University of Patras -

Department of Business Administration of Food and Agricultural Enterprises

PhD Dissertation

Examination of the price transmission and volatility of the main agricultural and food

product categories

Dimitris N. Pachis

October 2015

i

University of Patras - Business Administration of Food and Agricultural Enterprises

Dimitris N. Pachis

© 2015 – All rights reserved

ii

iii

Acknowledgments

I would like to thank the department of Business Administration of Food and

Agricultural Enterprises and professor Anthony Rezitis for giving me the opportunity to pursue

my PhD thesis in the University of Patras. Professor Rezitis was the supervisor of my PhD

thesis from September 2011 to August 2015. Moreover, I would like to thank assistant professor

Achilleas Kontogeorgos for administrating the process of my doctoral examination since

professor Rezitis was not eligible as he was on a leave of absence from the university.

Furthermore, I would like to thank the members of my advisory committee: professor

Konstantinos Adamidis and professor Dimitris Kirikos. I am also grateful to the personnel of

the department and especially Mrs. Anna Skepetari for their valuable help. Next, I would like

to convey my thanks to my university colleagues for their support during these four years. Those

that left and those that stayed. Special thanks go to Ourania Katsara for her help on the

proofreading of my manuscripts.

Last but not least, I would like to express my indebtedness to my family and wife for their

continuous support and encouragement over the last four years without which I would never

have the chance to complete the doctoral program.

This research has been co-financed by the European Union (European Social Fund – ESF)

and Greek national funds through the Operational Program "Education and Lifelong Learning"

of the National Strategic Reference Framework (NSRF) - Research Funding Program:

Heracleitus II. Investing in knowledge society through the European Social Fund.

iv

Abstract

The Greek agricultural sector constitutes a major part of the national economy thus the

study of the supply chain can offer valuable insights of the interactions among market

participants. Especially, the price interactions among the market levels mirror the interactions

between the market fundamentals namely supply and demand. Moreover, the study of the price

transmission mechanisms of the agricultural markets reflects the formation of the economic

decisions of the agents based on market prices which reveal the quantities they are willing to

buy and sell. Furthermore, the market prices monitoring by policy makers aims at assessing the

competition conditions in the market and assuming relative action when the welfare of market

participants is at stake especially when they do not have the means to accommodate price

changes.

This study investigates the price transmission mechanisms of the Greek agricultural

sector while focusing to three fresh vegetables namely potatoes, tomatoes and cucumbers. The

price mechanisms are examined in conjunction to the Common Agricultural Policy framework.

The price mechanisms are modeled via Vector Autoregressive processes and their Co-

integrated variants for the mean and via Generalized Autoregressive Conditional

Heteroskedastic processes for the variance.

Three models are utilized for the examination of the price mechanisms. The first model

investigated the price transmission mechanism between the producer and the consumer of the

Greek food market while assessing the effects of the decoupling principle of the Common

Agricultural Policy of the European Union. The empirical analysis uses the panel Vector Error

Correction model for the depiction of the price mechanism. The second model investigates the

price transmission mechanisms between the producer and the consumer for the fresh potatoes,

tomatoes and cucumbers. The three products are selected on the base that they comprise a

significant part of the vegetables production as well as of the total agricultural output while they

do not need big investments for their production. The empirical analysis makes use of the

Markov Switching Vector Error Correction model under the assumption that the producer and

the consumer respond differently to price increases and decreases. The third model investigates

the transmission of the volatility between the producer and consumer prices of fresh potatoes,

tomatoes and cucumbers in Greece. The transmission mechanism of the volatility of the prices

is modelled by the BEKK model while possible asymmetries are taken into consideration.

The results of the study reveal that for the total agricultural sector, the producer does

not respond to long-run price changes while the own effects of a shock are quickly absorbed,

v

in contrast to the spillover effects of the shock. The implementation of decoupling seems to

have benefited the consumer more in mitigating his responses to price shocks rather than the

producer. Finally, the analysis of the fresh vegetable markets reveals the complexity of the price

mechanisms and the diversified response of each product to price shocks.

vi

Περίληψη

Ο ελληνικός αγροτικός τομέας αποτελεί βασικό κομμάτι της εθνικής οικονομίας

συνεπώς η μελέτη της αλυσίδας τροφοδοσίας μπορεί να προσφέρει σημαντικές πληροφορίες

για την αλληλεπίδραση μεταξύ των συμμετεχόντων στις αγορές αγροτικών προϊόντων. Πιο

συγκεκριμένα, οι αλληλεπιδράσεις τιμών μεταξύ των διάφορων επιπέδων της αγοράς

αντικατοπτρίζει τα θεμελιώδη μεγέθη της αγοράς τα οποία είναι η προσφορά και η ζήτηση. Η

μελέτη των μηχανισμών μετακύλισης τιμών των αγροτικών προϊόντων αντανακλά το πώς

διαμορφώνονται οι οικονομικές αποφάσεις των συμμετεχόντων στις αγορές αποφασίζοντας τις

ποσότητες που θα αγοράσουν ή θα πουλήσουν σύμφωνα με τις επικρατούσες τιμές. Επιπλέον,

η παρακολούθηση των τιμών από τους διαμορφωτές πολιτικής έχει σαν στόχο την εκτίμηση

των συνθηκών ανταγωνισμού στην αγορά ώστε σε περίπτωση που η ευημερία των

συμμετεχόντων απειλείται να λαμβάνονται τα κατάλληλα μέτρα ειδικά στη περίπτωση που δεν

έχουν τη δυνατότητα να αντιμετωπίσουν μόνοι τους τις μεταβολές των τιμών.

Αύτη η εργασία μελετά τους μηχανισμούς μετακύλισης τιμών του ελληνικού αγροτικού

τομέα εστιάζοντας σε τρία νωπά λαχανικά τα οποία είναι οι πατάτες, οι ντομάτες και τα

αγγούρια. Οι μηχανισμοί μετακύλισης τιμών εξετάζονται σε συνάρτηση με το πλαίσιο που

θέτει η Κοινή Αγροτική Πολιτική. Οι μηχανισμοί μετακύλισης μοντελοποιούνται μέσω των

Αυτόπαλίνδρομων υποδειγμάτων και των αντίστοιχων συν-ολοκληρωμένων παραλλαγών τους

για τον μέσο ενώ για τη διακύμανση μέσω των υπό Συνθήκη Ετεροσκεδαστικών μοντέλων.

Τρία μοντέλα χρησιμοποιούνται για την μελέτη του μηχανισμού τιμών. To πρώτο

μοντέλο εξετάζει το μηχανισμό μετακύλισης τιμών μεταξύ του παραγωγού και του

καταναλωτή στην ελληνική αγορά τροφίμων ενώ αποτιμά τα αποτελέσματα των

αποδεσμευμένων ενισχύσεων της Κοινής Αγροτικής Πολιτικής της Ευρωπαϊκής Ένωσης. Η

εμπειρική ανάλυση βασίζεται σε ένα πάνελ υπόδειγμα Διόρθωσης Λάθους για την απεικόνιση

του μηχανισμού των τιμών. Το δεύτερο μοντέλο μελετά το μηχανισμό μετακύλισης τιμών

μεταξύ του παραγωγού και του καταναλωτή για τις νωπές πατάτες, ντομάτες και αγγούρια.

Αυτά τα τρία προϊόντα επιλέχθηκαν με βάση το γεγονός ότι αποτελούν σημαντικό μέρος της

παραγωγής λαχανικών και της αγροτικής παραγωγής γενικότερα ενώ δεν απαιτούν μεγάλες

επενδύσεις για τη παραγωγή τους. Η εμπειρική ανάλυση κάνει χρήση ενός Μαρκοβιανού

Υποδείγματος Διόρθωσης Λάθους υπό την υπόθεση ότι ο παραγωγός και ο καταναλωτής

ανταποκρίνονται διαφορετικά σε μια αύξηση ή μια μείωση τιμών. Το τρίτο υπόδειγμα

μοντελοποιεί τη μετακύλιση της μεταβλητότητας μεταξύ των τιμών του παραγωγού και του

καταναλωτή για τις νωπές πατάτες, ντομάτες και αγγούρια. Ο μηχανισμός μετακύλισης της

vii

μεταβλητότητας των τιμών υποδειγματοποιείται με ένα BEKK μοντέλο ενώ πιθανές

ασυμμετρίες λαμβάνονται υπόψη.

Τα αποτελέσματα της μελέτης αποκαλύπτουν ότι για το σύνολο του αγροτικού τομέα,

ο παραγωγός δεν αντιδρά σε αποκλίσεις από τη μακροχρόνια ισορροπία ενώ οι ίδιες επιδράσεις

ενός σοκ απορροφώνται γρήγορα σε αντίθεση με τις κατανεμημένες επιδράσεις του. Η

εφαρμογή των αποδεσμευμένων ενισχύσεων φαίνεται να ωφέλησε περισσότερο τους

καταναλωτές από ότι τους παραγωγούς σε ότι αφορά την αντίδραση τους στα διάφορα σοκ

τιμών. Τέλος, η ανάλυση των αγορών φρέσκων λαχανικών αποκαλύπτει τη πολυπλοκότητα

των μηχανισμών τιμών καθώς και τη διαφοροποιημένη αντίδραση κάθε προϊόντος στα σοκ

τιμών.

viii

Table of Contents

1. Introduction

1.1 …...Studying the prices of the major Greek agricultural products and product categories (1)

1.2 …………………………………………………………………………..........Objectives (3)

1.3 …………………………………………………………...........Organization of the Study (5)

2. Market description

2.1 ……………………………………………………………The Greek agricultural sector (6)

2.2 ……………………………........The main agricultural products and product categories (10)

2.2.1 …...…………………………………………………………………………...Cereals (10)

2.2.1.1 ……………………………………………………………………………........Rice (12)

2.2.2 ……………………………………………………………………Olive oil and olives (12)

2.2.3 …………………………………………………………………………………..Wine (13)

2.2.4 ………………………………………………………………….Vegetables and fruits (14)

2.2.4.1 ……………………………….Selected vegetables production, imports and exports (15)

2.2.4.2 …………………………………………………………………….........Citrus fruits (18)

2.2.5 …………………………………………………………………………Meat products (19)

2.2.5.1 …………………………………………………………………………………Eggs (20)

2.2.5.2 ………………………………………………………………………………...Milk (20)

2.3 …………………………………………………........The Common Agricultural Policy (22)

2.3.1 …………………………………………………...The Common Market Organization (24)

2.4 …………………………………………………..Studies of the Greek vegetables sector (25)

3. Price transmission along the Greek food supply chain in a dynamic panel framework:

Empirical evidence from the implementation of decoupling

3.1 …………………………………………………………………………….Introduction (26)

3.2 ……………………………………………………………..Econometric Methodology (30)

3.3 ………………………………………………………………………….…………Data (34)

3.4 ………………….…………………………………………………...Empirical Results (34)

3.4.1 ……………………….…………………….Unit Roots and Co-integration Analysis (34)

3.4.2 ……………………………….………………...The results of the panel VEC model (18)

3.5 ………………………………………….…………………………………Conclusions (49)

ix

4. Investigating the price transmission mechanisms of Greek fresh potatoes, tomatoes and

cucumbers markets

4.1 ………………………………………………….………………………….Introduction (52)

4.2 ………………………………………………………….…..Econometric Methodology (56)

4.2.1 …………...........The asymmetric Markov Switching Vector Error Correction model (56)

4.2.2 …………………………….………………The analysis of the error correction model (59)

4.3 ……………………………………………………………………………………..Data (61)

4.4 …………….…………………………………………………………Empirical Results (63)

4.4.1 ………………….………………………….Unit Roots and Co-integration Analysis (63)

4.4.2 ……………….The Asymmetric Markov Switching Vector Error Correction Model (65)

4.5 …………………….……………………………………………………….Conclusions (80)

5. Investigating the price volatility transmission mechanisms of selected fresh vegetable

chains in Greece

5.1 ………….…………………………………………………………………Introduction (84)

5.2 ………………….………………………………………………………..Methodology (88)

5.3 ………………………….…………………………………………………………Data (93)

5.4 ………………………………….…………………………………...Empirical Results (94)

5.4.1 ………………………………………………Unit Roots and Co-integration Analysis (95)

5.4.2 …………………….……………...The results of the estimation of the BEKK models (96)

5.5 ……………………………………………………………………………..Discussion (110)

5.6 ……………………………………….…………………………………...Conclusions (111)

6. Conclusions

x

Tables

Table 2.1 ……………The shares of Greek agricultural products in total output 2009-2013 (6)

Table 2.2 ……………………….The agricultural income per product category 2000 – 2005 (9)

Table 3.1 …………………………………………………………………...Causality tests (32)

Table 3.2 ………………………………………………….The in-sample asymmetric tests (33)

Table 3.3 The products under investigation and their weights in the Greek Producer Price

Index (34)

Table 3.4 …………………………………………..The results of the panel unit root tests (35)

Table 3.5 ……………………………………...The results of the panel co-integration tests (35)

Table 3.6 ……………..The co-integrating vector estimates according to FMLS and DOLS (36)

Table 3.7 The panel VEC estimation results of the Greek agricultural sector from 1995 to

2013 (38)

Table 3.7a ………………….The tests of short and long run causality from 1995 to 2013 (39)

Table 3.8 ………………….The asymmetric model estimation results from 1995 to 2013 (42)

Table 3.8a The asymmetric in-sample asymmetric tests for the Greek agricultural sector for

the period 1995-2013 (44)

Table 3.9 The panel VEC estimation results of the Greek agricultural sector from 1995 to 2006

(subsample 1) (44)

Table 3.9a ……………………The tests of short and long run causality from 1995 to 2006 (45)

Table 3.10 The panel VEC estimation results of the Greek agricultural sector from 2007 to

2013 (subsample 2) (45)

Table 3.10a ………………...The tests of short and long run causality from 2007 to 2013 (46)

Table 4.1 ………………………………………………….The in-sample asymmetric tests (60)

Table 4.2 ……….The tests of Granger Causality, Weak Exogeneity and Joint Significance (61)

Table 4.3 …………………………………………..The descriptive statistics of the prices (62)

Table 4.4 ….The results of the KPSS unit root test for potatoes, tomatoes and cucumbers (63)

Table 4.5 ……………………………………………The results of Johansen’s Trace Test (64)

Table 4.6 ……………………………………….The results of the co-integration analysis (64)

Table 4.7 ………………………….The asymmetric MSVEC estimation results: Potatoes (66)

Table 4.7a …………..The permanent (P) and transitory (T) effects for potatoes in state 1 (67)

Table 4.7b ………………………….The in-sample asymmetric tests for potatoes in state 1 (67)

Table 4.7c The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

potatoes in state 1 (68)

xi

Table 4.7d …………..The permanent (P) and transitory (T) effects for potatoes in state 2 (69)

Table 4.7e ………………………….The in-sample asymmetric tests for potatoes in state 2 (70)

Table 4.7f The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

potato in state 2 (70)

Table 4.8 ………………………...The asymmetric MSVEC estimation results: Tomatoes (71)

Table 4.8a ………….The permanent (P) and transitory (T) effects for tomatoes in state 1 (72)

Table 4.8b ………………………...The in-sample asymmetric tests for tomatoes in state 1 (72)

Table 4.8c The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

tomatoes in state 1 (73)

Table 4.8d ………….The permanent (P) and transitory (T) effects for tomatoes in state 2 (73)

Table 4.8e …………………………The in-sample asymmetric tests for tomatoes in state 2 (74)

Table 4.8f The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

tomatoes in state 2 (74)

Table 4.9 ………………………The asymmetric MSVEC estimation results: Cucumbers (76)

Table 4.9a ………..The permanent (P) and transitory (T) effects for cucumbers in state 1 (77)

Table 4.9b ……………………….The in-sample asymmetric tests for cucumbers in state 1 (77)

Table 4.9c The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

cucumbers in state 1 (77)

Table 4.9d ………..The permanent (P) and transitory (T) effects for cucumbers in state 2 (79)

Table 4.9e ……………………….The in-sample asymmetric tests for cucumbers in state 2 (79)

Table 4.9f The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

cucumbers in state 2 (80)

Table 5.1 …………………...The null hypothesis for the cross and asymmetry effects tests (93)

Table 5.2 The descriptive statistics for the differenced prices of the potatoes, tomatoes and

cucumbers (93)

Table 5.3 ……………..The results of the ERS test for potatoes, tomatoes and cucumbers (95)

Table 5.4 ……………………………………………The results of Johansen’s Trace Test (95)

Table 5.5 ……………………………………….The results of the co-integration analysis (96)

Table 5.6 …………………………...The results of the symmetric BEKK (1.1) for potatoes (96)

Table 5.6a …………………..The variance of the states of the MSVEC model for potatoes (98)

Table 5.7 …………………………The results of the symmetric BEKK (1.1) for tomatoes (100)

Table 5.7a ............................The variance of the states of the MSVEC model for tomatoes (101)

Table 5.8 ……………………….The results of the symmetric BEKK (1.1) for cucumbers (103)

Table 5.8a ……………….The variance of the states of the MSVEC model for cucumbers (104)

xii

Table 5.9 ……………….The results of the asymmetric BEKK (1.1) model for potatoes (106)

Table 5.9a ……..The Cross and Asymmetric effects of the BEKK (1.1) model of potatoes (107)

Table 5.10 ……………..The results of the asymmetric BEKK (1.1) model for tomatoes (107)

Table 5.10a …...The Cross and Asymmetric effects of the BEKK (1.1) model of tomatoes (108)

Table 5.11 …………...The results of the asymmetric BEKK (1.1) model for cucumbers (109)

Table 5.11a …The Cross and Asymmetric effects of the BEKK (1.1) model of cucumbers (109)

xiii

Figures

Figure 2.1 ……………………….The share of agricultural GDP in Greece and the EU-28 (6)

Figure 2.2 ………………………...The agricultural employment in Greece and the EU-28 (7)

Figure 2.3 The index of the real income of agricultural factors per annual work unit for

Greece and EU-28 (base year 2010 = 100) (8)

Figure 2.4 ….The agricultural wages in Greece from 1996 to 2013 (base year 2010 = 100) (8)

Figure 2.5 ………………………………………………………………………….Cereals (10)

Figure 2.5a …………….The per capita apparent consumption of cereals in Greece (in kg) (11)

Figure 2.6 ……………………………………………………………………………..Rice (12)

Figure 2.6a ………………...The per capita apparent consumption of rice in Greece (in kg) (12)

Figure 2.7 ……………………………………………………………...Olive oil and olives (13)

Figure 2.8 …………………………………………………………………………….Wine (17)

Figure 2.8a …………………………The per capita apparent consumption of wine (in kg) (14)

Figure 2.9 ……………………………………………………………Vegetables and fruits (15)

Figure 2.9a …………………The production, the imports and the exports of fresh potatoes (16)

Figure 2.9b ………………The production, the imports and the exports of fresh tomatoes (16)

Figure 2.9c …………….The production, the imports and the exports of fresh cucumbers (17)

Figure 2.10 ........................................................................................................Citrus fruits (18)

Figure 2.11 ………………………………………………………………….Meat products (19)

Figure 2.12 …………………………………………Eggs - Production (in thousands tons) (20)

Figure 2.13 ……………………………………………………………………………Milk (21)

Figure 2.13a …………………………………………..Apparent consumption of milk (kg) (21)

Figure 3.1a The impulse responses of the producer and consumer prices for vegetal products

(39)

Figure 3.1b The impulse responses of the producer and consumer prices for animal products

(40)

Figure 3.1c The impulse responses of the producer and consumer prices for other products

(41)

Figure 3.2a The asymmetric responses of the producer and consumer prices for vegetal

products (47)

Figure 3.2b The asymmetric responses of the producer and consumer prices for meat

products (48)

xiv

Figure 3.2c The asymmetric responses of the producer and consumer prices for other

products (49)

Figure 4.1 ………………………….The prices of fresh potatoes, tomatoes and cucumbers (62)

Figure 4.2 The impulses responses of the producer and the consumer price along with the out-

of-sample asymmetric responses for potatoes in state 1 (69)

Figure 4.3 The impulses responses of the producer and the consumer price along with the out-

of-sample asymmetric responses for tomatoes in state 2 (75)

Figure 4.4 The impulses responses of the producer and the consumer price along with the out-

of-sample asymmetric responses for cucumbers in state 1 (78)

Figure 5.1 ……………..The differenced prices of fresh potatoes, tomatoes and cucumbers (94)

Figure 5.2a The smoothed probabilities of potatoes price mechanism in state 1 with the prices

of the producer and the consumer in first differences (98)

Figure 5.2b ………..The volatility impulse responses of potatoes for 1999:10 and 1999:11 (99)

Figure 5.3a The smoothed probabilities of tomatoes price mechanism in state 1 with the prices

of producer and consumer in first differences (102)

Figure 5.3b ……...The volatility impulse responses of tomatoes for 2007:06 and 2007:07 (102)

Figure 5.4a The smoothed probabilities of cucumbers price mechanism in state 1 with the

prices of producer and consumer in first differences (105)

Figure 5.4b ……The volatility impulse responses of cucumbers for 2004:05 and 2004:06 (105)

xv

Abbreviations

ADF – Augmented Dickey Fuller

AIC – Akaike Information Criterion

ARCH – Autoregressive Conditional Heteroskedastic

BEKK – Baba, Engle, Kraft, Kroner

BFGS – Broyden, Fletcher, Goldfarb, Shanno

CAP – Common Agricultural Policy

CCC – Constant Conditional Correlation

CMO – Common Market Organization

DCC – Dynamic Conditional Correlation

DOLS – Dynamic Least Squares

DVECH – Diagonal VECH

ERS – Elliot, Rothenberg, Stock

EU – European Union

FMLS – Fully Modified Least Squares

GARCH – Generalized Autoregressive Conditional Heteroskedastic

GDP – Gross Domestic Product

HQIC – Hannan-Quinn Information Criterion

IRF – Impulse Response Functions

Kg – kilograms

KPSS – Kwiatkowski, Phillips, Schmidt, Shin

LM – Lagrange Multiplier

MSVEC – Markov Switching Vector Error Correction

OLS – Ordinary Least Squares

PO – Producer Organization

PP – Phillips, Perron

SFP – Single Farm Payment

SIC – Swartz Information Criterion

US – United States (of America)

VAR – Vector Autoregressive

VEC – Vector Error Correction

VECH – Vectorized Conditionally Heteroskedastic

1

1. Introduction

1.1 Studying the prices of the major Greek agricultural products and product categories

The Greek agricultural sector contributes a significant share to the Gross Domestic

Product, even though the share has declined from almost 10% in 1995 to 3% in 2013. The share

is much higher in Greece than the average share in the European Union, which has also declined

from almost 3% in 1995 to 1.5% in 2013. On average, vegetal production accounted for the

70% of the Greek agricultural output while meat products represented the 27% of the output

for the period between 2009 and 2013 (European Commission, 2015). In particular, vegetables

and fruits accounted for the 35% of the total agricultural production with vegetables comprising

about 18% and fruits 16.6% respectively. Thus, vegetables and fruits constitute the most

significant product categories of the Greek agricultural sector. It is evident that the Greek

agriculture plays an important role in the national economy. Therefore, the study of the food

supply chain is of high importance for policy makers as well as for producers and consumers.

The interaction between market fundamentals (supply and demand) for agricultural products is

mirrored upon their prices. The agricultural market stakeholders base their economic decisions

via price monitoring. More specifically, they decide upon the quantity that they are going to

sell or buy according to the prices set by a given market. Moreover, policy makers’ price

monitoring is used for assessing the competition conditions in agricultural markets and taking

relative action when the welfare of the market participants is at stake, especially when they do

not have the means to accommodate such price changes. The competitive structure of the

market determines whether the producer and the consumer conceive prices as exogenous or

endogenous. Under competitive conditions both the producer and the consumer are price takers,

whereas in non-competitive structures the price is usually set by one of the market participants.

The organization of the Greek supply chain of fresh agricultural products was dominated by the

central markets of Athens and Thessaloniki up to 2000. In the following years, retailers’

concentration got raised sharply, resulting in the decentralization of the production distribution.

After 2000, the major retailers, especially the big supermarkets, were the key players in the

supply chain resulting in ever more increased concentration of distribution and demand (Reziti,

2005; Reziti and Panagopoulos, 2008).

Agricultural prices are collected at the producer, manufacturing / wholesalers, and

retailing level by the Hellenic Statistical Authority. The Producer Price Index of Agricultural

Output reflects the prices of agricultural products at the producer level. The Food Consumer

2

Price Index shows the prices at the consumer level, while the Manufacturing Producer Price

Index shows the prices at both the manufacturer and wholesaler level. The weights attributed

to the food products under the Manufacturing Price Index are small thus indicative of the fact

that only a small percentage of agricultural products is being processed. It is noted that the

majority of the agricultural production reaches consumers directly (without having being

processed first), revealing consumers’ preference for fresh agricultural products. Given that

Greece is self-sufficient on the majority of the agricultural products (especially the vegetal

ones), the imported prices do not seem to affect the domestic ones.

The main concern of the agricultural market stakeholders is whether the agricultural

product price formation follows the market fundamentals or whether the non-competitive

market structures allow market levels of the food supply chain to exert influence over the

pricing behavior of the other. By investigating the pricing behavior of the market participants,

researchers attempt to make inferences on whether the market fundamentals of supply and

demand interact freely or if there are market forces that hinder the transmission of price changes

and thus the dissemination of price information in the market.

The price changes of the agricultural products are mainly attributed to the variations of

the market fundamentals, namely supply and demand. Price changes are usually referred to as

price volatility and can be decomposed in observable as well as unobservable. Among the main

factors that affect production and consequently supply are weather conditions or diseases. On

the other hand, demand varies because of changes in income, in prices of substitutes or because

of shifts in tastes. The extent to which given production and consumption shocks translate into

price volatility depends on supply and demand elasticities, which reflect the responsiveness of

producers and consumers to price changes. The supply and demand price elasticities are low in

the short run (during the crop period), especially if the stocks of agricultural products are low

(Wright and Williams [3], Deaton and Laroque [4]). Apart from market fundamentals, shifts in

policy (Christiaensen [5]), input prices, exchange rates as well as speculation result in price

volatility (Interagency Report [1], Gilbert and Morgan [2]).

The agricultural economics literature suggests various sources of asymmetric price

volatility transmission. The most common reason for asymmetric price transmission is the

market power of wholesalers and/or retailers. That is, non-competitive market structures allow

wholesalers and retailers to transmit price changes that squeeze their margin more quickly

and/or more fully than price changes that stretch their margin (Meyer and von Cramon-

Taubadel, 2004). The second most frequently cited reason for price asymmetry in the literature

accounts for the transaction costs. Transaction costs incorporate adjustment costs, menu costs,

3

inventory management or any other costs incurred when an economic agent participates in a

market. Therefore, when firms have to change the quantities or the prices of their inputs and

outputs, asymmetric price transmission may result. Another reason which can cause

asymmetries to transaction costs is the asymmetric flow of information between markets as

shown by Abdulai (2000). Finally, political intervention is also proposed as a cause of

asymmetries (Kinnucan and Forker, 1987). The lack of contractual farming (Apergis and

Rezitis, 2003) results in asymmetries in the price mechanism, as well.

As a member of the European Union (EU), Greek agricultural markets are regulated by

the Common Agricultural Policy (CAP) of the EU since 1981. CAP is constantly changing so

as to encompass the changing economic environment as well as to serve the European food

producers and consumers. In 1992, a major restructuring of the CAP philosophy was enacted.

In particular, after 1990 and for the next twenty years this new philosophy was gradually

implemented by schemes like Agenda 2000 and its mid-term review, which in 2003 was

encoded as the Fischler Reforms. The reforms of the CAP since 1992 and onwards exposed

domestic EU prices to the signals of the world markets and decoupled the direct payments to

producers from the production of a specific product.

The observed volatility of the agricultural prices is studied with univariate and

multivariate Auto-Regressive models and its Co-integrated variants, while the unobserved

volatility of the prices with Autoregressive Conditional Heteroskedastic models and its

Generalized versions in univariate as well as multivariate forms. The former model is the mean

of the joint distribution of the prices while the latter of the variance. Thus, price volatility

transmission models can take several specifications depending on their intended use. These

specifications account for asymmetric effects in the transmission of observed and unobserved

price volatility as well as for regime switching that accounts for the different states of the

economic environment. In comparison to the variance model, the mean model of prices had

been used far more frequently in the agricultural economics literature.

1.2 Objectives

The main goal of the present study is to examine the price transmission mechanisms

between the producer and the consumer with respect to the major agricultural products and

product categories in Greece. In order to accomplish that, the observed as well as the

unobserved price volatility is going to be modelled. The focus of this analysis will be on

vegetables accounting for the 10% of the Producer and the Consumer price indices (fresh

4

potatoes, tomatoes and cucumbers). Despite the fact that these vegetables have high added value

and do not require heavy investments on behalf of the producer, still the agricultural economics

literature on this subject is limited.

For the assessment of the price mechanisms of the Greek food sector, and more

specifically that of fresh potatoes, tomatoes and cucumbers, three essays are employed.

The first essay attempts to investigate the price transmission mechanism between the

producer and the consumer of the Greek food market while assessing the effects of the

decoupling principle of the Common Agricultural Policy of the European Union. The empirical

analysis uses the panel Vector Error Correction model for the depiction of the price mechanism.

The dynamics of the price mechanism are evaluated both in the short and long run with the

usage of causality tests as well as with impulse responses. The results of the causality tests show

that the producer does not respond to long-run price changes while the ones of the impulse

responses reveal that the own effects of a shock are quickly absorbed, in contrast to the spillover

effects of the shock. It is noted that the later take more time to decay. Finally, the

implementation of decoupling seems to have benefited the consumer more in mitigating his

responses to price shocks rather than the producer. Thus, the first essay assesses the observed

volatility of the prices of the Greek agricultural sector with a mean model that uses panel data.

