沙漠研究 22-1, 45-48 (2012 ) Journal of Ard Land Studes "ICAL 1 / DT X" -Refereed Paper- Techncal Effcency of Olve Growng Farms n Tunsa and Potental Demand for Olve Ol n Japan Kench KASHIWAGI 1), Atsush KAWACHI 1), Sam SAYADI 2) and Hroko ISODA 1) Abstract: In the face of growng nternatonal compettveness, one of the man challenges for olve seeds and olve ol sector n Tunsa s the enhancement of productvty and the level of technology as well as open up non-tradtonal, emergng markets such as Japan. Ths study nvestgates frm level techncal effcency of producton and ts determnants n a sample of olve growng farms n Tunsa. It also explores the Japanese consumer behavor on olve ol and ther potental demand n Japan. Results suggest olve growng farms n Tunsa can ncrease ther producton on an average by 29.8% through more effcent use of technology and producton nputs. The rrgated olve farms are effcent, and ntroducton of the rrgaton technque and the selecton of varety such as Chétou and Chemlal sgnfcantly contrbute to upgrade the techncal effcency. Japanese consumers prefer on extra-vrgn and vrgn to olve ol, sweet taste and green color products wth the prce between 800 to 1000 Yen per 500 ml, and showed hgher preference on olve ol that contans hgh concentraton of polyphenol. Key Words: Consumer behavor, Japan, Olve growng farms, Techncal effcency, Tunsa 1. Introducton The olve ol consttutes an mportant part of the Tunsan export, accountng for 120,000 tons per year and representng 70% of total producton. Tunsa occupes the fourth poston as olve ol exporter preceded by Span, Italy and Greece; however most of the olve ol s exported by bulk (99%). Whle Tunsan government sought to ncrease bottled olve ol exports up to 10% of total exports, ths goal has not been acheved. On the other hand, olve ol consumpton n Japan showed an ncreasng trend durng the last two decades due to detary and health concerns. Accordng to the offcal statstcs, the volume of mport of olve ol n Japan skyrocketed after the md-90s. The orgn of mport was dversfed durng the 90s ncludng Greece, France, Portugal, US, Turkey and Tunsa, however almost 99% of the mport was from Span and Italy (Japan Tarff Assocaton, 2007). In the face of growng nternatonal compettveness, tradtonal olve ol producng countres such as Italy, Span and Tunsa may have nterest to strengthen ther penetraton n the Japanese market. For Tunsa as one of the man exporters of olve ol, the man challenge for the olve ol sector s to enhance the level of productvty and effcency as well as to expand export destnaton to non-tradtonal emergng markets such as Japan. Investgaton on the techncal effcency of olve ol producton as well as ts marketng for desgnng preferable product may provde valuable nsghts nto potental Correspondng Author: kashwag.kench.fn@u.tsukuba.ac.jp mprovement of productvty and export capacty. A few studes are found n the lterature of productvty analyss on producton of olve seeds and olve ol. In Tunsa, Lachaal et al. (2005) analysed productvty and techncal effcency of the olve growng farms. Although the majorty nvestgated the level of techncal effcency tself, few studes has focused on the effect of non-conventonal nputs such as rrgaton and varetes as factors to mprove techncal effcency. Regardng the consumpton analyss on olve ol, few studes, except Mtmet et al. (2009), focus on Japanese olve ol market. Whle Mtmet et al. (2009) nvestgated the olve ol consumer behavour; healthy attrbute of olve ol was not explctly examned. Abaza et al. (2005) and Yamada et al. (2008) detected bologcal functons for cosmetc, medcal and medcnal use n the extract of Tunsan olve ol n partcular by applyng the boassay technque. Ths study nvestgates frm level techncal effcency of producton and ts determnants n a sample of olve growng farms n Tunsa. Effects of the ntroducton of rrgaton and dfference n varetes on techncal effcency are explctly examned. Also, ths study examnes Japanese olve ol consumer behavour by usng the conjont analyss technque. Healthy attrbute of olve ol ndcated by the components of polyphenol was ntroduced n the product desgn. The objectve s to dentfy the sources of techncal effcency of olve growng farmers n the frst. Second, the factors determned the varatons of techncal effcency among the olve growng farmers are dentfed. Thrd, consumer behavor on olve ol wth dfferent attrbutes s analyzed n Japan. 1-1-1 Tennoda, Tsukuba, Ibarak, 305-8572, Japan Tel: +81-29-853-3982 Fax: +81-29-853-5776 1) The Allance for Research on North Afrca (ARENA), Unversty of Tsukuba, Japan 2) Center of Botechnology of Sfax (CBS), Tunsa
