Productive efficiency of tea industry: A stochastic frontier approach

Transcription

1 African Jornal of Biotechnology Vol. 9(5), pp , 1 Jne, 1 Available online at ISSN Academic Jornals Fll Length Research Paper Prodctive efficiency of tea indstry: A stochastic frontier approach Md. Azizl Baten 1 *, Anton Abdlbasah Kamil 1 and Mohammad Anaml Haqe 1 School of Distance Edcation, Universi Sains Malaysia, 118 USM Pla Pinang, Malaysia. Department of Statistics, ShahJalal Universy of Science and Technology, Sylhet-3114, Bangladesh. Accepted 31 May, 1 In an economy where recorses are scarce and opportnies for a new technology are lacking, stdies will be able to show the possibily of raising prodctivy by improving the indstry s efficiency. This stdy attempts to measre the stats of technical efficiency of tea-prodcing indstry for panel data in Bangladesh sing the stochastic frontier prodction fnction, incorporating technical inefficiency effect model. It was observed that Translog Prodction Fnction is more preferable than Cobb-Doglas Prodction Fnction. The stdy estimates that the average technical efficiency of tea prodcing indstries in Bangladesh is 59%. Therefore, the reslts indicated that there is a great potential exists for tea indstry to frther increase the vale added by 41% sing the available inpt, technology and efficiency improvement, thereby redcing the cost of prodction. The stdy identifies that the mean efficiency of tea indstries for vale added vary among the regions and year-wise mean efficiency seems to be nstable dring the stdy period and therefore, contined efforts to pdate technologies and eqipment are reqired in prs of efficiency in tea indstry. Key words: Technical efficiency, stochastic frontier, translog prodction, likelihood ratio test, tea indstry. INTRODUCTION The tea prodcing indstry has been tradionally regarded as one of the major agro-based labor intensive indstry and occpies an important role in the national economy of Bangladesh. The role of Bangladesh tea indstry in global context is insignificant. It is only 1.68% of the global tea prodction and.58% of the world tea export. It seems that s export is gradally declining. If this trend contines, Bangladesh will trn into a tea importing contry by 15 (Monjr, 4; Mahmd, 4). As a reslt, international comparisons of the tea indstry s efficiency have been of great interest to firms in the indstry as well as policymakers. The large tea prodcing contries like India and Sri Lanka prodce more than Bangladesh, where India and Sri Lanka s prodction level is 16 and 1 times higher than Bangladesh (BCS, ). It was fond that in 1998, on an average only 1,145 kg of tea was prodced per hectare in Bangladesh. Whereas, in the same year, prodction level per hectare *Corresponding athor. baten_math@yahoo.com. Tel: Fax: in India and Sri Lanka was 178 and 3 kg, respectively (Majmder, 3). The concept of the technical efficiency of firms has been pivotal for the development and application of econometric models of frontier fnctions. Althogh technical efficiency may be defined in different ways [see, example, Fare et al. (1985)], we consider the definion of the technical efficiency of a given firm (at a given time period) as the ratio of s mean prodction (condional on s levels of factor inpts and firm effects) to the corresponding mean prodction if the firm tilized s levels of inpts most efficiently [see Battese and Coelli, 1988]. Efficiency is an important factor of prodctivy growth as well as stabily of prodction in developing agricltral economics. In view of slow growth and increasing instabily in tea prodction in Bangladesh (Monjr, 4), the tea economy of Bangladesh is expected to be benefed to a great extent from the stdy on technical efficiency stdies. Estimates on the extent of inefficiencies cold help decide whether to improve efficiency or to develop new technology to raise tea prodctivy in Bangladesh. There are some stdies that have been carried ot to analysis for the measrement of efficiency of tea indstries.

