Assessment of Price Volatility in the Fisheries Sector in Uganda

Transcription

1 Volume 48, Issue Assessmen of Price Volailiy in he Fisheries Secor in Uganda James O. a a Professor of Resource Economics, College of Agriculural, Life, and Naural Sciences, Alabama A&M Universiy, 4900 Meridian Sree, Normal, AL james.bukenya@aamu.edu Absrac This paper examines price volailiy in he African cafish (Clarias gariepinus) supply chain in Uganda. The volailiy process in he cafish markes was analyzed based on monhly price daa from January 006 o Augus 03. A GARCH model is used o esimae he volailiy parameers. Empirical resuls revealed ha he value of he firs-order auoregressive erm and he value of he firs-order moving average erm were significan for boh aquaculure and wildharves cafish supply chains. The observed long persisence of volailiy in boh supply channels suggess a fundamenal level of uncerainy and risk in he cafish subsecor over he sudied period. Keywords: aquaculure, cafish, GARCH model, price volailiy Corresponding auhor. 8

2 Inroducion In sub-saharan Africa, several price volailiy sudies have explored oupu markes for saple foods (Sukai 03; Mino, 04; Ngare, Simowe, and Massingue, 04), bu very lile research has been done in he fisheries secor. This paper explores volailiy persisence in Ugandan cafish markes. An undersanding of he srucure of price volailiy in Uganda s cafish supply chain is of grea ineres because cafish has become an imporan raded species, wih expors o regional markes rising even faser han producion ( and Ssebisubi, 04). There have also been exensive effors by he governmen and inernaional donors o increase he counry s fish producion hrough invesmens in aquaculure, and he African cafish has become he predominan culured species. However, he consequences of increased cafish producion from aquaculure subsecor developmen on price sabiliy in he domesic marke have ye o be sudied. If monhly flucuaions can be deeced and measured, i will be easier o make predicions abou prices and o undersand heir behavior over ime. Ideally, well-funcioning markes ransmi price signals, which allow changes in demand o be me by supply. When demand is greaer han supply, producers increase producion in response o price signals; his increased producion, in urn, helps sabilize prices. Background Uganda is a small, landlocked counry in Eas Africa surrounded by Kenya, Tanzania, Rwanda, he Democraic Republic of Congo, and Souh Sudan. Fisheries resources are among is mos significan naural resource endowmens. Because abou 0% of is surface area is covered wih waer, Uganda has enormous fisheries resources poenial for capure fisheries and aquaculure producion (Deparmen of Fisheries Resources, 0). Capure fishery is basically arisanal and is suppored by small-scale fishing communiies around he lakes. The African cafish has recenly emerged as he mos favored species for aquaculure, accouning for more han 60% of aquaculure producion. Farmed cafish is primarily produced by farmers who pracice fish farming as one of many oher farming aciviies. Wih improved marke prices, governmen inervenion for increased producion, and sagnaing supply from capure fisheries, aquaculure has araced enrepreneur farmers seeking o exploi he business opporuniy provided by he prevailing demand. Alhough he operaion of he local markeing sysem has been he subjec of previous sudies, he disribuion of fish and fish producs has improved over he las fifeen years, wih increased channels involving middle agens supplying fish o facories involved in indusrial fish processing and expor and raders supplying fish o rural and urban markes. Pricing is mainly by negoiaion, as here are no binding conracs beween chain acors and markes are open access. Capure cafish currenly a low volume is mainly consumed locally, while some farmed cafish finds is way ino he regional expor marke. 8

3 Mehodology Daa The ime series daa used in his analysis consis of monhly farm-raised/aquaculure and wildharves cafish prices from January 006 o Augus 03. The daa are aken from secondary source daa recorded by he Aquaculure Managemen Consulan (03). All prices, expressed in Uganda Shillings per kilogram, were deflaed using a consumer price index (CPI) deflaor o adjus for inflaion over he period covered. CPI daa were obained from he Uganda Bureau of Saisics (Uganda Bureau of Saisics, 03). Table presens he characerisics of he daase. Boh farm-raised and wild-harves price series are moderaely skewed o he righ, indicaing ha he daa have longer righ ails han lef ails. The kurosis values are lower han 3, implying ha he series disribuion produces fewer and less exreme ouliers han does he normal disribuion. The large value of sandard deviaion in mean price suggess wide flucuaions in he cafish price series. I is always good pracice o plo he ime series while searching for poenial ouliers, rends, srucural breaks, and he general characerisics of he daa-generaing process. Visual inspecion of he series (Figure ) clearly suggess ha volailiy was presen a several poins in ime. Farm-raised cafish prices are more unsable, paricularly beween 008 and 0. Table. Descripive Saisics Farm-Raised Wild-Harves Mean 5,995 3,8 Maximum 8, 4,88 Minimum 4,53,899 Sd. Dev Skewness Kurosis Observaions 9 9 Saionariy Tess The basic assumpion in ime series economerics is ha he underlying series is saionary in naure. The es for saionariy of he cafish price series under consideraion was done using Augmened Dickey Fuller (ADF) and Phillips-Perron (PP) es saisics. The ADF es relies on parameric ransformaion of he model, while he PP es uses nonparameric saisical mehods o ake care of he serial correlaion in he error-erms. The opimal number of lags was deermined using he Schwarz crierion informaion crieria. The ADF and PP ess were found o be insignifican a he 5% level of significance for boh price series (Table ), confirming he non-saionariy of he level series. However, on differencing he series once, boh ess were found o be highly significan a he % level, confirming saionariy. Therefore, he need of firs differencing of he series was fel for proper modelling of he cafish price series. 83

