Spatial Price Efficiency in Mozambique s Post-Reform Maize Markets. Emílio Tostão and B. Wade Brorsen

Transcription

1 Spaial Price Efficiency in Mozambique s Pos-Reform Maize Markes Emílio Tosão and B. Wade Brorsen Conac Auhor: B. Wade Brorsen Deparmen of Agriculural Economics Oklahoma Sae Universiy Sillwaer, OK brorsen@oksae.edu Emílio Tosão is a Ph.D. suden and B. Wade Brorsen is a Regens Professor and Jean and Pasy Neusad Chair in he Deparmen of Agriculural Economics a Oklahoma Sae Universiy. Parial funding from he Unied Saes Agency for Inernaional Developmen is graefully acknowledged. L:\Home\Brorsen\PAPERS\Oldpapers Feb 05\Mozambique\Mozambique_final1.doc

2 Spaial Price Efficiency in Mozambique s Pos-Reform Maize Markes Absrac Maize (Zea mays) shorages and maize price insabiliy coninue o occur in Mozambique in spie of policy reforms and grea capaciy for increased producion in norhern Mozambique. This sudy measured he efficiency of spaial maize price arbirage in Mozambique s pos-reform period. Spaial price inefficiency does no seem o explain why food shorages coninue o occur in Mozambique, suggesing non-price marke consrains such as poor infrasrucure and poor marke insiuions as more likely explanaions. Baulch s spaial efficiency es indicaed ha i is no profiable o ship maize from norhern surplus maize regions o souhern Mozambique. Records of inerregional maize shipmens show ha indeed grain is no being shipped from he norh o he souh. Marke liberalizaion may have improved spaial price efficiency, bu high ransfer coss sill limi rade and poenial benefis from freeing markes. Under hese circumsances, food shorages and price insabiliy are likely o coninue. JEL Classificaion Codes: Q13, F14 Key Words: Food securiy, Mozambique, spaial arbirage, swiching regression. 2

3 Inroducion Throughou he pas 15 years, Mozambique reformed food markes, shifing from a planned economy o a marke economy. The procuremen, disribuion, and processing of maize (Zea mays), he main food saple, ha had been exclusively underaken by subsidized governmen-owned companies is now carried ou by privae enrepreneurs. Maize prices are now deermined by marke forces. Ye, maize shorages and maize price insabiliy coninue o occur in Mozambique (Jayne e al., 1999). Limied rade is a possible explanaion of why food securiy coninues o be a problem in Mozambique in spie of policy reforms and grea capaciy for increased producion in norhern Mozambique (Jeje a al., 1998; The Governmen of Mozambique, 1998). When spaial rade is efficien, food shorages in defici regions are ransmied o surplus regions via prices, and arbirage riggers flows of food across space. Through efficien spaial arbirage, he risk of crop failure in some regions is shared over a large marke area, prices are more sable, and food shorages may be prevened. The aim of his sudy is o measure spaial arbirage efficiency in Mozambique s pos-reform maize markes. The lieraure on spaial rade is vas, and he echniques used o assess spaial arbirage are increasingly sophisicaed. However, he sudies on spaial arbirage efficiency in Souhern Africa (SA), one of he mos food insecure regions in he world (Van Rooyen and Sigwele, 1998), are very few. Previous research by Tschirley and Sanos (1999), and Zavala (2001) used price correlaions and concluded ha maize markes are becoming inegraed in Mozambique. However, hese sudies rely solely on 3

4 price correlaions ha canno answer quesions regarding marke efficiency or marke inegraion (Harris, 1979; Baulch, 1997; Barre and Li, 2002). The erm marke inegraion is ofen used o describe spaial price efficiency (Barre, 2001). Here, markes are defined as inegraed if here are physical flows of produc from one o anoher. Inegraion, herefore, implies ransfer of excess demand from one marke o anoher noiceable hrough physical maize flows, or ransmission of price shocks from one marke o anoher, or boh (Barre, 2001). Alhough hey are closely relaed, marke inegraion and marke efficiency are no he same (Li and Barre, 2000). Arbirage is efficien if arbirageurs force maize price differences o converge oward ransfer coss whenever he maize price differenial exceeds ransfer coss. Thus, rade occurring is a sufficien condiion for marke inegraion, bu spaial arbirage efficiency requires ha price spreads be equal or less han ransfer coss. Because spaial rade depends upon prices, ransfer coss and commodiy flows, aemps o measure arbirage efficiency using only prices, like many popular mehods (e.g. bivariae price correlaions, coinegraion, error correcion models, hreshold auoregression, and coinegraion), can be hazardous (Harris, 1979; McNew and Fackler, 1997; Baulch, 1997; Li and Barre, 2000; Obsfeld and Taylor, 1997; Barre, 2001). The main inferenial difficuly is due o rade disconinuiies (due o a hird rading parner, and/or large ransacion coss) and nonsaionary prices and ransfer coss (due o inflaion, climae shocks or populaion) ha cause biased and inconsisen parameer esimaes (McNew and Fackler, 1997; Barre, 2001). When spaial efficiency ess rely solely on prices, well funcioning markes may be found saisically uncorrelaed and 4

