.93

−3.34 −3.24

9 7

−8.72 −10.82

8 5

The critical values for the ADF test without trend are −2.58 at the 10% significance level, −2.89 at the 5% significance level and −3.51 at the 1% significance level. The critical values for the ADF test with trend are −3.13 at the 10% significance level, −3.45 at the 5% significance level and −4.04 at the 1% significance level.

maximum likelihood (ML) estimation of cointegration is applied here. 2.3. Data This study empirically examines whether the tuna market in Hawaii is integrated with the tuna market in Japan. For both bigeye tuna and yellowfin tuna, the dataset used in this study consists of semi-monthly wholesale prices for the period from July 1997 to December 2004. The data for the Japan’s wholesale market are Tokyo auction data extracted from the website http://swr.nmfs.noaa.gov/fmd/sunee/twprice/jws.htm, which is maintained by the National Marine and Fisheries Service (NMFS). The dataset on Hawaii’s wholesale market is from the Honolulu auction as recorded by the Honolulu Laboratory of NMFS. Semimonthly price series include volume-weighted average prices, without any prior seasonal adjustment or smoothing. In addition, Japanese prices are converted to US dollars after adjusting for yendollar exchange rates. 3. Results 3.1. Univariate characteristics of the price series Before directly investigating market integration, the first priority is to examine the time series properties of the price series and to confirm whether the price series are non-stationary. This performed using the augmented Dickey–Fuller (ADF) test (Dickey and Fuller, 1979, 1981). For each individual price pit , the ADF statistic is measured using the following regression equation: pit = ˇo + ˇT + pit−1 +

k 

˛ pit− + εt ,

(4)

=1

where pit = pit − pit−1 , and T is a time trend. To determine whether pit is non-stationary, the t-statistic for  in Eq. (4) is used to test the null hypothesis of unit root, Ho :  = 1, i.e., whether the data series is non-stationary in levels. If the hypothesis is not rejected, the test is repeated using the first differences of each price series. In this case, the null hypothesis is non-stationary in first differences. In specifying Eq. (4), the lagged differences (Pit− ) are included to account for possible autocorrelation, and the lag length k is set to capture white noise in error terms. Including too few lags will cause the test to change in an unknown manner, while including too many lags will reduce the power of the test. In this paper, the lag length k is chosen to minimize the Akaike information criterion (AIC). The results are reported in Table 3. The test statistics indicate a rejection of the null hypothesis for both the Hawaii and Japan markets; i.e., the series are stationary.

H. Huang, P. Leung / Fisheries Research 109 (2011) 351–359

355

Hawaii

-2.5 -3 -3.5 -4 -4.5 -5 -5.5 -6 -6.5 -7

1

2

3

4

5

6

7

8

9

10

11

Test Statistic

Test statistic

A. With trend in levels

No. of lags

-1.5 -2 -2.5 -3 -3.5 -4 -4.5 -5 -5.5 -6

Japan 1

2

3

4

5

6

7

8

9

10

11

9

10

11

12

No. of lags

0 -2 -4 -6 -8 -10 -12 -14 -16 -18

Hawaii 1

2

3

4

5

6

7

8

9

10

11

No. of lags

Japan

0

12

1

Test Statistic

Test Statistic

B. Differenced 2

3

4

5

6

7

8

-5 -10 -15 -20

No. of lags

Fig. 2. Augmented Dickey–Fuller tests. At levels the ADF statistic rejects the null hypothesis for short lag lengths, but the statistic becomes insignificant as lag length increases. At different lag lengths, the ADF statistic is significant for all lag lengths examined.

