Estimating price rigidity in coffee markets: A cross country comparison

Transcription

1 Esimaing price rigidiy in coffee markes: A cross counry comparison Iqbal Syed Ph.D Candidae Kevin Fox Ph.D Thesis Supervisor School of Economics Universiy of New Souh Wales Preliminary Commens Welcome i.syed@suden.unsw.edu.au Absrac Price rigidiy heories explain why a firm may no change is prices in response o changes in demand and cos condiions. Undersanding rigidiy and is deerminans is imporan in evaluaing he poenial effeciveness of moneary policy. This paper uses error correcion models o esimae he degree of price rigidiy in he coffee markes of 16 differen counries. A high degree of rigidiy was found in seven counries, while relaively responsive prices were found in he remaining counries. A general conclusion ha can be drawn from he sudy is ha he price adjusmen occurs wihin wo quarers afer he shock is imposed. The observed rigidiy seems o be largely deermined by he high degree of marke concenraion. The economic margin resuling from imperfec markes allows a firm no o adjus prices o cos changes. I would like o hank Simon Angus, Daniel Buncic, Denzil Fiebig, Lorraine Ivancic, Nigel Sapledon and Minxian Yang for helpful commens and discussions.

2 1. Inroducion How prices are adjused by firms o exernal shocks has generaed considerable ineress o indusrial organizaion heoriss and has imporan implicaion for new Keynesian macroeconomics. I is crucial o undersand he naure of he price adjusmen processes for he appropriae design and conduc of moneary policy. Despie is cenral imporance in micro and macro heories and relevance in economic policies, here are only a limied number of empirical sudies (Blinder e al., 1998; Dua e al., 2002; Dhyne e al., 2005). Price rigidiy heories provide hypoheses abou why a firm may no choose o change is prices in response o changes in he demand and cos condiions i faces. There are a number of differen explanaions for he exisence of nominal price rigidiy, each following differen heoreical roues o explain he sluggishness of price changes. Blinder (1991) and Blinder e.al (1998) poined ou ha he economic research has been singularly unsuccessful in conducing empirical ess on he validiy of paricular rigidiy heories. Alhough i would be nice o know boh he degree of price rigidiy and he underlying explanaions for is exisence, a measure of he degree is imporan on is own sake. There are a number of ways in which price rigidiy has been measured in he empirical lieraure. Perhaps he mos popular mehod is simply o look a he size and he frequency of price changes in a given period. Unil recenly, some of he bes sources of informaion for he size and frequency of price changes for paricular caegories of producs were Carlon (1986), Viqueira (1991), Lach and Tsiddon (1992), Kashyap (1995) and Blinder e al. (1998). A number of recen sudies include Baharad and Eden (2004), Bils and Klenow (2004), Herrmann e al. (2005), Alvarez e al. (2005), Fabiani e al. (2005) and Nakamura and Seinsson (2006a). The sudy of Alvarez e al. is based on produc level price daa used o consruc he Consumer Price Index (CPI) and Producer Price Index (PPI) in he counries in he Euro area. Nakamura and Seinsson used similar se of daa used for he consrucion of CPI and PPI in US. These sudies are he firs wo sudies which covered wide range of producs boh in he consumer and producer caegories. The summary finding of hese sudies is ha price changes approximaely once a year. Noably, his finding is similar o he finding of he sudies underaken in he 1980s and 1990s menioned above, as poined ou by Nakamura and Seinsson, []he convenional wisdom from his lieraure was ha prices adjused on average once a year (p.1). The economeric analysis of individual price daa has provided wo alernaive measures of price rigidiy. One of hem is obained from he discree choice modeling framework and measures he probabiliy of price change wih respec o cos and demand shocks. The sudies ha fall in his caegory include Cecchei (1986), Dhalby (1992), Owen and Trzepacz (2002) and Campbell and Eden (2004). These sudies aemp o examine he facors ha deermine he incidence and size of price changes. Wihin he framework of dynamic economeric modeling such as vecor auoregression (VAR) and vecor error correcion models (VECM), Pelzman (2000) and Dua e al. (2002) esimae he ime dimension of price changes, i.e. he number of period price adjusmen akes place afer a cos or demand shock is being iniiaed. This noion of price rigidiy is of considerable ineres o macroeconomiss. In he lieraure, i is also referred o as he price ransmission or pass hrough effec. In his paper we examine he price rigidiy across 17 counries using publicly available daa on coffee prices a differen sages of producion and disribuion (farm, wholesale and reail secors). The frequency of observaion is monhly covering he period from 1976:1 o 2004:12. The counries considered in alphabeical order are Ausria, Belgium, Luxemburg, Denmark, 1

3 Finland, France, Germany, Ialy, Japan, he Neherlands, Norway, Porugal, Spain, Sweden, Swizerland, Unied Kingdom (UK) and Unied Saes of America (USA). 1 However, since he source of daa merges ogeher he reail coffee price daa of Belgium and Luxembourg, we consider hem as a union (referred o as Bellux).Using VECMs, he ransmission of shocks o he reail secor of coffee in hese counries was modeled. This paper conribues o he empirical lieraure of price rigidiy in a number of ways. Firs, here are only a limied number of ime series sudies which provide an empirical analysis of price rigidiy, as emphasized by Carlon (1986); Gordon (1990); Kashyap (1995); Dua e al. (2002). Second, we examine he price rigidiy in he way ha has macroeconomic policy implicaions. According o Blinder (1991), a more imporan quesion, from a macroeconomic perspecive, was o deermine how far price adjusmen lags behind demand and cos shocks. In his sudy we follow Blinder s suggesion as we believe ha his ype of sudy can provide us wih a beer undersanding of (1) he moneary policy ransmission process and (2) underlying inflaion. Price rigidiy, in our sudy, is measured as he effec of cos and foreign price changes on he domesic reail prices. In his conex he higher he lag he greaer is he degree of price rigidiy. Carlon and Perloff (1994) defined price rigidiy as a siuaion when price does no respond o variaion in coss and demand. This implies ha a price can have high frequency in erms of he number of imes i changes in a given period, bu i can sill be subjec o high degree of rigidiy if i does no respond o underlying changes in demand and supply. Conversely, a price wih a low degree of rigidiy may experience infrequen price changes. For he purpose of moneary policy and measures of underlying inflaion, i is imporan o idenify wheher he price response is due o underlying pressure of he economy. Dua e al. (2002, p.1876) observed ha.a leas some of he frequen price changes we observe in he daa may no be in response o changes in supply and demand condiions and herefore, may no be informaive abou he exen of price rigidiy/flexibiliy. Third, he sudy uses prices of a widely consumed produc, coffee. Coffee is currenly he second mos raded commodiy in he world (afer crude oil). The reason i is so widely raded is because of high volume of rade of boh coffee bean and processed coffee. There are some advanages in sudying price rigidiy wih a produc subjec o high degree of inernaional rade. The heory of spaial compeiion would sugges similar pricing behavior across counries. Therefore, variaion in price rigidiy could be aribued o he upsream sages of disribuion such as difference in he marke feaures, policy and economic environmen specific o ha paricular imporing counry. We conrol, a leas o some exen, for he differences in he feaures ha may resul in he variaion of price rigidiy up o he manufacuring level. Alhough he reail prices of coffee in all he counries do no relae o he same iem, hese reail prices are average and lised prices of coffee colleced by he respecive counries saisical agencies for he purpose of calculaion of he consumer price index. Ideally, we would have preferred o have ransacion price daa of some brands of coffee because of heir appropriaeness in he sudy of price rigidiy (Carlon, 1986; Lach and Tsiddon, 1992; Wynne, 1995; Levy, 2002). However, he reail price daa we are using are qualiy adjused and reflec he movemen of coffee prices in he respecive counries. Finally, unil recenly, here were no sudies ha examined he variaion of price rigidiy across counries. Mos of he exising sudies have examined he variaion across differen varieies 1 The choice of he counries has been dicaed by he ease of availabiliy of daa. 2

4 wihin a produc group in a single counry (for example, Kashyap, 1995; Dua e al., 2002; Levy e al., 2002) or across differen produc groups in a single counry (for example, Carlon, 1986; Kashyap, 1995; Bils and Klenow, 2004; Herrmann e al, 2005, Campbell and Eden, 2005). The few sudies ha looked a he variaion across counries are Alvarez e al. (2005) and Fabiani e al. (2005) 2. However, our sudy is differen in many respecs. In he Alvarez e al. sudy, a key measure of rigidiy ha was used for he basis of comparison was he frequency of price change. In our sudy, he comparison across counries is drawn wih respec o he dynamic impac of reail prices o exernal shocks. However, he firm inerview sudy of Fabiani e al. aemped o examine he naure in which he firms behave o cos shocks based on relaed inerview quesions. Thus he finding is based on he percepion of he responden on a paricular scenario, and herefore, may no be appropriae for his kind of analysis. The firm level survey may be more useful o undersand he underlying raionale of he behavior of he price seing pracices and make a cross secional comparison. However, he surveys aken a one poin of ime, and herefore, are no suiable for capuring dynamic feaures of he price adjusmen process. The paper is organized as follows. A descripion of he daa is provided in Secion 2. Secion 3 ses ou he economeric model used for esimaing price rigidiy. A number of sensiiviy ess were conduced on he model and hese ess and he resuls are described in secion 4. Secion 5 is devoed o a discussion on he possible causes of price rigidiy and variaions across counries. We conclude in secion Daa Descripions The sudy uses 18 monhly ime series daa on nominal prices. Among hese series, one series relaes o farm prices of coffee beans, one relaes o wholesale prices and he remaining 16 series relae o reail prices of coffee a 16 differen counries sudied in he paper. The daa is colleced from he Inernaional Coffee Organizaion (ICO) websie. 3 In addiion, he paper uses 16 series of consumer price indices (CPI) of he 16 respecive counries. 4 Among differen varieies of coffee beans produced in differen counries, for he farm price we used he price of Brazilian Arabica produced in Brazil. This is because (1) he variey is one of he dominan varieies 5 and (2) Brazil is he larges coffee producing counry in he world. In 2004, Brazil expored approximaely hiry per cen of he oal coffee bean expors of he world. 6 2 In recen years, a number of sudies were underaken by he Inflaion Persisence Nework (IPN) of he European Cenral Bank (ECB). One se of sudy is based on firm level surveys conduced in nine counries in he euro area. Fabiani e al (2005) summarizes he findings of he survey of 11,000 firms in Ausria, Belgium, France, Germany, Ialy, Luxembourg, he Neherlands, Porugal and Spain. These counries ogeher accouned for around 94 per cen of he GDP in Anoher se of sudy is based on he produc level CPI and/or PPI daa of differen counries in he Euro area. A summary of hese sudies are provided in Alvarez e al. (2005). 3 The ICO websie, is a rich source of daa for coffee producion, consumpion, rade and prices. Among ohers, i has various ime series on nominal prices of coffee beans a he farm and wholesale level, and nominal reail prices in differen counries. All prices are in US dollars. 4 The CPIs are CPI: All Iems, colleced from he OECD websie. 5 The Columbian mild is he nex imporan variey in erms of producion. We esimaed he same models wih he Columbian variey. The resuls obained were similar. 6 Auhor s own calculaion using expor daa obained from ICO websie. 3

