Introduction

The EU import regime for fresh fruit and vegetables (F&V)is rather complex and changes according to products, partnercountries, and seasonality. There are several reasons explainingsuch complexity arising from the circumstance that the EU isat the same time the largest importer in the world and one ofthe most relevant producer. Therefore, the F&V import regimepursues different objectives that in some situations could alsobe conflicting. The protection and stabilization of revenues ofEU producer of F&V; the supply of large and differentiatedprovisions of F&V to EU consumers at reasonable price; theintegration of the import regime within the international rela-tionships that the EU is promoting, particularly with developingand neighboring countries, are the most relevant aims that theimport regime should help to attain.

One of the most controversial features of the import regimeis certainly the entry price system (EPS) that was introducedin 1995 after the signing of the Uruguay Round Agreement onAgriculture, replacing the old reference price system. The EPSis applied only to a limited number of products that are the mostrelevant to EU producers, while imports of other F&V productsare only subject to duties.

The main objectives of the old reference price system intro-duced in the 1972 Common Market Organization of F&V, thatare also pursued by the existing EPS, were to give a contribu-tion both to the stabilization of EU domestic prices of F&Vproducts and in the prevention of market crises that in the F&Vsector are rather frequent. The stabilization effects of the EPS,in the sense of reducing price variability cutting the lower tailof price distribution, may arise from the avoidance of importsfrom a partner country whose import price or, more exactly, anindex built on it, called Standard Import Value (SIV), is belowthe trigger entry price (TEP). The effect would be the conse-quence of the fact that in days in which the SIV of a productimported from a country is lower than the TEP of that productby less than 8%, besides the tariff, imports from that countryare also charged of a specific duty that is roughly equal to thedifference between the TEP and the SIV. If the SIV is below92% of the TEP, the specific duty applied besides the tariff is themaximum tariff equivalent (MTE). The amount of the MTE forthe different products is generally so high that its charge wouldmake imports unprofitable. However, since the system workson a consignment basis, it is possible to avoid the paymentof the specific tariff showing that the actual sale price of theconsignment was such that a lower duty was to be paid(Swinbank and Ritson, 1995; Agrosynergie, 2008). Moreover,importers may also avoid the payment of the specific duty wait-ing for custom clearance when the SIVs are higher than theTEP.

Shortly after the introduction of the EPS, Swinbank andRitson (1995) comparing the new import regime to the pre-vious, were rather skeptical about its ability to increase signifi-cantly the EU F&V market openness. In recent years, there hasbeen a growing number of papers and articles devoted to theanalysis of the EPS. Some of them tried to analyze its effectson trade flows of F&V in comparison to the previous importregime. Cioffi and Dell’Aquila (2004) highlighted that duringthe first years of implementation the EPS showed a selectioneffect on the growth of EU imports of F&V, preventing importsof low quality/price produce. However, a recent evaluation re-port on the EPS demonstrated that in recent years imports ofF&V products covered by the import regime grew at a ratenot differing from that shown by F&V not covered by the EPS(Agrosynergie, 2008). The econometric analysis by Emlingeret al. (2008) through a gravity model approach showed thatthe import regime had effects on the EU import flows of F&V,although for some products other factors should also be takeninto account. Goetz and Grethe (2009) by means of a multivari-ate statistic analysis approach showed that the relevance of theEPS is not homogeneous among different products and origins,being wider for more perishable products and for neighboringpartner countries.

Since the effects of the EPS on EU import flows of F&V arenot clear-cut, while the issue of the destiny of the EPS afterthe Doha round is open, it is worth to get deeper insight onthe different effects it produces. Particularly, the question ofunderstanding if the EPS contributes to EU domestic pricesstabilization, which is the main goal of the EPS, is achieved, isstill unresolved.

Recent papers by Garcia-Alvarez-Coque et al. (2009, 2010)found that the removal of the EPS, as well as the reduction ofthe TEP and of the specific tariff while keeping alive the EPS,would have moderate impact on prices of EU domestic products.Although the stabilization issue is not directly addressed in thisarticle, such findings also imply that the EPS would be effectivein price stabilization. These results were obtained simulatingchanges in the border measures with partial equilibrium modelsof four products that acknowledge the full effectiveness of theEPS in sheltering the EU domestic markets. The results are notvery dissimilar from those of Ant ´on-L ´opez and Atance-Muniz(2007), which make use of a simpler methodology based ontruncating price distributions.

The problem addressed by this article is twofold: on one hand,we assess the effectiveness of the EPS, trying to evaluate howEU F&V domestic prices determination process changes whenthe SIVs fall below the TEP and if the EU market becomesisolated from the competition of partner countries’ products;on the other hand, we measure the stabilization effects of theEPS. Of course, the size of the stabilization effect of the EUdomestic price of a product is linked either to the occurrence ofthe EU market isolation and to the size of imports with respectto the quantity of products traded in the EU internal market.The econometric approach we propose and the results that arederived by its application to the cases of study we analyzedrepresent the main contributions to the current debate on thefunctioning of the EPS.

In the next section, we outline shortly the features of EUmarket and import regimes of tomatoes and lemons on whichthe study is focused and that are more relevant to the followinganalyses. The econometric model is presented in Section 3,while results are set out and discussed in Section 4. Conclusionsand indications for further research are developed in Section 5.

The main features of the EU market and import regimeof tomatoes and lemons

In order to pursue the two objectives of our article, that isto evaluate the effectiveness of the EPS and measure the sta-bilization effects of the EPS, we had to consider how pricesof domestic productions and of imported goods are related toeach other. The task is not easy because of the peculiarities ofthe price discovery processes in the F&V markets. Since worldprices cannot be obtained by registering price of transactionson foreign markets and adding transportation costs to obtaincif prices as it happened in the past for the other EU importregime, in implementing the EPS, they are estimated by calcu-lating the SIVs starting from prices registered on the main EUmarkets.