However, for yellowfin tuna, the test statistics indicate that the null hypothesis cannot be rejected for both the Hawaii and Japan’s markets. It is worth noting that in this case, the lag length generated by the AIC procedure appears longer than what seems reasonable in economic terms (Gordon and Hannesson, 1996), particularly for the Hawaii’s market. The problem is that with a long lag length, the existence of a unit root is ambiguous. Following Gordon and Hannesson’s (1996) method, the behavior of the ADF statistic as lag length changes is graphed in Fig. 2. For both price series, the ADF statistic rejects the null hypothesis for short lag lengths, but the statistic becomes insignificant as lag length increases. Particularly, for the Hawaii’s market, the AIC generates a lag length for the ADF statistic in the region of values that cannot be rejected, although more values of the ADF statistic over different lag lengths lie in the rejection region for the null hypothesis; in other words, the price series is stationary in levels. The concern is that if the Hawaii’s price of yellowfin tuna is stationary in levels whereas the Japan’s price is not, a cointegration regression including both price series will be spurious and thus may lead to incorrect conclusions. With this caveat in mind, this study proceeds to test for stationary in first differences. Table 3 also reports the ADF test results based on the firstdifference transformation of both markets. The null hypothesis is rejected for both markets using AIC procedures for choosing lag length. The different values of the test statistic at alternative lag lengths for both markets are also shown in Fig. 2 to evaluate the behavior of the ADF statistic at different lag lengths. Now, for each price series, the ADF statistic is significant for all lag lengths examined. This is taken as evidence that each price series is stationary in first differences, and thus, this study first proceeds to test for cointegrating vectors.

Table 4 Tests for the law of one price, bigeye tuna. Static

Hawaii/Japan Japan/Hawaii *

Dynamic

F-test

p-Value

64.70* 209.64*

0.000 0.000

F-test 0.29 36.41*

p-Value 0.593 0.000

Significance at the 1% level.

3.2. Market integration test for bigeye tuna

first run for the simple two-variable static equation (i.e., Eq. (1) from Section 2.2.2) and then repeated for the dynamic lagged model (i.e., Eq. (2) with the inclusion of seasonal dummies and time trending). Regarding the use of dynamic lagged models to test for market integration, in economic terms the lagged effects in the model are likely to arise from sluggishness in price adjustment, delays in transportation and information transmission, cold storage inventory holdings, and the formation of expectations under price uncertainty (Elston et al., 1999). Squres et al. (1988) and Elston et al. (1999) selected a lag length of six for a market linkage study of thornyhead species in Japan and Alaska using monthly data, as they believed that this relatively long lag length allows for ex-vessel markets after the Alaskan harvest seasonally tapers off over the winter months. As for Hawaii’s and Japan’s markets, a lag length of six is selected using semi-monthly data, as this lag length should accommodate the possible effects of commodity flows in both markets.1 Hence, the models were assessed for equal lag lengths in both markets for lengths of two, three, four, five, as well as six. The influence of seasonality is accounted for by quarterly dummy variables for winter, spring, summer, and fall. Finally, the possibility of long-term effects from volume changes in trade between the two regions is tested with the use of a linear time

As the price series are stationary in levels for bigeye tuna, stationary data procedure rather than cointegration models are used to test for market integration. This study tests both the static and dynamic specifications of the Granger causality models. The test is

1 A lag length of up to twelve is also tested in the models. A lag length longer than six does not significantly change the results.

356

H. Huang, P. Leung / Fisheries Research 109 (2011) 351–359

Table 5 Parametric F-tests for the Granger causality of the bigeye tuna market. Dynamic number of lags

Causality hypothesis

J causes H H causes J

Causation finding

2

3

4

5

6

11.61* (0.000) 0.07 (0.794)

6.45* (0.002) 0.08 (0.927)

3.55** (0.016) 0.68 (0.564)

3.34** (0.012) 0.35 (0.843)

2.76** (0.021) 0.28 (0.925)

J causes H H does not cause J

p-Values are reported in parentheses. * The statistical significance of F-test parameters at the 1% level. ** The statistical significance of the F-test parameters at the 5% level.

trend. Therefore, Eq. (2) becomes2 pAt = a +

m  j=1

bj pAt−j +

n  i=1

ci pBt−i +

4 

Table 6 The Engle–Granger test for cointegration.

ds Ds + fT + et .

