Effects of Policy Reforms on Price Transmission and Price Volatility in Coffee Markets: Evidence from Zambia and Tanzania

Size: px
Start display at page:

Download "Effects of Policy Reforms on Price Transmission and Price Volatility in Coffee Markets: Evidence from Zambia and Tanzania"

Transcription

1 Aus dem Insiu für Ernährung und Verbrauchslehre der Chrisian-Albrechs- Universiä zu Kiel Effecs of Policy Reforms on Price Transmission and Price Volailiy in Coffee Markes: Evidence from Zambia and Tanzania Disseraion zur Erlangung des Dokorgrades Agrar-und Ernährungswissenschafliche Fakulä der Chrisian-Albrechs-Universiä zu Kiel Vorgeleg von MSBS. Rhoda Mofya-Mukuka Aus Kasama, Zambia Kiel, May 2011 Dekan: Prof. Dr. Uwe Laacz-Lohmann 1. Bericher: Prof. Dr. Awudu Abdulai 2. Berichersaer: Prof. Dr. Dr. Chrisian Henning Tag der Mündlichen Prüfung:

2 Gedruck mi Genehmigung der Agra-und Ernährungswissennschaflichen Fakulä der Chrisian-Albrechs-Unversiä zu Kiel Diese Arbei kann als pdf-documen uner: hp://e-diss.uni-kiel.de/agrar-fak.hml aus dem inerne geladen werden

3 This Thesis is dedicaed o God almighy, he one who makes all hings possible and o my lae parens Mr and Mrs Mofya in heir loving memory ii

4 Acknowledgemens I owe my deepes graiude o my Supervisor, Professor Awudu Abdulai, Chair of he Insiue of Food Economics and Consumpion Sudies of he Chrisian-Albrechs Universiy Kiel, for his guidance and suppor hroughou he program. Alhough I remain responsible for any shorcoming in his sudy, his hesis would no have been possible wihou his valuable ideas and encouragemen ha made me undersand he subjec, dig deeper ino my research and remained moivaed. The financial suppor, provided by he Kaholischer Akademischer Ausländer- Diens (KAAD) is grealy acknowledged. In paricular, I hank Dr Marko Kuhn, head of he African Deparmen a KAAD and Mrs Simone Saure who were always ready o aend o my needs. I also hank Mrs Sabina Kamp, he chaplain for Kaholic Sudens a Kiel Universiy for supporing my applicaion o KAAD. The suppor of Dr Kassi, Dr Becerril, Dr Ali and Dr Ogundari, o whom I always ran o for quesions, is highly acknowledged. I also hank my colleagues Kai-Bri, Rakhshanda, Rebecca, Chrisian, Jan, Daniela and Dr Thiele for a beauiful and friendly working environmen. Special hanks o Kai-Bri for he German version of he summary of his hesis. I also hank he amicable secreary a he insiue, Mrs Nicola Benecke. The unique menoring program of he African Women in Agriculural Research and Developmen (AWARD) made me realize my career goals conribuing o he success of his hesis. I am graeful o AWARD for acceping me as heir fellow from 2009 o I especially hank Dr Aslihan Arslan, my AWARD menor for he monhly meeings and for aking her ime o read hrough his hesis. In Zambia, I acknowledge he saff a Zambia Coffee Growers Associaion, Zambia Coffee Board and he Minisry of Agriculure for providing me wih all he informaion I needed. Lasly, I offer my hearwarming acknowledgemen, love and blessings o my husband John and my children Chewe, Kasongo and Andela for heir love and paience. I also hank my siblings, he Mofyas, for he encouragemen, suppor and prayers. Grea appreciaion also goes o he Pfeiffer family of Kronshagen for he grea hospialiy and for heir endless prayers. May God coninue blessing you. iii

5 Table of Conen Dedicaion Acknowledgemen Table of Conen Lis of Tables Lis of Figures Summary Zusammenfassung Lis of Acronomy ii iii vi vii viii ix xi xiv Chaper 1: General Inroducion Problem Seing and Moivaion Objecives of he Sudy Significance of he Sudy Hypoheses Daa Sources Coffee Producion in Tanzania and Zambia- an Overview Organizaion of he Thesis 11 Chaper 2: Policy Reforms and Governance Srucures in Coffee Value Chains: A sudy of Zambia and Tanzania Absrac Inroducion Background Informaion on Value Chain Analysis and Governance The Value Chain Theory Theory of Governance in Value Chains Theory of Income Disribuion in Value Chain Coffee Inernaional Markes The Inernaional Coffee Value Chain Governance in Inernaional Coffee Value Chains Tanzania Coffee Value Chains On-farm Processing Coffee Cooperaives Exporing Sage iv

6 2.4.4 Governance Srucures and Price Shares Tanzania Fair Trade Coffee Zambia Coffee Value Chains Discussion and Conclusion Policy Recommendaions Chaper 3: Asymmeric Price Transmission in Coffee Markes: Impacs of Economic Reforms for Zambia and Tanzania Absrac Inroducion Modelling Asymmeric Price Transmission Daa Descripion Resuls and Discussion Uni Roo Tess Resuls of he Symmeric Coinegraion Model Threshold Coinegraion Resuls Impulses Response Conclusion Chaper 4: Impac of Economic Reforms on Coffee Price Volailiy in Zambia and Tanzania Absrac Inroducion Economic Reforms and Coffee Markes in Tanzania and Zambia Theoreical Framework for Undersanding Commodiy Price Volailiy GARCH Model Esimaion Procedure Daa Descripion Resuls Conclusion Chaper 5: Response of Coffee Supply in Zambia: Implicaions for Policy Absrac Inroducion Theoreical Framework v

7 5.3 Model Esimaion Daa Descripion Raionale for Variable Selecion Resuls Symmeric Coinegraion and Error Correcion Resuls Asymmeric Coinegraion and Error Correcion Resuls Chaper 6: Conclusion Sudy Focus Summery of Resuls Policy Implicaions APPENDICES Appendix A: Srucural Break Uni Roo Tes Theory Appendix B: Threshold Vecor Error Correcion Resuls Appendix C: Coffee Processing Sages Appendix D: Curriculum Viae vi

8 Lis of Tables Table 3.1: Uni Roo Tes Resuls.59 Table 3.2: Resuls of Engle-Granger Coinegraion Table 3.3 Threshold Auoregression Resuls Figure 3.4 TAR Error Correcion Model resuls Table 3.5: Resuls of ECM for MTAR...64 Table 3.6: Resuls of ECM for MTAR. 66 Table 4.1: Descripive saisics for he series..93 Table 4.2: Uni Roo Tess...94 Table 4.3: ARCH es resuls...96 Table 4.4: Volailiy Esimaes of GARCH and TGARCH Models.. 98 Table 5.1: Uni Roo Tess for Supply Response Variables.122 Table 5.2: Johansen Coinegraion Tes Table 5.3: Symmeric Error Correcion Esimaes..125 Table 5.4: Error Correcion Esimaion above he Threshold Table 5.5: Error Correcion Esimaion below he Threshold Table B1: Asymmeric Error Correcion Esimaions for Zambia and World Prices (Full sample) Table B2: Asymmeric Error Correcion Esimaions for Tanzania and World Prices (Full sample) Table B3: Asymmeric Error Correcion Esimaion for Supply Response below he Threshold..145 vii

9 Lis of Figures Figure 1.1: Tanzania Coffee Producion Expor and Value 3 Figure 1.2: Coffee Producion, Expors and Value for Zambia 3 Figure 2.1: Arabica and Robusa Coffee Prices (1980 o 2009) 25 Figure 2.2: Producer and Reail Coffee Prices in US$ cens per pound 28 Figure 2.3: Coffee Value Chain in Tanzanian and Zambia 36 Figure 3.1 A Threshold of 0.04 Idenified from he Minimum RSS 55 Figure 3.2 Zambia and CIP price rends 57 Figure 3.3 Tanzania and CIP price rends 57 Figure 3.4 Response of Posiive and Negaive Shocks for Zambia- Full Sample 68 Figure 3.5 Response of Posiive and Negaive Shocks for Zambia Pos- Economic Reforms 68 Figure 3.6 Response of Posiive and Negaive Shocks for Tanzania- Full Sample 69 Figure 3.7 Response of Posiive and Negaive Shocks for Tanzania before Economic Reforms 70 Figure 4.1: Seps in Modelling Volailiy 90 Figure 4.2: Coffee Prices for Zambia and Tanzania in US$/ per Lb 91 Figure 4.3: Reurn Price Series for Zambia and Tanzania 92 Figure 4.4: Auocorrelaions of Price Reurns 95 Figure 4.5: Volailiy of Zambia and Tanzania prices Compared o World prices 99 Figure 5.1: Coffee Producion in Zambia 107 Figure 5.2: Supply Response variables 118 Figure 5.3 Zambia Annual Inflaion Rae ( ) 121 Appendix C: Coffee Processing Sages 150 viii

10 Summary Price is he mos criical facor in deermining he way markes funcion. In a perfec marke a price shock is ransmied o oher verically or horizonally conneced markes wihin a given period of ime. However, when markes do no funcion perfecly he speed and he magniude of price ransmission may be hampered. Sae conrolled agriculural commodiy markes preven efficien price ransmission, as a resul, producer prices canno no reflec world prices. Many developing counries have implemened measures o improve price ransmission hrough agriculural marke liberalisaion policies and providing incenives for farmers. However, insead of prices being ransmied symmerically, monopoly or oligopoly markes, governmen policies and high ransacion coss may lead o asymmeric price ransmission. The lead firms in commodiy value chains end o become relucan o pass on prices ha squeeze heir margins. As a resul, producers may no benefi from consumer price increases and similarly, consumers may no benefi from producer price decreases. This sudy akes a look a hese issues in he analysis of coffee producer prices for Zambia and Tanzania. Coffee is an imporan expor commodiy in boh of hese counries and conribues significanly o he creaion of foreign exchange and employmen even if he oal producion is relaively low. Boh counries have liberalised heir coffee markes during he economic reforms of he lae 1990s o differing exens. The effecs of hese reforms on coffee price ransmission, price volailiy and on coffee supply response are examined in his sudy. This disseraion consiss of four papers ha invesigae various aspecs of coffee markes for Zambia and Tanzania. Firs, he sudy invesigaes coffee value chains in boh counries, paying aenion o he governance srucure and is implicaions for producer prices. Second, price ransmission beween coffee world prices and producers prices in Tanzania and Zambia is examined. The hird paper examines price volailiy in boh counries and assesses he impacs of rade liberalisaion policies on price volailiy. The fourh paper analyses coffee supply response o coffee prices in Zambia. ix

11 Resuls for Zambia indicae improved price ransmission afer he implemenaion of economic reforms. As expeced, he ransmission is asymmeric where price decreases are passed on quicker han price increases. In he case of Tanzania, resuls do no show any improvemens afer he economic reforms. Similar resuls are obained from he volailiy sudy where economic reforms led o an increase in coffee price volailiy in Zambia, wih negaive shocks inducing more volaile prices han posiive shocks. However, in Tanzania he inconsisen marke reforms have had no significan effecs on coffee price volailiy. The sudy of coffee supply response for Zambia shows ha in he long-run, coffee prices have negaive alhough insignifican impacs on supply. This could be relaed o he fac ha as a perennial crop supply ends o reach he marke when prices are on he decline. However, in he shor-run, coffee supply response changes are only significan for price decreases, indicaing ha coffee supply falls faser han i rises wih respec o price changes. On he oher hand, supply responds negaively o a sronger local currency and o price of maize, he main compeing crop. However, he 1990s economic reforms in Zambia have had posiive effec on coffee supply. These findings have imporan policy implicaions as hey reveal a shor and efficien value chain and governance srucure ha enables he producers o receive larger shares of he final coffee price. In addiion, he resuls discuss favourable policies for improving ransmission from world markes o producer prices over he pos economic reform period alhough he ransmission remains asymmeric. An efficien price ransmission may work o he disadvanage of he growers in he shor-run as prices become exposed o volaile world prices, bu in he long-run, i may yield he desired oucome. The findings also show he imporance of a liberalised currency for increased coffee supply. x

12 Zusammenfassung Der Preis is der enscheidendse Einflussfakor, wenn es darum geh zu versehen, wie Märke funkionieren. Handel es sich um eine perfeke Marksiuaion, so wird ein Preisschock eines Markes innerhalb einer gegebenen Zeispanne auf andere verikal oder horizonal mieinander verbundene Märke überragen. Wenn Märke jedoch nich perfek funkionieren, kann die Geschwindigkei und das Ausmaß von Preisransmission behinder werden. Agrarmärke, die saalich konrollier werden, können Preisransmission zum Teil vorbeugen, was dazu führ, dass die Produzenenpreise nich den Welpreisen ensprechen. Viele Enwicklungsländer haben darum Maßsäbe eingeführ, um die Preisransmission zu verbessern, indem Agrarmärke liberalisier und Anreize für die Landwire geschaffen werden. Nichsdesoroz führen zum einen Monopole und Oligopole in der Werschöpfungskee, zum anderen hohe Transakionskosen zu einer asymmerischen ansa zu einer symmerischen Preisransmission. Die führenden Firmen in der Werschöpfungskee endieren dazu, nur widerwillig die Preise zu überliefern, die ihre Gewinnspanne in die Enge reiben. Demzufolge können weder die Produzenen von den seigenden Konsumenenpreisen, noch die Konsumenen von den reduzieren Produzenenpreisen profiieren. Diese Sudie behandel diese Themaik, indem die Produzenenpreise von Kaffee in Sambia und Tansania unersuch werden. Kaffee sell ein wichiges Exporproduk in diesen beiden Ländern dar und räg signifikan dazu bei, dass es Fremdwährung und Arbeispläze gib, auch wenn die gesame Kaffeeprodukion in diesen Ländern relaiv niedrig is. In beiden Ländern fand eine unerschiedlich sark ausgepräge Liberalisierung der Kaffeemärke im Zuge der ökonomischen Reformen in den späen 90er Jahren sa. In dieser Sudie werden die Auswirkungen dieser Reformen auf die Preisransmission bei Kaffee, die Preisvolailiä und die Angebosenwicklung von Kaffee unersuch. Diese Dokorarbei beseh aus vier Teilen, die verschiedene Aspeke der Märke für Kaffee in Sambia und Tansania behandeln. Zuers wird die Werschöpfungskee von Kaffee in den beiden Ländern genauer berache. Dabei wird insbesondere ein Auge xi

13 auf die jeweilige Regierungssrukur geworfen und es werden die daraus resulierenden Auswirkungen hinsichlich der Produzenenpreise erörer. Als nächses wird die Preisransmission von den Welpreisen zu den Produzenenpreisen für Kaffee in Tansania und Sambia analysier. Im drien Teil wird die Preisvolailiä der beiden Länder unersuch und es werden die Auswirkungen der Liberalisierung des Handels auf die Preisvolailiä berechne. Der viere Teil ermiel die Angebosenwicklung von Kaffee in Sambia. Die Ergebnisse für Sambia zeigen eine verbessere Preisransmission nach den ökonomischen Reformen. Wie erware is diese Preisransmission asymmerisch, wobei Preissenkungen schneller überragen werden als Preisseigerungen. Im Falle Tansanias lassen die Ergebnisse keinerlei Verbesserungen der Preisransmission nach den ökonomischen Reformen erkennen. Ähnliche Resulae zeigen die Unersuchungen der Preisvolailiä. Die ökonomischen Reformen führen zu einem Ansieg der Preisvolailiä von Kaffee in Sambia, wobei negaive Schocks volailere Preise verursachen als posiive Schocks. Im Gegensaz dazu haben die inkonsisenen Markreformen in Tansania keinerlei signifikane Auswirkungen auf die Preisvolailiä von Kaffee gehab. Die Sudie über die Reakion des Kaffeeangeboes in Sambia zeig, dass Kaffeepreise auf lange Sich einen negaiven, wenn auch nich signifikanen Einfluss auf das Angebo haben. Dies könne daran liegen, dass Kaffee einen besändigen Errag liefer, sodass das Angebo den Mark genau dann erreich, wenn die Preise fallen. Jedoch is auf kurze Sich die Reakion des Kaffeeangeboes nur bei Preissenkungen signifikan, was darauf hindeue, dass das Kaffeeangebo bezüglich Preisänderungen schneller fäll als dass es seig. Auf der anderen Seie reagier das Angebo negaiv auf eine särkere lokale Währung und auf den Preis von Mais, der als Haupkonkurrenzproduk berache werden kann. Jedoch haen die ökonomischen Reformen der 1990er Jahre in Sambia einen posiiven Einfluss auf das Kaffeeangebo. Aus den Ergebnissen lassen sich insofern wichige poliische Implikaionen ableien, als dass sie eine kurze und effiziene Werschöpfungskee und eine Regierungssrukur offenbaren, die es den Produzenen ermöglich, größere Aneile des endgüligen Kaffeepreises zu erhalen. Des Weieren werden mi Hilfe der Ergebnisse wünschenswere poliische Maßnahmen diskuier. Diese beschäfigen xii

14 sich mi einer verbesseren Transmission von den Welpreisen hin zu den Produzenenpreisen im Zuge der Zeispanne nach den ökonomischen Reformen, obwohl die Preisransmission asymmerisch bleib. Auch wenn eine effiziene Preisransmission den Landwiren kurzfrisig Nacheile erbring, da die Produzenenpreise volailen Welpreisen ausgesez sind, so kann sie langfrisig dennoch das erwünsche Ergebnis erzielen. Ferner zeigen die Ergebnisse wie wichig hohe und sabile Preise und eine sabile Währung hinsichlich eines geseigeren Kaffeeangeboes sind. xiii

15 Lis of Acronyms AIC APT ARCH ARMA CBZ CIP CV ECM FAO FAO GARCH GDP GVC ICA ICO IMF KNCU MTAR SSR SETAR TMB TAR TCB TCSO TECM TGARCH ZCSO Akaike Informaion Crieria Asymmeric Price Transmission Auoregression Condiional Heeroskasiciy Auoregression Moving Average Coffee Board of Zambia Composie Indicaor Price Coefficien of Variaion Error Correcion Food and Agriculural Organisaion Food and Agriculural Organisaion Generalised Auoregression Condiional Heeroskasiciy Gross Domesic Produc Global Value Chain Inernaional Coffee Agreemen Inernaional Coffee Organisaion Inernaional Moneary Fund Kilimanjaro Naive Cooperaive Union Momenum Threshold Auoregression Sum of Squared Residuals Self Exciing Threshold Auoregression Tanzania Markeing Board Threshold Auoregression Tanzania Coffee Board Tanzania Cenral Saisics Office Threshold Error Correcion Models Threshold Auoregression Condiional Heeroskasiciy Zambia Cenral Saisics Office xiv

16 Chaper 1 General Inroducion 1.1 Problem Seing and Moivaion Coffee is a commodiy of criical economic imporance in many African counries, such ha even counries like Zambia and Tanzania ha have very small shares of he world marke depend highly on i for foreign exchange and rural employmen. Coffee remains he mos raded commodiy by poor counries wih a global annual expor of 438,000 meric onnes in 2010 (FAO, 2010). Expor values ranged beween $5 billion and $12 billion beween 1997 and 2005 (ICO, 2007). Over 2.25 billion cups of coffee are consumed per day (approximaely 800 billion cups per year) by 40 per cen of he world s populaion mainly in he indusrialized counries. I is also esimaed ha more han 25 million people are direcly employed in he coffee secor worldwide (Pone, 2004). However, despie is economic imporance, he performance of he indusry in he producing counries and he impac on he incomes and welfare of he producers has been unsaisfacory (Oxfam, 2002a). Pas research shows ha he more dependen he counry is on coffee expors he lower he per capia income (Fier and Kamplinsky, 2002). Acually, coffee producer prices showed he greaes fall during he las 20 years, as prices of major agriculural commodiies fell beween 50 and 86 per cen (Slob, 2006). During he same period, global coffee markes have ransformed rapidly. The emergence of new consumpion paerns such as ehical consumerism have no only increased demand for specialy and high qualiy coffee, bu also given rise o new governance srucures along value chains. Large ransnaional corporaions increasingly coordinae and conrol producion and processing in an effor o saisfy changing consumer preferences. However, firm consolidaion a roasing and reailing sages of he coffee value chains has given rise o oligopoly and monopsony powers, wih a few firms seing sandards and virually deermining producer prices (Pone, 2002a). Conversely, aggregaed producer power which was refleced in coffee markeing boards in mos producing counries, has weakened subsanially following 1

17 he abolishmen of markeing boards as par of marke liberalisaion policies. Kaplinsky (2004) argues ha bargaining power shifed from he poor producers o inernaional raders and roasers. Marke liberalisaion has also led o more efficien pass hrough of prices from world prices o producer prices in almos all coffee producing counries, however, he ransmission is mosly asymmeric, such ha price decreases are ransmied faser han price increases (Krivonos, 2004). The improvemen in price ransmission has, neverheless, exposed producers o he high volailiy of world prices. Empirical research by Forenbery and Zapaa (2004) confirms ha price risk for coffee is even higher for smaller expor counries (like Zambia) han in he overall secor. The auhors argue ha he choice of an effecive price risk managemen sraegy can be challenging for low-volume producing counries because hey lack he capaciy o influence prices. Valid concerns have been raised regarding he changing power srucure in he coffee indusry including he increasing marke power of inernaional raders, he asymmeries in he price ransmission, he declining producer prices and he exensive shor-erm producer price volailiy (Kaplinsky; 2004, Oxfam, 2002a; Bacon, 2004; Slob, 2006). Many researchers are concerned abou he effecs of hese facors on he livelihoods of he producers as well as on long-erm coffee producion. In counries like Zambia and Tanzania, invesmen in coffee producion declined considerably in he las decade, a probable consequence of declining and volaile prices. In some cases farmers have resolved o uprooing coffee rees replacing hem wih food crops like maize (Baffes, 2005). In Tanzania s case, coffee revenue declined from $200 million in he lae 1980s o less han $50 million in he 2004 (Piroe e. al, 2006). For Zambia, expors declined subsanially from 6,700 meric onnes in 2004 o less han 2,000 meric onnes by 2009 (see Figure 1.1 and Figure 1.2). Several large coffee esaes in Zambia have closed down, leaving a large rural populaion ou of seasonal or permanen employmen. A he global level, coffee producion has been increasing a declining raes of 0.5 percen beween 1998 and 2010 from 1.7 percen rae in he previous decade (FAO, 2010). Roasers and reailers are unable o find enough quaniies of some coffee varieies needed for cerain coffee blends. 2

18 Figure 1.1: Tanzania coffee producion expor and value Producion Expors Value Source: Own Presenaion based on Daa from FAO and TCB Figure 1.2: Coffee producion, expors and value for Zambia MT / / / / / / / / / / / / /09 PRODUCTION MT QUANTITY EXPORTED MT AVERAGE VALUE US$/MT Source: Own Presenaion based on daa from Zambia Coffee Growers Associaion (ZCGA) Value chain approach has become an increasingly useful approach o gain a comprehensive view of he srucure and barriers in commodiy markes. As a resul, he body of lieraure on Global Value Chains (GVC) has expanded considerably in he las wo decades. Much empirical and concepual analysis has focused on chain governance and producer upgrading (Gereffi, e. e, 2005; Schmiz, 2004; Pone 2004; Humphrey and Schmiz, 2005), marke power and disribuion of gains (Talbo, 1997a; Pelupessy, 2001; Fier and Kaplinsky, 2001; Kaplinsky and Morris, 2001; 3

19 Gilber, 2006; Swinnen e. al, 2007), sandards (Pone, 2004) and more recenly, fair rade (Slob, 2006; Kamplinsky, 2006). This sudy relaes o his body of lieraure in examining he effecs of value chain and governance srucures on producer prices. Furher, he sudy examines coffee price ransmission, volailiy and supply response in order o provide a wider perspecive of undersanding naure and effecs of producer price changes. While many sudies have examined he inernaional coffee marke srucure including supply and demand elasiciy, empirical evidence on asymmeric coffee price ransmission and volailiy is limied. Krivonos (2004) and Warako (2008) examine asymmeric price ransmission in he coffee marke using he Houck (1977) model, where dummy variables for price increases and price decreases are inroduced. This mehod can be misleading in cases where agens do no adjus prices immediaely bu only do so when he price change reaches a cerain hreshold. In realiy, price adjusing coss, such as menu or communicaion coss, preven agens from adjusing coninuously unil he price difference beween wo markes reach a cerain hreshold. A similar argumen can be applied o supply response analysis, where for example, changes in prices mus reach a cerain hreshold before inducing a change in supply of a commodiy. This sudy applies Threshold Auoregression (TAR) class of models ha enable he idenificaion of hresholds in he analysis of asymmeric price ransmission. Von Cramon and Meyer (2000) applied TAR models o commodiy price ransmission, bu hey used zero as a hreshold. TAR model wih hreshold equal o zero does no display a significan degree of asymmery, and ignore he possibiliy ha he hreshold could be differen from zero (Enders, 2004). Enders (2004) also menions ha, a non-zero hreshold has more advanages in ha i capures sraegic behaviours and adjusmen coss ha are rarely observed wih small changes. Mos price ransmission sudies also overlook he possibiliy of srucural breaks in uni roo hypohesis esing. Several researchers including Perron (1989), Zivo and Andrews (1992), Lumsdaine and Papel (1997), Lee and Srazicich s, (2003) and Narayan and Smyh (2005) found evidence of false non-rejecion of he null hypohesis in he radiional Augmened Dickey Fuller (ADF) and Phillips-Perron (PP) afer applying srucural break uni roo models o sysems which are acually 4

20 saionary wih srucural breaks. Glynn e al. (2007) discussed wo advanages of applying srucural break uni roo esing: Firs, i prevens es resuls ha are biased owards non-rejecion. Second, since he procedures can idenify when he possible presence of srucural break occurred, i provides valuable informaion for analysing wheher a srucural break on a cerain variable is associaed wih a paricular governmen policy, economic crisis, regime shifs or oher facors. Besides, srucural breaks can lead o he rejecion of he null hypohesis of symmeric ransmission more ofen han appropriae (von Cramon-Taubadel and Meyer, 2000). These are some of he criical overlooked aspecs in price ransmission sudies ha are addressed in his sudy. Overall, his sudy expands he exising lieraure on agriculural commodiy prices by aking ino accoun value chain srucures, hreshold price movemens and endogenous srucural breaks in order o conribue o more effecive and efficien policy formulaion. 1.2 Objecives of he Sudy The aim of his sudy is o examine coffee producer price movemens for Zambia and Tanzania, and o esablish he effec of hose price changes on coffee supply. To his effec, four specific objecives have been idenified: i. Invesigae coffee value chains and governance srucures and he implicaions on producer price changes in he wo counries. ii. Evaluae he effecs of rade policy changes on world-o-producer price ransmission, aking ino accoun hreshold price movemens. iii. Measure and explain coffee price volailiy wih respec o coffee marke liberalisaion policies. iv. Examine supply response of coffee o price movemens in Zambia. Compared o Tanzania, Zambia s coffee producion declined significanly in he las five years. The firs objecive is addressed using a global value chain analysis approach wih a focus on governance srucures a boh he inernaional and he local marke levels. The second objecive is ackled using hreshold coinegraion and error correcion models. The issue is wheher coffee prices from world markes are symmerically 5

21 ransmied o producers, and how he economic reforms have changed he naure of he price ransmission. GARCH models are used o address he hird objecive wih due aenion o asymmeric volailiy using hreshold GARCH models. The fourh objecive is also deal wih using hreshold coinegraion and error correcion models in order o ake ino accoun poenially asymmeric effecs of coffee prices on coffee supply. 1.3 Significance of he Sudy A key challenge facing mos counries in Sub-Saharan Africa is achieving susainable economic growh sufficien enough o reduce or even eliminae povery. Given he abundan land and a ropical climae suiable for mos agriculural producion, susainable economic growh in Sub-Saharan Africa can be achieved hrough enhanced agriculural producion and expors. Sable prices ha enable farmers o efficienly plan producion are essenial for susainable success of he agriculural indusry. Given ha he coffee value chain is buyer-driven, mos growers are basically price akers. I is, herefore, imporan ha policy makers and farmers in he producing counries undersand how he value chain hey feed ino operaes, how his influences producer prices, how producer prices respond o changes in oher markes, and how his affecs supply. Undersanding hese facors become criical for he developmen of policy inervenions especially hose relaed o price risk managemen sraegies. This sudy is novel in several aspecs: Firs, i is he very firs sudy o examine he coffee indusry from an economic perspecive in Zambia. Coffee has received very lile aenion from researchers and policy makers despie having poenial for improving economic diversificaion in Zambia s copper dominaed economy. Second, his paper exensively employs he analysis of value chains and governance srucure, which are criical elemens in undersanding efficiency and disribuional effecs. Third, alhough he lieraure on asymmeric price ransmission has increased in he recen pas, his is one of he few sudies ha apply hreshold coinegraion models o coffee price ransmission. To he bes of my knowledge, his is he firs sudy ha applies hreshold coinegraion o a supply response model. Moreover, his 6

22 is also he only sudy ha uses srucural break uni roo ess in he analysis of coffee price ransmission. 1.4 Hypoheses To achieve he sudy objecives, various hypoheses concerning coffee price ransmission and supply response have been formulaed based on previous heoreical and empirical findings in he economic lieraure. The hypoheses are lised below: i. Marke liberalisaion policies induce closer coinegraion relaionships beween coffee producer prices and world prices, bu lead o high price volailiy. ii. Considering he high concenraion of firms in he inernaional coffee markes, coffee price ransmission from he world marke o producers is asymmeric where agens pass on price decreases faser han price increases, such ha negaive shocks o coffee prices are more persisen han posiive shocks. iii. Negaive price shocks induce higher price volailiy a producer level han a world marke. iv. An increase in coffee prices lead o an increase in coffee supply, bu he response is asymmeric, where supply responds more o negaive price shocks han o posiive price shocks 1.5 Daa Sources The main source of daa was he Inernaional Coffee Organisaion (ICO). The ICO was se up in London in 1963 under he auspice of Unied Naions because of he grea economic imporance of coffee. As he main iner-governmenal organisaion for coffee, one of he aciviies of ICO is providing informaion on he world coffee secor by means of saisics and marke sudies (ICO, 2010). The organisaion collecs average price daa from member counries and compiles daily, monhly, quarerly and annually. Addiional daa was obained from Food and Agriculure Organisaion of he Unied Naions (FAO), Zambia Cenral Saisics Office (ZCSO), Tanzania Cenral 7

23 Saisics Office (TCSO) and he Zambia Coffee Growers Associaion (ZCGA). Comprehensive background informaion on coffee producion in Zambia and in he region was obained during a research say a he Coffee Board of Zambia and he ZCGA in Sepember Coffee Producion in Tanzania and Zambia- an Overview Tanzania is locaed in Easern Africa bordering Kenya and Uganda in he norh, Zambia, Malawi and Mozambique in he Souh and Congo DR, Rwanda and Burundi in he wes. To he eas lie he Indian Ocean and he islands of Pemba, Zanzibar and Mafia. Tanzania s populaion is esimaed o be 30 million, 80 per cen of which is engaged in agriculure. Agriculure remains he backbone of Tanzania s economy accouning for 50 percen of GDP and 40 percen of he expors. As he larges secor in he economy, agriculure has a significan effec on naional revenues, household incomes and povery levels. The secor is dominaed by small-holder farmers culivaing average farm sizes of 0.9 hecares o 0.3 hecares. Approximaely 5.1 million hecares (85 percen of he oal arable land) is under food crop producion composed of mainly maize, sorghum, whea, mille, rice, beans, planain (bananas), poaoes, and cassava. Tanzania also produces a variey of agriculural expor crops such as coffee, coon, cashew nu, obacco, sisal, ea and some horiculural crops. Coffee is he main expor crop being he counry s second larges expor afer he mining indusry. Tanzania is acually he fourh larges producer of coffee in Africa afer Ehiopia, Uganda and Ivory Coas. Coffee, which was inroduced as an esae crop in he 1920s, is now largely culivaed by Tanzania s small-scale farmers. I is esimaed ha more han 400,000 small-holder farmers are responsible for growing 94 percen of Tanzania s coffee, and derive mos of heir livelihoods from coffee. Approximaely, 160,000 hecares of land is under coffee culivaion in Tanzania s main coffee growing areas of Kilimanjaro, Arusha, Mbeya, Kigoma and Kagera (Newmann, 2006). Robusa coffee, which makes up approximaely 30 percen of Tanzania s coffee producion, is mainly grown in Kagera, while he res of he regions grow Arabica. Up o 75 percen of Tanzania s coffee is expored o Germany, Neherlands and Japan. Despie is connecion o niche markes like Japan, Tanzania s coffee has suffered from very low prices compared o oher counries in 8

24 he region. Tanzania s price rends in he las 20 years have been far below he world prices (see Figure 2.2 of Chaper 2). One of he consequences of low prices has been he abandonmen of coffee producion in preference o maize. As a resul, he counry s coffee expors reduced significanly from 36 percen of he oal expors in 1985, down o 17 percen in 2004 (Piroe e. al., 2006). Turning o Zambia, he counry is locaed in Souh-cenral Africa. Zambia is compleely land locked and covers an area of 752,612 square kilomeres. I is bordered by Boswana and Zimbabwe o he souh, Malawi and Mozambique o he Eas, Tanzania o he norh, Namibia and Angola o he souh wes and Congo o he Norh Wes. Zambia has a populaion of 12 million (2006 esimaes), which is almos hree imes smaller han Tanzania s populaion. Agriculure conribues only 18 per cen o GDP despie employing 60 per cen of he populaion. The economy of Zambia is heavily dependen on is mining aciviies (mainly copper), which consiue 78 per cen of all expors. The lack of economic diversiy subjecs Zambia o economic shocks arsing from flucuaing copper prices. Consequenly, povery levels remain high; wih 63.8 per cen of he populaion living below he povery line (2004 esimaion). Coffee in Zambia is one of he crops ha he Governmen inroduced in he lae 1970s as a non-radiional expor crop in order o implemen he expor diversificaion policy ha aimed a moving away from complee reliance on copper expors. Specifically, he crop was inroduced as an esae crop and is sill largely produced by large-scale farmers, who also go ino conracual arrangemens wih small-scale farmers. I is esimaed ha 99 percen of Zambia s coffee comes from large-scale esaes, while he 150 small-scale farmers only conribue 1 percen. Coffee is mainly grown in he Norhern Province (close o Tanzania) in he high aliude rural disrics of Kasama, Isoka, and Nakonde. Large esaes are also found on he copper-bel province and in he souhern province areas around Mazabuka Disric. Currenly only 3,100 ha of land is under coffee culivaion in Zambia, which is relaively small compared o oher counries in Easern and Souhern Africa. 9

25 Alhough Zambia remains a very small coffee producing counry conribuing only 0.02 percen o he world marke, is coffee indusry is probably he second agriculural indusry wih a large workforce employing beween 17,000 and 15,000 seasonal workers in he rural areas (ZCGA, 2007). In 1984, Zambia was allowed o become a member of he ICO on grounds ha i would grow and expor high qualiy washed Arabica coffee, which was by hen in shor supply (ZCGA, 2007). Zambia s coffee producion rose from abou 1,600 meric onnes in 1995/96 o almos 7,000 meric onnes in 2005/06, bu hen declined significanly o less han 2,000 meric onnes in 2008/09 (see Figure 2.4 of Chaper 2). Various facors have been associaed wih he drasic decline in Zambia s coffee supply including depressed world green coffee prices beween 2000 and 2005, he lack of long erm finance, he appreciaion of he Kwacha (Zambian currency) and a drough in Given ha he coffee rees ake up o four years before he firs harves, he curren low producion could largely be a consequence of farmers low invesmen in planing and crop managemen during he price decline and drough periods. To conclude, he coffee indusry in boh counries seems o face a number of challenges. Firs, being a perennial crop, he revenue is no realised immediaely compared o alernaive cash crops like obacco, coon, beans, groundnus, maize and horiculural producion. Even when he plan becomes producive and remains so for fifeen o weny-five years, farmers are unable o predic fuure prices. As a consequence, some end o uproo he coffee plans replacing hem wih oher crops. Second, since coffee is no produced for he local marke, farmers rely on various supply chains o access he expor markes in he high income counries. The consisence and reliabiliy of hese supply chains remains uncerain o he farmers. Furher, poor infrasrucure in he rural areas, where coffee is mosly produced, leads o high ransacion coss making arbirage difficul as well as hindering he flow of price informaion. There is also a lack of lieraure on he economic aspecs of he commodiy, paricularly for Zambia because coffee is a relaively new and a very small secor; herefore, having no basis for policy formulaion and farmers decision making. 10