5 deemed o be funcionally isolaed; and markes wihou rade links can be found o be price correlaed and deemed efficien (Harris, 1979; Baulch, 1997; Barre, 2001). To es for arbirage efficiency in Mozambique s maize markeing, we use Baulch s (1997) pariy bound model (PBM). The PBM is a hree- regime swiching regression ha accouns for nonsaionary ransfer coss, and recognizes he exisence of disconinuous rade paerns (Baulch, 1997; Li and Barre, 2000; Barre, 2001). Maize Trading and Marke Consrains Whie maize is boh he main source of calories and he larges produced crop in Mozambique (DPDS-UEM-IFPRI, 1998). Inerregional maize rading is imporan because souhern Mozambique is a defici maize region in all years and norhern Mozambique is a maize surplus region even during regional droughs (Tschirley and Sanos, 1999). Before marke reforms in Mozambique (which began in 1987), maize surpluses in norhern Mozambique were channeled o defici regions in he souh by subsidized sae-owned monopolies/monopsonies ha conrolled maize procuremen, disribuion, and processing. In spie of massive subsidies, he sysem failed o provide food securiy and caused profound budge deficis. While he massive subsidies are now gone, food securiy coninues o be a problem in Mozambique. Maize raders in Mozambique face ransporaion resricions (The Governmen of Mozambique, 1998, 1999b). Mos of he ransporaion sysem in Mozambique was buil during he colonial period and is in poor condiion. Maize is mainly ranspored by rucks and only 19 % of he roads are paved in Mozambique (The Governmen of Mozambique, 2001). Furher, mos roadways are impassable during he rainy season. 5

6 In 2001, maize ransporaion coss ranged beween $0.03 o $0.21/on/kilomeer and were abou 50 % of he maize price in Mapuo. Currenly, here is no direc road connecion beween norhern Mozambique and oher regions of he counry. The road beween maize surplus regions in norhern Mozambique, and maize defici regions in souhern Mozambique, sops a Caia on he border beween Sofala (Beira) and Zambézia (Mocuba) provinces (see figure 1). A ferry is used o cross he Zambezi River unil Zambézia where he road resumes again. However, he ferry is no always available. For example, in early 2001 he ferry was shu down for nearly wo monhs because he Zambezi River was flooding, and in July 2001 he ferry service was inerruped again because here was no enough waer in he river (AIM, No. 211, 2001). In much of he rainy season, road raffic beween Zambézia and oher regions of cenral Mozambique mus go hrough Malawi (The Governmen of Mozambique, 2001). Maize prices remain very low in norhern Mozambique and explain why subsisence agriculure is a common pracice despie favorable agroecological condiions for inensive agriculure (Jeje e al., 1998). Spaial maize rading in Mozambique is summarized as follows. Traders (wholesalers/reailers) ravel o producion areas (usually by rucks) where hey inspec and buy maize on sie. Alhough some barering has been observed (The Governmen of Mozambique, 1999a), maize is usually purchased in cash. Wholesalers and reailers, who supply consumers and small millers, buy maize from small-scale assemblers or direcly from producers. Besides receiving maize from oher regions wihin Mozambique, souhern Mozambique depends on maize impors from Souh Africa. 6

7 Daa The es for spaial arbirage efficiency uses Baulch s (1997) PBM and monhly whie maize reail prices and monhly ransporaion cos daa. All maize price daa ( ) were gahered in he pos-reform period, and are from he Sisema de Informação de Mercados (SIMA, 2001), a public oulook program. The SIMA (2001) daa are gahered by inerviewing randomly seleced reailers in each province (hree inerviews per week). As able 1 indicaes, average prices in souhern Mozambique are almos double prices in cenral/norhern Mozambique, bu price volailiy around he mean is higher in norhern/cenral Mozambique. Souhern Mozambique is a maize defici region bu has beer marke infrasrucure han he norhern maize surplus regions. Usually, reail maize prices begin rising around Sepember reaching a maximum around March (figure 2). Seasonal maize price flucuaions are correlaed wih maize harves season, which exends from March o lae April. Table 1 also indicaes ha price daa have large informaion gaps. The number of missing values varies from one (Mapuo) o one hundred and seven (Lichinga). Missing values represen los informaion and make regression esimaion more difficul. Despie he difficuly, he price daa used in he regression esimaion uses only acual price daa and do no include any inerpolaed values. The price daa were esed for uni roos wih Phillips-Perron ess. The null hypohesis of a uni roo was rejeced for all prices excep Lichinga, which has he fewes observaions. Road ransporaion cos daa used o exrapolae eigh years of ime-series ransporaion cos are from he Minisry of Trade and Indusry (one observaion in June 7