The lack of prices on world markets, while the only availableinformation are EU domestic prices and SIVs, obliges to carryout the analysis on the effects of the EPS on EU prices tryingto evaluate what happens to their determination process whenthe SIVs fall below the TEP. This implies that the analysis willbe carried out only on products that show a large share of SIVsbelow the TEP, referred to countries whose exports to the EUhave a significant share of the market, contributing significantlyto the EU price determination process. On the contrary, theassessment of the stabilization effects of the EPS arising fromthe deterrence effects in the case of products whose SIVs arerarely lower than the TEP it is not possible.

Our empirical analysis hinges on the idea that if the EPS af-fects EU domestic prices of F&V, their daily data series cannotbe described by a random walk. We carry out an economet-ric analysis based on a model that derives its autoregressivestructure from the competitive storage model by Deaton andLaroque (1992, 1996), modified to take into account importsfrom other countries and the EPS. Starting from this model,we estimate and test a nonlinear threshold model as proposedby Tong (1978, 1990) and generalized in the multivariate casein Tsay (1998). In our analysis, the threshold is exogenousTable 1Preferential EP and TRQ granted to Morocco in 2006–2007 and monthly importsTariff Entry price MTE Pref. Tariff Pref. entry price MTE TRQ Import 06/07(%)(€/t )(€/t )(€/t )(€/t )(t)(t)October 14.4 626 298 0 461 298 10,000 10,198November 8.8 626 298 0 461 298 26,000 28,813December 8.8 626 298 0 461 298 30,000 34,780January 8.8 846 298 0 461 298 30,000 42,807February 8.8 846 298 0 461 298 30,000 45,513March 8.8 846 298 0 461 298 30,000 41,975April 8.8 1126 298 0 461 298 15,000 36,303May 14.4 726 298 0 461 298 4,000 12,671Jun–Sept 14.4 526 298 6,859Conditional quota 2006–2007 45,000and is given by the 92% of the TEP that distinguishes twodifferent regimes. By means of this nonlinear specification, wewill be able to identify price relationships of different marketsand products and how they change in the two regimes, assess-ing if the EPS insulates the EU domestic market when SIVs arebelow the TEP.

Since the estimation of threshold models requires an ade-quate number of cases for each regime, our analysis is confinedto products and partner countries that undercut frequently theTEP. For this reason, the analysis is concerning with four casesregarding the imports of tomatoes from relevant partner coun-tries as Morocco and Turkey and of lemons from Turkey andArgentina. It is worth to underline that in the study on the rele-vance of the EPS of Goetz and Grethe (2009) these four com-binations of products and origins belong to classes of higherrelevance of the EPS. Although the analysis is confined to alimited number of cases, we believe that this is fairly enoughto get useful insights on the effects of the EPS on EU domesticprices.

Tomatoes and lemons are particularly suited to the analysisthat we are going to develop, also because of the large number ofSIVs calculated and published by the EU Commission, since theEPS is applied all year long. In the case of tomato, the two mostrelevant EU partner countries are Morocco and Turkey, while forlemons the main partner countries are Argentina and Turkey.Given this choice, it must be deemed that the import regimeapplied to the tomatoes imported from Turkey and Moroccois modified by preferential trade agreements, although withdifferent rules.

EU imports of agricultural products from Turkey benefit froma preferential regime within the Custom Union agreement. Thepreferential regime is defined by Council Decision (1998) 1/98and for many fresh and processed F&V tariff exemptions orreductions are bound by tariff-rate quotas (TRQs) and importcalendars. EU imports of tomatoes from Morocco benefit froma zero tariff quota also subject to a reduced preferential TEPwhile the specific duty is the MFN one. The tariff quota was ini-tially agreed at 130,000 tons of tomatoes distributed in monthlyquotas from November to March (Council Decision, 1994).In subsequent years, the quota granted to tomato imports fromMorocco was gradually increased and spread in monthly quotasfrom October to May under the reduced TEP (Table 1).

Among EU members, Spain is the chief exporter of toma-toes, except in summer when the Netherlands takes over.Morocco is the main exporting country of tomatoes to the EU,with a share of about 80% of total exports. Import volumes oftomatoes from other partner countries are much smaller: Israeland Turkey have a share of about 7–8%. However, Turkey ex-ports tomatoes mainly during summer months, while exportsof Israel are made of different varieties of tomatoes. The com-petition between Spain and Morocco is very intense due to thegreat similarity of their production seasons, target markets, tech-nologies, and varieties (De Pablo Valenciano and Perez Mesa,2004). The period with the highest competition spans from Oc-tober to March, when imports from Morocco have zero tariffsif SIVs are above the preferential TEP and the volume of im-ports does not exceed the monthly quotas. Given these features,in our econometric analysis, we considered the tomato pricescollected in Almeria (ES) and the Moroccan SIVs (case a), aswell as Chateau-Renard (FR) prices and Turkish SIVs (case b).

As far as lemons are concerned, Spain is the main EU pro-ducer with an average harvested production in the three years2006–2008 of 681,400 tons according to Eurostat data. Spain isalso a net exporter of lemons to other EU countries. Globally,the EU is a net importer of lemons, around 400,000 tons peryear, with Argentina that is the main partner country supplying50–60% of total import while Turkey is the second partner coun-try with a share of 20%. Imports of lemons from Argentina aredistributed from May to October while those from Turkey spanfrom September to April. As far as the case study is concerned,we carried out the analysis on Murcia (ES) prices and TurkishSIVs (casec) and on Murcia (ES) prices and Argentinean SIVs(cased).

The econometric model

Considering that very often the SIVs of F&V products im-ported from partner countries are below their TEP only fora few days, to assess the effects played by the EPS on EUproducts prices, it is necessary to use daily data. Since pricesand SIVs are the only data available on a daily base, wehave to formulate a simplified market model in which theequilibrium affects only prices of domestic and importedproducts.

