Price interrelationships in the EU cow milk markets

Abstract

This work evaluates the integration of cow milk markets in the European Union (EU) using monthly wholesale prices from 16 member states over 2003 to 2017 and rank-based cointegration techniques. The empirical results suggest that the degree of spatial market integration is high since the prices in 13 of the 16 national markets move to a large extent in sync with the average EU price for cow milk. Exceptions are the prices in Poland, Portugal and Lithuania. It appears that the differences in price dynamics may be associated with the time of accession to the EU, with Eurozone membership and with country-specific factors such as the availability of substitute commodities.

Keywords

cow milk EU prices market integration rank tests

Introduction

The price interrelationships in spatial (geographically separated) food markets of the European Union (EU) has long been an important policy issue and a focal point of empirical economic research. For the European Commission, an integrated, single market of food will render Europe a more attractive and competitive place for enterprises, production and innovation. The Single Market program, launched in 1985, led to elimination of all tariff and non-tariff barriers between member states by 1993. The Single Market Review, initiated in 2006, placed its emphasis on understanding the price adjustment mechanisms to changing economic conditions, thus making the single market policy more impact-driven and result-oriented. Despite the progress made, especially with regard to the increase in intra-EU volume of trade, statistical evidence suggests that there are still considerable price differences for food products even between neighbouring and comparable countries. More importantly, while price convergence had been quite pronounced in 2003 to 2009, it has slowed down since then and even reversed for some food commodity categories (Eurostat, 2017a). It is not accidental, therefore, that monitoring and benchmarking price differences is a cornerstone of the EU policy for the food sector (EC, 2014).

Research economists have a keen interest on spatial price interrelationships since the strength and the pattern of price linkages allow them to determine whether geographically separated markets are globalized (integrated) or regionalized (fragmented). In globalized markets, prices tend to move in tandem and integration is a prerequisite for economic efficiency (e.g. Asche et al., 1999; Goodwin and Piggott, 2001; Serra et al., 2006).

Over the last 25 years, there has been a number of empirical studies on the integration of the European food markets (e.g. Emmanouilides et al., 2014; Sanjuan and Gil, 2001; Serra et al., 2006; Zanias, 1993) focusing mainly on meats and, to a lesser extent, on other commodities such as cereals and olive oil. Far less attention has been paid to the milk market. This is quite surprising for three reasons: (a) Milk is produced in every single member state without exception and it is the EU’s number one single product sector accounting for about 12% of the value of the final agricultural production in 2016 (Eurostat, 2017b); (b) Intra-EU trade of milk and dairy products is very significant. According to the European Dairy Association (EDA), more than 85% of all milk produced by member states is consumed and commercialized within the EU (EDA, 2016). (c) Cow milk, that accounts for 97% of the total milk production, is a highly homogeneous commodity and as such ideal for assessing price linkages without the complexities arising under quality differentiation.

To the best of knowledge, there have been only two past empirical works on the integration of spatial (national) EU cow milk markets. Katrakylidis (2008) relied on monthly prices over 1980 to 2003 from Germany, France, Denmark, the Netherlands and Belgium and on linear cointegration analysis. He found that the five markets were very well integrated. Recently, Benedek et al. (2017) analysed the evolution of cow milk trade flows in the EU-28 over 2001–2012 using network analysis. They found that trade connections (measured in terms of number of partners per national market) have been getting stronger with time.

This work revisits the issue of integration of cow milk markets in the EU using price data from a panel of 16 member states over 2003 to 2017 and rank-based cointegration tests. The panel considered here accounts for 95% of the total production, and it includes national markets from all regions of the Union as well old, new, Eurozone and non-Eurozone members. With the rank-based cointegration tests, a researcher does not have to be explicit about the functional form of a cointegrating relationship. This is an important advantage over the linear tests given that, in most cases, the economic theory does not provide any guide about the appropriate functional form. At the same time, non-linear price relationships may well arise in the long-run due to frictions (e.g. transaction costs), agent heterogeneity and the properties of the underlying reaction functions. The rank-based approach is far less prone to misspecification relative to its alternatives. It has been utilized extensively to test for the validity of the purchasing power parity (PPP) theory (e.g. Chang and Su, 2013; Chang et al., 2011; Haug and Basher, 2011) but, thus far, not for assessing the integration of spatial farm commodity markets.

In what follows, the second section presents the analytical framework and the third, the data, the empirical models and the results. The last section offers conclusions and suggestions for future research.