(5)

Regressand

s=2

The LOP test results are reported in Table 4. In the table, the first column defines the fish prices used, and the first price in each pair is defined as the dependent variable. Column two reports the LOP results for the static test. The results for the static equation show a rejection of the LOP at the 1% significance level. Both the strict version and the weak version of the LOP do not hold, since restrictions a = / 0 and b = 1 are not satisfied. To test  for a long-run  LOP relationship, the hypothesis that the restriction bj + ci = 1 holds is tested, and F-statistics are used to indicate whether there is a long-run LOP relationship. The results for the dynamic model are somewhat different; the LOP test for Japan regressed on Hawaii is rejected, but the LOP test for Hawaii regressed on Japan is not rejected. Therefore, the prices from the two markets adjust toward equilibrium in the long run, even though this process may not be significant in the short run. With dynamic models, one can also investigate whether the adjustment process is bidirectional or unidirectional through Granger causality tests. The Granger causality tests are reported in Table 5. Causality is observed in one direction, which indicates that the influence of Japan’s price on that of Hawaii is significant, whereas the hypothesis of price causality from Hawaii on Japan is rejected. Thus, causality is not mutual. These results are robust with respect to lag length. Thus, there is market integration with obvious price leadership of the Japan’s bigeye market. 3.3. Market integration test for yellowfin tuna Given the non-stationary price series of yellowfin tuna, cointegration procedures are the correct tools for the market integration test. First, the Engle–Granger cointegration tests are applied, while joint endogeneity is assumed for each fish price. It is worth noting that the cointegration tests are done using prices in level form and that the specifications of the cointegration equation are estimated using seasonal dummies and time trends. Ordinary Least Squares (OLS) is used in estimation and applied to each equation separately (Enders, 1995, p. 374). The cointegration results (Table 6) report ADF statistics, which are calculated from the estimated residuals (that is, Zt in Eq. (3)) for the chosen OLS regression. The results are consistent across two equations but provide no evidence of cointegration of prices among the two markets. In other words, the data

2 The residuals are analyzed for serial correlation, the absence of which is verified with the modified Q-statistic; see Greene (2008) for further discussion. The regression using prices does not exhibit correlation due to the lagged variables. The statistical significance of the linear trend and of the dummy variables is assessed for each unrestricted model for the a priori preferred lag length of six. The linear trend variable is statistically significant and is retained only for the parametric regression of Japan’s prices on Hawaii. Meanwhile, the seasonal dummy variables are statistically significant and retained for the regression of Hawaii on Japan.

Japan

Hawaii

−2.89 (9)

−2.56 (8)

The test is based on the model Zt = 0 + 1 T + Zt−1 +

10% critical value

k i=1

Zt is the error term from the cointegration Eq. (3). The chosen AIC lag length is in parentheses.

−3.56 ˛i Zt−i + εt , where

Table 7 Bivariate Johansen tests of cointegration for yellowfin tuna prices. Maximum eigenvalue test

Trace test

Null

-Max

Null

Trace

Rank = 0

2.87

Rank = 0

2.23

The critical test value 12.25 at the 5% significance level is obtained from OsterwaldLenum (1992).

provide no support for long-run price linkages of yellowfin tuna in Hawaii’s and Japan’s markets. From the initial results, there appears to be no evidence of an integrated price system for yellowfin tuna with respect to Hawaii’s and Japan’s markets. It is not certain, however, that this result is due to a lack of cointegration or because of weaknesses in the Engle–Granger testing procedure. The Johansen test, which is a multivariate generalization of the Dickey–Fuller test, provides an alternative procedure for testing cointegration and thus should yield additional information on the price linkages between the two markets. The Johansen regressions are carried out using a VAR length of two. The results are reported in Table 7. The value of the calculated statistics for the maximum eigenvalue as well as the trace test for testing the null hypothesis indicate there exists no cointegrating vector. The null hypothesis of rank zero is not rejected at the 5% significance level by either the maximum eigenvalue test or the trace test. The results indicate that no cointegrating vector exists for the pair of yellowfin tuna prices. This supports the Engle–Granger results reported earlier. Due to the earlier suspicion that the Hawaii’s yellowfin tuna price is probably stationary in levels and should not be included in a cointegrating regression, the stationary integration tests are applied in the second phase of testing. This replicates the tests conducted for bigeye tuna, and the results are reported in Tables 8 and 9. The results for the static equation show a rejection of the weak LOP using the pairwise tests. The results for the dynamic model are similar. Both tests reject causality at the 5% significance level. Given that Japan’s data are non-stationary, these empirical Table 8 Tests for the law of one price for yellowfin tuna, assuming prices are stationary in levels. Static