The dynamic structure of the econometric model we es-timate may be seen as a reduced form of the competitivestorage model proposed by Wright and Williams (1982) andDeaton and Laroque (1992, 1996). These models are based onthe idea that consumers can buy both goods that have beenstored from the previous periods as well as goods produced inthe same period. The cost of inventories to risk neutral hold-ers is given by the interest rate r paid on capital and by theshrinkage of stocks from one period to the next. Deaton andLaroque (1992, 1996) build and test a model using only prod-uct prices series whose dynamic is based on a unique stationaryrational expectation equilibrium, while the estimation proce-dures enables the identification of the parameters characteriz-ing the structural form of the model. According to the com-petitive storage model, prices follow an autoregressive processof order one, switching to a white noise process in stock outperiods.

The storage model could seem not appropriate to representprice determination processes in the case of F&V products,considering that they are generally highly perishable and itis only possible to store them for short periods whose lengthdepends on the products’ characters. Moreover, the storage ofF&V bears costs of refrigeration and conditioning that addto the shrinkage cost. The justification of our referring at thecompetitive storage model is based on the fact that we have toanalyze series of daily prices and therefore it seems plausibleto assume that products can be transferred from one day to theother.

To evaluate the effects of the EPS on EU domestic prices ofF&V, we must also introduce assumptions regarding the rela-tionships of such prices with that of imported products. At thisaim, we assume that price of imported products in the EU F&Vmarkets follows the model of determination in a large country.Moreover, since we have different prices for domestic and im-ported products, we also assume that the domestic and importedF&V products are imperfect substitute in the EU consumers de-mand. This would allow the presence of relationships betweenthe two prices and separate price determination models.

The reduced form representation of the price determinationmodel in the EU market is an AR(1) system of Eqs. (1) and (2)Pt = f (Pt −1, SIVt −1) + εIt, (1)SIVt = g(Pt −1, SIVt −1) +ε2t, (2)where Ptand Pt −1represent the daily prices of EU domesticproducts, respectively, at timet and t − 1, SIVtand SIVt −1arethe daily Standard Import Values, respectively, at time t and t− 1, f and g are two different functional forms, ε1tand ε2tareerror terms assumed to be identically independently distributedwith mean 0 and varianceσ2.

To assess the effect of the EPS on price determination of EUF&V markets, we adopted a threshold autoregressive (TAR)model. TAR models belong to the general class of nonlinearmodels. Introduced by Tong (1978) and later formalized byhimself (Tong, 1990), they have been widely used, because oftheir interpretability in many economic analysis (see amongothers Kapetanios and Shin, 2006). These models allow to in-clude nonlinearity by separating the data in two or more linearregimes according to a “threshold variable.” In our analysis,the two regimes are separated by the following exogenous anddeterministic switching variable (It):It =1 if SIVt −1 ≥ 0.92 · TEPt −10 if SIVt −1 < 0.92 · TEPt −1,where again TEP stands for trigger entry price. The variable Itallows to separate the data in two subsamples according to therelative position of SIVs with respect to the TEP.1

Including the indicator Itinto the system of Eqs. (1) and (2)results into a two-regimes threshold system specified by⎧⎪⎨⎪⎩Pt = It·{ f1( Pt −1 ,SIVt −1) + ε1t }+ (1 − It)·{ f2( Pt −1 ,SIVt −1) + ε3t}SIVt = It·{ g1( Pt −1 ,SIVt −1) + ε2t }+ (1 − It)·{ g2(Pt −1 ,SIVt −1) + ε4t} ,(3)whereεitrepresent the error terms assumed to be i.i.d. (0,σi2),wherei = 1, ...,4.

In our analysis, we are interested in testing if price seriesbehave differently when the SIVs are above (“normal” regime)or below the 92% of the TEP, that is when the maximum specificduty is applied. Through the specification (3), we will be ableto assess if the two regimes are different. The threshold modelcan be estimated if two conditions are satisfied: (1) a sufficientnumber of observations are attributed to each regime (Andrews,1993; Seo, 2008); (2) the estimated coefficients of the modelparameters differ in the two regimes.

A parametric form of system (3) may be obtained assum-ing an additive specification. The following threshold vectorautoregressive of order one (TVAR(1)) results:⎧⎪⎨⎪⎩Pt = It·αI1+ βI11Pt −1 + γI11SIVt −1 + εI1t}+ (1 − It)·αII1+ βII11Pt −1 + γII11SIVt −1 + εII1tSIVt = It·αI2+ βI21Pt −1 + γI21SIVt −1 + εI2t+ (1 − It)·αII2+ βII21Pt −1 + γII21SIVt −1 + εII2t,(4)where the superscript index is referred to the regime (I or II)and the two subscript indexes are referred to the i-th Eq. (1 or2) and to the lag order (one for specification 4).

For cases in which residuals autocorrelation might be anissue, we consider a specification of higher order. The TVARof n lags (n > 1) of Eq. (5) represents a general case:⎧⎪⎨⎪⎩Pt = It· αI1+nj −1βI1jPt −j +nj =1γI1jSIVt −1 + εI1t+ (1 − It)· αII1+nj =1βII1jPt −j +nj =1γI1jSIVt −1 + εII1tSIVt = It· αI2+nj −1βI2jPt −j +nj =1γI2jSIVt −1 + εI2t+ (1 − It)· αII2+nj =1βII2jPt −j +nj =1γI2jSIVt −1 + εII2t,(5)where the subscript j indicates the lag order. The order of lags inthe TVAR( n) is determined by the Schwarz Information Criteria(SIC).

In our analysis, TVAR models were estimated using thefull information maximum likelihood method, which is asymp-totically efficient among estimators of simultaneous equationmodel and, for well-specified model, is able to provide con-sistent estimates of parameters either with i.i.d. ∼ N(0,1) errorterms and with errors autocorrelation (Greene, 2004).