5 For he nominal wholesale price, we used he ICO composie index which is consruced from four oher indices, each corresponding o a variey of coffee bean. 7 Figure 1 plos he composie index and is four componen indices and able 1 provides he pairwise correlaion coefficien among hese indices. Meha and Chavas (2004) menion ha since wholesale price of differen varieies of coffee beans co-move, much can be learn abou he aggregae behavior from one of he wholesale price indicaors. The sample range for he farm and wholesale prices used in he sudy is 1976:1 2004:12. Inser figure 1 here. Inser able 1 here. The nominal reail prices of coffee of differen counries are colleced by he ICO from he respecive saisical agencies. Excep for six counries, all oher reail prices cover he period from 1976:1 o 2004:12, hus maching he sample period for he farm and wholesale prices. The daa used for he following counries cover shorer periods: Finland (1982:1-2004:12), Germany (1976:1-2003:12), Japan (1982:1-2004:12), Porugal (1978:1-2004:12), Spain (1977:1-2004:12) and Sweden (1979:1-2004:12). We provide some descripive saisical analysis o beer acquain ourselves wih he daa se. Figure 1 shows he presence of high degree of variaion of wholesale prices over he enire sample period. In addiion, here are some marked increases and decreases in some periods, possibly caused by adverse weaher condiions 8. The wholesale prices show a slighly negaive rend. In order o check wheher here have been any changes in he mean prices, we divide he whole sample period in wo ways. Firs, by dividing i ino hree sub-periods: 1976:1-1985:12 (period I), 1986:1-1995:12 (period II) and 1996:1-2004:12 (period III) 9 and, second, by wo periods: 1976:1-1989:6 (pre-ica period) and 1989:7-2004:12 (pos-ica period) 10. The calculaed means and heir sandard deviaions are repored in able 2. The means of farm prices are almos he same among he sample sub-periods. However, he means of wholesale prices decline gradually over differen sample periods; for example, i was cens in he pre-ica and cens in he pos-ica period. Figure 3 plos he nominal reail prices of 16 counries. One common feaure of he graphs is ha hey exhibi high degree of variaion wih some marked changes in some periods. Wihin hese ups and down, some of he reail prices appear o have rends. For example, while a posiive rend can be seen in prices of Ialy, Swizerland and UK, a negaive rend is visible in case of prices of France. Table 2 repors he means of he reail prices for he whole sample period and he sample sub-periods. In 9 ou of 16 counries, he mean price wen up in he pos-ica period as compared o ha in he pre-ica period. This is he case even hough he nominal wholesale price decreased beween hese wo periods. Inser figure 2 here. 7 ICO composie index is a weighed average of four oher componen indices. These four componen indices relae o four differen varieies of coffee: Brazilian Naural, Columbian Mild Arabicas, Oher Mild Arabicas and Robusas. 8 Excessive fros in , and 1997 had adverse impac on he producion of coffee beans. 9 This division of sample period ino hree sub-samples is arbirary. For mos counries, period 1 and 2 have 120 observaions each and period 3 has 108 observaions. 10 ICA refers o Inernaional Coffee Agreemen. ICA. 4

6 I is ineresing o look a he raio of wholesale o reail prices o undersand he relaive imporance of coffee beans o he processed coffee sold in he reail marke. We plo he graphs of he raios in figure 4. Alhough he decline in he raio is more pronounced in some counries han he ohers, wihou any excepion, all of hem exhibi negaive rends. The means of he raios in differen sample periods are provided in able 3. In some counries, he raio declined by more han half beween he pre-ica and pos-ica periods; from example, Ialy from 0.4 o 0.16, Japan from 0.16 o 7 per cen, UK from 0.18 o 7 per cen and USA from 1 o 0.25 per cen. One can conclude ha he relaive imporance of coffee beans in he oal value of producion of coffee has declined gradually over he sample period. Inser figure 3 here. Wih he excepion of Japan, he real reail prices have fallen seadily in all he counries. The real reail prices, shown in figure 4, are calculaed by dividing he nominal prices wih he counry specific consumer price indices. One should noe ha he fall in he real price of coffee beans, boh a he farm and he wholesale level, is expeced o be much more pronounced because much of he coffee beans are produced in counries experiencing relaively high rae of inflaion. Therefore, one can expec ha he raio of real wholesale o real reail prices fell a a faser rae han he corresponding raio of nominal wholesale o nominal reail prices. However, he fac ha real reail price of coffee declined over years indicaes he imporance of coffee bean prices in he deerminaion of coffee prices a he reail level. Inser figure 4 here. There exiss high degree of correlaion among he real reail prices. Table 4 shows he pair wise correlaion coefficiens among he nominal reail prices for differen sample periods. The correlaions among he real prices are provided in able 5. A comparison beween hese wo ses of correlaion coefficiens shows ha he real prices move ogeher much closer han he nominal prices. This implies ha he counry specific facors play an imporan role in he deerminaion of nominal prices. The augmened Dickey-Fuller (ADF) and Perron s es for srucural change (Perron) have been used o es for he saionariy of he series. While he ADF es shows ha 11 ou of 18 series are saionary, he Perron es finds only five series as saionary. The five series found o be saionary by Perron es are also saionary in he ADF es. Given his discrepancy, one may argue ha he resuls obained from he above wo ess are inconclusive. However, he resuls demonsrae ha he series in our sudy are eiher near uni roo or uni roo process, because he esimaed characerisics roos, as repored in he column 6 in able 7 and column 10 in able 8, are very close o one. For he purpose of modeling, his led us o use he firs difference of he respecive series. In conducing he ADF, regression equaions of he following form have been esimaed: y = a k 0 + a1 y 1 + i. i= 1. β y + ε (1) i The lags of he firs difference have been chosen using Akaike s informaion crierion (AIC). Rejecion of he null hypohesis a 1 =0 would mean ha he process is saionary. The resuls of he ADF ess repored in able 6 shows ha while he farm price is saionary, he wholesale price is non-saionary. Among a oal of sixeen reail prices, six are found o be non-saionary. Among he 7 series repored o be non-saionary, only one of hem is non-saionary on a borderline case and ohers are non-saionary wih very low -saisics. However, one should noe 5

7 ha he power of he es, he probabiliy of correcly rejecing he null hypohesis of a uni roo, is very low, paricularly when he esimaed characerisic roos are very close o uniy. The characerisic roos repored in column 6 of able 6 range beween and Inser able 6 here. In he presence of a srucural break, he ADF es saisics are biased owards he non-rejecion of uni roos. In order o es for he suspeced srucural break due o he collapse of Inernaional Coffee Agreemen in June 1989, we conduc Chow ess on equaion (1). Table 7 repors he esimaed F saisics and he respecive probabiliy values. According o he ess, evidence of a srucural break is found in four series; reail prices of Ialy, Spain, Swizerland and UK. When ADF ess were conduced on he whole sample for he above four series, he null hypohesis of a uni roo could no be rejeced only for he series of Spain. A common economeric procedure for esing he uni roo hypohesis wih a known break poin is o spli he sample and conduc he ADF es on each par. The problem wih his procedure is a very low power due o he reducion in he degrees of freedom in esimaion for each of he equaions (Perron, 1989). Perron developed a mehodology o es he uni roo hypohesis ha reas a known srucural break as exogenous. He showed ha he es has more power han conducing various ADF ess on spli samples. In order o check for he uni roo hypohesis we esimaed he following equaion which Perron referred o as he crash model. y = a k 0 + µ 1. DL + µ 2. DP + a1. y 1 + a2. + β i. y i + ε (2) i= 1 where, D L =1 for 1989:7 and zero oherwise; D P =1 for =1989:7 and zero oherwise. The null hypohesis of he es is a uni roo process having a one ime change in he level a he break. Therefore in our model, he null hypohesis refers o a 1 =0, a 2 =0 and µ 2 0. Under he alernaive hypohesis, he series is rend saionary wih a permanen one-ime break. In our equaion his relaes o a 1< 0 and µ 1 0. The rend coefficien a2 can ake any value including zero depending on wheher he series has a deerminisic rend or no. The resuls of he esimaed equaion are repored in able 10. The calculaed saisic for he null of uni roo is compared wih he Perron s criical values for he appropriae lambda value. 11 The resuls show ha only 5 of he 18 series are saionary wihou any evidence of srucural change a he break poin. These 5 series relae o reail prices of Ialy, Norway, Neherlands, UK and US. The remaining 13 series are uni roo process and, only hree of hese series, farm price, wholesale price and reail price of France, indicae having a break a he poin of collapse of he ICA. Inser able 8 here. The findings from Perron s crash model receive addiional suppor because of he similariy of he resuls across equaions. The esimaed characerisic roos are very close ranging beween 0.92 and The -saisics of a 1 are very similar; 14 of hem are in he range beween 3 and 5. One more similariy is relaed o deerminisic rend. The esimaed coefficiens and heir corresponding sandard deviaion of he rend erm indicae he absence of deerminisic rend in any of he equaion. 11 Lambda is he raio beween he break poin and he oal number of observaions. Perron s criical values depend on his raio. When he raio is 0 or 1, hen he criical values collapse o he corresponding Dickey- Fuller criical values. 6