Hawaii/Japan Japan/Hawaii *

Dynamic

F-test

p-Value

153.35* 154.55*

0.000 0.000

F-statistics are significant at 1% level.

F-test 7.10* 16.52*

p-Value 0.008 0.000

H. Huang, P. Leung / Fisheries Research 109 (2011) 351–359

357

Table 9 Parametric F-tests for the Granger causality of yellowfin tuna, assuming prices are stationary in levels. Causality hypothesis

J causes H H causes J

Dynamic number of lags

Causation finding

2

3

4

5

6

0.39 (0.534) 0.15 (0.702)

1.13 (0.367) 0.14 (0.870)

1.01 (0.390) 1.51 (0.213)

0.64 (0.633) 1.57 (0.184)

0.85 (0.517) 1.17 (0.325)

J does not cause H H does not cause J

p-Values are reported in parentheses.

The purpose of this paper is to test for market linkages between Hawaii’s and Japan’s fish markets with respect to bigeye tuna and yellowfin tuna. To investigate time-series price data, this study reviews LOP, price causality and cointegration models that can be used in the study of market integration. Traditional stationary procedures are used to test market integration for stationary data series while cointegration tests are appropriate for non-stationary data series. With respect to the empirical procedure in this study, potential problems with the analysis lie in the time-series data that are used. There are only a relatively small number of observations available, and the time span is relatively short. Moreover, product substitution in the long run most likely exists for different species of tuna, including bigeye, yellowfin, skipjack and albacore. Further, Japan’s and Hawaii’s markets might be part of a globally integrated market and as such may interact with tuna markets elsewhere. If this is the case, integration should be examined in a way that includes all tuna species throughout the entire world market. Given the limitations of these studies, the results should be interpreted with caution. However, these problems generally are not cause to reject the results outright (Nielsen, 2005). The empirical results from causality models used to test stationary data indicate market integration for the Japan and Hawaii’s bigeye tuna markets. Even though the two markets are geographically dispersed, spatial pricing relationships indicate that the prices are linked together and that the prices provide relevant market signals within and across regional boundaries. What is more, it appears that the two markets are spatially linked, with evidence suggesting that the Japan’s market plays a dominant role in influencing prices. In integrated markets, market forces will ensure the stability of prices by maintaining a regional balance between demand for and supply of bigeye tuna. This is particularly true for Hawaii’s bigeye market, as its market linkage with Japan is strong, and it is evident that prices are likely to be influenced by the Japan market. This linkage is interesting in relation to regional market policies. If market integration is absent, regulation polices may be effective due to a possible price response. However, if market integration is present, such polices may be less effective or altogether ineffective, as price levels can only be affected in the short term in this case. Thus, a second hypothesis emerges that an intervention by fishery regulators in Hawaii only has a short-run impact on prices in the bigeye tuna market, as prices eventually return to the price path determined by the wider interregional market. In this case, the recent swordfish closures within the local bigeye market may have a rather insignificant price impact in the long run. Statistically, similar price interdependence holds before and after the implementation of closure (see Appendix A). According to these