Analytical framework

Let $x_{t}$ and $y_{t}$ $(t = 1, 2, ..., T)$ be two I (1) time series, here, cow milk prices in two national EU markets. Let also the long-run linkages between $x_{t}$ and $y_{t}$ be described by the general non-linear functional relationship

g (y_{t}) = f (x_{t}) + u_{t}

where $u_{t}$ is a stochastic error term. If $u_{t}$ is an I(0) time series, then a (potentially) non-linear cointegration relationship is said to exist between $x_{t}$ and $y_{t}$ (Breitung, 2001; Chung and Su, 2013). Functions g and f are typically unknown. They can be, however, approximated by the ranked series of $x_{t}$ and $y_{t}$ . The ranked series for $x_{t}$ is defined as $R_{T} (x_{t}) = R a n k {of x_{t} among x_{1}, ..., x_{T}}$ and the ranked series of $y_{t}$ , denoted by $R_{T} (y_{t})$ , is defined accordingly.

Under cointegration, the series $R_{T} (x_{t})$ and $R_{T} (y_{t})$ will evolve in sync; otherwise, they will tend to diverge. Breitung (2001) proposed a number of alternative statistics for testing the null hypothesis of no cointegration that are based on the evolution of the distance between $R_{T} (x_{t})$ and $R_{T} (y_{t})$ over time. This work relies on the $Ξ_{T}$ statistic that involves the least restrictive assumptions about $f (x_{t})$ and $g (y_{t})$ . It is computed as

Ξ_{T} = \frac{\sum_{t = 1}^{T} {(u_{t}^{R})}^{2}}{T \sum_{t = 2}^{T} {(u_{t}^{R} - u_{t - 1}^{R})}^{2}}

where $u_{t}^{R}$ stands for the least squared residuals obtained from the regression of $R_{T} (y_{t})$ on $R_{T} (x_{t})$ . Given that the ranked series diverge when there is no stable long-run relationship between $x_{t}$ and $y_{t}$ , the null hypothesis is rejected for values of $Ξ_{t}$ that are smaller than their respective critical ones.

When the rank-based test provides evidence in favour of cointegration between the time series of interest, one may proceed further to testing whether the long-run relationship is linear or non-linear. As shown by Breitung (2001), the later test involves two steps; in the first, one obtains the residuals, $e_{t}^{l}$ , from the standard linear dynamic ordinary least squares (OLS) (DOLS) model

y_{t} = α_{0} + \sum_{j = 1}^{p} α_{j} y_{t - j} + β x_{t} + \sum_{j = - q}^{q} γ_{j} Δ x_{t - j} + e_{t}, y_{t} = α_{0} + \sum_{j = 1}^{p} α_{j} y_{t - j} + β x_{t} + \sum_{j = - q}^{q} γ_{j} Δ x_{t - j} + e_{t}

where p and q are finite lag/lead orders; in the second, she(he) regresses $e_{t}^{l}$ on all variables appearing on the right hand side of (3) plus $R_{T} (x_{t}),$

e_{t}^{l} = a_{0} + \sum_{j = 1}^{p} a_{j} y_{t - j} + b x_{t} + \sum_{j = - q}^{q} c_{j} Δ x_{t - j} + ω R_{T} (x_{t}) + v_{t}

The relevant test statistic is the $T R^{2}$ (i.e. a score statistic), where $R^{2}$ is the coefficient of determination of (4). Under the null hypothesis (i.e. the long-run relationship is a linear one), the $T R^{2}$ follows the $χ^{2}$ distribution with one degree of freedom. If linearity is not rejected, the cointegrating equation has intercept $α_{0}$ and slope β obtained from the linear DOLS regression in (3). If linearity is rejected, the relevant intercept and slope are also obtained from (3) using, however, $R_{T} (x_{t})$ and $R_{T} (y_{t})$ in the place of $x_{t}$ and $y_{t}$ , respectively.

Results and discussion

The data

The EU-28 is among the leading producers of milk globally. In 2016, the volume of all milk produced in the Union was 168.3 million tonnes. The largest part of it (157.1 million tonnes) was delivered to dairies and it was processed into several fresh and manufactured products; the rest was utilized otherwise. Table 1 presents the volume of cow milk collected by dairies in 2016, by member state. Germany accounts for 21% of the total deliveries to dairies, followed by France (16%), the United Kingdom (9.1%) and the Netherlands (9%).

Table 1.

Cow milk collected by dairies (EU-28).