The panel data set has the advantage over time series data that it takes into consideration each

product and product category individually instead of using a weighted average of the prices for

the whole agricultural sector.

The second essay investigates the price transmission mechanisms between the producer

and the consumer for the fresh potatoes, tomatoes and cucumbers. The three products are

selected on the base that they comprise a significant part of the vegetables production as well

as of the total agricultural output while they do not need big investments for their production.

The empirical analysis makes use of the Markov Switching Vector Error Correction model

under the assumption that the producer and the consumer respond differently to price increases

and decreases. The asymmetric effects are investigated with in-sample as well as out-of-sample

measures for each state. The response of the producer and the consumer to long run deviations

from the equilibrium and to lagged changes of their prices is investigated along with the nature

of the price shocks for each state. The empirical results show that potatoes and cucumbers give

rise to similar price mechanisms even though the underlying characteristics of these two

markets are different. Moreover, the empirical analysis shows that tomatoes and cucumbers

result in different price relationships between the producer and the consumer despite the similar

market characteristics. Furthermore, the observed price volatility of the three products is

5

modelled under the assumption of switching regimes and asymmetric effects in price increases

and decreases. The results reveal the complexity of the price mechanism and the diversified

response of each product to price shocks.

The third essay models the transmission of the volatility between the producer and

consumer prices of fresh potatoes, tomatoes and cucumbers in Greece. The transmission

mechanism of the volatility of the prices is modelled by the BEKK model while possible

asymmetries are taken into consideration. The asymmetric effects are evaluated by asymmetric

BEKK models as well as volatility impulse responses. The implementation of the volatility

impulse responses is based on a new approach. The results reveal that in the regulated markets

of tomatoes and cucumbers, the producer is less vulnerable to volatility shocks transmitted from

the consumer whereas in the non-regulated potato market, the producer faces significant

spillover effects from the consumer. The last essay models the unobserved price volatility of

the three products. The empirical results of the second and third essays indicate the different

effect that shocks have for the observed and unobserved price volatility for the same agricultural

products.

1.3 Organization of the Study

The rest of the study is organized as follows. Chapter 2 offers a brief description of the

Greek agricultural sector by describing the main characteristics of agricultural production, the

consumption of agricultural products, the distribution network and the CAP policies as they are

implemented in the Greek agricultural sector. Chapter 3 presents a model of the price

transmission mechanism between the producer and the consumer of the Greek food market

under the decoupling principle of the Common Agricultural Policy of the European Union.

Chapter 4 investigates the price transmission mechanisms between the producer and the

consumer for fresh potatoes, tomatoes and cucumbers while chapter 5 models the transmission

of the volatility between the producer and consumer prices for the aforementioned products.

Finally, chapter 6 concludes.

6

2. Market description

2.1 The Greek agricultural sector

The domestic agricultural sector constitutes one of the major economic activities in the

Greek economy. The share of the agricultural GDP in the total output is quite high though

decreasing in the last twenty years. The Greek agricultural GDP is one of the highest in the

European Union. More specifically, on average is double in size of the average agricultural

GDP of the 28 Member States of the European Union. Figure 2.1 shows the two trends from

1995 to 2013.

Figure 2.1 The share of agricultural GDP in Greece and the EU-28

Agriculture plays a vital role in the sustainability of rural areas since it utilizes the total

of cultivable land and employs the one third of the rural labor force. Fruits and vegetables, olive

oil and cereals constitute the three most important Greek agricultural product categories in

terms of employment, land utilization, production volume and value. Table 2.1 presents the

average share of each product category in the total production from 2009 to 2013.

Table 2.1 The shares of Greek agricultural products in total output 2009-2013

Vegetal products % Meat products % Other products %

Vegetables 18 Milk 11.2 Other 3.5

7

Fruits 16.6 Sheep and goats 8.3

Industrial corps 10.2 Cattle 2.5

Olive oil 9.3 Pigs 2.3

Cereals 7.5 Poultry 1.4

Forage plants 5.2 Eggs 1.2

Potatoes 2.9

Between 2000 and 2003, rural areas accounted for the 66% of the total land available, forests

accounted for the 28% while the rest 6% accounted for urban areas. The 38% of rural land was

utilized for cultivation while 5% for pasture. Furthermore, another 6% was left fallow. One

third of the arable land was irrigated whilst the most important cultivations in terms of land

utilization were cereals, tree plantations and industrial plants.

The agricultural labor force is one of the largest in the European Union. As is evident

from figure 2.2, the Greek agricultural labor force has declined from almost 18% of the total

labor force in 1995 to about 13% in 2013, although from 2004 the percentage remains stable at

about 12%. On the other hand, the agricultural labor force in the EU of the 28 Member States

is much lower. On average it varies between 7 and 6% of the total labor force. The percentage

of the EU-28 is also decreasing though in a more moderate trend.

Figure 2.2 The agricultural employment in Greece and the EU-28

The aforementioned trends of a declining share of agricultural GDP and percentage of

employment is verified also by a declining index of the real income of agricultural factors per

8

annual work unit. As it is shown in figure 2.3, the index of the real income was reduced from

130% in 1995 to less than 100% in 2006. Thus, the return of agricultural factors is in decline

but stabilized from 2007 to 2013 around 100%. In contrast, the average real income for the EU-

28 is increasing from 110% in 2005 to 140% in 2013.

Figure 2.3 The index of the real income of agricultural factors per annual work unit for

Greece and EU-28 (base year 2010 = 100)

On the other hand, as it is obvious from figure 2.4, the wages in the Greek agricultural sector

were increasing from 1996 and onwards until 2009. However, after this point and up to 2013

the agricultural wages were in decline.

Figure 2.4 The agricultural wages in Greece from 1996 to 2013 (base year 2010 = 100)

9

In order to shed more light to the income of the Greek agricultural sector, table 2.2

presents its formation from 2000 to 2005 per product category. The table reveals part of the

rationale behind this study, since it gives evidence on the necessity of investigating the

vegetables market. In particular, it is observed that vegetables provided the highest net added

value among the rest of the product categories presented, while utilizing the smallest arable

area. Furthermore, the potential of the vegetables sector is revealed from the fact that the sector

achieved high performance despite the lack of governmental financial aid.

Table 2.2 The agricultural income per product category 2000 – 2005

Labor (hours) Area utilised

(acres)

Total gross

production

value (€)

Net added

value (€)

Subsidies (%

of gross

yield)

Cereals, industrial plants, rice

2000-2001 2464 89.05 14448.5 10743.5 44.69

2004-2005 2542.4 93.4 17420.5 11721.5 31.52

Vegetables

2000-2001 4435.2 23.95 33075.5 18996.5 2.93

2004-2005 4872 23.85 48697.5 28772.5 3.15

Vineyards

2000-2001 2688 37.45 11979.5 10134.5 35.43

2004-2005 3024 39.95 18883.5 16549 24.21

10

Fruits, nuts

2000-2001 2620.8 40.7 11769.5 9461.5 20

2004-2005 2676.8 45.8 13891 11883.5 23.93

Milk

2000-2001 3102.4 61.6 42881.5 16124.5 14.22

2004-2005 3640 88.05 52628 20773.5 27.69

Animal production

2000-2001 3550.4 53.35 23980.5 15608.5 29.85

2004-2005 3651.2 65.4 27158 20357.5 35.74

Source: Kaditi and Nitsi (2010)

2.2 The main agricultural products and product categories

The subsequent sub-sections present basic information on the main product categories

of the Greek agricultural sector along with a brief discussion on the implementation of the

Common Market Organizations. More specifically, the data on area cultivated, quantities

produced and apparent consumption are reported for cereals, olive oil, wine, vegetables and

fruits as well as meat products. The presentation is based on data available in Kaditi and Nitsi

(2010).

2.2.1 Cereals

Cereals are an important cultivation for the Greek agricultural sector since they

significantly contribute to the producers’ income, secure the self-sufficiency of the country and

cover large areas of arable land. Below, figure 2.5 presents the area cultivated as well as the

produced quantities from 1990 to 2006.

Figure 2.5 Cereals

Area Cultivated (in thousand acres) Production (in thousand tons)

11

As it is obvious from the figure, the area cultivated as well as the produced quantity were in

decline for the period under examination. In particular, in the beginning of the period, the area

cultivated was 14.7 million acres but the observed decline resulted in 11.6 million acres of land

to be utilized in the end of the period. Despite the decline of cultivated land, the 11.6 million

acres still accounted for the 30% of the total area cultivated in Greece. In 2006, 53.69% of the

total utilized land was occupied by hard wheat, 20% by corn, 12.51% by soft wheat and 8.76%

by barley. The area covered by oats and rye was minimum. The produced quantities of cereals

amounted to 5.2 million tons in 1990 and to 4.5 million tons in 2006 of which 52.26% was corn,

31.17% hard wheat, 8.37% soft wheat and 5.55% barley. The remaining 2.65% accounts for

the production of rye and oats. In contrast to produced quantities, the per capita apparent

consumption of cereals is increasing for the period from 1992 to 2006 as is depicted in figure

2.5a.

Figure 2.5a The per capita apparent consumption of cereals in Greece (in kg)

In the European Union, the market of cereals is regulated by the Common Market Organization

framework. The aims of the CMO, up to 2003, were the support of prices and the protection

from international competition. However, after the reforms of 2003 the direct payments were

12

sharply reduced. Up to this point, the policy framework was characterized by interventions

buying and coupled direct payments.

2.2.1.1 Rice

The area cultivated and the produced quantity of rice are presented in figure 2.6. As it

is apparent from the figure, the area utilized for the cultivation of rice as well as the produced

quantities were increasing between 1990 and 2000, however for the rest of the period they

declined.

Figure 2.6 Rice

Area Cultivated (in thousand acres) Production (in thousand tons)

Next, figure 2.6a reveals that for the same period the per capita apparent consumption followed

a similar pattern. For the first years, the trend was increasing but for the rest of the period was

decreasing.

Figure 2.6a The per capita apparent consumption of rice in Greece (in kg)

2.2.2 Olive oil and olives

13

From 1995 to 2006 the production of olive oil and olives showed an increasing trend.

More specifically, up to 2000, the increase was more intense for olive oil and olives but in the

next years, oil’s production stabilized while the production of olives retained the increasing

trend. Figure 2.7 presents the aforementioned observations.

Figure 2.7 Olive oil and olives

Production of oil (in thousands) Production of olives (in thousands)

The consumption of olive oil in Greece amounted to 265 thousand tones in 2006. Olive oil is

mainly preferred by consumers in the European Union and in a lesser extent in the United

States. In particular, Spain and Italy are the biggest consumer countries of olive oil and they are

followed by Greece. 2006 was a year of low consumption for olive oil in EU and Greece.

Though, in the next two years consumption retained its original levels. The CMO of oil and

olives followed the regime change of the CAP in 2004 that imposed the decoupling of the

magnitude of production from direct payments. Moreover, the CMO rules did not allow

member states to diversify prices of the other vegetal oils so that the consumption of olive oil

to be increased however emphasis was placed on the promotion of quality olive oil.

2.2.3 Wine

As it is evident from figure 2.8, the areas used for vineyards were in decline up to 2003.

However, after this point, the areas utilized regained their original size. Regarding produced

quantities, even though, they were steady during the decline of cultivated areas due to improved

yields, the subsequent increase in area utilization did not improve production accordingly.

Figure 2.8 Wine

Area (in thousand acres) Production (in thousand hundred-litres)

14

Despite the decline in produced quantities, apparent consumption follows an increasing trend

since 2000 as it is obvious from figure 2.8a.

Figure 2.8a The per capita apparent consumption of wine (in kg)

The main goal of the CMO for wine was to balance supply and demand along with the

improvement of the competitiveness of the sector. These goals were attempted to be met by the

means of producers’ organizations and abolishment of the intervention buying. Furthermore,

producers were not allowed to plant new areas with vineyards unless certain prerequisites were

met while a premium was paid to producers that abandoned viniculture. Following the 2003

reforms of CAP, viniculture was subjected to the decoupling principle.

2.2.4 Vegetables and fruits

The total area for the cultivation of fruits and vegetables amounted to 1.1 million acres

for 2006 accounting for 5.2% of the total cultivable land. Out of the 1.1 million acres, the 500

thousand were occupied by greenhouses while 50% of this area was in Crete. Regarding the

open-air cultivations, tomatoes and potatoes accounted for the largest proportion. In aggregate

terms, as figure 2.9 shows, the utilized land for fruits and vegetables is in decline. On the other

hand, the per capita apparent consumption of fruits shows an increasing trend. What is more,

15

the production of vegetables takes place allover Greece, thus allowing production throughout

the year.

Figure 2.9 Vegetables and fruits

Fruits and vegetables cultivated

(in thousand acres)

Per capita apparent consumption of fruits

(in kg)

The CMO for fruits and vegetables was fundamentally reformed in 2007. The reform was

deepened in 2008 with main goal the enhancement of the competitiveness of the sector by

making it more market oriented, by countervailing the decrease of producers’ income in times

of volatile prices, by increasing consumption, by safe keeping the environment as well as by

encouraging producers to join Producers’ Organizations (POs). In particular, in times of volatile

markets, producers are allowed to withdraw a portion of their production as long as this action

does not intervene in the market functionality. Moreover, the decoupling principle was applied

to the fruits and vegetable sector though more progressively than the other sectors. Lastly, it is

worth noting that the sector of vegetables and fruits is the only sector that subsidies were not

reduced firmly while potatoes are the only vegetal product that is not regulated from the CMO

of fruits and vegetables.

2.2.4.1 Selected vegetables production, imports and exports

Since the focus of this study is the fresh potatoes, tomatoes and cucumbers; their market

fundamentals are presented in more detail that the rest of the main product categories of the

Greek agricultural sector. More specifically, figures 2.9a, 2.9b and 2.9c show the produced

quantities as well as the exports and the imports of the three products under examination. The

exports and the imports are depicted as percentages of the total production of the respective

product. Moreover, the regulative framework that governs the products is discussed further. It

16

is worth repeating that the three products are produced throughout Greece. The first product

discussed is the fresh potatoes.

Figure 2.9a The production, the imports and the exports of fresh potatoes

As figure 2.9a reveals, the production of fresh potatoes shows a declining trend in the period

under examination despite the fact that the range of the produced quantity stays between 1

million and 9 hundred thousand tons per year until 2009. After 2009, the decline becomes

steeper while the recovery of the production starts in 2013. Regarding imported quantities, an

increase is observed from the 10% of total production in 1991 to over 20% in 2012. The

comparison between the produced and the imported quantities reveals that the market share of

imports increases at the expense of domestic production. On the other hand, exports show a

stable evolvement by comprising on average the 3% of the domestic production.

Figure 2.9b The production, the imports and the exports of fresh tomatoes

In the tomatoes market, as figure 2.9b shows, the production follows a relatively stable trend

with some signs of decline till 2010. However, during the next two years production collapses

to the half because of a decease that was difficult to be encountered. The imported fresh

tomatoes reveal an increasing trend, especially from 2000 and afterwards, when the percentage

17

climbs to 5%. Exports, on the other hand, are more stable up to 2009 while in the next two years

they increase rapidly in absolute value as well as a percentage of domestic production.

Figure 2.9c The production, the imports and the exports of fresh cucumbers

Figure 2.9c shows that cucumbers’ production is relatively stable, even though there are signs

of a gradual decline till 2009, as was the case with tomatoes. In particular, production fell by

almost 20% in 2010 before recovering to its former levels in the next two years. In contrast to

potatoes and tomatoes, the imports are very low (about 1% of total production) whereas the

exports comprise on average the 14% of the total quantities produced.

Fresh potatoes, tomatoes and cucumbers are regulated by the Common Agricultural

Policy since 1981 when Greece became a member state of the European Union. However, as it

is already mentioned, fresh potatoes are not regulated by the Common Market Organization

(CMO) of fruits and vegetables since the enactment of CAP in 1962. In 1992 and 1995,

proposals were made for a CMO but to no avail, they were rejected. Thus, until 2004, producers

of fresh potatoes did not receive any financial support from CAP funds. The first time that

potato producers of EU-12 became eligible for supporting payments was when the reforms of

2003 were implemented, giving the opportunity to potato producers to get financial support

according to the decoupling principle. The decoupling principle states that direct payments from

the CAP budget will not be linked to the production of a specific product. The last reform of

the CAP in 2008 ended any special statuses for potatoes thus allowing producers to get direct

financial support from the CAP budget or the member states. On the contrary, the tomatoes and

the cucumbers are regulated by the CMO for fruits and vegetables since 1962. From its early

state the CMO of vegetables and fruits was market oriented while in an attempt to stabilize

producers’ income, institutional prices were fixed. However, the high costs of this policy led to

the reform of 1996 which consolidated the market orientation of the CMO by encouraging

producers to join Producer Organizations (POs) so as to strengthen their position in the market

18

and become able to satisfy the concentrated demand from wholesalers and retailers. The 2000s

CAP reform aimed at enhancing the goals of 1996 by strengthening POs and improving export

refunds management. Though, neither of these reforms was fully successful since producers

were faced with the ever increasing concentration of distribution and demand, strong

competition from third countries, market crises which were affecting their income while the

consumption in the EU was stagnating. Thus, in 2007 a new reform was enacted. The main goal

of this reform was to improve attractiveness of POs because many producers claimed that POs

were empty shells which managed only the direct payments of CAP. Moreover, new hedging

tools were made available along with the decoupled payments so as to improve the

confrontation of market crisis that affected producer’s income. Finally, the marketing standards

of the CMO were simplified while the export refunds were abolished.

Regarding Greece until 2000, the central markets of Athens and Thessaloniki dominated

the supply chain, however in the following years the retailers’ sector concentration raised

sharply resulting in the decentralization of production’s distribution. Thus after 2000, the

retailers and especially the supermarkets played a major role in the supply chain.

2.2.4.2 Citrus fruits

Citrus fruits are the second most important tree plantation in Greece after peaches in

terms of cultivated areas, produced quantities and value. As it is evident from figure 2.10, the

areas planted followed an increasing trend up to 2000 when growth was stabilized. In the same

fashion, per capita apparent consumption followed an increasing path for the most part of 90s

and the beginning of 00s but was stabilized in the next few years.

Figure 2.10 Citrus fruits

Areas planted (in thousand acres) Per capita apparent consumption (in kg)

19

In alignment with the other fruits, citrus are regulated by the CMO for fruits and vegetables

thus decoupling was implemented accordingly.

2.2.5 Meat products

The meat production in Greece accounted for 23.6% of the total agricultural production

in 2005. The percentage was stable for the period between 1995 and 2005. Since, meat

production accounts for the one fourth of domestic agricultural production, its contribution to

the economy is significant. In particular, sheep and goats account for 58.7% of the meat

production, bovine for 19.25%, poultry for 9.9%, swine for 7.8% and the rest of meat products

for 4.45%. Figure 2.11 presents meat production and apparent consumption in Greece between

1990 and 2006. During the presented period, the total meat production is decreasing, though

the apparent consumption is increasing. As it is obvious, the consumption doubles the produced

quantities in volume, revealing the disproportional small contribution of domestic production

to fulfilling domestic consumption.

Figure 2.11 Meat products

Production (in thousand tons) Apparent consumption (in thousand tons)

Up to 2008 each meat type (beef, pork, poultry, sheep and goats) had its own CMO. However,

after this point all the agricultural products were incorporated to a single CMO. The meat

products were eligible for direct payments for many years however due to the reforms of 2003

the subsidies of the meat sector were sharply decreased. The countermeasures available for each

meat type in cases of abnormal conditions vary. In the case of decreasing beef prices, producers

are allowed to stock a percentage of their production. For sheep and goats, each member state

is granted certain quotas. For pork, the private stocking is also available as in the case of beef,

while in the case of a disease a member state has the right to request appropriate market

20

intervention. However, for poultry, no intervention buying or private stocking is allowed.

During the last two decades the meat sector suffered major shocks due to health issues that

affected the animals. The Bovine Spongiform Encephalopathy, the Foot and Mouth Disease as

well as the bird flu are always a potential treat for meat producers. For instance, the frequent

occurrences of bird flu resulted in volatile consumption patterns. These shocks had as a result

the change of consumer tastes regarding their preference for each meat type though they are

considered as temporary. Finally, it is worth mentioning that since the middle of 90s, the poultry

sector is the most dynamic sector of the Greek meat industry, since it has overpassed beef in

growth. The regulative framework of CAP is considered to have enhanced the growth of the

sector.

2.2.5.1 Eggs

The production of eggs is in sufficient levels for satisfying domestic consumption

despite the decreasing trend that is observed between 1990 and 2006, as it is obvious from

figure 2.12. More specifically, the yearly consumption varied between 111 and 124 thousand

tons while the per capita consumption was between 10.6 and 11.7 kg.

Figure 2.12 Eggs - Production (in thousands tons)

The CMO of eggs aims at improving the quality of the product, the trading conditions as well

as the knowledge of the market fundamentals.

2.2.5.2 Milk

21

The produced quantities of milk from cows, sheep and goats in Greece are presented in

figure 2.13. As it is obvious from the figure, the quantities are increasing for cow and sheep

milk but are decreasing for goat milk after 2000.

Figure 2.13 Milk

Cow milk (in thousand tons) Sheep milk (in thousand tons)

Goat milk (in thousand tons)

Similarly, the apparent consumption presents an increasing trend for the period from 1992 to

2006, as it is shown to figure 2.13a.

Figure 2.13a Apparent consumption of milk (kg)

22

The CMO for milk and dairy products regulates the milk market. The CMO supports

intervention buying for butter and powder milk. Furthermore, for certain cheese types private

stocking is allowed which on occasion can be offered via the market. In order to balance supply

and demand, member states quotas have been established. These measures have helped in a

balanced powder milk market though in the butter market they have not been so efficient. The

quotas of Greek producers have been increased while the abandoning of land from low

subsidized products to higher ones, is expected to enhance the production of milk and dairy

products. Furthermore, the expected increased consumption will absorb the additional

production.

2.3 The Common Agricultural Policy

Greece is part of the common agricultural market as a member state of the European

Union (EU) since 1981. The common agricultural market is regulated by the Common

Agricultural Policy (CAP) since its establishment in 1961. Since its establishment, CAP is

under constant reforming so as to encompass the changing economic environment and to serve

the producers and the consumers of the European Union. The reforming process has been

accelerated and intensified during the last 20 years. Since 1992 and onwards the reforms that

are under process aimed at disentangling the direct aid to producers from production while

exposing European prices to the signals of world markets. These reforms were implemented by

the 2000 Agenda and 2003 mid-term review which is known as Fishler Reforms. The reforms

are based on two pillars: direct payments (pillar 1) which are constrained to certain limits and

cross compliance along with rural development (pillar 2) which are not subjected to any budget

constraints.

The first major reform of the CAP took place in 1992. This reform aimed at eliminating

market interventions since they were considered as a source of inefficiency and the emphasis

was placed on direct payments to farmers. Moreover, the member states had to protect the

environment and impose strict relevant legislation. The 1992 reform was adopted under the

pressure for complying with the international trade rules. Hence, since 1993, an increased

proportion of expenditure has been dedicated to direct subsidies to compensate farmers for the

absence of market-price supporting mechanisms. Also, it involved cutting budgetary

expenditures on export subsidies. However, no penalty was imposed for the producers that

violated the quotas set.

23

The Fishler reforms that were implemented with the Agenda 2000 and 2003 mid-term

review aimed at dealing with the cost of the direct payments which would almost double after

the EU enlargement from 15 to 25 member states in 2004. Up to this point, CAP expenditures

accounted for about 50% of the EU budget. Therefore, the reforms ought to control budget

expenditures through financial discipline. Moreover, the new legislation had as an ultimate goal

to make EU farmers more competitive and market-oriented, and at the same time to ensure that

environmentally sound practices are used in farming. The new rules let the EU to participate in

WTO negotiations with well-prepared producers. More specifically, the decoupling principle

was implemented with the Single Farm Payment (SFP). In this way, producers would be more

able to respond to market-price signals since they would be able to concentrate on production,

without being obliged to produce a specific commodity in order to receive direct payments.

Member states could choose among the historical model according to which payments are based

on individual historical reference amounts per farmer, the regional model according to which

flat rate payments are based on amounts received by farmers in a region in the reference period

and the hybrid model which was a mix of the two approaches.

In 2004, Greece agreed in adopting the decoupling principle based on the historical

model by 1st January of 2006. The historical model was adopted on the grounds that it would

not result to intense income redistribution in the agricultural sector while ensuring a fair

standard of living for producers. Apart from arable corps, decoupling was adopted also for

pulses, cereals, protein plants, rice, fodder, cotton, olive oil and meat products. Sugar beet and

bananas production payments were decoupled in 2007, vegetables in 2008 and vineyards in

2009. The producers are eligible for direct payments: i) if they were granted subsidization for

the historical period from 2000 to 2002, ii) if they have inherited a farm and iii) if they have

rights from the National Reserve or from buying and selling. The beneficiaries of rights should

own or rent the land for which they claim payment. In Greece, the decoupling principle was

fully adopted in fear of abandonment of rural areas. Moreover, coupled payments are also used

for the producers of specific corps. Hard wheat, rice, nuts with shell, energy plants, milk, cotton,

fodder and seeds are these products.

A producer in order to be eligible for decoupled payments should abide by the rules of

cross-compliance. These rules define the context of land management, proper farming and

environment protection. In particular, the following elements should be taken into consideration

by producers: the environment, public health, the health of animals and corps, publication of

animal deceases, the prosperity of animals and the sound farming and environmental practices.

Furthermore, the rural development is financed by a progressive reduction of direct decoupled

24

payments. Though, if a producer receives less than €5000 in direct payments, the amount

deducted is refunded. Moreover, the reduction of direct decoupled payments also finances the

improvement of quality and trading of agricultural products along with the protection of the

environment. Finally, the direct payments for the Greek producers amounted to the 79% of the

total Greek agricultural budget in 2007. The 62% of the direct payments were distributed among

four regions: the East Macedonia and Thrace (17.69%), Central Macedonia (16.53%), Thessaly

(16.9%) and Crete (10.72%).

The Health Check of 2008 aimed at simplifying further and increasing the effectiveness

of the 2003 reform. The SFP in the aforementioned form was eligible up to 2013 and was

extended up to 2015. The proposition of the Commission regarding the future of the CAP is not

a full scale liberalization without any kind of protection for producers. Thus, the direct financial

aid to producers is expected to continue with a constraint budget while payments will not be

based on a historic basis.

2.3.1 The Common Market Organization

Under the framework of CAP, the production and the marketing of agricultural products

is implemented by the Common Market Organizations (CMOs). CMOs function as instruments

for the fulfillment of the goals of the CAP. The main functions of a CMO is to stabilize markets,

offer an acceptable standard of living for producers and to enhance the productivity of the

agricultural sector. In particular, the CMOs allow the designation of common prices for all

European agricultural markets, the subsidization of producers and the establishment of

mechanisms for the control of production and trade with third countries while the creation of

Producers Organizations is encouraged.

The markets of cereals, tobacco, cotton, sugar, olive oil and olives, vegetables and fruits,

wine, bananas, pork – beef – sheep and goat meat, dairy products, eggs and poultry, fodder and

seeds are regulated by CMOs. As a consequence, CMOs account for the 90% of the total

European agricultural production. A CMO sets three different prices for an agricultural product:

i) The target price which is the price that the transactions are expected to take place. Even

though, the target price is an estimate, it usually represents the actual prices that are observed

in the European agricultural markets.

ii) The threshold price is the minimum price in which the imported products can be traded. The

threshold price is bigger than the intervention price so that European stakeholders to prefer the

domestic production than the one from third countries.

25

iii) The intervention price is the price at which an organization appointed by the member states

buys the produced quantities and stores them.

The control of the agricultural production for the restraining of surpluses is achieved by

quotas systems, nationally guaranteed produced quantities as well as the obligation for a certain

percentage of cultivable land to be left fallow.

2.4 Studies of the Greek vegetables sector

There are few studies of the Greek vegetable sector, even though vegetables correspond

to the 18% of the total agricultural production and offer high added value to producers without

requiring high investments. Reziti (2005) studied the price transmission mechanism between

the producer, the wholesaler and the consumer for potatoes, tomatoes, oranges and milk in

Greece. A Vector Error Correction model was used on monthly prices from 1995 to 2003. The

results revealed that a co-integrating relationship existed between the prices of the producer and

the consumer for all products. The causality tests revealed that the producer responded to lagged

price changes of the consumer for potatoes and tomatoes, however in the case of oranges and

milk the consumer responded to lagged price changes of the producer. Moreover, the empirical

analysis showed that asymmetries were present between the producer and the consumer while

symmetries were present between the producer and the wholesaler. Moreover, Reziti and

Panagopoulos (2008) investigated the price transmission mechanism between the producer and

the consumer for fruits, vegetables and the total of the Greek agricultural sector. The methods

used for the empirical analysis were the Vector Error Correction model and the General to

Specific approach for monthly data from 1995 to 2004. The results indicated that in the case of

vegetables the consumer does not respond to deviations from the long run equilibrium while

the opposite holds for the other two categories. Furthermore, asymmetric effects are present in

the price mechanisms of the whole sector and vegetables but not for the fruits sector. On the

other hand, Apergis and Rezitis (2003) examine the agricultural sector in aggregate terms but

with the focus on the transmission of the price volatility among inputs, producers and

consumers. The model used is the Generalized Autoregressive Conditional Hetreskedastic for

the period between 1985 and 1999. The results revealed that both the inputs and the consumer

exert positive and significant spillover effects to the producer. Moreover, the own effects of a

volatility shocks to the producer are positive and significant. Thus, producer prices are proved

to be more volatile than the ones of inputs and consumer.

26

3. Price transmission along the Greek food supply chain in a dynamic panel framework:

Empirical evidence from the implementation of decoupling

This chapter investigates the price transmission mechanism between the producer and

the consumer of the Greek food market while taking into consideration the decoupling policy

scheme of the Common Agricultural Policy of the European Union. The empirical analysis uses

the panel Vector Error Correction model for the empirical investigation of the price mechanism.