8 The above findings from he ADF and Perron s es show ha here exiss dissimilariy in characerizing wheher a series is saionary or non-saionary. However, he characerisics roos are very close o uniy according o boh ess, and he low power of he ess for poins close o uniy may be he reason behind he conflicing resuls. Enders (2004, p.210) wih he help of a Mone Carlo experimen showed ha he es has very low power o deec near uni roo series. However, having said all his, one can conclude ha he series are eiher borderline saionary (near uni roo) or uni roo process. Therefore, for he purpose of modeling, we used he firs difference of he series. The firs difference of a uni roo process de-rend he series. 12 For he near uni roo process, Mone Carlo experimens have shown ha shor-run forecass from differenced models are usually superior o he forecass from he series on levels (Lukepohl, 2006). IV. The economeric model In his secion, we describe he coinegraion analysis and he economeric model used for esimaion. Our primary ineres in his sudy are o evaluae he impac on he nominal reail prices o (1) cos shocks and (2) shocks originaing from large coffee consuming counries (referred o as foreign shock). Coffee beans are one of he major raw maerials in he producion of processed coffee and as such provide an imporan source of poenial supply side shocks. There are wo possible daa sources for coffee bean prices: (1) he farm gae prices or (2) he wholesale prices. As menioned before, we used ICO composie index for wholesale prices as he source of cos shock. The scenario here is similar o he n sages of producion linked by a chain of producion promulgaed proposed by Blanchard (1993). An implicaion of Blanchard s heory is ha, he closer he wo sages of producion are, he shorer he number of lags (in he number of periods) beween he cos changes and he price change. In oher words, he degree of price rigidiy will be calculaed as lower when cos shock originaes from he wholesale secor as compared o when i originaes from an upper sream sage of producion, such as farm secor. However, one should noe ha a change in he cos from any sage does no have o physically pass hrough all or any of he inermediae sages o have impac on prices a any downsream sages of producion. For example, i is possible ha a change in he farm level price can have a direc impac on he reail price raher han passing hrough he wholesale secor. Therefore, he impac of a price change a he farm level can impac on he reail price eiher direcly or indirecly. Due o he possibiliy of he exisence of muliple channel relaionship, Dua e al. (2002) esimaed a model including hree sages of producion, farm, wholesale and reail sages. Two of he larges coffee imporing counries of he world are he US and Germany. In 2004, he US and Germany impored around 20.1 and 15.2 per cen of he coffee bean imporing counries of he world. 13 We ook a weighed average of coffee prices in he US and Germany o ge a represenaive price of he large coffee consuming counries (Figure 5). In his sudy, we refer o his price as foreign price and any shock originaing from his marke as foreign shock. 14 The 12 To provide addiional suppor, we esed he uni roo hypohesis on he firs differences of he series. The null hypoheses of uni roo are rejeced in all cases a 1 per cen significance level. 13 The hird larges coffee imporing counry of he world is Japan, accouning for 6.3 per cen of he oal coffee bean impors. 14 The reail prices of he US and Germany may refer o differen ypes of coffee. Therefore, before we ook he weighed average, we demeaned he prices. For weighs, we use he yearly impor volumes obained from he ICO websie. 7

9 idea is o assess he impac of shock originaing from a large coffee consuming counry on he domesic reail price. Inser figure 5 here. We will model a hree equaion sysem. These equaions relae o wholesale, reail and foreign marke. The srucural vecor auo-regressive (SVAR) model is as follows: Sx p = Γ + Γ x + ε 0 i i (1) i= 1 r where, x =[x x f w x ] is a (3 X 1) vecor of prices, x r, x f and x w are he reail, foreign and wholesale prices, respecively. S is a (3 X 3) marix of conemporaneous coefficiens, S 1 = s fr swr s s rf 1 wf srw s fw 1 r Γ 0 is he (3 X 1) vecor of inerceps. ε = [ε disurbances. ε f ε w ] is a (3 X 1) vecor of whie noise A sandard form vecor auo-regressive (VAR) model is obained by pre-muliplying equaion (1) by S -1. p x = A0 + Ai x i + e (2) i= where, A0 = S Γ0, A i = S Γi and e = S ε. Since ε r, ε f and ε w are whie noise processes, i follows ha e r, e f and e w have zero mean, consan variance and are individually serially uncorrelaed. However, e r f, e and e w are correlaed wih each oher. Idenificaion resricion To analyze he impac of a srucural shock on he endogenous variables idenificaion resricions need o be imposed on he model. We se s rf, s rw and s fw equal o zero, which led o a model recursive in naure. Dua e al. (2002) imposed similar idenificaion resricion. Thus he marix of conemporaneous coefficiens of he esimaed model is, S 1 = s fr swr s 0 1 wf As he coffee producion chain is organized in a similar hierarchical way his resricion will also be applied o our model. The economic reasoning for such idenificaion follows from he fac ha we do no expec reail prices o have effec on he foreign and wholesale prices in he curren period and, similarly, foreign price o have an impac on he wholesale price in he curren period. One can jusify his reasoning on wo grounds. Firs, he marke becomes smaller as we move from wholesale o foreign, and hen foreign o reail marke. The wholesale coffee marke 8

10 caers o he coffee manufacurers/processors who supply o all he coffee imporing counries of he world, whereas a reail marke refers o only one of hose imporing counries. Noneheless, one may argue ha shocks from a large imporing counry could poenially affec he wholesale marke. However, noe ha he resricion only resrains he conemporaneous effec, hus allowing feedback in a backward direcion wih a one-period or more lag. The second reason arises if we hink abou he naure of he shocks expeced o originae from he reail marke. The reail markes which we model are all locaed in open economies, which reflec compeiive marke srucures and allow for impors a minimal or no ariff rae. Therefore, he supply shocks, if any, from hese markes are expeced o be small. On he demand side, coffee demand is expeced o be sable. Therefore, we would expec shocks of small magniude from smaller marke would no have conemporaneous effec on he upsream larger markes. The Error Correcion Model Granger and Newbold (1994) showed ha a regression model wih non-saionary variables migh resul in he esimaion of spurious regression. If he variables are inegraed of order one I(1) he firs difference de-rends he series. If he I(1) variables are no coinegraed, i is preferable o esimae VAR on firs differences raher han on levels (Enders, 2004). However, if he I(1) variables are coinegraed, hen esimaing a VAR in firs differences wihou he inclusion of a coinegraing relaion leads o misspecificaion error. Misspecificaion error is he resul of he exclusion of long run equilibrium relaionships among he variables. If his error is presen hen he resuling coefficien esimaes, inferences and innovaion accouning are no represenaive of he rue process (Enders, 2004). Therefore, i is imporan o check for he exisence of he coinegraing relaionship among he variables included in he VAR model. Alhough here are a number of alernaive approaches o esing for coinegraion Johansen s maximum likelihood approach (MLA) was used in his sudy for a number for reasons. Firs, i has he advanage of being able o esimae and es muliple coinegraing relaions. Second, we can es for he resriced version of he coinegraing vecors and he speed of adjusmen. In addiion o his, a comparaive sudy by Gonzalo (1989) on five alernaive mehods of finding long run relaionship found ha he Johansen s mehod provides he leas biased and mos symmerically disribued coefficien esimaes. This finding was robus o differen disribuional assumpions. The model in equaion (2) can alernaively be wrien as an error correcion model (ECM). p = a + x i= 1 x 0 x + e (3) i i where x = x x, = ( I ) and i = A j. I is a (3 X 3) ideniy marix. 1 p 0 A i i= 1 In equaion (3) Δx is saionary if 0 x 1 + i x i is saionary. This implies ha, i= 1 0 x 1, he linear combinaion of he variables (in levels) mus be saionary. The rank of Π 0 is equal o he number of independen coinegraing vecors. If rank (Π 0 )=0, hen equaion (3) is a VAR model in firs differences. Alernaively, if he rank of Π 0 is n, hen all variables are saionary. In our case, as n equals 2, here is a furher possibiliy ha Π is of rank 1. In his case p 1 p j= i+ 1 9

11 we would have a single coinegraing vecor and he expression. 0 x 1 would be he error correcion erm. Following Johansen, we define wo marices α and β, each of dimension (n X r), where r is he rank of Π 0 such ha Π 0 = αβ. Here β is he marix of coinegraing parameers and α is he weigh of each coinegraing vecor in he respecive equaion. α is commonly referred o as he speed of adjusmen parameers. The vecor β x -1 denoes he coinegraing relaion and, herefore, when i exiss is saionary. Johansen (1988) proposed wo ess for esimaing he number of coinegraing vecors, commonly referred o as λ race (r) and λ max (r). The λ race (r) ess he null hypohesis ha he number of disinc coinegraing vecors is less han or equal o r agains a general alernaive. The λ max (r) saisics es he null hypohesis ha he number of coinegraing vecors is r agains he alernaive of r+1 coinegraing vecors. The criical values abulaed by Johansen and Juselius (1990) have been used for hese ess. Using Johansen s MLA one can es for he coinegraing relaion and esimae he error correcion model simulaneously. Therefore, before applying he MLA one has o decide on he inclusion of exogenous variables and choose he appropriae lag lenghs for all he variables. The resuls of he ess can be quie sensiive o he lag lengh. The ess are also sensiive o he inclusion of he deerminisic erms. In our case, we included an inercep erm in he coinegraing vecor. The drif erm is no included because we do no expec ha he firs differences of he variables will have any rend (posiive or negaive). As for he inercep erm in he coinegraing relaion, i is a common pracice o include i unless here is a priori reason agains is inclusion. Lag lengh selecion The lag lengh selecion is an imporan sep as he es resuls can be sensiive o he chosen lag lengh. Enders (2004) saes ha he mos common procedure is based on he esimaion of vecor auoregression on levels. There are a number of mehods, such as Akaike Informaion Crierion (AIC), Schwarz Bayesian Crierion (SBC), Hannan-Quinn (HQ), Likelihood Raio es (LR), Final Predicion Error (FPE) which can be used o es for appropriae lag lengh. The selecion crieria menioned above provide differen ways of weighing improvemens in he explanaory power agains he loss of degrees of freedom. The resuls from hese ess can be used for coinegraion analysis and, consequenly, for he error correcion model. All hree mehods were used in his sudy. We sar wih a lag lengh of 13 monhs based on an a priori noion ha one year is a sufficienly long ime o capure he sysem s dynamics. The model was hen esimaed wih lag lenghs of 12 monhs, and subsequenly esimaed by reducing he lag lengh by one monh unil a lag lengh of one was reached. The lag lenghs chosen by hese hree crieria were raher small considering he fac ha he daa is monhly (Table 9). Among hese crieria, LR es has he highes lag lengh. However, we chose o esimae he base model wih he lag lenghs seleced by he AIC. 15 In 15 The above selecion crieria penalize loss of degrees of freedom due o inclusion of an addiional variable agains improvemen in he explanaory power. The fac ha SBC penalizes more from degrees of freedom loss han AIC resuls in AIC suggesing more lag lenghs han SBC. In order o guard agains omied variable bias, his may end one o choose lag lenghs suggesed by AIC crieria. Bu noe ha here is a cos o ha. Lukepohl (1990) suggesed ha a very high lag lengh in VAR specificaion may resul in 10