Member state	Cow milk volume^a	Member state	Cow milk volume^a	Member state	Cow milk volume^a
Belgium	3.88	France	24.55	Netherlands	14.32
Bulgaria	0.52	Croatia	0.49	Austria	3.09
Czech Republic	2.79	Italy	11.49	Poland	11.14
Denmark	5.36	Cyprus	0.20	Portugal	1.85
Germany	31.9	Latvia	0.81	Romania	0.95
Estonia	0.72	Lithuania	1.42	Slovenia	0.58
Ireland	6.85	Luxemburg	0.36	Slovakia	0.82
Greece	0.60	Hungary	1.55	Finland	0.24
Spain	6.88	Malta	0.04	Sweden	2.86
				United Kingdom	14.68

EU: European Union.

Source: Eurostat (2017c).

^aIn million tonnes.

The data for the empirical analysis are monthly wholesale cow milk prices (in Euro per 100 kg). They have been obtained from the Milk Market Observatory, they refer to the period January 2003 to December 2017 and they come from the 16 major cow milk-producing countries in the EU.¹ The national markets considered here are Germany (DE), France (FR), the United Kingdom (UK), the Netherlands (NE), Italy (IT), Poland (PO), Spain (SP), Ireland (IE), Denmark (DK), Belgium (BE), Austria (AU), Sweden (SE), Czech Republic (CZ), Portugal (PT), Hungary (HU) and Lithuania (LT). Taken together, these 16 markets account for about 95% of the total cow milk deliveries to dairies in the EU-28.

Figure 1 presents the evolution of cow milk prices in the 16 markets and in the EU-28 (weighted average of national prices) over 2003 to 2017. There are notable similarities among the prices series especially with regard to peaks and troughs but there are certain differences as well. For example, while in the majority of markets prices remained stable or declined up to 2007, prices in Poland, Lithuania, and in the United Kingdom showed generally upward trends. The behaviour of prices in the two new members (PO and LT) early in the sample is an indication that these markets had been in a process of catching up with the rest. Also, whereas prices in all countries exhibited a very strong recovery after the trough in 2009, prices in Spain remained at low levels until 2012.

Figure 1.

The evolution of cow milk prices over 2003 to 2017.

Table 2 presents summary statistics of cow milk prices for the period under study. Italy has been the market with the highest mean price followed by Sweden; Poland has been the market with the lowest mean price followed by Hungary. The Italian (Polish) cow milk producers received, on average, prices 12(15)% – higher (lower) – than the average EU-28 ones. Important differences appear to exist among the 16 markets with respect to price volatility (measured by the standard deviation). The prices in Lithuania and in the Netherlands have exhibited the highest volatility, while those in Italy and in Portugal the lowest. The ratio of the highest to the lowest volatility has been 2.01 indicating that the price risk for cow milk producers in the Lithuania has been (on average) 100% higher than that for their counterparts in Italy.

Table 2.

Summary statistics for cow milk prices.

National market	Mean	Standard deviation	Max	Min
DK	32.9	4.3	43.5	25.4
DE	31.3	4.7	42.5	22
IE	30.9	5.2	45.4	21.8
SP	31.5	3.4	45.4	26.9
FR	32.2	3.3	39.8	23.6
IT	35.3	2.8	41.8	29
NE	32.7	5.3	45.9	21.6
PO	26.8	5.1	37.2	14.6
UK	29.2	4.3	40	21.8
BE	30.5	4.9	43.2	20.8
CZ	28.3	4	38.3	21.1
LT	28.4	5.8	37	11
HU	27.7	3.9	36.9	19.8
AU	33	4.2	42.7	25.3
PT	31	3.2	39.5	26.2
SE	33.5	4.4	42.5	23.5
EU-28	31.6	3.7	40.2	24.4

DK: Denmark; DE: Germany; IE: Ireland; SP: Spain; FR: France; IT: Italy; NE: Netherlands; PO: Poland; UK: United Kingdom; BE: Belgium; CZ: Czech Republic; LT: Lithuania; HU: Hungary; AU: Austria; PT: Portugal; SE: Sweden; EU: European Union.

Source: Author’s calculations based on the time series available in the Milk Market Observatory.

The empirical models

Prior to cointegration analysis, the statistical properties of the price time series have been investigated using the augmented Dickey–Fuller and the Phillips–Perron unit root tests.² Note that for all markets but the Polish and the Lithuanian ones, the tests have been carried out without including any deterministic components. The lag length for the tests has been selected using the Bayesian information criterion (BIC). Table 3 presents the results according to which all the 17 time series are I(1).