The dynamics of the price mechanism are evaluated in the short and long run with causality

tests and impulse responses. The results show that the producer does not respond to long-run

deviations from the equilibrium. Moreover, a shock in the producer or the consumer prices

results in own disequilibrium effects that are quickly decayed and disequilibrium spillover

effects from one price level to the other that take time to decay. Finally, the implementation of

the decoupling scheme seems to have benefited the consumer more rather than the producer in

mitigating his responses to own and cross price shocks. The chapter is based on the working

paper Rezitis and Pachis (2015a).

3.1 Introduction

The Common Agricultural Policy (CAP) of the European Union (EU) has regulated the

agricultural markets of Europe since 1960s. Since then, CAP is constantly changing so as to

encompass the changing economic environment as well as to serve the European food producers

and consumers. In 1992, a major restructuring of the CAP conduct was enacted. In particular,

after 1990 and for the next twenty years new reforms were gradually implemented by schemes

like Agenda 2000 and its mid-term review which in 2003 was encoded as the Fischler Reforms

(Ihle et al, 2012).

According to the European Commission (2015), a key aspect of these reforms was the

adoption of the decoupling principle. The decoupling principle dictated that the direct payments

to producers from the CAP budget should not be linked to the production of a specific product.

The direct payments constitute a safety net for farmers with the aim to stabilize their income

stemming from market sales. In this way, producers have the opportunity to maximize their

profits by responding to market signals without the adverse effects of prices volatility that

would lead them to suboptimal decisions. What is more, member states were allowed to

implement the new policy regime according to their own discretionary agricultural policies, for

the first time. The reforms of the CAP since 1992 and onwards, apart from the adoption of the

27

decoupling, also aimed at exposing domestic EU prices to the price signals of the world

markets. Thus, the extent to which global prices are transmitted to European markets would

depend on how strongly integrated the latter are with the former. However, the price dynamics

of agricultural products are also affected by their characteristics. Therefore, the price

transmission mechanisms of European agricultural products are determined by international

markets as well as by the domestic markets conditions.

Since 2003 when the decoupling principle was for the first time adopted by the member

states, researchers tried to assess the impact of the new policy regime for the producers of

agricultural products. According to the Centre for Rural Economic Research (2003), ADAS

(2002) and Moss et al. (2002) the decoupling could result in reduced utilization of the economic

resources in agriculture. Therefore, production would be reduced and prices would be

increased. However, due to the expansion of the more efficient producers, agriculture would be

benefited in aggregate terms. These studies based their conclusions on future projections of

existing trends in agriculture. Similarly, more recent studies such as Viaggi et al. (2011) verified

that producers have not changed their behavior after decoupling. In similar fashion, Lobley and

Butler (2010) reported that the existing trends in agriculture are reinforced. On the other hand,

Ihle et al. (2012) showed that decoupling reduced the prices of calves while EU spatial price

relationships were impacted by the heterogeneity in the implementation of the policy. A similar

study by Prehn et al. (2015) shows that the heterogeneity in the implementation of the

decoupling led to artificial trading that infringed the concept of the single market of agricultural

products. However, decoupling was not fully implemented until 2007, thus the majority of the

empirical work so far is based on relatively few observations for studying its implications. This

chapter aims at contributing in this gap of the literature by investigating the price transmission

mechanism of the Greek food sector between the producer and the consumer under the impact

of the decoupling scheme for a larger data set.

The Greek food market participates in the common EU agricultural market since 1981.

However, the focus of this chapter would be the period from 1995 to 2013 which is the time

interval that the major reforms of the CAP took place. The data set used, includes seventeen

products and product categories that comprise 60% of the Greek Producer Price Index of

Agricultural Output and Food Consumer Price Index indicating the importance of the selected

sample for producers and consumers. Moreover, the selected products and product categories

account for just 7% of the domestic Manufacturing Producer Price Index revealing the

preference of Greek consumers for fresh agricultural products. The selected products and

product categories are classified in three categories. The vegetal products which consist of

28

potatoes, cucumbers, marrows, onions, tomatoes and lemons. The animal products that include

milk, beef, pork, sheep, goat, poultry and eggs. The other products that include cereals, wine,

olive oil and olives. Since 2008, the aforementioned products and product categories are

organized under the regulations of the single Common Market Organization (CMO) of the EU

except for potatoes. Before 2008, each product category was under its own CMO. The

decoupling principle was enacted in 2006 for Greece and included all the product categories

under investigation. The decoupling principle which was implemented in Greece was the Single

Payment Scheme (SPS) which was based on the historical model according to which farmers

got aid according to the financial support they received between 2000 and 2002. What is more,

the organization of the Greek supply chain of fresh agricultural products was dominated by the

central markets of Athens and Thessaloniki, up to 2000. In the following years, retailers’

concentration raised sharply resulting in the decentralization of production distribution. Thus,

after 2000, the retailers and especially the supermarkets were the key players in the supply chain

resulting in ever more increased concentration of distribution and demand (Reziti, 2005; Reziti

and Panagopoulos, 2008).

In agricultural economics literature, the modelling of the price transmission mechanism

takes place with Autoregressive models as well as with their co-integrated variants in univariate

and multivariate forms. One of the first attempts to study price transmission was undertaken by

Wolffram (1971). However, the first models of price transmission did not take into account the

properties of the price series. More specifically, they did not take into consideration the non-

stationarity of the data and did not incorporate the concept of co-integration as shown by von

Cramon-Taubadel (1998), among others. Thus, more advanced models were preferred such as

the ones from Enders and Granger (1998), Goodwin and Piggott (2001) and Krolzig (1997).

However, the limited length of time series data was preventing the clear identification between

a permanent response to a shock implied by a unit root and a response with extended half-life

(Doan, 2012). This shortcoming of time series data gave rise to panel Vector Autoregressive

(VAR) models and their co-integrated variants which extent the cross section dimension of the

data set instead of endlessly expanding the dimension of time. In the economic literature, there

is a growing body that utilizes the advances of these models. As Canova and Ciccarelli (2013)

state, economic analysis and policy evaluation can significantly be improved by looking into

the interdependencies across sectors, markets and countries. The majority of papers that apply

panel VAR models use macroeconomic and financial data sets. Caivano (2006) tested the

transmission of shocks from Euro area to the United States and vice versa. Beetsma and

Giuliadori (2011) study the transmission of shocks in the government spending of the EU. On

29

the other hand, Love and Zicchino (2006) investigated the relationship between firms’ financial

condition and degree of investment. Moreover, panel models are widely used in the

investigation of the relationship between energy and output in conjunction with relative

economic variables. Lee (2005) studied the relationship between the energy consumption and

output in developing countries with a co-integrated panel model. Apergis and Payne (2009)

examined the same relationship in the Commonwealth of Independent States. However, the

papers that study food price relationships are scarce. Nazlioglu and Soytas (2012) investigated

the relationship between agricultural commodity and oil prices jointly with US dollar exchange

rate by implementing a co-integrated panel VAR analysis. Whilst, Rezitis (2015) studied the

same relationship with a panel VAR model.

In this chapter, the empirical methodology implements a panel Vector Error Correction

(VEC) model in order to study the price transmission mechanism of the Greek food sector

between the producer and the consumer while assessing the impact of the decoupling. This

chapter implements a panel VEC model for the study of the vertical price mechanism of a

national agricultural market while giving new evidence on the impact of decoupling in a

member state of the EU. The panel VEC model is analyzed by testing for causality relationships

in the short and long run as well as by calculating the individual impulse responses of the

seventeen products and product categories. The causality tests results show that the producer is

not responding to long-run deviations from the equilibrium while the analysis of the impulse

responses reveals that the own effects of a shock to the producer or the consumer price die out

quickly whereas the spillover effects of the shock are more persistent. Moreover, the price

mechanism is examined for the existence of possible asymmetries. The in-sample asymmetric

tests verify the non-existence of asymmetric responses giving validation to the symmetric panel

VEC model. The effect of the decoupling on the price mechanism is examined by dividing the

sample into two sub-samples. The first subsample expands from 1995 to 2006 while the second

one from 2007 to 2013. The breakpoint is considered to be 2007, since it is the first year that

the decoupling principle is fully adopted by Greece. Moreover, according to the European

Commission (2013), in 2007, the decoupling principle is adopted in over 80% in the EU. Each

of the sub-samples is analyzed for causality effects in the short and long run along with the

estimation of impulse responses. The results of the impulses responses calculated for each

subsample show that the decoupling changed the dynamics of the price mechanism between the

producer and the consumer. More specifically, the implementation of the decoupling in the

Greek food sector in 2007 seems to have enabled the consumer to mitigate the effects of price

shocks smoother than the producer.

30

The rest of the chapter is structured as follows. In section 2, the methodology used is

presented. Section 3 presents the data set utilized while section 4 exhibits the empirical results

of the chapter. Finally, section 5 concludes.

3.2 Econometric Methodology

In order to investigate the price transmission mechanism between the producer and the

consumer in the Greek food market, firstly, the properties of the panel data series are examined

with panel unit roots. Then, the existence of a long-run relationship is investigated by panel co-

integration analysis. The co-integration analysis is implemented as it is described in Pedroni

(2000). According to the co-integration analysis, two sets of tests are used for the examination

of a possible co-integrating relationship while its estimation takes place with a regression

estimate. The Pedroni (2000) co-integration analysis assumes that the determinants of the long

run relationship are heterogeneous something that is verified with heterogeneity tests. This co-

integration analysis is preferred to one where the determinants of the long run relationship are

assumed to be homogeneous (Pesaran et al, 1999). More specifically, let it

P be the logged

producer prices and it

C the logged consumer prices where 1...t T , T is the time length and

1...i N , N is the number of the cross sections. The general structure used for the panel unit

roots is:

, 1 , ,1

(3.1)k

it i i t i j i t j i it itj

y y y d u

where ity is the logged producer ( )it

P or consumer ( )itC prices, is the first difference

operator, k is the lag length, it

d are the deterministic components and it

u is the error term. If

0i

then the y process has a unit root for individual i , while if 0i

then the process is

stationary around the deterministic part. As soon as, the non-stationarity of the variables under

investigation has been verified, the panel co-integration analysis takes place by estimating the

long-run relationship between the producer and the consumer according to Pedroni (2000).

Fully Modified Least Squares (FMLS) and Dynamic Ordinary Least Squares (DOLS) are used

for the regression estimate. The long-run relationship estimated is formulated as:

(3.2)it i it i it itC d P

31

The two sets of tests proposed by Pedroni (1999) for verifying the long-run relationship are

utilized. The two sets of tests are based on the within as well as the between dimension

approaches. The null hypothesis 0

( )H of the tests is that there is no co-integration. The testing

of the null hypothesis of no co-integration takes place by the following unit root form for the

error term it :

, 1 , ,

1

(3.3)iK

it i i t i k i t k it

k

v

where it is the error term from (2). If 0i then the residuals are stationary, thus, a co-

integration relationship exists. The heterogeneity of the slope coefficients of (2) is verified by

heterogeneity tests. The null hypothesis 0H of the heterogeneity test is that each individual’s

coefficient is equal to the average of all. In the next stage, the estimated co-integrating vector

is incorporated in the panel Vector Error Correction form. The model estimated is unrestricted,

though the Minnesota Bayesian shrinkage prior is employed so as to effectively reduce the

dimensionality of the coefficients vector (Canova and Ciccarelli, 2013). The model is given by:

1 11 , 12 , 1 , 1 1

1 1

2 21 , 22 , 2

1 1

(3.4)

k k

it i ip i t p ip i t p i i t it

p p

k k

it i ip i t p ip i t p i i

p p

P P C

C P C

, 1 2t it

where ,i t p

P

is the first-differenced lagged prices of the producer and ,

Ci t p

is the first-

differenced lagged prices of the consumer, , 1i t

is the one lagged error term obtained from (2).

The analysis of the panel VEC models takes place by causality tests as well as impulse

responses. The causality tests are conducted using the formal procedure suggested by Granger

(1969) for the short run while long run causality is examined in the notion of weak exogeneity.

The tests for the short run follow an F-distribution while the long run tests follow a t-student

distribution. The long-run causality implies that the producer (consumer) price does not adjust

to long-run deviations from the equilibrium however producer (consumer) price may still react

to lagged changes of the consumer (producer) price in the short run. Therefore, Granger

32

causality tests are utilized, which show whether the producer (consumer) still reacts to lagged

price changes of the consumer (producer). It should be noted that the existence of long run

causality might imply a possible attempt by the producer (consumer) to exert influence over the

pricing behavior of the consumer (producer) (Engle et al., 1983). More specifically, the Granger

causality test examines whether the lagged price changes of the producer, ,i t p

P

, (consumer,

,i t pC

) improve the forecasting of consumer (producer) price. The long run causality test checks

whether the coefficients of the error correction terms ( )it

are statistically insignificant. Table

3.1 shows the null hypothesis of the tests.

Table 3.1 Causality tests

Short run Long run

Producer 21,1 21,

... 0p ip

1 0i

Consumer 12,1 12,

... 0p ip

2 0i

The dynamics of the price mechanism are further investigated by the out-of-sample measure of

impulse responses. The impulse responses reveal how the producer and the consumer absorb

shocks through time while depicting the impact of these shocks to the rest of the variables under

consideration. In particular, a shock to the producer price results in own and spill-over effects.

The own effects depict the impact that the shock has to the producer price while the spill-over

effects reveal the effect of the shock to the consumer price. Similarly, a shock to the consumer

price would have own as well as spillover effects. The own effects would show the impact of

the shock to the consumer while the spillover effects would show the impact to the producer.

In order to investigate the presence of price asymmetries, the price transmission

mechanism should be allowed to take into consideration the different price adjustments that

might occur according to whether prices are increasing or decreasing. This is done by including

dummy variables in the following way: 1 if ΔP , ΔC , ect 0 0 otherwise

it it it

tD

and

1 if ΔP , ΔC , ect 0 0 otherwise

it it it

tD

. So, the Asymmetric Error Correction model takes the form:

33

( ) ( ) ( )

0 ,1, , ,1, , ,2, ,1 1 1

( ) ( ) ( )

,2, , , 1 , 11

( )

0 ,2, , ,2,1

(3.5)

K k Kp p p p

it i i k t i t k i k t i t k i k t i t kk k k

K p p p p

i k t i t k t i t t i t itk

Kc c

it i i k t i t k i kk

P D P D P a D C

a D C D ect D ect u

C a D C a

( ) ( )

, ,1, ,1 1

( ) ( ) ( )

,1, , , 1 , 11

(3.6)

K Kc c

t i t k i k t i t kk k

K c c c c

i k t i t k t i t t i t itk

D C D P

D P D ect D ect u

The terms , , , 1

, ,t i t k t i t k t i t

D P D C D ect

will be referred to as

, , , 1, ,

i t k i t k i tP C ect

and the terms

, , , 1, ,

t i t k t i t k t i tD P D C D ect

as

, , , 1, ,

i t k i t k i tP C ect

for ease of notation. Next, the equations of

the producer and the consumer are separately estimated with the Minnesota Bayesian shrinkage

prior. The in-sample asymmetric tests are presented in table 3.2.

Table 3.2 The in-sample asymmetric tests

:o

H Null Hypothesis of Symmetry

Short-Run Long-Run

Equations Distributed Lag Effect Autoregressive Lag Effect

Producer / it

P ( ) ( )

,2, ,2,1 1

K kp p

i k i kk k

p( ) p( )

,1, ,1,1 1

K k

i k i kk k

( ) ( )p p

it it

Consumer / it

C c( ) c( )

,1, ,1,1 1

K k

i k i kk k

c( ) c( )

,2, ,2,1 1

K k

i k i kk k

( ) ( )c c

it it

Finally, the effect of the decoupling in the Greek agricultural sector is investigated by

short- and long-run causality tests along with impulse responses for both subsamples. As it is

already mentioned, subsample 1 spans from 1995 to 2006 while subsample 2 extends from 2007

to 2013. In this way, the direction of the price relationship as well as the possible asymmetric

response of the producer and the consumer to a price shock in the period before decoupling and

the period after it is assessed. The asymmetric response of the producer and the consumer

between the two periods is calculated as the absolute value of the impulse response of the

producer or the consumer to a shock i

in subsample 1 minus the absolute value of the impulse

response of the producer or the consumer to the shock in subsample 2. That is:

1995 2007| ( ) | | ( ) |

i iasymmetry irf irf , where

1995| ( ) |

iirf is the impulse response of the

producer or the consumer in subsample 1 in absolute value and 2007

| ( ) |i

irf is the impulse

response of the producer or the consumer in subsample 2 in absolute value. In this way, the

34

result of the subtraction can be seen as a more intense response of the producer or the consumer

to a shock in the pre-decoupling period when the effect is positive asymmetric whereas a

negative asymmetric effect would imply a stronger response in the post-decoupling period.

3.3 Data

The national averages of monthly price indices of seventeen Greek products and product

categories for the producer and the consumer are used. The price indices were provided by the

Hellenic Statistical Authority. Producer price indices are part of the Producer Price Index of

Agricultural Output while consumer indices are part of the Consumer Price Index. The producer

and the consumer price indices expand from January 1995 to September 2013. Thus, the time

length T is 225 months per product. The total observations account to 3825. All price indices

are converted into their natural logarithms and they are nominal. Table 3.3 shows the products

that are used with their weights in the Producer Price Index.

Table 3.3 The products under investigation and their weights in the Greek Producer Price

Index

Vegetal

products

% weights Animal

products

% weights Other

products

% weights

Potato 2.591 Milk 12.007 Cereals 6.533

Cucumber 0.803 Beef 3.035 Wine 1.615

Marrow 0.601 Pork 3.144 Olive Oil 11.428

Onion 0.824 Sheep 4.365 Olives 1.652

Tomato 5.5 Goat 2.682

Lemon 0.405 Poultry 1.681

Eggs 1.495

3.4 Empirical Results

3.4.1 Unit Roots and Co-integration Analysis

Prior to the co-integration analysis, unit root tests are performed so that the stationarity

of the producer and the consumer price series can be examined. The unit root tests that are used

35

are the Levin et al. (2002) test, the Im et al (2003) test and the Breitung (2000) test. The first

two tests are based on bias-corrected estimators whereas the other one is based on unbiased

estimators. Additionally, Levin et al. and Breitung tests have as alternative hypothesis that i

in (1) are identical and negative while Im et al. test examines whether some of the cross-sections

can have a unit root. Despite, the different specifications of the tests the next step under the

finding of non-stationarity is co-integration analysis. The results of the three tests support the

existence of non-stationarity and are presented in table 3.4.

Table 3.4 The results of the panel unit root tests

Levin – Li – Chu Im – Pesaran – Shin Breitung

itP 2.756 (0.997) -0.110 (0.456) 2.204 (0.986)

itC 2.271 (0.988) 0.400 (0.655) 7.730 (1.000)

itP -11.38 (0.000) -22.58 (0.000) -14.15 (0.000)

itC -5.287 (0.000) -22.99 (0.000) -6.435 (0.000)

In the parenthesis p-values are reported. The deterministic part of the unit root processes is a constant for all

tests. The most appropriate unit root process for each one of the tests is chosen by Information Criteria and

Likelihood Ratio tests.

Since, the results of the unit root tests showed the existence of non-stationarity, the next step of

co-integration analysis takes place. The co-integration analysis is implemented as described in

Pedroni (1999, 2004). Two sets of tests are utilized. The panel tests which are based on the

within dimension approach and include the v-, ρ-, PP- and ADF- statistics and the group tests

which are based on the between dimension approach and include the ρ-, PP- and ADF- statistics.

All seven tests follow asymptotically the standard normal distribution and they confirm that the

producer and consumer prices are co-integrated as is evident by table 3.5.

Table 3.5 The results of the panel co-integration tests

Within dimension (panel statistics) Between dimension (group statistics)

Panel v-statistic 7.640 (0.000) Group ρ-statistic -15.870 (0.000)

Panel ρ-statistic -7.580 (0.000) Group PP-statistic -10.500 (0.000)

Panel PP-statistic -6.430 (0.000) Group ADF-statistic -5.770 (0.000)

Panel ADF-statistic -3.710 (0.000)

In the parenthesis p-values are reported.

36

Table 3.6 shows the results of the estimated co-integrating vectors for the group and the cross-

sections. The group presents the estimated co-integrated vector for the whole panel while the

cross sections present the estimated co-integrating vectors of the individual products and

product categories. The Fully Modified Least Squares (FMLS) and the Dynamic Ordinary Least

Squares (DOLS) have been used for the estimation process.

Table 3.6 The co-integrating vector estimates according to FMLS and DOLS

FMLS DOLS

constant trend tP constant trend

tP

group 2.305

(0.000)

0.001

(0.000)

0.444

(0.000)

group 2.222

(0.000)

0.001

(0.000)

0.464

(0.000)

Cereals 3.954

(0.000)

0.003

(0.000)

0.008

(0.373)

Cereals 3.887

(0.000)

0.003

(0.000)

0.023

(0.277)

Potato 0.987

(0.000)

5∙10-4

(0.050)

0.774

(0.000)

Potato 0.896

(0.001)

5∙10-4

(0.021)

0.794

(0.000)

Cucumber 1.690

(0.000)

0.001

(0.001)

0.591

(0.000)

Cucumber 1.163

(0.003)

6∙10-4

(0.027)

0.713

(0.000)

Marrow 1.639

(0.000)

-2∙10-4

(0.660)

0.663

(0.000)

Zucchini 1.763

(0.004)

-1∙10-4

(0.562)

0.633

(0.000)

Onion 1.651

(0.000)

0.001

(0.000)

0.594

(0.000)

Onion 1.392

(0.000)

0.001

(0.001)

0.658

(0.000)

Tomato 2.123

(0.000)

0.002

(0.000)

0.485

(0.000)

Tomato 2.218

(0.000)

0.002

(0.000)

0.461

(0.000)

Lemon 2.564

(0.000)

16∙10-4

(0.003)

0.378

(0.000)

Lemon 2.586

(0.000)

14∙10-4

(0.007)

0.377

(0.000)

Wine 3.676

(0.000)

0.003

(0.000)

0.069

(0.158)

Wine 3.484

(0.000)

0.003

(0.000)

0.117

(0.154)

Olive Oil 2.078

(0.000)

0.001

(0.000)

0.506

(0.000)

Olive Oil 2.044

(0.000)

0.001

(0.000)

0.514

(0.000)

Olives 2.689

(0.000)

0.001

(0.069)

0.395

(0.004)

Olives 2.736

(0.000)

0.001

(0.105)

0.384

(0.011)

Milk 1.9866 14∙10-4 0.526 Milk 1.596 0.001 0.618

37

(0.005) (0.000) (0.002) (0.031) (0.002) (0.002)

Beef 3.2672

(0.000)

0.002

(0.000)

0.2164

(0.003)

Beef 3.222

(0.000)

0.002

(0.000)

0.227

(0.004)

Pork 3.2173

(0.000)

13∙10-4

(0.000)

0.252

(0.000)

Pork 3.269

(0.000)

13∙10-4

(0.000)

0.241

(0.002)

Sheep 1.3715

(0.000)

7∙10-4

(0.000)

0.673

(0.000)

Sheep 1.334

(0.000)

7∙10-4

(0.000)

0.682

(0.000)

Goat 0.4487

(0.112)

1∙10-4

(0.408)

0.901

(0.000)

Goat 0.230

(0.310)

-1∙10-4

(0.856)

0.953

(0.000)

Poultry 1.8287

(0.000)

0.001

(0.000)

0.551

(0.000)

Poultry 1.845

(0.000)

13∙10-4

(0.000)

0.547

(0.000)

Eggs 4.0357

(0.000)

0.004

(0.000)

-0.039

(0.634)

Eggs 4.102

(0.000)

0.004

(0.000)

-0.055

(0.666)

DW (0.098) DW (0.097)

AIC -3.810 AIC -3.772

SIC -3.764 SIC -3.648

HQIC -3.791 HQIC -3.722

Heterogeneity tests

constant trend tP constant trend

tP

group 392.6

(0.000)

674.6

(0.000)

465.6

(0.000)

group 217.9

(0.000)

553.3

(0.000)

249.9

(0.000)

In the parenthesis p-values are reported.

The optimal model for each method was selected by the information criteria of Akaike (AIC),

Schwarz (SIC) and Hannan – Quinn (HQIC) as well as the implementation of likelihood ratio

tests. Moreover, the Durbin-Watson test for possible autocorrelation (DW) is used. The results

of the co-integration analysis according to both methods (FMLS and DOLS) reveals that the

producer and the consumer prices are co-integrated in the long run while the coefficients are

significant. Similarly, for each one of the 17 products, the producer and the consumer prices

are in a long-run relationship with statistically significant coefficients. However, three products

(cereals, wine and eggs) have a non-significant coefficient for the producer price. The co-

integrating vector that is estimated by FMLS is incorporated into the panel VEC model since

the information criteria indicate that is better specified. Moreover, the heterogeneity tests verify

38

that the individual co-integrating vectors are heterogeneous as shown in table 3.6. Thus,

justifying the utilization of Pedroni’s co-integration method instead of a homogeneous one.

3.4.2 The results of the panel VEC model

The most appropriate panel VEC model for the price transmission mechanism of the

Greek agricultural sector is determined according to the information criteria of Akaike (AIC),

Swartz (SIC) and Hannan-Quinn (HQIC) as well as likelihood ratio tests. The model that is

selected consists of 6 lags for the producer and the consumer while no signs of autocorrelation

are present. Table 3.7 presents the results of the panel VEC model.

Table 3.7 The panel VEC estimation results of the Greek agricultural sector from 1995 to

2013

Producer Consumer

Constant 0.003 (0.024) 0.004 (0.000)

, 1i tP -0.022 (0.282) 0.031 (0.029)

, 2i tP -0.204 (0.000) -0.063 (0.000)

, 3i tP -0.123 (0.000) -0.068 (0.000)

, 4i tP -0.088 (0.000) -0.035 (0.011)

, 5i tP -0.123 (0.000) -0.023 (0.086)

, 6i tP -0.193 (0.000) 0.026 (0.050)

, 1i tC 0.136 (0.000) 0.250 (0.000)

, 2i tC 0.131 (0.000) 0.017 (0.397)

, 3i tC -0.068 (0.017) 0.008 (0.685)

, 4i tC 0.021 (0.448) 0.022 (0.261)

, 5i tC -0.060 (0.026) -0.092 (0.000)

, 6i tC 0.001 (0.963) -0.190 (0.000)

, 1i t -0.012 (0.616) -0.253 (0.000)

DW (0.163) (0.149)

AIC -4.472

39

SIC -4.448

HQIC -4.463

In the parenthesis p-values are reported.

The dynamics of the panel model are investigated by causality tests as well as impulse

responses. The results of the short and long run causality are presented in table 3.7a. The tests

reveal that the producer does not adjust to deviation from the equilibrium in the long run but he

reacts to the lagged price changes of the consumer in the short run. On the other hand, the

consumer adjusts to the long run equilibrium as well as to lagged price changes of producer in

the short run.

Table 3.7a The tests of short and long run causality from 1995 to 2013

Short run Long run

Producer 11.28 (0.000) -0.501 (0.616)

Consumer 10.79 (0.000) -14.71 (0.000)

The out-of-sample measure of impulse responses for vegetal products is presented in figure

3.1a. As it is observed from the figure, a shock to the producer price has a similar own effect

for all six products that lasts for two to three months. Though, the spillover effect is more

complex, thus characterized by memory effects and duration of six months. It is notable that

cucumbers have a negative response to the shock. On the other hand, a shock to the consumer

price results in a similar spillover response for all six products that lasts for about three months

with intense memory effects. The own effect of the shock is similar for the five out of the six

products while spanning for three months. Though, lemons need a whole of twelve months for

eliminating the shock.

Figure 3.1a The impulse responses of the producer and consumer prices for vegetal products

Producer shocked (own effect) Consumer (spillover effect)

40

Producer (spillover effect) Consumer shocked (own effect)

The impulse responses for animal products are presented in figure 3.1b. As is evident from the

figure, the own effect of a shock to the producer price lasts for two to three months for all

animal products before fading away. Moreover, milk as well as sheep and goat meat exhibit

cyclical memory effects. However, the spillover effect of the shock is more unstable as in the

case of vegetal products and it lasts for two to three months. Though, milk in a smaller degree

and sheep and goat meat in a bigger degree follow a response path which is the opposite of the

other meat products. On the other hand, a shock to the consumer results in smooth own effects

which last for almost three months before fading away. The spillover response of the producer

to the shock is more complex even though all products follow the same response path. The

shock decays in almost four months while displaying slight memory effects.

Figure 3.1b The impulse responses of the producer and consumer prices for animal products

Producer shocked (own effect) Consumer (spillover effect)

41

Producer (spillover effect) Consumer shocked (own effect)

The impulse responses for other products are presented in figure 3.1c. A shock to the producer

results in own effects that last for two to four months. Moreover, cereals and wine exhibit

memory effects in their path to equilibrium. The spillover effect of the shock is more intense

for olive oil and olives, showing prolonged memory effects before convergence. In contrast,

cereals do not show any response to the shock while wine follows the response of oil and olives

but with the opposite sign. On the other hand, a shock to the consumer results in own effects

that last for just two months for all four products. However, the spillover effects of the shock

are much more unstable presenting memory effects for all four products that last for almost

seven months with the exception of wheat where the memory effects are prolonged for three

more months.