26 1.7 Organizaion of he Thesis The nex chaper describes he influence of value chains and chain governance srucures on producer prices in Tanzania and Zambia. This chaper is generally descripive relying on pas sudies as well as informaion gahered during he auhor s four-week research say a he Coffee Board of Zambia and he Zambia Coffee Growers Associaion. Chaper 3 examines price ransmission beween coffee producer prices for Tanzania and Zambia and world coffee producer prices over a period of 20 years using monhly price daa. Chaper 4 is dedicaed o price volailiy analysis for he wo counries. This chaper focuses on he influence of marke policies on price volailiy. In order o have a wider view of volailiy in oher markes, he chaper compares he volailiy a he producer level wih he volailiy a he reail level using coffee reail prices in Germany. Monhly price daa is also used for his analysis. Chaper 5 examines coffee supply response o coffee price insabiliy in Zambia. Oher variables ha play a role in coffee supply such as prices of alernaive crops, real exchange raes and economic reforms are also included. The chaper generally focuses on Zambia, which has high price insabiliy and a large price range over he period under consideraion. Chaper 6 concludes and makes some recommendaions for policy formulaion. 11

27 References Bacon, C. (2004). Confroning he Coffee Crisis: Can Fair Trade, Organic, and Specialy Coffee Reduce Small-Scale Farmer Vulnerabiliy in Norhern Nicaragua. World Developmen, vol. 3, No. 3, pp Baffes, J (2005). Tanzania s Coffee Secor: Consrains and Challenges. Journal of Inernaional Developmen, vol. 17, pp Enders, W (2004). Applied Economeric Time series. Second ediion. Willey series in Probabiliy and Saisics. Book published by John Wiley and sons. USA. FAO (2010). Medium-Term Prospecs for Agriculural Commodiies. Available a hp:// Rerieved on 3rd March, Fier, R. and R. Kamplinsky (2002). Who Gains from Produc Rens as he Coffee Markes becomes more Differeniaed. A value Chain Analysis. IDS working paper. Forenburg, R.T. and H.O Zapaa (2004). Developed Speculaion and Under Developed Markes The role of fuures rading on Expor Prices in Less Developed Counries. Saff paper No Universiy of Agriculural and Applied Economics. Gereffi, G., J Humphrey and T. Surgeon (2005), The Governance of Global value Chains. Review of Inernaional Poliical Economy Vol. 12(1) Gilber, C (2006). "Value Chain Analysis and Marke Power in Commodiy Processing wih Applicaion o he Cocoa and Coffee Secors." UNCTAD PRESS in /2007/017. Glynn, J., N. Perera and R. Verma (2007). Uni Roo Tess and Srucural Breaks: A survey wih Applicaions. Universiy of Pablo Olavide, Seville. Houck, J.P. (1977). An Approach o Specifying and esimaing non-reversible Funcions. American Journal of Agriculural Economics, Vol. 62, pp

28 Humphrey J. and H. Schmiz (2001). Chain Governance and Upgrading: Taking Sock, in Huber Schmiz (2004) Local Enerprises in he Global Economy. Issues of Governance and Upgrading. Insiue of Developmen Sudies, Universiy of Sussex, UK. Inernaional Coffee Organisaion (2007), Prices paid o growers. available on hp:// downloaded on 14/08/07 Inernaional Coffee Organisaion (2010), hp:// Rerieve on 14h February Kaplinsky, R. and M. Morris (2001). A handbook for value marke chain research. Canada: IDRC. Kaplinsky, R. (2004). Compeiions and Policy and he Global Coffee and Cocoa Value Chains. Paper prepared for Unied Naions Conference for Trade and Developmen (UNCTAD). Kaplinsky, R. (2006). How can Agriculural Commodiy Producers Appreciae a Grea Share of value Chain Incomes? Chaper Published in Sarris Alexander and Hallam David, Agriculural Commodiy Markes and Trade: New Approaches o Analysing Marke Srucure and Insabiliy. Published by FAO and Edward Elger, Ialy. Krivonos, E. (2004). The Impac of Coffee Marke Reforms on Producer Prices and Price Transmission. World Bank Policy Research Working Paper Lee, J. and M.C Srazicich (2003). Minimum LM Uni Roo Tess wih Two Srucural Breaks, Review of Economics and Saisics, 63, pp Liunga, L.M (2005). The Saus of Conrac Farming and Conracual Arrangemens in Zambian Agriculure and Agribusiness. Paper prepared for FARNPAN. Universiy of Zambia. 13

29 Lumsdaine, R.L. and H. Papell (1996). Muliple Tread Breaks and he Uni-Roo Hypohesis. The Review of Economics and Saisics, vol. 79, no. 2, pp Narayan, P. K. and R. Smyh (2005). Srucural Breaks and Uni Roos in Ausralian Macroeconomic Time Series. Pacific Economic Review- The Blackwell Publishing, vol. 10(4), pp Newman, S. (2005). Commodiy Marke Srucures and Marke Deerminaion. Oxfam (2002a). Mugged Povery in your Coffee Cup, Oxfam Inernaional, Oxford. Pelupessy, W. (2001). Marke Failures in Global Coffee Chains. Paper presened a he conference on The Fuure of Perennial Crops. Yamoussoukro, Ivory Cos, November 4-9, Perron, P. (1989). The grea crash, he oil price shocks, and he uni roo hypohesis, Economerica, 57, pp Piroe, G., G. Pleyers and M. Poncele (2006). Fair Trade Coffee in Nicaragua and Tanzania: a Comparison. Pone, S (2004). Sandards and Susainabiliy in he Coffee Secor: A Global Value Chain Approach. Documen of UNCTAD and IISD. Pone, S. (2002a). Lae Revoluion? Regulaion, Markes and Consumpion in he Global Coffee Marke. World developmen, Vol. 30 no. 7, pp Schmiz, H., (2004). Local Enerprises in he Global Economy. Issues of Governance and Upgrading. Insiue of Developmen Sudies, Universiy of Sussex, UK. Slob, B. (2006). A Fair Share for Smallholders; A value chain analysis of he Coffee Secor. SOMO- Cenre for Research on Mulinaional Corporaions. Amserdam. 14

30 Swinnen, J. F. M., A. Vandeplas and M. Maerens (2007). Governance and Surplus Disribuion in Commodiy Values Chains in Africa. Paper Presened a he 106h Seminar on Pro-poor developmen in Low Income Counries: Food, agriculure, Trade and Environmen. Monpellier, France, Ocober 25-27, Talbo, J.M. (1997a). Where Does Your Coffee Dollar Go? The Division of Income and Surplus along he Coffee Commodiy Chain. Comparaive Inernaional Developmen Vol 32(1) pp USAID (2006). Direc Expor of Premium Coffee from Tanzania. Changes in Markeing regulaions creaes opporuniies for small scale famers. Accessed on v. Cramon, T. and J. Meyer (2000), Asymmeric Price Transmission: Fac or Arefac? Working Paper. Universiy of Göingen, Insiue of Agriculural Economics. Warako T.K, van Schalkwyk, H.D and Ayele, G (2008), Producer Price and price ransmission in a deregulaed Ehiopian coffee marke, Agrekon, vol. 47, no. 4 pp Zambia Coffee Growers Associaion (ZCGA) (2007), The Coffee Indusry in Zambia. Saus Quo, Sepember Lusaka, Zambia. Zivo, E. and K. Andrews (1992), Furher Evidence on he Grea Crash, The Oil price Shock, and The Uni Hypohesis, Journal of Business and Economic Saisics, 10(10), PP

31 Chaper 2 Policy reforms and Governance Srucures in Coffee Value Chains: A sudy of Zambia and Tanzania Absrac Coffee, one of he mos governed commodiy value chains, demonsraes a high asymmeric power srucure which has raised conenious debae wih regards o ren disribuion, producer prices and consequenly producer welfare. This paper, applies he concep of Governance in Value Chains o undersand how coffee value chains are coordinaed and how his can have a srong bearing on small scale producer prices in Africa. The sudy noes ha coffee producers in Zambia and Tanzania find hemselves in capive relaions, where lead firms (he roasers) se he rules under which he growers operae. The cos of swiching o oher buyers is high, and moving verically up he value chains is hindered by barriers se by he acors a hose levels. In heory, high coordinaion leads o unequal profi disribuion along he chain, while a large number of inermediaries in such consumer-driven chains lead o lower producer prices. According o he findings of his sudy, he highly coordinaed and more complex value chain in Tanzania s coffee secor dominaed by small-scale farmers, explains Tanzania s low producer prices o a large exen. In conras, a less complex value chain governance srucure in Zambia s case has enabled he farmers o receive high prices. The paper concludes ha, he curren wo-ier privae cooperaive union srucure for small-scale farmers in Tanzania, if well managed, can raise coffee incomes hrough value adding and a sronger bargaining power. This would resul in more balanced power symmery along he chains. Key Words: Value Chains, Governance, Coffee Producer Prices, Zambia, Tanzania 16

32 2.1 Inroducion Coffee is an imporan foreign exchange earner in mos African counries. In Zambia and Tanzania, like in many oher producing counries, he coffee secor has undergone significan srucural changes following economic liberalizaion. This has brough abou shifs in he conrol and coordinaion of he value chains, fundamenally moving away from governmen conrolled chains, hrough various markeing boards, o privae secor driven chains hrough Transnaional Cooperaions (TNC). Whils hese changes have led o more efficien markes in he case of Zambia, where 99 percen of he crop is produced by large-scale farmers, here has been disappoining secoral performance in Tanzania, where 94 percen of he crop comes from small holder farmers (Bargawi,2008). Marke liberalizaion in Tanzania has induced a privae secor influx in he coffee indusry giving rise o he number of inermediaries in he value chain, each demanding a share of he expor price. In some cases, he removal of markeing services by he cooperaives has lef even he smalles farmer o deal wih large TNC, and a he same ime exposing producer prices o global price volailiy (Kaplinsky, 2006). In conras, while coffee reail prices in he imporing counries have been escalaing, producer prices have no increased a he same rae, creaing a widening gap which has raised wide concerns among researchers (e.g. Pone, 2004; Kaplinsky, 2004; Slob, 2006). However, he effecs of he new global and local value chain governance srucures on coffee producer prices remains less invesigaed. As Schmiz (2001) explains; here is limied knowledge on how rade is organized and how his affecs he producers. The curren endency owards firm concenraion in he high income counries, he differeniaion of producs, he increased number of sandards and he new markeing sraegies have araced aenion in coffee value chain sudies (see e.g. Pone, 2004: Slob, 2006). These facors have been associaed wih globalisaion, which has seen TNC moving from local o global sourcing. Governance, which describes boh power relaions in commodiy value chains and he insiuions which mould and wield his power, has, herefore, become a disincive feaure in value chains (Moris and Kaplinsky, 2004). According o Swinnen e. al. (2007), he governance of food markes and commodiy chains is a crucial elemen for efficiency and disribuional 17

33 effecs, and ha chain governance becomes endogenous in an environmen of weak enforcemen and imperfec markes. The coffee value chain is among he highly governed commodiy chains due o a number of facors. Firs, is marke is highly concenraed especially a he rading and roasing sages, creaing imperfec markes. 1 Gilber (2007) observes ha, while he coffee value chain is relaively simple, here is considerable concenraion a he laer sages of he value chain giving rise o poenial exercise of monopoly and monopsony powers. Second, as chain governance largely depends on he value in he chain (Swinnen, 2007), i is likely o be high for coffee given is value relaive o oher agriculural expor commodiies. Third, coffee has a producion-consumpion paern whereby producion akes place mainly in developing counries while consumpion is mainly in rich counries. 2 This implies ha coffee producers depend on esablished supply neworks o access he markes in consuming counries. Mos coffee producers, especially in Africa are predominanly small-scale farmers, culivaing on less han 5 hecares. Farmers are mosly locaed in isolaed remoe areas (Pone, 2002a, 2002b), such ha a direc link beween producers and consumers in rich counries does no exis. Given heir small scale, hey are required o sell hrough a complex of inermediaries ha includes local raders, exporers, inernaional raders, roasers and reailers before geing o he consumer. In view of he complexiy of he coffee value chain and he increasing fragmenaion in he geographies of coffee producion, chain governance has srenghened in he las wo decades (Bacon, 2004). The disribuion of rens along he chain and subsequen effec on he incomes and welfare of small-scale farmers as well as on he expor revenues of he producing 1 By 2007, almos 40% of he global rade was being conrolled by four companies: Neumann Kaffee Gruppe (A German Group wih operaions in 17 coffee producing counries): Volcafe Agroindusrial Corp Ld (A Swiss- Spanish Group wih commercial operaions in 13 coffee exporing counries); Dreyfus (a Global Conglomerae Group wih operaions in more han 53 counries and is engaged in processing, rading and merchandising a range of agriculural commodiies). A Roasing sage, 4 roasers conrol 45% of he global marke: Nesle SA (Swizerland), Kraf Foods Inc (USA) Procer & Gamble (USA) and Sara Lee Corporaion (USA) which also brands as Douwe Egbers in Europe. 2 Europe accouns for 40% of he global coffee demand. The US accouns for 24% while Japan jus over 10% (Fier and Kaplinsky, 2001). More han 70% of he world s coffee is produced in Lain America, Asia and Africa (Oxfam, 2001). 18

34 counries has raised concerns among researchers in recen years (see e.g Bargawi 2008; Pelupessy, 2001; Kaplinsky, 2004; Oxfam, 2001; Pone; 2002a; 2002b; 2004; Slob, 2006). Mos sudies argue ha he larges shares of value chain rens accrue o he chain acors ouside he producing counries, mainly o inernaional raders, roasers and manufacurers/reailers (see Oxfam, 2002a, Bacon, 2004). While an average of 20 percen of oal income was reained by producers in he 1970s, in 2003/2004 only an average of 13 percen was reained and he remaining 78 percen accumulaed in he consuming counries (Pone, 2004). Large amoun of work has been done on coffee markes, especially on he effecs of deerioraing and volaile prices on he welfare of he producers (e.g. Pone, 2004; Kaplinsky, 2004; Slob, 2006). However, lieraure focusing on issues of governance in coffee value chains, which may well provide insighs ino price movemens and he disribuion of rens along he chain and consequen effecs on producer welfare, is limied. The sudy by Muradian and Pelupessy (2005) aemped o examine governance in coffee value chains focusing on he role of volunary regulaory sysems. The curren paper focuses on changes in governance srucures in coffee value chains and how his affecs ren disribuion along he chain, and subsequen effecs on small-scale producer prices. The paper applies a global value chain approach because i enables us o decompose he oal value chain reurns ino hose arising from, for example, producion, markeing, roasing and reailing. Essenially, we are able o explain how he reurns accrue o which acors in he chain, and why. We hypohesise ha, governance in value chains is a concealed facor, ye i has significan bearing on ren disribuion and subsequenly on producer prices. Because i is a concealed facor in value chains, we observe ha governance has been less invesigaed especially wih regards o he coffee secor. As Newman (2005) explains, he undersanding of facors behind changes in disribuion of rens along commodiy chains is criical o undersanding he mechanisms of price deerminaion in commodiy markes. Similarly, Fier and Kamplinsky (2001) argue ha, enry ino global markes ha allows for susainable income growh requires knowledge of dynamic facors wihin he whole value chain and no only profi margins. 19

35 For his sudy, we refer o wo counries in Easern and Souhern Africa: Tanzania and Zambia. The raionale behind he selecion of he wo counries is o compare he ypes of value chains and governance forms beween wo differen scales of coffee producion. While Tanzania s coffee is predominanly grown by small-scale farmers, Zambia, a neighbouring counry o Tanzania, has 99 percen of he coffee grown on large esaes. In fac, Zambia is a unique case in Africa where coffee producion has barely been adoped by small-scale farmers. Second, Coffee markeing in Zambia is compleely liberalised where producers sell direcly o roasers. On he oher hand, alhough Tanzania had liberalised is domesic coffee markes in he mid-1990s, cooperaive unions sill dominae he indusry, wih high governmen regulaion hrough he Tanzania Coffee Board (TCB). The TCB is also mandaed o conduc all coffee aucions (see Pone, 2002; Baffes, 2005). Examining he wo counries, wih differen scales of coffee producion and differen degrees and rajecories of marke liberalisaion, ensures a wider viewpoin. The res of he paper is organised as follows. Secion wo draws aenion o he heory of value chains and governance in order o provide a background for undersanding how coffee value chains operae and how hey are governed. A descripion of inernaional coffee markes and governance in value chains is hen given in secion hree. In secion four, we discuss Zambia and Tanzania coffee producion and markeing focusing on how hey fi ino he global coffee value chains and how hey are governed. In secion five, we discuss and make conclusions on some consrains and opporuniies for coffee producers based on how he value chain and governance heory relaes o he siuaion in coffee markes in he wo counries. Secion six provides some policy recommendaions. 2.2 Background Informaion on Value Chain Analysis and Governance The Value Chain Theory The lieraure on value chains has expanded considerably boh empirically and heoreically in las wo decades. The heory of value chains can be raced back from Wallersein s concep of world sysems approach o economic analysis (Wallersein 1974). Laer, Porer (1985), ook an inra-firm approach and inroduced a new area of 20

36 aenion, focusing on inerrelaions beween inra-firm aciviies and resources ha bring a produc o is final form. According o Porer s definiion, value chains are sysems of inerrelaed economic aciviies wihin a firm. Based on Porer s heory, economiss embraced an iner-firm approach involving an analysis of whole range of aciviies and acors from producion o consumpion. A broader approach was necessary as more firms began o specialise in specific producion aciviies linked o aciviies by oher firms. In view of hese developmens, researchers like Gereffi (1994) suggesed a Global Value Chain (GVC) approach, which has formed a foundaion for mos curren value chain analyses. For example, Gilber (2006) defines a value chain as consising of full range of inerrelaed producive aciviies performed by firms in differen geographical locaions o bring ou a produc or a service from concepion o complee producion and delivery o he final consumers. A global value chain, herefore, consiss of muliple business parners across counries ha add on o he value ha is ulimaely presened o he buying public (Gereffi, 1994). Value chain analysis considers issues of marke power, regulaion and supply chain resrucuring o sudy heir implicaions upsream. The difference beween a value chain and a supply chain analysis is ha he value chain analysis is concerned wih he added value a each node of he chain and how he acors inerrelae in adding value. Supply chain analysis, on he oher hand, is only concerned abou how goods move from one acor o anoher. Gereffi (1994) and Gereffi e. al. (2005) idenify hree key dimensions of commodiy chains: i) he inpu-oupu srucure and geographical coverage; ii) he form of governance; and iii) he insiuional framework. While he inpu-oupu srucure and he geographical srucure help us undersand he processing of he commodiy and ransacion coss, chain governance explains he level of firm concenraion, providing insighs ino marke power Theory of Governance in Value Chains Governance in value chains is bes described in erms of conrol and coordinaion of aciviies in he chain. The quesion of who coordinaes and conrols he value chain 21

37 and how he chains are coordinaed and conrolled could parly be undersood hrough he wo broad caegories of value chains: i) producer-driven value chain, in which he key governors are producers (suppliers) embedded in he producion chain and commanding core echnologies and; ii) he buyer-driven value chain, where he reins of power are held by he key buyers also referred o as lead firms (Gereffi, 1994; 2005). These caegories deermine he naure of he access of producers o final consumers. While producer-driven chains focus on aaining economies of scale, buyer driven chains are characerised by dominaion of reail companies and brandnamed merchandise. The key buyers or lead firms deermine he naure of access of producers o final consumers. Hence hey are he governors of he chain. The decisions of lead firms (governors) creae winners and losers in an indusry. For example, research on he UK-Africa horiculure chain suggess ha small growers are marginalised no because of he efficiency advanage of large growers bu because of he lead firms sourcing sraegies (Oxfam, 2002b). Traceabiliy is criical for he lead firms if hey are o mee healh, safey, environmenal, and labour sandards demanded by consumers, NGOs and governmen agencies (Dolan and Humphrey, 2000; 2004). Small farmers ofen canno ge ino hese expor markes because hey canno mee he demand for raceabiliy. According o Humphrey and Schmiz (2001), he concep of governance is cenral o he global value chain approach as he erm is used o express ha some firms in he chain se and/or enforce parameers under which ohers in he chain operae. These parameers are: 1) wha o produce? Referring o produc design and specificaions; 2) how o produce i? This involves he definiion of producion process, which can include elemens such as he echnology o be used, qualiy sysems, labour sandards and environmenal sandards; 3) how much o produce? and; 4) when o produce? This is basically producion scheduling and logisics. These parameers help o undersand he influences of chain governance on producer aciviies. In view of variaions of ypes of value chains and he chain acors, disinguishing beween forms of governance (ype of relaionship) ha exiss wihin a paricular chain becomes vial in value chain analysis. Gereffi e. al. (2005) and Schmiz (2004) disinguish beween four ypes of chain relaionships in global value chains: The firs is arm s lengh marke relaions, where enerprises deal wih each oher in arms 22

38 lengh ransacions. Maradian and Pelupessy (2005) posied ha his form of governance is no more han marke ransacion where coordinaion is low or missing. There is low informaion exchange mediaed only by prices and sandard aribues of producs. Paries can easily swich o oher commercial parners because he cos of swiching is very low (Gereffi, 2005). However, high power asymmery in he case of oligopoly and monopsony is likely o exis in marke ransacions, even when here is no coordinaion (Maradian and Pellupessy, 2005). The second caegory is balance neworks, where enerprises co-operae and have complemenary compeences bu have no conrol over each oher. According o Maradian and Pelupessy (2005), balance neworks are characerised by low monioring coss for buyers and low coss of swiching o oher commercial parners (boh for buyers and suppliers). The hird caegory is capive neworks in which lead firms se parameers under which ohers in he chain operae his relaionship is quasi-hierarchical (Gereffi e al, 2005). According o auhors, suppliers in capive neworks face significan swiching coss and are, herefore, held capive. These relaionships are characerised by a high degree of monioring. Besides, capive suppliers are confined o a narrow range of asks and are dependan on he lead firm for complemenary aciviies. In his case muual dependence is likely. The fourh caegory of chain relaionships is hierarchy governance, where he lead firm akes direc ownership of some operaions in he chain. The case of inra-firm rade beween ransacional companies and heir subsidiaries falls ino his caegory. Recenly a erm referred o as homologaion has been closely linked o global governance in value chains. The erm refers o a sysem, where uniform global rules are applied o he small-scale producers by he lead firms. These rules include qualiy measures, specific grades of producs and environmenal sandards. Kamplinsky (2004) argues ha hese rules govern he inegraion of small producers ino he global value chain. I is also imporan o noe ha, in addiion o lead firms governance, here are several agens exernal o he chain ha regulae produc design and manufacure, no only wih a view o consumer safey, bu also o creae ransparen markes (e.g. by defining sandard weighs and sizes, echnical norms). I is argued ha, where clearly defined sandards and sysems are enforced (for example, cerificaion 23

39 sysems) he need for governance by lead firms is reduced (Swinnen, 2007). Insead of monioring performance along he chain, buyers can rely upon exernal monioring and verificaion o guaranee produc and process sandards. In many cases, nework acors conrol opporunism hrough he effecs of repea ransacion and repuaion ha are embedded in paricular geographic locaion or social groups (Gereffi e. al, 2005) Theory of Income Disribuion in Value Chain The heory of income disribuion along value chains is adequaely explained in Schumpeer s heory of enrepreneurship and barriers o enry (Schumpeer, 1961). According o Schumpeer s heory, he abiliy o insulae aciviies can be encapsulaed by he concep of ren, which arises from he possession of scarce aribues and involves barriers o enry and ha scarciy can be consruced hrough purposive acion. In ha way, an enrepreneurial surplus can accrue o hose who creae his scarciy. This implies ha, he primary reurns o value chain rens accrue o hose paries, who are able o proec hemselves by creaing barriers o enry. Thus, income growh can be susained hrough an enduring barrier o enry for some chain aciviies. This is closely relaed o roasers and manufacurers in coffee value chains. The changes in governance srucure of he chain are a resul of sruggle for rens as acors aemp o increase he income and profi derived by paricipaion (Muradian and Pelupessy. 2005). Thus he degree of verical coordinaion in supply chains influences economic oucomes, in paricular efficiency and equiy (Swinnen e. al, 2007). Besides, his could also lead o a large value added accumulaing in hose segmens obaining a lower share of he oal reail value hrough producer upgrading. Rens can be relaed o produc and markeing, echnology, financial, resource, infrasrucural and policy (Newman 2005). In his sudy, he focus is on governance in coffee value chains because i deermines he exen o which producers can paricipae in he chain. For example, involvemen in roasing adds value o green coffee. Schmiz (2005) explains ha working o he specificaions of large global buyers ofen provides a fas rack o upgrading processing and producs. This subsequenly influences he price of a commodiy. In addiion, he forms of governance in value chains deermine he speed and 24

40 magniude of price ransmission along he chain. As Norh (1995) argues, price formaion and ransmission are no only influenced by supply and demand bu also insiuional srucures. If marke insiuions allow efficien ransmission of price, producers can hen ake advanage of he opporuniies o increase heir producion. Therefore, he analysis of forms of governance in value chains esablishes he indusry s araciveness and provides insighs ino how prices will evolve in he fuure. 2.3 Coffee Inernaional Markes Coffee is possibly he mos raded perennial crop in he world. The commodiy is broadly caegorised ino wo species: Arabica (Coffea Arabica) and Robusa (Coffea Conefora). The beans of he wo varieies differ considerably in ase and flavour. Arabica bean is considered o produce beer qualiy coffee han he Robusa bean, and herefore, feches a higher price (Figure 2.1). Figure 2.1: Arabica and Robusa coffee prices (1980 o 2009) Source: Own Presenaion based on Daa from ICO(2010). 25

41 The wo species also vary in heir agro-ecological requiremens. While Arabica coffee grows beer in higher aliudes of semi-ropical climaes, Robusa coffee is mainly produced in low laying areas in ropical climaes. Arabica is furher differeniaed ino hree flavours; Brazilian Arabicas, Columbian Milds, and Oher Milds. Each ype makes differen blends and feches differen prices. For example, Zambia s Oher Milds is used as a asemaker and i is highly priced (Pelupessy, 2001). Cerified coffee such as Fair Trade, Organic, Uz Kapeh, Bird Friendly, Rainfores Alliancecerified and many oher cerificaions, have slowly gained marke shares in he coffee world marke. The success of specialiy coffee on he world marke is largely associaed wih new consumpion paerns, such as ehical consumerism, ha have emerged in he indusrialised counries (Pone, 2004). Coffee is produced in 85 counries in Africa, Asia, Lain America and Ausralia. Brazil, which produces Arabica coffee, is he larges coffee producing counry in he world accouning for 31.5 percen of he oal volumes of coffee produced globally based on 2009 esimaions by he Inernaional Coffee Organisaion (ICO) daa 3. By he year 2000, Vienam emerged o be he second larges coffee producing counry in he world afer Brazil, surpassing mos Lain American counries such as Cosa Rica and Columbia. Currenly, Vienam produces 17.7 percen of he oal global producion according o ICO producion daa and i only produces Robusa coffee. Ehiopia, he larges coffee producer in Africa accouns for only 2 percen of he oal global producion. Tanzania has only 1.2 percen, while Zambia has 0.02 percen share of he oal global producion. Coffee consumpion is mainly in he indusrialised counries. The Unied Saes of America (USA) is he larges consumer of coffee accouning for 18 percen of he oal global consumpion. Brazil, he second larges, consumes 13 percen of he global coffee consumpion. Germany is he hird larges wih 9 percen share of all he coffee consumed worldwide. Alhough, he demand for coffee has sagnaed over he las few years, demand for high qualiy coffee has increased considerably (Slob, 2006). The laer observaion is sill growing across he secor bu i is esimaed o be less han 10 percen of he oal coffee produced on he world marke. Recenly, here has been a growing demand for specialy coffee by China as a resul of a growing 3 downloaded on 4h February,

42 middle class. The Chinese demand was expeced o increase he world demand for coffee by 70 percen by 2008 (Slob, 2006). In he las wo decades, coffee exporing counries have faced a crisis as a resul of low world producer prices. The world price of green coffee beween 1998 and 2002 dropped from US$1.20/lb o beween US$0.40 and US$0.75 (Bacon, 2005). In February 2002, he prices reached he lowes levels since 1930s. I is esimaed ha he price fell by 25 percen beween lae 1990s and early 2000 s (Common Fund for Commodiies, 2000). In conras, reail prices of coffee in high income counries increased over he same period. As he coffee bean price fell by 25 percen, he margins of coffee roasers rose by more han 50 percen (Common Fund for Commodiies, 2000). This rend is aribued o increased marke power, which originaes from firm consolidaion a manufacuring and reail sages resuling in oligopoly markes in he coffee indusry (see e.g. Fier and Kamplinsky, 2001; Slob, 2006). I is ofen argued ha, as a resul of loss of bargaining power by he farmers following he abolishmen of he cooperaive boards, incomes which accrued in he producing counries have been ransferred o rich counries (Kamplinsky 2006, Pone, 2004). Similarly, coordinaion and conrol of coffee supply shifed from he governmen-run cooperaives o mulinaional companies (See Swinnen e. al. 2007) The Inernaional Coffee Value Chain Unlike many oher agriculural commodiies, mos of he pos-harves processing of coffee akes place on he farm. The farmers harves red (ripe) coffee cherries which are eiher we processed or dry processed. We processing is ypically applied o Arabica coffee. I involves soaking and fermening he cherries in order o remove he coffee bean. The bean is hen dried in he sun o a cerain moisure level. Tanzania and Zambia boh apply we processing, which yield a beer qualiy. The coffee bean is hen sold as parchmen coffee o privae buyers, cooperaives (as in he case of Tanzania) or direcly o exporers or roasers based in he consuming counries. The major players in he global coffee value chain are producers, cooperaives or associaions, local raders, inernaional raders (exporers), imporers/roasers, manufacures/reailers and he consumers. Mos of he inermediaries like local 27

43 raders, cooperaives or exporers do no add any significan value o he coffee bean oher han incurring ransporaion and oher ransacion coss. I is no unil he bean reaches he roasers ha i is roased and blended wih oher varieies according o consumer preferences. The roasers hen sell he coffee eiher as roased beans or as ground coffee o manufacurers/reails or direc o coffee oules such as supermarkes, resaurans and cafés. Coffee processing differs from several oher commodiies like coon, in ha he large par of coffee processing is done by he farmers hemselves. However, several sudies have shown ha despie mos of he processing aking place on he farm, he producers receive a very small share of he reail price, such ha in some cases i is only 9 percen (e.g. Oxfam, 2002b; Kaplinsky, 2004). The sudies show ha approximaely 30 percen of he value of he reail price goes o he roasers. The local raders and exporers ogeher ge approximaely 15 percen share of he reail price while reailers may reain up o 15 percen. Figure 2.2 shows coffee prices a producer level for Zambia and Tanzania in comparison o world producer prices and reail prices aking he Germany reail prices. Figure 2.2: Producer and reail coffee prices in US$ cens per pound WORLD TANZANIA ZAMBIA Source: Auhor s Presenaion based on Daa from ICO 28

44 2.3.2 Governance in Inernaional Coffee Value Chains The Inernaional Coffee Agreemen (ICA), which was basically a se of agreemens on producion and consumpion quoas, governed he global coffee indusry for mos of he period from 1962 o The primary objecive of he ICA was o raise and sabilise prices in member producing counries ha comprised 99 percen of he coffee exporers. The idea was o raise prices in consumer counries which subsequenly raised prices in producing counries (Bohman and Jarvis, 1990). When he producer price indicaor calculaed by he ICO rose over he se price, quoas were relaxed; when i fell below he se price, quoas were ighened. In some cases when prices rose exremely, quoas were abandoned unil prices dropped back wihin a cerain band (Pone, 2001b). I is ofen argued ha he breakdown of he ICA in 1989, combined wih he collapse of governmen-run cooperaives in mos developing counries following liberalisaion policies, led o he drasic drop of coffee producer prices in he early 1990s (Bacon, 2004). The prices improved in he period , and hen plummeed again in he early 2000s before surging during he second half of he las decade. Brown e.al (2008) argue ha alhough he ICA was relaively successful for wo decades wih acive supply managemen objecives, i evenually succumbed o common flaws such as insufficien financing and unrealisic price arges in exended periods of low prices and increasing supplies. During he ICA regime, he global coffee chain was no conrolled or driven by any acor (Pone, 2002b). In principle, neiher he producers nor consumers conrolled he coffee value chain. However, during he pos ICA period, he coffee value chain has mosly exhibied characerisics of a buyer-driven chain. To be more specific, Pone (2002b) refers o i as a roaser-driven chain. Following rade liberalisaion in mos coffee producing counries ha led o he wihdrawal of minisries and governmenowned markeing boards from coordinaing coffee producion, markeing and qualiy conrol, governmens los inernaional negoiaing power (producer drive). Kamplinsky (2004) argues ha he aggregaed producer power which had been refleced in hese markeing boards has weakened, and small-scale producers, who previously linked o final markes hrough he various forms of markeing boards, found hemselves selling direcly ino volaile world markes. Large-scale 29

45 ransnaional raders and roasers quickly moved in o fill he gaps lef by he governmens markeing board, while gaining negoiaing power. Swinnen e. al. (2004) observes ha he combinaion of marke liberalisaion and increased coffee producion coincides wih raes of ransnaional corporaion concenraion. Noably, mos roasers have moved o dealing wih only a few raders who abide by heir condiions. The adapion of supplier-managed invenory (SMI) has been siffened by he requiremens ha raders have o abide by if hey have o qualify o supply coffee o he roasers. These requiremens have led raders o have more supervision over producers. Furhermore, a he roasing sage, here are various sraegic barriers such as seing of minimum quaniies needed for a paricular ype of coffee o be included in a cerain blend or paen righs o cerain processing echnologies ha have been creaed in he las en years (Pone, 2002b). In addiion, he ICO, he Inernaional Coffee Council and several oher regional and naional bodies play an imporan role in seing sandards for coffee producion and markeing. In some counries, he governmen has specific sandards se wih regards o consumer safey (e.g. EurepGAP). 2.4 Tanzania Coffee Value Chains On-farm Processing Coffee in Tanzania is harvesed by hand as red ripe cherries grown on small plos of 1 o 5 acres. The bean is removed from he cherries using a hand pulping machine. The wo halves of he seed, referred o as he coffee beans are hen dried for approximaely 11 days before i is sold as parchmen coffee o he primary socieies. In is green coffee form, he coffee can say up o 12 monhs before i can sar deerioraing in flavour. Roasing reduces shelf life; a reason for mos roasers preferring o carried i ou shorly before consumpion, preferably in he consuming counries. 30

46 2.4.2 Coffee Cooperaives Coffee in Tanzania has a producion and markeing sysem which is hisorically and closely linked o cooperaive movemens. Currenly, Tanzania s coffee indusry is dominaed by a wo-ier cooperaive sysem, whereby farmers from muliple village level primary socieies (groups of 100 o 1,000 persons) form cooperaive unions. The unions have he primary responsibiliy of inpu supply (as loans), free exension services and purchase of produce. In addiion, he unions are responsible for financing, ransporing, markeing and supervising he sale of coffee supplied by heir primary socieies (Parrish e. al., 2005). The cooperaives buy, sore and process coffee using heir own faciliies. Prior o economic reforms (before 1991), all cooperaives were governmen-owned and all small-scale farmers were obliged o belong o one. Presenly, he cooperaives operae as privae eniies, owned and managed by members, and are suppose o compee wih privae raders. Membership of he cooperaives is no mandaory anymore, while a lo of services such as inpu supply and exension services have been wihdrawn (Chachage, 2004; Cooksey, 2004 cied in Mhando and Mbeyale, 2010). Presenly, mos farmers rely on governmen paid exension agens, which has also drasically reduced from hree in 1985 o one per four villages of abou 1,200 households in 2007 (Mhando and Mbeyale, 2010). Afer he primary socieies have purchased a minimum required quaniy of parchmen coffee, i is sold o he cooperaive unions such as he Kilimanjaro Naive Cooperaive Union (KNCU). Afer marke liberalisaion in 1994, privae raders were allowed o buy coffee direcly from he farmers, hereby compeing wih exising cooperaives in he purchase of he crop a village level (Bargawi, 2008). 4 The privae raders purchase direcly from farmers and some of hem even process i in heir own facories before sending i o he TCB operaed aucion. However, despie he marke liberalisaion, only 20 percen of he farmers sell direcly o privae raders, he remaining 80 percen sell heir coffee o cooperaive unions or o small privae local raders (Newman, 2006, Bargawi, 2008). In some areas like Mshiri in Kilimanjaro 4 In he 2001, he Coffee Indusry Ac, privae raders were mandaed o choose only one licence; purchasing parchmen coffee from he farmers, curing or exporing. However, in realiy hese raders have designed a mechanism whereby hey operae a many sages of he along he hrough parner companies (Mhando and Mbeyale, 2010). 31