Import price of bigeye World

Japan

1.6

USD per kg

4. Discussion

additional tests, the LOP for Japan regressed on Hawaii is rejected, but that for Hawaii regressed on Japan is not rejected for both before and after the implementation of swordfish closure. Therefore, equilibrium in the markets exists in the long run. Causality is observed in one direction, which indicates the influence of Japan’s price level on that of Hawaii; moreover, causality is significant in this case, while Hawaii’s influence on Japan’s market is rejected. These results are robust with respect to lag length before the implementation of swordfish closure but are only robust for two lag lengths after the implementation of swordfish closure. The notion that the price leadership of Japan is less evident for the period after the swordfish closure resulting from the fact that regional policy has a short-term impact and long-run relationship between two markets is not as evident under studies following a shorter time period. In sum, the findings that bigeye tuna prices in Hawaii’s and Japan’s markets are integrated and that Japan is the price leader imply that Japan serves as a stabilizing factor for Hawaii’s tuna prices. Furthermore, when Hawaii’s local market is connected through prices to a foreign market, local production, marketing decisions and management cannot be isolated from the foreign market. Since price leadership exists in Japan’s market, management of local fisheries should maintain an awareness of developments in Japan, including consumer tastes and preferences, trade policies, and even Japan’s fishery polices. For example, the possible existence of excess harvesting capacity in Hawaii-based pelagic fishery as reported by Balsiger et al. (2008) may require the control of fishing capacity in Hawaii, which will limit landings in the Hawaii’s market. Even if the Hawaii’s bigeye tuna price proves to be responsive to the volume of local landings as well as local demand and other price-setting

1.4 1.2 1

0.8 0.6 0.4 1990

1992

1994

1996

1998

2000

2002

2004

2006

2002

2004

2006

Import price of yellowfin World

Japan

1

USD per kg

results are consistent with analytical results in earlier literature, which showed that causal models tend to over-reject the LOP for non-stationary data (Asche et al., 2004). Given the number of tests employed, the evidence for market integration is statistically weak. This study thus concludes that evidence for the existence of longrun price linkages is not strong for yellowfin tuna across Hawaii’s and Japan’s markets.

0.8 0.6 0.4 0.2 0 1990

1992

1994

1996

1998

2000

Source: Fishstat Plus, Globefish , FAO Fig. 3. Import prices of tuna, World and Japan. The import price movement of the world market overlapped that of Japan’s market. However, for yellowfin, the price movement of Japan’s market deviated from the world price movement to some degree. Source: Fishstat Plus, Globefish, FAO.

358

H. Huang, P. Leung / Fisheries Research 109 (2011) 351–359

factors, the price is not insulated from movements in the Japan’s bigeye price. Management of fishing capacity that ultimately only affects landings in Hawaii for a constant demand cannot result in price changes in Hawaii in isolation from Japanese events to the extent that Japan’s market transmits these influences through its bigeye price. Thus, for example, decreased import volumes in Japan for a given level of demand could affect Honolulu’s bigeye prices even if the fishing capacity affecting landings in Hawaii is controlled. Based on the above analysis, it can be further posited that if the local market is understood as a separate market without considering the integrated market, the price flexibility of the local market as the first derivative in relation to quantity may be over-estimated. This is especially true for Hawaii’s bigeye market, as the price is significantly influenced by Japan’s market. When prices are formed within an integrated market and quantities in the different parts of the market are independent (i.e., Japan’s bigeye tuna quantity and Hawaii’s quantity vary randomly in relation to each other over time), Hawaii’s local price changes are affected by local supply as well as supply from the integrated market. Therefore, the studies covering a single market that suppose that price changes are influenced only by local quantity changes tend to overestimate price flexibilities within a local market given the presence of market integration. While the results show that the bigeye market is strongly integrated, in the case of the yellowfin markets, no price linkages between Hawaii and Japan markets are observed. That is, the existence of long-run market integration is not supported by either cointegration or stationary procedures. Based on the data used, the results appear ambiguous due to the possible presence of different stationary features of the datasets from the two markets. Otherwise, the results indicate that the prices of yellowfin tuna between two regions may deviate substantially with no common factors forcing price linkage. In other words, changes in Japan’s price levels appear to have a minimal impact on Hawaii’s markets and vice versa. The results for yellowfin tuna suggest two different market systems for the two regional yellowfin markets. That being true, the price effects from yellowfin tuna conservation and management