Figure 3.1c The impulse responses of the producer and consumer prices for other products

Producer shocked (own effect) Consumer (spillover effect)

42

Producer (spillover effect) Consumer shocked (own effect)

In order to shed more light to the dynamics of the price generation in the Greek

agricultural sector, the price mechanism is investigated for possible asymmetries using in-

sample asymmetric tests. The most appropriate asymmetric model is determined according to

the information criteria of Akaike (AIC), Swartz (SIC) and Hannan-Quinn (HQIC) and the

likelihood ratio tests. The model selected consists of 6 lags for the producer and the consumer

while no signs of autocorrelation are present. Table 3.8 presents the estimated coefficients of

the asymmetric error correction models for the producer and the consumer.

Table 3.8 The asymmetric model estimation results from 1995 to 2013

Producer Consumer

constant 0.006 (0.012) 0.005 (0.003)

1tP

-0.110 (0.000) -0.009 (0.666)

2tP

-0.168 (0.000) -0.064 (0.002)

3tP

-0.092 (0.003) -0.084 (0.000)

4tP

-0.061 (0.047) 0.018 (0.382)

43

5tP

-0.152 (0.000) 0.006 (0.760)

6tP

-0.184 (0.000) 0.017 (0.402)

1tP

0.0514 (0.118) 0.047 (0.036)

2tP

-0.234 (0.000) -0.070 (0.002)

3tP

-0.145 (0.000) -0.034 (0.132)

4tP

-0.131 (0.000) -0.096 (0.000)

5tP

-0.100 (0.005) -0.064 (0.004)

6tP

-0.202 (0.000) 0.035 (0.105)

1tC

0.114 (0.005) 0.206 (0.000)

2tC

0.172 (0.000) 0.052 (0.066)

3tC

-0.118 (0.004) 0.029 (0.290)

4tC

-0.056 (0.170) -0.010 (0.712)

5tC

-0.107 (0.007) -0.241 (0.000)

6tC

0.114 (0.004) -0.067 (0.015)

1tC

0.216 (0.000) 0.348 (0.000)

2tC

0.045 (0.352) -0.033 (0.327)

3tC

-0.004 (0.930) -0.017 (0.596)

4tC

0.099 (0.036) 0.041 (0.212)

5tC

0.014 (0.763) 0.120 (0.000)

6tC

-0.121 (0.006) -0.342 (0.000)

1tect

0.008 (0.820) -0.236 (0.000)

1tect

-0.014 (0.717) -0.252 (0.000)

DW (0.168) (0.164)

AIC -1.234 -1.791

SIC -1.199 -1.756

HQIC -1.222 -1.779

In the parenthesis p-values are reported.

44

The in-sample asymmetric tests for the Greek agricultural sector are presented in table 3.8a. An

inspection of the table reveals that, in the short run, the producer and the consumer respond

symmetrically to price increases or decreases. Similarly, the responses of the producer and the

consumer to deviations from the long-run equilibrium are symmetric. Thus, the results of the

tests reveal that the price mechanism is not characterized by asymmetries justifying the

estimation of the symmetric panel VEC model in the previous step.

Table 3.8a The asymmetric in-sample asymmetric tests for the Greek agricultural sector for

the period 1995-2013

Short-Run Long-Run

Distributed Lag Effect Autoregressive Lag Effect

Producer 0.013 (0.909) S 0.001 (0.974) S 0.001 (0.974) S

Consumer 0.005 (0.942) S 0.002 (0.960) S 0.001 (0.974) S

In the parenthesis p-values are reported, S denotes symmetry.

The empirical results that are used for the assessment of the price dynamics before and

after the full implementation of the decoupling are presented next. Table 3.9 shows the

estimated coefficients of the panel VEC model for the subsample from 1995 to 2006. The most

appropriate model was chosen according to the information criteria and likelihood ratio tests.

The model consists of 3 lags for the producer and the consumer while no signs of

autocorrelation are present.

Table 3.9 The panel VEC estimation results of the Greek agricultural sector from 1995 to 2006

(subsample 1)

Producer Consumer

Constant 0.003 (0.145) 0.004 (0.015)

, 1i tP -0.047 (0.070) -0.051 (0.006)

, 2i tP -0.191 (0.000) -0.099 (0.000)

, 3i tP -0.103 (0.000) -0.104 (0.000)

, 1i tC 0.193 (0.000) 0.393 (0.000)

, 2i tC 0.163 (0.000) 0.090 (0.000)

, 3i tC -0.062 (0.062) 0.101 (0.000)

45

, 1i t -0.073 (0.009) -0.403 (0.000)

DW 0.115 0.118

AIC -3.866

SIC -3.847

HQIC -3.859

In the parenthesis p-values are reported.

The results of the short- and long-run causality for the subsample from 1995 to 2006 are

presented in table 3.9a. The tests reveal that the producer and the consumer adjust to deviation

from the equilibrium in the long run and react to their lagged price changes. Thus, they

participate in a feedback relationship in the short and long run.

Table 3.9a The tests of short and long run causality from 1995 to 2006

Short run Long run

Producer 22.49 (0.000) -2.614 (0.009)

Consumer 18.96 (0.000) -20.06 (0.000)

In the parenthesis p-values are reported

Table 3.10 presents the estimated coefficients of the panel VEC model for the subsample from

2007 to 2013. The most appropriate model was chosen according to the information criteria and

likelihood ratio tests. The model that is selected consists of 1 lag for the producer and the

consumer while no signs of autocorrelation are present.

Table 3.10 The panel VEC estimation results of the Greek agricultural sector from 2007 to

2013 (subsample 2)

Producer Consumer

Constant 80∙10-3 (0.712) 0.002 (0.107)

, 1i tP 0.193 (0.000) 0.234 (0.000)

, 1i tC 0.045 (0.385) 0.161 (0.000)

, 1i t -0.041 (0.180) -0.157 (0.000)

DW (0.135) (0.146)

AIC -5.677

SIC -5.661

46

HQIC -5.671

In the parenthesis p-values are reported.

The results of the short- and long-run causality for the subsample from 2007 to 2013 are

presented in table 3.10a. The tests reveal that the producer does not adjust to deviation from the

equilibrium in the long run and he does not react to lagged price changes of the consumer. On

the other hand, the consumer adjusts to long-run deviations while responding to lagged changes

of the producer in the short run.

Table 3.10a The tests of short and long run causality from 2007 to 2013

Short run Long run

Producer 0.754 (0.385) -1.342 (0.179)

Consumer 154.6 (0.000) -9.446 (0.000)

In the parenthesis p-values are reported

Next, the out-of-sample measures of asymmetric responses of the Greek agricultural sector

between the two periods, before and after the decoupling of payments, are presented in figures

3.2a, 3.2b and 3.2c for the three food categories vegetal, animal and other products,

respectively. The asymmetric responses of the producer and the consumer for vegetal products

between the two subsamples are presented in figure 3.2a. From the figure it is obvious that a

shock to the producer results in negatively asymmetric own effects for the first two to three

months for the four out of the six products. Tomatoes and in a lesser extent lemons present

positive asymmetries until the decay of the shock. Moreover, cucumbers turn to strong positive

asymmetry just after the second month. Thus for the vegetal products, the producer responded

more to shocks during the decoupling period rather than the previous regime with the exception

of tomatoes, lemons and to some extent cucumbers for which producer’s response is more

intense during the pre-decoupling period. The spillover effect of the shock results also in

negative asymmetric effects during the first two to three months and positive for the next three

to four months. Consequently for all products, consumer’s response is stronger in the

decoupling period for the first months but then it gets stronger in the pre-decoupling period. On

the other hand, a shock to the consumer price results in positive asymmetric effects for both the

producer and the consumer. Particularly, the spillover effect of the shock is positive asymmetric

for the four out of the six products however is negatively asymmetric for lemons and to some

extent for tomatoes. Thus for vegetal, producer response is more intense during the pre-

47

decoupling period apart from lemons. Similarly, consumer response is positively asymmetric

indicating stronger effects for a shock before the enactment of the decoupling principle.

Moreover, the shock for lemons is permanent.

Figure 3.2a The asymmetric responses of the producer and consumer prices for vegetal

products

Producer shocked (own effect) Consumer (spillover effect)

Producer (spillover effect) Consumer shocked (own effect)

The asymmetric responses of the producer and the consumer to animal products price shocks

between the two sub-periods are presented in figure 3.2b. The own effects of a producer price

shock are positively asymmetric for about six months with the exception of pork where the

response is negatively asymmetric for three months. Thus, producers of meat products respond

more intensively in price shocks in the pre-decoupling period apart from pork producers who

react stronger in the post-decoupling period. In contrast, the spillover effects of the shock are

negatively asymmetric except for the sheep and goat prices which exhibit positive asymmetries.

Therefore, for the five out of the seven meat products consumer prices are more responsive in

the post-decoupling period with the exception of sheep and goat meat that are more responsive

48

during the pre-decoupling period. On the other hand, a shock to the consumer price results in

spillover and own effects that are positive asymmetric for the producer and the consumer,

respectively. More specifically, the spillover effects of the shock decay in about four months

for all seven products whereas the own effects of the shock last for a bit longer for four out of

the seven products. Moreover for poultry, eggs and beef, the shock’s duration is nine months.

Therefore, a price shock in the consumer results in both the producer and the consumer to react

more intensively during the pre-decoupling period.

Figure 3.2b The asymmetric responses of the producer and consumer prices for meat

products

Producer shocked (own effect) Consumer (spillover effect)

Producer (spillover effect) Consumer shocked (own effect)

The asymmetric responses of the producer and the consumer to other products price shocks

between the two sub-periods are presented in figure 3.2c. As is observed from figure 3.2c, a

shock to the producer results in negative own asymmetric effects for cereals and oil whereas

positive for wine and olive oils. In particular, cereals, oil and olives price shock fades away in

three to four months whereas wine needs nine months. Thus, for cereals and oil price shocks,

49

the response of the producer is stronger in post-decoupling era while for wine and olives in the

pre-decoupling period. The spillover effects of the shock result in negative asymmetric effects

for three months. Consequently, the consumer price responds more to shocks in the post-

decoupling period rather than to shocks in the previous regime. On the other hand, a shock to

the consumer price has similar own and spillover asymmetric effects for both the consumer and

the producer. More specifically, the price shock to the consumer results in positive asymmetric

effects for both the producer and the consumer however with different time duration. The

spillover effect of the shock to the producer lasts for about four months for cereals and oil whilst

eight months for wine and olives. However, the duration of the own effects of the shock are

longer. Thus, the response of the producer and the consumer to the shock was more intense

during the pre-decoupling era.

Figure 3.2c The asymmetric responses of the producer and consumer prices for other

products

Producer shocked (own effect) Consumer (spillover effect)

Producer (spillover effect) Consumer shocked (own effect)

3.5 Conclusions

50

The present chapter has analyzed the effect of the decoupling in the price transmission

mechanism of the Greek food sector between the producer and the consumer for the period

from January 1995 to September 2013. The price mechanism is modeled under the panel VEC

framework since it provides new insights to the issue of decoupling that the more conventional

time series analysis is not. More specifically, the empirical results did not reveal any signs of

asymmetric price transmission giving evidence on the use of a symmetric panel Vector Error

Correction (VEC) model. The causality analysis of the symmetric model showed that even

though the producer and the consumer interact for the price formation in the short run, in the

long run the producer does not respond to deviations from the long-term equilibrium. On the

other hand, the causality analysis of the panel VEC models of the two subsamples (i.e.

subsample 1 from 1995 to 2006 and subsample 2 from 2007 to 2013) revealed that in the pre-

decoupling period the producer and the consumer had a feedback relationship in the short and

long run whereas in the decoupling era, after 2006, producer did not react neither to consumer

price changes in the short run nor to the long-run deviations from the equilibrium.

The dynamics of the price mechanism were further explored by individual impulse

responses for the seventeen products and product categories which were grouped into vegetal,

animal and other agricultural products. The analysis of the impulse responses of the three

product categories for the entire sample (i.e. 1995 – 2013) revealed that the own effects of a

shock fade away on average in 3 months and the response of the producer and the consumer to

the shock is smooth. In contrast, the spillover effects of the shock are characterized by longer

duration and memory effects. Thus, a shock to the producer or the consumer price results in

own effects that are absorbed in a timely and smooth manner, however the spillover effects of

the shock are sluggish and complex. Moreover, the responses of the producer and the consumer

to price shocks before and after the implementation of the decoupling is investigated. For the

three product categories, the analysis shows that a shock to the producer results in a mixture of

positive and negative asymmetries for both the producer and the consumer. In contrast, a shock

to the consumer gives rise to only positive asymmetries. Thus, the implementation of the

decoupling seems to have decreased the response of the producer and the consumer to price

shocks that originate from the consumer. However, it has resulted in mixed effects regarding

the response of the producer and the consumer to price shocks that come from the producer

which depend to the specific product.

The aforementioned empirical results contradict the findings of previous research which

has showed that the producers did not change their expectations after the enactment of the

51

decoupling. In particular, the empirical findings of chapter 3 show that the response of both

market levels to producer price shocks is mixed and depends on the product, thus indicating

that the producer has changed his pricing behavior in the Greek agricultural sector. Moreover,

this chapter assessed the effect of the decoupling on the pricing behavior of the consumer along

the food supply chain providing useful empirical results. Additionally, the data set that is used

is a panel one, thus allowing for the investigation of the effects of the decoupling for the main

agricultural products of a single country instead of focusing on a single product. Furthermore,

the results show the changing pricing relationship between the producer and the consumer

before and after the implementation of the decoupling. More specifically, the causality tests

reveal that during the pre-decoupling era the producer and the consumer were in a feedback

relationship in the short and long run whereas after the implementation of decoupling the

producer did not respond to lagged price changes of the consumer and to deviations from the

long-run equilibrium.

In concluding, the main finding of this chapter is that the effects of decoupling on the

producer and the consumer regarding their response to price shocks depends on the source of

the shock. If the shock originates from the consumer both the producer and the consumer seem

to respond more to price shocks in the pre-decoupling period. If the shock comes from the

producer the response of the producer and the consumer depends on the individual

characteristics of the product that is investigated and the intensity of their response varies

between the two periods. Thus, these findings provide evidence against the trend of CAP to

eliminate product-specific policy measures by substituting them with more general ones for the

sake of simplicity of the regulative framework. The trade-off between the generality of the

policy framework for simplicity and the complexity of the product-specific measures should

probably reexamined.

52

4. Investigating the price transmission mechanisms of Greek fresh potatoes, tomatoes and

cucumbers markets

Chapter 4 investigates the price transmission mechanisms between the producer and the

consumer for three major Greek agricultural products: fresh potatoes, tomatoes and cucumbers.

The empirical analysis uses a Markov Switching Vector Error Correction model which accounts

for asymmetric responses of the producer and the consumer to price increases and decreases.

The asymmetric effects are tested with in-sample as well as out-of-sample measures for each

state. The response of the producer and the consumer to long-run deviations from the

equilibrium and to lagged price changes is investigated along with the nature of the price shocks

for each state. The empirical results show that potatoes and cucumbers give rise to similar price

mechanisms even though the underlying characteristics of these two markets are different.

Moreover, the empirical analysis shows that tomatoes and cucumbers result in different price

relationships between the producer and the consumer despite the similar market characteristics.

This chapter is a reproduction of Rezitis and Pachis (2015).

4.1 Introduction

The asymmetric price transmission literature examines agricultural markets for

rigidities in the price transmission mechanism. In particular, asymmetric price transmission can

take place when the producer price decreases but this decrease does not reach consumers in full

or reaches them at a slow pace. Similarly, when an increase in the consumer price takes place,

the producers do not benefit. This issue is always at the top of the agenda of policy makers

while it is a major concern of both consumers and producers (von Cramon-Taubadel, 1998).

In the agricultural economics literature, a number of reasons have been proposed for

explaining the asymmetric effects in the price transmission mechanisms. The most common

reason proposed is the market power of wholesalers and/or retailers. That is, non-competitive

market structures allow wholesalers and retailers to transmit price changes that squeeze their

margin more quickly and/or more fully than price changes that stretch their margin (Meyer and

von Cramon-Taubadel, 2004). A frequently cited study that makes an attempt to link price

asymmetry with market power is one undertaken by Peltzman (2000). Moreover, Neumark and

Sharpe (1992) show empirically how market power leads to price asymmetries in the banking

sector of the United States. Borenstein et al. (1997) also suggest that market power causes price

53

asymmetry in the United States gasoline market. However, they make the assumption that the

oligopolistic structure of the market results in the accumulation of market power by the retailers

without showing it empirically. The second most frequently cited reason for price asymmetry

in the literature refers to transaction costs. Transaction costs incorporate adjustment costs, menu

costs, inventory management or any other costs incurred when an economic actor participates

in a market. Therefore, when firms have to change the quantities or the prices of their inputs

and outputs, asymmetric price transmission may result. Bailey and Brorsen (1989) show how

adjustment costs can result in asymmetric price transmission. Balke et al. (1998) indicate how

accounting rules can lead to asymmetries. Furthermore, a related reason to transaction costs that

can cause asymmetries is the asymmetric flow of information between markets as shown by

Abdulai (2000). Finally, agricultural policy is also proposed as a cause of asymmetries by

Kinnucan and Forker (1987).

One of the first attempts to study asymmetric price transmission was undertaken by

Wolffram (1971) and later on this approach was improved by Houck (1977) and Ward (1982).

However, these models did not take into account the properties of the price series. More

specifically, they did not take into consideration the non-stationarity of the data and could not

incorporate the concept of co-integration as shown by von Cramon-Taubadel (1998), among

others. In that article, von Cramon-Taubadel proposed a model that is consistent with the

properties of the price series. In a similar vein, Enders and Granger (1998) proposed the concept

of threshold co-integration. They suggested that it is highly probable for price series to be co-

integrated not in a linear form but in a non-linear one. These types of studies exploit the error

correction model in conjunction with threshold co-integration. In a recent paper, Sun (2011)

examines the price transmission mechanism of furniture in the US with a threshold co-

integration model. A more contemporary approach to asymmetric price transmission involves

the regime-switching models. Goodwin and Piggott (2001) proposed the threshold vector error

correction model by investigating the price linkages between the soybean and corn markets in

the United States. Moreover, Ben-Kaabia and Gil (2007) used a threshold vector error

correction model to explore the Spanish lamb market. They found that retailers benefit from

shocks that affect the marketing channel of lamb. Krolzig (1997) suggested the Markov

Switching Vector Error Correction (MSVEC) model, which was a generalization of Hamilton’s

model (1989). As Frey and Manera (2007) discuss, a regime-switching model is part of the

family of models in which the relationship between the variables of interest depends on the state

of a variable t

s , which can either be part of the explanatory variables or not. This variable is

54

called the transition variable. The level of t

s , with respect to a threshold , describes different

states of the world or regimes. Therefore, the regime-switching model has the property to

account for the behavior of economic actors under different circumstances. A further

categorization defines two classes of models according to the nature of the state variable, which

can be deterministic or stochastic. In the case of the MSVEC model, the price transmission

mechanism passes through different states according to an unobserved transition state variable,

which follows a first-order ergodic Markov chain. Thus, the transition variable is stochastic and

the shift from one regime to another is random. Additionally, the transmission variable is

defined by the data itself and not by the researcher. In the case of the threshold vector error

correction model, the transition variable is deterministic and is defined by the researcher.

Brummer et al. (2009) implement an MSVEC model for the examination of the price

transmission mechanism between wheat and flour in Ukraine. They demonstrate that the price

transmission mechanism has been affected by the policy interventions of the Ukrainian

government.

The aim of this chapter is to investigate the price transmission mechanisms of fresh

potatoes, tomatoes and cucumbers in Greece. As a member of the European Union (EU), Greek

agricultural markets are regulated by the Common Agricultural Policy (CAP) of the EU. In

particular, the survey of the European Parliament (2011) reports that the tomato and the

cucumber markets are regulated by the Common Market Organization (CMO) for fruits and

vegetables. The CMO is market-oriented and the CAP reforms of 1996 and 2000 encouraged

producers to join Producer Organizations (POs). However, fresh potatoes have not been

organized in a common market since the enactment of the CAP in 1962 (European Committee,

2007). For the period under examination, the Greek fresh potato market is characterized by

increasing imports, however the self-sufficiency of the country amounts to over 80%. Similarly,

the Greek tomato and cucumber markets are characterized by small imported quantities

resulting in a self-sufficiency of 95% and 99%, respectively. The self-sufficiency of Greece for

the three products implies that imports are not expected to affect domestic prices. Moreover,

the three products account for 9% of the Greek Producer Price Index of Agricultural Output,

indicating their high importance for producers. Furthermore, according to the above surveys,

Greece has the largest vegetable consumption in the EU while fresh vegetables and potatoes

account for 9% of the Greek Food Consumer Price Index, verifying the importance of fresh

vegetables for Greek consumers as well. However, the three products account for just 0.096%

of the domestic Manufacturing Producer Price Index, indicating the preference of Greek

55

consumers for fresh agricultural products. Up to 2000, the Greek supply chain of fresh

agricultural products was dominated by the central markets of Athens and Thessaloniki.

However, in the following years, retailers’ concentration increased sharply, resulting in the

decentralization of production distribution, which now was dominated by the supermarkets

(Reziti, 2005; Reziti and Panagopoulos, 2008). The above discussion shows that even though

the reforms of the CAP since 1992 have been aimed at exposing domestic EU prices to the

signals of the world markets, the strong demand of Greek consumers for potatoes, tomatoes and

cucumbers of domestic origin has dampened the effects of price transmission from international

to domestic prices for the products under consideration. Therefore, the domestic marketing

characteristics of the three fresh agricultural products are expected to play a major role in the

way shocks are absorbed domestically.

Even though fresh potatoes, tomatoes and cucumbers constitute a major part of Greek

consumers’ diet in terms of quantity and value as well as a large proportion of Greek vegetable

production, they have not been studied thoroughly. The only attempts to study Greek fresh

agricultural products were undertaken by Reziti (2005) and Reziti and Panagopoulos (2008)

with an asymmetric error correction model. Both studies find that asymmetry is present in the

long run. Therefore, chapter 4 aims to enrich the literature on fresh vegetables by using an

MSVEC model that allows for asymmetric adjustment of the price mechanism to positive and

negative price shocks. Cologni and Manera (2009) also propose an asymmetric Markov

switching model in a univariate autoregressive framework, without taking into consideration

the possible non-stationarity of prices. The empirical analysis shows that the prices of the

producer and consumer of fresh potatoes, tomatoes and cucumbers are co-integrated and their

price mechanisms switch between two states. Each state of the price mechanism of the three

products is characterized by different patterns of asymmetries, non-responses to long-run

deviations from the equilibrium and lagged changes of the prices along with different price

shocks. The methodology proposed brings inference of Markov models a step further by

providing the opportunity to agricultural researchers to shed new light on the price mechanisms

of agricultural products under switching stochastic states by the use of well-known methods.

The rest of the article is structured as follows. In section 2, the econometric methodology

of the asymmetric Markov Switching Vector Error Correction model is presented, along with

the tests that are utilized for the analysis of the price mechanism. Section 3 demonstrates the

data set that is used for the empirical estimation. The empirical results of the estimation are

described in section 4. Finally, in section 5 the findings and the conclusions of the chapter are

presented.

56

4.2 Econometric Methodology

4.2.1 The asymmetric Markov Switching Vector Error Correction model

The aim of this chapter is the investigation of the price transmission mechanisms of

fresh potatoes, tomatoes and cucumbers between the producer and the consumer. In order to

accomplish this goal, first, it is necessary to examine the properties of the price series. As soon

as, the non-stationarity of the price series is confirmed then the existence of a long-run

relationship between the producer and the consumer is investigated with co-integration

analysis. The co-integration analysis is implemented as described in Juselius (2006). A

complete discussion of the derivation of the theory behind this co-integration analysis takes

place in Johansen (1996). According to the co-integration analysis, a vector error correction

model is estimated and then by the use of a likelihood ratio test, which is called trace test, the

number of the co-integrating relationships is determined. Let ( )t t t

Y P C , where ( )t t

P C is

the logged producer (consumer) prices, be the 2-dimensional vector of the variables of interest,

where 1t T and T the sample size. The 2-dimensional vector autoregressive model would

be:

1 1, (0, ) (4.1)

t t p t p t t tA A D

where: 1t t p

are the lagged variables of t

, and t

D is the vector of deterministic variables.

Equation (1) is reformulated in the error correction form, in order to distinguish between

stationarity created by linear relationships between producer and consumer prices and

stationarity created by differencing.

1 1 1 1 1... (4.2)

t t t p t k t tD

where: 1t t p

are the differenced lagged values of t

, 1

p

i ni

A I

, 1

p

i ii i

,

1

1

( ) (1 ) (1 )k

p ii

z z I z z

the characteristic polynomial of the process and '

.

57

Finally, the co-integrating relationship, which represents the long-run relationship between

producer and consumer prices, is given by the product of the matrices '

and 1t

. This co-

integration analysis is preferred to a simple OLS regression estimate as it takes into account the

error structure of the underlying process (Johansen, 1988b). Moreover, this procedure allows

for the incorporation of deterministic variables such as a constant, a linear trend or seasonal

dummy variables. Furthermore, the procedure can incorporate stationary stochastic variables

that are weakly exogenous or that can be excluded from the co-integrating space (Juselius,

2006).

In the next step, a Markov Switching Vector Error Correction (MSVEC) model is

formulated. The MSVEC model is formulated by letting ( )t t t

Y P C be the 2-

dimensional vector of the variable of interest where 1t T and T the sample size and

( 1, , )t

s i i M is the M state unobserved variable that follows a first-order ergodic Markov

chain. The number of the states of the unobserved variable is countable. The 2nd order MSVEC

model will be written as:

0 1 1 1( ) ( ) ( ) ( ) ,

(4.3)(0, ( ))

t t t t p t t p t t t

t t

A s i A s i A s i B s i ect u

u s i

where: 0( ) ( )

t p tA s A s are the state dependent coefficient matrices, ( )

tB s the state dependent

coefficient matrix of the error correction term 1t

ect

and t

u is the state dependent error term of

the equation. The Markov chain undergoes transitions from one state to another with a specific

probability, the transition probability. The matrix of transition probabilities is given below:

11 1

1

M

M MM

p pP

p p

with 1

1M

kjj

p

and 0, , {1, , } (4.4)kj

p k j M

In the case of this chapter, the price transmission mechanism is modeled with the following

MSVEC model:

0 1, 2, 11 1

( ) ( ) ( ) ( ) ( )

K Kp p p p p

t t k t t k k t t k t t t tk k

P s i s i P a s i C s i ect u s i

0 2, 1, 11 1

(4.5)

( ) ( ) ( ) ( ) ( )K Kc c c c c

t t k t t k k t t k t t t tk k

C s i s i C a s i P s i ect u s i

58

where: ( )t t

P C are the first differences of the producer (consumer) price at period t ,

0 0( ) ( ( ))

p c

t ts i s i is the state dependent intercept term of the producer (consumer)

equation, ( )t k t k

P C

are the lags of the producer (consumer) price, 1t

ect

is the error correction

term at period 1t and ( ) ( ( ))p c

t t t tu s i u s i is the state dependent error term of the estimated

equation of the producer (consumer). In this chapter, each term of the vector error correction

model is considered to be affected by the state variable as it is expected that the economic actors

treat all the variables according to the state of the world that they are in. So, each of the variables

is allowed to vary between the states and none of the variables is kept fixed.

In order to investigate the presence of asymmetries in the price transmission mechanism,

equation (5) should be allowed to take into consideration the different price adjustments that

might occur according to whether prices are increasing or decreasing. This is done by including

in equation (5) dummy variables in the following way: 1 if ΔP, ΔC, ect 0 0 otherwiset

D and

1 if ΔP, ΔC, ect 0 0 otherwiset

D . In the data set there are no observations with zero value that is why

only inequalities are used. So, the Asymmetric Markov Switching Vector Error Correction

model takes the form:

( ) ( ) ( )

0 1, 1, 2,1 1 1

( ) ( ) ( )

2, 1 11

( ) ( ) ( ) ( )

( ) ( ) ( ) ( )

K k Kp p p p

t t k t t t k k t t t k k t t t kk k k

K p p p p

k t t t k t t t t t t t tk

P s i s i D P s i D P a s i D C

a s i D C s i D ect s i D ect u s i

( ) ( ) ( )

0 2, 2, 1,1 1

(4.6)

( ) ( ) ( )K Kc c c c

t t k t t t k k t t t k kk k

C s i a s i D C a s i D C

1

( ) ( ) ( )

1, 1 11

( )

( ) ( ) ( ) ( )

K

t t t kk

K c c c c

k t t t k t t t t t t t tk

s i D P

s i D P s i D ect s i D ect u s i

From now on, the terms 1

, ,t t k t t k t t

D P D C D ect

will be referred to as

1, ,

t k t k tP C ect

and

the terms 1

, ,t t k t t k t t

D P D C D ect

as

1, ,

t k t k tP C ect

for ease of notation. The equations of

the producer and the consumer are simultaneously estimated, as a system, with the use of the

Expectation Maximization algorithm which uses Maximum Likelihood for the estimations of

the coefficients. The algorithm was developed by Dempster et al. (1977) and was adapted to

the MSVEC framework by Krolzig (1996). Apart from the estimation of the parameters of the

MSVEC model this procedure gives the filtered and smoothed probability of the price

mechanism being in state i at period t . The filtered probabilities are estimated based on

59

information up to the previous period 1t whereas the smoothed probabilities are estimated

based on information that are available for the whole period.

4.2.2 The analysis of the error correction model

In order to identify whether a price change will cause a transitory or a permanent

deviation from the long-run equilibrium, the direction of the short-run impact of the price

change is contrasted with its long-run impact for each state of the price mechanisms. Thus, let

us assume that, according to model (6), a consumer price increase ( C )t k

occurs which is

equally distributed among the k lags while everything else is held constant. Moreover, let us

note that an increase in the consumer price ( C )t k

or a decrease in the producer price ( )

t kP

results in a positive deviation from the long-run 1

( )t

ect

, whereas a producer price increase

( )t k

P

as well as a consumer price decrease ( )

t kC

lead to a negative long-run deviation

1( )

tect

. As a result, according to the producer’s equation, if the sign of

p( )

2,1

( )K

k tk

s i

is identical

to the sign of ( )

( )p

ts i

, this consumer price increase results in a permanent (P) deviation

from the long-run equilibrium. Otherwise, if the sign of the coefficient of the cumulative effect

is opposite to the sign of the coefficient of the error correction term, the long-run deviation is

transitory (T). A similar analysis takes place for a consumer decrease as well as for a producer

price increase and decrease for both equations.