47 region, 94.1 percen of he farmers sell o cooperaives and only 5.9 percen sell o privae raders (Mhando and Mbeyale). A major reason for he farmers preference o sell heir coffee o cooperaives is ha he cooperaives pay in hree insalmens (Sepember/Ocober, December/January and February/March). Farmers who go for his choice also avoid sorage coss such ha awaiing furher paymen from he cooperaives unions may only be an opion for he mos affluen coffee producers (Bargawi, 2008). Even when privae raders offer a higher price han he iniial price offered by he cooperaives, mos farmers sill prefer he insalmen paymen sysem by he cooperaives because prices end o increase for he second and hird paymens. The second and hird paymens also help he farmers mee oher cash needs laer in he season such as paymen of school fees for heir children or purchase of inpus for he nex season (Bargawi, 2008). In addiion, cooperaives also pay a premium on op of he bes price offered. Bargawi (2008) noed ha he coninued funcioning of he cooperaive unions is seen as medium of sabiliy o he producers wih a more sable producer price. However, in some cases, for example, in Kiruweni and Wanri villages in Kilimanjaro region, privae raders have resored o compee wih cooperaives on qualiy raher han price, such ha he poor grades ha are rejeced by he cooperaives is wha he privae raders buy (Bargawi, 2008). Primarily, he choice o sell o privae raders is linked o urgen cash needs as mos privae raders bough he coffee early in he season, when he prices are sill low Exporing Sage A he cooperaive union level, he parchmen coffee is milled and re-graded before i is aken for curing and re-grading a he curing facories. 5 Recenly, verically inegraed exporers working as subsidiaries of mulinaional companies have emerged in Tanzania and now own all he processing facories in he counry (Temu e. al 2001 cied in Maradian and Perupessy, 2001). A hese curing facories, he 5 The coffee in Tanzania is graded ino classes ranging from 1 o 13. The bes grade is 1 while 13 is he wors grade (Mhando and Mbeyale, 2010). 32

48 coffee is dehulled and re-graded in readiness for he weekly coffee aucion. The TCB organises and faciliaes he coffee aucion and sends graded samples o licensed exporers, before conducing aucions hrough he Moshi Coffee Aucion, a privae company owned by he TCB. The aucion company buys coffee from cooperaives, privae raders or direcly from he famers. The TCB aucion is primarily a markeing agen charging 1.6 percen of he aucion sale (Baffes, 2005). Figure 2.3 shows he value chain srucure in Tanzania Governance Srucures and Price Shares The exporers coordinae and conrol all he value adding aciviies along he value chain wihin he counry on behalf of he large mulinaional roasing companies based in he consuming counries. The exporers decide he qualiy of coffee o buy and how i should be produced. The exporers are he ones who bid a he aucion wih capabiliy o influence he prices. Moreover, i is esimaed ha he larges share of he coffee expor price goes o he exporers as hey are well informed of he price being offered by he roasers (Mhando and Mbeyale, 2010). 6 Alhough he TCB is he main regulaory body of he coffee secor in Tanzania, is monioring of aciviies in he secor is no very effecive (Mhando and Mbeyani). For example, he privae raders are no allowed o buy parchmen coffee from he farmers a he farm gae; insead hey are supposed o se up purchasing poins, where farmers could bring heir coffee. However, his is no he case as he raders sill buy a he farm gae. In addiion, privae raders sar buying coffee even before he buying season is officially auhorized. In 2003/04, he TCB inroduced direc sale of coffee o buyers oversees in order o eliminae inermediaries. However, direc sales seem o apply only o premium coffee or specialy coffee such as fair rade coffee. The res of he producers largely sell o cooperaives and privae raders, who end o make profis wihou necessarily adding any value o he commodiy. Mos farmers canno sell direcly o roasers in developed counries because hey are no able o mee he exporing coss involved, i.e. curing, ransporaion, communicaion, packaging and expor duies. Addiionally, 6 The exporers sell heir coffee o roasers mosly in Germany, Neherlands and Japan. The hree counries consiue 75% of Tanzania s expors. 33

49 issues of raceabiliy become criical if small-scale farmers have o deal direcly wih reailers or consumers. Mos roasers and manufacures in he consuming counry own propery righs for cerain processing echnologies creaing barriers for enry for coffee producers. As privae raders were allowed o purchase direcly from farmers a he village level, he cooperaives ha provided inpus o he farmers los he monopsony purchasing power. Farmers obained subsidised inpus from he cooperaives, bu hen sold heir crop o privae buyers offering higher coffee prices. The deserion by he farmers lef he cooperaives wih large debs, unable o coninue wih he inpu provision services (Bargawi, 2008). This coupled wih he removal of ferilizer subsidies by governmen from 70 percen (of he oal ferilizer requiremen by he farmers) in 1990/91 o zero in 1994/95 drasically reduced ferilizer applicaion and pesicide use. Esimaes show ha, while 51 households used pesicides, insecicides and herbicides in Kilimanjaro region in 1994/95, he number had reduced o 35 by 2002/2003 (Unied Republic of Tanzania, 1996; Unied Republic of Tanzania, 2006 cied in Bargawi, 2008). Consequenly, he qualiy of Tanzanian coffee has significanly reduced following economic reforms (Baffes, 2005). Some researchers relae he declining prices of Tanzania coffee o low qualiy of coffee produced (e.g Baffes, 2005, Pone, 2002). Qualiy is also los hrough processing when small-scale producers use dry processing insead of he we processing which yields beer qualiy Tanzania fair rade coffee Fair rade is a commercial parnership aimed a creaing greaer equaliy in he curren liberalised markes hrough offering a minimum price. A guaraneed minimum price of $1.26/Ib for rade fair coffee was agreed upon beween producers and Fair Trade Organisaions (FTO) in June On op of ha, a price premium of 5 o 10 percen is paid above he fair rade price. The premium price is mean for communiy developmen programs. Despie he fac ha i was he firs counry in sub-saharan Africa where i was inroduced in 1990 However, fair rade coffee represens only a small fracion of he oal coffee producion in Tanzania (Piroe e. al. 2006). The wo larges co-operaives which are acive in fair rade coffee producion are: Kagera Cooperaive Union (KCU) and Kilimanjaro Naive Co-operaive Union (KNCU) accoun for only 5 percen of he counry s coffee expors (Piroe e. al, 2006). The amoun of 34

50 coffee expored o fair rade marke significanly dropped in he las few years from 52,380kg in 2006/07 o 32,760kg in 2007/08 (Mhando and Mbeyale (2010). Alhough all coffee expors in Tanzania have o go hrough he aucion, he TCB recenly adjused some rules o allow fair rade coffee producers o expor direcly o roasers by passing he aucion (Parrish e al, 2005). 2.5 Zambia Coffee Value Chains The coffee value chain in he case of Zambia is less complex han he Tanzania value chain (see Figure 2.3). For Zambia, Parchmen coffee from he farmers is expored direcly o roasers in he consuming counries hrough he Zambia Coffee Growers Associaion (ZCGA). Unlike he case of Tanzania, Zambia produces mainly specialiy coffee and does no rely on he large TNC s roasers for is markes. The producers have esablished relaions even wih some small buyers in he consuming counries. The coffee markeing in Zambia is compleely liberalised and governmen does no inerfere in price seing. All coffee producers (large- and small-scale producers) are, by law, members of he ZCGA which provides markeing, qualiy conrol, milling, warehousing, shipping, exension and secrearial services o he growers. Alhough he ZCGA has been delegaed markeing funcions by he Zambia Coffee Board (ZCB), he former also gives licences o some members ha wan o marke heir own producs. The associaion also issues cerificaes of qualiy o all expor shipmens. The ZCGA is an operaing wing of, and supervised by he ZCB, whose members represen he governmen, small-scale farmers, large-scale farmers, and agriculural research and exension services. On-farm coffee processing in Zambia is similar o Tanzania s we processing. However in he case of Zambia, he parchmen coffee is washed, sored, graded and bagged righ on he farm, before being aken o he ZCGA for aucioning. A he ZCGA, he coffee undergoes milling and re-grading before he samples are sen o he roasers o prepare for he aucion. The grading and expor presenaion is in grades AAA, AA, AB, AB, PB and many small grades making up o welve coffee grades. The grades are based on bean size wih AAA being he larges size achieved hrough using proper field managemen pracices. 35

51 Figure 2.3: Coffee value chain in Tanzanian and Zambia a. Zambia Coffee Value Chain Tanzania Coffee Value Chain Smallholder Farmers (94%) Primary Socieies Large Scale Farmers Privae Traders (20%) Smallholder Farmers (1%) Large Scale producers Cooperaive Unions (80%) (e.g KNCU) Milling, Grading, Bagging e.g Fair Trade Milling/Curing Privae Companies ZCGA (milling, curing, re-grading, aucioning, shipping (FOB)) TCB Aucion (Governmen) (e. g Moshi Coffee Exchange) Cooperaive Unions Privae Exporers/Inernaional Expor Marke Expor Marke Source: Auhor s own design based on discussions wih he Zambia Coffee Growers Associaion (ZCGA) and he Zambia Coffee Board (ZCB) as well as various oher lieraure sources. 36

52 All physical expors are, however, handled by he ZCGA and are usually roued via Durban in Souh Africa or Dar es Salaam in Tanzania using road and rail. As a land locked counry, Zambia uses he pors of is neighbours for exporing is coffee. Ulimaely, a large share of profis is used o pay inernaional ranspor services and duies. Zambia s coffee expors go o Europe (94%), he Unied Saes (1%) and he remaining 5% is expored o Souh Africa, Ausralia, and Japan (ZCGA, 2007). The pricing sysem a he ZCGA organised aucions are based on he rading price a he New York Board of Trade (BOT) 7. A presen, he associaion uses a silen elephone aucion sysem. This sysem, however, is no only expensive, bu also limis he number of bidders. Mos buyers end o know each oher whereby ransparency can be limied. The mos imporan seback of his sysem is ha bidders can collude and avoid high bids. However, Zambia s coffee is one of he highly priced coffees in Africa and porrays closer movemens o he world price as shown in Figure 2.2 above. In 2009, Zambia received he highes producer prices in he whole of Africa (Mafusire e. al., 2010). Tradiionally, Zambian coffee farmers grow a variey of Arabica coffee called Bourbon, which, despie being highly suscepible o pess and diseases, produces high qualiy specialiy coffee. Because of Zambia s lae sar in he coffee indusry, he growers immediaely adoped some laes echnologies for culivaion pracices including composie manual from he coffee pulp and pes managemen hrough chemical and biological mehods. In addiion, hey use sophisicaed mehods for irrigaion, ferilizaion, and chemical applicaion. The applicaion of he appropriae echnologies and good managemen pracices has enabled he counry o produce high qualiy specialiy coffee and o penerae niche markes in Europe, USA and Japan. In 2007/2008 he ZCGA sold 74 meric onnes o specialy markes (ZCGA, 2009). 7 Renamed Inerconinenal Exchange in Sepember

53 2.6 Discussion and Conclusion The paper has analysed value chains and governance srucures in Zambia and Tanzania and how producers access world coffee markes. Despie he differences (in qualiy and value chain srucure) beween he wo counries, which are largely due o differences in scales of producion and levels of marke liberalisaion, here is commonness in he wo counries in ha coffee producers in boh counries feed ino a concenraed world coffee marke. However, he characerisics of governance srucures in Tanzania s value chain reflec ha of capive relaions. In hese ypes of relaionships, lead firms se he rules under which ohers operae and he coss of swiching o oher buyers is very high. The producers, as seen in he case of Tanzania s small-scale producion are confined o narrow asks, which is basically producion of parchmen coffee while he curing companies have dominaed he indusry a processing level. Essenially, here are high barriers for producers o move up he chain. Even if he Tanzania coffee markeing sysem is supposedly a compeiive marke, where cooperaives compee wih privae raders in purchasing coffee a village level, farmers prefer o sell o he cooperaives because he sysem allows hem o access second and final paymen, aking advanage of any seasonal increases in coffee prices. The sraegy by he cooperaives o pay he farmers in insalmens, have earned hem comparaive advanage over he privae buyers. This, in a way, hinders he privae raders from purchasing direcly from he farmers, on he grounds ha hey canno compee wih cooperaive unions on issues of sabilising seasonal prices. Evidenly, beween he firs and second acors in he value chain, i.e.; he famers and he cooperaives; he cooperaives are he price seers. A he second level of he value chain, i.e.; beween he cooperaives and he exporers/inernaional raders, he price seers are he exporers/inernaional raders acing on behalf of roasers ha are based in he consuming counries. An imporan observaion in Tanzania s value chain is he emerging of verically inegraed curing companies and exporers, working on behalf of mulinaional companies, who now own all he coffee processing companies in he counry. The exporers, hrough heir large invesmen in processing machinery, benefi from economies of scale, creaing barriers for producers o engage ino processing. In so doing, hey manage o capure he larges share of he reail price in comparison o oher acors operaing wihin he counry. 38

54 Turning o Zambia coffee markes, he inerpreaion of relaions beween he famers and he roasers is close o free marke or arm s lengh. In free markes, sandards and price become he medium of communicaion beween he suppliers and he buyers. There is very lile supervision from he buyers because he supplier mees he required sandards, which deermine he price. An essenial observaion in he Zambia coffee value chain is he small number of acors which has enabled high prices received by he producers in comparison o oher counries on he coninen. Much of he success of he coffee markes in Zambia is due o a well organised insiuional se-up consising of he ZCGA, which represens he privae secor and he ZCB which represens he public secor. Given ha he ZCGA provides mos services including ransporaion, he number of acors in he chain is reduced. As he coffee is sold direcly o roasers in he consuming counries, he cos of swiching o oher roasers is low and he farmers can sell o any buyer depending on he price offers. Thus, he farmers are able o sell a prices ha reflec world producer prices. A visual inspecion of price movemens for Zambian coffee over he pas weny years shows ha he prices move closely wih world producer prices, an indicaion ha he price changes for Zambian reflec changes in world. Since Zambia s coffee is usually of good qualiy, here is always demand for i, such ha bidders are forced o bid higher prices. However, one shorcoming in he Zambia aucion sysem is he aucion mehod. Telephone aucions limi he number of roasers paricipaing, consequenly limiing higher bids. A inernaional level, he marke srucure for coffee beans is in general oligopolisic in naure, where growers have lile power and are herefore unable o capure large pars of he generaed surplus. The high level of concenraion a he roasing sage gives he roasers a leeway o be slow o pass on price increases o producers. This ulimaely explains he ever widening gap beween producer prices and reail prices. As observed in he case of ransnaional buyers, who have increasingly, began o operae in he producing counries, now capure large shares of he value of he commodiy slowly gaining monopoly of he indusry. Power asymmery, which is largely suppored by high marke concenraion a he roasing and manufacuring sages, is eviden in coffee markes. This aricle suppors 39

55 previous asserions ha marke power in he coffee indusry has resuled in imbalances of ren disribuion along value chains (e.g. Kaplinsky, 2004). In consequence, he price for green coffee, raher han being deermined by marke forces, is largely deermined by he roasers in he chain. In he case of Tanzania for example, wheher farmers sell o he cooperaives or privae raders, he world price is deermined by large ransnaional roasing companies hrough he naional aucion marke. The cooperaives and privae raders offer prices according o how hey anicipae offer prices by exporers/inernaional raders a he bi-monhly coffee aucions. As ofen argued, privae regulaory sysems should lead o sronger coordinaion, since hey increase he amoun and complexiy of non-marke informaion exchange alhough hese resul in reduced rens a he lower levels of he chain. In comparison o large-scale producers in Zambia, coordinaion for small-scale farmers in Tanzania is high, he effec being unequal disribuion of rens across he chain. As ofen argued (e.g. Oxfam, 2002b), he price capured by he farmer largely depends on how much processing is done a a local level, hence he power shifs and producion rends have drasically reduced producing counry s share of he final reail price. However, going ino processing o consumer specificaions for small-scale farmers is ofen faced wih several barriers such as lack of capial, equipmen and he required skills. Such barriers o enry creae imbalances in ren share disribuion as explained in he Schumpeer heory. 2.7 Policy Recommendaions As discussed in he heory secion above, in order o undersand he ren disribuion along value chains, i is vial o map ou which aciviies susain high incomes in he chain. Given he high profiabiliy a he laer sages of he value chain (roasing and manufacuring); carrying ou hese aciviies by he producers would creae value added for he producing counries. This would no only increase economic aciviies in he producing counries bu also raise expor revenues. However, he paper has shown ha here are high barriers for he producers o go ino hese aciviies. The roasers, hrough heir subsidiaries based in he producing counries, have dominaed he processing, including aci knowledge of processing and inellecual propery 40

56 righs as mos of hem own paens on processing echnologies. Oher limiaions o producers going ino processing include he lack of capaciy o source oher ypes of coffees from oher pars of he world for blending o consumer specificaions. Furher, issues of raceabiliy become criical for consumers dealing direcly wih small-scale farmers. Therefore, governmen policies on improving coffee incomes hrough value addiion should be direced owards removing such barriers, especially reducing ransporaion coss by improving infrasrucure. In addiion, capaciy building for producers in erms of building skills and developing echnology for various value adding aciviies should be a prioriy. A more equal disribuion of profis along value chain can be achieved if cooperaives manage o break hese barriers and are able o compee successfully wih he local subsidiaries of TNC. Alernaively, primary cooperaives can mobilise hemselves o sell direcly a he aucion like in he case of Mruwia and Maeruni primary cooperaives in Kilimajoro. In addiion, fair-rade coffee growing should be srenghened among small-scale farmers in Tanzania. The fair-rade markes mus be sricly moniored o ensure ha agens buy all he fair-rade coffee produced and avoid cases where fair-rade coffee is sold a he lower conversional prices due o lack of demand. Anoher sraegy would be o srenghen coordinaion among producers wihin he counries or across counries wih common coffee varieies. For example, srenghening souh-souh relaionships in order o gain bargaining power hrough agreeing on a minimum price becomes necessary. The already exising coffee producing counries associaion would be a saring poin of discussion 41

57 References Bacon, C. (2005). Confroning he Coffee Crisis: Can Fair Trade, Organic, and Specialy Coffee Reduce Small-scale Farmer vulnerabiliy in Norhern Nicaragua? World Developmen. Vol. 33. No. 3. Pp , Baffes, J. (2005). Tanzania s Coffee Secor: Conrains and Challenges. Journal of Inernaional Developmen, Vol. 17. Pp Bargawi, H and S. Newmann (2008). From fuure markes o farm-gae: assessing he impac of financial crisis along coffee chains. Available on hp:// Dowloaded??/08/2010 Bellachwe, B. (2008). Some Observaions on Coffee Producion in Zambia. Field Trip Repor. InerAfrica Coffee Organisaion (IACO, 2008). Brown, O., A. Crawford and J. Gibson (2008). Boom or Bus: How Commodiy Price Volailiy Impends Povery Reducion, and Wha o do abou i. Inernaional Insiue for Susainable Developmen (iisd), Canada. Dolan, C. and J. Humphrey (2000), Governance in Trade in Fresh Vegeable, he Impac of Uk Supermarkes on African Horiculure Indusries. Journal of developmen Sudies, vol. 37, no. 2, pp Fier, R. and R. Kamplinsky (2002). Who Gains from Produc Rens as he Coffee Markes becomes more Differeniaed. A value Chain Analysis. IDS working paper. Forenburg, R.T. and H.O Zapaa (2004). Developed Speculaion and Under Developed Markes The role of fuures rading on Expor Prices in Less Developed Counries. Saff paper No Universiy of Agriculural and Applied Economics. Gereffi G. (1994). The Organisaion of Buyer-Driven Global Commodiy Chains: How U.S Reailers Shape Overseas Producion Neworks. ICCO quarerly Bulleing of Cocoa Saisics. USA. 42

58 Gereffi, G., J. Humphrey and Surgeon, T. (2005). The Governance of Global value Chains. Review of Inernaional Poliical Economy Vol. 12(1) Gilber, C (2006). "Value Chain Analysis and Marke Power in Commodiy Processing wih Applicaion o he Cocoa and Coffee Secors." UNCTAD PRESS in /2007/017 Huber S. (2004). Local Enerprises in he Global Economy. Issues of Governance and Upgrading. Insiue of Developmen Sudies, Universiy of Sussex, UK. Humphrey J. and H. Schmiz (2001). Chain Governance and Upgrading: Taking Sock, in Huber Schmiz (2004) Local Enerprises in he Global Economy. Issues of Governance and Upgrading. Insiue of Developmen Sudies, Universiy of Sussex, UK. Inernaional Coffee Organisaion, hp:// downloaded on 14/08/07. Kaplinsky, R. (2000). Globalisaion and Unequalisaion: Wha can be Learned from Value Chain Analysis? Journal of Developmen Sudies, 37(2), Kaplinsky, R. (2006). How can Agriculural Commodiy Producers Appreciae a Grea Share of value Chain Incomes? Chaper Published in Sarris Alexander and Hallam David, Agriculural Commodiy Markes and Trade: New Approaches o Analysing Marke Srucure and Insabiliy. Published by FAO and Edward Elger, Ialy. Kaplinsky, R. and M. Morris (2002). A Hand Book For Value Chain Research. Insiue of Developmen Sudies. Lazaro, E. A., M. Jeremiah and T.F Kilima (2008). Susainabiliy Sandards and Coffee Expors from Tanzania. DIIS. Danish Insiue for Inernaional Sudies. Working Paper no. 2008/1. 43

59 Mafusire, A., A. Salami, B. Kamara and F.E. Lawson (2010). Coffee Producion in Africa and he Global Marke Siuaion, The African Developmen Bank Group. Commodiy marke Brief, Vol. 1, no. 2. Mhando, D. G. and G. Mbeyale, (2010). An analysis of he Coffee Value Chain in he Kilimanjaro Region, Tanzania. NCC Norh-Souh Dialogue, no. 27. Micheal E. Porer (1985). Compeiive Advanage: Creaing and Susaining Superior Performance. New York, USA. Oxfam (2002b). Mugged- Povery in your cup. Oxfam Inernaional, Oxford. Parrish, B. D., V. A. Luzadis and R. B. William (2005). Wha Tanzania s Coffee Farmers Can Teach he World: A Performance-Based Look a he Fair-Trade Debae. Susainable Developmen, Vol. 13, pp Pelupessy (2001). Marke Failures in Global Coffee Chains. Paper presened a he conference on The Fuure of Perenial Crops. Yamoussoukro, Ivory Cos, November 4-9, Piroe, G., G. Pleyers and M. Poncele (2006). Fair Trade Coffee in Nicaragua and Tanzania: A Comparison. Developmen in Pracice, Vol. 16, No. 5, Augus Pone, S. (2001). Coffee Markes in Eas Africa: Local Responses o Global Challenges or Global Responses o Local Challenges? Working Paper Sub-series on Globalisaion and Economic Resrucuring in Africa no. XV. Cenre for Developmen Research (CDR) Working Paper Pone, S. (2002a). Lae Revoluion? Regulaion, Markes and Consumpion in he Global Coffee Marke. World developmen, Vol. 30 no. 7, pp

60 Pone, S. (2002b). Brewing a bier cup? Deregulaion, Qualiy and Re-organising of Coffee Markeing in Eas Africa. Journal of Agrarian Change, Vol. 2(2), Pone, S (2004). Sandards and Susainabiliy in he Coffee Secor: A Global Value Chain Approach. Documen of UNCTAD and IISD Schmiz, H.(2005). Value Chain Analysis for Policy Makers and Praciioners. Inernaional Labour Office (ILO), Geneva. Schumpeer, J (1961). The Theory of Economic Developmen. Oxford: Oxford Universiy Press. Slob B (2006). A fair Share for Coffee Producers. Chaper published in A.Oserhaus (ed), Business Unusual: Successes and Challenges of Fair Trade. Brussels Fair Trade Advocacy office, Belgium. Swinnen, J. F. M., A. Vandeplas and M. Maerens (2007). Governance and Surplus Disribuion in Commodiy Values Chains in Africa. Paper Presened a he 106h Seminar on Pro-poor developmen in Low Income Counries: Food, agriculure, Trade and Environmen. Monpellier, France, Ocober 25-27, Verissimo, P (2001). Implicaions of he EU Banana Trade Regime in Seleced Impor Markes: Economic analysis and Poliical Dimensions. Wallersein, I. (1976). The Modern World-Sysem: capialis Agriculure and he Origins of he European World Economy in he sixeenh Cenaury. New York: Academic Press, pp Zambia Coffee Growers Associaion hp:// downloaded on 12/07/07 ZCGA (Zambia Coffee Growers Associaion) (2009), Markeing Repor for he Firs Quarer Board Meeing, 2008/2008). 45

61 Chaper 3: Asymmeric Price Transmission in Coffee Markes: Impacs of Economic Reforms for Zambia and Tanzania Absrac Applying hreshold auoregression models, we examine effecs of marke liberalisaion on coffee price ransmission from world prices o grower prices in Zambia and Tanzania. Conrary o previous sudies, srucural breaks due o agriculural policy shifs have been idenified endogenously o deermine he rue effecs on he daa. Generally, resuls confirm ha price ransmission improved in he case of Zambia where coffee markeing is fully liberalised alhough he ransmission is asymmeric. In ha case, producer prices are able o adjus o correc price decreases more han price increases over an idenified hreshold. Resuls indicae ha examining price ransmission wihou aking ino accoun srucural breaks ineviably leads o false rejecion of he null hypohesis of symmeric ransmission. Key words Economic Reforms, Coffee Markes, Asymmeric Price Transmission, Threshold Auoregression, Endogenous Srucural Breaks.

62 3.1 Inroducion During he 1980s and 1990s, mos counries in sub-saharan Africa underook exensive economic reforms o move o a more marke-based developmen sraegy (Whie and Leavy, 2001). Because agriculure plays a significan role in mos of he counries, marke reforms in he agriculural secor ook a key posiion in he economic reforms (FAO, 2003). The reforms which included removal of price conrols, rade liberalisaion, and privaisaion of governmen owned agriculural enerprises, were aimed a improving producer prices and enhancing rade efficiency. However, he poor performance of he agriculural secor (sagnan growh, poor and volaile prices, unsable markes) coupled wih increasing povery levels on he coninen, has raised quesions abou he effecs of he reforms (see e.g. Whie and Leavy, 2001; Widner, 1994). While some sudies have mainained ha rade liberalisaion has led o improvemens in he agriculural secor, here is growing body of lieraure saing ha he reforms have, ineviably, led o he curren hardships in Africa. Ohers argue ha he reforms paid less aenion o fundamenals of marke funcioning, which he poor, ofen uneducaed farmers, could inexorably be exposed o. In he Coffee indusry paricularly, mos empirical sudies have found ha alhough economic reforms have improved price ransmission from world prices o producer prices, he ransmission has become more asymmeric wih price decreases in he world price being ransmied o producer prices faser han price increases (Krivonos, 2004; Warako, 2008, Fafchamps and Hill, 2008). Ohers have argued ha coffee marke liberalisaion has led o increased producer prices, hough he prices have become more volaile (Hill, 2010). On he conrary, oher sudies have linked marke liberalisaion and fall of cooperaives o reduced farmers bargaining power, leading o low producer prices which do no reflec changes in world prices (Kaplinsky, 2004; Slob, 2006). In heory, when price adjusmens are no efficienly conveyed o producers or consumers, marke inermediaries are benefiing from imperfecions and are reducing marke ransparency (Le Goulven, 2001). Paricularly he ever widening gap beween coffee reail prices in he high income counries and producer prices in he growing counries means ha eiher he producers are no benefiing from price increases or he consumers are no benefiing from price 47

63 decreases. The fac ha coffee is mainly produced in low income counries while consumpion is mosly in he high income counries means a more complex price pass hrough ha involves he norh and he souh. Alhough he number of sudies on he impac of economic reforms on producer prices has increased in he recen pas, many price ransmission aricles published over he pas wo decades have no paid aenion o possible anomalies, such as srucural breaks and asymmeries (Abdulai 2007). Even hough here is high evidence of asymmeric price ransmission (APT) in agriculural commodiies (Meyer and von Cramon Taubadel, 2004), here has been very lile empirical invesigaion in he coffee indusry. The few sudies ha have aemped o discuss APT in coffee include Krivonos (2004), who invesigaed impac of economic reforms on price ransmission from world o various producing counries. Krivonos found ha he speed of adjusmen of coffee producer prices improved in sub-saharan Africa following economic reforms compared o Souh America hough price increases were ransmied slower han price decreases- an indicaion of APT. Anoher sudy by Fafchamps and Hill (2008) on Uganda s Robusa coffee found evidence suggesing ha price increases in he inernaional marke were ransmied o local raders bu no o producers. Warako e. al. (2008) found ha he share of producer prices in world price subsanially increased for all ypes of coffee in Ehiopia since he inroducion of economic reforms alhough he ransmission was symmeric. However, he sudies moioned above did no ake ino accoun he fac ha a hreshold value may have o be reached before price changes in one marke are provoked in anoher marke, considering ha ransacion coss such as menu coss preven agens from adjusing prices coninuously. Second, he sudies paid no aenion o possibiliies of endogenous srucural breaks in he daa. This sudy fills in hese gaps by firs idenifying hresholds in he price movemens ha can poenially aler he speed and magniude of he ransmission. Thus, he price ransmission is examined according o wheher he prices are increasing or decreasing. Second, his sudy idenifies he economic reform periods using srucural break uni roo ess. This is based on he undersanding ha srucural breaks can lead o more rejecion of he null hypohesis of symmeric ransmission han appropriae (von Cramon-Taubadel and Meyer, 2004). 48

64 This sudy specifically examines he impac of he policy changes on he rae and magniude of price ransmission in a compleely liberalised coffee indusry counry Zambia as well as Tanzania, a counry ha is sill under considerable markeing regulaions in he indusry (Baffes, 2004). Given ha policies were implemened over a period of ime, we idenify he srucural breaks endogenously o ensure he correc effec of policy shifs on he daa. As lieraure would reveal, his is he firs sudy ha has employed endogenous srucural breaks in asymmeric price ransmission analysis. Differen sources of APT have been examined in he lieraure. Marke power ha rises from imperfec markes is he mos widely sied source of APT. Raionally, under imperfec markes; any price movemen ha squeezes he margin is ransmied more rapidly han any price movemen ha enlarges he margin (Meyer and von Cramon-Taubadel, 2004). Depending on he price relaionship, wo ypes of APT can be observed. Eiher marke agens ransmi price decreases from world prices o producers faser han hey ransmi price increases or a siuaion where price increases from he producers are ransmied faser o he consumers han price decreases. In his case of coffee, price decreases from he world marke are expeced o be ransferred faser o he producers han price increases. The high level of marke concenraion a he roasing and rading sages of coffee pave way for roasers o be slow or less likely o pass on price decreases o consumers or reail price increases o producers. The inernaional coffee marke is highly concenraed such ha, he op five imporers, i.e. Kraf General Foods Jacobs Suchard, Nesle, Douwe Egbers, Tchibo and Eduscho, accoun for over 40 per cen of oal global rade, while he op en accoun for more han 60 per cen (Fier and Kamplinsky 2001; Slob, 2006). Ward (1982) also argues ha marke power can lead o APT if he oligopolisic raders are relucan o risk losing marke share by increasing ouside prices. Anoher form of APT arises from high ransacion coss, paricularly menu coss ha preven economic agens from adjusing prices coninuously. Adjusmen only akes place afer a cerain hreshold. Threshold price ransmission is based on he undersanding ha prices in one marke may no be ransferred o anoher, verically or horizonally conneced marke unil he price difference reaches a cerain hreshold 49

65 (Enders and Granger, 1998). A relaed cause of APT is asymmeric movemen of informaion from a cenral marke o peripheral marke (Abdulai, 2000). By virue of being a he cenre of he nework of informaion, he cenral marke price may end o be less responsive o price changes in he peripheral markes. 3.2 Modelling Asymmeric Price Transmission Various economic models for examining APT have been employed in Lieraure. The models can broadly be classified ino wo. The firs class of models are hose ha segmen he price variables ino posiive and negaive regimes. Houck (1977) inroduced his class of models using a saic model. One shorcoming of Houck s procedure is ha i implicily assumed ha he properies of he series included lineariy and saionariy, in such a way ha, where hose properies did no apply, he resuls would carry incorrec implicaions for inference abou marke symmery (Gauhier, 2003). Alhough Houck s model was laer modified by Ward (1982) o include lagged values of he independen variable in order o accoun for dynamic responses, i sill assumed saionariy and lineariy in he daa. Sudies ha have applied his class of APT models include von Cramon (1998) and Aguiar and Sanana (2002). The second class of APT models include hose ha are exensions o he basic coinegraion heory and deals wih nonlineariies in coinegraion relaionships. The models ake ino accoun ime series properies of saionariy. They include hreshold auoregression models (TAR) wih is exensions; SETAR (self exciing hreshold auo regression) which disinguishes regimes in ime series dynamics wih poenially differen parameers (and hus dynamic properies) of each regime; STAR (smooh adjusmen auoregression) which allows for smooh ransiion of adjusmens beween regimes (Chan and Tong, 1986); and he MTAR (momenum hreshold auoregression) which allows he degree of auoregressive decay o depend on he sae of he firs difference of he variable in a model. While he TAR model can capure asymmerically deep movemens in a series, he MTAR capures he possibiliies of asymmerically sharp or seep movemens in a series (Enders and Granger, 1998). In his sudy, boh TAR and MTAR models are employed and compared. Several sudies including ha of Enders and Siklos (2001) and Abdulai 50

66 (2002) found he power of he MTAR es o be many imes higher han ha of he symmeric Enders-Granger coinegraion ess. Coinegraion models are based on he heoreical undersanding ha explains convergence of residuals of a pair of prices o long run equilibrium (Granger, 1983). The models imply ha wo or more inegraed ime series of any order have a linear combinaion of a lower order of inegraion. Tha is, if wo or more series, each of which are I(1) are coinegraed hen here exiss a saionary represenaion ha is called he error-correcion represenaion (Engle and Granger,1987). The models sugges ha prices move closely in he long run, alhough in he shor run hey may drif apar. The Engle and Granger (1987) discuss a symmeric coinegraion model ha includes an error correcion model o examine shor run adjusmens o he longrun equilibrium. Taking P 1 as producer price in Zambia or Tanzania a ime and P 2 as world prices a ime, he Engle granger wo-sep symmeric coinegraion model is represened as: P µ 1 = β o + β1p2 + (1) Where P i mus be non-saionary and inegraed of he same order e.g. I(1), β i are he parameers o be esimaed, and correlaed. µ i is he disurbance erm ha may be serially In he second sep, he residual from equaion (1) are used o esimae ρ in he following relaionship; k 1 + ρ i i 1 µ = ρµ v (2) Where, v is a whie noise process. I hen follows ha, if ρ = 0 he null hypohesis of no coinegraion is no rejeced while if 2 < ρ < 0, hen he residuals in (1) are saionary wih mean 0. In his case (1) is an aracor such ha is pull is sricly 51

67 proporional o he absolue value of µ (Enders and Siklos, 2001). Tesing he null hypohesis of no coinegraion is similar o esing for uni roo. Enders and Siklos, 2001 inroduce several saisics o es he residuals of he OLS regression of he levels daa for uni roo. Error correcion model supplemens he coinegraion model by capuring he effec of adjusmen of he dependan variable when i deviaes from he long run equilibrium. We presen he error correcion model in wo simulaneous equaions aking each of he wo prices as a dependan variable. The model is presened as: P k k 1 = δ ( P1 a P2 ) + βi 1 P1 i + β 2i P2 i + v (3) P2 i= 1 i= 1 where v is a whie noise disurbance erm, k is he lag lengh and he error correcion mechanism is given by he erm in parenhesis. P 1 and P 2 are coffee producer and world prices respecively. The k lag is se o correc for serial correlaion using a combinaion of mehods such as he Akeike Informaion Crieria (AIC) and he Bayesian Informaion Crieria. A model wih smaller values of AIC is considered beer han one wih a larger value. The AIC values from he TAR and MTAR models are herefore examined o esablish he bes model. The Durbin Wason (DW) value is used o check if here is no serial correlaion. Ideally, when here is no serial correlaion, he DW value should be around 2. Values greaer han 2 indicae negaive serial correlaion, and hose below 2 are an indicaion of posiive serial correlaion. A major limiaion o he Granger and he Engle-Granger coinegraion models is ha hey implicily assume a linear adjusmen mechanism which could lead o misleading resuls in he presence of asymmeric adjusmens (Enders and Siklos 2001). The models assume ha endency o move owards long run equilibrium is presen every ime which may no be he case. According o Enders and Granger (1998) he coinegraion ess from Engle-Granger are misspecified if he adjusmen o long run equilibrium is asymmeric. Anoher weakness of he symmeric coinegraion model is ha a saisically significan coefficien may be due o common rends in he price 52