4 8,000 7,000 Price per kg (UG Shs.) 6,000 5,000 4,000 3,000,000 Farm-raised Wild-harvesed, Time period (Monhly) Figure. Price Movemen in he Cafish Supply Chain. Source: Aquaculure Managemen Consulan (03). Table. Saionariy and LM Tes Resuls. Farm-Raised Wild-Harves Levels ADF -0.3 [] 0.4 [0] PP -0.8 (7).34 () Firs Difference ADF -3.67*** [0] -0.9*** [0] PP -9.75*** (8) -.8*** (5) LM Tes F-sa Obs*R Prob Noes: [ ] represens lags while ( ) represens bandwidh, 0.0 criical values: -.59, Lag Lengh- based on SIC, maxlag=. Price Volailiy Volailiy refers o variaions in economic variables over a period of ime. Large variaions in prices ha do no reflec marke fundamenals become problemaic because hey can lead o incorrec decisions. The focus in his sudy was on variaions in he cafish price series over ime. The series are said o be volaile when a few error erms are larger han he ohers and are responsible for he unique behavior of he series. This phenomenon is known as 84

5 heeroscedasiciy. The popular and non-linear model for dealing wih heeroscedasiciy is he auoregressive condiional heeroscedasic model proposed by Engle (98) and exended by Bollerslev (986). Auoregressive Condiional Heeroscedasic (ARCH) Models The ARCH(q) model for he series { ε } is defined by specifying he condiional disribuion of ε given he informaion available up o ime. Leing ψ denoe his informaion, i follows ha ψ consiss of he knowledge of all available values of he cafish series and anyhing ha can be compued from hese values (e.g., innovaions, squared observaions, ec.). ε given he I can be said ha he process { } available informaion ψ is ε is ARCH(q) if he condiional disribuion of { } (.) ε ψ ~ N(0, h ) and (.) h = a + a i ε, 0 i q i= where a > 0, a 0 0 i of { ε } given he nex observaion { } q for all i and < a i i= ( 0, h. Equaion (.) implies ha he condiional disribuion ψ is normal, N ). In oher words, given he available informaion ψ, ε has a normal disribuion wih a (condiional) mean of E[ ε / ψ ] = 0, and a (condiional) variance of var[ ε / ψ ] = h. Equaion (.) specifies he way in which he condiional variance h is deermined by he available informaion. Noe ha h is defined in erms of squares of pas innovaions. This, ogeher wih he assumpions ha a > 0 and 0 0, guaranees ha h is posiive, as i mus be since i is a condiional variance. a i The GARCH Model The GARCH model proposed by Bollerslev (986) is an exension of he ARCH model, in which condiional variance is also a linear funcion of is own lag. In his sudy, he GARCH (,) model was employed o measure he exen of price volailiy in he cafish price series. The model was specified as (.) Y = X θ + ε (.) σ ω + α + βσ = 85

6 where he mean equaion given in equaion (.) is wrien as a funcion of exogenous variables wih an error erm. Since σ is he one-period ahead forecas variance based on pas informaion, i is called he condiional variance. The condiional variance equaion specified in equaion (.) is a funcion of hree erms: a consan erm, ω ; news abou volailiy from he previous period, measured as he lag of he squared residual from he mean equaion, (he ARCH erm); and las period's forecas variance, σ (he GARCH erm), while he error in he squared residuals is given by v = σ. Subsiuing for he variance in he variance equaion and rearranging he erms, he model can be wrien in erms of he errors as (.3) = ω + α + β ) + v βv. ( Thus, he squared error follows a heeroscedasic ARMA (,) process. The auoregressive roo ha governs he persisence of volailiy shocks in he price series is he sum of α and β. The ARCH parameer corresponds o α and GARCH parameer o β. If he sum of he ARCH and GARCH coefficiens is close o, his implies ha volailiy shocks are quie persisen. Resuls The firs sep in he specificaion and selecion of he model was o es for ARCH effecs in he series. This was accomplished using he ARCH Lagrange muliplier (LM) es on he square of he residuals obained afer fiing he ARIMA model on he wo price series. The idea here was o es wheher residuals do in fac remain consan. The resuls es (Table ) revealed he presence of he ARCH effec for boh price series. The implicaion of hese resuls was ha boh cafish price series were volaile and needed o be modeled using he Generalized ARCH model (GARCH). The esimaed univariae GARCH (,) parameers for he variance equaions are repored in Table 3. In his model, he sum ( α + β ) measures he degree of volailiy persisence in he marke, which reveals he degree of efficiency in he marke. If a marke is compleely efficien i should immediaely correc o any shock. The observed volailiy in he monhly cafish price series of wild-harves supply chain revealed ha boh he values of he firs-order auoregressive erm ARCH (α = 0.458) and he value of he firs-order moving average erm GARCH ( β = 0.404) were saisically significan a he % level. The observed volailiy coefficien (α + β ) was quie persisen of he order of 0.86 (Table 3). Similarly, boh ARCH and GARCH erms (α = 0. and β = 0.780, respecively) for he monhly cafish price series of farm-raised supply chain were saisically significan a he 5% and % levels, respecively, and he persisen volailiy was measured a he order of The quie large value of he GARCH erm compared o ARCH erm in he farm-raised supply chain shows reasonably long persisence of volailiy in he price series over he sudied period. The resuls sugges ha he wild-harves cafish price series display a larger degree of efficiency han 86