8 2001) and from SIMA (2001) (weekly daa gahered beween Augus and Sepember 2001). When he average per uni ransfer cos beween he wo sources of daa (SIMA (2001) and Minisry of Trade and Indusry) is differen, he value used o projec imeseries cos daa is he highes of he wo, excep when he average ransfer cos from he SIMA (2001) is based on more observaions. This choice allows conservaive esimaes of spaial price inefficiency. As ables 2 and 3 show, ranspor coss vary beween U.S. $0.03 and U.S. $0.21 per kilomeer. Per uni maize ransporaion coss decrease wih he disance and he size of he ruck. Nampula, which has he poores roads, has he highes ransporaion charges per kilomeer. By comparison, U.S. ruck ransporaion raes, assuming no backhaul, are abou U.S. $0.07/on-kilomeer. Thus Mozambique s ruck ransporaion coss are roughly comparable o U.S. raes. Mos U.S. grain, however, ravels by barge or rail when he disance is large. U.S. barge raes are abou 10% of ruck raes. As suggesed by Baulch (1997), a consumer price index (CPI) was used o exrapolae available cos daa ino eigh years of monhly ime-series ransporaion cos daa (a proxy for ransfer cos). The monhly CPI for all Mozambique is from he Banco de Moçambique. Monhly diesel prices are used as an alernaive deflaor, o provide a es of fragiliy. The Direcção Nacional de Energia ( ), provided he diesel prices. Diesel prices were unusually high during he daa collecion period (Figure 3). Therefore, using diesel prices as a deflaor is expeced o resul in less spaial arbirage inefficiency. The exrapolaed cos daa, which are used as iniial ransfer-cos esimaes on he PBM, seem o be reasonable. For example, using he CPI as a deflaor, he exrapolaed 8

9 9 ransfer cos for he roue Chimoio-Mapuo in December of 1995 (U.S. $0.07/on/kilomeer) is slighly higher han Couler s (1995)esimae of U.S. $ $0.05 on/kilomeer for he same period. The Governmen of Mozambique s (1998) esimae of coss o ranspor maize from Chimoio o Mapuo in December of 1998 (369 Meicais per kilogram) is also slighly lower han he deflaed value of (487 Meicais/on per kilomeer) used in his sudy. If he exrapolaed values overesimaed he coss, hey would in urn underesimae spaial arbirage inefficiency. Pariy Bound Model Baulch s Pariy Bound Model (PBM) is used o deermine spaial price efficiency in Mozambique s pos-reform maize markes. The problem consiss of esimaing he probabiliy of each rade regime, given maize prices and ransfer cos daa. Regime I and II are consisen wih efficien spaial arbirage, and regime III indicaes marke inefficiency. The likelihood funcion of he PBM is defined as (Baulch, 1997): ] ) (1 [ T f f f L λ λ λ λ + + = = where λ 1 and λ 2 are he probabiliies for regimes I and II when price spreads equal and fall below ransfer coss respecively. The funcion f 1 is he densiy funcion for regime I (i.e. spreads equal ransfer coss) defined as: = e e K Y f σ φ σ 1 1 ; f 2 is he densiy funcion for regime II (i.e. spreads less han ransfer coss), and is defined as: + Φ + + = 2 1 / / / ) ( ) ( 1 ) ( ) ( 2 u e e u u e u e K Y K Y f σ σ σ σ σ σ φ σ σ ; (4) (5) (6) (1)

10 and f 3 is he densiy funcion for regime III when price spreads exceed ransfer coss. The densiy funcion for regime III is: f ( σ e + σ u ) Y K φ 2 2 ( σ e + σ u ) ( Y K ) σ v 1 Φ 2 2 ( σ e + σ v ) = 1 / 2 1 / 2 1 / 2 σ e (7) where Y is he naural logarihm of he absolue value of price difference beween markes i and j; σ e, σ u, and σ v are sandard deviaions of he error erm, respecively e, u, and v ; K is he logarihm of he nominal ransporaion cos in period, and φ(.) and Φ (.) represen he sandard normal densiy and disribuion funcion respecively. The esimaes of he parameers λ 1, λ 2, σ e, σ u andσ v are obained by maximizing he likelihood funcion (14) using TSP version 4.4. (1997) The seleced opion ML(HCOV = D, HITER = D, TOL = 0.01) calls for he Davidon-Flecher-Powell algorihm (ha uses numeric derivaives and is helpful wih bad saring values) wih a convergence crierion of 0.01 (TSP User s Guide). The opion ANALYZ is used o compue he sandard errors of he unconsrained regime probabiliies and conduc (Wald) -ess. Mozambique has poor roads and informaion lags so here is almos cerain o be some inefficiency. Therefore, insead of esing he null of perfec spaial efficiency (λ 1 + λ 2 = 1), we focus on he magniude of he regime probabiliies. The sum of probabiliies of regime I (i.e. spreads equal o ransfer coss) and II (i.e. price spreads less han ranspor coss) is inerpreed as he frequency of marke efficiency. Tha is, he more efficien spaial arbirage is, he closer o zero is he index of marke inefficiency λ 3 (λ 3 = 1-λ 1 + λ 2 ). Because here is a possibiliy of a local opimum, en ses of saring values, 10

