Scholarly article on topic 'Price adjustment in world wine markets: A cointegration analysis'

Price adjustment in world wine markets: A cointegration analysis Academic research paper on "Economics and business"

CC BY-NC-ND
0
0
Share paper
Academic journal
Wine Economics and Policy
OECD Field of science
Keywords
{Cointegration / "Export prices" / Wine}

Abstract of research paper on Economics and business, author of scientific article — Juan Sebastián Castillo-Valero, Mª Carmen García-Cortijo

Abstract World wine trade has undergone an exponential dynamic in recent years because of the fall in domestic demand of the main traditional producing countries. This study aims to measure the degree of price integration in the international wine market, within a framework where review and re-adaptation of strategies and behaviors is continuous in a scenario of increasing globalization. Prices from the principal Old World exporting countries have been taken into account, and those from New World exporters. The methodology adopted is based on estimating the Error Correction Vectors, linear and with thresholds. Results obtained show that export prices of Old World countries in the EU are homogenous and seek equilibrium within the same cointegration space; and, on the other hand, that New World exporters do not share a common behavior in their exporting dynamics. France appears as the “leader” of Old World countries, although its leadership and trend is not followed or shared by the New World exporters. However, Italy and particularly Spain are the ones cointegrated, linearly and non-linearly, with markets from New World countries, USA and Argentina. Therefore, France is reference within the EU, while New World exporters countries take Italy and Spain as reference competitors.

Academic research paper on topic "Price adjustment in world wine markets: A cointegration analysis"

Author's Accepted Manuscript

PRICE ADJUSTMENT IN WORLD WINE MARKETS: A COINTEGRATION ANALYSIS

Sebastian Castillo-ValeroJuan Sebastián Castillo-Valero, Ma Carmen García-Cortijo

m mm±

fe Econcfmics Policy31

www.elsevier.com/locate/wep

PII: S2212-9774(15)00019-8

DOI: http://dx.doi.org/10.1016Zj.wep.2015.05.004

Reference: WEP53

To appear in: Wine Economics and Policy

Received date: 8 January 2015 Revised date: 29 April 2015 Accepted date: 29 May 2015

Cite this article as: Sebastian Castillo-ValeroJuan Sebastián Castillo-Valero, Ma Carmen García-Cortijo, PRICE ADJUSTMENT IN WORLD WINE MARKETS: A COINTEGRATION ANALYSIS, Wine Economics and Policy, http://dx.doi.org/ 10.1016/j.wep.2015.05.004

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1 PRICE ADJUSTMENT IN WORLD WINE MARKETS: A

2 COINTEGRATION ANALYSIS

4 Juan Sebastián Castillo-Valeroa; Ma Carmen García-Cortijoa*

5 aRegional Development Institute, Castilla-La Mancha University, Campus Universitario s/n, 02071

6 Albacete, Spain

10 ii i2

20 2i 22

26 * Correspondence to: Campus Universitario s/n, 02071 Albacete, Spain

27 Tel.: +34 967599200 ext 2619.

28 E-mail address: garcia.cortijo@gmail.com

31 Abstract

32 World wine trade has undergone an exponential dynamic in recent years because of the

33 fall in domestic demand of the main traditional producing countries. This study aims to

34 measure the degree of price integration in the international wine market, within a

35 framework where review and re-adaptation of strategies and behaviors is continuous in

36 a scenario of increasing globalization. Prices from the principal Old World exporting

37 countries have been taken into account, and those from New World exporters. The

38 methodology adopted is based on estimating the Error Correction Vectors, linear and

39 with thresholds. Results obtained show that export prices of Old World countries in the

40 EU are homogenous and seek equilibrium within the same cointegration space; and, on

41 the other hand, that New World exporters do not share a common behavior in their

42 exporting dynamics. France appears as the "leader" of Old World countries, although its

43 leadership and trend is not followed or shared by the New World exporters. However,

44 Italy and particularly Spain are the ones cointegrated, linearly and non-linearly, with

45 markets from New World countries, USA and Argentina. Therefore, France is reference

46 within the EU, while New World exporters countries take Italy and Spain as reference

47 competitors.

49 Keywords: Cointegration, Export Prices, Wine

1. Introduction

During the past two decades, the global tendency of the wine industry has experienced fundamental changes (Cassi et al., 2009). The so-called New World (NW) countries: USA, Chile, Australia, Argentina and South Africa, have now joined the stable productive and commercial pattern of wine production concentrated in a selected number of European countries called Old World (OW), namely France, Italy and Spain, which used to dominate the international market. In recent years the NW countries have positioned themselves in the international world market designing lowering price strategies to compete with the traditional European producer countries (Anderson, 2004). OW countries have a developed, strongly regulated industry, with traditional practices and high amortization. NW countries are young producers in general and have experienced great growth in recent decades without much concern for controlled designations of origin but with a strong drive to incorporate new technology (Moreira et al, 2011; Villanueva, 2011).