Next, the existence of asymmetry is tested in the short and long run for each state of the

price mechanisms. For the short-run, two measures are used: the in-sample measure of a Wald

test and the out-of-sample measure of impulse responses after a first-period shock. The long-

run asymmetries are tested with a Wald test as well. Table 4.1 presents the in-sample

asymmetric tests for the short- and the long-run. More specifically, in the short run, asymmetries

are tested on the grounds that the positive cumulative effects of the lagged prices of the producer

or the consumer might have a different impact on the current prices of the producer and the

consumer than the negative ones. Similarly, in the long run, it is tested whether a positive or

negative deviation from the equilibrium affects the current prices differently. These tests reveal

whether the producer or consumer responds more intensively and/or quickly to a price increase

or a price decrease. In the short run, the asymmetric response of the producer (consumer) to a

consumer (producer) price change is given by the distributed asymmetric lag effect, while the

asymmetric effect of the producer (consumer) price on its own price is given by the

60

autoregressive asymmetric lag effect. In the long run, the asymmetric tests show whether the

long-run equilibrium is adjusted more intensively and/or quickly to positive or negative

deviations, as is presented in the third column of table 4.1.

Table 4.1 The in-sample asymmetric tests

:o

H Null Hypothesis of Symmetry

Short-Run Long-Run

Equations Distributed Lag Effect Autoregressive Lag Effect

Producer /

tP

( ) ( )

2, 2,

1 1

( ) ( )K k

p p

k t k t

k k

s i s i

p( ) p( )

1, 1,

1 1

( ) ( )K k

k t k t

k k

s i s i

( ) ( )

( ) ( )p p

t ts i s i

Consumer /

tC

c( ) c( )

1, 1,

1 1

( ) ( )K k

k t k t

k k

s i s i

c( ) c( )

2, 2,

1 1

( ) ( )K k

k t k t

k k

s i s i

( ) ( )

( ) ( )c c

t ts i s i

On the other hand, the out-of-sample asymmetric measure assesses the asymmetric response of

the producer and the consumer to a particular shock i

as the sum of the impulse response for

this particular shock and the impulse response for the same shock but with the opposite sign.

That is: ( ) ( )i i

asymmetry irf irf , where ( )irf is the impulse response with a

positive shock and ( )irf is the impulse response with a negative shock. This measure has

the advantage of showing the persistence of a shock while taking into consideration the impact

of the shock on the other variables in the model. However, if the duration of a state of the price

mechanism of a product is short, then the impulse responses cannot model properly the effect

of a shock to the producer and the consumer price (Enders, 2010). Therefore, the out-of-sample

asymmetric measure is not calculated and thus it is not presented for such states.

Finally, the tests of weak exogeneity (Johansen, 1992) and Granger causality (Granger,

1969) are conducted for each state of the price mechanisms by the use of a Wald test. The weak

exogeneity of the producer (consumer) price implies that the producer (consumer) does not

adjust to long-run deviations but may still react to lagged changes in the consumer (producer)

price in the short run. Therefore, Granger causality tests are utilized, which show whether the

producer (consumer) still reacts to lagged price changes of the consumer (producer). It should

be noted that the existence of weak exogeneity might imply a possible attempt by the producer

(consumer) to exert influence over the pricing behavior of the consumer (producer) (Engle et

al., 1983). Moreover, since the feedback between the producer price and the consumer price

depends both on the lagged changes in the producer (consumer) price and on the coefficients

61

of the error correction terms, a joint significance test is utilized. In a notion, the joint

significance test represents a simplistic Granger causality test under the vector error correction

framework. Table 4.2 presents the null hypothesis (0

H ) of the three tests. More specifically, the

Granger causality test examines whether the lagged changes of the producer, , 1,...,t k

P k K

,

(consumer, , 1,...,t k

C k K

) improve the forecasting of the consumer (producer) price. The

weak exogeneity test checks whether the coefficients of the error correction terms ( ( ))t

s i

are statistically insignificant. Finally, the joint significance test checks whether the lagged

values of the producer , 1,...,t k

P k K

(consumer , 1,...,t k

C k K

) together with the error

correction term, 1t

ect

, of each equation improve the forecasting of the consumer (producer)

price.

Table 4.2 The tests of Granger Causality, Weak Exogeneity and Joint Significance

Granger Causality Weak Exogeneity

Producer c( ) c( ) c( ) c( )

1,1 1,k 1,1 1,( ) ... ( ) ( ) ... ( ) 0t t t k ts i s i s i s i

( ) ( )

( ) ( ) 0p p

t ts i s i

Consumer ( ) ( ) ( ) ( )

2,1 2, 2,1 2,( ) ... ( ) ( ) ... ( ) 0p p p p

t k t t k ts i s i s i s i

( ) ( )

( ) ( ) 0c c

t ts i s i

Joint Significance

Producer c( ) c( ) c( ) c( ) c( ) c( )

1,1 1,k 1,1 1,( ) ... ( ) ( ) ... ( ) ( ) ( ) 0t t t k t t ts i s i s i s i s i s i

Consumer ( ) ( ) ( ) ( ) p( ) p( )

2,1 2, 2,1 2,( ) ... ( ) ( ) ... ( ) ( ) ( ) 0p p p p

t k t t k t t ts i s i s i s i s i s i

4.3 Data

The national averages of the monthly prices of fresh potatoes, tomatoes and cucumbers

for the producer and the consumer are used for the estimation procedure. The prices are

constructed by monthly price indices that are available by the Hellenic Statistical Authority.

The producer prices are represented by the Producer Price Index of Agricultural Output, while

the consumer prices are represented by the Consumer Price Index. The price indices of producer

are transformed into prices by utilizing the annual producer prices, which are published by the

Hellenic Ministry of Rural Development and Food. In the case of the consumer, the weekly

prices are used, which are recorded by the Price Observatory of the Hellenic Ministry of

Development. The prices of fresh tomatoes and cucumbers cover the period from January 1995

to April 2012. The prices of fresh potatoes extend from January 1991 to April 2012. All prices

62

are converted into their natural logarithms and they are nominal. Table 4.3 shows the descriptive

statistics of the prices of the three products.

Table 4.3 The descriptive statistics of the prices

Producer Prices Consumer Prices

Mean Deviation Skewness Kurtosis Mean Deviation Skewness Kurtosis

Potato 0.283 0.095 -0.112 2.376 0.505 0.150 -0.160 2.190

Tomato 0.618 0.191 0.431 2.752 1.198 0.312 0.072 2.612

Cucumber 0.480 0.165 1.007 3.747 1.175 0.309 0.768 4.604

Figure 4.1, below, presents the prices of potatoes, tomatoes and cucumbers for the producer as

well as for the consumer.

Figure 4.1 The prices of fresh potatoes, tomatoes and cucumbers

potatoes tomatoes

cucumbers

63

As figure 4.1 shows, the prices of the three products are highly volatile. Moreover, each of the

products follows its own price path during the period that is depicted. However, for all three

products it is evident that the producer and the consumer prices follow its pair quite closely

indicating the presence of co-integration. Moreover, table 4.3 shows that, the highest average

price of the three products is observed for tomatoes whereas the lowest is observed for potatoes.

Finally, it is observed that the average price of tomatoes and cucumbers doubles by the time it

reaches consumers.

4.4 Empirical Results

4.4.1 Unit Roots and Co-integration Analysis

Prior to the co-integration analysis, unit root tests are performed so that the stationarity

of the producer and the consumer price series can be tested. The unit root tests used are the

Augmented Dickey Fuller (ADF) test (Dickey and Fuller, 1979), the Phillips-Perron (PP) test

(Phillips and Perron, 1998) and the Kwiatkowski et al. (KPSS) test (Kwiatkowski et al., 2002).

For potatoes, the tests showed that the natural logarithms of the producer and the consumer

prices are non-stationary, whereas the first differences of prices are stationary. For tomatoes,

the ADF test and the PP test showed that the natural logarithms of the producer and the

consumer prices were stationary. In the same fashion, the PP test favored the stationarity of

cucumbers prices. However, the KPSS test showed that the natural logarithms of the prices of

both products were non-stationary while the first differences of the prices were stationary.

Hence, the results of the KPSS test for the natural logarithms of the prices of the three products

present sufficient evidence in favor of non-stationarity leading to the next step of co-integration

analysis. The results of the KPSS test for the three products are presented in table 4.4.

Table 4.4 The results of the KPSS unit root test for potatoes, tomatoes and cucumbers

Potato Tomato Cucumber

lnt

P 1.780 (0.463) 0.289 (0.146) 0.261 (0.146)

lnt

C 1.758 (0.463) 1.708 (0.463) 1.490 (0.463)

lnt

P 0.031 (0.463) 0.074 (0.463) 0.065 (0.463)

lnt

C 0.033 (0.463) 0.064 (0.463) 0.058 (0.463)

In the parenthesis 5% critical values are reported.

64

Before carrying out the co-integration rank test, a vector error correction model is estimated for

each product. A 2 lag model with 12 centered seasonal dummies is selected for potatoes and a

6 lag model with again 12 centered seasonal dummies is chosen for tomatoes and cucumbers.

Then, the number of co-integrating relationships between the producer and the consumer prices

is defined by the trace test. The implementation of the test leads to the conclusion that one long-

run relationship between the producer and the consumer prices exists for each of the products

under consideration, as shown in table 4.5.

Table 4.5 The results of Johansen’s Trace Test

number of co-integrating

relationships Potato Tomato Cucumber

0 0.000 0.000 0.001

1 0.194 0.118 0.064

P-values are reported.

The optimal model for each product was selected by the information criteria of Schwarz (SIC)

and Hannan – Quinn (HQIC) as well as with the implementation of likelihood ratio tests.

Moreover, mis-specification tests for autocorrelation, heteroskedasticity and normality of the

error term were used. The results of the co-integration analysis for the three products are given

by table 4.6 where tect represents the long-run relationship.

Table 4.6 The results of the co-integration analysis

Potato Tomato Cucumber

tect

(-8.202) (-22.439)

ln 0.447 0.882ln

t tC P

(-14.219) (-12.077)

ln 0.656 0.961ln

t tC P (-11.883) (-10.307)

ln 0.787 0.837ln

t tC P

Number of lags 2 6 6

SIC -9.334 -7.507 -7.241

HQIC -9.601 -7.975 -7.709

LM autocorrelation (1) 0.115 0.962 0.216

LM autocorrelation (4) 0.779 0.607 0.350

LM autocorrelation (6) 0.000 0.672 0.447

65

LM autocorrelation (12) 0.000 0.289 0.000

Normality 0.000 0.970 0.435

LM heteroskedasticity (1) 0.846 0.670 1.000

LM heteroskedasticity (4) 0.120 0.243 0.944

LM heteroskedasticity (6) 0.000 0.370 0.712

LM heteroskedasticity (12) 0.000 0.156 0.170

In parenthesis t-values are reported. For LM and normality tests’ p-values are reported. The number of the lags of

the LM tests is reported.

As table 4.6 reveals, the coefficients of the co-integrating vectors are significant. Moreover, the

analysis of the residuals of the three models shows that the vector error correction models are

well specified. Especially, for the case of tomato and cucumber there are not any signs of

autocorrelation or heteroskedasticity and they are normally distributed. On the other hand, for

potato there are some signs of autocorrelation and heteroskedasticity while the null hypothesis

of normality is rejected.

4.4.2 The Asymmetric Markov Switching Vector Error Correction Model

In the next step of the empirical analysis, for each of the three products, an asymmetric

Markov Switching Vector Error Correction model is estimated. The best model for potatoes,

tomatoes and cucumbers price transmission mechanisms is chosen by the information criteria

of Akaike (AIC) and Hannan – Quinn (HQIC) as well as by likelihood ratio tests.

The model that is selected for potatoes consists of 4 lags and is characterized by 2 states.

As shown in table 4.7, state 1 is characterized by higher variance ( ) for both the producer and

the consumer equations compared with state 2; therefore, it is considered to be the high-

volatility state. The average duration ( )d of the-high volatility state (state 1) is 8.3 months,

while the average duration of the low-volatility state (state 2) is 4 months. The duration of each

state is calculated by 1(1 )d p , where p is the probability of the price mechanism remaining

in the same state. Additionally, the number of observations ( )obs for each state with a

probability higher than 0.5 is reported. In the high-volatility state (state 1), there are 166

observations, whereas in the low-volatility state (state 2), there are 85 observations. Finally, the

covariance ( cov ) of the producer and consumer prices is reported.

66

Table 4.7 The asymmetric MSVEC estimation results: Potatoes

State 1 State 2

Producer Consumer Producer Consumer

variable Coefficient Coefficient Coefficient Coefficient

constant 0.009 (0.757) 0.027 (0.070) 0.001 (0.914) -0.011 (0.032)

1tP

0.144 (0.270) 0.023 (0.787) 0.085 (0.729) 0.109 (0.382)

1tP

0.535 (0.03) 0.380 (0.001) 1.336 (0.000) 0.389 (0.000)

2tP

0.236 (0.139) 0.010 (0.935) -0.084 (0.241) 0.006 (0.881)

2tP

0.180 (0.263) -0.038 (0.739) 0.188 (0.137) 0.170 (0.020)

3tP

0.127 (0.374) -0.094 (0.282) -0.117 (0.064) 0.019 (0.577)

3tP

0.041 (0.783) 0.043 (0.650) -0.233 (0.129) -0.317 (0.000)

4tP

0.024 (0.869) -0.069 (0.337) 0.294 (0.000) 0.222 (0.000)

4tP

0.100 (0.628) -0.001 (0.993) -0.163 (0.228) -0.160 (0.047)

1tC

-0.331 (0.197) -0.093 (0.580) 0.431 (0.066) 0.264 (0.043)

1tC

0.025 (0.931) 0.502 (0.002) -1.251 (0.000) -0.060 (0.518)

2tC

-0.742 (0.007) -0.172 (0.327) 0.475 (0.002) 0.059 (0.499)

2tC

-0.443 (0.144) -0.407 (0.016) -0.045 (0.766) -0.133 (0.131)

3tC

0.463 (0.113) 0.217 (0.186) -0.232 (0.085) -0.113 (0.143)

3tC

-0.107 (0.740) 0.099 (0.616) 0.352 (0.082) 0.278 (0.009)

4tC

-0.442 (0.079) -0.234 (0.123) -0.271 (0.167) 0.058 (0.547)

4tC

-0.198 (0.458) -0.160 (0.330) 0.346 (0.011) 0.081 (0.310)

1tect

0.564 (0.001) 0.089 (0.356) 3.141 (0.000) 0.283 (0.000)

1tect

0.399 (0.031) -0.166 (0.181) -0.075 (0.632) -0.203 (0.018)

0.017 (0.000) 0.006 (0.000) 0.001 (0.000) 0.0004 (0.000)

cov 0.006 (0.000) 0.0005 (0.000)

d 8.333 4.048

obs 166 85

p 0.880 0.753

AIC -9.469

67

HQIC -10.461

In the parenthesis p-values are reported.

Next, the nature of price changes for fresh potatoes is examined. As shown in table 4.7a, when

the price transmission mechanism of potatoes is in the high-volatility state (state 1), the

deviations from the long-run equilibrium are transitory (T). More specifically, an increase in

the consumer price ( )t k

C

affects negatively the producer and consumer prices in the short

run. That is, the cumulative effects of the producer (-1.052) and the consumer (-0.282) in

conjunction with the positive price change (t n

C

) lead to a decrease in the producer and

consumer prices ( ,t t

P C ). In the long run, the increase in the consumer price causes a positive

deviation from the equilibrium (1t

ect

), which results in an increase in the prices of the producer

and the consumer ( ,t t

P C ) when contrasted with the long-run coefficients of the producer

(0.564) and the consumer (0.399). Thus, an increase in the consumer price is transitory since

the direction of its impact in the long run is opposite to its direction in the short run. This is also

the case for a decrease in the consumer price as well as for an increase or decrease in the

producer price.

Table 4.7a The permanent (P) and transitory (T) effects for potatoes in state 1

Short-Run Long-Run

t kC

t kC

t kP

t kP

1tect

1tect

tP -1.052 / (T) -0.725 / (T) 0.531 / (T) 0.856 / (T) 0.564 0.399

tC -0.282 / (T) 0.034 / (T) -0.130 / (T) 0.384 / (T) 0.089 -0.166

The in-sample asymmetric tests for potatoes in the high-volatility state (state 1) are presented

in table 4.7b. An inspection of the table reveals that, in the short run, the producer and the

consumer respond symmetrically to producer or consumer price increases or decreases, with

the exception of an asymmetric response of the consumer to a producer price change. In this

case, the consumer reacts more to a producer price decrease than to an increase, so the

asymmetric response is considered to be a negative one. Regarding the long-run, the responses

of the producer and the consumer to deviations from the long-run equilibrium is symmetric,

irrespective of their sign.

Table 4.7b The in-sample asymmetric tests for potatoes in state 1

68

Short-Run Long-Run

Distributed Lag Effect Autoregressive Lag Effect

Producer 0.157 (0.691) S 0.917 (0.338) S 0.381 (0.536) S

Consumer 6.285 (0.012) - 0.560 (0.454) S 2.385 (0.122) S

In the parenthesis p-values are reported, S denotes symmetry, + denotes positive asymmetry, - denotes negative

asymmetry.

The results of the Granger causality and weak exogeneity tests for potatoes, in the high-

volatility state (state 1), are shown in table 4.7c. The tests reveal that the producer adjusts to

deviations from the equilibrium in the long run and also reacts to the lagged price changes of

the consumer. On the other hand, the consumer does not adjust to deviations in the long run.

However, he reacts to the lagged price changes of the producer in the short run. The joint

significance test verifies that there is a feedback relationship between the producer and the

consumer in the high-volatility state.

Table 4.7c The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

potatoes in state 1.

Granger Causality Weak Exogeneity Joint Significance

Producer 26.419 (0.000) 21 (0.000) 32.986 (0.000)

Consumer 28.603 (0.000) 2 (0.300) 55.623 (0.000)

In the parenthesis p-values are reported.

The out-of-sample measure of asymmetric responses for potatoes in state 1 is presented in figure

4.2. A shock to the producer price lasts for eight months before both the producer and the

consumer reach convergence. During this period, the producer exhibits negative asymmetric

behavior for the first three months whereas for the rest of the period this reaction is positively

asymmetric. The consumer exhibits similar behavior; however, the turning point between a

negative and a positive asymmetric response is prevalent during the fifth month. Thus, the

reaction of the producer and the consumer to a shock in the producer price can be regarded as

symmetric in the long run with a memory effect in the short run. On the contrary, when a shock

to the consumer price occurs, it lasts for almost a year before reaching a steady state for both

the producer and the consumer. More specifically, the producer and the consumer react in a

similar way to the shock with a cyclical asymmetric pattern before the equilibrium; however,

in the long run, their behavior is symmetric with memory effects in the short run.

69

Figure 4.2 The impulses responses of the producer and the consumer price along with the out-

of-sample asymmetric responses for potatoes in state 1.

P is producer. C is consumer. (+1) denotes a positive shock. (-1) denotes a negative shock.

On the other hand, when the price mechanism is in the low-volatility state (state 2), the

deviations from the long-run equilibrium are mostly permanent, as shown in table 4.7d. More

specifically, for the producer, a permanent deviation occurs when the consumer price changes

or when the producer price increases. On the other hand, permanent deviations for the consumer

take place when the producer or the consumer price increases.

Table 4.7d The permanent (P) and transitory (T) effects for potatoes in state 2

Short-Run Long-Run

70

t kC

t kC

t kP

t kP

1tect

1tect

tP 0.403 / (P) -0.598 / (P) 0.178 / (P) 1.128 / (T) 3.141 -0.075

tC 0.268 / (P) 0.166 / (T) 0.356 / (P) 0.082 / (T) 0.283 -0.203

The in-sample asymmetric tests for potatoes in the low-volatility state (state 2) are provided by

table 4.7e. The results show that, in the short run, the producer is the one who responds

asymmetrically to price changes. More specifically, the producer reacts more to price decreases

than to price increases, so the asymmetric effect is negative. On the other hand, the consumer

responds symmetrically to any price change. In the long run, both the producer and the

consumer respond asymmetrically to deviations from the equilibrium. More specifically, the

asymmetry is positive; thus, the producer and the consumer respond more intensively to positive

deviations from the long run, that is, when the margin of the retailer is stretched.

Table 4.7e The in-sample asymmetric tests for potatoes in state 2

Short-Run Long-Run

Distributed Lag Effect Autoregressive Lag Effect

Producer 6.185 (0.012) - 8.176 (0.004) - 242.8 (0.000) +

Consumer 2.190 (0.138) S 0.211 (0.645) S 15.11 (0.000) +

In the parenthesis p-values are reported, S denotes symmetry, + denotes positive asymmetry, - denotes negative

asymmetry.

The results of the Granger causality and weak exogeneity tests for the low-volatility state (state

2) are presented in table 4.7f. The table shows that the producer adjusts to deviations from the

long-run and also reacts to the lagged price changes in the consumer price. In the same fashion,

the consumer adjusts to the long-run equilibrium and reacts to short-run price changes of the

producer. This feedback relationship between the producer and the consumer is also verified by

the joint significance test.

Table 4.7f The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

potato in state 2

Granger Causality Weak Exogeneity Joint Significance

Producer 111.559 (0.000) 583.130 (0.000) 138.457 (0.000)

Consumer 127.909 (0.000) 16.538 (0.000) 1063.592 (0.000)

In the parenthesis p-values are reported.

71

The out-of-sample measure of asymmetric response for potatoes in the low-volatility state (state

2) is not calculated, since its short duration (four months) results in impulse responses that

cannot model properly the effect of a shock to the producer and the consumer price.

Regarding tomatoes, the estimated model consists of 3 lags and is characterized by 2

states, as observed in table 4.8. State 1 is characterized as the low-volatility state whereas state

2 as the high-volatility state. In the low-volatility state, the average duration is 1.046 months

while having 23 observations. In the high-volatility state, the average duration is 8.547 months

and it has 181 observations.

Table 4.8 The asymmetric MSVEC estimation results: Tomatoes

State 1 State 2

Producer Consumer Producer Consumer

variable Coefficient Coefficient Coefficient Coefficient

constant 0.479 (0.000) 0.132 (0.000) 0.072 (0.010) 0.067 (0.001)

1tP

0.310 (0.158) 0.560 (0.000) -0.139 (0.241) -0.215 (0.012)

1tP

0.229 (0.399) 0.505 (0.001) 0.394 (0.000) 0.015 (0.855)

2tP

-0.980 (0.000) -0.047 (0.691) -0.432 (0.000) -0.449 (0.000)

2tP

-0.450 (0.172) -0.232 (0.220) 0.069 (0.573) -0.087 (0.317)

3tP

0.278 (0.192) -0.009 (0.941) 0.161 (0.233) -0.132 (0.197)

3tP

0.491 (0.108) 0.777 (0.000) -0.182 (0.160) -0.022 (0.814)

1tC

-0.435 (0.328) -0.410 (0.109) -0.249 (0.086) 0.047 (0.658)

1tC

-0.464 (0.221) -0.357 (0.105) -0.053 (0.730) 0.304 (0.017)

2tC

0.650 (0.013) 0.954 (0.000) -0.106 (0.488) -0.147 (0.149)

2tC

2.542 (0.000) 0.890 (0.004) -0.130 (0.409) 0.249 (0.019)

3tC

-2.111 (0.000) -0.972 (0.000) -0.074 (0.636) 0.114 (0.347)

3tC

0.079 (0.783) -0.279 (0.094) -0.109 (0.509) -0.242 (0.042)

1tect

-0.463 (0.030) -0.677 (0.000) 0.110 (0.388) -0.101(0.282)

1tect

0.320 (0.140) -0.578 (0.000) 0.515 (0.000) -0.071 (0.530)

0.003 (0.003) 0.001 (0.002) 0.028 (0.000) 0.015 (0.000)

72

cov 0.002 (0.003) 0.016 (0.000)

d 1.046 8.547

obs 23 181

p 0.044 0.883

AIC -2.589

HQIC -2.142

In the parenthesis p-values are reported.

The permanent and transitory effects for fresh tomatoes in the low-volatility state (state 1) are

shown in table 4.8a. When the price transmission mechanism is in the low-volatility state, the

deviations from the long-run equilibrium are transitory and permanent. More specifically, an

increase in the producer price results in a permanent increase in the consumer price. In the same

fashion, a decrease in the consumer price leads to a permanent decrease in the producer price

whereas a decrease in the producer price has as an effect the permanent decrease in the producer

and the consumer price.

Table 4.8a The permanent (P) and transitory (T) effects for tomatoes in state 1

Short-Run Long-Run

t kC

t kC

t kP

t kP

1tect

1tect

tP 5.388 / (T) 14.125 / (P) 0.960 / (T) 0.404 / (P) -0.463 0.320

tC 0.830 / (T) 0.584 / (T) 4.719 / (P) 18.259 / (P) -0.677 -0.578

Next, the in-sample asymmetric tests for tomatoes, in the low-volatility state (state 1), are

presented in table 4.8b. In the short run, the producer and the consumer respond symmetrically

to price changes with the exception of the producer when a consumer price change takes place.

The asymmetric reaction of the producer is negative, as the producer reacts more to a consumer

price decrease. In the long run, again, the producer is the one who responds asymmetrically to

deviations from the equilibrium. The producer responds more to positive deviations, that is,

when the margin of the consumer stretches.

Table 4.8b The in-sample asymmetric tests for tomatoes in state 1

Short-Run Long-Run

Distributed Lag Effect Autoregressive Lag Effect

73

Producer 15.239 (0.000) - 1.350 (0.245) S 6.415 (0.011) +

Consumer 2.720 (0.099) S 1.284 (0.257) S 0.308 (0.578) S

In the parenthesis p-values are reported, S denotes symmetry, + denotes positive asymmetry, - denotes negative

asymmetry.

Table 4.8c shows that the producer as well as the consumer adjust to deviations from the long-

run equilibrium and react to the lagged price changes of each other in the short run. Again, the

joint significance test reveals a feedback relationship between the producer and the consumer.

Table 4.8c The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

tomatoes in state 1.

Granger Causality Weak Exogeneity Joint Significance

Producer 74.644 (0.000) 6.680 (0.035) 368.551 (0.000)

Consumer 77.811 (0.000) 53.283 (0.000) 78.373 (0.000)

In the parenthesis p-values are reported.

The out-of-sample measure of asymmetric response for tomatoes in the low-volatility state

(state 1) is not calculated, due to the short duration (one month) of the state. Regarding the high-

volatility state (state 2) of fresh tomatoes price mechanism, the deviations from the long-run

equilibrium are permanent and transitory as shown in table 4.8d. More specifically, the

permanent price changes for the producer occur when the consumer price changes. On the other

hand, the consumer permanent price changes take place when there is a producer price change.

Table 4.8d The permanent (P) and transitory (T) effects for tomatoes in state 2

Short-Run Long-Run

t kC

t kC

t kP

t kP

1tect

1tect

tP 2.650 / (P) 1.566 / (P) 2.959 / (T) 2.507 / (T) 0.110 0.515

tC 0.005 / (T) 3.216 / (T) 21.561 / (P) 0.519 / (P) -0.101 -0.071

The in-sample asymmetric tests for tomatoes in the high volatility state (state 2) are presented

in table 4.8e. The results reveal that the producer is the one that causes asymmetries in the price

mechanism. In the short run, the producer price changes lead to positive asymmetric effects to

the producer and negative asymmetric effects to the consumer. Again, in the long run, the

74

producer reacts negatively asymmetric to deviations from the equilibrium whereas the

consumer response is symmetric.

Table 4.8e The in-sample asymmetric tests for tomatoes in state 2

Short-Run Long-Run

Distributed Lag Effect Autoregressive Lag Effect

Producer 0.137 (0.711) S 5.206 (0.022) + 3.104 (0.078) -

Consumer 9.705 (0.001) - 1.159 (0.281) S 0.031 (0.858) S

In the parenthesis p-values are reported, S denotes symmetry, + denotes positive asymmetry, - denotes negative

asymmetry.

Table 4.8f shows that the producer adjusts to the long-run equilibrium and to the lagged price

changes of the consumer, while at the same time, the consumer shows no adjustment to the

long-run equilibrium and no reaction to the lagged price changes of the producer. However, the

joint significance tests for the producer and the consumer reveal a feedback relationship.

Table 4.8f The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

tomatoes in state 2

Granger Causality Weak Exogeneity Joint Significance

Producer 50.353 (0.000) 15.287 (0.000) 50.564 (0.000)

Consumer 5.888 (0.435) 2.207 (0.331) 19.111 (0.014)

In the parenthesis p-values are reported.

The out-of-sample measure of asymmetric response for tomatoes in state 2 is presented in figure

4.3. The asymmetric response of the producer and the consumer to a producer price shock is

negatively asymmetric during the first three months and positively asymmetric for the next

three months with convergence being achieved at the sixth month. Next, the memory effect dies

out, leading to long-run symmetry. In contrast, in the case of a shock to the consumer price, the

producer demonstrates negative asymmetry during the first two and a half months whereas the

consumer responds negatively asymmetric for a period of three and a half months. For the rest

of the period until the convergence point (i.e. ninth month) the producer and the consumer

respond with a positive asymmetry resulting in long-run symmetry after the ninth month.

75

Figure 4.3 The impulses responses of the producer and the consumer price along with the out-

of-sample asymmetric responses for tomatoes in state 2.