11 randomly drawn from a uniform disribuion are used. The esimaes repored are he ones wih he highes likelihood funcion value. To accoun for seasonaliy, we esimaed an unresriced PBM considering wo seasons ha are predominan in Mozambique (cool and dry, and warm and we). The we season sars early Sepember and ends in lae March, and he cool season runs from April o Augus. Seasonal effecs are capured by a dummy variable included in equaion (4). However, if daa on ranspor coss were available, seasonal effecs would be beer capured using varying values of ranspor coss wih equaions (5)-(7). Limiaions of he PBM The PBM has limiaions ha are worh noing. Firs, he qualiy of he PBM esimaes depends heavily on he exrapolaed ransfer cos daa. The exrapolaed daa were close o esimaes by Couler (1995) and The Governmen of Mozambique (1998). However, since hese daa were esimaed similarly, hey could all suffer from he same bias. Sensiiviy ess showed ha he resuls are no fragile wih respec o which of he wo deflaors are used, bu are hey are sensiive o an increase or decrease in ransfer coss of 50 % 1. Secondly, while he PBM deecs violaions of spaial arbirage condiions ha indicae inefficiency, i does no idenify he reasons for such violaions. Spaial arbirage inefficiency may indicae ha some marke consrains sill exis. 1 Abou 77% of he marke pairs are sensiive o changes in ransfer coss (increases or decreases by 50%). A 50% increase in ransfer coss (TC) led o zero inefficiency in all marke pairs bu one (Mapuo- Chimoio). The percenage change in price inefficiency due o a 50% increase in TC varies from 0.5 o infiniy wih a mode a 2 (69% of elasiciies are equal o 2), while he change in price inefficiency due o a price decrease of 50% varies from 0.4 o infiniy. In summary, he qualiy of ransfer cos daa affecs he PBM esimaes. PBM esimaes were more sensiive o a 50 % decrease han a 50% increase. 11

12 Third, he disribuional assumpions such as independence beween ransfer cos daa and maize prices, half-normal error erms, and no auocorrelaions may no hold exacly. The swiching regression approach fis a disribuion as opposed o esimaing a predicive model. Barre and Li (2002) show ha resuls can be sensiive o violaions of he half normaliy assumpion. Resuls The esimaes of he PBM using monhly maize reail prices and ransporaion coss deflaed by he CPI for all Mozambique are presened in able 4. As indicaed by he sum of he probabiliies of regime I and II, spaial arbirage is efficien more han 55 % of he ime beween markes wihin souhern Mozambique (Mapuo, Xai-Xai, and Maxixe), and over 84 % of he ime beween markes wihin cenral Mozambique (Tee, Beira, and Chimoio). Resuls also sugges ha spaial arbirage beween souhern and cenral Mozambique is efficien more han 80 % of he ime, excep for he roue Chimoio - Mapuo ha is only efficien 18 % of he ime. Price spreads beween maize markes in norhern Mozambique and hose in cenral/souhern Mozambique fall below ransporaion coss abou 100 % of he ime. The PBM esimaes using diesel prices as a deflaor of ransporaion cos observaions are no repored because hey were similar and showed even less inefficiency. As able 5 indicaes, here are flows of maize wihin and beween souh/cenral Mozambique, bu here are no maize shipmens from norhern Mozambique o cenral/souhern regions. Taken ogeher, he PBM esimaes and inerregional maize shipmens sugges ha spaial arbirage is near efficien wihin and beween 12

13 souhern/cenral Mozambique. Markes in norhern Mozambique are funcionally isolaed from he oher markes. Hence, souhern Mozambique will coninue o rely on maize impors from cenral Mozambique or from Souh Africa. Furher, smallholders in norhern Mozambique will coninue o see low maize prices, and hence coninue wih subsisence agriculure, unless hey can expor o neighboring counries. Neighboring counries have impored considerable maize from Mozambique when faced wih bad harvess (Tschirley, 1998). High probabiliies of being inside pariy bounds (regime II) beween norhern Mozambique cenral/souhern Mozambique suggess ha ransporaion coss are prohibiive (Baulch, 1997), and is consisen wih poor road access in Mozambique. Roads are poor (abou 56% of he roads in Mozambique are feeder roads), and here is no direc road connecion beween norhern and cenral/souhern Mozambique. The size of spaial arbirage efficiency (as measured by sum of regimes I and II) in Mozambique is similar o he resuls obained for rice markes in he Philippines by Baulch (1997), and hose obained by Li and Barre (2000) while sudying soybean meal rade beween he Unied Saes, Canada, Japan and Taiwan. Thus marke efficiency does no seem worse in Mozambique han elsewhere. The only wo roues showing much inefficiency are he roues from Chimoio o Mapuo and from Mapuo o Maxixe. Roads along hese roues have undergone recen improvemens. Exrapolaing curren ranspor coss back o a ime when roads were no in as good of condiion may have underesimaed ranspor coss in he early years of he daa period. The resuls for hese roues may have more o say abou he imporance of qualiy roads han abou marke inefficiency. 13