Therefore, a continuous process of the restructuring of actors in the global market has taken place, making the rhythm of their trend and of their strategies change constantly (Calderón and Blanco, 2005). In this context, numerous studies are focused on analyzing the new dynamics of the international market: Campbell (2000), Anderson (2001), Green and Pierbattisti (2002), Anderson, Norman and Wittwer (2003), Pesenti (2011) and Villanueva (2011), Bentzen and Smith (2002) and Triguero (2002). However, one aspect of special interest is the study of the spatial relation of prices based on market integration, as its analysis helps explain the global operation of the markets (Sanjuan and Gil, 1997). Market integration refers to the price behavior of one product in different locations. Integrated markets are those where price variations in one market

77 are related to price variations in another market. That is, markets in which prices move

78 in a synchronized rather than independent way (Monke y Petzel, 1984). The Engle-

79 Granger Cointegration Theory (1987) is an econometric technique which allows for

80 viewing market spatial integration as a long-term balanced relationship with price

81 adjustment taking place in the short term; integration is confirmed if prices are

82 cointegrated. (Sanjuan and Gil, 1997).

84 This study aims to analyze the integration level of the international wine market

85 specifically: 1) within the OW, 2) within the NW and 3) between OW-NW. To that end

86 we will use the relationships between export prices as an indicator of the level of

87 connection between both markets (Hernández et al, 2002).

89 2. Materials and Methods

90 2.1. Sample and variables

92 The database used is composed of wine exports price series. The series has a monthly

93 time step and the time period runs from January 2005 through to February 2013. Prices

94 are in €/l in constant units1 and have been obtained as the ratio between export values

95 and export volumes. Data comes from Trade Map (http://www.trademap.org/). The

96 product studied is Wine of fresh grape, including fortified wines; grape must, except for

97 the 2009 consignments (code 2204 of the European Commission product classification

1 The set of exports deflated with the World Bank index. http://datos.bancomundial.org/.

98 system ). The wine classified in this consignment is exclusively the final product of the

99 alcoholic fermentation of the must and the product resulting from fresh grape treading. . 100

101 The countries included in the database are the main exporting countries according to the

102 OEMV in 2012: France, Italy, Spain, Australia, Chile, USA, Argentina and South

103 Africa. The fact that not all wine selling countries are included does not imply

104 significance problems in the results. According to the formula for finite samples,

Z2*p*(l-p)*N

N*e2+Z2*p*(l-p)

, the sample error (e) of not considering all selling countries is

106 virtually null for a 0.05 p-value. Thus, for the export value variable the error (e)

107 represents 0.3%4 and for the export volume variable the error represents 0.4% .

108 Furthermore, the sample covers eight complete years (January 2005- February 2013).

109 This number is sufficient to apply the cointegration technique given that following

110 Pulido and López (1999), time series covering information for 6 to 10 complete years

111 are suitable for presenting monthly data.

113 The descriptive statistics of the variables that are included (Table 1) show a range of

114 variation in the results. In the first place, prices from the three European countries are

115 conditioned in their high spectrum by the prices of French champagne, Italian frizzante

http :// exporthelp. europa. eu/thdapp/ nomenclature/Nomenclatures ervlet?action=nomen&section=stat&tari cCode=2204000000&prodLine=80&limitLevel= 8 &simDate=20100101 &languageId=es

3 http://www.armonizado.com/imprimir_notasexplicativas.php?pagina=2204

4 Z2*p*(l-p)*N 1

In n =--—-----l::n=20,167.1 million € ( export values of the 8 countries in the sample in

N*e2+Z2*p*(l-p)J F

2012), N= 25,282.6 million € ( world exports), p= 0.5, e= estimation error and Z= 1.96, ( normal value

at the 95% confidence level). OeMv data.

5 r Z2*p*(l-p)*N 1

In In =--—-----l: :n=8,232.2 million litres (export volumes of the 8 countries in the sample

L N*e2+Z2*p*(l-p)J F

in 2012), N= 9,391.7 million litres (world exports), p= 0.5, e= estimation error and Z= 1,96, (normal

value at the 95% confidence level) OeMv data.

116 and Spanish cavas. In the second place, because New World countries have had to

117 compete at low prices against the European producing countries in order to penetrate the

118 markets where the European traditionally had a dominion position.

120 121 122

Please insert table 1

2.2. Method

In order to carry out the study we used the linear and threshold cointegration technique, as in studies by Baldi et al. (2010, 2013) for the global wine industry represented by France, USA, Chile, China and Argentina; Mencet et al. (2006) for wine exports from France, Greece and Turkey; Pinilla and Ayuda (2002) for Spanish wine exports; and Orgun and Temiz (2011) for the Turkish wine industry. The reason we used this technique is that following Barret (2001); Fackler and Goodwin (2001); Meyer and v.Cramon-Taubadel (2004); Serra et al (2006); Ihle et al (2009), Stephens et al (2010), Orgun and Temiz (2011) and Espostia and Listortib (2012) we regard it as suitable for the study of agricultural markets integration. Furthermore, Gujarati (1995) points out that one of the main advantages of the Cointegration technique is that it does not require distinguishing between endogenous and exogenous variables as each variable is affected by the others and affects the other variables capturing the feedback, impact and adjustment of some markets over others. The traditional econometric techniques of multiple equation models do not allow this, nor do the classic and modern time series techniques; the former because they do not allow for all the variables in a model to be a function of the others and the latter because they only study the behavior of a variable on the basis of itself.