12 order o check for he robusness of our findings, we conduced sensiiviy analysis wih lag lenghs seleced by oher crierion. Exogenous variables The farm secor is considered o be exogenous o he sysem and, herefore, farm prices are included as exogenous variables in he equaions. Our iniial analysis shows ha in mos cases wholesale and reail prices do no Granger cause farm prices. Like many oher agriculural commodiies, coffee bean prices end o be more sensiive o supply han demand facors. One of he major influences on prices is he weaher condiions ha prevail when crops are growing. According o one esimae, he global demand for coffee is relaively sable, growing by around 10 per cen per annum during So if here is a feedback effec from consumer demand for he final produc o he upsream sages of producion ha feedback will be capured in he wholesale secor hrough changes in he prices which, in urn, will influence reail prices. The firs difference of he counry specific consumer price index was included for all iems (CPI) in he reail equaion of he models. Table 3 shows he raios of whole sale o reail prices of he sixeen counries sudied. These figures indicae ha a subsanial porion of he cos of processed coffee accrues from sources oher han coffee beans. Among ohers, hese coss include manufacuring, disribuion, soring and adverising coss. The problem lies in he fac ha coffee manufacuring, from roased coffee o differen varieies available in supermarke shelves, may ake place in differen pars of he world. Therefore, he CPI of he imporing counry is unlikely o capure he variaion of all he coss. However, i can be expeced ha CPI can provide a crude measure of he coss (such as manufacuring, disribuion, adverising) incurred in ha paricular imporing counry. Pelzman menions ha he CPI and he Producer Price Index (PPI) summarize he impac of economy wide nominal demand and cos changes. Considering he impracicabiliy of searching and including he lis of conrol variables for each of he hundreds of markes he analyzed, Pelzman included curren and lagged changes of he log of CPI and PPI in his equaions. In our case, he counry specific CPI appears o be a good measure of cos and demand. Figure 2 shows he nominal monhly reail prices for he period beween 1976:01 and 2004:12 and figure 4 shows he real prices for each of hese counries obained by deflaing nominal prices by he respecive CPIs. The heories of spaial compeiion sugges ha variaion of prices of similar goods/close subsiues should have similar paerns. Alhough he nominal prices of each of he counries have similariies, paricularly if we look a he peaks and he roughs, he similariies become even more apparen when we deflae he nominal prices by he respecive CPIs. This indicaes ha a subsanial porion of he price is deermined by real facors, such as cos and demand shifers, of each of he counries raher han nominal facors. The correlaion coefficiens of real prices wih each oher are higher han hose of nominal prices (Tables 5 & 6). Seasonal Dummies In addiion o he above variables, we included seasonal dummies in each of he equaions. However, our preliminary analysis shows ha he seasonal dummies, monhly or quarerly, do no have any significan impac on eiher he wholesale or he reail equaions. Therefore, weighing agains he loss of degrees of freedom resuling from heir inclusion, we decided o drop he seasonal dummies alogeher from he base model. imprecise esimaes of he coefficiens. In our case, given ha we have monhly daa, he lag lengh is no large. 11

13 Coinegraion ess Coinegraion ess were conduced for he 16 counries in a hree-variable seing (wholesale, foreign and reail prices). Table 9 repors he λ race (r) and λ max (r) saisics for r=0, r 1 and r 1, and heir significance levels. The resuls based on λ race (r) saisics indicae ha he null of no coinegraion (r=0) can no be rejeced only in Japan. The null of no coinegraion can be rejeced for he oher counries. Similar conclusion can be drawn using he λ max (r) saisics. In erms of he number of coinegraing vecors, he λ race (r) poins o one vecor for Ausria, Denmark, France, Norway, Spain, UK and he USA. The equaions of he oher counries (excep Japan) have wo coinegraing vecors. Table 13 repors he coinegraing vecors and he adjusmen coefficiens, denoed β and α in a common noaion, for he eigh counries where Johansen s es indicaed he presence of coinegraing relaions. Inser able 9 here. Table 10 repors he adjusmen coefficiens, denoed by α in a common noaion, for he 15 counries where Johansen s es indicaed he presence of coinegraing relaions. The α coefficiens measure he degree o which he variable in quesion adjuss o he deviaion from long run equilibrium relaionship. The firs coinegraing equaion is normalized wih respec o reail he equaions and he second, where applicable, is normalized wih respec o he foreign price equaions. Therefore, for a reail equaion, which is our primary ineres, we would expec he speed of adjusmen corresponding o he firs coinegraing equaion o have a negaive sign. In able 10, he α 11 coefficiens refer o he speed of adjusmen corresponding o he firs coinegraing equaion. I shows ha he speed of adjusmen have he expeced sign in hireen counries. The signs are posiive in Ialy and he USA, however hey are no significan according o chi-square ess a 5% significance level. If he null hypohesis ha α equals zero can no be rejeced, hen his would indicae ha he disequilibrium in he long run relaionship does no feed back ono he associaed variable. If his is he case, we can exclude he error correcion erm from ha equaion. The α 11 coefficiens are found o be significan in 8 counries; Bellux, Finland, France, Neherlands, Porugal, Spain, Sweden and Swizerland. The magniudes of adjusmen are found o be quie high in Bellux, Finland, he Neherlands and Sweden. They are -12.7, -9.6, and per cen, respecively. For he sake of inerpreaion, for example, in he case of Bellux, he magniude of adjusmen of indicaes ha in each monh he reail price adjuss by 12.7 per cen of he disequilibrium in he long run relaionship. In France, he adjusmen is very low. The adjusmens are beween 4 and 5 per cen in Porugal, Spain and Swizerland. Inser able 10 here. V. The Model Resuls The cumulaive response funcions for he base models of he sixeen counries are shown in figures 6 and 7. The response funcions in figure 6 race he dynamic impac of reail prices o one per cen change in coss originaing from he wholesale secor. In his case he magniude of he response is inerpreed as he elasiciy of he reail price wih respec o wholesale inpu price The response funcions in figure 7 show he dynamic impac of he same reail prices o a one per cen change in he price of a large coffee producing counry. Here, he magniude of he response can be inerpreed as he elasiciy of he reail price wih respec o foreign price. Figure 6 shows ha in nine ou of sixeen cases prices respond o upsream cos changes and, in seven cases, prices 12

14 exhibi complee rigidiy. One he oher hand, figure 7 shows ha prices of all he counries respond, almos in a similar way, o a foreign price shock. INSERT figure 6 here. INSERT figure 7 here. In his analysis reail prices were found o respond o inpu cos changes in nine ou he sixeen counries. These counries included Belgium, Denmark, France, Germany, he Neherlands, Norway, Sweden, UK and he USA. We will label hese counries responding counries in he following discussion. A consisen finding across responding counries was found wih respec o periods aken for price adjusmen afer he cos shock was iniiaed. In seven ou of he nine responding counries, price adjusmen was found o ake place over around wo quarers. In he case of Germany, he adjusmen process was found o be relaively fas, wih an adjusmen lags of less han one quarer. On he oher hand, in France he adjusmen process is relaively slow for he reail price, coninuing for around five quarers. The magniude of reail price adjusmen was found o vary beween 0.15 and 0 per cen. As menioned above, he magniude can be inerpreed as he elasiciy of he reail price wih respec o wholesale cos shocks. The magniude was found o be he highes in France, he Neherlands, Norway and he USA, a around 0 per cen. In oher words, in hese hree counries, a one per cen change in he cos resuls in around 0 per cen change in he price. On he oher hand, he smalles magniudes were found in Germany and UK, a only 0.15 and 0.20 per cen, respecively. In he oher hree responding counries, Belgium, Denmark and Sweden, prices were found o adjus by 0.25, 0.30 and 0.35 per cen, respecively. I may be of ineres o compare he proporion of cos change o he reail price and he magniude of he price change. If hey are equal, i indicaes ha he cos changes have passed hrough o he consumers hrough a proporional increase in reail prices. Table 18 shows ha he proporion of coss and he magniude of price change are similar in Belgium, Denmark, France and he Neherlands. Whereas in he oher responding counries such as Germany, Norway, Sweden, UK and he USA, cos and reail prices changes are no similar. However, i should be menioned he proporions have decreased consisenly for all counries over he sample period, alhough he rae a which hey have declined varies across he counries. When he confidence inervals of he responses include zero hen a counry is said o exhibi complee price rigidiy in response o a cos shock. This was found o be he case for seven counries including Ausria, Finland, Ialy, Japan, Porugal, Spain and Swizerland. Therefore, we conclude ha in hese seven counries prices do no respond o cos shocks. 16 This finding implies ha he firms locaed downsream in he disribuional channel absorb all of he wholesale cos changes. In order o assess how he shocks of he large foreign markes pass hrough o he domesic marke, we included foreign prices in he models. An example of a foreign shock is an inflaionary pressure in a large coffee marke and we would like o race he dynamic impac of price changes in he large foreign marke on he domesic marke. As discussed before, he foreign price is obained from aking he weighed average of he US and German reail prices. The cumulaive response funcion in he figure 7 shows he dynamic response of he domesic reail marke o one per cen change in prices in he foreign marke. In he models for he US and Germany, he equaions for foreign price refer o German and US prices, respecively. 16 Dua e al. (2002) inerpreed he price as rigid when he confidence inerval included zero. 13