14 To accoun for seasonaliy, we esimaed he PBM allowing parameer esimaes o vary across seasons (dry and we seasons). The resuls are presened in able 6. There is slighly more price inefficiency during he dry season han in he we season. In he we season, wo maize roues (15%) show saisically significan price inefficiency. During ha season, price inefficiency is zero in all maize roues, excep wo (Tee and Beira, and Chimoio and Mapuo). In he dry season, hree maize roues (23%) show price inefficiency. The number of maize roues showing price inefficiency in one season or he oher is 5 ( 40%). Seasonal effecs maered lile for he markes where price inefficiency was zero. For he oher markes, seasonaliy reveals some addiional inefficiency. Resuls of Price Based Mehods One of he weaknesses of he PBM is is high dependence on he qualiy of ransfer cos daa. To complemen our analysis, we used a bivariae vecor auoregression (VAR) o es for Granger causaliy beween pairs of markes presened in able 4. Time series procedures are applied o all maize roues used in he esimaion of he PBM excep Xai- Xai, Pemba and Mocuba, which did no have enough usable observaions for ime series mehods. Weekly daa are used because monhly price series had insufficien degrees of freedom for hypohesis esing. Oher possible relaionships are no included in eiher he PBM or ime series esimaion because of lack of sufficien daa or rare maize shipmens beween regions. The VAR s were esimaed in levels because he null hypohesis of a uni roo on maize prices was rejeced on he basis of Dickey Fuller and Philips Perron uni roo ess (Tosão, 2002). 14

15 The lag lenghs used in he VARs were deermined hrough a wo-sep (Bayesian Informaion Crieria, BIC) procedure. Firs, each maize price is regressed on is own lags, and he lag order wih minimum BIC is seleced. Second, each maize price is regressed agains he lags of oher prices (he maximum lag allowed was eigh) given is own lag lengh seleced in he firs sep. The lag lengh wih minimum BIC is again seleced, and used o conduc causaliy ess. Since he lag lenghs varied across equaions, seemingly unrelaed regression was used o esimae he parameers. Causaliy ess are presened in able 7. The lag lenghs used in causaliy ess have a mode of 1 week. These lag lenghs are shorer han hose found by Schroeder and Goodwin, and by Baulch (1997). Schroeder and Goodwin found ha slaugher seer prices in he U.S. cale indusry have lagged price responses beween wo and hree weeks. Baulch (1997) found lags beween one and wo monhs in Philippine s rice indusry. Bu, he shorer lags found here could be parly due o he small number of observaions available. Corroboraing wih he PBM esimaes, Granger causaliy ess indicaed ha here are price links beween mos pairs of markes (67%) presened in able 4. Maize prices are unlikely o be discovered in Nampula (norhern Mozambique), and Beira seems o be among he leaders in he price discovery process (Beira has feedback relaionships wih wo markes in cenral Mozambique - Tee and Chimoio). Beira is locaed in he hear of he Beira Corridor (a 100 km wide inernaional rade channel linking Zimbabwe, Mozambique, Zambia, Malawi and Souh Africa o he por of Beira on he Indian ocean) in cenral Mozambique. The causaliy ess show more significance 15

16 han he PBM ess. The differences in resuls are no surprising since he mehods are no measuring he same ype of efficiency. Policy Implicaions The resuls of his sudy say more abou wha he soluion o he hunger problem is no han wha he soluion is. The problem does no seem o be a lack of raders o ship grain from low price areas o high price areas. The problem is ha ranspor coss are so high ha i is no profiable o ship he grain. Given he low qualiy of Mozambique s roads he ransporaion coss are no unreasonable. The ideal soluion would be o develop low-cos rail or barge ransporaion. Bu, here are sill some policies ha can reduce ruck ranspor coss. The mos obvious is o coninue o improve roads. Transpor coss are subsanially high in areas wih poor roads. Daa on inerregional shipmens sugges ha surplus markes in norhern Mozambique are no conneced wih defici markes in souhern Mozambique. One opion for isolaed regions like norhern Mozambique is o focus on high value crops, so ha people will have he money o impor food when crops fail. Souhern Mozambique may have o impor food from oher counries. There has been a long debae in he developmen lieraure abou wheher he key o developmen was geing he prices righ or infrasrucure developmen. This aricle finds some price inefficiency in Mozambique, bu no enough o explain why food securiy remains a problem. The bes hope for Mozambique remains in developing infrasrucure such as roads, and financial and legal insiuions and o promoe economic 16

17 freedom. Price suppors and subsidies are cosly programs ha have been ried and failed. Summary and Conclusions Maize shorages and maize price insabiliy coninue o occur in Mozambique in spie of policy reforms and grea capaciy for increased producion in norhern Mozambique. This sudy measured spaial arbirage efficiency in he pos-reform period in Mozambique. Spaial inefficiency alone does no seem o explain why food shorages coninue o occur in Mozambique. Baulch s (1997) efficiency ess indicaed ha maize markes in cenral Mozambique are he mos efficien (84 o 100% of he ime), followed by markes in souhern Mozambique (55 o 100 % of he ime). Spaial arbirage beween cenral and souhern Mozambique is efficien 80 o 100 % of he ime. Price spreads beween maize markes in norhern Mozambique and hose in cenral/souhern Mozambique fall below ransporaion coss nearly all of he ime. The PBM esimaes indicae ha i does no pay o ship maize from he norhern surplus maize regions o souhern Mozambique. Records of inerregional maize shipmens show ha indeed grain is no being shipped from norhern o souhern Mozambique. Nearly 40% of maize roues show price inefficiency in one season or he oher. Marke liberalizaion seems o have improved spaial price efficiency, bu high ransfer coss sill limi rade and poenial benefis from freeing he markes. Under hese circumsances, food shorages and price insabiliy are likely o coninue. Poor roads and rader s lack of access o capial are possible explanaions of limied spaial rade beween maize surplus regions and maize defici regions in 17