7 Table 3. GARCH (, ) Esimaes. Variable Coefficien Sd. Error Prob. Volailiy Half-Life Variance Equaions (α + β ) (Monh) Wild-Harves Consan ** ARCH *** GARCH *** Farm-raised Consan ARCH 0.9** GARCH *** Noes: Double and riple aserisks (**, ***) indicae significance a he 5% and % levels. he aquaculure price series. The observed degree of persisence in he respecive supply chains was used o esimae he half-life of a volailiy shock, [log(0.5)/log( +β)], which measures he ime i akes for a shock o fall o half of is iniial value. The resuls (Table 3) show half-life esimaes of 4.7 monhs for he wild-harves cafish supply chain and 89.7 monhs for farmraised supply chain. Conclusion Price levels of farm-raised and wild-harves cafish supply chains in Uganda have increased over he period of sudy. The large value of sandard deviaion in mean price suggess wide flucuaions in cafish price levels during Empirical resuls of he GARCH model revealed ha he value of firs-order auoregressive erm ARCH and he value of firs-order moving average erm GARCH were significan for boh supply chains. The quie large value of he GARCH erm in comparison o he ARCH erm in he aquaculure supply chain showed reasonably longer persisence of volailiy. Based on hese resuls, a reliable marke informaion sysem and up-o-dae informaion on supply, demand, and socks may help in reducing price volailiy. Governmen acion is needed o suppor effors geared a increasing he capaciy of he fisheries secor o underake sysemaic monioring of fish producion, improved shor-run producion forecass, and marke analysis. As noed by previous sudies, adequae fish sock is a necessary componen of a well-funcioning marke, paricularly o smooh ou seasonal flucuaions and ime lags in he fish rade (FAO e al., 0). Limiaion: The daa used in his analysis are for a period of almos eigh years, a limied se of daa o which o apply GARCH models. The findings should herefore be reaed cauiously. Acknowledgemens This research is a componen of AquaFish Innovaion Lab, suppored by USAID (CA/LWA No. EPP A ) and by conribuions from paricipaing insiuions. The AquaFish succession number is 463. The opinions expressed herein are hose of he auhor and do no necessarily reflec he views of AquaFish or he U.S. Agency for Inernaional Developmen. 87

8 References Aquaculure Managemen Consulan. 03. Monhly Cafish Prices. Kampala, Uganda: Aquaculure Managemen Consulan, Ld Bollerslev, T Generalized Auoregressive Condiional Heeroscedasiciy. Journal of Economerics 3: , J.O., and M. Ssebisubi. 04. Price Inegraion in he Farmed and Wild Fish Markes in Uganda. Fisheries Science 80(6): Deparmen of Fisheries Resources. 0. Annual Repor 00/0. Governmen of Uganda, Minisry of Agriculure, Animal Indusry, & Fisheries. Available online: hp:// ANNUAL REPORT 0.pdf Engle, R. F. 98. Auoregressive Condiional Heeroscedasiciy wih Esimaes of he Variance of UK Inflaion. Economerica 50: FAO, IFAD, IMF, OECD, UNCTAD, WFP, he World Bank, he WTO, IFPRI, and he UN HLTF. 0. Price Volailiy in Food and Agriculural Markes: Policy Responses. hp:// n_food_price_volailiy.pdf Mino, N. 04. Food Price Volailiy in Sub-Saharan Africa: Has I Really Increased? Food Policy 45: Ngare, L., F. Simowe, and J. Massingue. 04. Analysis of Price Volailiy and Implicaions for Price Sabilizaion Policies in Mozambique. European Journal of Business and Managemen 6(): Sukai, M. 03. Measuring Maize Price Volailiy in Swaziland using ARCH/GARCH approach. MPRA Paper No Available online: hps://mpra.ub.unimuenchen.de/5840/. Uganda Bureau of Saisics. 03. Saisical Absracs Available online: hp:// 88