Table 3.

Unit root tests on cow milk prices.^a

National market	ADF		PP
National market	Level	First difference	Level	First difference
DK	−0.196	−7.829^b	−0.113	−7.533^b
DE	−0.139	−6.49^b	0.036	−6.063^b
IE	−0.089	−7.988^b	−0.004	−7.938^b
SP	−0.052	−7.614^c	−0.123	−5.495^b
FR	0.098	−2.514^c	−0.02	−10.199^b
IT	0.018	−10.222^b	−0.052	−10.593^b
NE	−0.129	−10.649^b	−0.044	−10.61^b
PO	−2.401	−7.226^b	−1.961	−7.301^b
UK	0.107	−9.286^b	0.177	−9.244^b
BE	−0.141	−6.662^b	−0.199	−5.866^b
CZ	0.056	−5.102^b	0.154	−9.399^b
LT	−2.343	−7.932^b	−2.503	−5.226^b
HU	−0.122	−7.359^b	−0.224	−11.715^b
AU	0.075	−6.442^b	0.172	−10.546^b
PT	−0.446	−6.958^b	−0.455	−11.042^b
SE	−0.089	−9.398^b	−0.058	−9.314^b
EU-28	0.174	−7.138^b	0.045	−5.694^b

ADF: augmented Dickey–Fuller; PP: Phillips–Perron; DK: Denmark; DE: Germany; IE: Ireland; SP: Spain; FR: France; IT: Italy; NE: Netherlands; PO: Poland; UK: United Kingdom; BE: Belgium; CZ: Czech Republic; LT: Lithuania; HU: Hungary; AU: Austria; PT: Portugal; SE: Sweden; EU: European Union.

^aThe critical values for the tests without deterministic components are −2.578, −1.943 and −1.616 at the 1, 5 and 10% level of significance, respectively. The tests for PO and LT include a drift in the auxiliary regression. The critical values, therefore, are −3.467, −2.877 and −2.575 at the 1, 5 and 10% level of significance, respectively.

^bStatistically significant at the 1% or less.

^cStatistically significant at the 5% or less.

Following earlier empirical works on price interrelationships across space (e.g. Dang et al., 2017; Jana and Sulganova, 2012), the weighted average of the EU-28 prices has been employed in the present study as the benchmark (numeraire) price. Therefore, in calculating the various test statistics and in estimating the cointegrating equations, X will be the numeraire price and Y will be the price in a given national market. Also, because the sample size (180 observations) is not very large to allow employing the asymptotic critical values of Breitung (2001) for the $Ξ_{T}$ statistic, sample size–specific critical values have been obtained using Monte Carlo simulation based on 50,000 replications. The Monte Carlo study and all necessary estimations and testing have been carried out using R.

Table 4 presents the results of the pairwise rank-based cointegration tests. The null hypothesis of no cointegration has been rejected strongly (at the 1% level or less) for nine national milk markets (i.e. DK, DE, IE, FR, IT, NE, HU, AU and SE); it has not been rejected, however, for three markets (i.e. PO, LT and PT). Also, for four markets (i.e. SP, the UK, BE and CZ), although the empirical values of the $Ξ_{t}$ statistic are higher than the critical value at the 5% level, they are considerably lower than the critical value at the 10% level. Thus, for SP, the UK, BE and CZ, there is evidence of (relatively weak) integration.

Table 4.

Pairwise rank-based cointegration analysis.

National market	Sample value of the Ξ_Τ statistic^a	National market	Sample value of the Ξ_Τ statistic
DK	0.013^b	UK	0.027^c
DE	0.011^b	BE	0.028^c
IE	0.011^b	CZ	0.026^c
SP	0.028^c	LT	0.036
FR	0.008^b	HU	0.011^b
IT	0.014^b	AU	0.012^b
NE	0.007^b	PT	0.033
PO	0.037	SE	0.018^b

^aThe sample size–specific critical values are 0.0188, 0.0257, 0.0306, at the 1, 5 and 10% level of significance, respectively.

^bStatistically significant at the 1% level.

^cStatistically significant at the 10% level.