measures taken in one area are confined to that region. The findings are somewhat surprising because, even if there is no obvious substitution in the two regional markets on the consumption side, producers can change their allocation of raw fish in response to changes in the prices of final products. This ought to prevent prices from diverging significantly apart. A subtle explanation of these findings is suggested by a comparison of the world tuna market with Japan’s tuna markets. Although more yellowfin than bigeye that are landed in Hawaii are exported to Japan, the price linkage of bigeye is statistically stronger than yellowfin, perhaps simply because Japan is a more internationally dominant market for bigeye. As illustrated by Fig. 3, Japan took up almost all of the world’s import of bigeye tuna by quantity and by value. Also, the import price movement of the world market overlapped that of Japan’s market, which indicates that Japan is actually the world price leader. However, for yellowfin, the export markets in the world are diversified. According to Fig. 2, Japan’s import of yellowfin accounts for a lower percentage of the world import as compared to that of bigeye in terms of both quantity and value. Therefore, the price movement of Japan’s market deviated from the world price movement to some degree (Fig. 3). Given these findings, it is more likely that the integrated Japan-Hawaii market interacts with tuna markets elsewhere, and hence, integration is not statistically evident.

Acknowledgements The authors thank the editors and two anonymous referees for helpful comments and suggestions. Insightful comments from James Moncur, Sam Pooley, Jerry Russo and Harry Ako are also gratefully acknowledged.

Appendix A. See Tables A1–A4.

Table A1 LOP tests for bigeye tuna before swordfish closure, May 1997–April 2001. Static

Hawaii/Japan Japan/Hawaii *

Dynamic

F-test

p-Value

F-test

p-Value

47.35* 110.03*

0.000 0.000

1.90 30.62*

0.172 0.000

Significance at the 1% level.

Table A2 Parametric F-tests for Granger causality for bigeye tuna before swordfish closure, May 1997–April 2001. Causality hypothesis

Dynamic number of lags 2

J causes H H causes J

3

6.04* (0.003) 5.22* (0.002) 2.02 (0.138) 2.26 (0.087)

Causation finding 4

5

6

3.75* (0.008) 1.93 (0.113)

4.35** (0.016) 1.96 (0.090)

4.08** (0.015) 1.98 (0.080)

J causes H H does not cause J

p-Values are reported in parentheses. * Statistical significance of F-test parameters at the 1% level. ** Statistical significance of F-test parameters at the 5% level. Table A3 LOP tests for bigeye tuna after swordfish closure, April 2001–December 2004. Static

Hawaii/Japan Japan/Hawaii *

Significance at the 1% level.

Dynamic

F-test

p-Value

F-test

p-Value

30.09* 124.69*

0.000 0.000

0.13 14.51*

0.722 0.000

H. Huang, P. Leung / Fisheries Research 109 (2011) 351–359

359

Table A4 Parametric F-tests for Granger causality for bigeye tuna after swordfish closure, April 2001–December 2004. Causality hypothesis

J causes H H causes J

Dynamic number of lags

Causation finding

2

3

4

5

6

1.38 (0.157) 1.30 (0.278)

2.36** (0.048) 2.33 (0.081)

1.95** (0.032) 1.95 (0.112)

1.81 (0.123) 1.83 (0.119)

1.95 (0.057) 1.91 (0.093)

J causes H H does not cause J

p-Values are reported in parentheses. ** Statistical significance of F-test parameters at the 5% level.