2. Models 2.1. Techncal effcency In ths paper, the Battese and Coell (1995) model of stochastc fronter producton functon s appled n the context of cross-sectonal data. Assumng a Cobb-Douglas form, a standard producton functon s wrtten as follows: Y βk βl βm βo v u = α K L M O e, (1) where α = exp(β 0 ); Y denotes olve producton expressed n ton; K s the number of agrcultural machnery (tractors, cultvators, sprayer) utlzed n the producton; L denotes the number of days of famly and hred labour used; M s the volume of ntermedate nputs (costs of fertlzer, pestcdes, ploughng, water and other expenses measured by Tunsan Dnar) devoted to olve producton; O denotes the number of olve trees; β I ( = 0, k, l, m, o) s an unknown parameter to be estmated; v refers to statstcal random dsturbance terms, assumed to be an ndependently and dentcally dstrbuted N (0, σ v 2 ). u represents non-negatve random varables, assumed to be d as N (0, σ u 2 ) wth truncatons at zero. Ths model s specfed n logarthmc form, as n (2). Y = β + β ln K + β ln L + β ln M + β ln O + v u, (2) ln 0 k l m o In ths specfcaton, exp ( u ), vares between 0 and 1, s a measure of techncal effcency of the th farm. Followng Battese and Coell (1995), the techncal effcency of producton of the th farm s defned by the rato of the observed output to the correspondng fronter output: TE = exp( -u ), 0 TE 1. (3) To nvestgate the relatonshp between techncal effcency and nonconventonal nputs, the techncal effcency of th farm can be endogensed as follows: LND D D D + w, (4) TE = δ 0 1IRR 2 EDU 3TEC 4 5 CT 6 CM 7 OF where δ ( = 0 7) s an unknown parameter to be estmated; IRR s the share of rrgated olve trees to total; EDU s the dscrete varable that represents the level of educaton of employee on a scale of 1 to 5 (1: llterate, 2: quranc educaton, 3: prmary, 4: secondary, 5: unversty); TEC s the dscrete varable that represents the level of technology on a scale of 1 to 3 (1: tradtonal, 2: modern, 3: advanced); LND denotes total area of land devoted to olve cultvaton; D CT s the dummy varable of Chétou varety that equals 1 f the farm cultvates ths varety and zero f otherwse; D CM s the dummy varable of Chemlal varety that equals 1 f the farm cultvates ths varety and zero f otherwse; D OF s the dummy of off-farm that equals 1 f the farm engages n off-farm actvtes and zero f otherwse; w represents the error term. 2.2. Conjont analyss The conceptual foundaton of conjont analyss arses from Lancaster (1966). Krystalls and Ness (2003), emprcally appled ths framework to the analyss on olve ol consumpton. Followng Mtmet et al. (2009), an ordered probt model has been selected for estmaton. The model s bult around a latent regresson as follows: y = λ + ε, (5) x where y denotes unobserved dependent varable; λ s a vector of unknown parameters to be estmated; x s the vector of ndependent varables, whch corresponds n ths study to the olve ol attrbutes; ε s the error terms vector assumed to be a normal dstrbuton. As y s unobserved, what we do observe s the rankng varable y : y = 1 f y µ 1, y = 2 f µ 1 y µ 2, y = 3 f µ 2 y µ 3,.. y = 8 f µ 7 y, (6) where µ j (0 µ 1 < µ 2 < < µ j-1 ) are unknown parameters (thresholds) to be estmated wth λ. 3. Data 3.1. Techncal effcency of olve growng farms n Tunsa The data used n ths study were drawn from a survey conducted n Tunsa n February and Mach 2011. The olve growng farms surveyed were located n the northern governorate (Beja) of the country. The olve growng farms were randomly selected and the questonnares were drected to the owner of olve growng farms. In total, 40 questonnares were completed. 3.2. Olve ol consumer behavor n Japan The survey was conducted by the NTT Resonant Inc. whch provdes nternet research servces n Japan. Randomly selected 3127 montors were questoned through the web-questonnare durng the last week of March 2011. A total sample of 532 persons responded to the questonnare. The conjont cards were descrbed by sx attrbutes: country of orgn, prce, olve ol type, taste, colour and component (Table 1). The choce of these attrbutes was based on the analyss on Tunsan olve ol by Abaza et al. (2005) and Yamada et al. (2008). The combnaton of the sx attrbutes correspondng to a total number of 384 hypothetcal products. To reduce the number of cards, an orthogonal desgn procedure (SPSS. 13) s used to lead a fnal desgn wth 16 products. To avod respondents fatgue, we blocked the 16 cards nto two groups wth 8 cards each.