68 pairs from facors such as populaion growh, inflaion or climae paerns raher han price inegraion (Abdulai, 2007). To exploi he possibiliy of asymmeries in price coinegraion, we employ a hreshold coinegraion model ha recognises he fac ha a shock may have o reach a significan level before a response is provoked. A hreshold regime swiching model known as he hreshold auoregression was firs proposed by Tong (1978) and laer discussed in deail by Tong and Lim (1980), Tong (1983), Balke and Fomby (1997), Enders and Granger (1998) and Enders and Siklos (2001). Laer imporan applicaion of he model in he form of auo regression o accoun for poenial nonlineariies and asymmeries in he adjusmen of individual prices and providing more informaion regarding he dynamics of he daa include Abdulai (2000,2002;2007), Srikholm and Teräsvira (2005), Hansen and Seo (2002) and Gonzales e. al. (2003). A deailed explanaion of he model is found in Enders (2004) and Zapaa and Gauhier (2003). TAR is a nonlinear regime swiching model ha allows he researcher o differeniae beween wo periods when he spread is away from he long run equilibrium, ha is, when he sysem divers o levels above or below he equilibrium, which is essenially, he esimaed hreshold value (Enders, 2004). Threshold models assume ha endency o move o long-run equilibrium is no always presen due o he presence of ransacion coss ha may preven economic agens from adjusing coninuously (Abdulai, 2002). The model suggess ha adjusmen owards equilibrium akes place only if he equilibrium error ges larger han a cerain hreshold value. In oher words, as long as he deviaions from equilibrium are small, he variables evolve independenly and become coinegraed once he disequilibrium is subsanial. Balke and Fomby (1997), Enders and Granger (1998) Enders and Siklos (2001) inroduce a wo-sep approach o examine hreshold coinegraion; residuals are obained from a linear coinegraion analysis in he firs sep and in he second sep, hreshold auo-regression is employed o ake ino accoun asymmeric movemens of he residuals. To accoun for shor erm dynamics, a es for coinegraion wih asymmeric error correcion model (ECM) can be employed. Granger and Lee firs 53

69 inroduced asymmeric error correcion specificaion where hey segmen posiive and negaive componens of he firs differences. Laer exensions of he ECM o hreshold adjusmens have been discussed by Enders and Granger (1998) and Enders and Siklos (2001), and more recenly Wolffram (2005). Again using he residuals from equaion (1), a 2-regime TAR model can be presened as; = I ρ1 ( µ 1 ) + (1 I) ρ 2 ( µ 1 ) + β µ + µ (4) µ 1 1 I = if 0 µ µ 1 1 τ < τ (5) In his regard, he firs lag of he residuals is aken as he hreshold variable, where he sysem will be 1 when he deviaions from long-run equilibrium is above he hreshold or equal o he hreshold and 0 when he deviaion is below he hreshold. There are several echniques developed o esimae a consisen hreshold (e.g Tsay, 1998; Chan, 1993; 2004) Modifying equaion (4) o fi in he firs difference of he lagged variable, he MTAR model is given by: 1 I = if 0 µ µ 1 1 τ < τ (6) Similar o he Engle-Granger coinegraion model, an error correcion model can be esimaed once hreshold coinegraion is confirmed. We presen he hreshold ECM as follows: P1 = P2 k k i P1 i + βi P2 i + γ 1Z _ plus 1 + i= 1 i= 1 α γ Z _ min us (7) Where Z _ plus and Z _ min us are dummy variables represening he sae when price differences are above and below he hreshold respecively. α i, β i, γ

70 and γ 2 are parameers o be esimaed. γ 1 and γ 2 represen he speed of adjusmen of he dependan variable in each sae. v is a whie noise disurbance erm, k is he lag lengh. P 1 and P 2 are producer and world prices respecively. Many researchers se Zero as a hreshold coinciding wih long run equilibrium, bu Enders (2004) poins ou ha a non-zero hreshold has an advanage in ha i capures sraegic behaviours and adjusmen coss ha are rarely observed wih small changes. He argues ha a TAR model wih hreshold equal o Zero does no display significan degree of asymmery, possibiliy ha he hreshold could no be Zero. Chan(1993) inroduces a mehod of esimaing a consisen hreshold hrough grind search over all possible values. Enders (2004: pp 413) explains he applicaion of he Chan procedure TAR models. Firs, he hreshold variable is sored in ascending order. Ideally, he firs and las 15 percen values of he hreshold variable are excluded such ha he search is limied o he middle 70 percen. Then a search is done over he poenial hreshold in order o minimise he Sum of Squared Residuals (SSR). The esimaed hreshold ha minimises he SSR is he opimal hreshold as illusraed in Figure 3.1. Figure 3.1 A hreshold of 0.04 idenified from he minimum RSS 4.00 Residual Sums of Squares hreshold Source: Own compuaion based on MTAR esimaion for Tanzania and World Prices 55

71 If here are more han one hreshold (more han wo regimes) hen here will be several minima of he SSR (Srikholm and Teräsvira, 2004). 3.3 Daa Descripion Monhly coffee prices are employed o examine price ransmission. The series include monhly observaions of Arabica coffee producer prices for Tanzania and Zambia, measured in U$ cens per pound (lb) 8. The producer price is he acual price received by he farmers which have been obained from he inernaional Coffee Organisaion (ICO). The response of producer prices is examined in relaion o he world prices aking he producer composie indicaor prices (CIP) 9. The CIP is calculaed by ICO based on marke share of expors of each group of coffee weighed. All he price series have 273 monhly observaions covering he period January 1986 o Sepember During his period, world prices had an average of $1.10 per pound, which was much higher han he wo producer prices. Zambian prices had an average of $0.80 per pound while Tanzanian prices have been he lowes wih an average of $0.58 (see Figure 3.1 and 3.2). For Zambia, coffee markeing has been fully liberalised where producers sell direcly o raders. Even before he economic reforms, Governmen inervenion in he coffee indusry was very limied. Neverheless, reforms in oher secors of he economy could have some direc impac on he coffee indusry, such as currency liberalisaion and he resrucuring of Governmen insiuions. Therefore, changes in price ransmission from world prices are expeced over he observed period. The preeconomic reforms in Zambia have had negaive implicaions on agriculure during 1980s due o several exogenous and domesic policy componens ha had a srong ani-rade bias and an esimaion of he real exchange-rae disorion (IFPRI, 1993). 8 Lb is he abbreviaion for Libra which is he roman word for pound. One pound is kilo grams (kgs). 9 CIP is he price calculaed based on marke share of expors of each group of coffee weighed in accordance wih Annex 1 of EB-3776/01 rev. 1 of he Inernaional Coffee Organisaion (ICO). 56

72 Figure 3.2 Zambia and CIP price rends showing he srucural break in 1998: ZAMBIA CIP Figure 3.3 Tanzania and CIP price showing he srucural breaks in 2000: TANZANIA CIP For Tanzania, alhough he reforms led o improvemen of privae secor paricipaion, he funcioning of inpu markes deerioraed as provision of credi declined. This, coupled wih he decline in qualiy and quaniy of services such as research and exension, resuled in drasic decline of overall qualiy of coffee. As a resul Tanzanian coffee prices coninue o rail far below world prices and remain he lowes 57

73 in sub-saharan Africa. Krivonos (2004) noes ha Tanzania is he only counry where he arge share of producer prices in he world marke price did no increase afer marke liberalisaion. Prior o he economic reforms, farmers had been required o sell heir expor coffee hrough he Governmen-run Moshi Coffee Aucion. USAID (2006) described he Moshi Coffee Aucion as impracical because of is 22,000 pounds expor minimum which effecively barred small farmers from paricipaing as individuals10. In 1994 cooperaives los heir monopoly afer he Governmen passed a legislaive o allow mulinaional and domesic buyers from Individual farmers. However in 1999/2000, he Governmen reversed he liberalised rade policies as i was fel ha he policies did no benefi he small scale farmers (Baffes, 2004). Currenly, all he coffee produced is Tanzania is raded hrough he Governmen-run aucions. The issuing of rading license also remains resricive (Krivonos, 2004). 3.4 Resuls and Discussion The sequence of reporing he resuls are as follows: In he firs par he resuls of uni roo ess are discussed. Nex, he resuls from he Engle and Granger coinegraion and ECM are discussed before urning o he resuls of he TAR and MTAR models and he hreshold ECM. The las par examines impulse response of he producer prices o a shock provoked by changes in he world prices Uni roo ess The hypohesis ha he price series are nonsaionary is esed using boh he Augmened Dickey-Füller (ADF) and Lee and Sracizichi Lagrange Muliplier (LSLM) srucural break uni roo ess suggesed by Lee and Sracizichi (2004). The Akaike informaion crierion was employed o deermine he appropriae lag lengh which varied across he series. The DW values also confirmed absence of auocorrelaion. 58

74 All he series were non-saionary a level bu saionary a firs difference from boh he ADF and he LSLM ess. The idenified srucural break for Zambia is May 1998 which coincides wih he compleion of economic reforms in he agriculural secor. For Tanzania a srucural break is affeced in he daa in June 2000, a period when he coffee marke liberalizaion policy was reversed giving back o he cooperaives he monopoly of purchasing coffee from he farmers. This implies ha, he acual liberalisaion policies ha ook place in Tanzania in he lae 1980s and early 1990s did no have significan effec on he coffee price movemens. Therefore, he idenified srucural break in he case of Tanzania is acually he reverse of liberalisaion policies. For World prices, a srucural break occurred in June 1989 coinciding wih he liberalisaion of he inernaional coffee marke following he collapse of he inernaional coffee agreemen in 1989 which had regulaed inernaional coffee supply and price. The resuls of he uni roo ess wih srucural breaks are presened in Table 3.1 ogeher wih he resuls of normal ADF ess. Table 3.1: Uni roo es resuls ADF(Inercep, no rend) LSLM (Inercep no rend) Level 1s Srucural Level 1s difference Break Differences Log Zambia 9 lags *** 1998: *** Log Tanzania 13 lags *** 2000: *** Log World Price (CIP) 10 lags *** 1989: *** ***, **,* denoes rejecion of he null hypohesis of saionariy a 1%, 5% and 10% significance levels respecively. Criical values for ADF ess are from MacKinnon (1996). Criical Values for LSLM Uni Roo Tes (Crash model) are from Schmid and Phillips (1992) 59

75 3.4.2 Resuls of he symmeric coinegraion model Table 3.2 repors resuls of he Engle-Granger symmeric coinegraion model esimaion (equaion 1). For Zambia, he null hypohesis of no coinegraion beween he Zambian price and he world price is overwhelmingly rejeced in he full sample. The -saisic value of is much higher han he 1% criical value of indicaing rejecion of he null hypohesis. However, examining he relaionship while aking ino accoun he economic reforms, he resuls show ha in he firs subsample, he wo price series did no co-move. The -saisic value of is lower han he 10% criical value of -2.57; hence he no-coinegraion null hypohesis canno be rejeced. A long-run relaionship is however observed afer economic reforms. The null hypohesis of no coinegraion is highly rejeced a 1% given a - saisics value of The resuls indicae ha economic liberalisaion, which included rade and currency liberalisaion, among oher reforms, has led o improvemens in price ransmission. These finding suppor mos findings from oher coffee producing counries showing improvemen in price ransmission afer economic reforms (see Krivonos, 2004). The Esimaed β 1 coefficiens shows ha in long-run a one uni change in he world prices leads o a increase in he Zambian coffee prices afer he economic reforms. However, in he pre-reform period, he change only led o a increase. Table 3.2: Resuls of he Engle and Granger coinegraion Zambia-World Prices 2 Sample β R 1986: : : : : :09 a β (5.433)*** (2.616)* (-0.127) Sample β : : : : : :09 (3.039) (1.193) (-2.320) (3.096)** *** (2.667)* *** (13.705)*** Tanzania-World Prices b β ρ *** (21.389)*** (-4.093) *** (16.210)*** (-3.301) (10.145)*** (-2.422) ***, **,* denoes rejecion of he hypohesis a he 1%, 5% and 10% significance levels respecively. CIP is Composie Indicaor Price calculaed by ICO represening world producer prices b ρ DW (-3.595) (-2.024) (-3.845)

76 Z represens he Tanzanian coffee price and Z represens he Zambian coffee price Turning o Tanzania, here is a long-run relaionship wih world prices in he enire sample. When his relaionship is examined in wo sub-periods, he esimaions show ha he wo prices are coinegraed only in he firs sub-sample and no he second sub-sample. According o he Tanzanian coffee rade policies, he firs sub-sample is he period he governmen allowed he privae raders o buy direcly from he farmers, reducing he monopoly of cooperaives. The coffee marke became liberalised during his period. However, for he period beween 2000:06 and 2008:09, here was no long-run relaionship beween he wo price series. This could be due o he fac ha, in he year 2000 he governmen reversed he liberalisaion policy, giving back he monopoly of buying coffee o he cooperaives. As discussed earlier, high governmen inervenion in commodiy markes prevens price ransmission. The high r-squared values indicae ha he models have a good fi, while he Durbin Wason (DW) values, which are close o 2, are an indicaion ha here is no auocorrelaion in he residuals Threshold Coinegraion Resuls This secion presens coinegraion and error correcion resuls for boh he TAR and MTAR model. However examining he AIC and BIC values for boh models shows ha he MTAR model provides a beer fi han he TAR model. Therefore, conclusions of his sudy are drawn only from he resuls of he MTAR coinegraion and ECM. Supply response funcions are also examined based of he rae of adjusmen from he TAR model. TAR Model Resuls Resuls for TAR which are presened in Table 3.3 indicae ha Zambian Coffee producer prices and world prices are no coinegraed over he enire sample. However, he null hypohesis of no-coinegraion is srongly rejeced in he posreform period, bu could no in he pre-reform period. This means ha afer he reforms he prices changes in world prices are ransmied o he producers while prior o he reforms his ransmission was no significan. The coefficiens on boh negaive and posiive shocks show ha adjusmen improved afer he reforms. For 61

77 example, while he Zambian coffee prices adjused by 0.15% o correc a deviaion from long-run equilibrium due o a decrease in he world prices prior o he reforms, his adjusmen increased o increase o 2.45% afer he economic reforms. This implies ha economic reforms improved price ransmission. Tes for symmery in he adjusmen was done only for he pos economic period in which he coinegraion es is significan. According o he F-saisic value of , which is way above he 1% criical value of 8.35, he null hypohesis for ρ 1 = ρ 2 could no be rejeced. In ha case price ransmission is asymmeric (indicaing presence of a hreshold). These resuls confirm he imporance of aking ino accoun srucural breaks in he daa. Making conclusions based only on resuls from he enire sample which show no price ransmission, would be misleading as he ransmission is, in acual fac, esablished in he pos economic reform period. Table 3.3 Threshold Auoregression Resuls Zambia-World Prices Sample τ a b c ρ 1 ρ φ 2 ρ 1 = ρ 2 DW 1986: :09 (-2.348) (-2.000) 1986: : (-1.128) (-0.518) : : (1.990) (-5.220) *** *** Tanzania-World Prices Sample τ a b c ρ 1 ρ φ 2 ρ 1 = ρ : : (-3.265) (-2.921) 8.604*** : : (-2.452) (-4.157) *** 8.976*** : : (0.218) a. ρ 1 and ρ 2 are he esimaed coefficiens from he hreshold coinegraion regression. The numbers in parenheses are -saisics (-0.205) b. φ are F-saisics esing he null hypohesis of no-coinegraion (i.e. = ρ 0 ρ ) 1 2 = ρ = ) c. Are F-saisics esing he null hypohesis of symmeric coinegraion (i.e. 2 *, **, *** denoes rejecion of he null hypohesis a he 10%, 5% and 1% level respecively. Criical values a he 1%, 5% and 10% level of significance are 8.35, 6.29 and 5.39 respecively for a model wih consisen hreshold (Enders, 2004) --- indicaes ha no ess was carried ou because here is no coinegraion beween he price pairs. 1 ρ 62

78 For Tanzania, resuls show evidence of a long-run bu symmeric relaionship wih world prices in he full sample. The F-saisics value of indicaes rejecion of he null hypohesis of no-coinegraion a 1% significance level. On he oher hand, he null hypohesis of symmeric adjusmen could no be rejeced a conversional levels over he enire period. However, he ess from he firs sub-sample, during he ime ha he coffee rade was liberalised, show a long-run and asymmeric relaionship beween Tanzania and world prices. On he conrary, he resuls from he second sub-sample, he period when he liberalisaion rade policy was reversed, shows no-coinegraion beween he wo price series. Again hese resuls indicae ha economic reforms ha suppor rade liberalisaion lead o improvemen in price ransmission from world prices. TAR ECM resuls Table 3.4 shows he resuls of he shor-run adjusmen o he long-run equilibrium. Since he error correcion esimaions examine rae of adjusmens were price pairs are coinegraed, only he full sample for Zambia, he full sample for Tanzania and he firs sub-sample for Tanzania are examined as repored in Figure 3.4. Figure 3.4 TAR Error Correcion Model resuls Sample 1998: :09 Sample 1986: : : :06 Zambia-World Prices Zambia World Prices Z Plus Z minus Z Plus Z minus (-2.608) (-3.266)** (-0.807) (-0.019) Tanzania-World Prices Tanzania World Prices Z Plus Z minus Z Plus Z minus 0.016(0.519) (-2.203) (-1.877) (-2.915)** (-1.900) (-3.422)** 0.081(1.963) (-3.148)** The resuls indicae quicker adjusmen for negaive deviaions from long-run equilibrium in all he hree cases. I is also quie eviden ha he Zambian prices do 63

79 no influence world prices bu he oher way round given he insignifican resuls from he model when he world price is he independen variable. Wihou even carrying ou causaliy ess, i can be concluded ha causaliy is in one direcion wih Zambian prices being caused by world prices. This is as expeced because Zambia, being a very small exporer of coffee, is no likely o have significan influence on he world price. For Tanzania, here is causaliy in boh direcions given he significan coefficiens from he model when he world price s he independen variable. However, as menioned earlier, his sudy also examines he MTAR where he overall conclusions are drawn from. MTAR Resuls From he MTAR model, unlike in he TAR model, he null hypohesis of ρ = ρ = (no coinegraion) is rejeced in every case. Similarly, he null hypohesis of ρ 1 = ρ 2 (symmeric adjusmen) is rejeced, apar from he pre-reform period for Zambia. Table 3.5: MTAR model esimaion resuls Sample τ a ρ 1 ρ : : : : : : ** (-2.702) * (-1.988) (-1.235) Zambia-World Prices *** (-3.869) ** (-2.897) *** (-4.315) b φ Tanzania-World Prices c ϕ DW *** 6.760** * *** *** Sample τ a ρ 1 ρ : : : : : : ** (-3.546) (-1.639) ** (-3.692) (-1.249) *** (-4.331) (-0.337) b φ c ρ 1 = ρ * 5.048* *** 9.191*** ** 6.043* ***, ** and * represens significance a 1%, 5% and 10% respecively z-plus are F-saisics value esing he hypohesis ha all lagged price increases above he hreshold are joinly equal o 0 64

80 z-minus are F-saisics value esing he hypohesis ha all lagged price decreases below he hreshold are joinly equal o 0 The same resuls are observed in he firs sub-sample for Tanzania. However, differen resuls are observed for Tanzania in he full sample and he second subsample where ρ1 `is larger han ρ 2. These resuls mean ha negaive shocks are no passed on o he producers due o price sabilisaion policies. In ha way he shocks are absorbed by Governmen which may no be susainable in he long-run. The esimaes for Zambia in he full sample and he second sub-sample show ha ρ 2 is more persisen han han posiive shocks. ρ 1` an indicaion ha negaive shocks are more persisen MTAR ECM Having esablished he long-run relaionship as well as he symmery of adjusmen, he rae of adjusmen is examined using he ECM. The resuls, which are presened in Table 3.5 indicae ha adjusmen owards he long-run equilibrium is quicker when he price spread deviaes below he equilibrium for Zambia in he enire sample and he second sub-sample. For Zambia, he pre-economic reform period, hreshold ECM was no esimaed because here was no evidence of asymmeric adjusmen. As expeced, he rae of adjusmen o negaive world price shocks is quicker han o he posiive shocks in enire sample and he pos-economic reform period. While negaive shocks in he world price are ransmied a a rae of 30% o producers, posiive shocks are ransmied much slower a a rae of 10%. Similarly for he poseconomic reform period, negaive shocks are ransmied a a rae of 75 percen, while price increases are ransmied a 31% percen speed in every period. These finding sugges ha due o marke power and high ransacion coss, agens are more likely o pass on negaive price changes while relucan o pass on price increases o he producers. On he oher hand, he resuls for Tanzania-World markes are quie differen wih he excepion of he second sub-sample, he liberalized period. Whereas 21 percen of a posiive deviaion is eliminaed wihin a monh, he corresponding figure for negaive deviaions is jus abou 5 percen, suggesing ha posiive deviaions are eliminaed faser han negaive deviaions. However, when he policy reforms are considered, 65

81 he resuls appear o differ for he differen policy regimes. Specifically, he esimaed parameers for he pre-reform period indicae ha only 8 percen of posiive deviaions are eliminaed wihin a monh, while 39 percen of negaive deviaions are eliminaed. Afer he reforms, as much as 28 percen of posiive deviaions appear o be eliminaed wihin monh, while jus abou 2 percen of negaive deviaions are eliminaed. Table 3.6: Resuls of ECM for MTAR Sample 1986: : : : : :09 Sample 1986: : : : : :09 Zambia-World Prices Producer price World Prices Z Plus Z minus Z Plus Z minus (-2.918)** (-3.416)** 0.002(0.245) (-1.827)* (-2.477)** (-0.727) 0.013(0.869) 0.023(0.879) (-2.233)** (-4.146)*** (-0.518) 0.001(0.014) Tanzania-World Prices Producer price World Price Z Plus Z minus Z Plus Z minus (-3.298)** (-3.235)** (0.439) (-1.552) (-1.811)* (-6.033)*** 0.024(0.672) (-2.450)** (-3.046)*** (-1.202) (-1.703) (0.052) ***, **,* denoes rejecion of he hypohesis a he 1%, 5% and 10% significance levels respecively. τ is he hreshold value deermined along wih he values ρ and 1 ρ 2 φ are F-saisics values for TAR and MTAR wih unknown hreshold esing he null hypohesis ρ 1 = ρ2 = 0 ρ 1 = ρ 2 are F-saisics values esing he null hypohesis of symmeric adjusmens i.e. ρ 1 = ρ2 CIP is Composie Indicaor Price calculaed by ICO represening world producer prices Z represens he Tanzanian coffee price and Z represens he Zambian coffee price The resuls are in line wih he policy reforms in he wo counries. In Zambia, reforms were implemened o ensure privaizaion of he coffee secor and o allow privae firms o purchase coffee from producers and sell on he world markes. In Tanzania, 66

82 reforms ha ook place in 2000 raher reversed he previous privaizaion policy, and raher inroduced policies o ensure sable prices for farmers. This involved ensuring ha farmers benefied from higher world marke prices, while price declines on he world marke were absorbed by he governmen hrough sable prices. Like he error correcion esimaions from he TAR models, MTAR error correcion coefficiens presened in Table 3.5, show faser adjusmens a a rae of 33.3 percen for price negaive shocks han posiive shocks where adjusmen is only by 10.7 percen in he enire sample Impulses response Impulse response measures he ime profile of he effec of a shock on he behaviours of a series (Koop e. al, 1996). Therefore, he inerpreaion of dynamic inerrelaions among prices in differen markes can be bes analysed hrough impulse response funcions (Porer, 1995). Impulse response is examined if price relaions are found o be coinegraed in order o examine how producer prices respond o negaive and posiive deviaions from he long-run equilibrium. In his subsecion, he impulse response is examined only for he price relaions ha are coinegraed in he MTAR model and exhibi asymmeric price ransmission. These include he Zambia- world price coinegraion in full sample and in he second subsample (afer economic reforms) and he Tanzania-world price in he full sample and boh sub-samples. Figure 3.4 shows he response of Zambian coffee prices o negaive and posiive shocks as a resul of changes in coffee world prices. Alhough he impulse is observed over 100 monhs, he prices rever back o he normal rend wihin he firs 50 monhs (5 years). Response is however asymmeric in ha, a negaive shock in he world prices is fully ransmied o he producers wihin 16 monhs while i akes 4 years (48 monhs) for he producers o fully adjus o a posiive shock in world markes. The magniude of he posiive shocks is, however, larger han he magniude of he negaive shocks. This is an indicaion ha he Zambian prices are likely o rise by a larger value when hey are hi by a posiive shock bu decline by a smaller value if hi by a negaive shock of he same magniude. As discussed earlier, 67

83 i is imporan o noe here ha, asymmery exiss in he form of speed and magniude and also varies wheher he shock is posiive or negaive. Figure 3.4 Response of posiive and negaive shocks for Zambia- full sample 0,500 0,400 0,300 0,200 0,100 0,000-0,100-0,200-0,300-0, The impulse response of he Zambian prices afer he economic reforms has a similar picure. As shown in Figure 3.5 he negaive shocks are larger in magniude and are more persisen han he posiive shocks. Figure 3.5 Response of posiive and negaive shocks for Zambia afer economic reforms 0,400 0,300 0,200 0,100 0,000-0,100-0,

84 While i akes approximaely 5 monhs for price decreases in world prices o be ransmied o he coffee producers in Zambia, he price increases ake up o 15 monhs before hey are ransmied. This affirms asserions ha prices end o be sickier downwards han upwards in liberalised markes. As discussed earlier, following inernaional coffee marke liberalisaion afer he collapse of he ICA, he markes have become oligopolisic whereby a few firms dominae he indusry. Turning o Tanzania, he impulse response of he coffee prices o deviaions from long-run equilibrium differs from he response in he case of Zambia in he full sample as illusraed in Figure 3.6. While negaive shocks are more persisen han posiive shocks in he Zambian case, posiive shocks end o persis more han he negaive shocks in he case of Tanzania. This implies ha, for Tanzania, posiive deviaions from he long-run equilibrium are correced faser han he negaive deviaions. I akes 27 monhs for producer prices o reac o price world price increases and up o 37 monhs for price world price decreases o be ransmied. In such cases, he price decrease from world prices end o be absorbed by Governmen in order o achieve price sabiliy policies. Figure 3.6 Response of posiive and negaive shocks for Tanzania- full sample 0,800 0,600 0,400 0,200 0,000-0,200-0,400-0, The paern of he price impulse response o deviaions from long-run equilibrium for Tanzania before economic reforms is differen from he paern in he full sample 69

85 (Figure 3.7). While in he full sample posiive shocks end o persis over a longer period, in he pre-economic reform period, negaive deviaions from long-run equilibrium persis longer. Whereas posiive shocks only ake abou 9 monhs o reurn o equilibrium, posiive shocks ake up o 50 monhs (more han 2 years) before he prices ge back o he normal rend. I should be noed ha, for Tanzania, his is he period when he coffee rade was liberalised and he cooperaives los he monopoly of purchasing coffee. Figure 3.7 Response of posiive and negaive shocks for Tanzania before economic reforms 0,700 0,600 0,500 0,400 0,300 0,200 0,100 0,000-0,100-0, Like in he case of Zambia, he resuls srongly demonsrae asymmeries in he ransmission of world price o producer price, a reflecion of imperfec markes among oher facors. In he pos-economic reform period, when he liberalisaion policy was reversed, he Tanzanian prices reac slowly o world price decreases bu swifly o world price increases. As illusraed in Figure 3.8, price increases are fully ransmied o he producers wihin 9 monhs while price decreases ake up o 22 monhs before hey are ransmied. 70

86 Figure 3.8 Response of Posiive and Negaive Shocks for Tanzania Pos Economic Reforms 0,300 0,200 0,100 0,000-0,100-0,200-0,300-0,400-0, Conclusion Several conclusions can be drawn from he resuls of he sudy. Firs, i has come ou clearly ha when daa is examined wihou paying aenion o srucural breaks, coffee producer prices in Tanzania and Zambia are found o move ogeher wih world prices. However, when he sample is divided according o he pre and poseconomic reforms period, esimaions from he TAR model show ha signals from world prices were no being ransmied o Zambia prices in he pre-reform period. As price ransmission was only esablished afer economic reforms, i clearly shows ha rade liberalisaion does connec producer prices. As demand for coffee is likely o increase, following China s economic boom, he producers are in his regard likely o benefi from increased prices in he long-erm. If Governmens wan o aim a improving prices, especially in he case of Tanzania, less conrol of he markeing sysem should be considered. However, policies should ake ino accoun he fac ha improved ransmission also means exposing he prices o he volailiy of he world price. Therefore, measures should be pu in place o ensure ha farmers ake advanage of marke-based risk minimising sraegies for coffee prices. 71

87 Overall, he sudy has found ha asymmeric price adjusmens eviden in coffee markes. In Zambia s case, he sudy has shown ha alhough marke liberalisaion policies improve price ransmission from world prices, price decreases are ransmied faser han price increases. High ransacion coss and lack of marke informaion are some of he facors ha lead o asymmeric price ransmission. As an inervenion o ensure efficien price ransmission, farmers need o be informed on ime, he world marke changes. This can only be possible if infrasrucure is developed especially in he rural areas where he coffee is grown. Infrasrucural developmen also reduces he ransacion coss, anoher imporan facor conribuing o asymmeric price ransmission. In he case of Tanzania, he sudy has shown ha, conrolled marke policies, alhough could lead o sable prices, do no help producers benefi from price increases because price ransmission is eiher slow or nonexisen. Trade policies ha encourage greaer inegraion wih world prices can be beneficial o he producers in he long-run, alhough shor-run price volailiy can discourage he producers. Second, our sudy found MTAR wih more power o rejec he hypohesis as compared o TAR and he Engle-Granger symmeric coinegraion ess. The AIC and SBC values also confirm ha he MTAR is a beer model. However, boh hreshold models provide beer explanaion of shor run adjusmens and asymmeries compared o he symmeric Engle-Granger coinegraion model. In any case, researchers should apply boh models because while he TAR model can capure asymmerically deep movemens in he series, he MTAR capures he possibiliy of asymmerically sharp or seep movemens in a series. A combined applicaion of he wo models has been missing in mos aricles of asymmeric price ransmission ha have been published in he recen empirical lieraure. 72

88 References Abdulai, A. (2000). Spaial Price Transmission and Asymmery in he Ghanaian Maize Marke. Journal of Developmen Economics Vol. 63, no.2, pp Abduali, A. (2002). Using Threshold Coinegraion o Esimae Symmeric Price Transmission in he Swiss pork marke. Applied Economics Vol.34, Abdulai, A. (2006), Spaial Inegraion and Price Transmission. Book Chaper published in Sarris Alexander and Hallam David, Agriculural Commodiy Markes and Trade: New Approaches o Analysing Marke Srucure and Insabiliy. Published by FAO and Edward Elger. Rome, Ialy. Abdulai, A. (2007). Spaial and Verical price Transmission in food saples marke chains in Easern and Souhern Africa: Wha is he evidence? Paper presened a he FAO Trade and Markes Division Workshop on Saple Food Trade and Marke Policy Opions for promoing Developmen in Easern and Souhern Africa, Rome, March African Developmen Bank group (2010). Coffee Producion in Africa and he Global Marke Siuaion. Commodiy Marke Brief, Volume 1, Issue 2. Aguiar, D.R. and J.A. Sanana (2002). Asymmery in Farm o Reail Price Transmission: Evidence from Brazil. Agribusiness, Vol. 18 (1), pp Baffes, J (2005). Tanzania s Coffee Secor: Consrains and Challenges. Journal of Inernaional Developmen, vol. 17, pp Balke, N.S., and T. B. Fomby (1997). Threshold Coinegraion. Inernaional Economic Review vol. 38(3), pp , Augus. Chan, K.S and H. Tong On Esimaing Thresholds in Auoregressive Models. Journal of ime series 7,

89 Chan, K.S Consisency and Limiing Disribuion of he Leas Squares Esimaor of a Threshold Auoregressive Model. Annals of Saisics 21, Edners, W. and P.L. Siklos (2001). Coinegraion and Threshold adjusmen. Journal of Business and Economic Saisics Vol. 19(2) pp Enders, W (2004). Applied Economeric Time series. Second ediion. Willey series in Probabiliy and Saisics. Book published by John Wiley and sons. USA. Enders, W., and C. W. J. Granger (1998). Uni-Roo Tess and Asymmeric Adjusmens Wih an Example Using he Term Srucure of Ineres Raes. Journal of Business & Economic Saisics Vol. 16 pp FAO, Trade Reforms and Food Securiy, Concepualising he Linkages. The Food and Agriculural Organisaion of he Unied Naions, Rome, Ialy. Fafchamps, M. and R.V. Hill (2008). Price Transmission and Trade Enry in Domesic Commodiy. Economic Developmen and Culural Change Vol. 56, no. 4, pp Fier R. and Kaplisnsky R (2001). Who Gains from Producion rens as he Coffee Marke becomes more Differeniaed? A Value Chain Analysis. Universiy of Sussex, UK. Gilber, C (2006). Value Chain Analysis and Marke Power in Commodiy Processing wih Applicaion o he Cocoa and Coffee Secors. UNCTAD PRESS in /2007/017 Giovannucci, D., and F.J. Koekoek The Sae of Susainable Coffee: A Sudy of Twelve Major Markes. Inernaional Coffee Organizaion: London; Inernaional Insiue of Susainable Developmen: Winnipeg; Unied Naions Conference on Trade and Developmen: Geneva. 74

90 Gonzales, F., Guilloreau, P., Le Grel. L. and M. Simioni (2003). Asymmery of Price Transmission wihin he French Value Chain of Seafood Producs. Casier de Recherche Granger, C.W. (1981). Some Properies of ime series daa and heir use in Economerical Specificaion. Journal of Economerics, vol. 16, pp Granger, C.W.J., and T. Teräsvira (1993). Modelling Nonlinear Economic Relaions, Oxford, U.K.: Oxford Universiy Press. Granger, W.C and P. Newbold (1977). Forecasing Economic Time Series, New York, Academic Press. Hansen, E.B and B. Sao (2002). Tesing for wo regime Threshold Coinegraion in Vecor error Correcion mehods. Journal of Economerics Vol. 110, pp Hill, R. V. (2010). Liberalisaion and Producer Price Risk: Examining Subjecive Expecaions in he Ugandan Coffee Marke. Journal of African Economics, Vol. 19, no.4, pp Houck, J.P. (1977). An Approach o Specifying and esimaing non-reversible Funcions. American Journal of Agriculural Economics, Vol. 62, pp Imai, K., Gaiha, R and G. Thapa (2008). Transmission of World Commodiy Prices o Domesic Commodiy Prices in India and China. BWPI working paper 45. The Universiy of Macheser, Brooks World Povery Insiue. Inernaional Coffee Organisaion. hp:// downloaded on 14/08/07 Kaplinsky, R. and M. Morris (2001), A handbook for value marke chain research. Canada: IDRC. 75

91 Krivonos, E (2004). The Impac of Coffee Marke Reforms on Producer Prices and Price Transmission. World Bank Policy Research Working Paper Larry Luxner, Zambia Coffee Repor. Tea and Coffee Trade Journal. Published online on 20 h January, Le Goulven, K (2001). Insiuion and Price Transmission in he Vienamese Hog Marke. Inernaional Food and Agribusiness Managemen Review, 2(3/4): Liunga, L.M (2005). The Saus of Conrac Farming and Conracual Arrangemens in Zambian Agriculure and Agribusiness. Paper prepared for FARNPAN. Universiy of Zambia. Meyer, J. and S. von Cramon-Taubadel (2004). Asymmeric Price Transmission: A Survey. Journal of Agriculural Economics- Vol. 55, Number 3, pp Norh, D.C. (1995). The New Insiuional Economics and Third World Developmen. J. Harriss, J.Huner and C.M.Lewis. London. Parrish, B. D., Luzadis, A. V. and W.R. Benley (2005). Wha Tanzania s Coffee farmers can each he World: A Performance-Based Look a he Fair Trade debae. Susainable Developmen, vol.13, pp Pelzman, S. (2000). Prices Raises faser han hey Fall. Journal of Poliical Economy, Vol. 108, no. 3, pp Perron, P. (1989). The grea crash, he oil price shocks, and he uni roo hypohesis. Economerica, 57, pp Phillips, P.C.B. and P. Perron (1988). Tesing for a uni roo in ime series regression. Bimerica, vol. 75, pp Pone, S. (2004). Sandards and susainabiliy in he Coffee Secor: A Global Value Chain Approach. Insiue for Susainable developmen, Denmark. 76

92 Poer, (1995). A non-linear approach o US GNP. Journal of applied Economics, vol. 10, pp Said, E and D. Dickey (1982). Tesing for uni roos in Auoregressive-Moving Average Models, Biomerika vol. 71, no. 3, pp Slob B (2006), A fair Share for Coffee Producers. Chaper published in A.Oserhaus (ed), Business Unusual: Successes and Challenges of Fair Trade. Brussels Fair Trade Advocacy office. Belgium. Srikholm, B. and T. Teräsvira (2006). A Sequenial Procedure for Deermining number of Regimes in A hreshold Auoregressive Model. Economerics Journal, Vol. 9 pp Talbo, J.M. (1997). Where Does Your Coffee Dollar Go? The Division of Income and Surplus along he Coffee Commodiy Chain. Comparaive Inernaional Developmen Vol. 32(1) pp Tsay, R. S. (1998). Tesing and Modelling Mulivariae Threshold Models. Journal of he American Saisical Associaion, Vol. 93, USAID (2006). Direc Expor of Premium Coffee from Tanzania. Changes in Markeing regulaions creaes opporuniies for small scale famers. Accessed on Warako T.K, van Schalkwyk, H.D and Ayele, G (2008), Producer Price and price ransmission in a deregulaed Ehiopian coffee marke, Agrekon, vol. 47, no. 4 pp Ward, R.W. (1982). Asymmery in Reail, Wholesale and Shipping Poin Pricing for fresh Vegeables. American Journal of Agriculural Economics, Vol. 62, pp