2 Baten et al. 389 They inclde among others: Hazarika and Sbramanian (1999) for Asam tea indstry in India; Ariyawardana (3) for vale added tea prodcers in Sri Lanka; Basnayake and Gnaratne (), and Rohan Jayatilake (9) for tea small holdings in Sri Lanka; Baten et. el. (9) for seven tea regions of Bangladesh; and Nghia Dai Tran (9) for different tea prodction systems in Thai Ngyen Province. Besides Daniela et al. (8), and Fahr and Snde (5) investigated empirically the spatial variation of prodctivy across Brazilian regions applying stochastic frontier analysis to manfactring data. Haqe (6, 7) examined and compared the vale chain models that are adopted by the tea indstries of Bangladesh and Japan sing some descriptive statistics analysis. These stdies do not adopt a stochastic frontier model, which is generally thoght as an essential for prodctivy analysis and for measring technical efficiency of tea indstry. The stochastic frontier prodction fnction, which was independently proposed by Aigner et al. (1977), Meesen and van den Broeck (1977) has been a significant contribtion to the econometric modeling of prodction and the estimation of technical efficiency of firms. The stochastic frontier involved two random components, one associated wh the presence of technical inefficiency and the other being a tradional random error. Applications of frontier fnctions have involved both cross-sectional and panel data. These stdies have made a nmber of distribtional assmptions for the random variables involved and have considered varios estimators for the parameters of these models. Srvey papers on frontier fnctions have been presented by Forsod et al. (198), Schmidt (1986), Baer (199) and Battese (199), the latter article giving particlar attention to applications in agricltral economics. Beck (1991) and Ley (199), have compiled extensive bibliographies on empirical applications of frontier fnctions and efficiency analysis. However, a few empirical researches have been carried ot to estimate the technical efficiency of the tea indstries in Bangladesh sing stochastic frontier model. Therefore, there is a great need to research the prodction efficiency of the tea indstries, which may contribte largely to the present low performance of the tea indstry in Bangladesh. The aim of this stdy is to estimate the inefficiency of tea indstries in Bangladesh and identify the factors casing technical inefficiency of tea indstries. In this stdy, an effort has been made to analyze in measring technical efficiency of tea indstry sing the stochastic frontier prodction fnction model specified by Battese and Coelli (1995), for the panel data. To determine the sorces of inefficiency to improve the existing sations in tea indstry are also of or interest. MATERIALS AND METHODS In stochastic frontier analysis, the assmption is that the prodction fnction of the flly efficient firm is known. Fried, Lovell et al., (1993), have shown that econometric approaches like the stochastic frontier analysis can distingish the effects of noise from the effects of inefficiency. Since one of the objectives of this research is to examine the prodction efficiency (scores) of tea indstries in Bangladesh, the Stochastic Frontier Analysis was selected as the tool to measre efficiency in this stdy. We employed a stochastic prodction frontier approach introdced by Battese and Coelli (1995), and can be wrten as ( V U ), i 1,,... Y = β X + =..., N ; t = 1,,..., T.....( 1) Where Y the logarhm of otpt of the h tea indstry is in tth period X is a vector of inpt qanties; V s random variables which are assmed to be i.i.d., N (, v ) and independent of U ; U s are non-negative random variables which are assmed to accont for technical inefficiency in prodction and to be independently distribted as trncations at zero of the N( µ, ) distribtion; where ; 1 p vector U = Z δ where; Z is a ( ) of variables which may inflence the inefficiency of tea indstry and δ is a ( p 1) vector of parameters to be estimated. The parameterization from Battese and Corra (1977), are sed, replacing and v wh = v + and the parameters are estimated by Maximm Likelihood approach. The Technical inefficiency effect U in the stochastic frontier model is specified as follows; U = Z δ + W Where, the random variable....