Eurozone membership, the time of accession to the EU (‘Old’ vs ‘New’ member) and the characteristics of national markets appear to play some role in the existence and the strength of price cointegration. With just two exceptions, namely PT and LT, the milk prices in Eurozone markets have turned out to be relatively strongly or weakly cointegrated with the EU-28 average. The milk markets, however, of older but not Eurozone members such as those of DK and SE appear to be very well integrated with the aggregate EU market. At the same time, prices in new and non-Eurozone members appear to have different dynamics; for example, prices in CZ and HU have evolved in sync with that at the EU-28 average while prices in PO have not. Finally, given that SP is an old Eurozone member and among the biggest cow milk producers, it is a bit surprising that its market is not very strongly interconnected with the rest. A possible explanation lies in the special characteristics of the Spanish milk market. Spain is the most important producer of sheep milk in the Union, meaning that national cow milk prices are influenced both by the cow milk prices elsewhere in the EU as well as from the national supply and demand for sheep milk.

Table 5 presents the results of the test for linear versus non-linear cointegration in 13 national markets; the 9 for which the null of no cointegration was rejected strongly and the 4 borderline cases where the rejection was weaker. To conduct the test, the p (lag) and q (lead) orders have been selected using the BIC. The null of linearity has not been rejected in all but two national markets, namely, FR (at the 5%) and BE (at the 10%). The findings of the present study (i.e. predominance of linear cointegration) are in line with those reported by Chang and Su (2013) for the PPP in East Asian countries but contrast with the results of Chang et al. (2011) for the PPP in G-7 countries.

Table 5.

Linear versus nonlinear cointegration.

National market	Sample value of the TR² statistic^a	National market	Sample value of the TR² statistic^a
DK	2.268	UK	0.140
DE	0.403	BE	2.978^b
IE	0.446	CZ	0.002
SP	0.287	HU	0.021
FR	3.962^c	AU	0.027
IT	0.226	SE	0.307
NE	0.876

^aThe critical values are 6.635, 3.841 and 2.706, at the 1, 5 and 10 percent level of significance, respectively.

^bStatistically significant at the 10% level.

^cStatistically significant at the 5% level.

Table 6 presents the estimated cointegrating vectors. For the linear cointegration case, the slope coefficient shows the response of milk price in a given national market to a unit change in the EU-28 average. The prices in the NE and in DE appear to be the most responsive while those in SE and in CZ the least. For the non-linear cointegration, the slope coefficient shows the change in the position (rank) of a price in a given national market induced by a change in the rank by one unit of the price at the EU-28 level. Prices in FR are much more responsive than those in BE. The magnitudes of slope coefficients are generally consistent with the empirical values of the $Ξ_{t}$ statistic in Table 4. A smaller value indicates stronger cointegration that translates into a higher price responsiveness.

Table 6.

Cointegrating vectors.

National market	Intercept	Slope	National market	Intercept	Slope
(a) Linear cointegration
DK	−1.371^a (0.026)	0.407^b (0.000)	BE	−1.059^c (0.072)	0.279^b (0.000)
DE	−4.406^b (0.000)	0.662^b (0.000)	CZ	−0.031 (0.951)	0.117^b (0.002)
IE	−3.498^b (0.000)	0.446^b (0.000)	HU	−1.765^a (0.022)	0.318^b (0.000)
SP	−0.823^c (0.076)	0.087^b (0.000)	AU	−0.592 (0.332)	0.281^b (0.000)
NE	−5.051^b (0.000)	0.749^b (0.000)	SE	0.127 (0.845)	0.148^b (0.000)
UK	−2.211^b (0.001)	0.234^b (0.000)
(b) Nonlinear cointegration
FR	5.801^b (0.000)	0.579^b (0.000)	BE	4.342^b (0.007)	0.219^b (0.000)

^aStatistically significant at the 5% level; p values in parentheses.

^bStatistically significant at the 1% level; p values in parentheses.

^cStatistically significant at the 10% level; p values in parentheses.

Conclusions

The objective of the present work is to assess the integration of national (spatial) cow milk prices in the EU. This has been pursued using monthly wholesale price data from the 16 most important cow milk-producing countries over 2003 to 2017 and a rank-based approach that allows cointegration analysis without imposing a priori a functional form on the underlying price relationship. The empirical results suggest that prices in nine national markets evolve strongly in sync with the EU-28 average price. Among these are leading producers such as Germany, France, the Netherlands and Italy. There are also four more markets the price developments in which are to a certain extent in line with those at the aggregate EU level. These include Spain and the United Kingdom. The price dynamics in Poland, Lithuania and Portugal, however, have turned out to be divergent from the rest. The 13 markets where prices are strongly or weakly cointegrated with the average of the EU-28 account for about 89% of the cow milk deliveries to dairies. One, therefore, is justified to conclude that, generally speaking, the national cow milk markets in the EU are well integrated. With regard to five markets (i.e. Germany, France, the Netherlands, Belgium and Denmark), the results here are in line with those of the earlier study by Katrakylidis (2008). Moreover, they appear to be consistent with the findings of Benedek et al. (2017) since stronger trade connections is a necessary condition for market integration.