142 The concept of integration allows clarifying the type of spatial relationship across

143 markets: two markets will have a high degree of spatial integration if price variations in

144 one of them are transferred to the other, or they can be segmented if they there is no

145 connection of any kind between their prices (Goodwin and Schroeder, 1991). Moreover,

146 in the former case, if the transfer of price variation from one market to another is

147 immediate in time, we have linear cointegration. If movement towards balance happens

148 only when price variations exceed a given threshold, cointegration will be non-linear.

150 The model which gives shape to Cointegration is the so-called Vector Error Correction

151 Model (VECM), which can be functionally summarized as follows:

152 AP, = f (PtAP,-k, ft) (1)

153 That is, price variation (APt) in a market composed of a set of countries for a specific

154 product and at a given moment is a function of the prices available in that market

155 (Pt_1) and the variation experienced by those prices (APt_k). Random shock is

156 represented by (et).

158 Here we will study two markets: 1.Old World, which includes Spain, France and Italy

159 and 2. New World, which includes the USA, Australia, Chile, South Africa and

160 Argentina. Thus, the object pursued through the use of Cointegration and its VEC

161 models is to study the transfer of price adjustments in the short term to reach balance in

162 the long term and maintain competitive dynamics, a) within the Old World, b) within

163 the New World and c) between the Old World and the New World.

Finally, we have used the Revolution R Enterprise 6.1 software to obtain statistical and non-linear econometric results, and Gretl and Eviews 6 for the linear.

3. Results

3.1. Econometric properties of the series: non-stationarity and cointegration.

The first stage of the methodology consists of defining whether or not the series met the econometric requirements of integration. The variables have to be integrated of order one, I(1), that is, non-stationary at level but stationary at first difference. To test for stationarity of the variables, the Dickey and Fuller test (1979, 1981) was used, and that of Ng and Perron (2001). For all the series the null hypothesis of non-stationarity is accepted in levels and rejected in first differences. They are integrated variables of the first order, I(1), with the exception of the South African price series that proved to be stationary in levels, therefore it was excluded from the cointegration analysis. Table 1 shows the test results.

With the I(1) series, the next step consisted in contrasting whether they were linearly cointegrated by using the Johansen test (1988). In order to apply the Johansen test, first we need to define optimum lag length using the Hannan-Quinn information criterion (HQC). The results are confirmed with the Doornik and Hansen test (1994).

Contrasts were applied to the two groups of countries: Old World (France, Italy and Spain) and New World (Argentina, Australia, Chile and USA). The Hannan-Quinn results indicate (k=2) for the system of Old World countries (HQ= -0.811893) and (k=

190 1) for the New World (HQ=-9.052701). The correct specification of the two groups was

191 confirmed with the contrast of Doornik and Hansen, with a p-value lower than the level

192 of significance of 5 % (p-value = 1.7626e-038 and p-value = 0.00723032, respectively).

193 Therefore, the conditions for contrasting the cointegration of the series were fulfilled.

195 Once lag length was defined the Johansen test was applied, specifically the trace test.

196 The Johansen test (1988), for a level of significance of 5 %, allowed accepting the null

197 hypothesis of cointegration for the group of Old World countries (p-value of 0.0146

198 and critical value 21.13162) and rejecting it for the New World countries (p-value of

199 0.4260 and critical value 47.85613).

201 In short, the Old World countries belong in the same integration space and maintain a

202 dependent price transfer dynamics. However, New World countries do not belong in the

203 same space and the price dynamics observed in Old World countries are not seen; prices

204 rather fluctuate in an independent way. In view of the results, it was decided to study

205 the linear cointegration relationship between Old World countries and later, the

206 relationship of crossed pairs between countries of the Old and the New World.

207 Integration amongst New World countries is not studied given that, as explained before,

208 this relationship does not take place.

210 3.2. Cointegration across Old World countries.

212 The linear cointegration relationship between Spain, France and Italy was estimated by

213 the Maximum Likelihood method. The results are shown in Table 2. Following Gujarati

214 (1995), one of the problems of this method is that the estimation results are difficult to

215 interpret, thus it is restricted to the analysis of the significance of the estimated p and A.

216 The significance of parameters p and A is shown with the contrast of exclusion and weak

217 exogeneity, respectively. On the one hand, the contrast of exclusion found that

218 coefficients fi of the 3 countries were significant with a p-value close to zero; this

219 implies that the 3 are part of the same space of cointegration, determining a relationship

220 of equilibrium between them. On the other hand, the contrast of weak exogeneity

221 showed that the France and Italy markets (X(2, 1), A,(3, 1)) fulfill the condition of

222 exogeneity (with a p-value over 0.05; 0.875993 and 0.243085 respectively), but not

223 Spain (^(1,1)). Spain was the most dependent market (with a p-value lower than 0.05,

224 specifically 0.000162), adjusting to the shocks that are produced. Furthermore, in terms

225 of France and Italy, France is in a leadership situation in the context of the EU given

226 that according to the exclusion contrast the probability associated with the French

227 variable X(2, 1), 0.875993, is higher than that associated with the Italian variable X(3, 1),

228 with a value of 0.243085. Therefore, France is more independent than Italy, although

229 both of them are independent with a probability over 0.05.