Table 1. Selected olve ol attrbutes and ther correspondng levels. Attrbutes Regon of orgn Prce (500 ml bottle) Olve ol type Taste Colour Component Levels Italy Span Tunsa Medterranean 600 Yen 800 Yen 1000 Yen 1200 Yen Extra vrgn Vrgn Olve ol(refned) Btter Sweet Green Yellow Hgh polyphenol Low polyphenol Table 2. Parameter estmates and t-values of the stochastc fronter producton functon of a sample of olve growng farms n Tunsa.,, ndcate sgnfcant at the 10% level, 5% level, 1% level, respectvely. Independent varables Estmates t-values Constant -6.130-22.272 ln K 0.070 1.703 ln L -0.107-0.882 ln M 0.207 3.211 ln O 1.021 8.949 Varance parameters σ 2 0.273 3.423 γ 0.940 16.632 Log-lkelhood -10.002 4. Results and Dscusson 4.1. Techncal effcency of olve growng farms n Tunsa Maxmum lkelhood estmates of the parameters of the stochastc fronter producton are obtaned usng the computer program FRONTIER 4.1. In Table 2, estmated coeffcents of number of agrcultural machnery, ntermedate nputs and the number of olve trees are postve and statstcally sgnfcant. The negatve coeffcent of the labour nput s nsgnfcant. The estmate of varance parameter, γ, s postve and statstcally sgnfcant at 1% level. Thus, the stochastc fronter functon s emprcally justfed. In Table 3, the average level of techncal effcency s 70.2%, rangng from a mnmum of 31.6% to a maxmum of 93.9%. Twenty three farms represented 57.5% n the sample are more effcent than the average; however 7.5% of the sample shows relatvely neffcent wth the scores showed less than 40%. These results suggest the possblty of enhancng effcency for upgradng the productvty and compettveness. The OLS estmaton results to nvestgate the relatonshp between techncal effcency and nonconventonal nputs are presented n Table 4. The estmated coeffcent of the rate of rrgated olve trees s postve and statstcally sgnfcant. Ths result clearly ndcates that the rrgaton of olve trees contrbutes to ncreases n techncal effcency. The postve sgn of the dummy varable of Chétou (D CT ) and Chemlal Table 3. Frequency dstrbuton of techncal effcency for a sample of olve growng farms n Tunsa. Techncal effcency (%) Number of Farms Percentage 0 < TE 40 3 7.50 40 < TE 70 14 35.00 70 < TE 100 23 57.50 Mean effcency 70.2 Mn. effcency 31.6 Max. effcency 93.9 Table 4. Parameter estmates and t-values of sources of varaton n techncal effcency of a sample of olve growng farms n Tunsa.,, ndcate sgnfcant at the 10% level, 5% level, 1% level, respectvely. Independent varables Estmates t-values Constant 0.460 2.973 IRR 0.476 3.510 EDU -0.085-2.311 TEC 0.071 1.345 LND -0.001-1.716 D CT 0.286 2.385 D CM 0.357 2.250 D OF 0.167 2.523 Adjusted R 2 0.359 F-statstcs 4.13 Observatons 40 Techncal effcency (%) 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 5 10 15 20 Producton per olve tree (kg) Fg. 1. The relaton between productvty and techncal effcency of a sample of olve growng farms n Tunsa. TE n non-rrgated olve farms TE n rrgated olve farms (D CM ) confrm the postve effect of cultvaton of these varetes on the techncal effcency. The postve sgn of the estmated coeffcent of the off-farm dummy (D OF ) ndcates the nvolvement wth non-agrcultural actvtes has postve effect on techncal effcency. The negatve sgn of area of olve cultvaton (LND) ndcates the effcency tends to declne as the area become large. Fgure 1 shows the relaton between the level of techncal effcency and productvty. Whle the level of techncal effcency and productvty vary across farms, t s clear that the rrgated olve farms are effcent. Comparng the most effcent group between the rrgated and the non-rrgated, the producton per olve tree of the rrgated s hgher than non-rrgated. Ths result mples that the rrgaton contrbutes to further enhance the productvty.