93 Wichern, R., Hausner, U. and K.C. Chiwele (1999). Impedimens o Agriculural Growh in Zambia. TMD Discussion Paper No. 47. Inernaional Food Policy Research Insiue (IFRI), New York. Whie, H. and J. Levy (2001). Economic Reforms and Economic Performance: Evidence from 20 developing Counries. European Journal of Economic Research, Vol. 2(13), pp Wolffram, R Esimaion of shor- and Long-erm irreversible Relaions. Working Paper. Deparmen of Agriculural Policy, Marke Research and Economic Sociology. Universiy of Bonn. Zambia Coffee Growers Associaion hp:// downloaded on 12/07/07 Zambia Coffee Growers Associaion (ZCGA) (2007). The Coffee Indusry in Zambia. Saus Quo, Sepember Lusaka, Zambia. Zapaa. H and W.M. Gauhier (2003). Threshold Models in Theory and Pracice, Seleced Paper prepared for presenaion a he Souhern Agriculural Economics Associaion Annual Meeing, Mobile, Alabama, February 1-5, Zivo, E. and K. Andrews (1992). Furher Evidence on he The Grea Crash, The Oil price Shock, and The Uni Hypohesis, Journal of Business and Economic Saisics, 10(10), PP

94 Chaper 4: Impac of Economic Reforms on Coffee Price Volailiy in Zambia and Tanzania Absrac During he 1990s, Zambia and Tanzania boh implemened economic reforms ha included agriculural marke liberalisaion wih differing levels of deph and impac on coffee markes. This paper invesigaes he effecs of hese economic reforms on coffee producer price volailiy in he wo counries. Unlike mos previous work which employ only symmeric GARCH models, his paper employs Threshold GARCH (TGARCH) models o examine asymmeric effecs of shocks on volailiy, aking ino accoun he fac ha posiive and negaive shocks may no have he same effec on volailiy. This is essenial for price inervenions policy. The resuls show ha marke libearlisaion led o an increase in coffee price volailiy in Zambia. The volailiy is asymmeric such ha negaive shocks lead o more volaile prices han posiive shocks of he same magniude. On he conrary, in Tanzania where coffee marke liberalisaion has been inconsisen, he reforms have had no significan effecs on coffee price volailiy. Key Words: Volailiy, Coffee Markes, GARCH, TGARCH, Economic Reforms 79

95 4.1 Inroducion Designing economic policies o achieve price sabilisaion has long been a challenge for mos counries in Africa. Agriculural commodiy prices have coninuously remained unsable, leading o devasaing effecs on food securiy and he general welfare of he producers, who are mosly small-scale poor farmers (World Bank, 2009). Low-income counries end o be vulnerable o price volailiy in heir expor commodiies as hey have limied ways o proec hemselves from global price volailiy (Akiyama e. al, 2003). High price volailiy makes i difficul for producers o plan producion since hey do no know in advance how inernaional prices will be. Consequenly, governmens canno predic expor revenues, and hence, canno engage in sable socio-economic policies. Economic reforms in Africa significanly changed he way primary agriculural commodiies were markeed (Akiyama e. al, 2003). While price sabilisaion was he policy objecive prior o he reforms, he main objecive afer he reforms was o enhance privae secor paricipaion in order o enable farmers receive a larger share of he expor price. The agriculural reforms inroduced by world bank and IMF aimed a reducing sae conrol of agriculural markes hereby opening up o he privae secor, hus fosering compeiive markes which would lead o increased agriculural producion. In general, he reforms included 1) liberalisaion of inpu and oupu prices by eliminaing subsidies on agriculural inpus such as ferilizer and credi, by bringing domesic crop prices in line wih world prices, and by ending he pracice of imposing a single price for all regions and seasons; 2) reducing overvalued exchange raes by parially liberalising he marke for foreign exchange; 3) encouraging privaesecor aciviy hrough removal of regulaory conrols in inpu and oupu markes; 4) resrucuring and/or privaising public enerprises and resricing markeing boards aciviies such as providing marke informaion and mainaining securiy socks (Kherellah, 2000, pp9). Wheher producer prices essenially improved in magniude and sabiliy in he pos reform era is a quesion ha mos economiss have aemped o address in recen sudies. Alhough in-deph empirical sudies on coffee price volailiy are few, he evidence available shows ha price volailiy significanly increased afer economic 80

96 reforms despie he fac ha farmers shares of he reail prices considerably increased (e.g. Karanja e al., 2003 Krivonos, 2004; Newman, 2006; Gemech and Sruhers, 2007; Hill, 2010). In oher counries such as Uganda, sudies show ha coffee prices became very vulnerable o inernal price flucuaions (Fafchamps and Hill, 2008). In general, he economic reforms in Sub-Saharan Africa in he 1990s were less resricive han elsewhere, resuling in sudden dissoluion of producer cooperaives and consequen reducion in he producer s bargaining power, such ha farmers were now exposed o price risks ha were previously absorbed by he governmen (Krivonos, 2004). However, for commodiies ha were axed prior o reforms, he welfare consequences of reform may be offseing, ha is, he reforms ha boos producers share of expor prices may well compensae for increased price volailiy (Akiyama e al. 2003). Crop commodiies in paricular, are highly suscepible o price volailiy because of he naure of he producion and markeing cycle. 11 When prices increase, farmers are moivaed o increase producion, such ha excess producion may lead o price decline. When prices decline, farmers ge discouraged from invesing in producion, leading o low yields and less supply. In he end, prices rise again due o low supply. In he case of Coffee, volailiy is exacerbaed by he fac ha coffee is a perennial crop wih lags beween planaion and harvesing varying beween 18 o 24 monhs. Peak yields are only experienced afer 5 o 7 years. This implies ha, while inpu invesmens respond quickly o price changes, supply response in coffee markes is very slow. Consequenly, addiional supply ends o reach he marke when prices are on he decline, leaving he farmers wih no choice bu o sell a any price. Moreover, following he fall of he Inernaional Coffee Agreemen (ICA) 12 in 1989, which had pu in place a pricing mechanism o sabilise prices, coffee producer pries a inernaional level became very volaile (ICO daa). Some researchers have also linked he deregulaion of he inernaional coffee marke o increased coffee supply 11 There are various oher facors associaed wih coffee price volailiy. Three facors have come ou srongly in lieraure; 1) Marke deregulaions- a domesic and inernaional levels, 2) failure of inernaional cooperaion o sabilise prices 3) he developmen of commodiy exchange markes- wih fuure deals making up o 80% of coffee rade. I is agued ha price volailiy encourages fuures rading, which in urn magnifies volailiy (Gilber and Brunei,1996). 81

97 on he marke, especially afer he emergence of Vienam as second larges producer afer Brazil. Consequenly, inernaional coffee prices fell drasically in he las wo decades hiing lowes prices in 1992, a siuaion which severely affeced many producing counries. In Tanzania, many producers were forced o diversify heir income-generaing aciviies from coffee in order o reduce price risks (Piroe e al., 2006; Newman, 2006). In some cases, coffee rees were being uprooed and he fields replaced wih oher crops such as maize. Similarly for Zambia, a number of large esaes closed down, leaving ens of housands of he rural populaion ou of employmen. Zambia s coffee producion also drasically declined from approximaely 7000m o less han 2000m in he las four years. 13 While he body of lieraure on commodiy price volailiy has increased in he las wo decades, specific lieraure on coffee price volailiy is raher scany. Currenly here are only wo sudies ha have empirically examined coffee price insabiliy and he causes; Newman (2006) sudied he exen o which changing marke srucures have caused rise in coffee price volailiy while Krivonos (2004) invesigaed he impac of rade reforms on coffee price movemens. Krivonos concluded ha, he improvemen in coffee price ransmission from world markes has exposed producer prices o he risk ha was previously absorbed by he governmen. The Newman (2006) sudy even included a volailiy analysis in he case of Tanzania bu did no provide saisical evidence of increased price volailiy in he pos economic reforms as compared o he pre-economic reform period. While he Krivonos (2004) sudy only carried ou price ransmission analysis bu did no examine he volailiy. Furher, none of hese sudies considered hreshold or asymmery effecs of volailiy. Ever since Glosen e. al. (1993) inroduced asymmery volailiy models, mos of is applicaion has been confined o financial daa save for a few sudies ha have exended he model o commodiy markes. One example is he sudy by Guilda and Mringe (2004) ha applied symmeric and asymmeric volailiy models in financial and commodiy price series. They found ha asymmeric models led o beer forecass han he symmeric one. Anoher ineresing sudy was by Shively (2001) who examined hreshold volailiy in spaial maize markes in Ghana. He found ha grain prices in Ghana followed a hreshold process characerized by low variance

98 and high variance regimes. In his sudy, auo-regression condiional heeroskedasiciy (ARCH)-ype models are employed, paricularly he Threshold Generalized ARCH (TGARCH) in order o examine asymmeric effecs of shocks on volailiy. There has been very lile applicaion of asymmeric models in volailiy analysis in mos economic lieraure, despie evidence of asymmeric price ransmission in agriculural commodiy markes (See e.g Krivonos, 2004). More over, he TGARCH model enables he assessmen of asymmeric effecs of he shocks on price volailiy, an aspec ha mos sudies have overlooked. Volailiy of coffee prices for Zambia and Tanzania is examined paying aenion o asymmeric behaviour and he impac of coffee marke liberalisaion. In boh counries, significan marke reforms occurred especially in agriculural commodiies. However, he liberalizaion process, paricularly in he coffee indusry, varied beween he wo counries, boh in he scope of he reforms and heir consequences (Akiyama e. al. 2003). While Zambia s coffee marke is fully liberalized where producers sell direcly o raders, Tanzania s coffee marke sill has some governmen inervenion such as mandaory coffee aucioning (Jeffes, 2004; Akiyama e al. 2003; Newman, 2006). The diversiy beween he wo counries provides a beer undersanding of impacs of economic reforms under differen marke srucures. Coffee is Tanzania s larges agriculural expor commodiy wih half a million small scale farmers direcly earns heir living from he commodiy. For Zambia, coffee has a very small share of agriculural expors, alhough i couns as one of he op en non-radiional expors. Zambian coffee (a mild Arabica variey) feches very high bu unsable prices compared o coffee from mos African counries. On he oher hand, Tanzanian coffee prices (boh for Arabica and Robusa varieies) are among he lowes on he coninen bu very sable (See Figure 4.2 in secion 4.5). Moreover Tanzania is he only counry in Africa where he arge share of producer prices of he world marke price did no increase following liberalizaion (Krivonos, 2004). This sudy invesigaes price volailiy in coffee markes for Zambia and Tanzania. The analysis includes an examinaion of he effecs of economic reforms on coffee price volailiy in he wo counries. More imporan, he sudy examines he impac of negaive and posiive shocks o he prices on price volailiy. 83

99 The res of he paper is organised as follows: In he nex secion, an overview of economic reforms in Zambia and Tanzania is given, paricularly hose ha direcly or indirecly relae o he agriculural secor and he coffee indusry in paricular. Secion hree discusses he concepual framework including a discussion of he models used and he mehodology of his sudy. Secion four describes he daa and he resuls are discussed in secion five. The las secion gives a conclusion. 4.2 Economic Reforms and Coffee Markes in Tanzania and Zambia Prior o he 1990s economic reforms in Tanzania, all coffee markeing was handled by he Tanzania Markeing Board (TMB) and he cooperaive unions, who also provided inpus, ransporaion and processing services. In 1990, Tanzanian governmen ook he firs seps in resrucuring he coffee indusry. A noable improvemen of he reforms was he promp paymen (wihin hree weeks) o he unions by he Coffee Board (Baffes, 2005:25). By 1992 he cooperaives were allowed o decide prices paid o growers. In 1993 legislaion allowing privae secor paricipaion in markeing and processing coffee was passed, which also furher reduced governmen s conrol of pricing. In 1994 cooperaives los heir monopoly afer he Governmen passed a legislaion o allow mulinaional and domesic buyers o buy direcly from individual farmers. Alhough his move led o improvemens in privae secor paricipaion, he funcioning of inpu markes deerioraed as he provision of credi declined. This, coupled wih he decline in he qualiy and he quaniy of services such as research and exension, resuled in a drasic decline in he overall qualiy of coffee. Subsequenly, in 1999/2000, he Coffee Board adjused he policies in order o help he small-scale farmers, who, i seemed, had no benefied from rade liberalizaion. Currenly only abou 20% of he coffee is sold direcly o inernaional raders (mainly by large esaes), while 80% has o go hrough he governmen run aucions (Newman, 2006). For Zambia, prior o economic reforms, he Zambian economy was characerized by srong sae inervenion in agriculural markes, which involved he fixing of producer prices, he provision of ransporaion, sorage and inpus, and subsidizing of credi 84

100 for agriculural commodiies (Wichern and Hausner, 1999). The exchange rae was also consanly overvalued, which had an indirec impac on he producers of radable commodiies. Privae raders of all agriculural commodiies were no allowed and even discouraged by he fixed marke margins by he Governmen. Afer 1990, governmen implemened a series of liberalizaion policies ha aimed a deconrolling prices, privaizaion of sae-owned enerprises, reducing inflaion and inroducing marke-based exchange and ineres raes. The privaizaion included large scale coffee esaes, which under governmen managemen had become insolven as menioned above. Alhough he liberalizaion policies were consanly reversed for some crops (especially for maize afer some shocks such as droughs), he coffee indusry in Zambia remained fully liberalized where producers sell direcly o raders. I is quie eviden ha coffee price volailiy in he wo counries is influenced by he rade policies. Since he policies deermine how price shocks in world markes are ransmied o producer prices, he level and persisence of he volailiy will largely depend on he marke policies. In order o undersand how he level and persisence of volailiy is measured, he nex secion gives a heoreical background for undersanding volailiy in commodiy markes. 4.3 Theoreical Framework for undersanding commodiy price volailiy Mos commodiy price daa end o exhibi volailiy persisence, leverage effecs 14 and he endency of large residuals o cluser ogeher in such a way ha large changes follow large changes and small changes follow small changes- a fashion known as volailiy clusering. The changes are largely unprediced boh in he magniude and he sign. Large disurbances become par of he informaion se used o consruc he variance forecas of he nex period s disurbances. As a resul, large shocks eiher negaive or posiive can persis, influencing fuure prices for several periods. Therefore, volailiy shows he variance/sandard deviaion of he error erm and wha makes hem large. 14 Leverage effecs are a kind of asymmeric effecs of he previous periods variance on volailiy. I is a siuaion where negaive reurn sequences are associaed wih increases in he volailiy of he prices. 85

101 Commodiy price volailiy is modelled using Auoregressive Condiional Heeroskedasiciy (ARCH)-ype models such as he Generalised Auoregressive Condiional Heeroskedasiciy (GARCH) (Bollersler, 1986). Ideally, he goal of he ARCH/GARCH model is o provide a volailiy measure, such as he sandard deviaion, ha can be used for making decisions concerning risk and derivaive pricing (Engle, 2001). ARCH/GARCH models consider he variance of he curren errors o be a funcion of he acual size of he previous errors. Enders (2004) explains ha since he condiional heeroskedasiciy of { ε } in equaion (1) will resul in heeroskedasiciy in { y }, ARCH-ype models are able o explain periods of volailiy and ranquilliy. Because of he presence of heerosdasiciy in he error erm, ARCH-ype models have an advanage over oher volailiy measuring mechanisms such as he coefficien of Variaion (CV). 15 ARCH-ype models have widely been used in differen sudies (e.g. Boleslaw, 1990; Shivery, 2001; Brooks, 2002 and Linon, 2008). Conrary o he homoskedasiciy assumpion, 16 ARCH/GARCH models are based on he expecaion ha no all daa have all error erm values ha are he same a any given ime. Especially for agriculural commodiies and in paricular producer prices ha rely on exernal markes, he variance of he error erm is no likely o be consan over ime. In his case a problem would arise as heeroskedasiciy or daa in which he variances of he error erm are no equal, resuling in errors and confidence inervals esimaed by leas squares being narrow, giving false sense of precision (Engle, 2001). To avoid his problem, ARCH models rea heeroskedasiciy as a variance o be modelled. In his way, he deficiencies of leas squares are correced and a predicion is compued for variance of each error. The basic ARCH-ype model is he ARCH model iself and is composed of wo equaions, which are esimaed simulaneously. The firs equaion is he condiional mean equaion which describes he expeced value of he sochasic process y. I is 15 The CV is he absolue value of raio of he sandard deviaion o he mean. 16 For a homoskedasiciy innovaion, he error erm, wih mean zero is assumed o have a consan variance over a period of ime, presened as: q ε = a 0 + a jε j + u where u ψ 1 N(0, σ ) j= 1. 86

102 assumed ha he ime series y is saionary wih consan variance. The second equaion is he condiional variance equaion and he variance is assumed o be heeroskedasic. Since volailiy is unobservable characerisic of a series, a proxy is chosen for i which is he varianceσ. y = ' X θ + ε 2 where ε ψ 1 ~ N(0, σ ) (1) p i= σ = ω + α ε (2) i While in he ARCH model, he condiional variance depends on he squared residuals of he las period, in he GARCH model he condiional variance erm will depend upon he lagged variances as well as he lagged (squared) residuals. This allows for persisence in volailiy wih a relaively small number of parameers. Presence of ARCH effecs does no imply absence of GARCH effecs. ARCH effecs indicae presence of auocorrelaion, such ha, high order models are required. To circumven such misspecificaion, volailiy is beer modelled using GARCH models, which combines he ARCH (q) and variance (p) equaion ino a non-linear ARMA (p,q) process presened as: q i= 1 2 i + p 2 σ = ω + α ε βσ (3) i i= 1 2 i I follows ha, if here are no GARCH effecs he sum of he coefficiens should be equal o zero, i.e. q i= 1 p α β σ = 0 (4) i i= 1 i i 2 In ha case, he long run variance ω = σ. Since variance is sricly posiive, sufficien condiions o ensure non-negaiviy are; 87

103 ω > 0 α i 0 β 0 i (5) q 2 2 where i=1, 2,.., p If α ε β σ < 1 in equaion (4), hen volailiy is i= 1 i i p + i= 1 i saionary, ha is any shocks o he sysem will dissipae or vanish over ime. Bu if q i= 1 p + i 2 2 α ε β σ 1 hen he shocks will accumulae or persis over ime. In a i i i= 1 i i q 2 2 special case when α ε β σ = 1, shocks will persis indefiniely such ha i= 1 i i p + i= 1 i i arbirage will no be able o adjus he level of volailiy o a long run equilibrium. Engle and Bollerslev (1986) refer his o Inegraed GARCH (IGARCH). Due o nonlineariy of mos ime series daa, ARCH/GARCH models are esimaed using he maximum likelihood. 17 A challenge facing he ARCH/GARCH model is he implicaion ha posiive and negaive residuals have a symmeric impac on he condiional variance. GARCH models assume ha good and bad news have he same effec on volailiy, an assumpion which is ofen violaed (Black, 1976). Glosen e al. (1993) and Rabemanayjara and Zakoian (1993) discuss hreshold models ha allow for negaive residuals o affec he condiional variance differenly from he posiive residuals. Noicably, ever since Glosen e al. (1993) inroduced asymmery volailiy models, mos of is applicaion has been o financial daa analysis, excep for a few sudies ha exended he model o commodiy markes. One example is he sudy by Guilda and Maringe (2004) ha applied symmeric and asymmeric volailiy models o financial and commodiy price daa. They found ha asymmeric models led o beer forecass han he symmeric one, based on values of various model fi saisics. Anoher sudy examined hreshold volailiy in spaial maize markes in Ghana (Shively, 2001). The sudy found ha grain prices in Ghana followed a hreshold 17 The maximum likelihood is given by: L n = n 1 2 In( h = s+ 1 2 ε + h (6) where s = max{p,q} 88

104 process characerized by low variance and high variance regimes. The Threshold GARCH model is specified as: p p q = α 0 + α iε i + λiε = id 1 + βiσ i i= 1 i= 1 i= 1 σ (7) where λ i is he parameer capuring asymmeric effecs in he model such ha ess for asymmery include a es of all λ i. If he λ i is saisically differen from Zero, he daa conain a hreshold effec. The hreshold in his case is ε = 1 0 such ha he effecs of he shocks greaer han he hreshold will be differen from hose below he hreshold. d 1 is an indicaor funcion for ε 1 such ha d = 1 1 when ε 1 is negaive and d = 1 0 when ε 1 is posiive. Tha means ha negaive values of ε 1 end o increase he variance more han posiive values. This is because he effec of he shock on he variance when ε 1 is negaive and d = 1 1 will be α i + λi bu when ε 1 is posiive and d = 1 0, he effec will be only he α i. Therefore a posiive value of λ i means ha a pas negaive reurn has a larger impac on condiional volailiy han a pas posiive reurn of he same ampliude- a siuaion commonly referred o as leverage effec (Enders, 2004: pp. 141). This implies ha he bad news has more effec on he variance han he good news. To examine he effec of economic reforms on volailiy while a he same ime, aking ino accoun possibiliy of asymmeric movemens, I modify he TGARCH model o include a dummy variable for he period of he economic reforms as idenified endogenously using Lagrange muliplier uni roo ess. Because he reforms were implemened over a period of ime, he srucural break is idenified endogenously o ensure he rue effec of he reforms on he daa generaing process. Equaion (4) is herefore modified as; p p q = i i i i i i i i= 1 i= 1 i= σ ω + α ε + λ ε d + β σ + δrf (8) where RF is a dummy variable for Economic reforms such ha δ = 1 for he period before he reforms and δ = 0 for period afer he reforms. 89

105 4.4 GARCH Model Esimaion Procedure Figure 4.1 illusraes he mehodology for esimaing GARCH models as suggesed by Moledine e al. (2007). The firs sep is uni roo esing. Ideally, saionary series should be used in modelling volailiy in order o mee he ime series modelling assumpion of zero mean. If he series is non-saionary, hen i should be differenced o aain saionariy. Figure 4.1: Seps in Modelling Volailiy Fail o rejec null Hypohesis of Firs difference series Rejec Hypohesis of no ARCH Esimae GARCH/ TARCH/ GARCH Perform uni roo ess Box-Jekins mehod o deermine ARCH es Rejec hypohesis of uni roo Series remains in Levels Fail o rejec hypohesis Esimae ARIMA Source: Moledine e. al The second sep is he idenificaion of he model (number of lags) using Box-Jekins approach along wih informaion crieria such as he Akaike Informaion Crieria (AIC). Afer idenifying he number of lags, he model is esed for ARCH effecs. If he hypohesis of no ARCH effecs is rejeced, hen here is heeroskedasiciy in he error erm. I is hen clear ha he model should no be esimaed wih ordinary ARMA models, raher, wih GARCH models which do no assume homoskedasiciy (Engle, 2001). 90

106 4.5 Daa Descripion The sudy uses monhly observaions of coffee prices measured in US$ per pound (lb) for Tanzania and Zambia. Boh series have 273 monhly observaions covering he period January 1986 o Sepember. The price series are presened shown in Figure 4.2. Figure 4.2: Coffee Producer Prices for Zambia and Tanzania in UD$/ per Lb TANZANIA ZAMBIA Source: Auhor s Presenaion based on Daa from ICO In order o compue he daa in a coninuous way as well as ensure saionariy, prices are in naural logarihms of reurns. The price reurn is calculaed as logarihm price a a given ime less he logarihm price a he previous ime presened in R InP InP (9) = 1 Where R is he price reurn, P is price a ime and P 1 is price a he previous period. In his way he variance of he price reurn is aken as he risk level of he reurns (Engle, 2001). Presence of heeroskedasiciy herefore implies ha some 91

107 ime periods are riskier han ohers, where he magniude of he variance is higher in some periods and lower in oher periods. A visual inspecion of price reurn series (Figure 4.3) shows some evidence of heeroskedasiciy given ha he variance is no he same over ime. The ampliude of he variance is higher in some periods while very low in oher periods. For example, he boh producer prices experienced very high volailiy around 1994, a ime when major economic reforms were aking place. Figure 4.3 Price reurns for Zambia and Tanzania 1.5 Zambia 1.6 Tanzania World Saisical properies of he series are repored in Table 4.1. If he sandard deviaion is aken as a measure of volailiy, Zambia prices are he mos unsable wih a sandard deviaion of Tanzania series is less volaile han Zambia prices bu is sandard deviaion is 0.03 higher han world price. Addiionally, all he hree producer price series show evidence of fa ails since hey are all above he normal 92

108 disribuion value of 3. Reail prices on he oher hand show evidence of normal disribuion wih kurosis less han 3. Since volailiy clusering, which is a ype of heeroskedasiciy, is responsible for excess kurosis (fa ails), high volailiy for he hree producer prices compared o he reail price is expeced bu no for reail prices. Table 4.1: Descripive saisics for he series Zambia Tanzania World Mean Sd Dev Skewness Kurosis (excess) Skewness is posiive and significanly differen from zero for Zambia and world prices, an indicaion ha here are more values above he zero mean han below. I is somehow abnormally high in he Tanzania series, probably due o non free rade policies in ha counry during mos of he ime period under review. I is also posiive bu relaively smaller for reail prices compared o producer prices, an indicaion of normaliy. 4.6 Resuls Volailiy analysis requires ha he variables are saionary in order o obain meaningful resuls. Uni roo ess o examine saionariy were performed using Augmened Dickey Fuller (ADF) es and Lagrange muliplier srucural break ess. I was necessary o apply srucural break uni roo ess because he observed period experienced policy changes a domesic and inernaional level, which could resul in srucural breaks in he daa. Lee and Srazichich (2004) srucural break uni roo es, which simulaneously indenifies a srucural break and ess for uni roo, was applied o all he four series. Resuls show ha all series are non-saionary a level while saionary a firs difference (Table 4.2). 93

109 Table 4.2: Uni roo ess Coffee Prices ADF LSLM (Inercep no rend) 1986:1 o 2008:9 Level 1s Srucural Level 1s difference Break Differences Log Zambia 9 lags ** 1998: ** Log Tanzania 13 lags ** 2000: ** Log World Price (CIP) 10 lags ** 1989: ** According o Lee and Srachizic srucural break uni roo ess, Zambia coffee prices experienced a srucural break in May, Given he occurrence of agriculural marke policy changes, 1998 coincides wih compleion of agriculural liberalisaion iniiaed around 1991 (Rainer Wichern, 1999). In Tanzania s case, a srucural break occurred in June From he occurrence of coffee rade policy changes, Tanzania adoped liberalised rade sysem in However, in 2000/2001 he coffee Board revoked he buying licenses of privae raders, effecively handing he monopsony power back o he unions (Krivonos, 2004). To esablish he number of lags o include in he model, we examine he auocorrelaion using Box-Jekins model. As presened in Figure 4.4, he series have he p-values which are almos zero, apar from he firs lag which is almos one. In his case, he models should include one lag o accoun for he auocorrelaion. The models are hen GARCH (1, 1) and TGARCH (1, 1). 94

110 95

111 Figure 4.4: Auocorrelaions of price reurns 1.00 Zambia Squared Residuals auocorrelaions DPCORRS PDPCORRS 1.00 Tanzania Squred residuals AuoCorrelaions AUTOCORRS PARTCORRS 1.00 Cip Squared Residuals auocorrelaions DPCORRS PDPCORRS 96

112 Saisical analyses of ARCH effecs confirm heeroskedasiciy for boh Zambia and Tanzania a 1% significance level. The ARCH resuls are presened in Table 4.3. Table 4.3: ARCH es resuls F-saisic Significance Zambia *** Tanzania *** CIP *** The F-saisics show ha Zambia experienced he mos volaile coffee producer prices over he observed period, which is even more han world prices. However, Tanzania prices are less volaile han world prices. The volailiy is significan in all series, an implying presence of heeroskedasiciy. GARCG models can hen be applied o examine volailiy. For his sudy, GARCH (1, 1) and TGARCH(1,1) models are esimaed for Zambia, Tanzania and world prices, o esablish impac, persisence and symmeric effecs of price shocks on volailiy. The value of α parameer (he ARCH componen of he model) in equaion (8), which is he sum of squared residuals, measures he impac of price shocks on volailiy. The variance parameer β (he GARCH componen of he model) and he hreshold parameer λ show persisence and symmery of he price shocks respecively. The resuls of he esimaion and he saisical verificaion of he wo models (GARCH and TGARCH) are summarized in Table 4.4. The esimaed coefficiens for GARCH(1,1) have he correc sign for Zambia and are all saisically significan. Ward ess for he resricion α + β = 1 which ess he null hypohesis of no GARCH effec is rejeced a 1 percen for all he hree producer prices. The esimaed coefficiens affirm preliminary resuls discussed earlier, which show evidence of high volailiy for Zambia compared o he oher price series. Any shock o he sysem riggers 58 percen volailiy in he case of Zambia. The shocks hardly have any effecs on Tanzania prices, where a shock o he prices, leads o only 0.08 percen volailiy. However, even if Zambia experiences he larges impac, persisence of he volailiy is relaively low. The price volailiy due o a shock persiss for only 15 97

113 monhs before disappearing. Volailiy persisence is however very high in he case of Tanzania lasing up o 51 monhs. The coefficien for he domesic economic reforms is no significan in boh counries in he GARCH (1, 1) model. The resuls from he TGARCH presen a differen picure. For Zambia, almos all he coefficiens are posiive and saisically significan. Unlike in he GARCH esimaion, he impac of he shocks on volailiy is very low for Zambia in he TGARCH (1, 1) esimaion. The shocks rigger only 0.04 percen volailiy which is no saisically significan. The volailiy is however more persisen such ha he prices remain volaile due o a shock in one period for he nex 59 monhs (almos 5 years). In ha case, persisence of even small impacs is very high. Similar o GARCH resuls, he impac of shocks on volailiy of prices in Tanzania is very low in TGARCH compared o Zambia. In Tanzania s case any shock riggers only 0.8 percen, alhough ha persiss up o 38 monhs. Again, a plausible explanaion for he differences could be he exen of liberalisaion (alhough ha could no be he only facor). The sign for he dummy variable represening economic reforms is posiive and saisically significan for Zambia in he TGARCH model bu no in he GARCH model. The resuls indicae ha price volailiy increased afer he economic reforms. However, for Tanzania, he impac of he economic reforms is no significan probably due o he fac ha he reforms have been consanly reversed as discussed earlier. The λ parameer in he TGARCH models is significanly differen from zero for Zambia confirming exisence of asymmeric effecs. Since he coefficien is posiive, a conclusion can be made ha bad news (negaive shocks) has more effec on volailiy han good news. However, he hypohesis of no asymmeric effec is no rejeced for Tanzania, an indicaion ha posiive and negaive shocks had he same effec on volailiy. Again his could be explained by high governmen inervenion in seing prices, which does no reflec he acual movemens of world producer prices. To verify previous empirical findings ha Asymmeric TGARCH models are beer han symmeric GARCH model (Guida and Maringe, 2004) differen saisics ha explain he bes fi for he model are examined. 98

114 Table 4.4: Volailiy esimaes of GARCH and TGARCH models Coefficien Zambia GARCH Zambia TGARHG Tanzania GARCH Tanzania TGARCH World GARCH World TGARCH Mean (-0.234) (-2.369) 0.008(0.849) 0.008(0.718) (-1.142) (-0.965) C 0.036(7.421)*** 0.010(3.177)** 0.018(4.206)*** 0.017(2.616)* (-1.23) 0.003(2.955) α 0.581(3.510)*** 0.037(1.217) 0.077(1.231) 0.377(3.293)** 0.309(3.178) 0.387(2.767) β 0.148(2.099)* 0.594(10.012)*** (-1.820) (-0.918) 0.378(3.112) 0.324(2.242) λ (3.311)** (0.336) Reform (-0.571) 0.074(2.970)** (-1.227) (-1.047) F-sas( α + β + λ = 0) LL 30.43(0.000)*** (0.000)*** (0.000)*** (0.004)** (0.000) (0.000) AIC SBC The -saisics are given in parenheses while f-saisics are given in brackes. ***, **, * denoe significance a 1%, 5% and 10% respecively. 99

115 Figure 4.5: Volailiy of Zambia and Tanzania prices compared o he world price a. Volailiy of Zambian and World Prices b. Volailiy of Tanzanian and World Prices World Prices Zambia Tanzania World Prices 100

116 The hree include he Log Likelihood (LL), Akaike Informaion Crierion (AIC) and Schwarz Bayesian Crierion (SBC). In his case, a lower saisic in absolue values is an indicaion of a beer model fi. Resuls (presened in he las 3 rows of Table 4.4) show ha in each of he hree price series, TGARCH has lower values compared o GARCH models. Thus he bes resuls are achieved by TGARCH, which also akes asymmeric effecs of shocks ino accoun. Overall, he sudy has confirmed ha coffee producer prices a producer level and world marke level are more volaile han reail prices. Visual inspecion of price volailiy in Figure 4.5 above, shows ha whereas he volailiy of he Tanzanian prices are relaively on a raher consan low level, he volailiy of he Zambian prices shows huge up-and down urns. On he hand he volailiy for world prices is more consan compared o Zambia, bu less so compared o Tanzania. 4.7 Conclusions This paper has invesigaed he impac and persisence of shocks on coffee price volailiy in differen markes. Special aenion was given o he impac of economic reforms ha included coffee marke liberalizaion in Zambia and Tanzania. Srucural breaks around he imes of economic reforms were idenified using endogenous Lagrange muliplier uni roo ess. Alhough he paper applied wo ARCH-ype models (GARCH and TGARCH), he conclusion is mainly drawn on he resuls of he TGARCH model, which was found o provide a beer fi of he model, based on absolue values of he log likelihood, he AIC and BIC. TGARCH resuls show discernible effec of economic reforms on price volailiy in compleely liberalised markes. I is eviden ha producer prices in Zambia, which has a compleely liberalized coffee marke, showed lower impac bu longer persisence of shocks on volailiy compared o producer prices in Tanzania. The persisence of shocks on volailiy in he Zambian case was even longer han for world prices. However, world prices had a higher impac of he shocks on volailiy han Zambia and Tanzania. Alhough cerainly no he only facor, economic reforms have inescapably exposed coffee producer prices o world prices, as concluded by several earlier sudies. Oher facors could be ha, small producers like Zambia, are 101

117 ineviably price akers wih neiher capaciy nor governmen ineress in conrolling price movemens. On he oher hand, he Governmen in Tanzania pays much aenion o ensuring price sabiliy, even afer he implemenaion of economic reforms, given he significan role coffee plays on he economy (as he second expor commodiy). Alhough price sabiliy could be o he advanage of he producers for sable incomes in shor-run, he price inervenion policies have negaively affeced Tanzanian coffee prices in he long-run. As he sudy has shown, Tanzanian prices have in he long-run lagged far below world producer prices, and are even much lower han Zambian prices (again here could be oher facors such as qualiy of coffee produced ha could conribue o price levels and volailiy). Moreover, his sudy found ha he effec of economic reforms was no significan for Tanzania, indicaing ha coninued governmen inervenion in price sabilisaion prevened he inended economic reforms o have any impac on price volailiy. Wha governmens should focus on is a holisic approach o ensuring price sabiliy especially marke based price sabilisaion policies. The direc Governmen inervenion in price sabilisaion may no be susainable in he long-run. In addiion, infrasrucure developmens in order o improve communicaion and informaion disseminaion become imporan in minimising price volailiy. Curren price inervenions in coffee markes such as Fair rade iniiaives are anoher opion ha can enable farmers obains a fair share of he reail prices. 102

118 References Akiyama T, Baffes J, Larson D.F and Varangis P (2003). Commodiy Marke Reform in Africa: Some Recen Experience. World Bank Policy Research Working Paper Bo Sjö (2008). Tesing for ARCH and GARCH: Modelling he volailiy of Elecrolux sock. Brook, A double-hreshold GARCH model for he French/Deuschmark exchange rae. Journal of forecasing, 20, pp Enders, W. (2004). Applied Economeric Time series. Second ediion. Willey series in Probabiliy and Saisics. Book published John Wiley and sons. Unied Saes. Engle R. (2001). The use of ARCH/GARCH models in applied Economerics. Journal of Economic Perspecives, Vol. 15 pp Engle, R.F. and R. Susmel, Common Volailiy in Inernaional Equiy Marke. Journal of Business and Economic Saisics, vol. 11 no. 2, pp Fafchamps, M. and R.V. Hill (2008). Price Transmission and Trade Enry in Domesic Commodiy. Economic Developmen and Culural Change Vol. 56, no. 4, pp Gilber, C and C. Brunei (1996). Speculaions, Hedging and Volailiy in he Coffee Marke, Queen Mary and Wesfield College, Universiy of London; Glosen, L., J. Ravi and E. David (1993). On he Relaion Beween he expeced Value and he Volailiy of Nominal Excess Reurns on Socks. Journal of Finance. Vol. 48, no.5, pp Guida, T. and O. Maringe (2004). Applicaion of GARCH models in Forecasing he Volailiy of Agriculural Commodiies. Minisry of Economy, Avan projec de 103