231 Please insert table 2

233 3.3. Cointegration between Old World and New World countries

235 The analysis is completed with the study of crossed pairs between Old World and New

236 World countries, in order to gain a better perspective of how the integration dynamics

237 works in the wine export market. To this end, the optimal delay that relates them is

238 calculated, in the first place (Table 3). Afterward, the linear and non-linear relationships

239 were sought, through the Johansen test (1988) and the Hansen and Seo test (2002),

240 respectively (Table 3). The results of the Cointegration tests show that the countries

241 most related in terms of price transfer are: 1) Italy and USA, 2) Spain and USA, 3)

242 Spain and Argentina

243 Please insert table 3

244 The results of the estimation of the three pair of countries are presented below. The

245 model that connects Italy-USA is a VEC of one threshold and two delays, based on the

246 criterion of Residual Sum of Squares (RSS) with a value of 0.9512823 (with k=1 the

247 RSS was 0.9819776). The result of estimation (Table 4) shows that the USA, with a

248 negative sign for parameter a2, (a2 =-0.0041) in the Down regime, adjusts its prices

249 when wt-1 < 0.548, although not significantly. That is, the USA attempts to seek

250 competitiveness confronting Italy. When the differential is more than 0.548 points (

251 wt-1 > 0.548) the interest for the competitor's price is reduced for both countries,

252 possibly generating an imbalance process in the c/p (positive sign of the a1, a2, of

253 D(LIT) and D(LUSA)), although not significant. Italy's reaction, with a higher a is

254 greater than the USA's.

256 The VEC model between Spain-USA with two delays, 1 threshold is the one that better

257 adjusts in face of the one with 1 delay and 1 threshold (for k=1 the RSS was 1.90789

258 and for k=2 the RSS was 1.764306). Spain seeks competitiveness in face of the USA in

259 the Down regime (a1 = -0.755)) and in the Up regime (a1 = -0.4089), the first parameter

260 is significant at 1 % and the second at 10 % (Table 4).

262 Finally, the VEC model between Spain and Argentina shows two integrated markets,

263 with homogeneous prices, adjustment of possible imbalances of the prices in the short

264 term and equilibrium in the l/p (Table 4). The contrast of exclusion found that the P

265 coefficients of the 2 countries were significant with a p-value close to zero (Table 4);

266 this means that Spain and Argentina are part of the same space of cointegration,

267 determining an equilibrium relationship between them. The contrast of weak exogeneity

268 (Table 4) showed that no market fulfills the condition of exogeneity, no country is

269 found in a situation of predominance over the other in price transmission. Spain and

270 Argentina follow the same trend in price dynamics, fundamentally because they are

271 specialized in a low range wine. They both adjust to price imbalances.

272 Please insert table 4

273 4. Discussion

275 The preceding results reveal that Old World and New World countries follow different

276 trends in the transfer of prices linked to various productive and commercial strategies.

277 These conclusions are in line with those in Campbell (2000), Anderson (2001), Green

278 and Pierbattisti (2002), Anderson, Norman and Wittwer (2003) and Villanueva (2011).

279 In spite of normally being considered a commercial block, New World countries, unlike

280 Old World countries, do not share common behavior in their exporting dynamics as a

281 result of their different commercial positioning strategies and sale prices, as seen in

282 Villanueva (2011).

283 OW countries have a common export wine price dynamic; that said, France is in a

284 position of leadership followed by Italy. A similar conclusion was obtained by Bentzen

285 and Smith (2002) and Triguero (2002), with these three countries and the wine market,

286 and where France is the most independent market of the three countries. On the other

287 hand, Spain is the most dependent market and adjusts itself to price variations as they

288 happen. However, in the results by Bentzen and Smith (2002), Spain would not have

289 been so dependent on France and Italy in past decades.

291 On the other hand, although France is the "leader", it presents a dynamic that is not

292 followed by New World countries, because of its specialization in wines of the

293 Premium segments. Surprisingly, it is Italy and Spain that are cointegrated, linear and

294 non-linearly, with markets of New World countries, especially USA and Argentina. The

295 relationship between Spain-USA is similar to that of Italy-USA in the sense that the

296 relationship between their prices is non-linear, that is, prices need to reach a certain

297 threshold before the other market reacts. The relationship between Spain and Argentina,

298 on the other hand, is linear. Therefore, the reaction of one country in the face of price

299 variations in the other one takes place regardless of the threshold reached.

301 Prices in Italy and USA follow homogeneous trends in the long term, but in the short

302 term the reactions to prices are asymmetrical, possibly allowing price imbalances. USA

303 attempts to directly seek competitiveness against Italy. The degree of relationship

304 between these two markets is in accordance with the degree of integration that agrarian

305 markets have in the USA and Italy, which has been estimated in other studies

306 (Vasciaveo and Rosa, 2012). On the other hand, Spain attempts to directly seek

307 competitiveness against USA, fundamentally to penetrate its market and the English-

308 speaking profile. Italy is not that sensitive to price variation in the USA. In this same

309 line are the results obtained by Thach and Cuellar (2007) when they point out that the

310 price of Spanish wine is very sensitive to changes in the USA price, although also to the

311 French and Italian.

313 Lastly, Spain and Argentina are two integrated markets, with homogenous prices,

314 adjustments for possible price imbalances in the short term and equilibrium in the 1/p,

315 given that both countries compete strongly in the lower range wines, sold in bulk (with

316 containers that start at 2 liters) (Bernal y Mercado, 2006; Villanueva 2011).