For the sake of simplicity, we present the interpretation of re-sults referring to the TVAR(1) specification of Eq. (4) since it iseasily extendable to the TVAR(n) model. If the price determina-tion processes of EU F&V products depend on the imports froma partner country and EPS affects such processes, the functionalforms of the first and the second regime would differ from eachother. However, since the threshold variable discriminates theregimes according to the relative position of SIVs with respectto the TEP, the different competitive behaviors at different pricelevels might contribute to change the parameters of price trans-mission. Under the assumptions in the system of Eqs. (1) and(2), SIVs and EU domestic prices should influence each other inthe first regime. If the influence of SIVs on EU domestic price isobserved in the first regime (γI11> 0) and it is also maintained inthe second regime ( γII11> 0), we cannot conclude that the EPSdoes isolate the EU market since the relationship stands in bothregimes. According to model (4), if the coefficients of SIVs γI11and γII11in the equations will be, respectively, higher than zeroand zero, we can say that the EPS is effective in avoiding cheapimports from a country, isolating the EU domestic market fromlow-price imports. Finally, when the estimated coefficients donot show an evidence that the SIVs affect the EU domesticprices in both regime we cannot conclude on the effectivenessof the EPS.

As far as tomato imported from Morocco is concerned, sincethe TRQ is binding, changes in the quota size agreed betweenEU and Morocco may have had effects on the market pricedetermination process. In order to capture the effects of thequota expansion from 150,676 to 175,00 tons in 2003 and ofthe introduction of a further conditional quota (45,000 tons) by2006, we modified the specification (4) accordingly⎧⎪⎨⎪⎩Pt = It·αI1+ βI11Pt −1 + γI11SIVt −1 + δI11D1+ δI12D2 + εI1t(1 − It) ·αII1+ βII11Pt −1+ γII11SIVt −1 + δII11D1 + δII12D2 + εII1tSIVt = It·αI2+ βI21Pt −1 + γI21SIVt −1 + δI21D1+ δI22D2 + εI2t(1 − It) ·αII2+ βII21Pt −1+ γII21SIVt −1 + δII21D1 + δII22D2 + εII2t,(4b)whereD1and D2are, respectively, dummy variables assumingvalue 1 for period from 2003 to 2006 and from 2006 to the endof the series. The TRQ expansion, allowing increased importsfrom Morocco, should lower the SIVs level, while the effecton EU domestic prices could be negligible due to the smalldimension of additional quota with respect to the total marketedquantities. Furthermore, if the EPS is effective in isolating theEU price the quota expansion might not influence the price andSIVs in the second regime.

The relationships among EU domestic prices and SIVs ofdifferent products and partner countries are quantified and com-pared through the coefficient η that normalize the regime-/product-specific impact multipliers.2

Empirical results

The daily prices data used to carry out the analysis wereextracted from the Agriview database of the European Com-mission, which includes daily prices of F&V collected on EUwholesale markets of different member countries. Data on dailySIVs, proxy of border prices of imports, are calculated by theEU Commission. All prices are reported in euro and expressedin current terms. Since price series collected on the differentEU markets within the Agriview database have several missingdata, the selection of markets on which to carry out the anal-ysis has been forced by data availability. The criteria adoptedin the selection were to pick the longest series with the smallernumber of missing price data. Choosing the market, using thecriteria of data availability, may have the limit that market inte-gration relationships could have hidden possible relationshipsbetween the SIVs and EU prices observed on other markets.Concerning the analysis, when price series we choose still havemissing data, they were omitted to obtain a full and continuoustime series.

The effectiveness of EPS in stabilization of EU tomato priceshas been analyzed through the cases of imports from MoroccoTable 2Descriptive statistics per regimesTomatoIregime IIregimeMean Median St. dev. Mean Median St. dev.(a) EU price (Almeria) 82.18 76.56 27.07 52.13 51.21 8.44SIVs (Morocco) 70.77 64.1 25.98 37.18 38.6 4.19(b) EU price (Chateau-Renard) 94.83 83.85 38.98 84.57 80 33.31SIVs (Turkey) 92.58 86 28.35 66.34 60.3 24.42LemonIregime IIregime(c) EU price (Murcia) 56.74 55 13.04 49.24 48.72 10.01SIVs (Turkey) 61.39 60 10.01 42.36 43 6.31(d) EU price (Murcia) 60.44 58.75 10.45 56.15 55 9.36SIVs (Argentina) 63.59 62 6.36 50.81 51.8 4.44and Turkey, which account, respectively, for 83% and 6% ofthe total fresh tomato imports from extra-EU countries. SinceEU imports from Morocco are mainly spread from November toMarch, when the import monthly quotas are effective and wider,the econometric analysis is built on daily price data related tothese months, using a time series starting on January 2000 andending on February 2007. On the other hand, since importsfrom Turkey are distributed from April to October also in thiscase daily data are limited to these months. The EU domestictomato price was collected on the Almeria (ES) wholesale mar-ket. Such market is of high relevance for a tomato-producingarea that is affected by the competition of products importedfrom Morocco. On the other hand, to analyze the relationshipsregarding the SIV of tomato imported from Turkey, we usedprices collected from the French market of Chateau Renard.3

To analyze the effectiveness of the EPS in the case of lemon,we considered the SIVs of imports from Argentina and Turkeywhile the EU domestic prices were collected on the Murcia(ES) wholesale market. This market is of high relevance for theSpanish lemon-producing area. The SIVs of lemons importedfrom Turkey and Argentina show a high share of values belong-ing to the second regime (17%, 124 out of 729, and 35%, 213out of 611, for Turkish and Argentinean imports, respectively).Conversely, as far as tomato imports are concerned, 13% and11% of SIVs of products imported, respectively, from Moroccoand Turkey pertain to the second regime.

Summarizing, the econometric analysis was applied to twocase studies related to the tomato market (casea is related toAlmeria [ES] prices and Moroccan SIVs, caseb is regardingChateau-Renard [FR] prices and Turkish SIVs) and two casesof the lemon market (case c stands for Murcia [ES] pricesand Turkish SIVs, case d concerns Murcia [ES] prices andArgentinean SIVs) (Table 2). Time series of daily prices andSIVs refer to weekdays from Monday to Friday and contain datafor the season in which transactions are registered: November–March (a); April–October (b); October–May (c); May–October(d). Prices from different years are combined to obtain a uniquesample and cover the periods 2000–2007 (case a), 2000–2004(caseb) and 1998–2006 (cases c and d) (Graph 1).