119 documen inerimaire de sraegie de reducion de la pauvree, Republique Togolaise, Lome, pp. 82p. Hill R.V (2010). Liberalisaion and Producer Price Risk: Examining Subjecive Expecaions in he Ugandan Coffee Marke. Journal of African Economies, Vol. 19, number 4, pp Inernaional Coffee Organisaion (ICO, 2005). Coffee Price Volailiy Sudy. Inernaional Coffee Council, Niney-fourh session. Salvador, Brazil. Kherallah, M., C. Delgado, E. Gabre-Madhin, N. Mino and M. Johnson (2000). The Road Half Travelled: Agriculural Marke Reforms in sub-saharan Africa. Food Policy Repor. IFPRI, Washingon, D.C. Krivonos, E (2004). The Impac of Coffee Marke Reforms on Producer Prices and Price Transmission. World Bank Policy Research Working Paper Karanja A, A. Kuyvenhoven and H. Moll (2003). Economic reforms and evoluion of producers prices in Kenya: An ARCH-M Approach. African Developmen review, vol. 15, pp Li Yu and Yuan C (2010). Geographical Srucure Based High Frequency Daa Disribuion GARCG Model and Empirical Analysis. Paper presened a he Second Inernaional Conference on Compuer Modeling and Simulaion. IEEE Compuer Sociey. Lee, J. and M.C Srazicich (2003). Minimum LM Uni Roo Tess wih Two Srucural Breaks. Review of Economics and Saisics, 63, pp Moledina A., T. Roe, and M. Shane (2004). Measuring commodiy price volailiy and he welfare consequences of eliminaing volailiy. Paper presened a he AAEA, Denver Colorado. 104

120 Newman, S. (2006). Terms of rade, price volailiy and disribuion in commodiy markes: An examinaion ino impacs of marke srucure on commodiy prices movemens. World rade insiue. Swizerland. Piroe, G., G. Pleyers and M. Pocele (2006). Fair-rade coffee in Nicaragua and Tanzania: a comparison. Developmen in Pracice, Vol.16, No.5, pp Ranbemananjara, R. and J. Zakoian (1993). Threshold ARCH models and Asymmeries in Volailiy. Journal of Applied Economerics 8: Shively, G. (2001). Price hresholds, price volailiy, and he privae coss of invesmen in a developing grain markes. Economic Modelling, volume 18 pp Wichern, R., u. Hausner and K. Chiwele (1999). Impedimens o Agriculural Growh in Zambia. TMD discussion papers No. 47. Trade and Macroeconomics Division, Inernaional Food Policy Research Insiue Washingon, D.C. World Bank (2009). Issue Briefs, Global Food Crisis, hp://go.worldbank.org/f28z

121 Chaper 5: Response of Coffee Supply in Zambia: Implicaions for Policy Absrac This aricle examines coffee supply response o variance incenives in Zambia. Resuls show ha in he long run coffee supply does no respond o coffee prices, bu ha i responds negaively o local currency (Kwacha) appreciaion. A he same ime, economic reforms which were implemened in 1998 have had a posiive effec on coffee supply. Vially, resuls show ha in he long-run, real exchange rae has had he highes impac on coffee supply. This implies ha currency liberalisaion has posiive effec on coffee producion in Zambia. To examine asymmeric shor-run supply adjusmens, unlike previous research, his sudy employs a dynamic hreshold error correcion model based on hreshold auo regression esimaion. Resuls confirm asymmeric shor-run adjusmen o long-run equilibrium. Paricularly supply response is significan for posiive shocks while insignifican o negaive shocks. These resuls show ha coffee supply does no adjus immediaely, unil he shocks reach a cerain hreshold. This sudy fills in a gap in he supply response lieraure, which has virually failed o analyse asymmeric supply response so far. 106

122 5.1 Inroducion High and susainable agriculural growh, mainly driven by agriculural produciviy is needed for African counries o reduce povery and foser economic developmen. Agriculural growh in Africa remains a key facor in economic developmen as i accouns for large shares of naional income, employmen and foreign rade (Forum for Agriculural Research in Africa, 2006, pp.7). The growh in agriculural produciviy is likely o depend on, among oher hings, adequae access o producive resources, well funcioning markes, infrasrucure, and a conducive policy environmen (e.g. sable macro economic policies). In he las wo decades, many counries in Africa implemened srucural adjusmen programs, which included agriculural policy changes, in he effor o provide conducive environmen for increasing oupu. Despie hese economic reforms, mos counries in sub-saharan Africa, including Zambia, coninue o face he challenges of declining agriculural produciviy compared o oher coninens, raising concerns on supply responsiveness. Coffee, an essenial expor commodiy in mos easern African counries, has suffered declining levels of producion amids declining and unsable prices. 18 Zambia, in paricular, experienced drasic decline in coffee producion from abou 6800 meric onnes (m) in 2005 o less han 2000 meric onne in 2009 (Figure 5.1). The causes of he decline in coffee producion in he counry have never been invesigaed. A plausible explanaion could be he decline in coffee producer prices or he srenghening of he Zambian currency (Kwacha) in he las decade. Despie he fac ha, economic policies in Zambia have, for a long ime, emphasised expor diversificaion from he radiional expor commodiy, copper, o non-radiional expors like coffee, here has no been any sudy ha has looked ino he economic aspecs of coffee. To he bes of my knowledge, his is he firs sudy o examine supply response of an agriculural expor commodiy in Zambia. Despie an increasing body of lieraure on supply response in sub-saharan Africa in he recen pas, very lile similar work has been done for Zambia. The few supply response sudies in Zambia focused on Maize, obviously because i is he larges produced crop as well as he saple food (see e.g. Wold, 1997, Xu e. al. 2006; Nyairo, 2009). 18 Inernaional Coffee Organizaion (ICO) price daa

123 Figure 5.1: Coffee Producion in Zambia MT / / / / / / / / / / / / /09 PRODUCTION MT QUANTITY EXPORTED MT AVERAGE VALUE US$/MT An undersanding of agriculural commodiy supply responsiveness becomes a useful guide in economic policy formulaion, paricularly hose relaing o incenives for producion. I is imporan for policy makers o know precisely he supply responses of commodiies if effecive policies are o be implemened. Supply response sudies become relevan in providing empirical evidence for policy makers o idenify key variables ha are imporan in deermining agriculural commodiy supply. Basically, agriculural supply response explains he elasiciy of oupu adjusmen o various policy and oher producion incenives. As Rao (1989) explains, he exen o which farm producion decisions respond o informaion on various incenives should be cenral in policy planning. The raional expecaions of supply uilises all informaion o generae predicions of producion incenives. For example, high prices should provide incenives for more producion such ha quaniy supplied should move linearly and in he same direcion as price. 108

124 However, wih recen revelaion of asymmeric behaviour of mos economic and financial ime series, he heoreical assumpions of linear adjusmen of supply o price or oher incenives become inappropriae and may lead o misleading conclusions. The implicaion of a linear adjusmen is ha a shock o he price is assumed o lead o he same response in he oupu, regardless of wheher he shock refleced a price increase or a price decrease. Ye in mos cases, he reacion of commodiy supply o increases in incenives is differen from is reacion o decreases in he incenives. I is in his regard ha his sudy focuses on asymmeries in supply response, a missing subjec maer in mos supply response lieraure. A large body of lieraure exis on symmeric supply responsiveness for differen agriculural commodiies across Africa. Presumed facors affecing supply response vary widely across commodiies as well as across counries. Abdulai and Rieder (1995) concluded ha cocoa supply in Ghana is significanly influenced by real producer prices of cocoa, real maize prices (as a subsiue crop), real exchange rae and supply of manufacured goods. Muchapondwa (2009) found ha price incenives did no affec aggregae agriculural supply in Zimbabwe beween 1970 and In his/her sudy, Molua (2010) found ha irrigaion, expor promoion and access o affordable finance significanly affec produciviy and supply of rice. Rahji and Adewumi (2007) aribue he increase in local rice supply in Nigeria o a ban on rice imporaion coupled wih provision of producion incenives especially of cerified seeds, ferilizers and agro-chemicals. In Uganda coffee farmers were responsive o producer prices, such ha when prices increased, hey responded by raising producion (Oim and Ngaegize, 1993). Kidane (1999) argued ha coffee supply in Ehiopian was acually responsive o he real exchange rae, despie argumens by some economiss ha small scale farmers in developing counries do no respond o price incenives in a raional and predicable manner. In view of he oucome of heses sudies, i is quie eviden ha facors ha influence supply vary widely depending on he commodiy as well as he counry. The mehodologies applied in hese sudies also vary widely. While mos sudies employ he Nerlove (1958) model, more recen sudies (e.g. Abdulai and Rieder, 1995; Alemu e. al. 2003) have adoped a modelling echnique ha recognises he use of coinegraion and error correcion o overcome he problem of spurious resuls 109

125 arising from he use of inegraed series in he Nerlove model. As noed earlier, hese sudies employed linear coinegraion and error correcion esimaion o arrive a conclusions of supply elasiciy. Linear esimaes canno differeniae he effecs of posiive and negaive price shocks on supply. As lieraure would reveal, very lile aenion has been paid o asymmeric supply response. Shirvin and Wilbrae (1999) examined asymmeric response of domesic prices o impor prices in differen developed counries using asymmeric error correcion models. Their sudy however, differs from his sudy in he segmenaion of he error correcion erm. In his sudy, a hreshold is idenified in a hreshold variable, which is he error erm, using Threshold Auoregression (TAR) models o examine hreshold coinegraion. Shor-erm asymmeric supply adjusmens are hen analysed using hreshold error correcion models. The overriding objecive of his sudy is o esimae an asymmeric coffee supply response o price incenives. However, i would be difficul o obain robus resuls if oher imporan incenives o supply are no conrolled for. Therefore, he sudy includes prices of compeiive crops (specifically maize) and real exchange rae (given ha coffee is grown specifically for expor) in an asymmeric auoregression model. Moivaed by he work of Deaon and Laroque (2003), his sudy no only akes ino accoun asymmeries, bu also he naure of he coffee circle as a perennial crop. Thus, I include lags in he supply response analysis considering he fac ha coffee akes up o four years before he firs harves as well as he fac ha mainaining curren rees may have implicaions on supply. Supply response is herefore, likely o go over one period. A he same ime, i is imporan o noe ha, while economics and economerics assume a sable policy srucure, which clearly is no he case in Africa, where poenial srucural breaks in fundamenal economic behaviour make modelling problemaic. Unlike previous sudies, his sudy pays aenion o endogenous idenificaion of srucural breaks, paricularly hose relaing o economic reform periods. The sudy is designed o provide necessary informaion for policy planning on he response of coffee supply o various incenives following policy regime changes in he agriculural secor. The aim is o inform policy makers abou he effecs of an increase or a decrease in he price of coffee and oher incenives will have on coffee supply. 110

126 The res of he paper is organised as follows: The nex secion discusses he economic heory in which supply response is embedded. Secion hree describes he mehodology used in modelling supply response, and is followed by a descripion of daa in secion four. The resuls of he analysis are presened and discussed in secion five which is followed by a conclusion in secion six. 5.2 Theoreical Framework Modelling supply response has is roos in he parial adjusmen, adapive adjusmen and raional expecaion heories embedded in he heory of a firm 19. The parial adjusmen and adapive adjusmen model developed by Nerlove (1958) capures he dynamics of agriculural supply wih regards o price and oher expecaion. Specifically, he models asses he farmers opimisaion behaviour and how hey reac o changing arges wih respec o changes in oher incenives. The parial adjusmen model shows a dynamic parial adjusmen where an observable oupu variable y is adjused, hrough ime, o an expeced or desired bu unobservable oupu variable y * in he equaion below: y y = β * ( y y 1) 1, 0 < β < 1 (1) 19 A profi maximising firm is observed how i adjuss is supply in response o price expecaions. Given ha supply response is only focused on he oupu supply funcion, and no necessarily on he demand funcion, he model only considers he firm decisions on oupu supply. Since any profi maximising firm assumes ha oupu opimisaion has already been achieved in he inpu space, i will produce oupu up o he poin where i equaes marginal revenue o is marginal cos. The cos minimising behaviour equaes marginal coss o price, which he firm canno influence in a compeiive marke, as he case is wih farmers. Being a price aker, he firm sees he marke price for is produc, assumes i will remain he same regardless of how much or how lile i sells and makes is plans accordingly. Since he main objecive is o maximise profis, he firm chooses he level of oupu and he combinaion of facors ha solve he equaion Max pf (y) - x 0 i = 1 subjec o i. The soluion o he equaion explains how much oupu he firm will sell and how much of which inpus i will buy. n w i x i 111

127 Equaion (1), posulaes ha change in oupu beween he curren and previous is a fracion of he difference beween he oupu opimum level and he previous year s oupu. In oher words, a any paricular ime (period ), only a fixed fracion ( β ) of he desired adjusmen is accomplished. Equaion (1) shows ha he adjusmen of he oupu beween period and -1 is equal o β y ) where he coefficien β ( * y i measures he speed of adjusmen assuming he values beween 0 and 1. The larger he value of β, he faser he adjusmen such ha when β = 1 adjusmen is insananeous and he smaller he value he longer i akes for oupu o adjus o he desired level, implying a higher adjusmen lag. However, he argeed or desired oupu * y is no observed, bu is influenced by various oher observed facors x such ha: * y x = (2) where x are curren or lagged variables of oher economically imporan variables. Subsiuing 2 ino 1 gives y y = β ( x y ) 1 (3) 1 where x is a vecor of variables influencing he firms decision o inves in aaining he opimal or desired oupu. These variables could be lagged prices, lagged oupu, and lagged prices of subsiue crops. A major limiaion o he Nerlovean approach is he possibiliy of spurious regression arising from no aking ino accoun saionariy properies of ime series. Furhermore, he long run price elasiciy canno be esimaed using he Nerlove model, unless assumpions of wheher i is a parial adjusmen or price expecaion model are made. This implies no forward looking behaviour by he farmers. 112

128 Nerlovean model has been has been exensively applied in modelling supply response in a wide range of agriculural empirical research. 20 Neverheless, he Nerlovean model is used in connecion wih he adapive expecaion. In adapive expecaions, he farmer makes his/her expecaions based on wha happened in he pas. In he equaion below, if * y is he expeced value of some variable which is unobserved. In he adapive expecaion model, he adjusmen of expecaions from ime o -1 is presened as: y * γ δ where 0<δ <1 (4) * = x + ( 1 ) y 1 Where x is he observed value of x in period, raio or he value of x expeced o prevail in he period -1, * y 1 is he arge level of a paricular * y is he value of he raio for or he value of x expeced o prevail in period. δ is he coefficien of expecaions, he proporion of he curren change in he indusry mean/median aken o be permanen raher han ransiion. Unlike he adapive expecaions, in raional expecaions heory, he farmer akes ino accoun all informaion available o make his expecaions. I is ofen argued ha raional expecaions are model consisen in ha i assumes ha marke oucomes ha are forecased do no deviae widely from he equilibrium, bu ha he deviaion becomes consisen wih assumpion of random sandard error in modelling expecaions (See Muh, 1961; Seay e. al, 2004). This sudy examines supply response in he framework of raional expecaions heory. Supply response basically examines speed and magniude of changes in planned oupu in response o anicipaed oupu prices. However, neiher planned oupu nor anicipaed price is observable, because firs, weaher and oher environmenal facors can make observed oupu deviae from planned oupu and second, because he farmer only knows pas and curren prices (Rao, 1989). Proxies 20 See Askari and Cummings (1977) for a survey of he supply response approaches in agriculural commodiies. 113

129 for hese variables, herefore, need o be employed, alhough having in mind ha he choice of he proxy has significan bearing on he resuls. For example, several economiss sugges he use of acreage as proxy for oupu because i is hough o be more subjecive o farmers conrol (Askari and Cummigs, 1977; Rao, 1989). However, if yield improves given unchanged acreage, his proxy can be misleading. So far, empirical work of supply response o price of agriculural commodiies is found o be consisen wih raional expecaion heory (Rao, 1989). In heory, a rise in marke price of a commodiy encourages producers o produce more, such ha, an increase in price means a rise in supply. Mos ime series sudies find posiive supply response for specific crops wih respec o relaive price changes (Rao, 1989). In general, shor-erm supply elasiciy of oal agriculural oupu o changes in prices ranges from 0.1 o 0.3 (Meerman, 1997). In he long-run, because of he mobiliy of facors of producion such land, labour and Capial, he producion response o improved prices for he enire agriculural secor is much higher- 0.3 o 1.2 (Rao, 1989). Such ha, when prices are no aracive in he long-run, no only are he resources reallocaed o oher compeing crops, bu also o oher uses. Abdulai and Rieder (1995) argued ha if farmers receive high prices for heir commodiies relaive o oher goods, hey would be encouraged o increase oupu wihin heir limis. Wih regards o economic reforms, mos empirical findings are consisen wih heory depicing ha agriculural producion and expors increases in response o srucural adjusmen programs (Meerman, 1997). However, one consrain o faser producion response is he inconsisence of he reforms. Meerman (1997) argues ha, he privae secor become relucan o inves if Governmen ownership and he persisence of he reforms are in doub, such as in siuaions where reforms have ofen imes been reversed or haled and fall shor of full liberalisaion. Closely linked o economic reforms and commodiy rade liberalisaion is currency liberalisaion. The heory of geing he price righ hrough exchange rae devaluaion implies an increase in prices and profiabiliy of expor crops relaive o locally consumed commodiies, hence an imporan incenives for increasing oupu supply (Meerman, 1997). In a sudy of cocoa in Ghana, Abdulai and Rieder (1995) found ha overvaluaion of domesic currency serves as a disincenive o producion of radable, paricularly expor commodiies. They argue ha, o provide incenives o producers 114

130 of expor crops, he general macro-economic environmen ha deermines real exchange rae needs o be pu on a sound base. 5.3 Model Esimaion In order o address he Nerlove model limiaions and o give consisen esimaion of shor and long-run supply elasiciy, his sudy applies coinegraion and error correcion models. Two models are esimaed and resuls compared. The firs model employed in he analysis i.e. he radiional coinegraion and error correcion model (ECM) models assumes linear or symmeric adjusmen o long-run equilibrium, while he second model considers asymmeric adjusmen. In he second phase, hreshold coinegraion and error correcion models is employed o assess asymmeric adjusmens of supply o various incenives. In general, error correcion specificaions examine shor-run adjusmens o a long-run equilibrium. Several researchers including Abdulai and Rieder (1995), Mc Kay (1999) and Thiele (2003), have applied coinegraion and ECM in supply response analysis in differen commodiies. Unlike he Nerlovean model, coinegraion models demand ha he variables in quesion mus be inegraed of he same order in order for he long-erm relaionship o be esablished wihou giving spurious resuls. According o Engle and Granger (1987), wo or more non-saionary variables may wonder apar in he shor-run, bu in he long-run here exiss a relaionship which iself is saionary. The long-run relaionship beween coffee supply and coffee price, maize prices, real exchange raes and policy changes is given by: Q c m = β 0 + β1p + β 2P + β3rer + β 4RF + ε (4) where: Q Logarihm of coffee producion a ime c P = Logarihm of coffee prices a ime m P = Logarihm of maize prices a ime RER = Logarihm of real exchange rae for he kwacha dollar RF = Dummy variable for economic reforms 115

131 and β i are he parameers o be esimaed. All he variables in equaion (4) mus be non-saionary and inegraed of he same order e.g. I(1). ε is a disurbance erm ha may be serially correlaed. If a long-run relaionship exiss, hen he error erm ε is saionary even when he individual variables are non-saionary. The well known augmened Dickey-Fuller (ADF) saisic is used o ascerain wheher he residuals ε are saionary in he following equaion: p 1 + γ ˆ iε 1 + i= 1 ˆ ε = ρεˆ υ (5) where υ is a whie-nose disurbance. The null hypohesis of uni roo is no rejeced if ρ = 0 agains an alernaive hypohesis of 2 < ρ < 0. In he case of equaion (4), he series are said o be coinegraed if he null hypohesis of uni roos in ε is rejeced. Tha means presence of a unique long-run equilibrium beween coffee supply on he one side and coffee producer prices, real exchange raes and maize prices on he oher side. The ECM is presened as; Q k k k k k c m i j + β1i Q1 j + β2i P j + β3i P j + β4i RER j + 5i RF + j= 1 j= 1 j= 1 j= 1 J = = δ ( ECM ) β v (6) The erm ( ECM i ) j represens he error correcion erm υ which is derived as he residuals from he coinegraion esimaion from equaion (4). A major limiaion of he linear coinegraion models is ha hey implicily assume a linear adjusmen mechanism conrary o he characerisics of mos ime series daa (Enders and Siklos 2001). According o Enders and Granger (1998) he symmeric or linear coinegraion ess are misspecified if he adjusmen o long run equilibrium is asymmeric A deailed discussion of symmeric and asymmeric coinegraion mehodologies is given in Chaper

132 To accoun for asymmeric supply response, hreshold error correcion models (TECM) is applied wihin he framework of hreshold auoregression (TAR). Building on he works of Enders and Granger (1998), Granger and Lee (1998), Enders and Siklos (2001) and Enders (2004), a mulivariae hreshold error correcion specificaion is developed in his sudy o asses possible asymmeries of shor-run supply adjusmens o is long-run equilibrium based on changes in coffee prices, maize prices, real exchange raes and economic reforms. To examine hreshold coinegraion and error correcion adjusmens, he residuals ( ε ) obained from he long-run relaionship (equaion (4)) are used in he esimaion of he TAR (equaion (7)). To esablish a hreshold, his sudy follows he procedure by Chan (1990, 1993) and Enders (2004) where he hreshold is seleced from all possible values of a hreshold variable by a grid search. The selecion of he hreshold is done alongside esimaion of β where all Sums of Squared Residuals (SSR) from he TAR models are recorded. The TAR model wih a hreshold ha minimises SSR is seleced. A TAR model is presened as follows: k k Z = I α + α iz i + 1 (1 I) α 2o + α 2iZ i + µ i= 1 i= 1 (7) where I is he Heaviside indicaor funcion such ha: 1 I = if 0 Z Z 1 1 τ < τ Z is he fis difference of he residuals derived from he long-run coinegraion esimaion in equaion (4); I = 1 if Z τ ; I = 0 if Z < τ. This implies ha when Z 0 1, I = 0 an d (1- I )= Examining hreshold coinegraion is equivalen o esing he null hypohesis of no coinegraion by esing he join resricion ha α α 0 in equaion (7), based 1 i = 2i = 117

133 on F-saisics and is significance level. The F-saisics values are abulaed in Enders and Enders (1998). If he null hypohesis of no-coinegraion is rejeced, hen asymmeric adjusmens can be esed. Examining asymmeric supply adjusmens is equivalen o examining he presence of a hreshold which is done by esing wheherα = in equaion (7). If he hypohesis is rejeced, hen he sysem has 1i α 2i asymmeric adjusmen effecs. In ha case, asymmeric shor-run adjusmens o he long-run adjusmen are esimaed using a Threshold Error Correcion Model (TECM). As suggesed by he raionale expecaion heory, asymmeric adjusmen in supply response is reaed as a shor-run phenomenon. Therefore, he conclusions are mainly drawn on he shor-run esimaions. In order o render he resuls robus wih respec o he asymmeric effec of he dependen variables on supply, his sudy develops a TECM for each regime, i.e. for he adjusmens above he hreshold and hose below he hreshold. Each model is hen esimaed as a normal symmeric ECM given in equaion (6). The wo models are represened as: k k k k c m 1 + α k Q k + β k P k + δ k P k + λk j= 1 j= 1 j= 1 j= 1 + Q = λ Z + RF + v (8) k k k k + c m 1 + α k Q k + β k P k + δ k P k + λk j= 1 j= 1 j= 1 j= 1 Q = λ Z + RF + v (9) Where + Z 1 and Z 1 are derived from (7) depending on wheher he sysem is above or below an esimaed hreshold. + Z 1 are posiive shocks above a hreshold and Z 1 are he negaive shocks below a hreshold. Similarly + λ and λ represen he adjusmen coefficiens when he sysem is above and below he hreshold respecively. 5.4 Daa Descripion Annual series daa covering he period 1983 o 2008 is used for he empirical analysis. The variables include annual coffee producion in meric onnes, coffee 118

134 producer prices in US cens per pound (lb), 22 real maize prices ZMK per kg and real exchange rae for he Kwacha currency. Coffee producion and price daa were obained from he Inernaional coffee organisaion (ICO), daa on real exchange raes and maize prices ( ) where obained from Bank of he Zambia (BOZ), while he oher par of he maize prices (1983 o 1994) where obained from a sudy conduced by Wold (1997). A dummy variable idenifying he economic reform is included o conrol for a poenial srucural break. This srucural break was deermined endogenously using he Lee and Srazichici (2003) srucural break uni roo ess in he real exchange raes variable. I choose o use he exchange rae variable o deermine economic reform because currency liberalizaion ook a cenral sage during he economic reforms. The original daa on exchange rae and for maize prices where in he local currency needed o be deflaed wih a price index. Consumer price index (CPI) wih he year 2000 as he base was used o deflae exchange raes for he enire period and maize prices for he period 1994 o Maize prices which were obained from he sudy by Wold (1997) covering he period 1984 o 1994 were already in real erms, deflaed using low income CPI. Figure 5.2 plos hese variables. Figure 5.2: Supply Response Variables 140 CPRICE 7000 CPRODUCTION RER 1200 REALMPRICES Lb is 0he abbreviaion for Libra which is he roman word 0 for pound. One pound is kilo grams (kgs). 119

135 Cprice are Zambian coffee prices in US cens/ib; Cproducion is coffee producion for Zambia in meric onnes; RER is he Real exchange rae expressed as Zambian Kwacha (ZMK) per USD Raionale for Variable Selecion The selecion of he variables was based on heoreical undersanding of supply response discussed in secion 5.3. In he firs place, farmers make heir producion decisions based on pas price informaion. As noed by Askari and Cummings (1997), crop specific price risk increases supply responsiveness paricularly of peasans whose livelihood may be hreaened by down-side risks. This is especially so wih expor crops like coffee which demonsrae acue period o period volailiy. In heory, high prices are supposed o moivae farmers o produce more. Thereby, prices having a posiive effec on producion of a paricular commodiy. However, in perennial crops, where supply ends o reach he marke when prices are on he decline, he price coefficien may no be posiive. For coffee in paricular, when prices increase, farmers are moivaed o increase producion, such ha excess producion may lead o price decline. When prices decline, farmers ge discouraged from making invesmens in new rees or mainaining he old ones, resuling in low yields and less supply. In he end, prices rise again due o low supply. This is exacerbaed by he fac ha coffee is a perennial crop wih lags beween planaion and harvesing varying beween 18 o 24 monhs. Peak yields are only experienced afer 5 o 7 years. This implies ha, while inpu invesmens respond quickly o price changes, supply response is very slow. Consequenly, addiional supply ends o reach he marke when prices are on he decline. Second, alernaive crops prices may influence coffee producion. As maize is Zambia s main sable crop, is successes and failures in erms of yield and price may have a bearing on he farmers decisions o grow an alernaive cash crop such as coffee. Askari and Cummings, 1997 explain ha where muliple cropping is possible and rigid paerns of land use are no dicaed by subsiue requiremens, farmers se of choices is wider. Hence, responsiveness is likely o be greaer. In his case, coffee supply response o maize prices is expeced o be significan bu negaive. 120

136 Third, Zambia embarked on economic reforms in he 1990s when agriculural rade was liberalised. The change from a conrolled sysem ha included price fixing could have significan effec on he coffee supply. As explained by Rao (1989), he analysis of long-run response mus disinguish changes in supply condiions ha are brough abou by he decenralised privae acions from changes ha resul from cenralised public acions. Given ha rade liberalisaion encourages privae raders which could lead o increased and more efficien markes, he coefficien of he dummy variable for economic reforms is expeced o be posiive. Wheher he coefficien would be significan is subjec o invesigaion because rade liberalisaion did no direcly affec he coffee indusry. Even before he reforms he governmen did no much conrol of he indusry as i was relaively small and insignifican. However, he reforms in oher secors of he economic such as exchange raes, privaisaion of sae owned esaes and decenralisaion of he agriculural insiuions could have an impac on he coffee indusry. Fourh, exchange raes influence farmers decision o increase supply. A sronger currency makes expors more expensive han a relaively weaker currency relaive o prices of locally consumed commodiies. The rend in mos cases is ha a sronger currency resuls in a decline in expors. Abdulai and Rieder (1995) argue ha overvaluaion of he domesic currency serves as disincenive o producion of radable, paricularly of expor commodiies. In he las decade Zambian Kwacha seadily gained srengh agains he US Dollar. Hypoheically, a srong currency reduces profiabiliy of expors crops. As such a posiive coefficien for real exchange rae is expeced in boh he long and shor-run model esimaion. Figure 5.4 shows Zambia s annual inflaion rae in he las eigh years. 121

137 Figure 5.3 Zambia Annual Inflaion Rae ( ) Inflaion Rae Source: Bank of Zambia (BOZ) daa a Resuls The use of coinegraion demands ha he variables be non-saionary and inegraed of he same order. To assess saionariy of each variable, he ADF ess (equaion 5) uni roo ess are employed. The number of lags o include in he ADF model is seleced using he Akaike Informaion Crieria (AIC). 23 A general-o-specific approach is also employed o verify he AIC selecion. 24 The resuls presened in Table 5.1 show ha all variables are non-saionary in levels while heir firs differences are saionary. 23 Akaike s (1987) Informaion Crieria (AIC) and Schwarz s (1983) Bayesian Crieria (SBC) are commonly used o deermine he number of parameers o include in a model. The model giving he smalles value of AIC and SBC is considered bes and mos likely gives he bes goodness of fi. Based on heoreical explanaions and various simulaion sudies, SBC is preferred for large samples because he AIC ends o selec models wih oo many parameers when he sample size is large. However, he number of observaions in his sudy is 26 for all he variables jusifying he use of AIC. 24 In a general-o-specific approach, differen lags are included in he uni roo es unil he appropriae lag is reached. The procedure involves saring wih many lags and reducing unil he appropriae number of lags is reached, based on he resuls of he uni roo es as well oher indicaive values for auocorrelaion such as he Durbin Wason (DW) value. 122

138 Table 5.1: Uni Roo Tess for Supply Response Variables Wih Consan Wih Consan and Drif Level Firs difference Level Firs difference Coffee Producion (2) *** * 3.555*** Coffee Price(2) *** *** Maize Prices(5) *** *** Real Exchange raes(1) ** *** *,** and *** denoe significance a 10%, 5% and 1% respecively. The number of lags for each variable are indicaed in parenheses. According o he -saisics, he null hypohesis of uni roos canno be rejeced a he 1 percen level for all variables in levels. However, he null hypohesis of uni roos is highly rejeced when applied o firs differences of he variables. Tha means ha all he variables are inegraed of order one or are I (1). Coinegraion analysis has been carried ou using boh he Engle and Granger (1987) and Johansen (1988) ess, which presumes a linear coinegraion, and he TAR model which akes ino accoun possibiliies of asymmeric coinegraion. Furher, boh symmeric and asymmeric error correcion models are esimaed o examine he shor run adjusmens o long run equilibrium Symmeric Coinegraion and Error Correcion Resuls The Engle-granger coinegraion procedure involves wo seps. In he firs sep longrun relaionships are esablished beween coffee supply a ime and he explanaory variables (coffee producer prices, maize prices, he real exchange rae and economic reforms) as specified in equaion (4). The resuls are presened in equaion (10) where -saisics obained from ADF disribuion are given in parenheses. 123

139 c m Q = P P RER RF (10) (8.910) (-0.317) (0.086) (9.031) (2.373) 2 R = AIC = The mos significan variable in deermining supply response is he exchange rae in real erms. A posiive coefficien of he exchange rae means ha when he Zambian Kwacha weakens agains he dollar (an increase in he exchange rae) coffee supply increases. Given ha coffee is grown solely for expor, a srong currency makes he commodiy less compeiive, hence unaracive for farmers o plan more or inves more in he already growing rees. As equaion (10) shows, one uni appreciaion of he Zambia Kwacha leads o 0.33 percen increase in coffee supply in he long-run. Thus, he 2004 peak coffee producion ha Zambia experienced can be associaed wih he 21.4 percen annual inflaion rae in he same year (Figure 5.4 above). In addiion, he resuls show ha economic reforms, which occurred in 1998 (according o a srucural break in he real exchange rae), have had a posiive impac on coffee producion. Since he -saisic value of on he economic reform coefficien is higher han he 5 percen criical value of 1.708, he impac is significan. The coefficien is posiive, an indicaion ha coffee producion increased. On he conrary, he effec of coffee prices on coffee producion in Zambia in he long-run is no significan. A plausible explanaion for he insignifican coffee prices is ha he prices are in dollars and ha is wha he farmers receive. Therefore, he incomes for he farmers, in he local currenly, grealy depend on he exchange rae. As long as he local currency coninues o weaken, coffee farmers will always find i aracive o produce more, even if price remains he same. Similarly, price for maize, he compeing crop wih coffee, has no significan impac on supply. An explanaion for his oucome is ha land may no be a facor for he large scale farmers, such ha maize is no grown as an alernaive crop, bu as complemenary o maize. Neverheless, he second sep of he Engle Granger coinegraion approach, is o es saionariy of he residuals ε. As menioned earlier, ADF ess are employed for his 124

140 es (equaion (5)). The null hypohesis of non-saionary presened by ρ = 0 was rejeced a 0.01 given a -saisic of This value is greaer han he 1 percen criical value of which jusifies he rejecion of he null hypohesis. 25 Given he possibiliy ha here may be more han one coinegraion vecor, he mulivariae coinegraion approach proposed by Johansen and Julius (1990) was also applied o examine he number of coinegraion vecors. Unlike he Engle and Granger coinegraion es, he Johansen es does no assume a single coinegraion relaionship. I herefore becomes necessary o apply he Johansen es as well o esablish he number of coinegraion relaionship. The resuls are presened in Table 5.2. The procedure involved esing for he number of coinegraion vecors beween coffee supply on one side and coffee prices, real maize prices, real exchange rae and a dummy of economic reforms on he oher. Resuls indicae wo coinegraion equaions significan a 5 percen level. The Johansen resuls provide furher evidence supporing he resuls of he Engle-Granger ess which shows evidence of coinegraion beween he variables as shown in equaions (10). As in he Engle granger coinegraion es, he Johansen es shows ha a mos wo variables have a significan effec on coffee supply. Table 5.2: Johansen Coinegraion Tes Unresriced Coinegraion Rank Tes (Trace) Hypohesized Trace 0.05 No. of CE(s) Eigenvalue Saisic Criical Value Prob.** None * A mos 1 * A mos A mos A mos Trace es indicaes 2 coinegraing eqn(s) a he 0.05 level * denoes rejecion of he hypohesis a he 0.05 level **MacKinnon-Haug-Michelis (1999) p-values 25 The criical values are for coinegraion relaionships for four variables wih a consan in he coinegraion vecor. The values are abulaed in Enders, 2004 pp

141 Having esablished he exisence of a long run relaionship, a symmeric ECM, is employed o capure shor-run supply adjusmens o he long-run equilibrium esimaed using equaion (4). The ECM presened in equaion (6) esimaes he shor-run adjusmens of supply oward he long-run equilibrium. The maximum order for he ECM, given he number of observaions for his sudy, is wo. The DW value confirms he absence of auocorrelaion when one lag is included for each variable. As Deaon and Laroque (2003) have shown, coffee supply response o increasing coffee prices can occur wihin wo years; hence a model wih wo lags is sufficien o capure coffee supply elasiciy. Table 5.3 shows he shor-run adjusmens. The error correcion erm (ECT) is negaive as expeced from ECM and significan a he 15% level. This means ha when all incenives increase, supply does no increase proporionally; insead i lags behind creaing a negaive disequilibrium which ges correced over ime. The deviaions from he long-run equilibrium are shor lived, such ha 0.41% of he previous year s disequilibrium from he long-run relaionship is correced in he curren year. Table 5.3: Symmeric Error Correcion Esimaes β -saisics Sandard Error i Φ * * Φ c p c p m p m p 2 RER 0.326* RER RF RF ECT * C * ** Significance a 15% 126