15 Figure 7 shows ha he impacs of a foreign marke shock o domesic prices across counries are quie similar. The average magniude of he immediae impac is around 0.28 per cen in response o a one per cen shock (sandard deviaion is 5). Afer he immediae impac, for mos counries, ransmission occurs wihin a period of wo quarers. I should be noed ha similar response period was found for he responding counries wih respec o he cos shocks. The average magniude of adjusmen a he end of wo quarers is 1 wih a sandard deviaion of The similariy in he impac, boh in erms of immediae impac and he adjusmen process over ime, is evidence of how well he markes are inegraed. In order o check for he robusness of he resuls, we esimaed he models for each counry in he following ways: (1) various sample periods; (2) differen selecions of lag lenghs; and (3) exclusion of some exogenous variable(s). In oal he model was esimaed/specified in 13 differen ways. The impulse response funcions from he esimaed models are provided in appendix figures 1-8. Each of hese figures (16 of hem) is for one counry and displays hireen differen impulse responses for ha counry under differen model specificaions. The op panel shows he impulse responses for cos shocks and he boom panel shows he impulse responses foreign shock. The diagram (a) in each figure is he response funcion obained from he base model, reproduced here from figures 6 and 7 for he purpose of comparison. The response funcions of diagrams (b)-(g) relae o differen sample periods. The diagrams (h)-(j) are he response funcions esimaed under differen selecions of lag lenghs. The las se of diagrams (k)-(n) are he impulse response funcions from he models esimaed o check for he significance of exogenous variables. The sample period for he base model is 1976:1-2004:12, he periods for which he models have been esimaed are 1981:1-2004:12, 1986:1-2004:12, 1989:7-2004:12, 1976:1-1989:6, 1976:1-1994:12 and 1976:1-1999:12. The diagram (b) depics he impulse response for he sample period 1981:1-2004:12, hus excluding he firs five years from he sample period of he base model. The esimaes and resuling impulse response appear o be similar o he base model. Noe ha he firs hree ess exclude he observaions from he beginning of he sample period; 5 years, 10 years and around 14 years, respecively. A comparison of he resuls wih he base model and oher sample period models indicae ha he response of reail prices o cos changes have no changed over he years. The nex hree diagrams, (e)-(g), exclude observaions from he end of he sample periods. The similariy of he resuls for boh he cos and foreign shocks reinforces he fac ha he model resuls are insensiive o he selecion of sample periods. In order o check wheher he model resuls are sensiive o he number of lags included, we esimaed he models for each counry wih differen lag specificaions. Table 17 shows he number of lag lenghs chosen by differen selecion crierion. The base models were esimaed wih he lag lenghs seleced by he AIC crieria. In he sensiiviy analysis, he models were esimaed for he whole sample period wih he number of lag lenghs seleced by he BIC crieria and he LR es. 17 The general conclusion is ha, wih he excepion of few cases, changes in he lag lenghs do no lead o a qualiaively differen response in reail prices. Wih respec o impac of cos shocks, if we compare he base models wih he models esimaed using BIC crierion, hen in four counries, Denmark, France, Neherlands and Norway, here are some (minor/major) changes in he magniude of he response. Comparison of he base models wih he corresponding LR models show ha in he case of Finland and he Neherlands, he 17 Noe ha he number of lag lenghs seleced by he HQ and FPE are in mos cases he same as hose of BIC and AIC. 14

16 responses are quie differen. Using he LR crierion he impulse responses for Finland and he Neherlands show ha price does respond o cos changes, alhough in he base case prices were unresponsive. In he case of he Neherlands, he response becomes more persisen; wih adjusmen occurring over five quarers and he magniude increasing form o 0.75 per cen. The response o cos shock beween he AIC and LR models appear o be similar for oher counries. The responses o foreign shock using AIC and BIC models are similar. The only difference can be seen in he case of France, where he BIC model where prices are found o be unresponsive unlike ha of AIC. However, for some counries, LR produces differen resuls han he base models. The responses seem o become more persisen and reach higher magniudes for Bellux, Ialy, Porugal and Swizerland. For a number of oher counries, including Finland, Japan, Neherlands, Spain, UK and USA, he confidence inervals of he responses become very large. The esimaed sandard deviaions from he coefficiens of he addiional lags are very large, leading o wide confidence inervals. The las se of sensiiviy analysis involves exogenous variables. The exogenous variables included in he models are farm prices, consumer price indices and he average of he reail prices of he sixeen counries. Alhough individually each of hese variables is no significan in all he counries, hey are found o be joinly significan. This is eviden in diagram (m), figures When all hree exogenous variables are excluded from he model impulse responses appear o be differen from he corresponding base models. In mos cases, he magniudes of he adjusmen appear o be higher relaive o he base model where all he exogenous variables are included. This indicaes ha he exogenous variables explain par of he variaion of he domesic reail prices. While reail prices of seven counries exhibi complee rigidiy, nine counries respond o cos shocks. There exiss a srong evidence ha he process of adjusmen coninues for wo quarers, alhough he magniude of adjusmen varies from beween 0.15 and 0 per cen. Wih respec o foreign shocks, he reail prices in all 16 counries are found o be responsive. There are similariies in he way prices of differen counries respond o foreign shocks; ypically, here is an immediae impac wih price, adjusmen coninuing for anoher wo quarers. The average magniude of price adjusmen a he end of he second quarer is approximaely per cen. INSERT able (summary able for impulse response) here. The evidence suggess ha, in general, reail prices are more responsive o foreign shocks han o cos shocks. No only do more counries respond o foreign shocks when compared o cos shocks, bu also he average magniude of adjusmen is higher for foreign shocks han cos shocks. A sriking similariy in he impac of he wo kinds of shocks is in he period i akes o adjuss afer he shock is iniiaed, wih price adjusmen occurring wihin wo quarers afer he shock is imposed. These findings are generally found o be robus in erms of variaion in sample sizes and number of lags of he variables. 6. Discussions Price rigidiy is esimaed for he reail secors of he coffee markes in sixeen counries wih respec o shocks originaing from he wholesale secor and he large coffee consuming counries. We esimaed hree aspecs of price rigidiy: (1) he speed of price change; (2) he magniude of price change; and (3) he ime pah of adjusmen. The following discussion will be limied o he firs wo aspecs. However, in erm of policy implicaions, all hree aspecs are imporan. 15

17 The speed a which he reail prices adjus o wholesale cos shocks depends on he acions of he agens locaed a he downsream sages of he wholesale marke. These agens esablish he link beween he wo markes. Examples of few of he agens include exporers, raders, manufacurers and ransporers. 18 The acions of each of hese agens are governed by many facors such as he marke srucure, echnology and he naure hey are linked wih each oher. For example, i can be expeced ha if here are few agens and each of hem operaes under compeiive environmen, cos changes will pass on o he reail secor relaively quickly. In his analysis, i was found ha, among he responding counries, generally i akes approximaely wo quarers (six monhs) for a wholesale cos shock o pass on o he reail secor. For comparison purposes, i should be menioned ha Dua e al. (2002) found a lag of four o six weeks (approximaely one and a half monh) in he response of reail prices of orange juice concenraes in he US marke o wholesale cos shocks. Blinder e al. (1994) found an average lag of hree o four monhs in heir firm survey sudy on wide range of producs in he US. Fabiani e al. (2005) repors ha an adjusmen process of one quarer in macro models for France, Luxembourg, Ausria and Porugal and of wo or more quarers for Spain seems o be jusified on he grounds of hese findings (p.29). The above comparison indicaes ha in general price rigidiy in he coffee markes are higher han in many oher markes of he sixeen counries sudied. There are a number of reasons for he delay in ransmission of cos changes o he nex sage of he disribuional channel. Fabiani e al. (2005) found ha in he euro area he mos imporan explanaions for rigid prices are provided by he cusomer marke models (nominal and implici conracs). The nominal conrac heory says ha he agens may prefer o esablish explici conracs ha se prices in advance in order o esablish a long erm relaionship (Fisher, 1977; Taylor, 1980). Even if here is no explici conrac, a implici conrac may deer he agens o change he price (Hicks, 1974; Okun, 1981; Schulze, 1985; Nakamura and Seinsson, 2006). Blinder e al. (1998) found ha wo of he mos imporan reasons for he exisence of price rigidiy in he US are explained by coordinaion failure and firms pursuance of cos based pricing. Coordinaion failure occurs because each firm unsure of is compeiors acion wais for he compeior o ake he lead. The idea of cos based pricing is ha he firms change prices only when coss of inpu changes. Blanchard (1983) showed ha in a chain of producion, he accumulaion of lags in he arrival of informaion in each sage may cause subsanial delays in responding o cos changes. Anoher popular explanaion is provided by differen varians of menu cos models. Menu cos models say ha here are coss associaed wih changing prices and if ha cos is higher han he benefi accrued from i, hen he firm will deer from changing he price (Barro, 1972; Danzinger, 1983; Sheshinski and Weis, 1977, 1983; and Golosov and Lucus, 2006). Alernaively, uncerainy abou he naure of he shock, wheher i is ransiory or permanen, may also be responsible for delayed response. The firm survey sudy shows ha wihin he same indusry here were variaions in he way he firms responded regarding he relevance of a paricular price rigidiy heory (Blinder e al., 1998; Apel e al., 2005; Hall e al., 1997; Fabiani e al., 2005). Thus, a any poin of ime a cerain number of firms will change heir prices while prices will remain fixed for some oher firms. This 18 Coffee beans have o pass hrough he hand of a large number of middlemen, oher han big players like manufacurers and imporers, before hey reach supermarke shelves (The Fairrade Foundaion, 2002). 16