P is producer. C is consumer. (+1) denotes a positive shock. (-1) denotes a negative shock.

Regarding cucumbers, the model of the price transmission mechanism consists of 3 lags

and 2 states. As shown in table 4.9, the average duration of state 1 is nearly 7.8 months and it

has 147 observations whereas the average duration of state 2 is almost 3.3 months and has 57

observations. The magnitude of the producer and the consumer variances is similar between the

two states; however, it can be observed that state 1 has similar characteristics to the high-

volatility state of potatoes and tomatoes price mechanisms. Accordingly, state 2 of cucumbers

has similar characteristics to the low-volatility state of potatoes and tomatoes.

Table 4.9 The asymmetric MSVEC estimation results: Cucumbers

76

State 1 State 2

Producer Consumer Producer Consumer

variable Coefficient Coefficient Coefficient Coefficient

constant -0.044 (0.315) 0.036 (0.186) 0.339 (0.000) 0.223 (0.001)

1tP

0.172 (0.372) -0.015 (0.896) -0.515 (0.162) -0.830 (0.001)

1tP

0.750 (0.000) 0.400 (0.000) 0.111 (0.777) 0.336 (0.246)

2tP

0.462 (0.017) -0.004 (0.974) -0.907 (0.010) -0.828 (0.001)

2tP

-0.090 (0.590) -0.178 (0.078) -0.400 (0.161) 0.491 (0.031)

3tP

-0.120 (0.538) -0.300 (0.014) -0.423 (0.108) 0.214 (0.292)

3tP

-0.020 (0.906) 0.009 (0.927) 0.389 (0.164) 0.509 (0.030)

1tC

-0.162 (0.493) 0.143 (0.317) -0.606 (0.135) 0.059 (0.841)

1tC

-0.478 (0.063) -0.079 (0.641) 0.470 (0.367) 0.154 (0.627)

2tC

-0.667 (0.003) -0.143 (0.312) 0.641 (0.093) 0.169 (0.562)

2tC

0.347 (0.140) 0.334 (0.020) -0.087 (0.845) -1.107 (0.003)

3tC

-0.336 (0.141) -0.027 (0.844) 0.377 (0.322) -0.821 (0.008)

3tC

-0.326 (0.164) -0.120 (0.384) -0.343 (0.372) -0.210 (0.490)

1tect

1.133 (0.000) -0.215 (0.197) -0.701 (0.175) 0.675 (0.108)

1tect

0.045 (0.839) -0.224 (0.106) 0.370 (0.272) -0.920 (0.001)

0.030 (0.000) 0.012 (0.000) 0.025 (0.000) 0.015 (0.000)

cov 0.015 (0.000) 0.014 (0.000)

d 7.751 3.289

obs 147 57

p 0.871 (0.000) 0.696 (0.000)

AIC -7.972

HQIC -8.962

In the parenthesis p-values are reported.

For the case of fresh cucumbers, table 4.9a shows that, if the price transmission mechanism is

at state 1, deviations from the long-run equilibrium are mostly transitory with the exception of

77

when the price margin of the consumer stretches either by an increase of consumer price or a

decrease of the producer price. Note that, this results in permanent consumer price changes.

Table 4.9a The permanent (P) and transitory (T) effects for cucumbers in state 1

Short-Run Long-Run

t kC

t kC

t kP

t kP

1tect

1tect

tP -1.165 / (T) -0.457 / (T) 0.514 / (T) 0.64 / (T) 1.133 0.045

tC -0.027 / (P) 0.135 / (T) -0.319 / (T) 0.231 / (P) -0.215 -0.224

The in-sample asymmetric tests for cucumbers in state 1 are presented in table 4.9b. In the short

run, the producer and the consumer respond symmetrically to price changes. In the long run,

the producer is the one who has a positively asymmetric reaction to deviations from the

equilibrium. That is, when consumer’s margin stretches, producer responds more than when it

is squeezed.

Table 4.9b The in-sample asymmetric tests for cucumbers in state 1

Short-Run Long-Run

Distributed Lag Effect Autoregressive Lag Effect

Producer 1.343 (0.246) S 0.055 (0.813) S 7.851 (0.005) +

Consumer 2.662 (0.102) S 0.174 (0.675) S 0.001 (0.967) S

In the parenthesis p-values are reported, S denotes symmetry, + denotes positive asymmetry, - denotes negative

asymmetry.

Next, table 4.9c shows that the producer adjusts to the long-run equilibrium and to the lagged

price changes of the consumer. However, the consumer does not adjust to deviations from the

equilibrium but he reacts to the lagged values of the producer in the short run. Furthermore, the

joint significance tests reveal again a feedback relationship between the two levels of the

market.

Table 4.9c The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

cucumbers in state 1.

Granger Causality Weak Exogeneity Joint Significance

Producer 28.379 (0.000) 18.954 (0.000) 37.282 (0.000)

78

Consumer 17.629 (0.007) 5.673 (0.058) 29.695 (0.000)

In the parenthesis p-values are reported.

Figure 4.4 presents the out-of-sample asymmetric response for cucumbers in state 1. The

graphic illustration shows that, a shock to the producer price causes both the producer and the

consumer to react with negative asymmetric effects for the first four months while for the next

six months, until the convergence point, the producers’ and consumers’ behavior is positively

asymmetric. Thus, a shock to the producer price lasts for ten months with a memory effect in

the short run. In the long run, it results in a positive asymmetric response for the producer and

in a symmetric response for the consumer. On the other hand, a shock to the consumer has

duration of twelve months for the producer and ten months for the consumer. More specifically,

both the producer and the consumer respond with a positive asymmetry during the first two

months and with a negative asymmetry for the next two months. For the rest of the period, the

producer keeps reacting with a negative asymmetry while the consumer exhibits a positive

asymmetric behavior.

Figure 4.4 The impulses responses of the producer and the consumer price along with the out-

of-sample asymmetric responses for cucumbers in state 1.

79

P is producer. C is consumer. (+1) denotes a positive shock. (-1) denotes a negative shock.

On the other hand, when the price mechanism is in state 2 deviations from the long-run

equilibrium are permanent and transitory, as shown in table 4.9d. In more detail, the producer

price is permanently affected when there is a price change at its own price or a consumer’s price

decrease. Moreover, a consumer price decrease also leads to a permanent effect for consumer.

Table 4.9d The permanent (P) and transitory (T) effects for cucumbers in state 2

Short-Run Long-Run

t kC

t kC

t kP

t kP

1tect

1tect

tP 0.412 / (T) 0.04 / (P) -1.845 / (P) 0.1 / (P) -0.701 0.370

tC -0.593 / (T) -1.163 / (T) -1.444 / (P) 1.336 / (T) 0.675 -0.920

Table 4.9e presents the results of the in-sample asymmetry tests for cucumbers in state 2.

According to the results, in the short run, the producer and the consumer react symmetrically

to consumer’s price changes. On the contrary, the producer and the consumer respond

asymmetrically to producer’s price changes and the asymmetry is positive. In the long run, once

again, the producer is the one who responds asymmetrically to deviations from the equilibrium

whereas the nature of the asymmetry is positive.

Table 4.9e The in-sample asymmetric tests for cucumbers in state 2

Short-Run Long-Run

Distributed Lag Effect Autoregressive Lag Effect

Producer 0.152 (0.696) S 3.196 (0.073) + 3.519 (0.060) +

Consumer 14.61 (0.000) + 0.797 (0.371) S 11.65 (0.000) S

In the parenthesis p-values are reported, S denotes symmetry, + denotes positive asymmetry, - denotes negative

asymmetry.

80

Next, table 4.9f presents the results of the weak exogeneity and Granger causality tests for state

2 of cucumbers price mechanism. The producer does not adjust to long-run equilibrium,

however, he reacts to lagged price changes of the consumer. On the other hand, the consumer

adjusts to long-run deviations but he does not react to lagged price changes of the producer.

The joint significance tests reveal that there is no feedback from the consumer prices to the

producer prices.

Table 4.9f The tests of the Granger Causality, Weak Exogeneity and Joint Significance for

cucumbers in state 2

Granger Causality Weak Exogeneity Joint Significance

Producer 41.533 (0.000) 3.599 (0.165) 43.085 (0.000)

Consumer 9.431 (0.150) 16.592 (0.000) 14.445 (0.070)

In the parenthesis p-values are reported.

The out-of-sample measure of asymmetric response for cucumbers in state 2 is also not

calculated, due to the short three-month duration of state 2 of cucumbers.

4.5 Conclusions

The present chapter examines the price transmission mechanisms of fresh potatoes,

tomatoes and cucumbers in Greece with a Markov Switching Vector Error Correction model

while allowing for asymmetric adjustments to positive and negative changes in the producer

and the consumer prices. The three products under investigation, besides their commonalities,

are also characterized by distinctive differences. Among the major common characteristics of

the three markets are: the self-sufficiency, the stable policy framework provided by the CAP,

the highly concentrated retailing sector and the low degree of processing. In addition, among

the minor common characteristics are the inputs used for the production process, the low

demand for substitute products such as prebaked potatoes or frozen vegetables and the small

percentage of imported quantities. On the other hand, the main sources of diversification among

the three markets are the regulative framework and the degree of perishability of the three

agricultural products. In particular, tomatoes and cucumbers are regulated by the CMO for fruits

and vegetables and they are highly perishable. In contrast, potatoes are more easily preserved,

however they are not organized under a common market. These diversifying market

81

characteristics are expected to determine the different patterns of price transmission in the three

markets.

Previous research (Reziti, 2005; Reziti and Panagopoulos, 2008) has found that the price

mechanisms of Greek agricultural products are characterized by long-run asymmetries and the

non-responsiveness of consumer price to deviations from the long-run equilibrium. However,

in the light of the findings of the present chapter it is evident that the price mechanisms switch

between two distinctive states, each one characterized by its own price dynamics between the

producer and the consumer. The analysis of the dynamics of the two different states of the price

mechanisms is based on four measures. The first one is the identification of the nature of the

price changes (i.e. permanent vs. transitory). The second is the examination of possible

asymmetries in the short and long run with in- and out-of-sample measures. The third one is the

response of the producer and the consumer to lagged changes of the prices, which is investigated

with Granger causality tests. Finally, the fourth examines their response to long-run deviations

from the equilibrium, which is assessed with weak exogeneity tests.

In particular, in state 1 of fresh potatoes’ price transmission mechanism, which lasts on

average for eight months, the producer and the consumer respond to lagged price changes of

each other in the short run while, in the long run, the consumer does not respond to deviations

from the equilibrium. Thus, the producer and the consumer are involved in a short-run feedback

price relationship. Moreover, the producer and the consumer react symmetrically to price

increases and decreases which are characterized as transitory. The symmetric and transitory

effects of the price mechanism in state 1 can be attributed to the feedback price relationship

between the producer and the consumer in the short run. However, the long-run pricing behavior

of consumer can be seen as an attempt to exert influence over producer to restore the price

equilibrium by adjusting his own price. Since, the consumer does not participate in the

restoration of the long run equilibrium for the duration of state 1, it can be supported that

consumer’s pricing behavior leads to the switch of the price mechanism to state 2. In state 2,

which lasts on average for four months, the response of the producer and the consumer to price

changes is asymmetric and they are characterized as permanent. Furthermore, the producer and

the consumer interact in the short and long run for the price formation. Since, the consumer

responds to deviations from the long-run in state 2, this change in his pricing behavior, can be

seen as the reason that reverts the price mechanism back to state 1 where symmetric and

transitory effects are the case. Fresh potatoes’ price transmission mechanism presents this

dynamics while not being under the influence of a regulative framework but being an

agricultural product that is easily preserved.

82

On the other hand, fresh cucumbers are regulated by the CMO for fruits and vegetables

and they are characterized by costly storing and transportation due to the perishability of the

product. Even though, these two cucumbers’ market characteristics are quite different from

potatoes, the price dynamics of their price mechanism is quite similar. More specifically, when

cucumbers price mechanism is in state 1, which has a duration of eight months, the producer

and the consumer respond to each other’s lagged prices in the short run while both respond to

deviations from the long-run equilibrium. Thus, it can be said that the producer and the

consumer are in a feedback price relationship in the short and long run. Moreover, price changes

are mostly transitory while both the producer and the consumer react symmetrically to them. In

state 2, for three and a half months, price changes are mostly permanent while the producer and

consumer respond asymmetrically to price increases and decreases. Furthermore, the producer

does not adjust to deviations from the long-run equilibrium but instead reacts to consumer

lagged price changes in the short run. The long-run pricing behavior of producer can be seen as

an attempt to exert influence over consumer to restore equilibrium. However, in the short run,

producer responds to consumer price changes. Therefore, it can be supported that the producer

and the consumer interact for the prices formation and in this way influence the price

mechanism to revert to state 1 where symmetric and transitory effects are the case.

Similarly to cucumbers, fresh tomato market is regulated by the CMO for fruits and

vegetables while the high perishability of the product results in costly preservation. However,

the price mechanism does not exhibit similar price dynamics to cucumber market. Particularly,

under state 2, which lasts for eight months, the producer reacts to consumer price changes in

the short run but the consumer is not responding to long-run deviations from the equilibrium.

Therefore, it can be said that the consumer attempts to exert influence on the pricing behavior

of the producer by allowing him to restore the long-run price equilibrium by adjusting his own

price. Moreover, the producer and the consumer respond asymmetrically to price changes. In

state 1 (one month duration), symmetries are evident while the producer and the consumer

interact in the short and long run. Furthermore, neither transitory nor permanent deviations are

prevalent in any of the two states. The analysis of the two states reveals that even though the

consumer does not interact with producers’ prices in state 2 since consumer does not react to

lagged producer prices and to long-run equilibrium deviations, the price mechanism switches

to state 1 where symmetric effects are prevalent for a short period. Therefore, it can be said that

the consumer seems to have a strong place in the tomato market.

The empirical analysis of fresh potato, tomato and cucumber prices with a stochastic

regime-switching model revealed the complicated nature of the price mechanisms of the three

83

products. More specifically, the price transmission mechanisms of fresh potatoes and

cucumbers follow a similar pattern in their response to price shocks in both states, depending

on the state of the economic environment, even though they are characterized by different

regulative frameworks and degree of perishability. Furthermore, the fresh tomato market

follows its own distinctive pattern of responses to price shocks, quite different from cucumbers,

even though they are both characterized by the same regulative framework and degree of

perishability. These results have policy implications for the Common Agricultural Policy in the

EU and especially for the effectiveness of the Common Market Organization for fruits and

vegetables. The apparent implication of this chapter is that the CMO for fruits and vegetables

does not affect agricultural products in the same way, even though they share common

characteristics like tomatoes and cucumbers. In particular, the cucumber market is found to be

more capable of transmitting price signals between the producer and the consumer and

absorbing price shocks. However, the fresh tomatoes market doesn’t seem to be able to absorb

price shocks efficiently since the consumer exerts influence on the producer’s pricing behavior.

Thus, European and domestic agricultural policy makers should take into consideration the

unique price dynamics of each agricultural product when designing policy schemes especially

when price mechanisms are governed by regime-switches. Finally, the empirical results of

potato market give further support to this conclusion since the price mechanism is able to

transmit price information in a timely and effective manner among the market participants even

though the market is unregulated.

84

5. Investigating the price volatility transmission mechanisms of selected fresh vegetable

chains in Greece

In chapter 5, the transmission of the volatility between the producer and the consumer

prices of fresh potatoes, tomatoes and cucumbers in Greece are modeled individually. The

transmission mechanism of the prices volatility is modelled by the BEKK model while

possible asymmetries are also included. The asymmetric effects are evaluated by

asymmetric BEKK models as well as volatility impulse responses. A new approach is

proposed for the implementation of the volatility impulse responses. The results show that

in the regulated markets of tomatoes and cucumbers, the producer is less vulnerable to

volatility shocks transmitted from the consumer whereas in the non-regulated potato

market, the producer faces significant spillover effects from the consumer. The chapter is

based on the working paper Rezitis and Pachis (2015b).

5.1 Introduction

In agricultural markets, one of the major concerns of producers and consumers is the

large and unanticipated price changes that are commonly known as volatility. Volatility and its

spillovers increase the uncertainty and risk in the market. Moreover, producers and consumers

tend to respond more to bad news than to good news, giving rise to concerns that economic

actors react asymmetrically to positive and negative volatility shocks. When the increased risk

aversion leads to suboptimal decisions while dealing with volatility is beyond market

participants’ capacity, it becomes an issue for policy response (Interagency Report 2011).

Therefore, price volatility and its asymmetric effects are always at the top of the agenda of

policy makers, practitioners and academics.

The aim of this chapter is to investigate the individual price volatility transmission

mechanisms of fresh potatoes, tomatoes and cucumbers in Greece. The volatility price

mechanisms are examined by the BEKK model. Possible asymmetries are thoroughly

investigated with volatility impulse responses and asymmetric BEKK models. In estimating

volatility impulse responses, the selection of a specific month during which a shock takes place

is necessary. Previous studies estimate volatility impulse responses either by selecting the

month of the shock based on their subjective judgment or by using methods that identify

deterministic structural breaks. However, in this chapter the selection process is carried out with

the use of a Markov Switching Vector Error Correction model (MSVEC). The MSVEC model

85

provides the months during which the volatility changes, that is, the price mechanism switches

from the high- to the low-volatility state. Apart from the methodological interest in investigating

price volatility with the proposed exposition, the results have important implications for the

implementation of agricultural policies since they reveal that unregulated markets (i.e. potatoes)

are more vulnerable to price volatility than regulated ones (i.e. tomatoes and cucumbers).

In agricultural economics, the modelling of the price volatility takes place with the

Autoregressive Conditionally Heteroscedastic (ARCH) model by Engle (1982) as well as the

Generalization of the ARCH model proposed by Bollereslev (1986). These models study

volatility per se, however, the understanding of the co-movements of price volatility takes place

with multivariate GARCH models. The Vectorised Conditionally Heteroscedastic (VECH)

model by Bollereslev et al. (1988) and its parsimonious version, the Diagonal VECH (DVECH),

were the first attempt to model the inter-dependencies of price volatilities. Next, Baba, Engle,

Kraft and Kroner (1991) proposed the BEKK model which was previously defined in Engle

and Kroner (1995) as another restricted version of the VECH model. Apart from the VECH

family of models, the co-movements of price volatility is studied by the conditional correlations

models. The simplest multivariate correlation model is the Constant Conditional Correlation

(CCC) model (Bollereslev 1990) while the generalization of CCC led to dynamic models like

Engle’s (2002) Dynamic Conditional Correlation model. The aforementioned multivariate

GARCH models take into consideration the facts that the unconditional distribution of prices is

much fatter-tailed than the normal one and that large residuals tend to cluster, however, they do

not account for probable asymmetric responses of economic actors to volatility shocks. These

shortcomings can be addressed either by GARCH models which incorporate asymmetric effects

or by estimating impulse responses that give a different response according to the nature of a

shock. Glosten et al. (1993) proposed a formulation that allowed the conditional variance to

respond differently to the past negative and positive innovations. This formulation is closely

related to the Threshold GARCH by Zakoian (1994) and to the Asymmetric GARCH by Engle

(1990). On the other hand, the impulse responses treat the volatility dynamics between the

variables of interest. Engle and Ng (1993) as well as Lin (1997) are two of the proposals that

have been made to trace the impact of shocks to volatility through time. More recently, Hafner

and Herwartz (2006) proposed an approach in the spirit of Koop et al. (1996) where the impulse

response depends on the history of the volatility apart from the shock.

The price volatility of the agricultural markets is mainly attributed to the variations of

the market fundamentals which are the supply and demand. According to Gilbert and Morgan

(2010), natural shocks caused by the weather conditions or diseases are the main factors that

86

affect production unexpectedly. On the other hand, consumption varies because of sudden

changes in income or, in the prices of substitutes or because of shifts in tastes. The extent to

which given production and consumption shocks translate into price volatility depends on

supply and demand elasticities which reflect the responsiveness of producers and consumers to

price changes. The supply and demand price elasticities are low in the short run (during the

crop period), especially if the stocks of agricultural products are low (Wright and Williams

1991; Deaton and Laroque 1992). Apart from market fundamentals, sudden shifts in policy

(Christiaensen 2009), input prices, exchange rates as well as speculation result in price volatility

(Interagency Report 2011; Gilbert and Morgan 2010). Furthermore, various sources of

asymmetric price volatility transmission have been suggested in the agricultural economics

literature. According to Assefa et al. (2013), the most common reasons for asymmetric

transmission are: market power of wholesalers and/or retailers (Serra 2011) and the lack of

contractual farming (Apergis and Rezitis 2003). Furthermore, Bénabou and Gertner (1993)

show that volatility affects the searching costs of consumers resulting in increased retailing

market power while Kroner and Ng (1998) show that a negative return shock (unexpected price

drop) would lead to higher subsequent volatility than would a positive return shock (unexpected

price increase) of the same magnitude.

As a member of the European Union (EU), Greece’s agricultural markets are regulated

by the Common Agricultural Policy (CAP) of the EU. More specifically, the tomato and

cucumber markets are organized according to the Common Market Organization (CMO) for

fruits and vegetables (European Parliament 2011). The CMO is market-oriented and the reforms

of 1996 and 2000 encouraged producers to join Producer Organizations (POs). However, for

fresh potatoes, there has been no common organization since the enactment of the CAP in 1962

(European Committee 2007). For the period under examination, the Greek fresh potato market

is characterized by increasing amounts of imports; however, the self-sufficiency of the country

amounts to about 80%. Similarly, the Greek tomato and cucumber markets are characterized

by small imported quantities resulting in self-sufficiency of 95% and 99%, respectively. The

near self-sufficiency of the three markets underlines the fact that the prices and subsequently

the volatility do not depend on fluctuations of the import prices. Moreover, the three products

account for 9% of the Greek Producer Price Index of Agricultural Output, indicating their

significance for producers. What is more, the above surveys reveal that Greece has the largest

vegetable consumption in the EU, while fresh vegetables and potatoes account for about 9% of

the Food Consumer Price Index, confirming the significance of fresh vegetables for Greek

consumers as well. However, the three products account for just 0.096% of the domestic

87

Manufacturing Producer Price Index, indicating the preference of Greek consumers for fresh

agricultural products. Up to 2000, the Greek supply chain of fresh agricultural products was

dominated by the central markets of Athens and Thessaloniki. In the following years, retailers’

concentration rose sharply, resulting in the decentralization of production distribution. Thus,

after 2000, the retailers and especially the supermarkets were the key players in the supply

chain, producing ever increasing concentration of distribution and demand (Reziti 2005; Reziti

and Panagopoulos 2008). The above analysis reveals that even though the reforms of the CAP

during the period under examination aimed to expose the domestic EU prices to the signals of

the world markets, the self-sufficiency of Greece resulted in immunity to the international prices

of vegetables. Therefore, the individual marketing characteristics of each product are expected

to show how volatility shocks are absorbed domestically.

The literature that studies the price volatility transmission mechanisms of vegetables is

dated. However, the research undertaken on grain and meat is substantial, although the focus is

on the United States. For instance, Natcher and Weaver (1999) study the transmission of the

price volatility in the US beef market. Furthermore, Yang et al. (2001) examine whether the

price volatility of major grains in the United States has decreased since the adoption of a

liberalization policy by the Government. Their results show that price volatility has risen, in

contrast to Crain and Lee (1996), who found that the volatility of wheat has decreased. On the

other hand, a limited number of studies investigate the price volatility transmission in European

food supply chains. The focus here is on meat as well. Rezitis and Stavropoulos (2011) study

the Greek broiler market, finding that the producer’s volatility is higher than the consumer’s.

Moreover, Apergis and Rezitis (2003) study the price volatility of the Greek food sector

between 1985 and 1999, finding that the spillovers of volatility from the consumer to the

producer are positive and significant, as are the own effects of the producer. Their main

conclusion is that, in the subsequent years, the reforms of the CAP increased the volatility for

Greek food producers as they adjusted to the market-oriented reforms. However, the admittance

of Greece into the European Monetary Union will decrease the volatility for the Greek economy

and consequently for agricultural production. Chapter 5 enriches the literature on price volatility

transmission in agricultural markets with a thorough investigation of the price mechanism of

three major Greek fresh vegetables. In particular, the results reveal that the Greek producers of

tomatoes and cucumbers are in a more advantageous position than the producers of potatoes

regarding their adjustment to volatility shocks. This finding adds to the debate on whether a

regulative framework results in increased or decreased price volatility by giving evidence for

the latter. Furthermore, the literature is enriched with a study that investigates asymmetric

88

effects with two different approaches that have never been used for the same data set. What is

more, these new findings contradict the results of previous research that supported the negative

impact of CAP reforms on producers of agricultural products. Finally, the implications of the

empirical results for economic theory are presented.

The rest of the article is structured as follows. In section 2, the methodology of the

estimation of the symmetric and the asymmetric BEKK model is presented, along with the

estimation of the volatility impulse responses. Section 3 demonstrates the data set that is used.

The results of the estimation of the symmetric and the asymmetric price volatility mechanisms

of the three products’ as well as the results of the impulse responses, are presented in section 4.

Finally, the findings of the chapter and their significance are presented and discussed in sections

5 and 6, respectively.

5.2 Methodology

In order to investigate the price volatility transmission mechanisms of the three markets

between producer and consumer the mean and the variance models are estimated with a quasi-

maximum likelihood procedure simultaneously. The specification of the mean model begins by

examining the properties of the price series. As soon as, the non-stationarity of the time series

is confirmed then the existence of a long-run relationship between the producer and the

consumer prices is investigated by co-integration analysis. The co-integration analysis is

implemented as is described in Juselius (2006). A complete discussion of the derivation of the

theory behind this co-integration analysis can be found in Johansen (1996). According to the

co-integration analysis, a vector error correction model is estimated and then by using a

likelihood ratio test, which is called trace test, the number of the co-integrating relationships is

determined. This co-integration analysis is preferred to a regression estimate as it takes into

account the error structure of the underlying process (Johansen 1988b). Thus, if it is assumed

that ( )t t t

Y P C where ( )t t

P C is the logged producer (consumer) prices of the 2-dimensional

vector of the variables of interest 1t T and T the sample size, then the vector error

correction model would be:

11

(5.1)k

t t j t j t tj

Y Y Y D u

89

where: t

Y is a 2 1 vector of the first differences of the logged producer and consumer prices

at period t , is a 2 2 matrix of long run coefficients, 1tY is a 2 1 vector of the error

correction terms, j is a k k matrix of the short run coefficients,

t jY is a 2 k matrix of k

lagged first differences of the logged producer and consumer prices, is a 2 M matrix of

coefficients where M is the number of seasonal dummies plus the constant, tD is a 1M vector

of the deterministic variables and tu is a 2 1 vector of the error terms. The vector of the error

terms is examined for GARCH effects with the GARCH-LM test proposed by Engle (1982)

which examines the possible clustering of large residuals.

The variance model that is utilized is the BEKK model:

' ' ' '

1 1 1(5.2)

t t t tH CC Au u A B H B

where: t

H is a 2 2 variance-covariance matrix, C is a lower triangular matrix of constants,

1tu is the 2 1 vector of the one lagged error terms, A is a 2 2 matrix of ARCH term

coefficients, B is a 2 2 matrix of GARCH term coefficients. The A coefficient matrix

represents the own and the cross recent shock transmission effects across the stages of the

chains. The B coefficient matrix represents the own and the cross past volatility transmission

effects across the stages of the chains. Thus, the A coefficient matrix represents the short run

effect of a shock whereas the B coefficient matrix shows the long run effect of the shock. The

BEKK model is estimated with robust standard errors. Finally, the starting values of the

estimation are polished by the use of the SIMPLEX algorithm. The algorithm that is used for

the estimation of BEKK model is the BFGS (Broyden, Fletcher, Goldfarb and Shanno). The

analytical form of the BEKK (1.1) model is given below.

2 2 2 2 2 2 2

11, 11 11 1,t 1 11 12 1, 1 2, 1 12 2, 1 11 11, 1 11 12 12, 1 12 22, 12 2 (5.3)

t t t t t t th c a u a a u u a u b h b b h b h

2 2 2 2 2 2 2 2

22, 21 22 21 1,t 1 21 22 1, 1 2, 1 22 2, 1 21 11, 1 21 22 12, 1 22 22, 12 2 (5.4)

t t t t t t th c c a u a a u u a u b h b b h b h

2 2

12, 11 21 11 21 1, 1 11 22 12 21 1, 1 2, 1 12 22 2, 1 11 21 11,t 1

11 22 12 21 12, 1 12 22 22, 1

( )

( ) (5.5)

t t t t t

t t

h c c a a u a a a a u u a a u b b h

b b b b h b b h

90

where: 1, 1tu

and 2, 1tu

are the one lagged error terms of the producer and the consumer equation,

respectively.

When the most appropriate BEKK model is selected by Information Criteria and

Likelihood Ratio tests, then the volatility impulse responses are calculated following the

method of Hafner and Herwartz (2006). Appendix A presents in detail the selection process for

the most appropriate BEKK model. In order to proceed with the estimation of the impulse

responses, first, the BEKK model should be transformed to its vech-representation as Engle and

Kroner (1995) suggest. The vech-representation is given by (8).

'

t t 1 t 1 t 1vech(H ) Q R vech(E E ) P vech(H ) (5.6)

where: Q is a matrix of constants, R and P are coefficient matrices. Vech is the operator that

stacks the lower triangular part of a square matrix to a vector. Below, the Q , R and P matrices

are linked to the parameters of the BEKK model.