142 The changes in he RER seem o have a larger impac on changes in supply, given a 15% significan coefficien. The coefficien on he RER is also posiive supporing he argumen ha a depreciaion of he currency (as he exchange rae value increases) aracs invesmen in agriculural expors. The coefficien on price is posiive alhough he shor erm adjusmens for wo lags are no significan. Similarly he shor run adjusmen for maize prices is no significan for he firs lag alhough i has he expeced negaive sign. The previous period s producion has relaive effec on curren period s producion given he significan coefficien a 15% level. Deailed vecor error correcion esimaion is presened in Table 5.4. The -saisics on he esimaed coefficiens give an indicaion ha coffee prices, exchanges raes and maize prices do no influence each oher in he shor-run Asymmeric Coinegraion and Error Correcion Resuls The resuls discussed above do no differeniae beween effecs of posiive and negaive shocks on supply response. Due o a possibiliy of asymmeric supply movemen, hreshold coinegraion was examined using he TAR model (equaion 7). 26 The resuls indicae he exisence of a hreshold of The selecion of his hreshold was based on grid search for he smalles SSR compued from TAR esimaions using all possible hreshold values. The null hypohesis of no hreshold coinegraion, which is equivalen o a es ha α α 0 in equaion (7) was srongly rejeced a 5%. The es gives a P-value of 1 i = 2i = wih level significance. This value is greaer han he 5% criical value of 3.55 wih 2,18 degrees of freedom. A es of asymmeric supply response o he longrun equilibrium involves a Wald es of he null hypohesis ha α = in equaion 1i α 2i (7). If supply response o all he variables is asymmeric, hen i is expeced ha α1i differs significanly from 1i α. The hypohesis is rejeced a 5% giving a P-value of which is greaer han The resuls mean ha coffee supply adjuss according o wheher a deviaion from he long-run equilibrium is above a hreshold of or below he hreshold. 26 The TAR model was esimaed using Regression Analysis for Time Series (RATS) program. 127

143 The resuls from he TAR model imply ha posiive shocks (represened as Z 1 in he ECM) and negaive shocks ( Z + 1 ), have differen impacs on he adjusmen of coffee supply in he long-run. Therefore, shor run adjusmens o he long-run equilibrium can be deermined based on esimaions of he wo ECMs presened in equaion (8) and (9). 27 Resuls, which are presened in Table 5.4 and 5.5 shows ha, for all he variables apar from coffee prices, shor-run adjusmens of coffee supply o is long-run equilibrium ends o occur only when he shocks are above he hreshold and no when he shocks are below he hreshold. Regarding he changes in coffee prices, resuls show ha abou 0.26% of he deviaions from long-run equilibrium are correced back whenever he price changes move above he hreshold. Below he hreshold, maize prices do no have any significan effec on supply. Wha his means is ha coffee price increases only lead o increases in coffee supply in he shor run if he deviaion from long-run equilibrium ges above he hreshold. However as explained earlier, given he naure of he coffee ree as a perennial crop, is bes examined in a long-erm. Changes in maize prices do no seem o have any significan effec on coffee supply in he shor run, wheher he sysem is above or below he hreshold. For real exchange rae, hreshold shor run adjusmen resuls show ha whenever he Zambian Kwacha depreciaes by one uni above he hreshold, coffee supply increases by 0.19%. 27 One lag each variable is included in he models. The model for posiive shocks gives an AIC value of 0.449, SBC of and a DW value of 2.24 while he AIC value for he negaive shock model is 0.516, he SBC is and he DW is Given ha Coffee is a perennial crop which can ake up o 4 years before he firs harves and ha he rees can las up o 50 years, more lags would be needed o explain he adjusmens o shocks ha occurred several years back. However, he bes model was seleced for one lag. This could be parly because of he small number of observaion in he sample. 128

144 Table 5.4: Error Correcion Esimaion above he Threshold Variable Coefficien -saisics Sandard Error Φ * p c * m p 1 RER 0.191* RF 0.330* Z 0.216* C 0.790** Obs 24 2 R DW 2.23 Table 5.5: Error Correcion Esimaion below he Threshold Variable Coefficien -saisics Sandard Error Φ 1 + p c 1 m p RER RF Z -o C 0.501* Obs 25 2 R 0.18 DW 2.50 *, **, *** represens significance a 10%, 5% and 1% respecively This demonsraes ha farmers find exporing profiable when he Zambian Kwacha depreciaes agains he US Dollar. 129

145 5.7 Conclusions The sudy invesigaed coffee supply response in Zambia using hreshold error correcion model. The main objecive was o address he quesion of wheher changes in coffee supply do respond asymmerically o various incenives, hus deparing from earlier sudies on his opic ha paid no aenion o asymmeric adjusmens. In his sudy asymmeric adjusmens are examined boh in he long-run and he shor-run, such ha he error erm is spli ino wo series; when he deviaion from equilibrium is above he hreshold and when i is below he hreshold. Several conclusions are drawn from he resuls. Firs, here is robus evidence ha a srong currency does no favour coffee producion and expor in boh he long-run and he shor-run. The resuls have shown ha when he Zambian currency (he Kwacha) appreciaes, coffee growers receive less Kwacha for he same quaniy of expors, alhough he value in US Dollars remains he same. This in reurn discourages he growers as mos producion expenses such as labour and oher inpus are paid for in he local currency. The resuls are consisen wih heory of geing he price righ hrough currency devaluaion, which leads o profiabiliy of expor crops relaive o locally consumed commodiies. As such, policies should focus on puing in place a general macroeconomic environmen ha deermines exchange rae needs. Over valuaion of he exchange rae discourage he producion of expor commodiies. Coffee prices which are hypoheically he main facor influencing producion have been found o have no significan effec in he case of Zambia. A conclusion can be made here ha in perennial crops like coffee, prices become endogenous o supply. The supply response o prices ends o be slow in ha i reaches he marke when prices are on he decline. Furhermore, he farmers are more concerned abou he exchange rae in heir producion decisions and no on he prices per se. This is because hey receive coffee revenues in US dollars while heir expenses are in he Zambian Kwacha. In addiion he coefficien for economic reforms, which is highly significan, demonsraes ha a liberalised economy favours he producion of coffee in he shor run. 130

146 Regarding asymmeric supply response, he sudy found ha posiive shocks o he variables such as increase in coffee prices have more impac on coffee producion han he negaive shocks. Such asymmeric response canno be capured in he linear coinegraion models. The applicaion of hreshold coinegraion models which has provided evidence ha here is no reason for presumpion ha coinegraion is linear. As observed in he sudy, supply may no coninuously adjus o is long-run equilibrium unil he changes in he influencing variables reach a cerain hreshold. Analysing supply response wihou aking ino accoun asymmeric effecs, can lead o misleading resul. Therefore, in comparison o he symmeric coinegraion models, a conclusion can be made ha he TAR mechanism provides a sraighforward and a more meaningful explanaion of ime-series daa adjusmens o shocks due o changes in he exogenous variables. Esimaes of a dynamic hreshold error correcion model wih wo lags have clearly shown ha coffee supply adjuss o shocks in coffee prices above he hreshold bu no o shocks below he hreshold. Since he model is only lagged for one year, his resul means ha when coffee prices increase in year -1, farmers increase heir invesmen in mainaining he coffee rees o be harvesed and expored in year. Furhermore, coffee rees may be subjeced o several shocks such as weaher or pess over ime, such ha a longer lag operaor can give misleading resuls, if such variables are no properly conrolled for. Subsanively, fuure research should consider a possibiliy of more han one hreshold in he error correcion erm such ha muliple regime analysis which be considered. Overall, he heory ha supply adjuss o price incenives may no apply o expor commodiies where he farmers receive heir incomes in a foreign currency. In ha case he moivaion o produce more highly depends on he exchange rae in relaion o locally consumed goods and wages. In ha case policies should focus on increasing non-price incenives and creaing an environmen for deermines real exchange raes. 131

147 References Abdulai, A. and Rieder (1995). Impacs of Agriculural Price Policy on Cocoa Supply in Ghana: an Error Correcion Esimaion, Journal of African Economies, Vol. 4, no. 3, pp African developmen bank group (ADB) (2010). Coffee Producion in Africa and he Global Marke Siuaion. Commodiy Marke Brief- 19/07/2010. Askari, H. and Cummings, J., (1977). Agriculural Supply Response: A survey of Economeric Evidence. Praeger, New York, 443 p. Chisala, V., Geda. A., Dagdeviren, H., McKinley, T., Saad-Filho, A., Oya, C., and Weeks, J., (2006). Economic Policies for Growh, Employmen and Povery Reducion. Case Sudy of Zambia. Publishe by he UNDP. Enders, W (2004). Applied Economeric Time series. Second ediion. Willey series in Probabiliy and Saisics. Book published by John Wiley and sons. USA. Engle, R. and Granger C.W.J. (1987). Coinegraion and Error Correcion: Represenaion, Esimaion and Tesing. Economerica 5: FARA (Forum for Agriculural Research in Africa) (2006). Framework for African Agriculural Produciviy, ACRA, Ghana. Johansen, S. (1988), Saisical Analysis of Coinegraion Vecors, Journal of Economic Dynamics and Conrol, Vol. 12, No. 2 3, pp Johansen, S. and K. Juselius (1990). Maximum Likelihood Esimaion and Inference on Coinegraion- wih Applicaions o he Demand for Money. Oxford Bullein of Economics and Saisics pp

148 Kidane, A. (1999). Real exchange rae price and agriculural supply response in Ehiopia: The case of perennial crops. AER Research Paper 99, African Economic Research Consorium, Nairobi. Lee, J. and M.C Srazicich (2003). Minimum LM Uni Roo Tess wih Two Srucural Breaks, Review of Economics and Saisics, 63, pp Lopez, R.A. (1986). The use of composie Price Expecaions in Supply response Models, Canadian Journal of Agriculural Economics, 34 pp Mckay, A., O. Morrisey and C. Vailla (1999). Aggregae Supply Response in Tanzanian agriculure. The Journal of Inernaional Trade and Economic Developmen 8(1), Meerman, J. P (1997). Reforming agriculural: The World Bank goes o Marke. The World Bank, Washingon D.C. Molua, E. L. (2010). Rice Producion response o Trade Liberalizaion in Cameroon. Research Journal of Agriculure and Biological sciences, Vol. 6, no. 2 pp Muh, J.F. (1961). Raional Expecaions and he Theory of Price movemens. Economerica, vol.29; no. 3, pp Muchapondwa, E. (2009). Supply response of Zimbabwean agriculure: Afjare Vol 3 no 1, pp Nerlove, M. (1958). The Dynamic of supply: Esimaion of farmers Response o Price. Johns Hopkins, Balimore. Nyairo, N. (2009). Facors affecing Supply of Agriculural Producs. Discussion Paper no. 36. Universiy of Helsinki, Finland. 133

149 Oim, S and P.K Ngaegize (1993). Uganda Coffee Supply Response and Expor Demand: An Economeric Analysis. African Crop Science Journal, Vol. 1, No. 2, pp Rahji, M. A.Y and M. O Adewumi, (2008). Marke Supply Response and Demand for Local Rice in Nigeria: Implicaions for Self-Sufficiency Policy. Journal of Cenral European Agriculure, Vol. 9, No. 3 pp Rao, J.M. (1989). Agriculural supply response: A survey, Agriculural Economics, no. 3 pp Shirvani H. and B. Wilbrae (1999). The asymmeric Response of Domesic Prices o Changes in Impor prices; a Coinegraion Tes of he Rache Effec. Journal of Macroeconomics, spring, 1999, Vol. 2, No. 2, pp Thiele, R. (2000). Esimaing aggregae agriculure supply response; a survey of echniques and resuls for developing counries. Kiel Working Paper No. 1016, Kiel Insiue of World Economics, Germany. Wichern, R., Hausner, U., Chiwele, D.K.(1999). Impedimens o Agriculure Growh in Zambia. Trade and Marke division of he Inernaional Food Policy Research Insiue, Washingon, D.C. TMD Discussion Papers No. 47. Wold, K.B (1997). Supply Response in a Gender-Perspecive: The case of Srucural Adjusmen in Zambia. Repors 97/23 Saisics, Norway. World Bank Group (2001). Supply Response o adjusmen in Low Income Counries, Lessons from Zambia, Independen Evaluaion Group Publicaions. Xu, Z., J. Govereh, T. Black, and S. Jayne (2006). Maize Yield Response o ferilizer and Profiabiliy of Ferilizer Use among Small Producers in Zambia. Paper Presened a he Inernaional Associaion of Agriculural Economiss Conference, Gold Coas, Ausralia. 134

150 Chaper 6: Conclusion This sudy examined coffee value chains, price ransmission, price volailiy and supply response in he presence of srucural breaks arising from economic reforms. The main economic challenges facing coffee markes are asymmeric power srucures arising from high firm concenraion a he rading and roasing sages of he coffee value chain which preven efficien price ransmission from world prices hereby prevening he produces from benefiing from increases in world producer prices. While on he oher hand, price decreases in world prices are quickly passed on o he producers. Consequenly, coffee producers in he Zambia and Tanzania, like in many oher producing counries, have experienced declining producer prices and exensive shor-erm producer price volailiy in he las decade. These challenges affec producion, incomes and consequenly he welfare of he producers who are mosly poor small scare farmers fragmened across developing counries. The coffee value chain, like many oher agriculural commodiies, is buyer driven. This implies ha producer prices are deermined by forces a higher sages of he chain. This Chaper firs gives an overview of he sudy by summarising he major findings. The conclusions are drawn based on he findings and he hypoheses of he sudy. The quesion of wha he resuls mean for policies relaing o he coffee secors and he welfare of he growers is hen addressed. 6.1 Sudy Focus The analysis of he resuls repored in his secion aimed a invesigaing he impacs of rade policy reforms on coffee price movemens i.e. he sabiliy and asymmery of price ransmission from inernaional producer prices o grower s prices in he producing counry. The main moivaion for sudying coffee price movemens is ha while coffee remains he mos imporan expor crop in Easern Africa, is prices declined he mos in comparison o oher expor commodiies in he las decade. Furher, coffee producer prices have remained he mos volaile compare oher agriculural commodiies. Undoubedly, undersanding he explanaions behind he 135

151 price rends, oher han basic marke forces of supply and demand, which could have an effec on supply, becomes vial in price policy formulaion. In order o achieve he objecives of he sudy, four specific sudies have been done: i) value chain and governance srucures and implicaions on producer prices, ii) asymmeries in Price ransmission and he effecs of economic reforms iii) price volailiy analysis and iv) supply response o price movemens. 6.2 Summery of Resuls A review of he inernaional coffee value chains and governance srucures in coffee markes gives an indicaion ha he coffee value chain is composed of complex inerrelaions among he acors. Despie Zambia and Tanzania having differen degrees and rajecories of coffee marke liberalisaion, here is somehing in commonness in ha boh counries feed ino a complex value chain ha is governed by mulinaional corporaions. In he case of Tanzania, where coffee is he main expor crop is mainly produced by small scale farmers, he chain is even more complex as coffee has o pass hrough several of inermediaries ha include cooperaives and raders. Despie he liberalisaion of coffee markeing, by law, all coffee expors in Tanzania go hrough he governmen owned coffee aucion. Due o high governmen inervenion in markeing, large number of inermediaries as well as qualiy issues, Tanzania s grower prices are far below he world producer indicaor prices. In conras, Zambia s coffee from he farmers only go hrough he privae Zambia coffee growers associaion before i is expored. As a resul Zambia producer prices much higher han mos counries in Africa albei being very volaile. This however is a real reflecion of world prices. To subsaniae he findings of he qualiaive analysis of coffee markes, an analysis of coffee price ransmission from inernaional markes o producers has been carried ou. The sudy uses 273 monhly observaions of coffee producer prices for Zambia, Tanzania and he world producer indicaor price. The analysis has been done for he enire sample as well as for he wo subsamples using hreshold auo regression and he momenum hreshold auo regression models. For each of he wo producer prices, coinegraed wih world prices is examined specifically, wheher he ransmission is symmeric or asymmeric. If he es for long-run relaion (implicaion 136

152 of coinegraion) is significan, asymmery is esed o esablish he naure of he ransmission. The resuls from he hreshold auoregression model indicae ha price ransmission for Zambia improved afer economic liberalisaion while in he case of Tanzania, he ransmission has no improved. The TAR model specifically examines deepness of price adjusmens o long-run equilibrium, while he MTAR examines seepness of he adjusmen. Comparing TAR and MTAR models, he resuls show evidence of high rejecion power of he null hypohesis of symmery for he MTAR model. In he second sage of he analysis, shor run adjusmens o long-run equilibrium are examined using hreshold error correcion models. Resuls show ha, afer he economic reforms in Zambia, price decreases have a larger impac on producer prices han he price increases. For example, a negaive shock on Zambian prices led o he prices adjusing by 58.3 percen while a posiive shock only led o 43 percen adjusmen in he Zambian prices. In mos cases he adjusmen for CIP in response for shocks iniiaed by changes in producer prices was no significan. For Tanzania, negaive shocks lead o 30 percen price adjusmen owards he long-run equilibrium while posiive shocks only iniiae 9 percen adjusmen. Price volailiy has been analysed o examine coffee price variaions in Zambia, Tanzania, he inernaional marke price and reail prices in Germany. The aim is o assess he effec of he fall of he ICA in 1989 and marke liberalisaion policies in he wo producing counries. Reail prices demonsraed homoskedasiciy such ha GARCH models could no be esimaed. I should be noed ha he idea of GARCH models is o esablish impac and persisence of shocks in variables wih heeroskedasiciy. According o hreshold GARCH resuls, he impac of shocks was highes in world prices where a shock o he prices increased volailiy by 35.2 percen. On he oher hand, a shock o he prices increased volailiy by 3.5 percen in Zambia while i acually reduced volailiy by 0.8 percen in Tanzania s case. However shocks were mos persisen in he case of Zambia such ha i ook up 59.3monhs before dying away while i only ook 32.7 monhs before i is eliminaed in he world prices and Tanzania prices respecively. Resuls also show ha volailiy was asymmeric for Zambia and World prices while symmeric in Tanzania s case. This means ha in Tanzania s case posiive and negaive shocks had similar effecs 137

153 on volailiy. Furher, beween TGARCH and GARCH models, he analysis finds TGARCH models o provide beer fi for he resuls. The las par of he sudy examined supply response of coffee in Zambia. The moivaion for his sudy is ha coffee, despie is poenial for increasing Zambia s expor diversificaion, is sill a very small secor compared o oher counries in easern and souhern Africa. Undersanding he facors ha influences coffee supply would render policy makers develop some pahways for increasing supply. Applying coinegraion and hreshold error correcion mehods, resuls show ha all he independen variables i.e. coffee prices, maize prices, exchange raes and dummy variables apar from economic reforms have significan impac on coffee supply. As expeced he price of maize he main compeing crop and real exchange raes have negaive effecs on supply. This means ha when maize prices increase farmers resen coffee and op for maize. Coffee expored also become discouraged when he local currency gains agains he dollar as exporing becomes more expensive. 6.3 Policy Implicaions The resuls of he sudy have significan policy implicaions. Firs, he effec of value chain governance srucures on producer prices is eviden alhough i has been less invesigaed in economic lieraure. Producer price share of he final price is much of a consequence of value chain governance srucures ha producers find hemselves in. Clearly, he rules for paricipaion in he chain, which are se by he governors of he chain and he barriers o enry, deermine he exen o which producers can paricipae, subsequenly, he price ha hey receive. In addiion, comparing he wo value chain srucures, i.e. he Zambian and he Tanzanian value chains, he sudy shows ha he more he inermediaries in a given chain, he less price growers are likely o obain. I is herefore imporan ha governmen policies be direced owards having fewer bu effecive players a inra-counry level. The cooperaive unions arrangemen in he case of Tanzania, if well organised like in Zambia s case can help Tanzania s coffee growers receive higher prices han wha hey are currenly receiving. The cooperaives should however operae as privae eniies free from sae inervenions like in Zambia s case. 138

154 Second, his sudy has confirmed asserions ha commodiy price movemens are fundamenally asymmeric whereby posiive and negaive shocks o prices have differen impacs on long-run adjusmen o equilibrium as well as on volailiy. As price decreases end o have large impacs and persiss longer in producer prices, any rade inervenion policy should be direced owards working on modaliies ha help eliminae negaive shocks quicker. Marke-based price risk inervenions like fuures become ideal for prevening negaive price shocks wihou disoring marke funcioning. Fair rade iniiaives may also provide soluions o negaive price shocks alhough empirical evidence of fair rade impac and exen is largely missing. Furher, i has been esablished from he TAR model ha, price ransmission in he case of Zambia improved afer economic reforms while i was no he case for Tanzania. These resuls confirm earlier findings ha Tanzania was he only coffee producing counry where price ransmission did no improve afer economic liberaions. From he resuls, i can be concluded ha economic liberalisaion lead o improved price ransmission. There is also srong evidence of negaive shocks increasing he persisence of volailiy more han posiive shocks. Zambia which is more exposed o world markes experiences higher volailiy han Tanzania and even more han he world prices. As negaive shocks lead o more volaile prices han posiive shocks, policies should be direced owards minimising he negaive shocks if sable prices are o be achieved. Again, marke-based iniiaives are recommended. Invesigaing supply responsiveness of coffee becomes criical for Zambia where producion significanly declined in he las 5 years. An imporan finding from his sudy is ha, conrary o heoreical asserions ha commodiy producion responds posiively o prices, i is no he case wih coffee in Zambia. Since coffee is mainly produced for expor and he farmers receive heir revenues in US Dallas, he exchange rae ends o play a key role in he farmers decisions o produce more coffee. Therefore, he sudy has shown ha farmers have no aken advanage of price increases o increase producion. For policy decisions, i is imporan o undersand ha non-price incenives play a key role in improving producion. 139

155 APPENDICES Appendix A: Srucural Break Uni Roo Tes Theory In Chaper 3, 4 and 5, srucural break uni roo ess have been employed o deermine he srucural break due o economic reforms in he daa. This secion discusses an endogenous uni roo ess. Lee and Srazicich s (2004) developed a Langrage Muliplier (LM) esing sraegy (LSLM) ha allowed for wo srucural breaks o be deermined endogenously under boh he null and he alernaive hypohesis. The LSLM is based on Perron (1989) s srucural break uni roo ess ha follows hree models: Model A allows for a break in he inercep, model B in he rend and model C in boh he rend and inercep. The LSLM model is basically an exension of he Perron (1989) model, which also allows for a srucural break under boh he null and he alernaive hypohesis. Considering daa-generaing process, he model is presened as; y = δ ' Z + e e β + ε = e 1 Where Z is a vecor of exogenous variables and 2 ε is an iid N (0, σ ). The wo srucural breaks under Model C28 are described by, [ 1,, D DT ] = for T + 1 Z, 1 1 B and zero oherwise, for DT j = TB j=1, 2 and zero oherwise. j D is he indicaor dummy variables for a mean shif occurring a imes TB j. TD j are he corresponding rend shif variables. The DGP includes breaks under he null ( β = 1). According o he LM (score) principle, uni roo es saisics are obained from y = δ + u 10 ' Z + φs ~ 1 28 Lee and Srazicich s (2003) model, like in he Perron C 28 model, allow for changes in he level and rend. 140

156 where ~ ~ S = y ~ ψ Zδ = 2,... T, ~ δ are coefficiens in he regression of y on Z ; ψ x is given by y ~ 1 Z 1 δ ; The Uni roo null hypohesis is described byφ = 0. Noe ha he esing regression (9) involves ~ correlaion errors, he augmened erms S, Z insead of Z. To correc for serial j j = 1... k are included o he equaion using he general specific lag selecion mehod suggesed by Perron (1998). The mehod involves selecing a number of lags P from a more general srucure lengh such ha he coefficien of he las lag is significan, and ha he coefficien in an auo-regression of order more han P is insignifican up o an opimal number of lag lenghs- where he error erm is saionary. Selecion of lag lengh is criical as oo few lags lead o auo correlaion and oo many lags will lead o inefficiency. The LM ~ es saisic is given by: S, -saisic for esing he uni roo null hypohesis ha j φ = 0. The locaion of he srucural break ( T B ) is deermined by selecing all possible break poins for he minimum -saisic as follows: ~ Inf ~ τ ( λ ) Inf ~ τ ( λ i = ), where λ = T B / T Lee and Srazicich (2004) argue ha he LM uni roo ess saisics which is esimaed by he regression according o he LM principle will no spuriously rejec he null hypohesis of uni roo. Tha i is also invarian o nuisance parameer, as such i does no require he assumpion of no break under he null. The advanage of he model is ha he parameers do no change regardless of wheher he series is saionary. 141

157 Appendix B: Threshold Vecor Error Correcion Resuls In Chaper 3, an analysis of shor-run adjusmens o long-run equilibrium has been carried ou. The resuls presened in Table 3.4 and 3.5, are a summery of vecor error correcion esimaion. Figure B1 and B2 presens he vecor error correcion esimaions explaining how he Zambian coffee prices and Tanzanian coffee prices adjus o world prices and vice versa. The error erm has been esimaed from TAR and MTAR models presened in Table 3.3. The Table B1 shows he resuls for he Zambia-world price error correcion esimaion. Up o 12 lags for boh variables have been included in he model. A general- o-specific lag selecion procedure was used o selec he 12 lags. The Durbin Wason (DW) value of confirms absence of auocorrelaion in he error erm for he 12 lags. The resuls of he error correcion esimaion from he TAR model indicae ha world coffee prices (denoed as Zambian coffee prices Pz Pcip ) are no influenced by he given he insignifican -saisics for mos of he lags in column 2 of Table B1. On he oher hand, he Zambian prices are influenced by world prices for mos of he lags. The Zambian prices also seem o be influenced by he prices in he previous periods. Similar resuls are observed from he MTAR model such ha when he -saisics for he coefficiens in he la wo columns of Table 3.5 are compared, he coefficiens for Zambia are more significan for almos all he lags. Asymmeric shor-run adjusmens have been observed where he Zambian prices adjus by 12.4 percen whenever he deviaion is above he equilibrium while adjusing by 14.8 percen he deviaion from long-run equilibrium ges below he hreshold. This means ha he negaive shocks have more effec on he Zambian prices in he shor run. For Tanzania, he coffee producer prices do no seem o be significanly influenced by he world prices in he shor-run. This is confirmed he insignifican coefficiens on he lags of world prices when regressed on Tanzania in he TAR model (Column 3 of Table B2). However he esimaions from MTAR model show significan influence of world prices on Tanzania. The adjusmens are also asymmeric where 22.7 percen of he deviaion from long-run equilibrium is correced back in he shor run. 142

158 Table B1: Asymmeric ECM Resuls for Zambia and World Prices (Full Sample) TAR MTAR P cip Pz Pcip PZ Consan (0.151) (1.396) (0.655) (1.400) P Z (0.521) (4.06) (1.085) (3.33) P Z (0.338) (1.566) (0.764) (1.074) P Z (0.332) (1.134) 0.005(0.251) (1.231) P Z (0.950) (1.99) 0.020(1.007) (1.949) P Z (0.611) (1.388) 0.012(0.593) (1.419) P Z (1.203) (1.960) 0.026(1.322) (1.866) P Z (1.741) (2.309) 0.037(1.878) (2.194) P Z (2.078) (0.331) 0.041(2.100) (0.321) P Z (1.600) (2.936) 0.034(1.731) (2.827) P 0.044( (1.100) 0.042(2.158) (1.169) Z 10 P Z 11 P z (-0.694) 0.044(2.338) 0.000(0.014) 0.179(2.777) (0.55) 0.046(2.457) 0.013(0.191) 0.186(2.918) P 0.212(3.207) 0.411(1.816) 0.200(3.055) 0.360(1.609) cip 1 P 0.007(0.107) (0.343) 0.001(0.019) (0.451) cip 2 P 0.078(1.194) 0.347(1.546) 0.070(1.081) 0.311(1.404) cip 3 P cip (1.452) 0.081(0.361) (1.442) 0.085(0.382) 4 P (1.519) (1.653) (1.388) (1.504) cip 5 P 0.005(0.082) 0.079(0.353) 0.007(0.115) 0.090(0.407) cip 6 P cip 0.009(0.135) (2.309) 0.014(0.220) 0.024(0.108) 7 P (1.024) (0.331) (1.365) 0.274(1.230) cip 8 P (0.209) (2.936) 0.009(0.139) 0.226(1.013) cip 9 P cip (0.269) (1.100) 0.004(0.059) 0.038(0.170) 10 P cip (1.218) 0.000(0.0147) 0.099(1.514) 0.289(1.295) P cip (0.263) (2.777) (0.441) (0.512) z-plus (0.684) (2.290) (0.171) (0.918) z-minus (0.649) (3.027) (2.410) (0.775) Q(26) 1.693(0.0226) 3.518(0.000) 1.913(0.006) 3.823(0.000) DW

159 Table B2: Asymmeric ECM Resuls for Tanzanian and World Prices (Full Sample) TAR P cip MTAR Pz Pcip Pz Consan (-2.156) (-0.011) (-0.941) 0.000(0.050) P TZ (-0.435) (0.843) (-0.577) 0.021(0.299) P TZ (2.703) (0.197) 0.113(2.222) 0.047(0.670) P TZ (-0.405) (0.338) (-1.045) (-0.782) P TZ (0.290) (-0.141) (-0.525) (-0.422) P TZ (0.325) 0.073(0.790) 0.012(0.260) (0.201) P TZ (2.151) 0.010(0.117) 0.095(2.169) (-0.297) P TZ (-0.440) 0.049(0.561) (-0.763) (-0.230) P TZ (-0.56) 0.047(0.541) (0.903) 0.002(0.034) P TZ (-0.121) 0.072(0.880) (-0.450) 0.017(0.259) P TZ (0.669) (-0.225) (-0.539) (-0.018) P cip (2.513) (0.063) 0.229(3.124) 0.078(0.707) P cip (-0.545) 0.226(1.556) (-1.759) 0.185(1.652) P cip (0.287) (-1.347) 0.068(0.921) (-0.875) P cip (-2.104) (-0.566) (-2.143) (-0.753) P cip (-0.635) 0.184(1.291) (-1.095) 0.219(1.953) P cip (-0.681) 0.022(0.155) (-0.319) 0.067(0.609) P cip (-0.283) (-0.687) (-0.378) (-0.538) P cip (-1.335) (-0.542) (-0.750) (-0.401) P cip (1.288) (-1.118) 0.028(0.394) (-0.807) P cip 0.065(-0.739) 0.167(1.230) 0.030(0.425) 0.183(1.732) 10 z-plus (-0.382) (-1.770) 0.033(0.880) (-2.592) z-minus (-4.105) (-3.436) (-2.568) (-2.577) Q(22,222) 2.533(0.001) 1.979(0.009) 2.410(0.001) DW In parenheses are -saisics values. P cip and Pz denoes World marke and Tanzanian coffee prices in heir firs differences. Pcip i and Pz i P cip i denoes firs differences of lagged values of he prices. and prices a ime and world prices (he composie indicaor price) a ime respecively. 144

160 Table B3: Symmeric Vecor Error Correcion for Supply Response below he Threshold Error Correcion: D(LOGCPRO) D(LOGCPRI) D(LOGRER) D(LOGMPRICE) D(RF) CoinEq ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] D(LOGCPRO(-1)) ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] D(LOGCPRO(-2)) ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] D(LOGCPRI(-1)) ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] D(LOGCPRI(-2)) ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] D(LOGRER(-1)) ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] D(LOGRER(-2)) ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] D(LOGMPRICE(-1)) ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] D(LOGMPRICE(-2)) ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] D(RF) ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] C ( ) ( ) ( ) ( ) ( ) [ ] [ ] [ ] [ ] [ ] R-squared Sandard errors in () and -saisics in [] 145

161 Appendix C: Coffee Processing Sages Producing Counry $0.70/ Ib Consuming Counry Markeed on-line as Newyork Zambia Coffee $11.95 (1 lb) Roaser Source: Auhors own presenaion. The green coffee price of $0.70/Ib is Zambia s average grower price from January 1986 o Sepember The price for New York Zambia Ground Coffee was obained from hp:// (downloaded ). A 5lb packe is selling a $

POLICY RELEVANCE SUMMARY

POLICY RELEVANCE SUMMARY POLICY RELEVANCE SUMMARY Ensuring Food Securiy in Ghana The Role of Maize Sorage Sysems. Paul W. Armah and Felix Asane 1 Research Findings and Policy Relevance-Summary 1. Research Framework and Objecives

More information

Analysis of Egyptian Grapes Market Shares in the World Markets

Analysis of Egyptian Grapes Market Shares in the World Markets American-Eurasian J. Agric. & Environ. Sci., 3 (4): 656-66, 008 ISSN 1818-6769 IDOSI Publicaions, 008 Analysis of Egypian Grapes Marke Shares in he World Markes Hamdi A. El- Sawalhy, Mohamed G.M. Abou

More information

INSTITUTIONAL INVESTOR SENTIMENT

INSTITUTIONAL INVESTOR SENTIMENT 9. INSTITUTIONAL INVESTOR SENTIMENT AND MARET RETURNS: EVIDENCE FROM THE TAIWAN FUTURES MARET Absrac Ralph Yang-Cheng LU 2 Hsiu-Chuan LEE 3 Peer CHIU 4 This sudy explores he dynamic relaionship beween

More information

Unravelling the underlying causes of price volatility in world coffee and cocoa commodity markets

Unravelling the underlying causes of price volatility in world coffee and cocoa commodity markets MPRA Munich Personal RePEc Archive Unravelling he underlying causes of price volailiy in world coffee and cocoa commodiy markes Noemie Maurice and Junior Davis UNCTAD 011 Online a hps://mpra.ub.uni-muenchen.de/43813/

More information

AN ECONOMIC EVALUATION OF THE HASS AVOCADO PROMOTION ORDER S FIRST FIVE YEARS

AN ECONOMIC EVALUATION OF THE HASS AVOCADO PROMOTION ORDER S FIRST FIVE YEARS AN ECONOMIC EVALUATION OF THE HASS AVOCADO PROMOTION ORDER S FIRST FIVE YEARS A REPORT PREPARED FOR THE HASS AVOCADO BOARD BY Hoy F. Carman Lan Li Richard J. Sexon 1 March 30, 2009 1 Hoy F. Carman is Professor

More information

Economic Computation and Economic Cybernetics Studies and Research, Issue 3/2016, Vol. 50

Economic Computation and Economic Cybernetics Studies and Research, Issue 3/2016, Vol. 50 Economic Compuaion and Economic Cyberneics Sudies and Research, Issue 3/2016, Vol. 50 Assisan Professor Murad A. BEIN, PhD E-mail: mbein@ciu.edu.r Deparmen of Accouning and Finance Faculy of Economics

More information

The Design of a Forecasting Support Models on Demand of Durian for Export Markets by Time Series and ANNs

The Design of a Forecasting Support Models on Demand of Durian for Export Markets by Time Series and ANNs AIJSTPME (20) 4(2): 49-65 The Design of a Forecasing Suppor Models on Demand of Durian for Expor Mares by Time Series and ANNs Udomsri N. Deparmen of Indusrial Engineering, Faculy of Engineering, King

More information

Supply and Demand Model for the Malaysian Cocoa Market

Supply and Demand Model for the Malaysian Cocoa Market MPRA Munich Personal RePEc Archive Supply and Demand Model for he Malaysian Cocoa Marke Amna Awad Abdel Hameed and Akram Hasanov and Nurjihan Idris and Amin Mahir Abdullah and Faimah Mohamed Arshad and

More information

PRODUCTION PERFORMANCE OF MAIZE IN INDIA : APPROACHING AN INFLECTION POINT

PRODUCTION PERFORMANCE OF MAIZE IN INDIA : APPROACHING AN INFLECTION POINT In. J. Agricul. Sa. Sci., Vol. 10, No. 1, pp. 241-248, 2014 ISSN : 0973-1903 ORIGINAL ARTICLE PRODUCTION PERORMANCE O MAIZE IN INDIA : APPROACHING AN INLECTION POINT Ranji Kumar*, K. Srinivas, Naveen Kumar

More information

Prices of Raw Materials, Budgetary Earnings and Economic Growth: A Case Study of Côte d Ivoire

Prices of Raw Materials, Budgetary Earnings and Economic Growth: A Case Study of Côte d Ivoire Prices of Raw Maerials, Budgeary Earnings and Economic Growh: A Case Sudy of Côe d Ivoire Nguiakam Sandrine 1 and Kabore Augusin 1 Ciaion: CTA and FARA. 2011. Agriculural Innovaions for Susainable Developmen.