( ), W follows trncated normal distribtion wh mean zero and variance, sch that the point of trncation is Z δ. Parameters of the stochastic frontier given by Eqation (1) and inefficiency model given by Eqation () are simltaneosly estimated by sing maximm likelihood estimation. After obtaining the estimates ofu the technical efficiency of the I, - th tea indstry at t - th time period is given by: ( ) ( δ ) TE = exp U = exp Z W...(3). Selecting the fnctional form of the prodction fnction In order to select the best specification for the prodction fnction (Cobb - Doglas or Translog), for the given data set, we condcted hypothesis tests for the parameters of the stochastic frontier prodction model sing the generalized likelihood - Ratio (LR) statistic defined by { ln[ L( H )/ L( H )]} = ln[ L( H )] ln[ L( )] { }. λ = 1 H1 (4) Where [ L( )] ln H the vale of the log likelihood fnctions for the

3 381 Afr. J. Biotechnol. stochastic frontier is estimated by pooling the data for all the seven regions nder nll hypothesis, and ln[ L ( H 1 )] is the sm of the vales of the log - likelihood fnctions for the seven stochastic prodction fnctions (North Sylhet + Jri + Lngla + Man-Doloi + Balisera + Lskerpore + Chtagong) estimated separately nder alternative hypothesis. Specification of the Stochastic Frontier Translog (Vale Added) Model The fnctional form of the stochastic frontier Translog prodction model is defined as: ln( Y 1 ) = β + β T + β ln A + β ln A β ln L β ln A 1 T + β lnl 1 ln L T + ( β T + β ln A + β ln L ) 11 + V U (5), Where, the sbscripts i and t represent the i-th tea indstry and the t-th year of observation, respectively; i = 1,,...,7 ; t = 1,,...,15 ; Y denotes the otpt variables (Vale added) of the h tea indstry in the t-th period in vales (taka); T represents time; A denotes area of h tea indstry in the t-th period; L represents labor of h tea indstry in the t-th period; ln refers to the natral logarhm; the β i s are nknown parameters to be estimated; N, v and U follows a trncations at zero of the N( µ, ) distribtion and garantees inefficiency to be posive only. V follows ( ) Identifying sorces of technical inefficiency and hypothesis tests The tea indstry specific inefficiency is considered as a fnction of some explanatory variables and the inefficiency effects model is defined as: U = δ + 1 δz1 + δz + 3 δz3 + W...(6), δ is the intercept term and ( j 1,,3) δ = is the where j parameter for the j-th explanatory variable and Z 1= Temperatre, Z = Rainfall, Z 3= Herfindahl-Hirschman index. The hypothesis tests are obtained sing the generalized likelihoodratio test statistic (4). This test statistic is assmed to be asymptotically distribted as mixtre of chi-sqare distribtion wh degree of freedom eqal to the nmber of restrictions involved. The restrictions imposed by the nll hypothesis are rejected when λ exceeds the crical vale (Taymaz and Saatci, 1997, p. 474). These are obtained by sing the vales of the log-likelihood fnctions for tea indstries and the stochastic frontier prodction fnction. Given the specification of the stochastic frontier prodction fnction, defined by (5), the nll hypothesis that technical inefficiency is not present in these model, is defined by H : γ =, where γ is the variance ratio, explaining the total variation in otpt from the frontier level of otpt attribted to technical efficiency and defined by γ = ( + v ) This is. done wh the calclation of the maximm likelihood estimates for the parameters of the stochastic frontier models by sing the compter program FRONTIER version 4.1 developed by Coelli (1996). If the nll hypothesis is accepted this wold indicate that is zero, hence that the U term shold be removed from the model, leaving a specification wh parameters that can be consistently estimated sing ordinary least sqare (OLS). Frther, the nll hypothesis that the technical inefficiency effects are time invariant defined as H : η =. If the nll hypothesis is tre, the generalized likelihood ratio statistic λ is asymptotically distribted as a chi-sqare (or mixed chi-sqare) random variable. Data description and variable constrction The data were collected from the varios isses of Annal Report of Bangladeshyio Cha Sansad (BCS) and International Tea Commtee (ITC) etc. Or stdy covers total Tea Indstry that is available, nder registered tea gardens of Bangladesh over the reference period 199 to 4. Vale added (Y) Vale added figres are sed in this stdy to represent otpt and is eqal to the vale of prodcts and is measred in vales (taka). This vale added figre is maniplated by the price of yield per hectare and is treated as gross prodction or gross otpt. To obtain the gross otpt series in constant prices, the yearly crrent vales were deflated by the indstry price index of the relevant year. In this analysis gross vale added (otpt) is dependent variable. Area (A) Area is one of the essential inpts in measring prodctivy. Gross fixed area nder tea is sed in this stdy. Labor (L) The nmber of employees directly or indirectly in prodction is sed in this stdy as a labor inpt. It covers all workers inclding administrative, technical, clerical, sales and prchase staff. Ths all prodction and non-prodction workers except temporary daily casals and on paid workers are inclded in the analysis. In brief, they inclde prodction workers, salaried employees, and working proprietors. The best measre of labor inpt is the nmber of hors worked. As no sch data are available for any indstry, employment figres were taken as the second measre and were weighted by the base year wage rate to obtain measre of labor inpt. Time (T) To find the prodctive efficiency of a i-th tea indstry over time we have sed time as the inpt variable. In this stdy we have sed data of 15 years from 199 to 4. Explanatory Variables which inflence the level of inefficiency is considered also in this stdy:

4 Baten et al Table 1. OLS and MLE estimates of stochastic frontier translog (Vale added) model. Variable Parameters Estimated OLS Estimates Estimated MLE Estimates Constant ** (1.448) * (.99) Time 1 (.16) (.17) Area ** (3.85 ) * (.748) Labor (1.881) * (.675) Time * Time 11 (.9 ) * (55.747) Area * Area ** (1.19) * (.547) Labor * Labor ** (.953) -.57 * (.568) Time * area 1 (.137) Time * Labor (.11) (.135) Area * Labor ** (.959).313 * (.545) Sigma - sqared.58 Log likelihood fnction N=15 and *, **, *** significance level at 1%, 5%,1% means insignificant, and vales in the parentheses indicate Standard Error. Temperatre (Z 1) Temperatre is sed as inflencing variables which are not deflated bt actal measrement and s n of measrement is Fahrenhe. Rainfall (Z ) Rainfall is sed as inflencing variables which are not deflated bt actal measrement and s n of measrement is millimeter. Herfindahl - Hirschman Index (Z 3) The Herfindahl-Hirschman index takes into acconts both the relative size and nmber of tea indstries. Mathematically, HHI is N described as follow: H H I = S i where N is the i = 1 nmber of indstries and S i is share of the i th tea. HHI is known as measre of competion which is measred as the sm of sqared of the otpt share of each tea indstry in the otpt of considered total tea indstries in Bangladesh. RESULTS AND DISCUSSION Selection of the translog prodction fnction We have tested the hypothesis whether the Translog prodction fnction is an adeqate representation of the data or not sing Eqation (5). The vales of the log likelihood for the Cobb-Doglas and Translog prodction frontiers are and 5.3, respectively. By employing Eqation (4), we have estimated the vales of Likelihood Ratio for the Cobb-Doglas and Translog prodction are and , respectively. These vales are χ compared wh the pper five percent points for ( 3,.5) and χ (9,.5) which are 3.85 and 1.5, respectively. Finally is conclded that the nll hypothesis H : β is strongly rejected and indicates that ij = Translog Prodction Fnction is more preferable than Cobb-Doglas Prodction Fnction. Estimating the stochastic frontier translog model The reslts of the Ordinary Least Sqare (OLS) and Maximm-likelihood Estimation (MLE), for the Translog prodction fnction as described in Eqation (5) are reported in Table 1. From the OLS estimation we have observed that a total of 4 coefficients ot of 9 are statistically significant at 5% level, indicating the importance of some of the interactions and non - lineary among variables. The direct effects of area, interaction effects of area and labor, sqare terms or second order parameters of area and labor are significantly