http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 Techncal Effcency of Manufacturng Frms n Cameroon: Sources and Determnants Ernest Ngeh Tngum 1 & Moses A. Ofeh 1 Faculty of Economcs and Management Scences, Unversty of Bamenda, Bambl, Cameroon Department of Economcs, Hgher Teacher Tranng College, Unversty of Bamenda, Cameroon Correspondence: Ernest Ngeh Tngum, Faculty of Economcs and Management Scences, Unversty of Bamenda. P.O Box 39, Bambl, Cameroon. Receved: Aprl 8, 017 Accepted: May 17, 017 Onlne Publshed: July 18, 017 do:10.5430/jfr.v8n3p17 URL: https://do.org/10.5430/jfr.v8n3p17 Abstract The prmary objectve of ths study s to analyze the determnants of effcency n manufacturng frms n Cameroon. The study used a stochastc fronter model employng RPED data of 319 frms from dfferent manufacturng ndustres. The data are mcro-level whch s the most adequate type of data used n the estmaton of these models. The model used s that outlned by Battese and Coell (1995) whch determnes the causes of neffcency n the manufacturng sector n Cameroon. The estmates of the stochastc producton fronter wth neffcency effects model ndcates that frms n Cameroon exhbt varous degrees of techncal neffcency for the sample of frms consdered. The results show that frm sze plays an mportant role n explanng techncal effcency n the sub-sector of food processng. However, large frms reduce techncal neffcency levels of frms n all the sub sectors. Another mportant varable whch has an effect n determnng techncal effcency level s the foregn ownershp varable. It s sgnfcant n food processng, wood processng, textle and garments as well as n the overall sample. Hence, t ncreases techncal effcency n all the sub-sectors. Fnally, snce an ncrease n age of frms leads to a reducton n effcency levels n manufacturng frms, polces should be adopted to replace the exstng captal n the large frms. Keywords: Cameroon, manufacturng, maxmum lkelhood estmates, stochastc fronter techncal effcency 1. Introducton and Background The manufacturng sector has played an mportant role n Cameroon snce ts ndependence wth productvty enhancement beng crucal to the drve for rapd ndustralzaton and economc growth (Njkam and Cockburn, 007). The sector employs around 9. percent of the total labor force, supples ts output both n domestc and foregn markets, generates foregn exchange recepts (up to 35 per cent of export recepts) and contrbutes up to 17.5 percent to the Gross Domestc Product (GDP) at current prces 3. Moreover, manufacturng nduces most of the lnkage effects on the other sectors of the economy, thus contrbutng to export dversfcaton, job creaton, and poverty reducton (Natonal Insttute of Statstcs (NIS), 009). However, the performance of the sector has been declnng n recent years. Accordng to NIS (009), the drop may be attrbuted to the declnng number of frms leadng to a contnuous fall n output. Evdence from lterature ponts to the declne n manufactured commodty prces, apprecaton of the Communauté Fnancère de l Afrque (CFA) franc relatve to the US dollar, and certan domestc dstortons such as hgh cost of nputs, a cumbersome admnstratve machnery, poor management of publc enterprses, poor macroeconomc polcy, and cutbacks n government subsdes to frms as the man causes of the fall n manufactured output (Njkam et al., 008). Therefore, low and nadequate manufacturng frms output n Cameroon may potentally be explaned by the fact that most of the frms are old (0 years and above) and stll use the obsolete technology wth lttle or no technologcal change. Ths may lead to both hgh techncal and allocatve neffcences. Although some of the factors that lead to low productvty n frms have been dentfed (Soderlng, 1999; Njkam, 003; 007), economc and nsttutonal factors that are expected to have sgnfcant effect on techncal effcency of manufacturng frms n Cameroon are stll not well emprcally establshed. The paper focuses on estmatng techncal effcency and factors nfluencng techncal neffcences of manufacturng frms n Cameroon. Publshed by Scedu Press 17 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 The rest of the paper s organzed as follows. In secton two, some emprcal lterature on productvty and effcency studes n Cameroon are revewed. Secton three provdes the methodology for the study. In secton four, the estmated coeffcents are reported and results are equally dscussed. Secton fve gves conclusons and polcy mplcatons.. Studes on Cameroon Manufacturng Frms Soderlng (1999) used frm level data coverng the perod 1980 1995 to present man developments n the manufacturng ndustry n Cameroon. The study lad more emphass on structural factors of compettveness. A producton functon and an export functon were estmated n order to study the determnants of total factor productvty (TFP) and export performance. The results provded evdence ndcatng that openness to trade, development of sklled labor and adequate management of the real exchange rate were crucal factors n the enhancement of productvty and exports. The smple model used to quantfy these mpacts revealed that the devaluaton of the CFA franc n 1994 had some apprecably benefcal effects on manufacturng productvty and exports. More so, Soderlng (1991) demonstrated a mutually renforcng relatonshp between productvty and export performance and constructed a model to assess the cost of Real Effectve Exchange Rate (REER) evaluaton, both n terms of productvty and exports. The study showed that performance of the manufacturng sector n Cameroon deterorated consderably after the md-1980s. The declne was to a large extent explaned by n-ward lookng polces n the manufacturng sector. Njkam (003) usng frm-level data to establsh the trade