18 causes he impac of shocks on he whole marke o be disribued over ime. The overall rankings found from hese surveys in he euro area, Sweden, UK and he US are provided in he appendix able 2. The sudy shows ha he magniude of he price adjusmen varies among he counries. Among he responding counries, he maximum and minimum magniudes are and 0.15 per cen, respecively, and four counries adjus in he viciniy of per cen. A number of reasons can be pu forward for hese variaions. The marke demand of he final good may vary among he counries resuling in a varying abiliy of he firms or reailers o pass on he cos change o he consumers; he higher he price elasiciy of demand, he less is he change in he price. Alernaively, firms in he shor run may make subsiuion in he use of inpus o relaively cheaper inpus of producion. Carlon (1989) argues ha marke may clear along dimensions oher han price such as delivery lags, produc qualiy, selling effor and level of services. For example, insead of raising price, firms may respond by reducing selling effors, he qualiy of he produc and service. These measures conribue owards absorbing increased coss. Carlon suggess ha an opimal mixing of all insrumens clear he marke. In he survey sudy of Blinder e al., he idea of non-price compeiion ranked hird among he welve compeing heories. However, i ranked poorly (8 h ) in he Fabiani e al. euro area sudy. In addiion, firms may choose o adjus invenory raher han prices (Reagan, 1982; Blinder, 1982). I implies ha wih a posiive cos shock, prices would rise less han i would be if he firms do no have invenories o vary. Anoher reason for he exisence of price rigidiy may be due he belief ha cerain prices have psychological significance o he consumers and firms may no wan o exceed ha barrier. However, Kashyap (1995) found only limied evidence supporing his heory. Among he survey sudies, he idea was appreciaed in UK (Hall e al, 1997, 2000), bu ranked poorly for he US and Sweden (Blinder e al., 1998; Apel e al. 2005). Marke srucure can play an imporan role in deermining firms reacion o cos changes. Dornbusch (1987) showed ha he impac of cos shock is larger he more compeiive he marke. He formulaed an oligopolisic marke srucure where he degree of compeiion was measured by he markup of price over marginal cos and he source of cos shock was exogenous movemens of exchange rae. He showed ha []he marke srucure impor share and concenraion are he key parameers ha explain he oucome [he magniude of price change] (p.97). The indusrial organizaion lieraure ells us ha marke concenraion is he hallmark of adminisered prices. Viqueira (1991) using individual price daa of a wide range of producs in Mexico found ha he more concenraed indusries end o have lower frequency of price adjusmen. Viqueira s finding was found o be robus o differen measures of marke concenraion. Hannan and Berger (1991) examined he deposi raes of differen banks and found ha banks in more concenraed marke exhibi greaer price rigidiy. As menioned above, Dua e al. found ha he reail prices of orange juice concenraes in he US are flexible, adjusing relaively quickly, ofen wihin 4-6 weeks. They repor ha he markes for orange juice concenraes are highly compeiive boh a he manufacuring and reail levels. McCorrison and Sheldon (1996) showed ha boh he number of verical sages and he degree of imperfec compeiion a any specific sage can negaively affec he degree of ariff pass-hrough. In he Fabiani e al. sudy, one of he key findings which appeared as a sylized fac is ha [m]ark-up (consan or variable) pricing is he dominan price seing pracice adoped by firms 17

19 in euro area. The lower he level of compeiion, he frequenly used his mehod is (p.15). The sudy found ha in each counry mark-up pricing dominaes o any oher pricing rules and, as a whole of he euro area, 54 per cen of he firms follow mark-up pricing rule 19. The variable markup rule is found o have aken he lead in he counries where quesionnaires allowed hem o make a disincion beween consan and variable mark-up pricing rules. The second imporan pricing rule is found o be price seing according o compeiors price (27 per cen). Anoher relaed finding consisen across counries is he exisence of negaive relaionship beween he share of firms following he mark-up rule and he degree of compeiion. The survey resuls promped Fabiani e al. o sugges ha models wih monopolisic compeiion, like New Keynesian models, may be a beer descripion for mos goods and service markes han hose ha assume perfec compeiion (p.5). They argue ha mark-up, which under general condiions is a feaure of imperfecly compeiive marke srucure, provides firms some room for no adjusing he price in variaion o coss. Therefore, price rigidiies can occur unlike in a perfecly compeiive marke where price equals marginal cos. 20 Roemberg and Woodford (1994) provide an overview of differen models wih exogenously and endogenously deermined ime-varying mark-ups and heir implicaions on macroeconomics. In he case of he US marke, Blinder e al. found ha he idea of coordinaion failure offered he mos aracive explanaions for price rigidiy for he majoriy of he firms. This implies ha mos of he US firms se price according o compeiors price. This suggess ha price rigidiy may be expeced o be lower in US han in he euro area. Alvarez e al. (2005) repors ha price duraions in he euro area is significanly longer han in he US. The coffee markes in he counries sudies are found o be highly concenraed. Table 13 shows he concenraion raios CR1 and CR4, where CR1 and CR4 refer o he share of he marke capured by he larges firm (one firm) and he larges four firms, respecively. The CR4 was found o be as high as 90.7 per cen in Porugal in 2004 wih he larges firm capuring 68 per cen of he marke. The lowes concenraion raios were found in US, he CR1 and CR2 being 18.8 per cen and 42.5 per cen, respecively. According o Scherer (1990), a CR4 of above 40 per cen is an indicaion of firms being enjoying marke power. The CR4 in all he counries are above 40 per cen indicaing ha coffee manufacurers or roasers are enjoying moderae o exreme degrees of marke power in hose counries 21. Inser he able on concenraion raio here 7. Conclusions Much of he empirical sudies of price rigidiy examined variaions wihin or across produc groups in a single locaion. In recen years, he lieraure has seen a number of cross counry sudies wihin he euro area covering a wide range of producs. This sudy, on he oher hand, is specific o only one produc (coffee). The produc is subjec o high degree of inernaional rade and, herefore, he markes among he counries are expeced o be well inegraed. The sudy used dynamic economeric models and examined price rigidiy by analyzing he ransmission of shocks o he reail secor. This way of examining price rigidiy, among differen alernaive 19 The resuls of Ausria and Luxembourg were no repored. 20 Alhough price rigidiy is more likely in less compeiive markes, in he exreme case of very high cos of adjusmen, even highly compeiive firms may fail o adjus o small inpu cos changes. 21 The coffee markes for he whole world in is differen disribuional channels are very concenraed. According o The Fairrade Foundaion (2002), he indusry is dominaed by four mulinaional companies: Procor and Gamble, Philip Morris, Sara Lee and Nesle accoun for 40 per cen of he worldwide reail sales. Similarly, he 40 per cen of he expor of coffee beans is conrolled by six mulinaional firms. 18

20 measures, is paricularly useful for macroeconomic purposes. A cross counry comparison was drawn among 16 counries including he counries in he euro area and he US. The finding of his sudy can be generalized o some oher producs, such as producs having similar marke feaures. Almos half of he counries do no respond o inpu cos shocks. A finding consisen among he responding counries was ha he adjusmen ook place over around wo quarers, alhough he magniude of adjusmen varies. Wih respec o foreign shocks, here is an immediae impac. Afer he immediae impac, for mos counries, he ransmission occurs wihin a period of wo quarers. Therefore, in boh cases, a general conclusion can be drawn ha he price adjusmen occurs wihin wo quarers afer he shock is imposed. However, he evidence suggess ha, in general, reail prices are more responsive o foreign shocks han o cos shocks. No only do more counries respond o foreign shocks when compared o cos shocks, bu also he average magniude of adjusmen is higher for foreign shocks han cos shocks. 19

21 Figure 1: Composie index and is four componen indices for wholesale prices Figure 2: Nominal reail prices of 16 differen counries (in US cens) 20

22 Figure 3: Raios of wholesale o reail prices in 16 counries Figure 4: Real reail prices of 16 differen counries 21

23 Figure 5: Accumulaed impulse response of reail prices o one per cen cos shock Ausria Belgium Denmark Finland Accumulaed Impulse France Neherlands Germany Norway Ialy Porugal Japan Spain Sweden Swizerland UK USA Monhs Monhs Monhs Monhs Figure 6: Accumulaed impulse response of reail prices o one per cen shock from a large coffee consuming counry Ausria Belgium Denmark Finland Accumulaed Impulse France Neherlands Germany Norway Ialy Porugal Japan Spain Sweden Swizerland UK USA Monhs Monhs Monhs Monhs 22