More information

Sustainability of external imbalances in the OECD countries *

Sustainability of external imbalances in the OECD countries * Susainabiliy of exernal imbalances in he OECD counries * Oscar Bajo-Rubio (Universidad de Casilla-La Mancha) Carmen Díaz-Roldán (Universidad de Casilla-La Mancha) Vicene Eseve (Universidad de Valencia

More information

Inter-regional Transportation and Economic Development: A Case Study of Regional Agglomeration Economies in Japan

Inter-regional Transportation and Economic Development: A Case Study of Regional Agglomeration Economies in Japan Iner-regional Transporaion and Economic Developmen: A Case Sudy of Regional Agglomeraion Economies in Japan Jepan Wewioo a and Hironori Kao b a Deparmen of Civil Engineering, The Universiy of Tokyo 7-3-1,

More information

International Trade and Finance Association THE EFFECT OF EXCHANGE RATE CHANGES ON TRADE BALANCES IN NORTH AFRICA: EVIDENCE

International Trade and Finance Association THE EFFECT OF EXCHANGE RATE CHANGES ON TRADE BALANCES IN NORTH AFRICA: EVIDENCE Inernaional Trade and Finance Associaion Inernaional Trade and Finance Associaion 15h Inernaional Conference Year 2005 Paper 46 THE EFFECT OF EXCHANGE RATE CHANGES ON TRADE BALANCES IN NORTH AFRICA: EVIDENCE

More information

Transmission of prices and price volatility in Australian electricity spot markets: A multivariate GARCH analysis

Transmission of prices and price volatility in Australian electricity spot markets: A multivariate GARCH analysis Transmission of prices and price volailiy in Ausralian elecriciy spo markes: A mulivariae GARCH analysis Auhor Worhingon, Andrew, Kay-Spraley, Adam, Higgs, Helen Published 2005 Journal Tile Energy Economics

More information

The Long-Run Volatility Puzzle of the Real Exchange Rate. Ricardo Hausmann Kennedy School of Government Harvard University

The Long-Run Volatility Puzzle of the Real Exchange Rate. Ricardo Hausmann Kennedy School of Government Harvard University The Long-Run Volailiy Puzzle of he Real Exchange Rae Ricardo Hausmann Kennedy School of Governmen Harvard Universiy Ugo Panizza Research Deparmen Iner-American Developmen Bank Robero Rigobon * Sloan School

More information

GROWTH AND CONVERGENCE IN THE SPACE ECONOMY : EVIDENCE FROM THE UNITED STATES

GROWTH AND CONVERGENCE IN THE SPACE ECONOMY : EVIDENCE FROM THE UNITED STATES GROWTH AND CONVERGENCE IN THE SPACE ECONOMY : EVIDENCE FROM THE UNITED STATES John I. CARRUTHERS *, Michael K. HOLLAR **, Gordon F. MULLIGAN *** Absrac - This paper invesigaes geographic relaionships in

More information

Testing for the Random Walk Hypothesis and Structural Breaks in International Stock Prices

Testing for the Random Walk Hypothesis and Structural Breaks in International Stock Prices Universiy of Wollongong Research Online Faculy of Business - Economics Working Papers Faculy of Business 200 Tesing for he Random Walk Hypohesis and Srucural Breaks in Inernaional Sock Prices S. Chanchara

More information

Título artículo / Títol article: Re-examining the risk-return relationship in Europe: Linear or non-linear trade-off?

Título artículo / Títol article: Re-examining the risk-return relationship in Europe: Linear or non-linear trade-off? Tíulo arículo / Tíol aricle: Re-examining he risk-reurn relaionship in Europe: Linear or non-linear rade-off? Auores / Auors Enrique Salvador Aragó, Chrisos Floros, Vicen Aragó Manzana Revisa: Journal

More information

NBER WORKING PAPER SERIES A SIMPLE TEST OF THE EFFECT OF INTEREST RATE DEFENSE. Allan Drazen Stefan Hubrich

NBER WORKING PAPER SERIES A SIMPLE TEST OF THE EFFECT OF INTEREST RATE DEFENSE. Allan Drazen Stefan Hubrich NBER WORKING PAPER ERIE A IMPLE TET OF THE EFFECT OF INTERET RATE DEFENE Allan Drazen efan Hubrich Working Paper 12616 hp://www.nber.org/papers/w12616 NATIONAL BUREAU OF ECONOMIC REEARCH 1050 Massachuses

More information

DOCUMENTOS DE ECONOMÍA Y FINANZAS INTERNACIONALES. Working Papers on International Economics and Finance

DOCUMENTOS DE ECONOMÍA Y FINANZAS INTERNACIONALES. Working Papers on International Economics and Finance DOCUMENTOS DE ECONOMÍA Y FINANZAS INTERNACIONALES Working Papers on Inernaional Economics and Finance DEFI 11-07 Ocober 2011 Susainabiliy of exernal imbalances in he OECD counries Oscar Bajo-Rubio Carmen

More information

The Influence of Earnings Quality and Liquidity on the Cost of Equity

The Influence of Earnings Quality and Liquidity on the Cost of Equity Inernaional Business Research; Vol. 8, No. 4; 2015 ISSN 1913-9004 E-ISSN 1913-9012 Published by Canadian Cener of Science and Educaion The Influence of Earnings Qualiy and Liquidiy on he Cos of Equiy Ming-Feng

More information

The Spillover Effects of U.S. and Japanese Public Information News in. Advanced Asia-Pacific Stock Markets

The Spillover Effects of U.S. and Japanese Public Information News in. Advanced Asia-Pacific Stock Markets The Spillover Effecs of U.S. and Japanese Public Informaion News in Advanced Asia-Pacific Sock Markes Suk-Joong Kim Absrac School of Banking and Finance The Universiy of New Souh Wales UNSW SYDNEY 2052

More information

Application of Peleg Model to Study Water Absorption in Bean and Chickpea During Soaking

Application of Peleg Model to Study Water Absorption in Bean and Chickpea During Soaking Applicaion of Peleg Model o Sudy Waer Absorpion in Bean and Chickpea During Soaking A. A. Masoumi * Assisan Professor, Deparmen of Farm Machinery. Isfahan Universiy of Technology, IUT Isfahan, Iran 8456-83

More information

Accounting Fundamentals and Variations of Stock Price: Forward Looking Information Inducement

Accounting Fundamentals and Variations of Stock Price: Forward Looking Information Inducement Accouning Fundamenals and Variaions of Sock Price: Forward Looking Informaion Inducemen Sumiyana Gadjah Mada Universiy Absrac This sudy invesigaes a permanen issue abou low associaion beween accouning

More information

A Macro Assessment of China Effects on Malaysian Exports and Trade Balances

A Macro Assessment of China Effects on Malaysian Exports and Trade Balances MPRA Munich Personal RePEc Archive A Macro Assessmen of China Effecs on Malaysian Expors and Trade Balances Tze-Haw Chan and Hooi Hooi Lean and Chee Wooi Hooy Graduae School of Business, Universii Sains

More information

Price convergence in the European electricity market

Price convergence in the European electricity market Price convergence in he European elecriciy marke Dr. E. Dijkgraaf Prof.dr. M.C.W. Janssen Erasmus Compeiion and Regulaion insiue Erasmus Universiy Roerdam Augus 7 7 Conac: Elber Dijkgraaf SEOR-ECRi Erasmus

More information

Volume-Return Relationship in ETF Markets: A Reexamination of the Costly Short-Sale Hypothesis

Volume-Return Relationship in ETF Markets: A Reexamination of the Costly Short-Sale Hypothesis Journal of Applied Finance & Banking, vol., no. 6,, -4 ISSN: 79-658 (prin version), 79-6599 (online) Scienpress Ld, Volume-Reurn Relaionship in ETF Markes: A Reexaminaion of he Cosly Shor-Sale Hypohesis

More information

Duration models. Jean-Marie Le Goff Pavie-Unil

Duration models. Jean-Marie Le Goff Pavie-Unil Duraion models Jean-Marie Le Goff Pavie-Unil Oher erms for duraion models Duraion models: economery Survival analysis : medical sciences, demography Even hisory analysis, ransiion analysis : social sciences

More information

Applicability of Investment and Profitability Effects in Asset Pricing Models

Applicability of Investment and Profitability Effects in Asset Pricing Models Disponível em hp://www.anpad.org.br/rac RAC, Rio de Janeiro, v. 21, n. 6, ar. 6, pp. 851-874, Novembro/Dezembro, 2017 hp://dx.doi.org/10.1590/1982-7849rac2017170027 Applicabiliy of Invesmen and Profiabiliy

More information

Market Overreaction and Under-reaction for Currency Futures Prices. January 2008

Market Overreaction and Under-reaction for Currency Futures Prices. January 2008 Marke Overreacion and Under-reacion for Currency Fuures Prices Sephen J. Larson and Sephen E. Wilcox* Minnesoa Sae Universiy, Mankao January 2008 Sephen J. Larson, Ph.D., CFP Minnesoa Sae Universiy, Mankao

More information

On the relationship between inventory and financial performance in manufacturing companies Vedran Capkun HEC Paris, Paris, France

On the relationship between inventory and financial performance in manufacturing companies Vedran Capkun HEC Paris, Paris, France The curren issue and full ex archive of his journal is available a www.emeraldinsigh.com/0144-3577.hm On he relaionship beween invenory and financial performance in manufacuring companies Vedran Capkun

More information

What Determines the Future Value of an Icon Wine? New Evidence from Australia. Danielle Wood

What Determines the Future Value of an Icon Wine? New Evidence from Australia. Danielle Wood Wha Deermines he Fuure Value of an Icon Wine? New Evidence from Ausralia Danielle Wood Produciviy Commission Melbourne dwood@pc.gov.au and Kym Anderson (corresponding auhor) School of Economics and Cenre

More information

Macro-Finance Determinants of the Long-Run Stock-Bond Correlation: The DCC-MIDAS Specification *

Macro-Finance Determinants of the Long-Run Stock-Bond Correlation: The DCC-MIDAS Specification * Macro-Finance Deerminans of he Long-Run Sock-Bond Correlaion: The DCC-MIDAS Specificaion * Hossein Asgharian, Lund Universiy + Charloe Chrisiansen, CREATES, Aarhus Universiy ++ and Ai Jun Hou, Sockholm

More information

Taking Advantage of Global Diversification: A Mutivariate-Garch Approach

Taking Advantage of Global Diversification: A Mutivariate-Garch Approach Taking Advanage of Global Diversificaion: A Muivariae-Garch Approach Elena Kaloychou *, Soiris K. Saikouras, Gang Zhao Cass Business School, Ciy Universiy London, 106 Bunhill Row, London EC1Y 8TZ Firs

More information

THE UNIVERSITY OF TEXAS AT SAN ANTONIO, COLLEGE OF BUSINESS Working Paper SERIES

THE UNIVERSITY OF TEXAS AT SAN ANTONIO, COLLEGE OF BUSINESS Working Paper SERIES THE UNIVERITY OF TEXA AT AN ANTONIO, COLLEGE OF BUINE Working Paper ERIE Dae May 15, 013 WP # 0046FIN-0-013 Commodiy Financializaion and Herd Behavior in Commodiy Fuures Markes Rıza Demirer Deparmen of

More information

Textos para Discussão PPGE/UFRGS

Textos para Discussão PPGE/UFRGS Texos para Discussão PPGE/UFRGS Programa de Pós-Graduação em Economia Universidade Federal do Rio Grande do Sul ENDOGENEITY AND NONLINEARITIES IN CENTRAL BANK OF BRAZIL S REACTION FUNCTIONS: AN INVERSE

More information

PRODUCTIVE EFFICIENCY OF PORTUGUESE VINEYARD REGIONS

PRODUCTIVE EFFICIENCY OF PORTUGUESE VINEYARD REGIONS MARTA-COSTA A., MARTINHO V., SANTOS M., Regional Science Inquiry, Vol. IX, (2), 2017, pp. 97-107 97 PRODUCTIVE EFFICIENCY OF PORTUGUESE VINEYARD REGIONS Ana MARTA-COSTA Corresponding auhor. Assisan Professor.

More information

Volatility and risk spillovers between oil, gold, and Islamic and conventional GCC banks

Volatility and risk spillovers between oil, gold, and Islamic and conventional GCC banks Volailiy and risk spillovers beween oil, gold, and Islamic and convenional GCC banks Walid Mensi a,b, Shawka Hammoudeh c,d, Idries Mohammad Wanas Al-Jarrah e, Khamis Hamed Al-Yahyaee b*, Sang Hoon Kang

More information

Paper for Annual Meeting 2015 Abstract. World Trade Flows in Photovoltaic Cells: A Gravity Approach Including Bilateral Tariff Rates * Abstract

Paper for Annual Meeting 2015 Abstract. World Trade Flows in Photovoltaic Cells: A Gravity Approach Including Bilateral Tariff Rates * Abstract Paper for Annual Meeing 2015 Absrac World Trade Flows in Phoovolaic Cells: A Graviy Approach Including Bilaeral Tariff Raes * Asuko Masumura (Tokyo Inernaional Universiy) Absrac This paper invesigaes he

More information

CO2 Emissions, Research and Technology Transfer in China

CO2 Emissions, Research and Technology Transfer in China MPRA Munich Personal RePEc Archive CO2 Emissions, Research and Technology Transfer in China Ang, James Monash Universiy 09. February 2009 Online a hp://mpra.ub.uni-muenchen.de/13261/ MPRA Paper No. 13261,

More information

Out-of-Sample Exchange Rate Forecasting and. Macroeconomic Fundamentals: The Case of Japan

Out-of-Sample Exchange Rate Forecasting and. Macroeconomic Fundamentals: The Case of Japan Ou-of-Sample Exchange Rae Forecasing and Macroeconomic Fundamenals: The Case of Japan Takashi Masuki and Ming-Jen Chang * ABSTRACT: The sudy explores he exchange rae forecasing abiliy of a number of macroeconomic

More information

Global financial crisis and spillover effects among the U.S. and BRICS stock markets

Global financial crisis and spillover effects among the U.S. and BRICS stock markets Acceped Manuscrip Global financial crisis and spillover effecs among he U.S. and BRICS sock markes Walid Mensi, Shawka Hammoudeh, Duc Khuong Nguyen, Sang Hoon Kang PII: S1059-0560(15)00227-0 DOI: doi:

More information

Essays on Board of Directors External Connections. Sehan Kim. B.A., Applied Statistics, Yonsei University, 2001

Essays on Board of Directors External Connections. Sehan Kim. B.A., Applied Statistics, Yonsei University, 2001 Essays on Board of Direcors Exernal Connecions by Sehan Kim B.A., Applied Saisics, Yonsei Universiy, 2001 M.S., Saisics, Purdue Universiy, 2008 Submied o he Graduae Faculy of The Joseph M. Kaz Graduae

More information

Investor Herds in the Taiwanese Stock Market

Investor Herds in the Taiwanese Stock Market Invesor Herds in he Taiwanese Sock Marke Rıza Demirer *, Chun-Da Chen ** and Ali M. Kuan *** * Souhern Illinois Universiy Edwardsville ** Tennessee Sae Universiy *** Souhern Illinois Universiy Edwardsville,

More information

The Interest Rate Sensitivity of Value and Growth Stocks - Evidence from Listed Real Estate

The Interest Rate Sensitivity of Value and Growth Stocks - Evidence from Listed Real Estate The Ineres Rae Sensiiviy of Value and Growh Socks - Evidence from Lised Real Esae This Version: January 25, 2017 Please reques he mos recen version from he auhors Chrisian Weis, Universiy of Regensburg

More information

This paper can be downloaded without charge from the Social Sciences Research Network Electronic Paper Collection:

This paper can be downloaded without charge from the Social Sciences Research Network Electronic Paper Collection: = = = = = = = Working Paper A Regime Shif Model of he Recen Housing Bubble in he Unied Saes Rober Van Order Sephen M. Ross School of Business a he Universiy of Michigan Rose Neng Lai Universiy of Macau

More information

TABIE l.~ Yields of Southern Peas In Relation to Seed Coat Color and Season. Pounds per Acre of "Whole-Pod F^asgT 19?5-196l#

TABIE l.~ Yields of Southern Peas In Relation to Seed Coat Color and Season. Pounds per Acre of Whole-Pod F^asgT 19?5-196l# HALSEY: SOUTHERN PEAS 233 SOUTHERN PEA VARIETIES, CULTURE AND HARVESTING AS RELATED TO PRODUCTION FOR HANDLING AND PROCESSING L. H. Halsey Florida Agriculural Experimen Saion Gainesville Much emphasis

More information

Monetary Policy Impacts on Cash Crop Coffee and Cocoa Using. Structural Vector Error Correction Model

Monetary Policy Impacts on Cash Crop Coffee and Cocoa Using. Structural Vector Error Correction Model Moneary Policy Imacs on Cash Cro Coffee and Cocoa Using Srucural Vecor Error Correcion Model By Ibrahim Bamba Michael Reed Preared for resenaion a he Meeing of American Agriculural Economiss Associaion,

More information

Employment, Family Union, and Childbearing Decisions in Great Britain

Employment, Family Union, and Childbearing Decisions in Great Britain Employmen, Family Union, and Childbearing Decisions in Grea Briain Arnsein Aassve Simon Burgess Carol Propper Ma Dickson Conens 1. Inroducion... 1 2. Daa... 3 Union formaion and dissoluion... 3 Employmen

More information

Working Paper

Working Paper Inernaional Nework for Economic Research Working Paper 010.1 Modelling he Cyclical Behaviour of Wine Producion in he Douro Region Using a Time-Varying Parameers Approach by Mario Cunha (Universidade do

More information

Milda Maria Burzała * Determination of the Time of Contagion in Capital Markets Based on the Switching Model

Milda Maria Burzała * Determination of the Time of Contagion in Capital Markets Based on the Switching Model D YNAMIC E CONOMETRIC M ODELS DOI: hp://dx.doi.org/10.12775/dem.2013.004 Vol. 13 (2013) 69 85 Submied Ocober 10, 2013 ISSN Acceped December 30, 2013 1234-3862 Milda Maria Burzała * Deerminaion of he Time

More information

ICC September 2018 Original: English. Emerging coffee markets: South and East Asia

ICC September 2018 Original: English. Emerging coffee markets: South and East Asia ICC 122-6 7 September 2018 Original: English E International Coffee Council 122 st Session 17 21 September 2018 London, UK Emerging coffee markets: South and East Asia Background 1. In accordance with

More information

IRREVERSIBLE IMPORT SHARES FOR FROZEN CONCENTRATED ORANGE JUICE IN CANADA. Jonq-Ying Lee and Daniel S. Tilley

IRREVERSIBLE IMPORT SHARES FOR FROZEN CONCENTRATED ORANGE JUICE IN CANADA. Jonq-Ying Lee and Daniel S. Tilley SOUTHERN JOURNAL OF AGRICULTURAL ECONOMICS DECEMBER, 1983 IRREVERSIBLE IMPORT SHARES FOR FROZEN CONCENTRATED ORANGE JUICE IN CANADA Jonq-Ying Lee and Daniel S. Tilley Canada is he mos imporan U.S. expor

More information

The Role of Infrastructure Investment Location in China s Western Development

The Role of Infrastructure Investment Location in China s Western Development The Role of Infrasrucure Invesmen Locaion in China s Wesern Developmen By Xubei Luo Developmen of he wesern region is vial o he balanced growh of China. Luo sudies he impacs of infrasrucure invesmen ha

More information

Deakin Research Online

Deakin Research Online Deakin Research Online This is he published version: Widjaja, Wany 2010, Modelling he cooling of coffee : insighs from a preliminary sudy in Indonesia, in MERGA 2010 : Shaping he fuure of mahemaics educaion

More information

Stock Market Liberalizations and Efficiency: The Case of Latin America

Stock Market Liberalizations and Efficiency: The Case of Latin America MPR Munich Personal RePEc rchive Sock Marke Liberalizaions and Efficiency: he Case of Lain merica João Paulo Vieio and Wing-Keung Wong and Sheung Chi Chow 06 Online a hps://mpra.ub.uni-muenchen.de/68949/

More information

Jordan Journal of Mathematics and Statistics (JJMS) 8(3), 2015, pp

Jordan Journal of Mathematics and Statistics (JJMS) 8(3), 2015, pp Jordan Journal of Mahemaics and Saisics (JJMS) 8(3), 015, pp. 57-70 FORECASTING THE TEA PRODUCTION OF BANGLADESH:APPLICATION OF ARIMA MODEL ** MD. MOYAZZEM HOSSAIN (1) AND FARUQ ABDULLA () ABSTRACT: Bangladesh

More information

Economics of grape production in Marathwada region of Maharashtra state

Economics of grape production in Marathwada region of Maharashtra state Volume 5 Issue 2 Sepember, 2014 179-183 e ISSN-2231-6434 Inernaional Research Journal of Agriculural Economics and Saisics Visi Us - www.researchjournal.co.in DOI : 10.15740/HAS/IRJAES/5.2/179-183 Research

More information

Modelling Financial Markets Comovements During Crises: A Dynamic Multi-Factor Approach.

Modelling Financial Markets Comovements During Crises: A Dynamic Multi-Factor Approach. Modelling Financial Markes Comovemens During Crises: A Dynamic Muli-Facor Approach. Marin Belvisi, Riccardo Pianei, Giovanni Urga February 24, 2014 We wish o hank paricipans in he Finance Research Workshops

More information

Tea Statistics Report 2015

Tea Statistics Report 2015 Tea Statistics Report 215 Introduction This report presents the scope and scale of the UTZ tea program in 215. Throughout this report tea also includes rooibos unless otherwise specified. The statistics

More information

Hi-Stat. Discussion Paper Series. Estimating Production Functions with R&D Investment and Edogeneity. No.229. Young Gak Kim.

Hi-Stat. Discussion Paper Series. Estimating Production Functions with R&D Investment and Edogeneity. No.229. Young Gak Kim. Hi-Sa Discussion Paper Series No.229 Esimaing Producion Funcions wh R&D Invesmen and Edogeney Young Gak Kim December 2007 Hosubashi Universy Research Un for Saisical Analysis in Social Sciences A 21s-Cenury

More information

The Determinants of Supply of Kenya s Major Agricultural Crop Exports from 1963 to 2012

The Determinants of Supply of Kenya s Major Agricultural Crop Exports from 1963 to 2012 Inernaional Journal of Buine, Humaniie and Technology Vol. 3 No. 5; May 13 The Deerminan of Supply of Kenya Major Agriculural Crop Expor from 1963 o 1 Leniy Kananu Maugu Lecurer in Economic, Chuka Univeriy

More information

Citation for published version (APA): Hai, L. T. D. (2003). The organization of the liberalized rice market in Vietnam s.n.

Citation for published version (APA): Hai, L. T. D. (2003). The organization of the liberalized rice market in Vietnam s.n. Universiy of Groningen The organizaion of he liberalized rice marke in Vienam Hai, L.T.D. IMPORTNT NOTE: You are advised o consul he ublisher's version (ublisher's PDF) if you wish o cie from i. Please

More information

QUARTELY MAIZE MARKET ANALYSIS & OUTLOOK BULLETIN 1 OF 2015

QUARTELY MAIZE MARKET ANALYSIS & OUTLOOK BULLETIN 1 OF 2015 QUARTELY MAIZE MARKET ANALYSIS & OUTLOOK BULLETIN 1 OF 2015 INTRODUCTION The following discussion is a review of the maize market environment. The analysis is updated on a quarterly 1 basis and the interval

More information

UTZ Coffee Statistics Report 2017

UTZ Coffee Statistics Report 2017 UTZ Coffee Statistics Report 2017 UTZ is the largest certification program in the world for sustainable coffee More than 480,000 coffee farmers and workers benefit from the UTZ program UTZ certified coffee

More information

HONDURAS. A Quick Scan on Improving the Economic Viability of Coffee Farming A QUICK SCAN ON IMPROVING THE ECONOMIC VIABILITY OF COFFEE FARMING

HONDURAS. A Quick Scan on Improving the Economic Viability of Coffee Farming A QUICK SCAN ON IMPROVING THE ECONOMIC VIABILITY OF COFFEE FARMING HONDURAS A Quick Scan on Improving the Economic Viability of Coffee Farming 1 OBJECTIVES OF STUDY Overall objective Identify opportunities for potential benefits to coffee farmers from improved farm profitability

More information

LIQUID FLOW IN A SUGAR CENTRIFUGAL

LIQUID FLOW IN A SUGAR CENTRIFUGAL LIQUID FLOW IN A SUGAR CENTRIFUGAL C.P. Please, N.D. Fowkes, D.P. Mason, C.M. Khalique, A. Huchinson and M. C. Rademeyer Indusry represenaives Richard Loubser and Seve Davis Absrac Massecuie is a mixure

More information

M03/330/S(2) ECONOMICS STANDARD LEVEL PAPER 2. Wednesday 7 May 2003 (morning) 2 hours INSTRUCTIONS TO CANDIDATES

M03/330/S(2) ECONOMICS STANDARD LEVEL PAPER 2. Wednesday 7 May 2003 (morning) 2 hours INSTRUCTIONS TO CANDIDATES c PROGRAMA IB DIPLOMA PROGRAMME PROGRAMME DU DIPLÔME DU BI DEL DIPLOMA DEL BI M03/330/S(2) ECONOMICS STANDARD LEVEL PAPER 2 Wednesday 7 May 2003 (morning) 2 hours INSTRUCTIONS TO CANDIDATES! Do not open

More information

SMALLHOLDER TEA FARMING AND VALUE CHAIN DEVELOPMENT IN CHINA

SMALLHOLDER TEA FARMING AND VALUE CHAIN DEVELOPMENT IN CHINA SMALLHOLDER TEA FARMING AND VALUE CHAIN DEVELOPMENT IN CHINA Intersessional Meeting of the Intergovernmental Group on Tea Rome, 5-6 May 2014 Cheng Fang, Economist, Trade and Markets Division, FAO Yanjiong

More information

The state of the European GI wines sector: a comparative analysis of performance

The state of the European GI wines sector: a comparative analysis of performance The state of the European GI wines sector: a comparative analysis of performance Special Report November 2017 1. Overview of a growing global wine market Wine is one of the most globalised products. The

More information

J / A V 9 / N O.

J / A V 9 / N O. July/Aug 2003 Volume 9 / NO. 7 See Story on Page 4 Implications for California Walnut Producers By Mechel S. Paggi, Ph.D. Global production of walnuts is forecast to be up 3 percent in 2002/03 reaching

More information

Coffee Supply Chain Development and Tourism in Timor-Leste

Coffee Supply Chain Development and Tourism in Timor-Leste Coffee Supply Chain Development and Tourism in Timor-Leste David Freedman, Asian Development Bank Country Economist, Timor-Leste. 18 July, 2016, Pacific Update Conference, Suva, Fiji. Today s Presentation

More information

Overview of the Manganese Industry

Overview of the Manganese Industry 39th Annual Conference Istanbul, Turkey 2013 Overview of the Manganese Industry International Manganese Institute Alberto Saavedra Market Research Manager June, 2013 Introduction Global Production Supply,

More information

WP Board 1035/07. 3 August 2007 Original: English. Projects/Common Fund

WP Board 1035/07. 3 August 2007 Original: English. Projects/Common Fund WP Board 1035/07 International Coffee Organization Organización Internacional del Café Organização Internacional do Café Organisation Internationale du Café 3 August 2007 Original: English Projects/Common

More information

Growing divergence between Arabica and Robusta exports

Growing divergence between Arabica and Robusta exports Growing divergence between Arabica and Robusta exports In April 218, the ICO composite indicator decreased by.4% to an average of 112.56, with the daily price ranging between 11.49 and 114.73. Prices for

More information

Pub mentors. Greets Inn, Warnham. The Great Lyde, Yeovil. The Elephant, Bristol. College Arms, Stratford

Pub mentors. Greets Inn, Warnham. The Great Lyde, Yeovil. The Elephant, Bristol. College Arms, Stratford Pub menors Offering help in overcoming he significan challenges facing pubs is he key heme of Pub Menor, he brand-new join iniiaive from Coca-Cola Enerprises Open More Business and he Publican s Morning

More information

Dairy Outlook. December By Jim Dunn Professor of Agricultural Economics, Penn State University. Market Psychology

Dairy Outlook. December By Jim Dunn Professor of Agricultural Economics, Penn State University. Market Psychology Dairy Outlook December 2015 By Jim Dunn Professor of Agricultural Economics, Penn State University Market Psychology The Class III market has taken a beating lately as cheese prices have drifted down and

More information

Ecobank s pan-african footprint. Africa-Asia trade flows

Ecobank s pan-african footprint. Africa-Asia trade flows Ecobank s pan-african footprint Africa-Asia trade flows Ghana, Côte d Ivoire and Nigeria: The changing face of West African cocoa Dr Edward George Head of Group Research, Ecobank Lagos, February 3rd 2016

More information

REMARKS BY PAUL BULCKE, GROUP CHIEF EXECUTIVE OFFICER, NESTLÉ S.A. MEDIA CONFERENCE, NAIROBI, FRIDAY, JULY 2, 2010

REMARKS BY PAUL BULCKE, GROUP CHIEF EXECUTIVE OFFICER, NESTLÉ S.A. MEDIA CONFERENCE, NAIROBI, FRIDAY, JULY 2, 2010 REMARKS BY PAUL BULCKE, GROUP CHIEF EXECUTIVE OFFICER, NESTLÉ S.A. MEDIA CONFERENCE, NAIROBI, FRIDAY, JULY 2, 2010 Disclaimer This speech might not reflect absolutely all exact words spoken. This speech

More information

UTZ Cocoa Statistics Report 2017

UTZ Cocoa Statistics Report 2017 UTZ Cocoa Statistics Report 2017 UTZ is the largest program in the world for sustainable cocoa There are more than 760,000 cocoa farmers in the UTZ program UTZ certified cocoa is produced in 21 countries

More information

Downward correction as funds respond to increasingly positive supply outlook

Downward correction as funds respond to increasingly positive supply outlook Downward correction as funds respond to increasingly positive supply outlook Coffee prices fell sharply at the end of April as institutional investors sold off their positions. The coffee market continues

More information

WP Board No. 934/03. 7 May 2003 Original: English. Executive Board May 2003 London, England

WP Board No. 934/03. 7 May 2003 Original: English. Executive Board May 2003 London, England WP Board No. 934/03 International Coffee Organization Organización Internacional del Café Organização Internacional do Café Organisation Internationale du Café 7 May 2003 Original: English Executive Board

More information

Contents 1. Introduction Chicory processing Global Trends in Production, Producer Prices and Trade of Chicory...

Contents 1. Introduction Chicory processing Global Trends in Production, Producer Prices and Trade of Chicory... i ii Contents 1. Introduction... 1 2. Chicory processing... 1 3. Global Trends in Production, Producer Prices and Trade of Chicory... 3 4. SA s Production, Producer Prices, Gross Value and Trade Patterns

More information

ETHIOPIA. A Quick Scan on Improving the Economic Viability of Coffee Farming A QUICK SCAN ON IMPROVING THE ECONOMIC VIABILITY OF COFFEE FARMING

ETHIOPIA. A Quick Scan on Improving the Economic Viability of Coffee Farming A QUICK SCAN ON IMPROVING THE ECONOMIC VIABILITY OF COFFEE FARMING ETHIOPIA A Quick Scan on Improving the Economic Viability of Coffee Farming 1 OBJECTIVES OF STUDY Overall objective Identify opportunities for potential benefits to coffee farmers from improved farm profitability

More information

2016 China Dry Bean Historical production And Estimated planting intentions Analysis

2016 China Dry Bean Historical production And Estimated planting intentions Analysis 2016 China Dry Bean Historical production And Estimated planting intentions Analysis Performed by Fairman International Business Consulting 1 of 10 P a g e I. EXECUTIVE SUMMARY A. Overall Bean Planting

More information

ANALYSIS ON THE STRUCTURE OF HONEY PRODUCTION AND TRADE IN THE WORLD

ANALYSIS ON THE STRUCTURE OF HONEY PRODUCTION AND TRADE IN THE WORLD ANALYSIS ON THE STRUCTURE OF HONEY PRODUCTION AND TRADE IN THE WORLD GU G., ZHANG Ch., HU F.* Department of Sericulture and Apiculture, College of Animal Science Zhejiang University, Hangzhou 310029, CHINA

More information

and the World Market for Wine The Central Valley is a Central Part of the Competitive World of Wine What is happening in the world of wine?

and the World Market for Wine The Central Valley is a Central Part of the Competitive World of Wine What is happening in the world of wine? The Central Valley Winegrape Industry and the World Market for Wine Daniel A. Sumner University it of California i Agricultural l Issues Center January 5, 211 The Central Valley is a Central Part of the

More information

Tanzania. Coffee Annual. Tanzania Coffee Annual Report

Tanzania. Coffee Annual. Tanzania Coffee Annual Report THIS REPORT CONTAINS ASSESSMENTS OF COMMODITY AND TRADE ISSUES MADE BY USDA STAFF AND NOT NECESSARILY STATEMENTS OF OFFICIAL U.S. GOVERNMENT POLICY Required Report - public distribution Date: GAIN Report

More information

EMBARGO TO ON FRIDAY 16 SEPTEMBER. Scotch Whisky Association. Exports of Scotch Whisky; Year to end of June 2016 (2016 H1)

EMBARGO TO ON FRIDAY 16 SEPTEMBER. Scotch Whisky Association. Exports of Scotch Whisky; Year to end of June 2016 (2016 H1) EMBARGO TO 00.01 ON FRIDAY 16 SEPTEMBER Scotch Whisky Association Exports of Scotch Whisky; Year to end of June 2016 (2016 H1) VOLUME UP 3.1% to 531 MILLION bottles VALUE DOWN SLIGHTLY BY 1.0% TO 1.70

More information

UTZ Tea Statistics Report 2017

UTZ Tea Statistics Report 2017 UTZ Tea Statistics Report 2017 More than 107,000 tea farmers and workers benefit from the UTZ program UTZ certified tea is produced in 13 countries In 2017, more than 4.6 billion cups of UTZ certified

More information

much better than in As may be seen in Table 1, the futures market prices for the next 12 months

much better than in As may be seen in Table 1, the futures market prices for the next 12 months Dairy Outlook December 2009 By Jim Dunn Professor of Agricultural Economics, Penn State University Market Psychology prices are higher than last month, especially Class IV. The outlook for dairy prices

More information

GLOBAL DAIRY UPDATE KEY DATES MARCH 2017

GLOBAL DAIRY UPDATE KEY DATES MARCH 2017 MARCH 2017 GLOBAL DAIRY UPDATE European milk production decreased for the seventh consecutive month, while the US remains strong. The rate of decline in New Zealand production is easing. US exports continue

More information

Record exports in coffee year 2017/18

Record exports in coffee year 2017/18 Record exports in coffee year 2017/18 Total coffee exports increased each year since 2010/11 with a new record reached in 2017/18 at 121.86 million bags, 2% higher than 2016/17. In the twelve months ending

More information

United Nations Conference on Trade and Development

United Nations Conference on Trade and Development United Nations Conference on Trade and Development 10th MULTI-YEAR EXPERT MEETING ON COMMODITIES AND DEVELOPMENT 25-26 April 2018, Geneva Cocoa Boom & Bust: A Review of Cocoa Market Developments By Laurent

More information

Cocoa Prepared by Foresight October 3, 2018

Cocoa Prepared by Foresight October 3, 2018 Cocoa Prepared by Foresight October 3, 2018 TABLES Cocoa Bean Price Forecast... P. 4 World Cocoa Supply/Demand, Crop Year... P. 7 World Cocoa Production... P. 8 Cocoa Crops in Major Producing Countries...

More information

THE AFRICAN COFFEE MARKET THE CASE OF MALAWI COFFEE INDUSTRY)

THE AFRICAN COFFEE MARKET THE CASE OF MALAWI COFFEE INDUSTRY) THE AFRICAN COFFEE MARKET THE CASE OF MALAWI COFFEE INDUSTRY) BY HARRISON B. KALUA CHIEF EXECUTIVE OFFICER MZUZU COFFEE PLANTERS COOPERATIVE UNION AND CHAIRMAN OF COFFEE ASSOCIATION OF MALAWI 2.0. BACKGROUND:

More information

Bean corridors: A novel approach to scale up national and regional trade in Africa

Bean corridors: A novel approach to scale up national and regional trade in Africa Bean corridors: A novel approach to scale up national and regional trade in Africa Photo: S. Malyon (CIAT) PABRA has developed a new commodity corridors approach, which aims to eliminate production bottlenecks,

More information

World of sugar PAGE 54

World of sugar PAGE 54 World of sugar More than 1 countries produce sugar, about 8% of which is made from sugar cane grown primarily in the tropical and sub-tropical zones of the southern hemisphere, and the balance from sugar

More information

Volume 30, Issue 1. Gender and firm-size: Evidence from Africa

Volume 30, Issue 1. Gender and firm-size: Evidence from Africa Volume 30, Issue 1 Gender and firm-size: Evidence from Africa Mohammad Amin World Bank Abstract A number of studies show that relative to male owned businesses, female owned businesses are smaller in size.

More information

World vitiviniculture situation

World vitiviniculture situation World vitiviniculture situation Surface area Grape Wine Global grape production Production Consumption Trade 2016 FAO-OIV Focus: Table and Dried Grapes 2 Global area under vines Area under vines in the

More information