18 Mozambique. Poor roads raise rader s ransporaion coss and limi access o local markes. Ineres raes are high for mos maize raders. High ineres raes force raders o ransac small quaniies of maize (which add o he already high ransporaion coss) resuling in high per-uni ransfer coss. The end resul is high prices in maize defici areas, and reduced access o food by households in maize defici areas. Invesmen in marke infrasrucure (paricularly norh-souh roads) and markeing insiuions (e.g. financial, and legal insiuions) would lower raders coss and help reduce food shorages and price insabiliies. 18

19 References Admisisração Nacional de Esradas. Unpublished elecronic documen, Andersen, P.P., Deparmenal seminar, Oklahoma Sae Universiy, Sillwaer, OK. Arnd, C., Schiller, R., and Tarp, F., Grain ranspor and rural credi in Mozambique: Solving he space-ime problem. Agric. Econ. 25, Barre, C.B., Measuring inegraion and efficiency in inernaional agriculural markes. Rev. Agric. Econ. 23, Barre, C.B. and Li, J.R., Disinguishing beween equilibrium and inegraion in spaial price analysis. Am. J. Agric. Econ. 84, Baulch, B., Transfer coss, spaial arbirage, and esing for food marke inegraion. Am. J. Agric. Econ. 79, Banco de Moçambique. Saisical Bullein. Various issues Couler, J.P., Esudo da comercialização e a políica de preços de milho em Moçambique. England: Naural Resources Insiue Cenral Avenue, Chaham Mariime Ken ME 4TB. Direcção Nacional de Energia DECO, Esruura de Preços de Combusíveis. Unpublished energy price repors, Mapuo. DPDS (Deparameno de População e Desenvolvimeno), UEM (Universidade Eduardo Mondlane) and IFPRI (Inernaional Food Policy Insiue), Undersanding Povery and Well Being in Mozambique: Firs Naional Assessmen. Mapuo: Minisério do Plano e Financas. Harris, B., There is mehod in my madness: Or is i vice versa? Measuring agriculural marke performance. Food Res. Ins. 2, Inernaional Food Policy Research Insiue (IFPRI), Research group calls for heads of sae o maser poliical will o feed he world s hungry: Urges governmens o reinvigorae Sub-Saharan Africa s failing agriculural sysems and inves in coninen s marginal areas. Washingon, D.C., available a hp:// accessed December 16, Jayne, T.S., Mukumbu, M., Chisvo, M., Tschirley, D., Weber, M.T., Zulu, B., Johansson, R., Sanos, P., and Soroko, D., Successes and challenges of food marke reform: Experience from Kenya, Mozambique, Zambia, and Zimbabwe. MSU Inernaional Developmen Working Paper No

20 Jeje, J.J., Machungo, C., Howard, J., Srasberg, P., Tschirley, D., Crawford, E., and Weber, M., Wha makes agriculural inensificaion profiable for Mozambique smallholders? An appraisal of he inpus ssubsecor and he 1996/1997 DNER/SG2000 program: Main repor. Minisry of Agriculure and Fishers, Direcorae of Economics, Research Repor No.32. Kilman, S. and Thurow, R., December 3, Diminishing reurns: Africa could grow enough food o feed iself; Should i? The Wall Sree Journal, p.a1. Li, J.R. and Barre, C.B., Disinguishing beween equilibrium and inegraion in marke analysis. ISSN , Working Paper No. 99, available a hp:// accessed Augus McNew, K. and Fackler, P.L., Tesing marke equilibrium: Is coinegraion informaive? J. Agric. Resour. Econ. 22, Mozambique News Agency, AIM Repors No. 211, available a hp:// accessed Ocober Obsfeld, M. and Taylor, A.M., Nonlinear aspecs of goods-marke arbirage and adjusmen: Heckscher s commodiy poins revisied. Naional Bureau of Economics Research, Working Paper No Schroeder, T. C., and Goodwin, B. K., Regional fed cale price dynamics. Wesern Journal of Agriculural Economics, 15, Sisema de Informaçãode Mercado (SIMA), Elecronic Daabase: Minisry of Agriculure and Rural Developmen, Mapuo. The Governmen of Mozambique, Análise dos Cusos de Transpore na Comercialização Agrícola em Moçambique Esudo de Casos dos Transpores de Milho das Zonas Nore e Cenro Para a Zona Sul de Moçambique. Minisério da Indusria e Comercio, DNC No. 18. The Governmen of Mozambique, Caracerização do Comércio de Milho na Região Cenro Os inervenienes e o Desenvolvimeno do Mercado. Minisério da Indúria Comércio e Turismo, DNCI No. 3. The Governmen of Mozambique 1999a. Opções Financeiras Para a Comercialização Agrícola em Moçamique Documeno Principal. Minisério da Induria Comércio e Turismo, DNCI No. 11. The Governmen of Mozambique,1999b Some Aspecs of Access o and Demand for Commercial Credi among Agriculural Traders. Minisério da Indusria Comercio e Turismo, DNCI No