In order to carry out the analyses, some preliminary testsand transformations of time series were applied. Nelson andPlosser (1982) showed that a vast majority of economic se-ries could be better characterized by a unit root process ratherthan by a deterministic trend. Furthermore, according to otherauthors (Fama, 1995), price series are likely to follow a ran-dom walk process, that is a nonstationary process in whichthe autocorrelation function is one everywhere. This con-strains the number of applicable econometric techniques to thenonstationary ones. Alternatively, time series should be trans-formed into stationary time series. If the data are generated bya unit root process, subtracting a deterministic time trend isnot sufficient to produce a stationary process, while a correcttransformation could be into different time series (Hamilton,1994).

The presence of unitary roots in price series is usually testedby the conventional tests proposed by Dickey and Fuller (1979))and Philips and Perron (1988). The two tests were derived forthe null hypothesis of unit roots in linear time series. Taylor(2001) suggests to replace the unit root null hypothesis with astationary null when time series are expected to have nonlinearadjustments. We performed two different tests to assess thestationarity of time series: the DF-DLSu(Elliott, 1999), whichassume under the null the presence of a local unit root, andthe KPSS test (Kwiatkowski et al., 1992), in which the nullhypothesis is the stationarity of the time series. Overall, thetests reject the hypothesis of unit roots and fail to reject thestationarity of time series at 10% (Table 3) suggesting thatprice does not need transformations.

4.1. Results: tomato market

Final specifications for tomato cases are parsimonious andinclude only coefficients significant at least at 5%. In all equa-tions, the estimated coefficients of the lagged dependent vari-ables (prices or SIVs) assume a larger absolute value in thesecond regime, which is an evidence of their tendency to returnin the normal regime. In this case, the larger the absolute valueof lagged dependent the closer the behavior of time series to anonstationary process, hence the larger the tendency to switchto the normal regime.

As far as the case of tomatoes imported from Morocco isconcerned, the two regimes contain, respectively, 566 and 85observations. According to estimates in Table 4, the two regimesare different:4in the first regime βI11, γI11, βI21, and γI21are greaterthan zero and statistically significant (at 1% level), while in thesecond regime only βII11and γII21are statistically significant,meaning that the reciprocal influence of the EU domestic priceand the SIVs is lost in the second regime. We can interpret theresults as an evidence of the effectiveness of the EPS in isolatingthe EU market from SIVs when the latter are below the TEP.As regard the coefficients of quota dummies, only δI21δI22inTable 3DF-DLSuand KPSS test statisticsTo m a t o e sSpain—Morocco(case a) France—Turkey(case b)DF-DLSuDF-DLSuKPSS KPSS DF-DLSuDF-DLSuKPSS KPSS(1) (2) (3) (4) (1) (2) (3) (4)Almeria 1.248 2.739 0.192 0.194 Chateau-Renard 2.854 3.299 0.058 0.048Morocco 0.437 1.621 0.458 0.062 Turkey 2.315 4.046 0.095 0.079LemonsSpain—Turkey (case c) Spain—Argentina (case d)Murcia 1.789 4.640 0.377 0.136 Murcia 2.342 3.621 0.139 0.131Turkey 2.663 4.551 0.248 0.016 Argentina 0.729 2.339 1.056 0.108Table 4Model results for tomato marketsAlmeria (ES)—Morocco (a) Chateau-Renard (FR)—Turkey (b)Iregime(SIV > Trig) I regime (SIV> Trig)Almeria Morocco Chateau Renard Turkeyα 2.451∗∗(1.152) 2.661 (1.669) 9.581∗∗∗(3.427) 22.186 (4.194)Pt −10.898∗∗∗(0.026) 0.121∗∗∗(0.027) 0.892∗∗∗(0.035) 0.203∗∗∗(0.038)SIVt −10.081∗∗∗(0.025) 0.831∗∗∗(0.025) 0.531∗∗∗(0.044)D1 −2.073∗∗∗(0.756)D2 −2.131∗∗(0.921)II regime (SIV ≤ Trig) II regime (SIV≤ Trig)α 2.531 (9.366) 5.611 (8.286) 5.946∗∗∗(19.806) 2.062 (14.823)Pt −10.966∗∗∗(0.177) 0.942∗∗∗(0.223) 0.418∗∗∗(0.133)SIVt −10.893∗∗∗(0.235) 0.636∗∗∗(0.191)D1D2R20.93 0.90 0.80 0.61the second equation are statistically significant, and negative,as expected. Moreover, δI21, the coefficient that captures theeffect of the expansion of quota from 150,676 to 175,000 tonsis (in absolute value) smaller than δI22, which also considers thefurther expansion of 45,000 tons. In other terms, the dummycoefficients allow to show that, ceteris paribus, larger quotaslowered the SIVs level while effects on EU domestic prices werenot significant. In the first regime, the effect of EU domesticprice on SIVs ( ηSIVIPI = 0.141) largely exceeds, as expected, theeffect of SIVs on price (ηSIVIPIequals 0.069) while no multiplierscan be computed in the second regime. The results enforce theidea that SIVs follow the EU domestic price but the linkagebetween them is lost in the second regime.

As regard the Turkish tomatoes, estimated coefficients sup-port the idea that the EU domestic price influences the TurkishSIVs ( βI21and βII21are statistically significant), but not viceversa ( γI11and γII11are not statistically different from zero).4These results do not lead to conclude that EPS is ineffectivesince the SIVs do not influence the EU domestic price in bothregime but, according to the interpretation we provided in pre-vious paragraph, we cannot conclude that the EPS isolates theEU market.The multiplier of EU price on SIVs (ηPSIV)islowerin the first regime (0.198) than in the second (0.367) indicatingthat the EPS strengthens the influence of EU domestic price onSIVs when they are below the threshold.