318 5. Conclusions

319 To conclude, it is safe to state that for the period studied and according to the

320 econometric technique used, that the international wine market follows two different

321 paths. On one hand, the OW countries represent an integrated market given that, even in

322 spite of the price adjustment dynamics that might take place in the short and medium

323 term, eventually a trend towards balance is observed and on the other hand, the wine

324 market of the NW exporters do not represent an integrated market.

326 France, Italy and Spain follow a common exporting dynamic, sharing the same space of

327 cointegration and adjusting their prices in the long term in face of deviations that may

328 occur. France, followed by Italy, leads the wine market in the EU. And although they

329 manifest an asymmetry in the behavior of their prices in the short term, they do not

330 provoke price imbalance processes in the wine sector. New World countries, including

331 USA, do not share a common behavior in their exporting dynamic, because of their

332 different principal axes of target countries in their destination markets: USA is directed

333 towards Canada and Asian countries, Australia towards the United Kingdom, Chile

334 towards UK and Continental Europe, Argentina towards USA, etc. Within this group, it

335 has been USA that proves to be the most elastic market in adjusting its prices to the

336 standard international dynamics.

338 Interestingly, although there is no convergence, some wine segments, such as cask wine

339 as a commodity, explain the dynamics of old markets like Spain and Argentina. But the

340 most important aspect is the creation of an integrated market in the traditional EU

341 producer countries. In short, the integration between OW countries confirmed by our

342 results is largely a consequence -in line with Sanjuan and Gil (1997) - of a common

343 policy with common regulations and objectives which benefits the whole group of

344 countries when it comes to competing with other countries such as Australia or Chile, or

345 in the case of the USA, where the possibilities for the development of the wine sector

346 are high in view of the production capacity and the large domestic market there. Thus,

347 the union and synchronization of the wine market represents a substantial advance. This

348 characteristic is not observed in the New World countries, which have different policies

349 and objectives which result in a non-integrated wine market, not even in the potential

350 derivations that might have emerged from the Commonwealth commercial practices.

353 6. References

354 Anderson, K. 2001. The globalization and regionalization of wine. Center for

355 International Economic Studies, Adelaide University, Discussion Paper 0125.

356 Anderson, K. 2004. The World's Wine Markets: Globalization at Work. Massachusetts,

357 Edward Elgar.

358 Anderson, K., D. Norman, and G. Wittwer. 2003. Globalization of the World's Wine

359 Markets. London: Blackwell Publishing Ltd.

360 Baldi, L., D. Vandone, and M. Peri. 2010. Is Wine a Financial Parachute? Paper

361 prepared for the International European Forum, Innsbruck-Igls, Austria, February

362 8-12.

363 Baldi, L., D. Vandone, and M. Peri. 2013. Investing in the wine market: a country-level

364 threshold cointegration approach. Quantitative Finance 13: 493-503.

365 Barret, C., 2001. Measuring Integration and Efficiency in International Agricultural

366 Markets. Review of Agricultural Economics 23(1): 19-32.

367 Bentzen, J., and V.Smith. 2002. An empirical analysis of the interrelations among the

368 export of red wine from France, Italy and Spain. Working Paper 02-8, University of

369 Aarhus, Department of Economics.

370 Calderón, M., and L. Blanco. 2005. Análisis multivariado aplicado al mercado mundial

371 de vinos. Comunicación presentada a la XL Reunión anual, La Plata, Argentina,

372 Noviembre 16-18.

373 Campbell, G. 2000. Domestic demand and export imperatives for French and Australian

374 wines: a historical overview. PRATIC Université de Avignon.

375 Cassi, L., A. Morrison, and A. Wal. 2009. The evolution of knowledge and trade

376 networks in the global wine sector: a longitudinal study using social network

377 analysis. Working Paper 09-09, University of Utrecht, Section of Economic

378 Geography.

379 Dickey, D.A., and W.A. Fuller. 1979. Distribution of the estimators for autoregressive

380 time series with a unit root. J. Amer. Statist. Assoc. 74: 427-431.

381 Dickey, D. A., and W.A. Fuller. 1981. Likelihood ratio statistics for autoregressive time

382 series with a unit root. Econometrica 49: 1057-1072.

383 Doornik, J., and H. Hansen. 1994. An Omnibus Test for Univariate and Multivariate

384 Normality. http://www.doornik.com/research/normal2.pdf(Website accessed:

385 September 10, 2013).

386 Engle, R. F. and C. W. J. Granger. 1987. Cointegration and Error-Correction:

387 Representation, Estimation and Testing. Econometrica 55 (2): 251-276.

388 Espostia, R. and G. Listortib. 2013. Agricultural price transmission across space and

389 commodities during price bubbles. Agricultural Economics 44: 125-139.