21 Tosão, E., 2002 Spaial arbirage and maize price dynamics in Mozambique. Unpublished M.S. hesis, Oklahoma Sae Universiy. Tschirley, D., Planning for drough in Mozambique: Balancing he roles of food aid and food markes. Minisry of Agriculure and Fishers, Direcorae of Economics, Research Repor No.29. Tschirley, D. and Sanos, A.P., The effecs of maize rade wih Malawi on price levels in Mozambique: Implicaions for rade and developmen. Minisry of Agriculure and Fishers, Direcorae of Economics, Deparmen of Policy Analysis, Research Repor No.34. TSP Sofware, Version 4.4. Copyrigh (c) 1997 by TSP Inernaional. Van Rooyen, J. and Sigwele, H., Towards regional food securiy in souhern Africa: A (new) policy framework for he agriculural secor. Food Policy, 23, Zavala, H. Análise da Inegração dos Mercados de Cereais e Feijões. Unpublished Bsc hesis, Universidade Eduardo Mondlane,

22 Figure 1. Type of Roads in Mozambique and Locaion of he Markes Included in he Sudy. N W E Lichinga Pemba S Nampula Mocuba Caia Chimoio Beira Maxixe Mapuo Xai-Xai Miles Paved Earhwork Counry Impassable Source: Adaped from he Adminisração Nacional de Esradas,

23 Figure 2. Monhly Reail Maize Prices in Cenral and Souhern Mozambique (Meicais/Kg) beween May 1994 and March Price Mapuo Beira Tee Mar_94 Sep_94 Mar_95 Sep_95 Mar_96 Sep_96 Mar_97 Sep_97 Mar_98 Sep_98 Mar_99 Sep_99 Mar_00 Sep_00 Mar_01 23

24 Figure 3. Monhly CPI for All Mozambique, and Monhly Diesel Price Indexes for Beira and Lichinga beween February 1995 and June Index Lichinga CPI Beira Feb_95 Jul_95 Dec_95 May_96 Oc_96 Mar_97 Aug_97 Jan_98 Jun_98 Nov_98 Apr_99 Sep_99 Feb_00 Jul_00 Dec_00 May_01 24

25 Table 1. Descripive Saisics for Weekly Maize Reail Prices in Mozambique (Meicais/Kilogram) beween July 2, 1994 and April 4, Variable N Mean Sandard Deviaion Minimum Maximum Number of Missing Values Mapuo Xai-Xai Maxixe Beira Chimoio Tee Mocuba Nampula Pemba Lichinga Monhly CPI 2001 = Noe: The monhly price index (CPI) for he period January May

26 Table 2. Inerregional Maize Transporaion Cos in Mozambique (MZM/on/KM) beween Augus 29 and November 29, Desinaion Marke Maize Source Mapuo Xai-Xai Maxixe Beira Tee Maxixe Mocuba Xai-Xai 762 (n = 1) Chimoio 713 (n = 8) 450 (n = 1) Manica 826 (n = 1) 1040 (n = 2) Tee 944 (n = 1) 633 (n = 1) Nampula 698 (n = 1) 762 (n = 4) 756 (n = 7) 1179 (n = 2) Lichinga 442 (n = 5) 363 (n = 1) Noes: The number of observaions is given in parenheses. The daa were compiled from he SIMA (2001) weekly bulleins (Augus 29 and November 29, 2001). The daa were convered from bags o ons assuming a bag weighed 100kg. US $1 16,000 MZM. 26

27 Table 3. Road Maize Transporaion Coss in Mozambique (MZM/on/Km), June 2001 Desinaion Source Mapuo Xai-Xai Maxixe Beira Chimoio Tee Mocuba Nampula Pemba Lichinga Mapuo 1,120 1, ,440 1,440 1,440 1,440 Xai-Xai 909 Chimoio 400 Tee Nampula 1,303 1,338 1,260 1,572 1,532 1,142 1,483 1,231 Pemba 905 1,198 Noe: The exchange rae during daa gahering was $1 = 16,000 MZM. Source: The Governmen of Mozambique, 2001