Table 5. Parameter estmates and standard errors of the ordered probt model of a sample Japanese consumers.,, ndcate sgnfcant at the 10% level, 5% level, 1% level, respectvely. Model 1 Model 2 Independent varables Estmates Standard error Estmates Standard error Constant -0.941 0.067-0.674 0.344 Type Vrgn 0.484 0.043 0.480 0.043 Extra vrgn 1.104 0.056 1.088 0.554 Orgn Italy 0.027 0.049 0.029 0.049 Span 0.001 0.050 0.002 0.049 Tunsa -0.093 0.051-0.095 0.051 Taste: btter -0.283 0.036-0.285 0.036 Colour: green 0.192 0.036 0.188 0.036 Component: hgh 0.465 0.036 0.463 0.036 polyphenol Prce 1.048 0.802 Prce 2-0.001 0.000 800 JPY 0.046 0.049 1000 JPY -0.223 0.050 1200 JPY -0.318 0.051 µ 2-0.404 0.065-0.139 0.344 µ 3 0.005 0.065 0.269 0.344 µ 4 0.372 0.065 0.636 0.344 µ 5 0.740 0.066 1.003 0.344 µ 6 1.149 0.067 1.412 0.344 µ 7 1.692 0.069 1.954 0.344 Observatons 3504 3504 LR1 χ 2 (11) 873.2 LR2 χ 2 (10) 863.5 4.2. Olve ol consumer behavor n Japan Parameter estmates and standard errors of the ordered probt model are gven n Table 5. The estmated coeffcents of the both models showed clear results. Frst, the respondents prefer vrgn and extra vrgn olve ol to olve ol (refned). Second, respondents prefer a sweet taste olve ol rather than a btter one. Thrd, respondents show also a hgher preference for a green coloured olve ol compared wth a yellow one. It s noteworthy that the postve and statstcally sgnfcant estmaton of the coeffcent of the hgh polyphenol ndcates respondents prefer olve ol that contans hgher concentraton of polyphenol. The coeffcents correspond to the Italy and Span orgns are nsgnfcant whle that of Tunsa presents the negatve sgn. Fnally, respondents preference tends to ncrease when prce ncreases from 600 and 800 Yen, but t decreases when prce ncreases at the level more than 1000 Yen. 5. Concluson Results of the estmaton of stochastc fronter producton suggest that olve growng farms n Tunsa can ncrease ther producton on an average by 29.8% through more effcent use of technology and producton nputs. The rrgated olve farms are most effcent than non-rrgated farms. It s also mpled that the ntroducton of the rrgaton technque and the selecton of varety such as Chétou and Chemlal sgnfcantly contrbute to upgrade techncal effcency. Regardng the results of conjont analyss, Japanese consumers prefer on extra-vrgn and vrgn to olve ol, sweet taste and green color products wth the prce between 800 to 1000 Yen per 500 ml. Whle the preference for the country of orgn ndcated at Tunsa s less compare wth the label of Medterranean, the consumers showed hgher preference on olve ol that contans hgh concentraton of polyphenol. The olve ol made of the varety of Chétou that contans plenty of polyphenol would have a potental to be exported to ths emergng market. Acknowledgement The authors gratefully acknowledged that ths research was supported by the JICA-JST Scence and Technology Research Partnershp for Sustanable Development Project: Valorzaton of Bo-resources n Sem-Ard and Ard Land for Regonal Development, and supported by the JSPS Asa Afrca Scence Platform Program: Establshment of Integratve Research Base by Humantes and Scences on Valorzaton of Useful Plants for Regonal Development n North Afrca. References Abaza L., Taamall W., Ben Temme S., Daoud D., Guterrez. F., Zarrouk M. (2005): Natural Antoxdant Composton as Correlated to Stablty of Some Tunsan Vrgn Olve Ols. Rvsta Italana delle Sostanze Grasse, 82: 12-18. Battese G.E., Coell T.J. (1995): A Model for Techncal Ineffcency Effects n a Stochastc Fronter Producton Functon for Panel Data. Emprcal Economcs, 20: 325-332. Japan Tarff Assocaton (2007): Japan Export & Imports: Commodty by Country, 2006, March, Japan Tarff Assocaton: Tokyo. Krystalls A., Ness M. (2003): Motvatonal and Cogntve Structures of Greek Consumers n the Purchase of Qualty Food Products. Journal of Internatonal Consumer Marketng, 16(2): 7-36. Lachaal L., Karray B., Dhehb B., Chebl A. (2005): Techncal Effcency Measures and Its Determnants for Olve Producng Farms n Tunsa: A Stochastc Fronter Analyss. Afrcan Development Revew. 17(3): 580-591. Lancaster K. (1966): A New Approach to Consumer Theory. Journal of Poltcal Economy, 74(2): 132-157. Mtmet, N., Kashwag K., Zabet L., Nagak M. (2009): Japanese Olve Ol Consumer Behavor: A Conjont Analyss Approach. In Sato eds., Consumer Behavour. Nova Scence Publshers, Inc., New York, 27-47. Yamada P., Zarrouk M., Kawasak K., Isoda H.(2008): Inhbtory Effect of Varous Tunsan Olve Ols on Chemcal Medator Release and Cytokne Producton by Basophlc Cells. Journal of Ethnopharmacology, 116, 279-287.