390 Fackler, P. and B. Goodwin. 2001. Spatial Price Analysis. Handbook of Agricultural

391 Economics 1 (17): 973-1021.

392 Green, R., and L. Pierbattisti. 2002. Principales tendencias del mercado mundial de

393 vinos. Etude réalisée pour l'Estacion Experimental Agropecuaria (EEA) de

394 l'Instituto Nacional de Tecnologia Agropecuaria (INTA), Mendoza, Argentine.

395 Goodwin B. K. y Schroeder T. C. 1991. Cointegration Test and Spatial Price Linkages

396 in Regional Cattle Markets. American Journal of Agricultural Economics 73(2):

397 452-464.

398 Gujarati, D. 1995. Basic Econometrics.McGraw-Hill, New York.

399 Hansen, B., and B. Seo. 2002. Testing for two-regime threshold cointegration in vector

400 error-correction models. Journal or Econometrics 110: 293-318.

401 Hernández, N., Pañeda, C. and E. Ruiz. 2002. Las relaciones dinámicas en el mercado

402 internacional de la carne de vacuno. Revista de Economía Aplicada Número 29:

403 137 -149.

Ihle, R., von Cramon-Taubadel, S. and S. Zorya. 2009. Markov-Switching Estimation of Spatial Maize Price Transmission Processes between Tanzania and Kenya. American Journal of Agricultural Economics 91(5): 1432-1439.

Johansen, S. 1988. Statistical Analysis of Cointegrated Vectors. Journal of Economics Dynamics and Control 12: 231-254.

Mencet, M.N., M.Z. Firat, and C. Sayin. 2006. Cointegration analysis of wine export prices for France, Greece and Turkey. Paper prepared for presentation at the 98th EAAE Seminar 'Marketing Dynamics within the Global Trading System: New Perspectives', Chania, Crete, 29 June - 2 July.

Meyer, J. and S. von Cramon-Taubadel. 2004. Asymmetric Price Transmission: A Survey. Journal of Agricultural Economics. 55, 581-611.

Monke, E. and T. Petzel. 1984. Market Integration: An Application to International Trade in Cotton. American Journal of Agricultural Economics 66 (4): 481-487.

Moreira, V.H., J.L. Troncoso, and B.E. Bravo-Ureta. 2011. Eficiencia técnica de una muestra de productores chilenos de uva vinífera: Un análisis con fronteras de producción estocástica. Ciencia e Investigación Agraria 38:321-329.

Ng, S., and P. Perron. 2001. Lag Length Selection and the Construction of Unit Root Tests with Good Size and Power. Econometrica 69: 1519-1554.

OEMV (Observatorio español del mercado del vino) 2012. Principales exportadores mundiales de vino - año 2012. http://www.oemv.es/esp/-oemv.php. (Website accessed: September 13, 2013).

Orgun, A., and D. Temiz. 2011. Supply response of the wine industry in Turkey. Asia Life Sciences 20: 271-287.

427 Pesenti, P. 2011. The wine industry, a globalized market.

428 https://www.academia.edu/4067250/the_wine_industry_a_globalized_market

429 (Website accessed: January 15, 2014).

430 Pinilla, V., and M.I. Ayuda. 2002. The political economy of the wine trade: Spanish

431 exports and the international market. European Review of Economic History 6: 51432 85.

433 Pulido, A. and A. López. 1999. Predicción y simulación aplicada a la economía y

434 gestión de empresas, Ediciones Pirámide, Madrid.

435 Sanjuan, A.I., and J.M. Gil. 1997. Integración espacial de mercados: Revisión empírica

436 y aplicación al sector porcino de la UE. Invest. Agr.: Econ. 12 (1,2 y 3): 277-297.

437 Serra, T.; Goodwin, B.; Gil, J. and A. Mancuso. 2006. Non-parametric Modelling of

438 Spatial Price Relationships. Journal of Agricultural Economics 57 (3): 501-522.

439 Stephens, E.; Mabaya, E.; v.Cramon-Taubadel, S. and C. Barrett. 2010. Spatial

440 Adjustment with and without Trade. Charles H. Dyson School of Applied

441 Economics and Management. November 2010.

442 Thach, L., and S. Cuellar. 2007. Trends and implications for Spanish wine sales in the

443 US market. International Journal of Wine Business Research 19: 63-78.

444 Triguero, A. 2002. Integración de Mercados en la OCM del Vino: 1984-1996. Un

445 Análisis de Cointegración. PhD diss., University of Castilla La Mancha.

446 Vasciaveo M., and F. Rosa. 2012. Volatility in US and Italian agricultural markets,

447 interactions and policy evaluation. Paper prepared for the 123rd EAAE Seminar,

448 Dublin, February 23-24.

449 Villanueva, E. 2011. El boom exportador del "nuevo mundo vitivinícola" (C. 1975450 2005). PhD diss., University of Barcelona.

453 Table 1. Descriptive statistics and Stationarity contrasts.

ADF(a) p-value in levels

N Min. Max. Mean Tip.