different from zero. These implied that there exists having a major role in tea prodction. The area remains the single most important inpt wh an otpt elasticy of 5.983, followed by labor , respectively. Reasonably enogh, for a labor srpls economy, labor has the negative otpt (vale added) elasticy and is fond to be insignificant in the prodction process. This implies that labor does not affect the yield of the tea significantly. The variable time, second order parameter of time, interaction of time and labor and interaction of time and area are fond to be insignificant. So we can say that the area and labor wh interaction to time do not affect on the vale added (prodction) in tea indstries of Bangladesh. The sign of coefficients of all variables in Eqation (5), when estimated wh MLE techniqe are negative bt significant except area and s interaction wh labor,

5 381 Afr. J. Biotechnol. Table. Region wise mean efficiency of vale added for the selected tea regions in Bangladesh. Year North Sylhet Jry alley Lngla Man-doloi Balisera Lskerpore Ctg. Dist. Mean efficiency Mean interaction in between area and labor which are posive. In this analysis, is fond that the variable time and s interaction wh area and labor are insignificant. The direct effects of area, labor, sqare terms or second order parameters of area and labor and interaction of area and labor are significantly different from zero. This indicates that the rejection of the Cobb-Doglas model as an adeqate representation of Bangladesh Tea Indstry is jstified, becase the fnction is non - linear in some dimensions and there are important interactions among the variables. The variables area and labor appear to be the major determinants of tea prodction. However, area remains the single most important inpt wh an otpt (vale added) elasticy of 6.141, followed by labor , respectively. Reasonably enogh, for a labor srpls economy, labor has the negative otpt (vale added) elasticy and is fond to be insignificant at 5% level in the prodction process. The coefficients of interaction of time wh area and labor are.98 and -.7, respectively indicating that vale added (prodction) is explained only by 9 and 7% by these interaction variables. So from this reslt we may conclde that the area and labor wh time interaction have low otpt (vale added) elasticy. We have observed that the variable area shows significant affect for both OLS and MLE estimation of the Translog stochastic frontier prodction fnction. The coefficient on the time trend variable indicates that there is a negative technological progress bt declines downwards wh an annal rate of 1.1% per annm and the effect is nonlinear, as indicated by the significant coefficients of the sqared terms. Overall these findings spport the reslts of Baten et al. (9). Table reveals that the technical efficiency of Bangladesh Tea indstry dring the period 199 to 4 is fond to be.59 ranging from a minimm of.8 to a maximm of 1. for vale added for the selected tea regions. This implies that 59 percent of potential vale added is being realized by the tea indstry of Bangladesh. In the present stdy, none of the estimates had achieved zero level efficiency, while only Balisera in 3 and Lskerpore in 1998 achieved fll level efficiency (1%). The findings also sggest that 41% technical inefficiency exists in the vale added of tea. In other words, 59% of the tea estates were able to prodce on the prodction frontier, and 41% were off-frontier of varying degrees for vale added. There is wide variation in the technical efficiencies among the different tea prodcing region. However, the overall vale added efficiency for all regions is steadily increasing over time except the year 4 presented in Figre 1. The highest mean efficiency was in 1998 and was more than 8% which is 5% higher than previos year. The average efficiency in 4 was 4% lower than 3. We observed that the vale added efficiency in tea prodcing regions (like north Sylhet, Jry and Ctg.) in Bangladesh dring the period have lower efficiency comparable to other regions. The year wise vale added average prodctive efficiency has been illstrated also by Figre separately. Year wise vale added (prodctive) efficiency seems to be nstable dring the stdy period. The efficiency for vale added was least for the year 1997 bt s highest efficiency for the year It is hope that there has been a general improvement occrred after the year Following the Figre 3, we have also observed that Balisera and Man - Doloi are most efficient in prodcing tea wh 77 and 7% respectively. This reslt indicates that big size (measring their total area, technology) regions are comparatively more efficient. The lowest efficiency is in the Lngla (4%). From the analysis we