4.2. Results: lemon market

The cross-correlogramms of residuals of estimated TVAR(1)models of lemon markets (Table 5) show that the autocorrelationof residuals is an issue, probably because for a product morestorable than tomato in daily data there is a longer memory ofthe price determination process.5Therefore, the next step in5Contrary, the cross-correlogramms of residuals estimated for time series incase (a) and (b) indicate that a TVAR(1) specification might be appropriate.our analysis of the series of lemon market was the estimationof a TVAR models of higher order. We employed the SIC forlag length selection.6Taking into account these results, weconsidered a TVAR(3) for case (c) and a TVAR(2) for (d).The preferred final specifications are parsimonious and includeonly coefficients statistically significant at least at 5% level.7The estimated coefficients of the lagged dependent variableassume a larger absolute value in the second regime, as weobserved for tomatoes, in all but one equation: the coefficientsof Argentinean SIVs are larger in the first regime.8

The TVAR model estimated to find the relationships betweenthe price of lemons on the Murcia market and the SIVs of Turkey(casec) confirms previous results: the estimated coefficient γI11is statistically significant, while γII11, γII12, and γII13are not statis-tically different from zero. Therefore, the estimates indicate thatEPS is effective in isolating the Spanish market from Turkishimports. The values of the two multipliers in the first regime,ηPISIVIand ηSIVIPIare, respectively, 0.079 and 0.032, leading toconsiderations similar to case a: in the normal regime the SIVsfollow the EU domestic price but the linkage between series islost in the second regime.

The results for case ( d) show that the only relationship be-tween series is the following: EU domestic price influencesArgentinean SIVs in the first regime (γI11equals 0.101 beingstatistically significant while the correspondent multiplier is0.091). Since the SIVs do not influence EU prices either in bothregimes in this case, we cannot conclude on the effectivenessof the EPS.Table 5Model results for lemon marketsMurcia (ES)—Turkey (c) Murcia (ES)-–Argentina (d)Iregime(SIV > Trig) I regime (SIV > Trig)Murcia Turkey Murcia Argentinaα 0.224 (1.028) 3.736 (2.679) 5.142∗∗∗(1.534) 11.560∗∗∗(3.366)Pt −10.963∗∗∗(0.013) 0.093∗∗∗(0.028) 0.913∗∗∗(0.020) 0.101∗∗∗(0.032)Pt −2Pt −3SIVt −10.027∗∗∗(0.012) 0.466∗∗∗(0.041) 0.386∗∗∗(0.048)SIVt −20.248∗∗∗(0.037) 0.316∗∗∗(0.033)SIVt −30.119∗∗∗(0.033)II regime (SIV ≤ Trig) II regime (SIV ≤ Trig)α 2.202 (2.315) 4.801 (4.626) 1.091 (1.247) 33.128∗∗∗(3.717)Pt −10.969∗∗(0.049) 0.984∗∗∗(0.026)Pt −2Pt −3SIVt −10.380∗∗∗(0.105)SIVt −20.365∗∗∗(0.070) 0.388∗∗∗(0.067)SIVt −30.230∗∗∗(0.064)R20.932 0.561 0.876 0.426

4.3. The stabilization effect of the EPS

The assessment of the stabilization effects of the EPS ispursued by evaluating changes of the first and second mo-ments of the distributions of interpolated EU prices and SIVsfrom estimated models. Analytically, we computed the meanand standard deviation of the samples under two scenarios:the first one simulates what is actually working with the EPS,while the second one is aimed at the simulation of a removalof the EPS under the assumption that the price determinationmodel estimated in the first regime would remain unchangedeven without the EPS. Therefore, under the two scenarios, wehave

:(1) The dynamics of time series are governed by the coeffi-cients estimated for the first and second regimes.

The dynamics of first regime are also true in the secondregime.

Computationally, the simulation of scenario 1 is made byinterpolating observed data using the estimated coefficientsof first regime when the observed SIVs are higher than thethreshold and the coefficients of second regime when the SIVsare below the threshold. The simulation of the second sce-nario is always made using coefficients estimated in the firstregime.

We expect EU prices and SIVs to be lower in mean and largerin standard deviation if no adjustments in price transmission areassumed (e.g., when coefficients of the first regime are adoptedto interpolate data either in the first and in the second regime).The results of simulation are expressed as percentage change9( %) of means and standard deviations computed under thesecond scenario with respect to those calculated in the firstscenario (Table 6), for the whole sample period of each modeland for each month we included in the sample.

Without the EPS the EU domestic prices and SIVs wouldbe lower on average in all but case (d) with the largest effectsdetected for cases (b) and (c). As regard price variability, wefound that for the EU domestic prices the removal of the EPSwould increase standard deviation in cases (a) and (c), respec-tively, by 0.12% and 0.64%. The variability of the SIVs wouldslightly increase in case (a) (0.04%) and more substantially incase (c) (3.5%). As far cases (b) and (c) are concerned, thesimulated removal of EPS shows that the variability would de-crease.

Looking at the simulated effects of a removal of the EPS atthe monthly level, we have decreases in EU domestic price inall months for cases (a) and (c) for which we detected a clearisolation effect played by the EPS. For cases (b) and (d), weobserved no clear effects. Monthly changes in price variabil-ity are consistent with observed changes in price means. Theremoval of the EPS would decrease monthly SIVs increasingtheir variability in all months only in case (c), while in case(a) there are monthly changes rather small of different signs. Incases (b) and (d), the simulation did not give clear-cut changes.

Summing up, we found that the EPS contributes to increasethe EU domestic prices means in three out of four cases and todecrease the standard deviations, stabilizing EU prices in cases(a) and (c) for which we detected a clear isolation effect due tothe EPS functioning.