References Asche, F., Grodon, D.V., Hannesson, R., 2004. Tests for market integration and the law of one price: the market for whitefish in France. Mar. Resour. Econ. 19, 195–210. Balsiger, J.W., Risenhoover, A., Boreman, J., 2008. Excess harvesting capacity in US Fisheries. A Report to Congress. National Marine Fisheries Service. Bayliff, W.H., 2001. Organization, functions, and achievements of the Inter-American Tropical Tuna Commission. IATTC Special Report 13. Inter-America Tropical Tuna Commission, La Jolla, California. Dickey, D.A., Fuller, W.A., 1979. Distribution of the estimators for autoregressive time series with unit root. J. Am. Stat. Assoc. 74, 427–431. Dickey, D.A., Fuller, W.A., 1981. Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica 49, 1057–1072. Elston, J.A., Hastie, J.D., Squires, D., 1999. Market linkage between the U.S. and Japan: an application to the fisheries industry. Jpn. World Econ. 11, 517–530. Enders, W., 1995. Applied Econometric Times Series. John Wiley, New York. Engle, R.F., Granger, D.E.A., 1987. Co-integration and error correction: representation, estimation and testing. Econometrica 55, 251–276. Gillett, R., McCoy, M., Rodwell, L., Tamate, J., 2006. Tuna: a key economic resource in the Pacific islands. A report prepared for the Asian Development Bank and the Forum Fisheries Agency. Food and Agriculture Organization of the United Nations. Goodwin, B.K., Grennes, T.J., Wohlgenant, M.K., 1990. Testing the law of one price when trade takes time. J. Int. Money Finance 5, 682–693. Gordon, D.V., Salvanes, K.G., Atkins, F., 1993. A fish is a fish is a fish? Testing for market linkages on the Paris fish market. Mar. Resour. Econ. 8, 331–343. Gordon, D.V., Hannesson, R., 1996. On prices of fresh and frozen cod. Mar. Resour. Econ. 11, 223–238. Greene, W., 2008. Econometric Analysis. Prentice Hall, New Jersey. Jeon, Y., Reid, C., Squires, D., 2008. Is there a global market for tuna? Policy implications for tropical tuna fisheries. Ocean Dev. Int. Law 39, 32–50. Jiménez-Toribio,

R., Guillotreau, P., Mongruel, P., 2010. Global integration of European tuna markets. Prog. Oceanogr. 86, 166–175. Johansen, S., 1988. Statistical analysis of cointegration vectors. J. Econ. Dyn. Control 12, 231–254. Johansen, S., 1991. Estimation and hypothesis testing of cointegration vectors in Gaussian autoregressive models. Econometrica 59, 1551–1580. Johansen, S., Juselius, K., 1994. Identification of the long-run and the short-run structure: an application to the ISLM model. J. Econometrics 63, 7–36. McConnell, K.E., Strand, I.E., 2000. Hedonic prices for fish: tuna prices in Hawaii. Am. J. Agric. Econ. 82, 133–144. Nielsen, M., 2005. Price formation and market integration on the European first-hand market for whitefish. Mar. Resour. Econ. 20, 185–202. NMFS, 1987–2008. Annual Trade Data. Fishery Statistics and Economics Division. Osterwald-Lenum, M., 1992. A note with quantities of the asymptotic distribution of the maximum likelihood cointegration rank test statistics. Oxf. Bull. Econ. Stat. 54, 461–471. Pan, M., Pooley, S., 2004. Tuna price in relation to economic factors and sea surface temperature in fresh tuna market. IIFET 2004 Japan Proceedings. Available from: http://www.pifsc.noaa.gov/pubs/sepub.php. Peterson, A., 2002. 1997 Hawaii fishery input–output model and methodology. SOEST 05-02. JIMAR Contribution 05-356. Ravallion, M., 1986. Testing market integration. Am. J. Agric. Econ. 68, 102–109. Slade, M.E., 1986. Exogeneity test of market boundaries applied to petroleum products. J. Ind. Econ. 34, 291–304. Squres, D., Herrick Jr., S., Hastie, J., 1988. Integration of Japanese and United States sablefish markets. Fish Bull. 87, 341–351. Squires, D., Jeon, Y., Kim, T., Clarke, R., 2006. Price linkages in Pacific tuna markets: implications for the south Pacific tuna treaty. Environ. Dev. Econ. 11, 1–21. WPRFMC (Western Pacific Regional Fishery Management Council), 2008. Pelagic Fisheries of the Western Pacific Region 2006 Annual Report. Honolulu, Hawaii.