6 Baten et al Efficiency Year North Sylhet Jry Lngla M on-doloi Balisera Lskerpore Ctg Figre 1. Vale added efficiency in tea prodcing regions in Bangladesh, Average Efficiency(Year wise) Year Figre. Year-wise vale added average efficiency in tea indstries of Bangladesh, observed that the Lngla valley and North Sylhet are so far lowered efficient in prodcing tea comparing to other regions. May be these less efficient regions are concentrating in other services rather than vale added. Estimating the inefficiency effect model In order to investigate the determinants of inefficiency, we have estimated the technical inefficiency model described

7 3814 Afr. J. Biotechnol. Mean Efficiency north sylhet jry valley lngla man-doloi balisera lskerpore ctg dist Year Figre 3. Region - wise vale added average efficiency in tea indstries of Bangladesh, Table 3. Inefficiency effects model for vale added. Variable Parameters MLE Coefficients Constant * (.959) Temperatre 1.5 ** (.85) (.88) HHI * (.155) sigma-sqared.41 * (.8) gamma γ.999 * (.1) *, **, *** indicate significance level at 1,5 and1% consectively, vales in the parentheses indicate S.E. indicates Insignificance. in Eqation (6) presented in Table 3. The sign of coefficients of the variable HHI is negative bt significant impact on tea prodction. These indicate that HHI variable is inversely related wh inefficiency. The variable temperatre significantly contribtes to improved technical efficiency in tea prodction and this implies that temperatre shold be one of the major variables in order to improve technical efficiency in tea prodction. The sign of the coefficient of rainfall indicates that rainfall is less efficient althogh the coefficient is not statistically significant. Using the composed error terms of the stochastic γ = + which frontier model, is defined by ( ) is a measre of level of the inefficiency in the variance parameter and ranges between and 1. It is observed that the MLE estimate of γ is.999 wh estimated v standard error of.1. The vale of γ is significantly different from one indicating that random shocks are playing a significant role in explaining the variation in tea prodction, which is expected in tea prodction where ncertainty is assmed to be the main sorce of variation. This implies that the stochastic prodction frontier is significantly different from the deterministic frontier, which does not inclde a random error. This indicates that the random component of the inefficiency effects does make a significant contribtion in the analysis. In the MLE estimation, γ is posive and significant at 1% level, implying that tea indstry specific technical efficiency is important in explaining the total variabily of vale added prodced. However, shold be noted that 99 percent of the variation in prodction is de to technical inefficiency and only 1 percent is de to the stochastic random error. Reslts of hypothesis tests The reslts of varios hypotheses tests for the specification model (5) are presented in Table 4. The vale of log likelihood fnction for OLS and MLE allow to test whether technical inefficiency exists or not. In case technical inefficiency does not exist then technically, there will be no difference in the parameters of OLS and MLE. The nll hypothesis which incldes the restriction that γ is zero does not have a chi-sqare distribtion, becase the restriction defines a point on the bondary of parameter space γ. The first nll hypothesis H : γ = for the