The evolution of cow milk prices in a given national market is shaped by factors that are market-specific and factors are common to all EU markets. The relative importance of the two types of factors is reflected in the strength of the co-movement between the price in the national market and the EU-28 average. Well-integrated national markets are more exposed to systemic (whole) EU market risk while for less integrated, the country-specific risk is more relevant. The results of the present study appear to offer some evidence that the time of accession to the Union, the participation in the Eurozone and the availability of substitutes to cow milk (e.g. sheep milk) may be among the factors influencing the price interrelationships across space.

This empirical study (as the overwhelming majority of similar studies in the past) assumes that the pattern of price co-movement between spatial market pairs remains stable over time. It is entirely possible, however, that price pairs exhibit similar dynamics in certain periods and diverging in others. A potential extension of this work, therefore, may involve rolling rank-based cointegration analysis that allows a researcher to determine periods where prices are cointegrated and periods where they are not. This, in turn, may shed some light to the question whether the process of price convergence in the EU has been recently intensified or reversed.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Asche

Bremnes

Wessels

(1999) Product aggregation, market integration, and relationships between prices. American Journal of Agricultural Economics 81 : 568–581.

Benedek

Bakucks

Falkowki

, et al. (2017) Intra-EU trade of dairy products: insights from network analysis. Studies in Agricultural Economics. DOI: 10.7896/j.1621.

Breitung

(2001) Rank tests for nonlinear cointegration. Journal of Business and Economic Statistics 19:331–340.

Chang

Lee

, et al. (2011) Revisiting the PPP for G7 using non parametric the rank test for cointegration. Applied Economic Letters 18:1795–1800.

Chang

(2013) Revisiting the PPP for Asian countries using the rank non linear cointegration. Applied Economics 45:2847–2852.

Dang

Yang

Chan

(2017) Yen or Yuan? The law of one price and economic integration in Asia. World Economy 41(1):171–193. DOI: 10.1111/twec.12529.

EDA (2016) Economic Report 2016/17. Available at: http://eda.euromilk.org/publications/read/article/edaewpa-economic-report-201617.html (accessed 14 November 2017).

Emmanouilides

Fousekis

Grigoriadis

(2014) Price dependence in the principal EU olive oil markets. Spanish Journal of Agricultural Research 12:3–14.

European Commission (2014) Available at: https://ec.europa.eu/growth/sectors/food/competitiveness/supply-chain-forum_en (accessed 14 November 2017).

10.

Eurostat (2017a) Available at: http://ec.europa.eu/eurostat/statistics-explained/index.php/Comparative_price_levels_for_food,_beverages_and_tobacco (accessed 5 December 2017).

11.

Eurostat (2017b) Available at: http://ec.europa.eu/eurostat/statistics-explained/index.php/File:Output_of_the_agricultural_industry,_EU-28,_2016.png (accessed 5 December 2017).

12.

Eurostat (2017c) Available at: http://ec.europa.eu/eurostat/statistics-explained/index.php/Milk_and_milk_product_statistics (accessed 29 November 2017).

13.

Goodwin

Piggott

(2001) Spatial market integration in the presence of threshold effects. American Journal of Agricultural Economics 83:302–317.

14.

Haug

Basher

(2013) Linear or non linear cointegration in the PPP relationship? Applied Economics 43:185–196.

15.

Jana

Sulganova

(2012) The price convergence of the Czech Republic and Eurozone countries. Journal of Competitiveness 4:49–68.

16.

Katrakylidis

(2008) Testing for market integration and the law of one price: an application to selected European milk markets. International Journal of Economic Research 5:93–104.

17.

Sanjuan

Gill

(2001) Price transmission analysis: A flexible methodological approach applied to European pork and lamb markets. Applied Economics 33:123–131.

18.

Serra

Gil

Goodwin

(2006) Local polynomial fitting and spatial price relationship: Price transmission in EU pork markets. European Review of Agricultural Economics 33:415–436.

19.

Zanias

(1993) Testing for integration in European Community agricultural product markets. Journal of Agricultural Economics 44:418–427.