Conclusions

This article presents an econometric analysis of the effectsof the EPS on the prices of EU F&V. It focuses on the cases oftomatoes and lemons, since for these two products the pricesof imports from the main EU partner countries frequently fallbelow the 0.92 of TEP, the condition under which the maximumduty is enforced. The following hypotheses were tested: whenthe price of imports fall below the 0.92 of TEP is there anyreaction in the prices of EU domestic produce? Is the EPSeffective in isolating the EU domestic market in such cases?What is the effect of the EPS in terms of price stabilization? Tothis aim, we specified a threshold model using the TEP as anexogenous threshold.

Since data used in the analysis are daily prices of domesticEU products sold on the main markets and the daily SIVs ofimported products from the main partner countries, we speci-fied a model whose autoregressive structure is derived from thecompetitive storage approach. Analysis of residuals from esti-mates showed that in the two cases of lemons the lag structureof the competitive storage model is not able to keep up with thedynamics of prices and SIVs asking for a different approach.The specification of TVAR models for lemons gave better re-sults from an econometric point of view. Overall, the performedeconometric analysis highlights that the EPS affects the pricedetermination processes of both tomatoes and lemons, in thesense that when SIVs are below the TEP the price determina-tion process of EU products follows a pattern different from theone shown when SIVs are higher.

In the cases of tomatoes and lemons imported, respectively,from Morocco and Turkey, while their SIVs affect the prices ofEU domestic product when they are higher than their threshold,this does not hold when SIVs are below this level. Econometricanalysis thus showed that, at least in these two cases, the EPSisolates the EU domestic prices determination process throughthe neutralization of the effects that low import prices couldexert. This relationship does not hold in the case of Turkishtomato imports, as well as of lemons imported from Argentina,since they never affect EU domestic prices. On the contrary, theEU domestic prices affect the SIVs, trough a linkage that is lostin the II regime in the case of Turkish tomato SIVs while it stillholds for Argentina lemons. These results may be due to thesmall import flows of tomatoes from Turkey compared to theEU domestic production and trade while in the case of lemonsfrom Argentina it may be due to a different seasonality betweendomestic supply and imports. However, this does not mean thatthe EPS is ineffective, indeed these results may be also due tothe effect of a poor integration between the markets on whichthese products are sold.

The analyses highlighted the effectiveness of the EPS in shel-tering the EU domestic market of F&V from low-priced importsonly in two out of four cases. However, the resulting stabiliza-tion effect, as well as the support effect on EU domestic prices,is rather small, particularly, in the case of tomatoes importedfrom Morocco. On this case, there are some evidences com-ing from other studies that could be compared with our results.Particularly, the recent study by Garcia-Alvarez-Coque et al.(2010) carried out through partial equilibrium models for someF&V produces, that includes tomatoes but not lemons, in thecase of tomatoes found that the removal of the EPS would havethe effect of a large increase in imports from Morocco thatin the period 2004–2006 would soar by a percentage between27.1% and 74.5% respectively under a “low” and “high” sce-nario. The last figure would produce a decrease of EU internalmarket prices in a range of−4.2% in October and −0.6% inMay. Similarly, Ant ´on-L ´opez and Atance-Muniz (2007) foundthat a 45% MTE cut would lead to cuts in domestic prices of upto 4.8% for tomatoes and 8.3% for lemons. These results showan impact of the EPS on EU domestic prices means larger thanthe one we found in our analyses. On the other hand, we foundthat the EPS has very limited impact on SIVs, while changessimulated by Garcia-Alvarez-Coque et al. (2010) on Moroccanprices are much wider.

Regardless of the difficulty in comparing results obtainedwith very different methodological approaches, the divergencecould be explained by means of two points

(1) Our econometric model is estimated using price data thatreflect a market in which the EPS is working while with-out it market agents would behave differently and there-fore estimated parameters would also be different;

(2) The price variations resulting from the simulation withthe partial equilibrium model of a removal of the EPS areobtained under an hypothesis of “high” elasticity scenarioto stress the effects of the different policy scenario, it istherefore possible, as Garcia-Alvarez-Coque et al. (2010)say, that such variations are overestimated and thereforewith a lower scenario they could be closer to those weestimated.

Ending this analysis, it is worth to ask what justifies tokeep working the EPS if its main objective, the stabilizationof EU domestic prices, is barely attained. The maintenance ofa complex system, as the one underlined by the EPS, cannotbe justified only on the ground that it improves market in-formation since other instrument could reach the same resultsmore efficiently. Probably, the removal of the system couldhelp to have more transparent rules to F&V tradewithout hurt-ing very much the level and the stability of EU F&V producers’incomes.

However, it must be acknowledged that price variability ofF&V, particularly low-price spikes when market crises occur,is still a relevant issue that deserves an appropriate set of pol-icy tools. However, the EPS does not seem belonging to suchset.

Acknowledgments

The authors would like to thank the anonymous referees andthe editor for their suggestions that contributed to substantialimprovements in this article.

Appendix

This appendix contains the detail on the calculation of thetime series means and standard deviations under the two sce-narios of presence and removal of the EPS.

Scenario 1EPt| I,II= It·αI1+nj =1βI1jPt −j +nj =1γI1jSIVt −j+ (1 − It) ·αII1+nj =1βII1jPt −j +nj =1γII1jSIVt −j,ESIVt| I,II= It·αI2+nj =1βI2jPt −j +nj =1γI2jSIVt −j+ (1 − It) ·αII2+nj =1βII2jPt −j +nj =1γII2jSIVt −j,St. devPt| I,II=St. devIt·αI1+nj =1βI1jPt −j +nj =1γI1jSIVt −j+ (1 − It) ·αII1+nj =1βII1jPt −j +nj =1γII1jSIVt −j,St. devSIVt| I,II=St. devIt·αI2+nj =1βI2jPt −j +nj =1γI2jSIVt −j+ (1 − It) ·αII2+nj =1βII2jPt −j +nj =1γII2jSIVt −j.