8 Baten et al Table 4. Likelihood-Ratio Test of Hypothesis of the Stochastic Frontier Translog Model. Nll Hypothesis Log-likelihood Fnction Test Statistic λ Crical Vale * Decision H : γ = H : β ij = H : η = Notes: All crical vales are at 5% level of significance. *The crical vales are obtained from table of Kodde and Palm (1986). Reject H Reject H Reject H Vale added specification model which specify that there is no technical inefficiency effects in the model. The vale of the logarhm of the likelihood fnction provides generalized likelihood ratio test statistic of , which is larger than the crical vale of 1.5. So the hypothesis is rejected and we can conclde that there is a technical inefficiency effect, given the specifications of the stochastic frontier and inefficiency effect model. Hence the stochastic frontier model does appear to be a significant improvement over an average prodction fnction that spports the reslts of Basnayake and Gnaratne (). The second nll hypothesis H : β ij = indicates that Cobb-Doglas Prodction Fnction is more preferable than Translog Prodction Fnction. From the otcome, is observed that the nll hypothesis is strongly rejected and Translog Prodction Fnction is statistically more favorable. The third nll hypothesis is, which specifies that the technical inefficiency H : η = effect does not vary significantly over time. The nll hypothesis is rejected indicating that the technical inefficiency effect varies significantly. There are not many stdies carried ot to estimate prodction efficiency sing tea indstries data in Bangladesh. Recently, Baten et. al., (9) sed panel data to estimate the prodction frontier and the technical inefficiency effects of tea prodction sing a Stochastic Frontier Analysis (SFA) methodology. Their stdies fail to consider vale added (otpt variable) for the measrement of tea prodctive efficiency. Or reslts are mostly compatible in measring indstries or firm s performance to some international stdies sch as Fahr and Snde (5) and Schettini et. al. (8). It was fond that the technical efficiency estimates are highly sensive to the fnctional form specified becase the Translog model yielded different technical efficiencies. However, the Translog specification is sed in the interpretation as is accepted by the data. The second stage analysis, which identifies the determinants of the inefficiency, shold be done for a meaningfl policy implication. Labors are fond to be more inefficient even when they are expected to be major determinants of tea prodction indstry. This may be becase their lacking of knowledge and information provided them the extension officers. Therefore is necessary to increase edcational facilies in the area. This stdy, however, emphasize the potential improvement of Bangladesh tea indstry throgh indstry efficiency improvement, which can allow Bangladesh to regain the competiveness in the world tea market. Conclding remarks This stdy focsed on the estimation of the technical efficiency of the tea prodcing indstries in Bangladesh, applying the Stochastic Frontier Approach and to identify the factors casing inefficiency over the reference period 199 to 4. The rejection of the Cobb-Doglas model as an adeqate representation of Bangladesh Tea Indstry was jstified, becase the fnction is non-linear in some dimensions and there are important interactions among the variables. The variables, both area and labor, disappeared to be the major determinants on the tea indstry prodction. According to the reslts obtained from the stochastic frontier estimation, the average technical efficiency of tea indstry given by the Translog model is 59%. This indicates that there is a scope to frther increase the otpt by 41% whot increasing the levels of inpts. From the inefficiency effects model, we have fond that the variable HHI shows negative bt impact on tea prodction and temperatre, significantly contribted to improve technical efficiency in tea prodction. We conclded that temperatre was one of the major variables in order to improve technical efficiency in tea prodction in Bangladesh, bt is srprising abot rainfall which was fond less efficient althogh, is not statistically significant. For the MLE, γ is estimated at.99, this can be interpreted that 99% of random variation is the Vale added among the tea indstry prodction de to inefficiency. ACKNOWLEDGEMENT The athors wish to acknowledge the spport provided by Research Universy Grant, No. 11/PJJAUH/ 81113, Universy Sains Malaysia, Penang, Malaysia for condcting this research.