2

11

11 21

2 2

21 22

,

c

Q c c

c c

2 2

11 11 12 12

11 21 11 22 12 21 22 12

2 2

21 21 22 22

2

2

a a a a

R a a a a a a a a and

a a a a

2 2

11 11 12 12

11 21 11 22 12 21 22 12

2 2

21 21 22 222

b b b b

P b b b b b b b b

b b b b

If the conditional variance is at an initial state 0

H and a shock '

0 1,0 2,0( , ) occurs, the

volatility impulse response t 0

V (Z ) is given by (9).

t 0 t t 1 0 t t 1V ( ) [vech(H ) | F , ] [vech(H ) | F ] (5.7)

Then, the volatility impulse response 1 0

V (Z ) is recursively computed by the following

relations:

1 1

'2 2

1 0 0 0 0 0 0V (Z ) R vech H Z Z H vech H (5.8)

1 0 t 1 0V (Z ) R P V (Z ), t 1 (5.9)

91

It should be noted that in estimating volatility impulse responses (Hafner and Herwartz

2006), the selection of a specific month during which a shock takes place is necessary. The

researcher can estimate the volatility impulse responses either by selecting the month of the

shock based on his subjective judgment or by implementing a method that identifies structural

breaks such as the Bai-Perron method (1998). However, in this chapter this is accomplished by

implementing a Markov Switching Vector Error Correction (MSVEC) model (Krolzig 1997)

which is a generalization of Hamilton’s model (1989). In this way, the data generating process

allows the most important shocks to become apparent. According to Hamilton (1989), the

turning point in a Markov model is a structural event that is inherent in the data generating

process. Therefore, the turning points of the smoothed probabilities of a Markov model can be

considered as stochastic structural breaks (since they are probabilistic) in contrast to the

deterministic structural breaks of the Bai-Perron (1998; 2003) method. Moreover, in the

Markov process there is no need for the researcher to designate the number of breaks that the

process should identify or the minimum distance between them. Thus, the Markov process has

the ability to identify structural breaks irrespective of their number and distances. What is more,

the Bai-Perron method is linear, thus incapable of taking into consideration possible non-

linearities of the data generating process.

The literature provides evidence of the shortcomings of the Bai-Perron method which

is used to identify structural breaks. In particular, according to Kar et al. (2013) the Bai-Perron

method has low power since it is incapable of discerning true breaks in growth regimes,

especially for countries where the series are highly volatile. This is known as the “true negative”

problem, which is also confirmed by Bai and Perron (2006). Thus, for chapter 5 the use of the

Bai-Perron method would be inadequate since the producer and the consumer prices are highly

volatile. Furthermore, the Bai-Perron method is ahistorical since it does not account for

previous breaks of the series under investigation when identifying current ones. This results in

missing out on changes in rates of economic growth that are of a significant magnitude. The

result is also supported by Pesaran and Timmermann (2002). In contrast, in a Markov process

the smoothed probabilities are calculated based on the whole sample period (Krolzig 1996).

Finally, the literature provides evidence that the Bai-Perron method is sensitive to changes in

the base year and length of a series (Dholakia and Sapre 2011). Rao’s (2010) investigation of

the consumption of spirits in the UK for structural breaks supports the instability of the Bai-

Perron method since increases and decreases of the break dates were needed in order to achieve

good results. In contrast, in the analysis the structural breaks identified by the Markov process

remain stable when the lag-length of the autoregressive processes changes. The result cannot

92

be generalized, however it is an indication of the stability of the calculated smoothed

probabilities.

The estimation of the MSVEC model gives as a by-product the smoothed probabilities

(Krolzig 1996), which show the probability of the price transmission mechanism being in state

i in period t based on information that is available for the whole period. The point at which

the smooth probability of the price mechanism is reduced below 50% is considered to be the

break point, where the price mechanism switches state. The months before and after the break

point designate the threshold between the high- and the low-volatility regime of the price

mechanism. Then, the two months are used for the calculation of the volatility impulse

response. The MSVEC models are formulated by letting ( )t t t

Y P C be the two-

dimensional vector of the variables of interest where 1t T (T is the sample size) and

( 1, , )t

s i i M is the M state unobserved variable that follows a first-order ergodic Markov

chain. The number of the states of the unobserved variable is countable. The 2nd order MSVEC

model will be written as:

0 1, 2, 11 1

( ) ( ) ( ) ( ) ( )

K Kp p p p p

t t k t t k k t t k t t t tk k

P s i s i P a s i C s i ect u s i

0 2, 1, 11 1

(5.10)

( ) ( ) ( ) ( ) ( )K Kc c c c c

t t k t t k k t t k t t t tk k

C s i s i C a s i P s i ect u s i

where: ( )t t

P C are the first differences of the logged producer (consumer) price at period t

, 0 0

( ) ( ( ))p c

t ts i s i is the state dependent intercept term of the producer (consumer)

equation, ( )t k t k

P C

are the lags of the producer (consumer) price, 1t

ect

is the error correction

term at period 1t and ( ) ( ( ))p c

t t t tu s i u s i is the state dependent error term of the estimated

equation of the producer (consumer).

Finally, the asymmetric effects of the price volatility transmission mechanisms are

investigated by asymmetric BEKK models. The asymmetries are incorporated to the model

according to the formulation of Glosten et al. (1993). This formulation defines t

v as

( 0)t t t

v u I u so that it it

v u if 0it

u and 0it

v otherwise, done component by

component. Thus, the asymmetric BEKK model is formulated as:

' '

1 1 1 1 1( ) ( ) (5.11)

t t t t t tH C A u u B H D v v

93

For the off-diagonal elements ( )ij , the asymmetry term is non-zero only if both , 1i t

u

and , 1j t

u

are negative. More specifically, the procedure differentiates between those data points that are

both negative and those where at least one is positive (Doan 2013). In the asymmetric BEKK

models the coefficients of the recent shocks matrix ( )A and of the asymmetry matrix ( )D do

not have a direct interpretation because of the way the two of them interact as is apparent. On

the other hand, the coefficients of the past volatility matrix keep their direct interpretation.

Whether the asymmetric effects are significant is assessed with two tests: the cross-variance

effects test and the asymmetric effects test. The former reveals whether the spillover effects and

the off-diagonal asymmetric coefficients are jointly significant whereas the latter examines the

significance of the asymmetric coefficients and are presented in table 5.1.

Table 5.1 The null hypothesis for the cross and asymmetry effects tests

Cross effects test 0 12 21 12 21 12 21:H a a b b d d

Asymmetry effects test 0 11 12 21 22: dH d d d

5.3 Data

The national averages of the monthly prices of fresh potatoes, tomatoes and cucumbers

for the producer and the consumer are used for the estimation procedure. The data set was made

available from the Hellenic Statistical Authority. The prices of fresh tomatoes and cucumbers

cover the period from January 1995 to September 2013. The prices of fresh potatoes extend

from January 1991 to September 2013. All the prices are converted into their natural logarithms

and they are nominal. Table 5.2 shows the descriptive statistics for the differenced prices of the

three products.

Table 5.2 The descriptive statistics for the differenced prices of the potatoes, tomatoes and

cucumbers

Producer

Mean Std. Deviation Skewness Kurtosis

Potatoes 0.002 0.145 1.213 11.64

Tomatoes 0.001 0.202 -0.235 3.806

Cucumbers 0.001 0.222 -0.240 3.249

94

Consumer

Mean Std. Deviation Skewness Kurtosis

Potatoes 0.003 0.084 -1.144 6.641

Tomatoes 0.002 0.150 -0.265 4.039

Cucumbers 0.2∙10-3 0.159 0.259 4.062

Figure 5.1, below, presents the differenced prices of potatoes, tomatoes and cucumbers for

producer as well as for consumer.

Figure 5.1 The differenced prices of fresh potatoes, tomatoes and cucumbers

potatoes tomatoes

cucumbers

As is observed by figure 5.1, the differenced prices of the three products are highly volatile

while they present clusters of volatility during specific periods.

5.4 Empirical Results

95

5.4.1 Unit Roots and Co-integration Analysis

Prior to the co-integration analysis, the stationarity properties of the producer and the

consumer price series were examined by the Elliot et al. (ERS) unit root test (Elliot et al. 1996).

As it is made evident from table 5.3, the results support the non-stationarity of the natural

logarithms of the producer and the consumer prices as well as the stationarity of the first

differences of the prices for all three products. Thus, the next step of co-integration analysis is

performed.

Table 5.3 The results of the ERS test for potatoes, tomatoes and cucumbers

Potato Tomato Cucumber

tP -1.398 (-1.942) -2.627 (-1.942) -0.350 (-1.942)

tC -0.965 (-1.942) 0.407 (-1.942) -2.527 (-1.942)

tP -2.665 (-1.942) -12.49 (-1.942) -11.50 (-1.942)

tC -1.380 (-1.942) -12.45 (-1.942) -11.74 (-1.942)

In the parenthesis 5% critical values are reported. The deterministic part of the unit root processes is a constant for

all the tests. The most appropriate unit root process for each one of the tests is chosen by Information Criteria and

Likelihood Ratio tests.

Before carrying out the co-integration test, a vector error correction model is estimated for each

product. A 2 lag model is selected for potato, a 6 lag model for tomato and an 8 lag model for

cucumber. All three models incorporate 12 centered seasonal dummies. The number of co-

integrating relationships between producer and consumer prices is defined by the trace test. The

implementation of the test leads to the conclusion that one long run relationship exists for each

one of the products, as shown in table 5.4.

Table 5.4 The results of Johansen’s Trace Test

number of co-integrating

relationships Potato Tomato Cucumber

0 0.000 0.015 0.001

1 0.262 0.153 0.089

P-values are reported.

96

The optimal model for each product was selected by the information criteria of Schwarz (BIC)

and Hannan – Quinn (HQIC) as well as by likelihood ratio tests. Moreover, mis-specification

tests for autocorrelation, heteroscedasticity and normality of the error term were used. The

results of the co-integration analysis for the three products are given by table 5.5 where ect

represents the long-run relationship.

Table 5.5 The results of the co-integration analysis

Potato Tomato Cucumber

tect

(-10.55) (-23.86)0.595 0.891

t tC P

(-8.909) (-7.761)0.920 1.289

t tC P

(-12.61) (-9.862)0.521 0.905

t tC P

Number of lags 2 6 8

BIC -9.444 -7.566 -7.181

HQIC -9.699 -8.008 -7.701

LM autocorrelation 0.118 0.931 0.428

Normality 0.000 0.695 0.309

LM heteroskedasticity 0.718 0.657 0.776

In parenthesis t-values are reported. For LM and normality tests’ p-values are reported.

As table 5.5 reveals, the coefficients of the co-integrating vectors are significant. Moreover, the

analysis of the residuals of the three models shows that the vector error correction models are

well specified. In particular, for the case of tomato and cucumber there are no signs of

autocorrelation or heteroscedasticity and they are normally distributed. However, for potatoes

the null hypothesis of normality is rejected.

5.4.2 The results of the estimation of the BEKK models

The most appropriate BEKK model for the volatility price transmission mechanism of

each product is determined according to the information criteria of Akaike (AIC), Swartz (BIC)

and Hannan-Quinn (HQIC) as well as likelihood ratio tests. The results of the mean models are

available on request. The volatility model that is qualified for fresh potatoes is a BEKK (1.1)

model. The results of the estimation of the variance model are presented in table 5.6.

Table 5.6 The results of the symmetric BEKK (1.1) for potatoes

coefficients Producer ( 1)i Consumer ( 2)i

97

Constants

1ic 0.054* (0.000)

2 ic 0.045* (0.000) 1.2∙10-6 (0.999)

Transmission of recent shocks

1ia 0.926* (0.000) 0.605* (0.000)

2 ia -0.480* (0.016) -0.121 (0.416)

Transmission of past volatility

1ib 0.844* (0.000) -0.097* (0.000)

2 ib -0.873* (0.000) 0.077 (0.118)

ARCH-LM 34.60 (0.000)

AIC -4.657

BIC -4.350

HQIC -4.534

Autocorrelation (0.373)

Heteroscedasticity (0.813)

* Denotes significant values at the 10% level of significance. In the parenthesis p-values are reported.

The analysis of the results of table 5.6 suggests that the volatility spillover effects between the

producer and consumer are significant and strong. Inference is based on comparing the

magnitude of coefficients in absolute value. In particular, a shock in the producer price volatility

is transmitted to the consumer price with a recent effect 12

( 0.605)a that is more intense than

the past volatility effect 12

( 0.097)b . Thus, the magnitude of the shock is mostly transmitted

in the short run, rather than in the long run. On the other hand, the transmission of a price

volatility shock from the consumer to the producer has a recent effect 21

( 0.480)a that is not

as strong as the past volatility effect; 21

( 0.873)b thus, the shock is transmitted in the long

run. It is worth mentioning that the negative coefficients of the off-diagonal elements of the A

matrix may imply that the volatility is more strongly affected when the shocks move in opposite

directions than when they move in the same direction, thus indicating the existence of

asymmetric responses to recent shocks (Doan 2013). Regarding the recent and past volatility

effect of a shock to the own-price volatility of the producer and consumer, the results indicate

that the producer is strongly affected 11 11

( 0.926, 0.844)a b whereas the consumer is not (the

98

effects are not significant 22 22

0.121, 0.077)a b . Thus, in the potato market a volatility

shock is transmitted upwards in the chain (from the producer to the consumer) more quickly

(the recent effect is stronger than the past volatility effect) than a shock that goes downwards

(from the consumer to the producer), while the consumers seem to transmit their volatility

shocks entirely to the producer. Moreover, the model seems to be well specified since the

analysis of the residuals does not reveal any sign of autocorrelation or heteroscedasticity while

the ARCH-LM test supports the existence of clustered residuals. Next, the volatility impulse

responses are calculated for the months that signal the transition of the price mechanism from

the high to the low volatility state. The MSVEC model that is estimated consists of two lags

and is characterized by two states. As table 5.6a shows, the comparison of the variances ( ) of

the producer and the consumer between the states reveals that in state 1 the producer and the

consumer variances are larger than state 2. Thus, state 1 is considered to be the high volatility

state, while state 2 the low volatility state.

Table 5.6a The variance of the states of the MSVEC model for potatoes

State 1 State 2

Producer Consumer Producer Consumer

0.034 (0.000) 0.007 (0.000) 0.005 (0.000) 0.003 (0.000)

Figure 5.2a depicts the smoothed probabilities of the potato price transmission mechanism for

state 1 in conjunction with the prices of the producer and the consumer in first differences. The

examination of the figure reveals that a big cluster of volatility in the producer prices expanded

from January 1991 to October 1999. During this period, the price mechanism is in the high

volatility state. The switch of the state from high volatility to low volatility occurred between

October 1999 (1999:10) and November 1999 (1999:11) where the smoothed probabilities are

0.55 and 0.37, respectively.

Figure 5.2a The smoothed probabilities of potatoes price mechanism in state 1 with the prices

of the producer and the consumer in first differences.

99

For these two transitional months the volatility impulse responses are calculated and presented

in figure 5.2b.

Figure 5.2b The volatility impulse responses of potatoes for 1999:10 and 1999:11.

1999:10 1999:11

The shock of 1999:10 affected producer volatility much more than the consumer’s. The effect

is positive for both the producer and the consumer. The spillover effect from the producer to

the consumer begins from a negative level with a decrease of the volatility transmission

depicted for the first month; however, for the rest of the time horizon the effect is positive. The

effect of the shock on producer’s and consumer’s volatility, as well as on the spillover effect,

100

dies out after a long period of almost 150 months, which equals the duration of the low volatility

state of the price mechanism for potatoes. Regarding the shock of 1999:11, the examination of

the graph reveals that the volatility of the producer decreases before the response to the shock

fades away after almost 100 months. Similarly, the consumer volatility as well as the spillover

effect from the producer to the consumer are negative; however, their persistence lasts for just

one month. Thus, for the two transitional months the patterns of the volatility responses are

opposite and asymmetric. In 1999:10, the high volatility month, the response of the producer

and the consumer is positive, whereas in 1999:11, the low volatility month, is negative. The

response of the consumer and the spillover effect are almost instantaneous in the low volatility

state in contrast to the high volatility state where they last for much longer.

Regarding tomatoes the results of the BEKK (1.1) model that depict the volatility price

transmission mechanism between the producer and the consumer are presented in table 5.7.

Table 5.7 The results of the symmetric BEKK (1.1) for tomatoes

coefficients Producer ( 1)i Consumer ( 2)i

Constants

1ic 5.9∙10-7 (0.999)

2 ic -4.4∙10-7 (0.999) 7.1∙10-7 (0.999)

Transmission of recent shocks

1ia 0.168* (0.027) 0.342* (0.000)

2 ia -0.275* (0.000) -0.235* (0.009)

Transmission of past volatility

1ib -1.364* (0.000) -0.971* (0.000)

2 ib 0.881* (0.004) 1.284* (0.000)

ARCH-LM 28.80 (0.000)

AIC -2.321

BIC -1.842

HQIC -2.128

Autocorrelation (0.842)

Heteroscedasticity (0.370)

* Denotes significant values at the 10% level of significance. In the parenthesis p-values are reported.

101

The analysis of the results presented in table 5.7 indicates that the transmission of volatility

between the producer price and the consumer price is significant, while the transmission occurs

in the long run. More specifically, a volatility shock of the producer price is transmitted to the

consumer price in the long run, since the recent effect of the shock 12

( 0.342)a is weaker than

the past effect of the shock 12

( 0.971)b . Similarly, the volatility spillover effect of a shock

from the consumer price to the producer price is transmitted with more intensity in the long run

21( 0.881)b than in the short run

21( 0.275)a . Furthermore, the negative coefficients of the

off-diagonal elements of the A matrix show the existence of possible asymmetries. The own

effects of a volatility shock to the producer and consumer are present in the long run

11 22( 1.364, 1.284)b b , since the effect is much smaller in the short run

11 22( 0.168, 0.235)a a . Therefore, a shock, whether it is upwards or downwards, is

transmitted with a lag, while it affects both the producer’s and the consumer’s own volatility in

the long run. Furthermore, the model does not show any signs of autocorrelation or

heteroscedasticity and the ARCH-LM test gives evidence of clusters of large residuals.

Continuing, the volatility impulse responses are calculated for the months that signal the state

switch of the price mechanism of tomatoes. The MSVEC model that is estimated is

characterized by one lag and two states. As shown in table 5.7a, state 1 is defined as the high-

volatility state, whereas state 2 is defined as the low-volatility state.

Table 5.7a The variance of the states of the MSVEC model for tomatoes

State 1 State 2

Producer Consumer Producer Consumer

0.040 (0.000) 0.022 (0.000) 0.019 (0.000) 0.011 (0.000)

Figure 5.3a depicts the smoothed probability of the price transmission mechanism for the high-

volatility state. The graph shows that the price mechanism of tomatoes shifts between states

frequently, though; there is a long period from October 1998 to June 2008 during which there

is a long volatility cluster. Therefore, the impulse response will be calculated for the months

during which the price mechanism shifts from the high-volatility state to the low-volatility state

at the end of the cluster. These months are June 2007 (2007:06) and July 2007 (2007:07), with

smooth probabilities of 0.72 and 0.44, respectively.

102

Figure 5.3a The smoothed probabilities of tomatoes price mechanism in state 1 with the prices

of producer and consumer in first differences.

For these two transitional months the volatility impulse responses are calculated and are

presented in figure 5.3b.

Figure 5.3b The volatility impulse responses of tomatoes for 2007:06 and 2007:07.

2007:06 2007:07

According to figure 5.3b, the response of the producer volatility is not as strong as the response

of the consumer volatility to the shock of 2007:06. The effect of the response is negative for

both the producer and the consumer while presenting an intense memory effect. In the same

103

fashion, the spillover effect from the producer to the consumer shows a negative response with

strong memory effects. The persistence of the shock lasts for almost 50 months for both the

consumer and the spillover effect before it converges to zero in the next 25 months. The effect

on the producer’s volatility dies out in 50 months, but it converges to a point a little below zero.

On the other hand, the 2007:07 shock has a positive effect on the volatility of the producer and

the consumer as well as on the spillover effect. The strong memory effect is still prevalent and

the consumer’s volatility is affected more. Furthermore, the persistence of the shock lasts for

50 months, then it fades away completely in the next 25 months. Thus, the switch of the price

mechanism from high to low volatility results in asymmetric responses of the producer and the

consumer to volatility shocks.

The price volatility transmission mechanism of cucumbers between the producer and

consumer is modelled by a symmetric BEKK (1.1) model as well. Table 5.8 depicts the

estimated coefficients for the variance model.

Table 5.8 The results of the symmetric BEKK (1.1) for cucumbers

coefficients Producer ( 1)i Consumer ( 2)i

Constants

1ic -1.3∙10-6 (0.999)

2 ic -7.1∙10-7 (0.999) 2.8∙10-7 (0.999)

Transmission of recent shocks

1ia -0.015 (0.724) -0.072* (0.042)

2 ia 0.130 (0.124) 0.195* (0.006)

Transmission of past volatility

1ib 0.986* (0.000) -0.009 (0.183)

2 ib 0.015 (0.393) 0.997* (0.000)

ARCH-LM 18.66 (0.028)

AIC -1.957

BIC -1.667

HQIC -1.840

Autocorrelation (0.102)

Heteroscedasticity (0.161)

* Denotes significant values at the 10% level of significance. In the parenthesis p-values are reported.

104

An examination of the results in table 5.8 reveals that the producer’s price volatility is

transmitted to the consumer but not vice versa. More specifically, a shock in the price volatility

of the producer is transmitted to the consumer 120.072a in the short run; however, in the

long run, there is no price volatility transmission 12

( 0.009)b . On the other hand, price

volatility spillovers from the consumer to the producer do not take place

21 21( 0.130, 0.015)a b . In the case of cucumbers, there are also negative coefficients in the

off-diagonal elements, although there are no shocks that move in opposite directions, so

asymmetries probably do not exist. The effect of a shock to the producer’s own volatility is

effective only in the long run 11

(b 0.986) , since the effect is insignificant in the short run

11( 0.015)a . Regarding the own effect of the consumer’s volatility, it is observed that the

effect is weaker 22

( 0.195)a in the short run than in the long run 22

( 0.997)b . Therefore,

upward shocks seem to be transmitted partly in the short run; however, downward shocks seem

to be absorbed at the consumer level. Similar to the case of potatoes and tomatoes, the model

is well specified while the residuals are clustered. Next, the dynamics of the price volatility

mechanism are examined by calculating the volatility impulse responses. The MSVEC model

that is qualified consists of one lag and two states. Table 5.8a shows that state 1 is the high-

volatility state, whereas state 2 is the low-volatility state.

Table 5.8a The variance of the states of the MSVEC model for cucumbers

State 1 State 2

Producer Consumer Producer Consumer

0.051 0.032 0.034 0.010

Figure 5.4a depicts the smoothed probability of cucumbers’ price transmission mechanism for

the high-volatility state. According to the graph, the price mechanism shifts from state 1 to state

2 during May 2004 (2004:05) and June 2004 (2004:06). The smooth probability of May is 0.67,

while the smooth probability of June is 0.46. Moreover, when the price mechanism is in the

high-volatility state, a long volatility cluster is observed from January 1995 to June 2004.

Figure 5.4a The smoothed probabilities of cucumbers price mechanism in state 1 with the

prices of producer and consumer in first differences.

105

For the two transitional months the volatility impulse responses are calculated and are presented

in figure 5.4b.

Figure 5.4b The volatility impulse responses of cucumbers for 2004:05 and 2004:06.

2004:05 2004:06

The volatilities of the producer and the consumer are positively affected by the shock of

2004:05. The response of both volatilities exhibits a memory effect as their magnitudes are

reduced nearly to zero after 160 months for the producer and after 120 months for the consumer,

106

respectively, but then they increase until they die out after approximately 380 months. The

spillover effect from the producer to the consumer follows a similar response with a memory

effect before diminishing to zero. In contrast, the producer and the consumer volatility are

negatively affected by the shock of 2004:06. The volatilities of the producer and the consumer

diminish after 170 and 110 months, respectively, but because of the memory effect, they

increase before fading away completely in about 400 and 300 months, respectively. The

spillover effect of the producer and the consumer follows a similar pattern. It decays until the

one hundred and thirtieth month, when an increase takes place before converging in the next

170 months. Thus, the paths of the responses of the producer’s and the consumer’s volatility

change between 2004:05 and 2004:06 from positive to negative; however, they are more

symmetric than not.

So far, the empirical analysis suggests the presence of possible asymmetries in the

transmission of price volatility of potatoes and tomatoes. More specifically, the negative off-

diagonal coefficients as well as the volatility impulse responses indicate asymmetric reactions

to shocks for potatoes and tomatoes. Thus, asymmetric BEKK models were estimated so as to

investigate these indications further. The models are qualified according to the information

criteria of Akaike (AIC), Swartz (BIC) and Hannan-Quinn (HQIC) as well as likelihood ratio

tests. The results of the mean models are available on request. Table 5.9 presents the results of

the estimation for potatoes.

Table 5.9 The results of the asymmetric BEKK (1.1) model for potatoes

coefficients Producer ( 1)i Consumer ( 2)i

Constants

1ic 0.051* (0.000)

2 ic 0.045* (0.000) -0.004* (0.015)

Transmission of recent shocks

1ia 0.018 (0.428) 0.140* (0.000)

2 ia 0.577* (0.000) 0.439* (0.000)

Transmission of past volatility

1ib 0.927* (0.000) -0.017* (0.000)

2 ib -0.965* (0.000) 0.025 (0.380)

Asymmetries

107

1id 1.083* (0.000) 0.631* (0.000)

2 id -0.761* (0.000) -0.315* (0.000)

ARCH-LM 34.60 (0.000)

AIC -4.705

BIC -4.345

HQIC -4.560

Autocorrelation (0.646)

Heteroscedasticity (0.799)

* Denotes significant values at the 10% level of significance. In the parenthesis p-values are reported.

As table 5.9 shows, the comparison of the past volatility matrices of the symmetric and the

asymmetric models reveals that the incorporation of asymmetric effects to the model does not

affect past volatility. However, according to the information criteria the asymmetric model

achieves a better fit than the symmetric. Thus, the existence of asymmetries is confirmed.

Moreover, the diagnostic tests support the implementation of the asymmetric BEKK model

while verifying the good fit of the model to the data. In order to shed more light to the

asymmetric effects, the cross and the asymmetric effects tests are used. The results are shown

in table 5.9a.

Table 5.9a The Cross and Asymmetric effects of the BEKK (1.1) model of potatoes

Cross effect 3448.8 (0.000)

Asymmetry effect 7734.6 (0.000)

In the parenthesis p-values are reported.

The results verify that the spillover effects and the asymmetric effects are significant and

improve the fit of the model.

Below, the results of the estimation of the asymmetric model for tomatoes are presented

in table 5.10.

Table 5.10 The results of the asymmetric BEKK (1.1) model for tomatoes

coefficients Producer ( 1)i Consumer ( 2)i

Constants

1ic 0.156* (0.000)

108

2 ic 0.059* (0.000) -1.3∙10-7 (0.999)

Transmission of recent shocks

1ia 0.352* (0.002) 0.354* (0.000)

2 ia -0.490* (0.031) -0.140 (0.348)

Transmission of past volatility

1ib -0.347 (0.589) -0.434 (0.180)

2 ib 0.707* (0.063) 1.082* (0.000)

Asymmetries

1id -0.136 (0.650) 0.403* (0.000)

2 id -0.439 (0.187) -0.757* (0.000)

ARCH-LM 16.83 (0.005)

AIC -2.320

BIC -1.906

HQIC -2.153

Autocorrelation (0.741)

Heteroscedasticity (0.431)

* Denotes significant values at the 10% level of significance. In the parenthesis p-values are reported.

The results indicate that the asymmetric model performs better than the symmetric one

according to the information criteria. In contrast to the case of potatoes, the estimated past

volatility of the asymmetric model of tomatoes is affected by the incorporation of the

asymmetric term. More specifically, the producer’s past volatility estimations are affected.

Moreover, the diagnostic tests show no signs of misspecification. Table 5.10a presents the

results from the cross and the asymmetric effects tests.

Table 5.10a The Cross and Asymmetric effects of the BEKK (1.1) model of tomatoes

Cross effect 32.94 (0.000)

Asymmetry effect 27.86 (0.000)

In the parenthesis p-values are reported.

The cross effect test as well as the asymmetry test verify that the spillover effects as well as the

asymmetries are significant and they should be incorporated in the model.

109

Finally, the results of the estimation for cucumbers are presented in table 5.11.

Table 5.11 The results of the asymmetric BEKK (1.1) model for cucumbers

coefficients Producer ( 1)i Consumer ( 2)i

Constants

1ic 5.6∙10-8 (0.999)

2 ic 1.6∙10-7 (0.999) 2.3∙10-8 (0.999)

Transmission of recent shocks

1ia 0.001 (0.984) -0.051 (0.279)

2 ia 0.092 (0.367) 0.156 (0.131)

Transmission of past volatility

1ib 0.987* (0.000) -0.007 (0.129)

2 ib 0.013 (0.335) 0.990* (0.000)

Asymmetries

1id -0.099 (0.198) -0.143* (0.042)

2 id 0.222 (0.161) 0.329* (0.050)

ARCH-LM 41.44 (0.000)

AIC -1.950

BIC -1.536

HQIC -1.783

Autocorrelation (0.725)

Heteroscedasticity (0.058)

* Denotes significant values at the 10% level of significance. In the parenthesis p-values are reported.

The results of the estimation of the asymmetric model show that the past volatility estimates

are almost identical to the ones of the symmetric model. Moreover, the information criteria

shows that the symmetric model is more appropriate for the modelling of the price volatility

transmission mechanism of cucumbers. This verifies the presence of symmetries in the

cucumber market. The misspecification tests prove once more how appropriate the model is.

Table 5.11a presents the results of the cross and asymmetric effect tests.

Table 5.11a The Cross and Asymmetric effects of the BEKK (1.1) model of cucumbers

110

Cross effect 54.53 (0.000)

Asymmetry effect 4.416 (0.352)

In the parenthesis p-values are reported.