ADF p-value in first differen ces

Ng-Perron p-value in levels

Ng-Perron p-value in first differences

France 98 3.98 27.00 12.01 6.90 0.3978 0.0001 -337.653 -571.460

Italy 98 2.05 12.17 5.89 3.54 0.3471 0.0000 -391.451 -470.252

Spain 98 0.22 7.73 2.78 1.86 0.1752 0.0000 -409.549 -665.434

Argentina 93 1.08 3.13 1.89 0.58 0.6481 0.0001 -266.189 -372.360

Australia 98 2.04 3.76 2.78 0.35 0.1064 0.0000 -900.239 559.428

Chile 98 1.22 2.09 1.56 0.19 0.2445 0.0001 -646.254 -127.610

South Africa 97 0.36 2.57 1.84 0.30 0.0000 0.0000 -426.334 -0.14348

USA 98 1.63 4.46 2.50 0.65 0.5618 0.0001 -122.126 -546.431

( ) p-value <0.5 rejects H0 of non-stationarity.

(b) contrasts on the MZa test. Asymptotic critical values (5% of significance is -8.10)

459 Table 2. VEC Results (Spain-Italy-France)

D(LES) D(LFR) D(LIT)

A -0.309514*** (8.81e-05) 0.007238 (0.8694) 0.059932 (0.2369)

D(LES(-1)) -0.330316*** (0.0022) -0.170810*** (0.0063) -0.193236*** (0.0070)

D(LES(-2)) 0.107957 (0.2947) -0.063567 (0.2902) -0.078292 (0.2561)

D(LIT(-1)) -1.081.924*** (0.0047) -0.021361 (0.9220) -0.138811 (0.5792)

D(LIT(-2)) -0.402938 (0.2465) -0.253374 (0.2119) -0.320917 (0.1683)

D(LFR(-1)) 1.761.330*** (0.0001) 0.004283 (0.9865) 0.153096 (0.5985)

D(LFR(-2)) 0.559155 (0.1770) 0.256024 (0.2883) 0.313240 (0.2574)

C 1.41133*** (0.0001) -0.0322978 (0.8743) -0.272266 (0.2465)

R-squared Sum sq. resids Durbin-Watson 0.380357 9.963278 2.080929 0.108905 3.389715 2.028627 0.124854 4.456385 1.974000

461 Cointegration Equation: 0(1)*LES(-1) + 0(2) *LFR(-1) +0(3)*LIT(-1)+ 0(4)

462 Estimated Cointegration Equation: 2.27*LES(-1) + 8.53*LFR(-1) - 7.93*LIT(-1) - 9.17

Contrast of exclusion

P-value. Coefficient

P(1) 0.000044 -2.27

P (2) 0.000293 8.53

P(3) 0.000194 -7.93

Weak exogeneity

P-value. Coefficient

X(1,1) 0.000162 -0.309514

X(2, 1) 0.875993 0.007238

X(3, 1) 0.243085 0.059932

464 In parenthesis p-value of t-statistics of coefficient estimates: * Denotes significance at the 10-percent

465 level; ** Denotes significance at the 5-percent level; *** Denotes significance at the 1-percent level.

466 (a) ES: Spain, FR: France, IT: Italy

Table 3. Selection of optimal delay and Johansen test and Hansen and Seo test, crossed pairs