24 Table 1: Raio of ICO wholesale composie index o nominal reail prices Period I: 1976:1-1985:12 Period II: 1986:1-1995:12 Period III: 1996:1-2004:12 Pre-ICA: 1976:1-1989:6 Pos-ICA: 1989:7 2004:12 Whole sample 1976:1-2004:12 Ausria 0.44 (0.12) 0.22 (0.10) 0.21 (6) 0.40 (0.13) 0.20 (7) 0.29 (0.14) Bellux 0.45 (9) 0.28 (9) 0.27 (5) 0.42 (0.10) 0.26 (7) 0.35 (0.12) Denmark 0.38 (9) 0.24 (8) 0.19 (5) 0.36 (9) 0.20 (5) 0.27 (0.11) Finland 8 (7) 0.33 (0.11) 0.29 (6) 0.49 (0.13) 0.30 (9) 0.36 (0.13) France 0.45 (9) 0.31 (0.12) 0.30 (9) 0.42 (0.10) 0.30 (0.12) 0.35 (0.12) Germany 0.33 (7) 0.21 (8) 0.19 (5) 0.31 (8) 0.19 (6) 0.25 (9) Ialy 0.44 (0.10) 0.21 (0.10) 0.15 (5) 0.40 (0.12) 0.16 (7) 0.27 (0.15) Japan 0.19 (1) 9 (4) 7 (2) 0.16 (4) 7 (2) 0.10 (5) Neherlands 0.49 (9) 0.30 (0.10) 0.24 (7) 0.46 (0.10) 0.25 (8) 0.35 (0.14) Norway 0.45 (9) 0.28 (8) 0.22 (7) 0.42 (0.10) 0.23 (8) 0.32 (0.13) Porugal 0.36 (6) 0.22 (9) 0.17 (5) 0.34 (7) 0.17 (6) 0.24 (0.10) Spain 0.47 (9) 0.29 (0.13) 0.24 (7) 0.44 (0.11) 0.25 (0.10) 0.33 (0.14) Sweden 0.43 (5) 0.26 (7) 0.22 (5) 0.39 (8) 0.23 (6) 0.30 (0.10) Swizerland 0.42 (0.11) 0.21 (8) 0.16 (6) 0.38 (0.12) 0.17 (7) 0.27 (0.14) UK 0.20 (9) 0.10 (4) 6 (2) 0.18 (8) 7 (2) 0.12 (8) USA 5 (0.12) 0.32 (0.10) 0.23 (7) 1 (0.12) 0.25 (7) 0.37 (0.17) Figures in he parenhesis are he esimaed sandard errors. Table 2: Augmened Dickey Fuller es on he log of nominal price series T K a0 a1 γ1=1+ a1 Uni roo Farm Price: Brazilian Arabica (3.324) -46 (-3.344) 0.954* No Wholesale: Composie Index (2.127) -19 (-2.149) Yes Reail Price: Ausria (2.967) -40 (-2.949) 0.960* No Bellux (3.359) -33 (-3.352) 0.967* No Denmark (3.477) -33 (-3.370) 0.967** No Finland (2.062) -22 (-2.058) Yes France (2.903) -26 (-2.892) 0.974* No Germany (2.732) -25 (-2.725) Yes Ialy (3.257) -18 (-3.208) 0.982* No Japan (2.036) -19 (-2.022) Yes Neherlands (4.489) -66 (-4.479) 0.934** No Norway (4.394) -56 (-4.384) 0.944** No Porugal (2.383) -24 (-2.382) Yes Spain (3.316) -37 (-3.327) 0.963* No Sweden (3.576) -42 (-3.567) 0.958** No Swizerland (3.168) -31 (-3.133) 0.969* No UK (3.455) -22 (-3.396) 0.978* No USA (4.534) -44 (-4.524) 0.956** No Noe: * and ** denoe ha he esimaed characerisics roos are significanly differen from uniy a 5 and 1 per cen, respecively. 23

25 Table 3: Perron s es for uni roo wih srucural change on nominal farm and whole sale prices of coffee beans and nominal reail prices of coffee T λ K a μ Farm Price: Brazilian Arabica Wholesale: Composie (3.27) (3.27) Index Reail Price: Ausria (3.33) Bellux (3.25) Denmark (3.49) Finland (2.06) France (3.25) Germany (2.95) Ialy (3.91) Japan (2.24) Neherlands (3.99) Norway (4.44) Porugal (2.75) 0 1 μ2 a2 a1 γ1=1+ a1 1 (4) -0 (-0.17) 2 (1.42) 0 (0.44) 1 (0.85) -0 (-0.39) -1 (-1.21) 1 (0.86) 1 (1.54) 2 (1.49) 1 (0.66) -1 (-0.69) 1 (1.16) 0 (-3.56) 8 (-3.85) 4 (0.62) 5 (1.50) 4 (1.43) -1 (-0.30) 7 (2.12) 2 (3) 4 (1.73) 1 (0.20) 4 (0.95) 3 (0.84) 4 (1.44) 00 (-0.63) -0 (-1.57) -0 (-1.84) -0 (-6) -0 (-1.22) -0 (-5) -0 (0) -0 (-1.61) -0 (-0.15) -0 (-1.79) -0 (-0.68) 0 (8) -0 (5) -5 (-3.36) -4 (-3.25) -5 (-3.24) -3 (-3.18) -3 (-3.44) -2 (-2.07) -4 (-3.33) -3 (-2.93) -3 (-3.83) -2 (-2.06) -7 (-3.99) -6 (-4.44) -3 (-2.79) Uni roo 0.95 Yes Yes Yes 0.97 Yes 0.97 Yes 0.98 Yes 0.96 Yes Yes 0.97* No 0.98 Yes 0.93* No 0.94** No 0.97 Yes Spain (2.40) 1 (0.72) 5 (3) -0 (-0.69) -3 (-2.41) 0.97 Yes Sweden (3.49) -0 (-7) 6 (1.46) -0 (-0.18) -5 (-3.51) 0.95 Yes Swizerland (3.55) UK (-1.61) USA (4.58) 1 (0.72) 0 (0.31) 0 (0) 5 (1.50) 6 (1.89) 2 (2) 0 (0.35) 0 (1.51) 0 (6) -4 (-3.50) -4 (-4.10) -5 (-4.53) 0.96 Yes 0.96* No 0.95** No Noe: * and ** denoe ha he esimaed characerisics roos are significanly differen from uniy a 5 and 1 per cen, respecively. The criical values are obained from Perron (1989). + indicaes he process having uni roo wih one ime change a he poin of break. 24

26 Table 4: Coinegraion es resuls Counry r=0 r 1 r 2 λrace λmax λrace λmax λrace Ausria 37.31** 28.18** Bellux 78.58** 53.72** 24.86** 16.39* 8.47** Denmark 46.63** 38.48** Finland 45.20** 27.36** 17.84* 14.74* 3.09 France 36.52** 21.83* * Germany 51.79** 33.76** 18.02* * Ialy 64.88** 49.25** 15.63* * Japan Neherlands 59.99** 34.83** 25.16** 20.39** 4.77* Norway 39.88** 24.99* * Porugal 39.01** * 14.35* 4.19* Spain 39.37** 24.93* * Sweden 40.46** 23.35* 17.11* * Swizerland 59.52** 40.33** 19.18* ** UK 47.13** 34.84** * USA 44.76** 25.20* 19.56* 15.26* 4.31* λmax ** * 5.74* 3.88* * 4.08* 4.19* 4.88* 5.19* 8.00** 4.76* 4.31* ** significan a 1% level; * significan a 5% level; significance level is deermined using MacKinnon- Haug-Michelis (1999) disribuion. Table 5: The coefficiens of he speed of adjusmen Counry Reail equaion Foreign equaion Wholesale equaion Ausria Bellux Denmark Finland France Germany Ialy Japan Neherlands Norway Porugal Spain Sweden Swizerland UK USA 1 s coin. (α ) * * -03* n.a * * -49* * -44* nd (α coin. 12 ) n.a 39 n.a 35* n.a n.a 39* n.a 09 n.a 49* 23* n.a n.a 1 s (α coin 21 ) 29* 34 67* 15-06* * n.a 45-17* * 0.102* 2 nd (α coin. 22 ) n.a -66* n.a -32 n.a -37* * n.a -55* n.a -33 n.a * n.a n.a 1 s (α coin 31 ) * 0.124* -11* 0.119* 33 n.a * 0.103* 0.104* * 24* nd (α coin 32) n.a -75* n.a n.a -65* -40 n.a -23 n.a -90* n.a * n.a n.a * significan a 5% per cen level. chi-square es is used o deermine he significance level n.a refers o no applicable 25

27 Table 6: Lenghs of lag seleced by differen selecion crieria Counry AIC BIC HQ LR FPE Ausria Bellux Denmark Finland France Germany Ialy Japan Neherlands Norway Porugal Spain Sweden Swizerland UK USA Noe: AIC: Akaike Informaion Crieria; BIC: Bayesian Informaion Crieria; HQ: Hannan-Quinn Crieria; LR: Likelihood Raio Tes; FPE: Final Predicion Error Table 7: Concenraion raios in coffee markes, 2001 and 2004 One-firm conc. (CR1) Four-firms conc.(cr4) Counry Ausria Belgium Denmark Finland France Germany Ialy Japan Neherlands Norway Porugal Spain Sweden Swizerland UK USA Source: Auhors own calculaion from he daa obained from he Global Marke Informaion Daabase published by Euromonior Plc. 26

28 References: Alvarez, L. J., E. Dhyne, M. M. Hoeberichs, C. Kwapil, H. L. Bihan, P. Lunnemann, F. Marins, R. Sabbaini, H. Sahl, P. Vermeulen and J. Vilmunen (2005), Sicky Prices in The Euro Area: A Summary of Micro Evidence, European Cenral Bank, Working Paper No Amiraul, D., C. Kwan and G. Wilkinson, A survey of he price seing behaviour of Canadian companies, Bank of Canada Review, Winer , pp Apel, M., R. Friberg and K. Hallsen (2005), Micro Foundaions of Macroeconomic Price Adjusmen: Survey Evidence from Swedish firms, Journal of Money, Credi and Banking, 37, pp Baharad, E. and B. Eden (2004), Price Rigidiy and Price Dispersion: Evidence from Micro Daa, Review of Economic Dynamics, 7, Barro, R. J. (1972), A Theory of Monopolisic Price Adjusmen, Review of Economic Sudies, 39, pp Bils, M. and P. Klenow (2002), Some Evidence on he Imporance of Sicky Prices, Journal of Poliical Economy, 112, pp Blanchard, O. J.(1983), Price asynchronizaion and price level ineria, In Dornbusch, R. and M. Simonsen (eds.), Inflaion Deb and Indexaion, The MIT Press, Cambridge, MA, pp Blinder, A.S. (1991), Why are prices sicky? Preliminary resuls from an inerview sudy, American Economic Review, 81, pp Blinder, A. S. (1994), On Sicky Prices: Academic Theories mee he Real World, InL Mankiw, N.G. (Ed.), Moenary Policy, Universiy of Chicago Press and Naional Bureau of Economic Research, Chicago, IL, pp Blinder, A.S., E. Canei, D.E. Lebow and J.B. Rudd (1998), Asking abou prices: a new approach o undersanding price sickiness, Russell Sage Foundaion, New York. Campbell, J. and B. Eden (2005), Rigid prices: Evidence from US scanner daa, NBER. Carlon, D.W. (1986), The Rigidiy of Prices, American Economic Review, 76, pp Carlon, D. W. (1989), The Theory and Facs Abou How Markes Clear: Is Indusrial Organizaion Valuable for Undersanding Macroeconomics?, In R. Schmalensee and R. D. Willig (Ed) Handbook of Indusrial Organizaion, Amserdam, Norh Holland. Carlon, D. and J. M. Perloff (1994), Modern Indusrial Organizaion, Harper Collins, New York. Checchei, S.G. (1986), The Frequency of Price Adjusmen: A Sudy of Newssand Prices of Magazines, Journal of Economerics, 31, pp Dahlby, B (1992), Price Adjusmen in an Auomobile Insurance Marke: A Tes of he Sheshinski-Weiss Model, Canadian Journal of Economics, 25, pp