28 Table 4. Esimaes of he PBM Regime Probabiliies for Mozambique Maize Reail Markes Using CPI as a Deflaor o Projec Transfer Coss. Probabiliies Maize Roue Regime I Regime II Regime III Chimoio Xai-Xai (0.165) Mapuo Beira (0.077) Chimoio Mapuo 0.1E-2 (0.3E-3) Pemba Mapuo (0.039) Xai-Xai Mapuo (0.125) Nampula Beira (0.507) Nampula Tee 0.1E-02 (0.2E-6 a ) Chimoio Mocuba (0.086) Tee Mapuo (0.077) Tee Beira (0.415) Chimoio Beira (0.097) Mapuo Maxixe (0.103) Nampula Maxixe 0.1E-2 (0.2E-2) (0.094) (0.077) (0.104) (0.039) (0.084) 0.1E-02 (0.503 a ) (0.2E-6 a ) (0.086) (0.077) (0.056) (0.097) (0.082) (0.018) (0.153) 0.7E-6 (0.2E-5) (0.104) 0.4E-10 (0.2E-9) 0.6E-2 (0.066) 0.2E-6 (0.4E-5 a ) 0.5E-11 (0.2E-10 a ) 0.2E-6 (0.4E-2) 0.7E-3 (0.6E-2) (0.424) 0.1E-11 (0.1E-10) (0.127) (0.016) Noes: Regime I, II, and III are defined respecively when price spreads equal, fall below, and exceed ransfer cos. Sandard errors are in parenheses. a These sandard errors may no be valid since he edge of he parameer space was encounered. 28

29 Table 5. Recorded Inerregional Maize Flows in Mozambique ( ). Percen of Desinaion Marke Exporing Marke Mapuo Inhambane Beira Chimoio Zambezia Nampula Mapuo Xai-Xai Maxixe Beira Chimoio Tee Mocuba Nampula Pemba Lichinga Ohers Toal Noes: Auhor s compuaions based on daa from he SIMA (2001) daabase. 29

30 Table 6. Esimaes of he PBM Regime Probabiliies for Mozambique Maize Reail Markes in he Dry and We seasons Using CPI as a Deflaor o Projec Transfer Coss. Probabiliies Maize Roue Regime I Regime II Regime III Dry season We season Dry season We season Dry season We season Chimoio Xai-Xai 0.1E-2 (0.3E-3) (0.126) (0.091) (0.126) (0.091) 0.2E-15 (0.1E-14) Mapuo Beira (0.098) (0.077) (0.095) (0.077) (0.078) 0.9E-10 (0.4E-9) Chimoio Mapuo 0.1E-2 (0.2E-3) 0.1E-2 a (0.2E-10) (0.087) a (0.064) (0.087) a (0.064) Pemba Mapuo 0.1E-2 a (0.2E-4) 0.1E-2 (0.9E-4) a (0.2E-4) (0.9E-4) 0.2E-8 a (0.1E-7) 0.6E-12 (0.1E-10) Xai-Xai Mapuo (0.083) (0.133) (0.083) (0.133) 0.1E-7 (0.1E-8) 0.3E-7 (0.1E-6) Nampula Beira 0.1E-2 a (0.1E-13) 0.1E-2 a (0.9E-4) a (0.2E-11) a (0.9E-4) 0.3E-12 a (0.2E-11) 0.4E-14 a (0.2E-12) Nampula Tee 0.1E-2 a (0.2E-5) 0.1E-2 a (0.1E-3) a (0.4E-5) a (0.1E-3) 0.2E-7 a (0.2E-5) 0.3E-7 a (0.5E-6) Chimoio Mocuba (0.143) (0.119) (0.143) (0.119) 0.9E-6 (0.1E-3) 0.1E-3 (0.1E-2) Tee Mapuo (0.141) (0.073) (0.141) (0.073) 0.9E-4 (0.4E-3) 0.1E-8 (0.5E-8) Tee Beira (0.108) 0.5E-02 (0.241) (0.108) (0.059) 0.6E-6 (0.2E-5) (0.263) Chimoio Beira (0.112) (0.165) (0.112) (0.164) 0.4E-10 (0.2E-09) 0.3E-3 (0.2E-2) Mapuo Maxixe (0.138) (0.092) (0.105) (0.092) (0.134) 0.3E-5 (0.8E-4) Nampula Maxixe 0.1E-2 0.1E (0.099) (0.9E-5) (0.099) 0.5E-7 (0.2E-5) 0.7E-8 (0.3E-7) Noes: Regime I, II, and III are defined respecively when price spreads equal, fall below, and exceed ransfer cos. Sandard errors are in parenheses. a These sandard errors may no be valid since he edge of he parameer space was encounered. 30

31 Table 7. F-saisics of Bivariae Causaliy Tess beween Reail Maize Prices Lagged Independen Variable Dependen Variable Mapuo Nampula Beira Tee Maxixe Chimoio Mapuo 6.11* (0.00) Nampula 3.21* (0.01) Beira 1.89 (0.17) Tee 4.53* (0.03) Maxixe 5.37* (0.02) 0.21 (0.65) 2.50 (0.11) 0.56 (0.45) 4.53* (0.00) 1.06 (0.35) 13.94* (0.00) 6.40* (0.01) Chimoio 7.04* (0.00) 4.46* (0.01) Noes: p-values are in parenheses. An aserisk indicaes significance a 5 % level * (0.00) 1.64 (0.12) 6.05* (0.01) 10.50* (0.00) 31