The results verify that asymmetries are not significant. Therefore, as is apparent, in the market

of cucumbers, the producer and consumer do not respond asymmetrically to volatility shocks.

5.5 Discussion

The aforementioned empirical analysis showed that in the tomato market, the volatility

shocks of the producer and the consumer are of a similar magnitude, with lagged time patterns,

while the asymmetric effects are not strong. In the cucumber market, the producer seems to

have a strong position in the chain, since the shocks in consumer volatility are not transmitted

to the producer, whereas part of the producer’s volatility is transmitted to the consumer.

Moreover, asymmetries do not seem to apply. In contrast, in the potato market, the producer’s

position is weak, since the consumer transmits volatility shocks entirely to the producer, and

the asymmetries are much stronger.

The different responses of fresh potatoes on the one hand and of fresh tomatoes and

cucumbers on the other in the absorption of volatility shocks depend on the individual market

characteristics. The common marketing characteristics of the three fresh products are: the self-

sufficiency of the markets, the stable policy framework provided by the CAP, the highly

concentrated retail sector resulting in decentralized production distribution and the low level of

processing of the products, since consumers favor their freshness. Furthermore, the inputs used

are similar for all three products; thus, their prices will affect volatility in a similar way;

substitute products like prebaked potatoes or frozen vegetables are not preferred by consumers,

while speculation and exchange rates do not seem to affect volatility since domestic

consumption is satisfied by domestic production. On the other hand, the differentiating market

characteristics are the Common Market Organization (CMO) for tomatoes and cucumbers and

the easy storability and transportation for potatoes. The empirical results show that tomatoes

and cucumbers seem to absorb volatility shocks much more easily than potatoes, which should

be attributed to the CMO and not to the perishability of the products, especially since potatoes;

easy preservation cannot help producers to deal with price volatility.

These results sheds new light on the study of the price volatility in the Greek food sector,

since previous research (Apergis and Rezitis 2003) has suggested that the producers are in a

111

difficult position, since volatility from other parts of the chain is transmitted to them and their

own response to volatility shocks is particularly intense. According to those results, the

producers’ predicament would become worse as a result of the 1992 CAP reform that liberated

the agricultural markets. However, the results of this chapter show that if producers are

organized co-operatively under the regulatory framework of the CAP, like the tomato and

cucumber producers, their place in the food supply chain is strengthened, even if the CAP aims

at more market-oriented agricultural markets that are more exposed to world prices. The current

results are important for producers and consumers of fresh agricultural products in Greece as

well as for policy makers. Producers and consumers benefit from the lower price volatility that

a regulated market entails. Furthermore, the lower price volatility leads to more optimal

decisions, improving the welfare of both producers and consumers, since the uncertainty and

risk are reduced. Therefore, policy makers should follow a scheme for incorporating potatoes

into the single CMO of the CAP.

Finally, the empirical results of this chapter for economic theory are presented. As it is

already stated, the unregulated market of fresh potatoes suffers from rigidity in the transmission

of price volatility from the producer to the consumer while it is characterized by asymmetric

volatility effects. On the other hand in the markets of fresh potatoes and cucumbers which are

regulated by the CMO for fruits and vegetables, volatility transmission and asymmetric effects

are not strong giving evidence against the expected theoretical result that a regulated market

would result in price rigidities. These results point out to the fact that, in essence, economic

theory cannot predict or explain the existence of asymmetric price adjustment. In particular, as

Jochen and v. Cramon-Taubadel (2004) state, asymmetric price transmission implies that

market participants do not benefited by a price reduction or increase, something that would

occur under symmetries. Therefore, asymmetric price transmission would result in a different

distribution of welfare by altering the timing and size of welfare changes. Similarly,

asymmetries in the price volatility would result in welfare changes quite different from the case

of symmetry. Since these asymmetries can be considered as market failure policy-wise, this

chapter provides evidence that a regulatory framework can result in the mitigation of possible

attempts from one level of the market to exert influence over the pricing behavior of the other.

5.6 Conclusions

Chapter 5 investigated the price volatility transmission mechanisms of fresh potatoes,

tomatoes and cucumbers in Greece by taking into consideration the existence of possible

112

asymmetric effects. The period that was investigated spans from January 1991 to September

2013 for potatoes and from January 1995 to September 2013 for tomatoes and cucumbers. A

new approach to the designation of volatility shocks that are worth examining by volatility

impulse responses is proposed. The proposed methodology designates the important shocks

according to the underlying data generating process by implementing a Markov process. The

empirical results have important implications for the implementation of agricultural policies

since they reveal the vulnerability of agricultural markets to volatility shocks when they are not

under a regulative framework as well as for economic theory.

113

6. Conclusions

This study examined the price transmission mechanism between the producer and the

consumer of the Greek agricultural sector while it focused to three major fresh vegetables:

potatoes, tomatoes and cucumbers.

Chapter 1, the introduction, presented the research question of the present study under

the light of the importance of the Greek agricultural sector for the national economy as well as

the objectives set by the study in order to achieve its goal.

Chapter 2 provided a brief description of the Greek agricultural sector and the main

agricultural product categories while it discussed in brief the literature that has studied the

research question. More specifically, the agricultural GDP, income and employment in

comparison to the EU-28 member states was presented. Furthermore, the historical values of

production and consumption of the main product categories are presented along with a brief

description of the implementation of the CAP via the CMOs’ to the Greek agricultural markets.

Finally, a general presentation of the CAP framework and its tools were presented by focusing

to the reforms of 1992 and 2003.

Chapter 3, modeled the price transmission mechanism between the producer and the

consumer of the Greek agricultural sector from 1995 to 2013 while the effect of the decoupling

to the pricing behavior of the producer and the consumer was assessed. The model used was a

panel VEC. The agricultural products and product categories utilized, account for the 60% of

the Producer Price Index of Agricultural Output and Food Consumer Price Index. In particular,

the product categories of cereals, fruits and vegetables, meat, wine and olive oil are included.

The price mechanism was examined for the direction of the causality relationship between the

producer and the consumer in the short run along with their adjustment to deviations from the

long run equilibrium. Furthermore, the existence of possible asymmetries was assessed. The

empirical results showed that the producer did not respond to deviations from the long-run

equilibrium while the price mechanism was found to be characterized by symmetric effects.

Moreover, price shocks to the producer or the consumer were found to have smoother own

effects in contrast to spillover effects. The examination of the price mechanism before and after

the implementation of the decoupling principle revealed that the causality relationship changed

from a feedback relationship to non-responsive behavior of the producer in the short and long

run. Furthermore, the response of the consumer to price shocks seems to have been mitigated

more than producers between the two periods (pre- and post-decoupling).

114

Chapter 4 examined the price transmission mechanism between the producer and the

consumer of fresh potatoes, tomatoes and cucumbers. The model used was a Markov Switching

Vector Error Correction model. The empirical analysis revealed that the price mechanisms of

the three products were subjected to regime changes. The analysis of the dynamics of the two

different regimes of the price mechanisms was based on the identification of the nature of the

price changes (i.e. permanent vs. transitory), the examination of possible asymmetries in the

short and long run with in- and out-of-sample measures, the response of the producer and the

consumer to lagged changes of the prices and long-run deviations from the equilibrium. Each

regime of the price mechanisms was characterized by a distinct pattern of asymmetric effects,

causality relationships as well as persistency effects. More specifically, the assessment of the

results showed that the potatoes and the cucumbers, despite their different characteristics have

similar price mechanisms, while fresh tomatoes and cucumbers despite their similar

characteristics have different price mechanisms.

Chapter 5, investigated the price volatility transmission mechanisms between the

producer and the consumer of fresh potatoes, tomatoes and cucumbers. The price volatility

mechanisms were modelled with symmetric and asymmetric BEKK models while volatility

impulse responses were estimated. The estimation of the volatility impulse responses has as a

prerequisite the selection of a specific month during which a shock takes place. Previous studies

have estimated volatility impulse responses either by selecting the month of the shock based on

their subjective judgment or by using methods that identify deterministic structural breaks.

However, in this study the selection process is carried out with the use of a Markov Switching

Vector Error Correction model. The empirical analysis revealed that in the tomato market, the

volatility shocks of the producer and the consumer are of a similar magnitude, with lagged time

patterns, while the asymmetric effects are not strong. In the cucumber market, the producer

seems to have a strong position in the chain, since the shocks in consumer volatility are not

transmitted to the producer, whereas part of the producer’s volatility is transmitted to the

consumer. Moreover, asymmetries do not seem to apply. In contrast, in the potato market, the

producer’s position is weak, since the consumer transmits volatility shocks entirely to the

producer, and the asymmetries are much stronger. This difference in the price volatility

mechanisms of fresh potatoes versus tomatoes and cucumbers is attributed to the fact that fresh

potatoes are not regulated by a CMO in contrast to tomatoes and cucumbers.

The three models used for investigating the price mechanisms of the Greek agricultural

sector shed new light to the dynamics of these processes. Each chapter examines the agricultural

sector from a different perspective. Chapter 3 investigates the price mechanism of the main

115

products and product categories of the Greek agricultural sector in a panel framework, thus

taking into consideration the interactions among the different products apart from their price

evolution through time. Chapter 4 and 5 focused to the examination of the price volatility of

fresh vegetables. In particular chapter 4, studied the observed price volatility of fresh potatoes,

tomatoes and cucumbers while chapter 5 examined the unobserved price volatility of the three

products.

The producers and the consumers are the two main links in the supply chain of the Greek

agricultural sector. The examination of the price relationship that binds them revealed the

complexity of the dynamics between them. The design of the study aimed at first acquiring a

view on the price mechanism of the Greek agricultural sector while at the same time to assess

the effect of the decoupling principle on the price mechanism. The examination of the Greek

agricultural sector revealed the gravity of producer’s role in the supply chain since he does not

respond to long run deviations from the equilibrium something that is in contrast to what it

would expected. Usually, the consumer is the one that exerts influence to the pricing behavior

of the producer. On the other hand, the observation that the own effects of a shock are more

easily absorbed in contrast to the spillover dimension of the shock according to the individual

impulse responses is considered as an anticipated result. However, the examination of the

effects of the CAP policy change in 2003 (the decoupling principle) showed that the consumer

appears to be more resilient to price shocks in contrast to the mixed effects that this policy had

for producers was also an unexpected finding. In the next phase, the study focused to the

investigation of fresh vegetables. Fresh vegetables is a neglected product category since is not

examined so frequently as cereals and meat products even though they offer high added value

without covering large areas of land or needing costly investments. These attributes are

important for the Greek agricultural sector since the availability of land and financing of

investments are in low levels. Fresh potatoes, tomatoes and cucumbers were selected as the

vegetables with the bigger weights in the Producer Price Index of Agricultural Output. The

observed as well as the unobserved price volatility was examined with MSVEC and BEKK

models respectively. On the one hand, the MSVEC models revealed that even though the price

mechanisms are subjected to regime switches the price mechanisms of fresh potatoes and

cucumbers have the ability to cope with anticipated price shocks. Though the case of fresh

tomatoes is different, since the price mechanism is not so competent in absorbing price shocks.

On the other hand, the BEKK models showed that fresh potatoes cannot efficiently absorb price

shocks. That is, the producers are susceptible to the transmission of volatility from the

consumers by responding to this influence asymmetrically. That is, unexpected price drops

116

would lead to higher subsequent volatility than would unexpected price increases of the same

magnitude. In contrast, fresh tomato and cucumber producers were not so prone to asymmetric

price volatility transmission from consumers.

The different responses of fresh potatoes on the one hand and of fresh tomatoes and

cucumbers on the other in the absorption of volatility shocks depend on the individual market

characteristics. The common marketing characteristics of the three fresh products are: the self-

sufficiency of the markets, the stable policy framework provided by the CAP, the highly

concentrated retail sector resulting in decentralized production distribution and the low level of

processing of the products, since consumers favor their freshness. Furthermore, the inputs used

are similar for all three products; thus, their prices will affect volatility in a similar way

substitute products like prebaked potatoes or frozen vegetables are not preferred by consumers,

while speculation and exchange rates do not seem to affect volatility since domestic

consumption is satisfied by domestic production. The main sources of diversification among

the three markets are the regulative framework and the degree of perishability of the three

agricultural products. In particular, tomatoes and cucumbers are regulated by the CMO for fruits

and vegetables and they are highly perishable. In contrast, potatoes are more easily preserved,

however they are not organized under a common market. These diversifying market

characteristics are expected to determine the different patterns of anticipated and unanticipated

price transmission in the three markets. More specifically, the fact that potatoes are unregulated

gives them the ability to handle anticipated price shocks however at the same time this is a

shortcoming since it hinders the market to absorb unanticipated price shocks. The regulated

markets of fresh tomatoes handle unanticipated price volatility more efficiently even though

the handling of anticipated price volatility is proved a difficult task for producers. Similarly

cucumbers achieve good performance in encountering anticipated and unanticipated price

volatility though in a much better degree than any other from the examined products.

The results of the present study provided new insights to the price transmission

mechanisms of the Greek agricultural sector. The main finding is that the response of the supply

chain to price shocks depends on the regulative framework as well as to the nature of the shock

(anticipated and unanticipated). Even though potatoes price mechanism can transmit anticipated

price shocks the lack of a regulative framework prohibits the mechanism from transmitting

efficiently unanticipated shocks. On the other hand, tomato price mechanism transmits

unanticipated price shocks effectively though cannot transmit anticipated price changes with

similar efficiency. Regarding cucumbers, their price mechanism is proved to be the most

efficient one regardless of the nature of the price volatility. Considering the agricultural sector

117

in aggregate terms, the analysis reveals that the producer are in a favorable position in the

supply chain, though the effects of decoupling have more favorable effects for the consumer.

These results can lead to certain implication for agricultural policy. Firstly, the markets

of fresh potatoes would be benefited by being incorporated to the CMO for fruits and vegetables

since producers would be protected from price volatility. Moreover, the observation that each

product has its own response to price shocks indicates the importance of balancing better the

generality of policy measures to the individual needs of each market. These findings provide

evidence that the trend of CAP to eliminate product-specific policy measures by substituting

them with more general ones for the sake of simplicity of the regulative framework has as a

side effect to create different marketing conditions for each product agricultural market. Thus,

the trade-off between the generality of the policy framework for simplicity and the complexity

of the product-specific measures should probably reexamined.

118

References

Abdulai A. 2000. Spatial price transmission and asymmetry in the Ghanaian maize Market.

Journal of Development Economics, 63:327-349.

ADAS. 2002. Assessment of the Potential Impact on Beef and Sheep Farming of Agricultural

Supply Chain. A Report prepared for Defra. London.

Apergis N., Rezitis A. 2003. Agricultural price volatility spillover effects: the case of Greece.

European Review of Agricultural Economics, 30:389–406.

Apergis N., Payne J.E. 2009. Energy consumption and economic growth: Evidence from the

Commonwealth of Independent States. Energy Economics, 31:641-647.

Assefa T.T., Meuwissen M.P.M., Oude Lansink A.G.J.M. 2013. Literature review on price

volatility transmission in food supply chains, the role of contextual factors, and the CAP’s

market measures. Working Paper No. 4, ULYSSES project, EU 7th Framework Programme,

30.

Baba Y., Engle R.F, Kraft D.F, Kroner K.F. 1991. Multivariate Simultaneous Generalized

ARCH. Department of Economics Mimeo, University of California, San Diego. Discussion

paper 92–5.

Bai J., Perron P. 1998. Estimating and testing linear models with multiple structural changes.

Econometrica, 66(1):47-78.

Bai J., Perron P. 2003. Computation and analysis of multiple structural change models. Journal

of Applied Econometrics, 18:1-22.

Bai J., Perron P. 2006. Multiple Structural Change Models: A Simulation Analysis. In: Corbea

D., Durlauf S., Hansen B.E. (ed) In Econometric Theory and Practice: Frontiers of Analysis

and Applied Research. Cambridge University Press, Cambridge, UK pp 212-237.

119

Bailey D., Brorsen B.W. 1989. Price asymmetry in spatial fed cattle markets. Western Journal

of Agricultural Economics, 14(2):246-252.

Balke N.S., Brown S.P.A., Yücel M.K. 1998. Crude oil and gasoline prices: An asymmetric

Relationship? Federal Reserve Bank of Dallas, Economic Review, First Quarter, pp. 2-11.

Beetsma R., Giuliadori M. 2011. The effects of government purchase shocks: Review and

estimates for the EU. Economic Journal, 121:4-32.

Bénabou R., Gertner R. 1993. Search with learning from prices: does increased inflationary

uncertainty lead to higher markups? Review of Economic Studies, 60:69-93.

Ben-Kaabia M., Gil J.M. 2007. Asymmetric price transmission in the Spanish lamp sector.

European Review of Agricultural Economics, 34(1): 53-80.

Bollerslev T. 1986. Generalized autoregressive conditional heteroskedasticity. Journal of

Econometrics, 31:307-327.

Bollerslev T. 1990. Modelling the coherence in short-run nominal exchange rates: A

multivariate generalized ARCH model. Review of Economics and Statistics, 72:498-505.

Bollerslev T., Engle R.F., Wooldridge J.M.A. 1988. Capital asset pricing model with time-

varying covariances. The Journal of Political Economy, 96:116-131.

Borenstein S., Cameron A.C., Gilbert R. 1997. Do gasoline prices respond asymmetrically to

crude oil price changes? Quarterly Journal of Economics, 112:305-339.

Breitung J. 2000. The local power of some unit root tests for panel data. Advances in

Econometrics, 15:161-177.

Brummer B., von Cramon-Taubadel S., Zorya S. 2009. The impact of market and policy

instability on price transmission between wheat and flour in Ukraine. European Review of

Agricultural Economics, 36:203-230.

120

Caivano M. 2006. The transmission of shocks between the U.S. and the Euro area. Bank of

Italy (manuscript).

Canova F., Ciccarelli M. 2013. Panel Vector Autoregressive models: A survey. Advances in

Econometrics, Volume 32.

Centre for Rural Economics Research. 2003. CAP Reform: Decoupling Arable Payments. A

Discussion Document for the Department for the Environment Food and Decoupling CAP

Aid Payments from Production, Prepared for Beef and Sheep Division. Defra, London.

Christiaensen L. 2009. Revisiting the global food architecture. Lessons from the 2008 crisis.

WIDER Discussion Paper 04, Helsinki: UNU-WIDER.

Cologni A., Manera M. 2009. The asymmetric effects of oil shocks on output growth: A

Markov-Switching analysis for the G-7 countries. Economic Modelling, 26:1-29.

Crain S.J., Lee J.H. 1996. Volatility in wheat spot and futures markets, 1950-1993: Government

farm programs, seasonality and causality. The Journal of Finance, 51(1):325-343.

Deaton A., Laroque G. 1992. On the behavior of commodity prices. The Review of Economic

Studies, 59:1-23.

Dempster A.P., Laird N.M., Rubin D.B. 1977. Maximum Likelihood from incomplete data via

the EM algorithm. Journal of the Royal Statistical Society, 39:1-38.

Dholakia R.H., Sarpe A.A. 2011. Estimating structural breaks endogenously in India’s post-

independence growth path: An empirical critique. Journal of Quantitative Economics, 9(2):73-

87.

Dickey D.A., Fuller W.A. 1979. Estimators for autoregressive time series with a unit root.

Journal of the American Statistical Association, 74:427-431.

Doan A.T. 2012. RATS handbook for panel and grouped data. Estima.

121

Doan A.T. 2013. Handbook for ARCH/GARCH and Volatility Models. Estima.

Elliot G., Rothenherg T.J., Stock J.H. 1996. Efficient Tests for an Autoregressive Unit Root.

Econometrica, 64:813-836.

Enders W., Granger C.W.J. 1998. Unit-root tests and asymmetric adjustment with an example

using the term structure of interest rates. Journal of Business & Economic Statistics, 16:304-

311.

Enders W. 2010. Applied econometric time series. 3rd ed. John Wiley and Sons, Inc.

Engle R.F., Hendry D.F., Richard J-F. 1983. Exogeneity. Econometrica, 51(2):277-304.

Engle R.F. 1982. Autoregressive conditional heteroscedasticity with estimates of the variance

of United Kingdom inflation. Econometrica, 50:987-1006.

Engle R.F. 1990. Discussion: Stock Market Volatility and the Crash of ’87. Review of Financial

Studies, 3:103-106.

Engle R.F. 2002. Dynamic conditional correlation: A simple class of multivariate generalized

autoregressive conditional heteroskedasticity models. Journal of Business and Economic

Statistics, 20:339-350.

Engle R.F., Kroner K.F. 1995. Multivariate simultaneous generalized ARCH. Econometric

Theory, 11:122-150.

Engle R.F., Ng V.K. 1993. Measuring and testing the impact of news on volatility. Journal of

Finance, 48:1749-78.

European Commission. 2007. The potato sector in the European Union. Brussels.

European Commission. 2013. Overview of CAP reform 2014-2020. Agricultural Policy

Perspectives Brief. No 5, December.

122

European Commission. 2015. Agriculture and Rural development: Direct Support.

European Parliament. 2011. The EU fruit and vegetables sector: Overview and post 2013 CAP

perspective. Brussels.

Frey G. Manera, M. 2007. Econometric models of asymmetric price transmission. Journal of

Economic Surveys, 21(2):349-367.

Gilbert C.L., Morgan C.W. 2010. Food price volatility. Philosophical Transactions of the Royal

Society B, 365:3023-3034.

Glosten L.R., Jagannathan R., Runkle D.E. 1993. On the Relation between the Expected Value

and the Volatility of the Nominal Excess Return on Stocks. Journal of Finance, 48(5):1779-

1801.

Goodwin B.K., Piggott N.E. 2001. Spatial market integration in the presence of threshold

effects. American Journal of Agricultural Economics, 83(2):302-317.

Granger C.W.J. 1969. Investigating causal relations by econometric models and cross spectral

methods. Econometrica 37:424-438.

Hafner C.M., Herwartz H. 2006. Volatility impulse responses for multivariate GARCH models:

An exchange rate illustration. Journal of International Money and Finance, 25:719-740.

Hamilton J.D. 1989. A new approach to the economic analysis of non-stationary time series

and the business cycle. Econometrica 57:357-384.

Houck J.P. 1977. An Approach to specifying and estimating nonreversible Functions. American

Journal of Agricultural Economics, 59:570-572.

Ihle R., Brummer B., Thomson S.R. 2012. Structural change in European calf markets:

decoupling and the blue tongue disease. European Review of Agricultural Economics.

39(1):157-179.

123

Im K., Pesaran M.H., Shin Y. 2003. Testing for unit roots in heterogeneous panels. Journal of

Econometrics, 115:53-74.

Inter-agency Report. 2011. Price Volatility in Food and Agricultural Markets: Policy

Responses.

Jochen M., von Cramon-Taubadel S. 2004. Asymmetric price transmission: A Survey. Journal

of Agricultural Economics, 55(3):581-611.

Johansen S. 1988b. Statistical analysis of co-integration vectors. Journal of Economic

Dynamics and Control, 12:231-254.

Johansen S. 1996. Likelihood-based inference in co-integrated vector autoregressive models.

2nd ed. Oxford University Press.

Johansen S. 1992. Testing weak exogeneity and the order of co-integration in UK money

demand data. Journal of Policy Modeling, 14(3):313-334.

Juselius K. 2006. The co-integrated VAR Model: Methodology and applications. 1st ed. Oxford

University Press, New York.

Kaditi H.A., Nitsi E.I. 2010. The agricultural sector in Greece. Centre of Planning and

Economic Research (KEPE). Athens.

Kar S, Pritchett L, Raihan S, Sen K. 2013. Looking for a break: Identifying transitions in growth

regimes. Journal of Macroeconomics, 28:151-166.

Kinnucan H.W. Forker O.D. 1987. Asymmetry in farm-retail price transmission for major dairy

products. American Journal of Agricultural Economics, 69:285-292.

Koop G., Pesaran M.H., Potter S.M. 1996. Impulse response analysis in nonlinear multivariate

models. Journal of Econometrics, 74:119-147.

124

Krolzig H.M. 1997. Markov switching vector autoregressions: modelling, statistical inference

and application to business cycles analysis. Springer – Verlang: Berlin.

Krolzig H.M. 1996. Statistical analysis of co-integrated VAR processes with Markovian regime

shifts. SFB 373 Discussion Paper 25/1996. Humboldt Universitat zu Berlin.

Kroner K.F., Ng V.K. 1998. Modeling Asymmetric Co-movements of Asset Returns. The

Review of Financial Studies, 11(4):817-844.

Kwiatkowski D., Phillips P.C.B., Schmidt P., Shin Y. 1992. Testing the null of stationarity

against the alternative of a unit root: how sure are we that the economic time series have a unit

root? Journal of Econometrics, 54:159-178.

Lee C.C. 2005. Energy consumption and GDP in developing countries: a co-integrated panel

analysis. Energy Economics, 27:415-427.

Levin A., Lin C.F., Chu S.S. 2002. Unit root tests in panel data: Asymptotic and finite-sample

properties. Journal of Econometrics, 108:1-24.

Lin W.L. 1997. Impulse response function for conditional volatility in GARCH models. Journal

of Business & Economic Statistics, 15:15-25.

Lobley M., Butler A. 2010. The impact of CAP reform on farmers’ plans for the future: Some

evidence from South West England. Food Policy, 35:341-348.

Love I., Zicchino L. 2006. Financial development and dynamic investment behavior:

Evidence from a Panel VAR. The Quarterly Review of Economic and Finance, 46:190-210.

Moss J., McErlean S., Kostov P., Patton M., Westhoff P., Binfield J. 2002. Analysis of the

Impact of Decoupling on Agriculture in the UK. Queens University, Belfast.

Natcher W., Weaver R.D. 1999. Transmission of Price Volatility in Beef Markets: A

Multivariate Approach. Presented at the Annual Meeting of the American Agricultural

Association, Nashville, TN.

125

Nazlioglu S., Soytas U. 2012. Oil price, agricultural commodity prices and the dollar: A panel

co-integration and causality analysis. Energy Economics, 34:1098-1104.

Neumark D., Sharpe S.A. 1992. Market structure and the nature of price rigidity: Evidence from

the market for consumer deposits. Quarterly Journal of Economics, 107:657-680.

Pedroni P. 2000. Fully modified OLS for heterogeneous co-integrated panels. Advanced in

Econometrics, 15:93-130.

Pedroni P. 2004. Panel co-integration: asymptotic and finite sample properties of pooled time

series tests with an application to the PPP hypothesis. Econometric Theory, 20:597-625.

Peltzman S. 2000. Prices rise faster than they fall. The Journal of Political Economy, 108:466-

502.

Pesaran M.H., Shin Y., Smith R.P. 1999. Pooled mean group estimation of dynamic

heterogeneous panels. Journal of American Statistical Association, 94(446):621-634.

Pesaran M.H., Timmermann A. 2002. Market timing and return prediction under model

instability. Journal of Empirical Finance, 9:495-510.

Phillips P.C.B., Perron P. 1988. Testing for a unit root in time series regression. Biometrika,

75:335-346.

Prehn S., Brummer B., Thomson S.R. 2015. Payment decoupling and intra-European calf

trade. European Review of Agricultural Economics, 42(1):1-26.

Rao B.B. 2010. Deterministic and stochastic trends in the time series models: A guide for the

applied economist. Applied Economics, 42(17):2193–2202.

Reziti I. 2005. An investigation into the relationship between producer, wholesale and retail

prices of Greek agricultural products. Centre of Planning and Economic Research, Athens –

Greece.

126

Reziti I., Panagopoulos Y. 2008. Asymmetric price transmission in the Greek agri-food sector:

some tests. Agribusiness, 24(1):16-30.

Rezitis A.N. 2015. The relationship between agricultural commodity prices, crude oil prices

and US dollar exchange rates: A panel VAR approach and causality analysis. International

Review of Applied Economics. 29(3):403-434.

Rezitis A.N., Pachis D.N. 2015. Investigating the price transmission mechanisms of Greek fresh

potatoes, tomatoes and cucumbers markets. Journal of Agricultural and Food Industrial

Organization. (Forthcoming)

Rezitis A.N., Pachis D.N. 2015a. Price transmission along the Greek food supply chain in a

dynamic panel framework: Empirical evidence from the implementation of decoupling.

Working Paper.

Rezitis A.N., Pachis D.N. 2015b. Investigating the price volatility transmission mechanisms of

selected fresh vegetable chains in Greece. Working Paper.

Rezitis A.N., Stavropoulos K.S. 2011. Price transmission and volatility in the Greek broiler

sector: A threshold co-integration analysis. Journal of Agricultural and Industrial Organization,

9:1-35.

Serra T. 2011. Food scares and price volatility: The case of the BSE in Spain. Food Policy,

36(2):179-185.

Sun C. 2011. Price dynamics in the import wooden bed market of the United States. Forest

Policy and Economics, 13:479-487.

v. Cramon-Taubadel S. 1998. Estimating asymmetric price transmission with the error

correction representation: an application to the German pork market. European Review of

Agricultural Economics, 25:1-18.

127

Viaggi D., Raggi M., Paloma S.G.y. 2011. Farm-household investment behavior and the CAP

decoupling: Methodological issues in assessing policy impacts. Journal of Policy Modeling,

33:127-145.

Ward R.W. 1982. Asymmetry in retail, wholesale and shipping point pricing for fresh

vegetables. American Journal of Agricultural Economics, 62:205-212.

Wolffram R. 1971. Positivistic measures of aggregate supply elasticities: Some new

approaches - some critical notes. American Journal of Agricultural Economics, 53:356-359.

Wright B.D., Williams J.C. 1991. Storage and commodity markets. Cambridge, UK:

Cambridge University.

Yang J., Haigh M.S., Leatham D.J. 2001. Agricultural liberalization policy and commodity

price volatility: a GARCH application. Applied Economics Letters, 8(9):593-598.

Zakoïan J.M. 1994. Threshold Heteroskedastic Models. Journal of Economic Dynamics and

Control, 18:931-955.