29 Danziger, L. (1987), Inflaion, Fixed Cos of Price Adjusmen and Measuremen of Relaive- Price Variabiliy: Theory and Evidence, American Economic Review, 77, pp Dhyne, E., L. Alvarez, H. Le Bihan, G. Veronese, D. Dias, J. Hoffman, N. Jonker, P. Lunnemann, F. Rulmer and J. Vilmunen (2005), Price seing in he euro area: some sylized facs from individual consumer daa, European Cenral Bank, Working Paper No Dornbusch, R. (1987), Exchange Raes and Prices, American Economic Review, 77, pp Dua, S, M. Bergen and D. Levy (2002), Price Flexibiliy in Channels of Disribuion: Evidence from Scanner Daa, Journal of Economic dynamics & Conrol, 26, pp Dua, S., M. Bergen, D.Levy and R. Venable (1999), Menu Coss, Posed Prices and Muliproduc Reailers, Journal of Money, Credi and Banking, 31, pp Enders, W. (2004), Applied Economeric Time Series, Wiley Series in Probabiliy and Saisics, Danvers, MA. Fabiani, S., M. Druan, I. Hernando, C. Kwapil, B. Landau, C. Loupias, F. Marins, T.Maha, R. Sabbaini, H. Sahl and A Sokman (2004), The pricing behaviour of firms in he euro area: New survey evidence, European Cenral Bank, Working Paper Series, No Fairrade Foundaion (2002), Spilling he Beans on he Coffee Trade, UK.( Fisher, S. (1977), Long-erm Conracs, Raional Expecaions, and he Opimal Money Supply Rule, Journal of Poliical Economy, 85, pp Gali, J., D. Lopez-Salido and J. Valles (2003), Technology shocks and moneary policy: Assessing he Fed s Performance, Journal of Moneary Economics, 50, pp Golosov, M. and R. E. Lucus (2006), Menu Coss and Phillips Curves, Working Paper, MIT, MA. Gonzalo, J. (1989), Comparison of Five Mehods of Esimaing Long-run Equilibrium Relaions, Universiy of Calirfornia, San Diego, Working Paper No. 55. Gordon, R. (1990), Wha is New-Keynesian economics?, Journal of Economic Lieraure, 28, pp Granger, C. and P. Newbold (1974), Spurious Regressions in Economerics, Journal of Economerics, 2, pp Hall, S., M. Walsh and A. Yaes (1997), How do UK Companies Se Prices?, Bank of England, Working Paper No. 67. Hall, S., M. Walsh and A. Yaes (2000), Are UK companies prices sicky, Oxford Economics Papers, 52, pp Hamilon, J. (1994), Time Series Analysis, Princeon Universiy Press, Princeon. 28

30 Hannan, T and A. Berger (1991), The Rigidiy of Prices: Evidence from he Banking Indusry, 81, pp Herrmann, R., Moeser, A. and Weber, S.A. (2005), Price Rigidiy in he German Grocery Reailing Secor: Scanner Daa Evidence on Magniude and Causes, Journal of Agriculural & Food Indusrial Organizaion, 3, pp Hicks, J. (1974), The Crisis of Keynesian Economics, Basic Books, New York. Johansen, S. (1988), Saisical analysis of coinegraion vecors, Journal of Economic Dynamics and Conrol, 12, pp Johansen, S. (1991), Esimaion and hypohesis esing of coinegraing vecors in Gaussian vecor auoregressive models, Economerica, 59, pp Kashyap, A. (1995), Sicky Prices: New Evidence from Reail Caaloes, Quarerly Journal of Economics, 110, pp Lach, S & Tsiddon, D. (1992), The Behavior of Prices and Inflaion: An Empirical Analysis of Disaggregaes Price Daa, Journal of Poliical Economy, 100, pp Levy, D., S. Dua and M. Bergen (2002), Heerogeneiy in Price Rigidiy: Evidence from a Case Sudy Using Microlevel Daa, Journal of Money, Credi, and Banking, 34, pp Lukepohl, H. (1990), Asympoic Disribuion of Impulse Response Funcions and Forecas Error Variance Decomposiions of VAR Models, Review of Economics and Saisics, 72, pp Lukepohl, H. (2006), New Inroducion o Muliple Time Series Analysis, Springer, Berlin. Mankiw, G. (1985), Small Menu Coss and Large Business Cycles: A Macroeconomic Model of Monopoly, Quarerly Journal of Economics, 100, pp McCorrison, S. and I. M. Sheldon (1996), Trade Policy in Verically-Relaed Markes, Oxford Economic Papers, 48, pp Meha, A. and J. Chavas (2004), Responding o he coffee crisis: Wha can we learn from price dynamics, Deparmen of Agriculural and Applied Economics, Universiy of Wisconsin, Madison. Meyer, J. and S. von Cramon-Taubadel (2004), Asymmeric price ransmission : A survey, Deparmen of Agriculural Economics, Gohigen. Nakamura, E. and J. Seinsson (2006a), Five Facs Abou Prices: A Reevaluaion of Menu Cos Models, Harvard Universiy, MA. Nakamura, E. and J. Seinsson (2006b), Price Seing in Forward Looking Cusomer Markes, Harvard Universiy, MA. 29

31 Okun, A. (1982), Prices and Quaniies: A Macroeconomic Analysis, The Brookings Insiuion, Washingon D.C. Owen, A. and D. Trzepacz (2002), Menu Coss, Firm Sraegy and Price Rigidiy, 76, Pelzman, S. (2000), Prices Rise Faser han hey Fall, Journal of Poliical Economy, 108, pp Perron, P. (1989), The grea crash, he oil price shock, and he uni roo hypohesis, Economerica, 52, pp Reagan, P. (1982), Price and Invenory Behavior, Review of Economic Sudies, 49, pp Roemberg, J. (1983), Aggregae Consequences of Fixed Cos of Price Adjusmen. American Economic Review, 73, pp Roemberg, J. and M. Woodford (1994), Dynamic General Equilibrium Modelswih Imperfecly Compeiive Produc Markes, in T. Cooley (ed.), Froniers of Business Cycle Research, pp Scherer, F. M. (1990), Indusrial Marke Srucure and Economic Performance, 3 rd Ediion, Chicago: Rand McNally. Schulze, C. (1985), Microeconomic Efficiency and Nominal Wage Sickiness, American Economic Review, 75, Sheshinski, E and Y Weiss (1977), Inflaion and Coss of Price Adjusmen, Review of Economic Sudies, 44, pp Sheshinski, E and Y Weiss (1983), Opimum Pricing Policy under Sochasic Inflaion and Coss of Price Adjusmen, Review of Economic Sudies, 50, pp Taylor, J. B. (1980), Saggered Wage Seing in an Macro Model, American Economic Review, 69, pp Viqueira, A P. (1991), Marke Srucure and Price Flexibiliy, Journal of Developmen Economics, 36, pp Wynne, M. (1995), Sicky prices: Wha is he evidence?, Economic Review, Federal Reserve Bank of Dallas, 1 s Quarer, Dallas. Zbaracki, M.J., M Rison, D Levy, S. Dua and M Bergen (2004), Managerial and Cusomer Coss of Price Adjusmen: Direc Evidence from Indusrial Markes, Review of Economics and Saiisics,86,

32 1

33 Esimaing price rigidiy in coffee markes: A cross counry comparison Ph.D candidae: Iqbal Syed Supervisor: Prof. Kevin Fox UNSW

34 Why are empirical measures of price rigidiy imporan? Beer undersanding of moneary policy ransmission Beer judgemen of differen measures of underlying inflaion Limied number of empirical work Unforunaely, his program [he sandard programs of scienific research in economics] has been singularly unsuccessful in he area of wage-price sickiness. (Blinder e al. 1998)

35 Definiion Price rigidiy is [o]fen nohing more han ha prices adjus less rapidly han Walrasian markeclearing prices.(blinder, 1998) Price rigidiy is said o occur when prices do no vary in response o flucuaions in coss and demand. (Carlon and Perloff, 1994) A number of heories based on he firms opimizaion rule provide explanaions for he exisence of price rigidiy

36 Empirical measures of price rigidiy Frequency of price change in a given period (Kashyap, 1995; Nakamura & Seinsson, 2006) Probabiliy of price change due o cos or demand changes (Cecchei, 1986; Campbell & Eden, 2004) The number of periods price response lags behind he shock (Pelzman, 2000; Dua e al, 2002) Focus of his paper Vecor error correcion model

37 Counries Ausria Belgium & Luxembourg Denmark Finland France Germany Ialy Japan Neherlands Norway Porugal Spain Sweden Swizerland UK USA

38 Marke Inegraion and coffee markes Economic and commercial policies promoe marke inegraion Coffee has high radable inpus Marke demand similar Poins o he fac ha coffee marke should be inegraed Marke srucure plays an imporan role heerogeneiy in price rigidiy

39 Daa Sources: ICO & OECD Period: Monhly, 1976:1 2004:12 (348 obs.) Endogenous variables Wholesale coffee bean price (weighed average of differen varieies of coffee beans) 1 series Domesic reail prices of 16 counries 16 series Exogenous variables Brazilian variey farmgae coffee bean prices-1 series CPI All Iems - 16 series Monhly average exchange raes for home currency in erms of US currency 15 series

40 Wholesale prices

41 Nominal reail prices of 16 differen counries (in US cens)