Argentina

Australia

France

AIC(2)=-5.355816

BIC(1)=-5.126179

HQC(2)=-5.24586

DH(1)=2.10861e-048

DH(2)=5.49129e-050

LC-P(1)=0.43 LC-P(2)=0.3049 TC-P(1)=0.4 TC-P(2)=0.8

AIC(2)=-5.688023

BIC(1)=-5.490414

HQC(1)=-5.591274

DH(1)=7.25101e-035

DH(2)=1.4876e-032

LC-P(1)=0.1648 LC-P(2)=0.1287 TC-P(1)=1 TC-P(2)=0.7

AIC(1)=-6.449839 BIC(1)=-6.286446 HQC(1)=-6.383865 DH(1)=3.56607e-036

LC-P(1)=0.1211 TC-P(1)=0.8

AIC(1)=-6.415959 BIC(1)=-6.252566 HQC(1)=-6.349986 DH(1)=6.75112e-058

LC-P(1)=0.2587 TC-P(1)=0.4

AIC(2)=-5.020061 BIC(1)=-4.777246 HQC(2)=-4.910774 DH(1)=2.02009e-072 DH(2)=9.76041e-071

LC-P(1)=0.4290 LC-P(2)=0.3870 TC-P(1)=0*** TC-P(2)=0 ***

AIC(1)=-5.402182 BIC(1)=-5.234409 HQC(1)=-5.334558 DH(1)=1.68285e-057

LC-P(1)=0.1602 TC-P(1)=0.4

AIC(1)=-6.252377 BIC(1)=-6.090039 HQC(1)=-6.186805 DH(1)=8.34424e-048

LC-P(1)=0.1115 TC-P(1)=0.4

AIC(1)=-6.153847 BIC(1)=-5.991504 HQC(1)=-6.088275 DH(1)=1.40985e-082

LC-P(1)=0.3402 TC-P(1)=0.3

AIC(2)=- 4.120703 BIC(2)=-3.850140 HQC(2)=-4.011416 DH(2)=1.82403e-009

LC-P(2)=0.1296 TC-P(2)=0 ***

AIC(2)=- 4.506388

BIC(2)=-4.246706

HQC(2)=-4.393680

DH(1)=3.97009e-008

DH(2)=9.28355e-008

LC-P(1)=0.1627 LC-P(2)=0.0895* TC-P(1)=0.2 TC-P(2)=0.3

AIC(2)=-5.287065 BIC(2)=-5.016502 HQC(2)=-5.177778 DH(2)=5.22607e-012

LC-P(2)=0.1278 TC-P(2)=0.4

AIC(2)=-5.1446

BIC(1)=-4.9148

HQC(2)=-5.03534

DH(1)=3.15376e-012

DH(2)=2.37223e-008

LC-P(1)=0.1839 LC-P(2)=0.1541 TC-P(1)=0.9 TC-P(2)=0.4

The value within parenthesis is the order of the delay

AIC: Akaike criterion, BIC: Bayesian Schwartz criterion, HQC: Hannan-Quinn criterion, DH: Test for

multivariate normality of residuals of Doornik-Hansen.

LC: Linear cointegration (Johansen test)

TC: Threshold cointegration (Hansen and Seo test)

* Denotes significance at the 10% level

** Denotes significance at the 5%level;

*** Denotes significance at the 1% level.

478 Table 4. VEC Results de los pares cruzados: Italy -USA/ Spain-USA/ Spain - Argentina

Italy-USA

Spain-USA

Spain - Argentina

Down Up Down Up

D(LIT) D(LUSA) D(LIT) D(LUSA) D(LES) D(LUSA) D(LES) D(LUSA) D(LES) D(LARG)

CointEql 0.0034 -0.0041 0.7050 0.0354 CointEql -0.755*** 0.0059 -0.4089* 0.0298 CointEql -0.25750*** -0.0459994**

(0.9163) (0.8022) (0.2430) (0.9069) (1.7e-05) (0.9171) (0.0953) (0.7214) (0.0100) (0.0390)

D(LIT(-1)) -0.0411 -0.0180 0.0056 -0.1638 D(LES(-1)) 0.0030 0.0166 -0.7058** -0.0958 D(LES(-1)) -0.263021** -0.0311030

(0.6970) (0.7351) (0.9932) (0.6228) (0.9808) (0.6938) (0.0049) (0.2559) (0.0366) (0.2664)

D(LUSA(-1)) -0.2009 -0.4682*** 12.286 -0.1474 D(LUSA(-1)) -0.0180 -0 4794*** 11.164 -0.3807 D(LES(-2)) 0.0329143 0.0220483

(0.3956) (0.0002) (0.1623) (0.7380) (0.9625) (0.0004) (0.0665) (0.0682) (0.7729) (0.3903)

D(LIT(-2)) -0.0266 0.0295 0.2367 0.0719 D(LES(-2)) 0.2168* 0.0114 -0.3162 0.0368 D(ARG(-1)) -0.263030 -0.181841*

(0.8433) (0.6637) (0.2330) (0.4711) (0.0364) (0.7447) (0.1600) (0.6325) (0.5778) (0.0890)

D(LUSA(-2)) 0.0011 -0.0991 -14.124 -0.0620 D(LUSA(-2)) 0.0644 -0.1233 0.0868 0.0983 D(ARG(-2)) 0.312463 -0.120698

(0.9961) (0.3893) (0.1313) (0.8947) (0.8646) (0.3429) (0.8951) (0.6638) (0.4890) (0.2352)

c 0.0171 0.0047 -0.4566 -0.0123 C -0.0248 0.0077 0.1987 -0.0148 c 0.409775** 0.0835814**

(0.1371) (0.4154) (0.2058) (0.9460) (0.1986) (0.2466) (0.1280) (0.7395) (0.0175) (0.0304)

Cointegrating vector: (1, - 1.344116 ) Threshold Values: 0.5487169= vt^ Cointegrating vector: (1, - 0.4657158 ) Threshold Values: 0.1749564 = W( | R-squared S.sq. resids 0.268619 1153786 0.189670 0.581636

Percentage of Observations in each regime 80% 20% Percentage of Observations in each regime 63.2% 36.8% DurbinWatson 2007885 2048164

AIC-1002.452 BIC-938.605 SSR 0.9512823 Down: ax =0.0034, a2 = -0.0041 Up: ax =0.7050, a2= -0.0354 AIC -932.0447 BIC -868 Down: ax = -0.755, a, = Up: a= -0.4089, a= 0. .1978 SSR 1.764306 0.0059 0298 Exclusion ß(U) P(l,2) P-value 0.000631 0.011582 Coefficient 1 1.463264339

Exogeneity P-value

KU) 0.010266 -0.257503

H2, 1) 0.039586 -0.0459994

479 In parenthesis p-value of t-statistics of coefficient estimates: * Denotes significance at the 10% level; ** Denotes significance at the 5%;

480 *** Denotes significance at the 1%. IT: Italy, ES: Spain, AR: Argentina