Scenario 2EPt| I=⎧⎨⎩αI1+n j =1βI1jPt −j +n j =1γI1jSIVt −j⎫⎬⎭,ESIVt| I=⎧⎨⎩αI2+n j =1βI2jPt −j +n j =1γI2jSIVt −j⎫⎬⎭,St. devPt| I=St. dev⎡⎣⎧⎨⎩αI1+n j =1βI1jPt −j +n j =1γI1jSIVt −j⎫⎬⎭⎤⎦,St. devSIVt| I=St. dev⎡⎣{⎧⎨⎩αI2+n j =1βI2jPt −j +n j =1γI2jSIVt −j⎫⎬⎭⎤⎦,wherePtand SIVtare, respectively, the interpolated EU pricesand interpolated SIVs, represents the information set, andmore precisely,I={αI, βI, γI} and II={αII, βII, γII} ,the bold Greek letters indicate the set of equation-specific pa-rameters (e.g., βI= βIi 1+···+βIin).

For cases in which the EPS is effective, we expect to observethe following:EPIItI <EPIItI,II ,E!SIVIItI"<E!SIVIItI,II",Std.dev.PIItI >Std.dev.PIItI,II ,Std.dev.!SIVIItI">Std.dev.!SIVIItI,II".

References

Agrosynergie, 2008. Evaluation of the system of entry prices and export re-funds in the fruit and vegetables sector. Accessed April 2008, available athttp://ec.europa.eu/agriculture/eval/reports/fruitveg/index_en.htm.Andrews, D. W. K., 1993. Tests for parameter instability and structural changewith unknown change point. Econometrica 61, 821–856.Ant´on-L ´opez, J. A., Atance-Muniz, I., 2007. Measuring domestic implicationsof tariff cuts under EU entry price regime, Contributed Paper, 1. Mediter-ranean Conference on Agro-food Social Scientists (103rd EAAE Seminar).Barcelona, Spain, April 23–25.Cioffi, A., Dell’Aquila, C., 2004. The effects of trade policies for fresh fruitand vegetables of the European Union. Food Policy 29, 169–185.Council Decision, 1994. Council decision of 19 December 1994 on the con-clusion of an agreement in the form of an exchange of letters between theEuropean Community and the Kingdom of Morocco on the regime for im-ports into the European Community of tomatoes and courgettes originatingin and imported from Morocco (95/35/EC).Council Decision, 1998. Council decision 1/98 of the EC-Turkey AssociationCouncil of 25 February 1998 on the trade regime for agricultural products.(98/223/EC). Off. J. Eur. Commun. L86/1.Deaton, A., Laroque, G., 1992. On the behaviour of commodity prices. Rev.Econ. Stud. 59, 1–23.Deaton, A., Laroque, G., 1996. Competitive storage and commodity price dy-namics. J. Pol. Econ. 104, 896–923.De Pablo Valenciano, J., Mesa, J. C. P., 2004. The competitiveness of Spanishtomato export in the European Union. Span. J. Agric. Res. 2, 167–180.Dickey, D. A., Fuller, W. A., 1979. Distribution of the estimators for autore-gressive time series with a unit root. J. Am. Stat. Assoc. 74, 427–431.Emlinger, C., Jacquet, F., Lozza, E. C., 2008. Tariffs and other trade costs:assessing obstacles to Mediterranean countries’ access to EU-15 fruit andvegetable markets. Eur. Rev. Agric. Econ. 35, 409–438.Elliott, G., 1999. Efficient tests for a unit root when the initial observation isdrawn from its unconditional distribution. Int. Econ. Rev. 44, 767–784.Fama, E., 1995. “Random walks in stock market prices.” Fin. Anal. J. 1, 75–80.Garcia-Alvarez-Coque, J. M., Mart´齨ez-Gomez, V., Villanueva, M., 2009. Atrade model to evaluate the impact of trade liberalisation on EU tomatoimports. Span. J. Agric. Res. 7, 235–247.Garcia-Alvarez-Coque, J. M., Mart ´齨ez-Gomez, V., Villanueva, M., 2010. Sea-sonal protection of F&V imports in the EU: impacts of the entry pricesystem. Agric. Econ. 41, 205–218.Goetz, L., Grethe, H., 2009. The EU entry price system for fresh fruits andvegetables – paper tiger or powerful market barrier? Food Policy 34, 81–93.Greene, W., 2004. Econometric Analysis. Pearson Prentice Hall, Upper SaddleRiver, NJ.Hamilton, J. D., 1994. Time Series Analysis. Princeton University Press, Prince-ton, NJ.Kapetanios, G., Shin, Y. C., 2006. Unit root tests in three-regime SETARmodels. Economet. J. 9, 252–278.Kwiatkowski, D., Phillips, P. C. B., Schmidt, P., Shin, Y., 1992. Testing thenull hypothesis of stationarity against the alternative of a unit root: how sureare we that economic time series have a unit root? J. Economet. 54, 159–178.Nelson, C. R., Plosser, C. I., 1982. Trends and random walks in macroeconomictime series: some evidence and implications. J. Monet. Econ. 10, 139–162.Philips P. C., Perron, P., 1988. Testing for a unit root in time series regression.Biometrika 75, 335–346.Seo, M. H., 2008. Unit root test in a threshold autoregression: asymptotic theoryand residual-based block bootstrap. Econ. Theory. 24(6), 1699–1716.Swinbank, A., Ritson, C., 1995. The impact of GATT agreement on fruit andvegetable policy. Food Policy 20, 339–357.Taylor, A. M., 2001. Potential pitfalls for the purchasing-power-parity puzzle?Sampling and specification biases in mean-reversion tests of the law of oneprice. Econometrica 69, 473–498.Tong, H., 1978. On a threshold model. In: Chen, C. H. (Ed), Pattern Recognitionand Signal Processing, Sijthoff & Noordhoff, Amsterdam.Tong, H., 1990. Non Linear Time Series: A Dynamical System Approach.Claredon Press, Oxford.Tsay, R. S., 1998. Testing and modeling multivariate threshold models. J. Am.Stat. Assoc. 84, 1188–1202.Wright, B .D., Williams, J. C., 1982. The economic role of commodity storage.Econ. J. 92, 596–614.