9 3816 Afr. J. Biotechnol. REFERENCES Aigner DJ, Lovell CAK, Schmidt P (1977). Formlation and estimation of stochastic frontier prodction fnction models. J. Econometrics. 6: Ariyawardana A (3). Sorces of Competive Advantage and Firm Performance: The Case of Sri Lankan Vale-Added Tea Prodcers. Asian Pac. J. Manage. : BCS ( ). Bangladeshio Cha Sangsad (BCS). Annal Report. Agargaon, Dhaka, Bangladesh. Bangladeshio Cha Sangsad (199-4). Annal Report. Agrabad Commercial Area. Chtagong. Basnayake BMJK, Gnaratne LHP (). Estimation of Technical Efficiency and It's Determinants in the Tea Small Holding Sector. Sri Lankan J. Agric. Econ. 4(1): Baten MA, Kamil AA, Haqe MA (9). Modeling Technical Inefficiencies Effects in a Stochastic Frontier Prodction Fnction for Panel Data. Afr. J. Agric. Res. 4(1): Battese GE, Coelli TJ (1988). Prediction of Gram-level Technical Efficiencies: Wh a Generalized Frontier Prodction Fnction and Panel Data. J. Econometrics, 38: Battese GE (199). Frontier Prodction Fnction and Technical Efficiency: A Srvey of Empirical Applications in Agric. Econ. 6: Battese GE, Coelli TJ (1995). A Model for Technical Inefficiency Effects in a Stochastic Frontier Prodction Fnction for Panel Data. Emp. Econ. : Battese GE, Corra G (1977). Estimation of a Prodction Frontier Model wh Application to the Pastoral Zone of Easter Astralia. Ast. J. Agric. Econ. 1: Baer PW (199). Recent Developments in the Econometric Estimation of Frontiers. J. Econometrics, 46: Beck M (1991). Empirical Applications of Frontier Fnctions: A Bibliography. mimeo, Joachim-Ringelnatz-Str. W-6 Wiesbaden, Germany, : p. 9. Coelli TJ (1996). A Gide to Frontier Version 4.1: A Compter Program for Stochastic Frontier Prodction and Cost Fnction Estimation. CEPA Working Papers. 7/96: School of Economics, Universy of New England, Armidale. Daniela S, Igliori DC, Azzoni CR (8). Prodctive Efficiency across Regions in Brazil: a Spatial Stochastic Frontier Analysis _en.pdf. Fare RS, Grosskopf S, Lovell CAK (1985). The Measrement of Efficiency of Prodction, Boston: Klwer-Nijhoff. Forsod FR, Lovell CAK, Schmidt P (198). A Srvey of Frontier Prodction Fnctions and their Relationship to Efficiency Measrement, J. Econometrics, 13: 5-5. Fried HO, Lovell CAK, Schmz SS (1993). The Measrement of Prodctive Efficiency: Techniqes and Application. Oxford Universy Press. Oxford. UK. Haqe SMR (6). Cost Factors Leading to Strategy Formation of Bangladesh Tea Indstry. Yokohama J. Soc. Sci. 1: Haqe SMR (7). Strategic Cost Management of Tea Indstry: Adoption of Japanese Tea Model in Developing Contry Based on Vale Chain Analysis. Yokohama J. Soc. Sci. 11: Hazarika C, Sbramanian SR (1999). Estimation of Technical Efficiency in the Stochastic Frontier Prodction Fnction: An Application to the Tea Iindstry in Asam. Indian J. Agric. Econ. 54: International Tea Commtee (199-4). Annal Blletin of Statistics. Sir John Lyon Hose London. Kodde DA, Palm AC (1986). Wald creria for jointly Testing Eqaly and Ineqaly Restrictions. Econometrica, 54: Ley E (199). A Bibliography on Prodction and Efficiency. mimeo, Department of Economics, Universy of Michigan, Ann Arbor, MI 4819, p. 3. Majmder PP (3). Working Condions in the Bangladesh Tea Plantation Indstry. newdelhi/download/plant4.pdf. Mahmd M (4). Tea in a New Brew. The Daily Star, p 1, Retrieved Janary 5, 6, from net/4/1/5/ d4151.html Meesen W, Van Den Broeck J (1977). Efficiency Estimation from Cobb-Doglas Prodction Fnctions wh Composed Error. Int. Econ. Rev. 18: Monjr M (4). Tea in a New Brew. The Daily Star (Bangladeshi National News Paper), Janary Nghia Dai Tran (9). Transion to Organic Tea Prodction in Thai Ngyen Province, Vietnam: Economic and Environmental Impacts. Research report (Economy and Environment Program for Sotheast Asia); ISSN ; 8-RR8 Co-pblished by the International Development Research Centre. Fahr R, Snde U (5). Regional Dependencies in Job Creation: An Efficiency Analysis for Western Germany. Discssion Paper Series, IZA DP No Rohan Jayatilake JC (9). The Estimation of Technical Efficiency of Tea Manfactring Firms in Sri Lanka: A Stochastic Frontier Analysis. Ann. Res. J. SLSAJ, Isse 6, Schmidt P (1986). Frontier Prodction Fnctions, Econometric Rev. 3: Taymaz E, Saatci (1997). Technical Change and Efficiency in Trkish Manfactring Indstries. J. Prod. Anal. 8: 474.