reform effcency on Cameroonan manufacturng frms reported a postve (but statstcally nsgnfcant) assocaton between the offcal tarff rates and the level of average techncal effcency acheved by frms. The author also found the assocaton between effectve protecton rate and the level of mean techncal effcency n the manufacturng frms to be postve but statstcally nsgnfcant. Further, the study observed a strong postve assocaton between mport penetraton rato and the level of mean techncal effcency acheved n the manufacturng ndustry. Even though the results obtaned by Njkam conformed to the a pror expectaton of a postve relatonshp between the two varables. However, the results were obtaned from a correlaton analyss whch does not provde a bass for measurng the mpact of one varable on the other. The results of Njkam (000) ndcated a postve and sgnfcant correlaton between manufacturng share of exports and average techncal effcency acheved n the Cameroonan manufacturng sector. The results showed that the hgher the share of manufacturng n total exports, the hgher the mean techncal effcency acheved n the manufacturng sector. The study also reported a postve and sgnfcant assocaton between changes n mport penetraton rate, export share, effectve rate of protecton and ntra-ndustry trade ndex and the mean techncal effcency acheved n the frms. Further, a negatve and nsgnfcant correlaton between changes n offcal tarff rates and the mean techncal effcency were found. Moreover, the results ndcated that, whle macroeconomc nstablty (nflaton) had a negatve and statstcally sgnfcant mpact on average techncal effcency acheved n the sector, the mpact of poltcal nstablty on the mean techncal effcency was also negatve but statstcally nsgnfcant. The author also revealed that the mpact of property rght protecton on mean techncal effcency s postve and statstcally sgnfcant. The results mply that poltcal and macroeconomc nstablty hndered effcency of manufacturng sector whle property rghts protecton promoted manufacturng sector s effcency n Cameroon. Njkam and Cockburn (007) used pooled pre and post reform perod data (from 1988/89 to 1991/9 and from 1994/95 to 1997/98) for Cameroon manufacturng frms to estmate a sngle stochastc producton fronter for each ndustral sector. Such a fronter was used to assess the effects of trade reforms n manufacturng frm-level techncal effcency. A Cobb-Douglas producton functon was specfed and estmated for the producton fronter. The lnk between trade reforms and frm-level techncal effcency was establshed usng a two-stage procedure. In the frst stage, the producton fronter parameters were estmated and frm-level techncal effcences derved. In the second stage, the derved frm-level techncal effcences were regressed on trade polcy and macroeconomc varables to assess the mpact of trade reform and macroeconomc varables. The results suggested that trade reforms provded an enablng envronment for mprovng frm-level techncal effcency. Average techncal effcency ncreased n sx of the eght sectors followng trade reforms. The pre-reform frm-specfc techncal effcences decreased on average at an annual rate of 0.76 percent, whle the post-reform frm-specfc techncal effcency ncreased on average at an annual rate of 1.4 percent. Lastly, factors that characterze frm-level techncal neffcency pror to trade lberalzaton, as showed by the Tobt and fxed effects results were macroeconomc nstablty and poltcal nstablty of the early 1990s, coupled wth restrcted trade regme. After the trade reforms, the potental determnants of frms techncal effcency were export share and Publshed by Scedu Press 173 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 mport penetraton rate (Njkam et al., 008). Cameroon ndustral polces rase dfferent questons whch the present study sheds lght on. In explanng the gap, the followng questons are asked: What are the determnants of Cameroon s manufacturng effcency, and usng such determnants, how effcent are the manufacturng frms n Cameroon across dfferent ndustres? 3. Methodology 3.1 The Sample of Cameroonan Manufacturng Frms and Varables The data set used n ths secton s obtaned from the Regonal Program Enterprse Development (RPED) dataset for Cameroon s manufacturng sector for the year 009 captured by the World Bank s RPED survey of year 010. The man objectve of these surveys n Afrcan countres s to ncrease the knowledge of the creaton process of Afrcan manufacturng frms and to shed some lght on the problems they face n ther development. The RPED defnes formal frms as those recorded n the trade regster. They are known to the government tax authortes and are potental taxpayers for all regular taxes resultng from ther commercal actvtes. The purpose of the survey n Cameroonan manufacturng was to capture busness perceptons on the man obstacles to enterprse growth, the relatve mportance of varous constrants to ncreasng employment and productvty, and the effects of the country s busness envronment on ts nternatonal compettveness. The sample conssted of 319 frms employng at least 5 permanent workers, and coverng the followng manufacturng sub-sectors: food processng, textle and garments, chemcals and pharmaceutcals, non-metallc, machnery and equpment, electroncs and wood processng. The fve sectors covered n the study represent approxmately 76.18 per cent of total manufacturng n Cameroon (RPED, 010). The food, wood and textle and garments sectors are the domnant sectors n terms of output and employment, followed by metals and machnery, electroncs, chemcal and pharmaceutcal ndustres among others. Durng the years of mport substtuton, most resources were nvested n the food sector, and later, durng the 1980s, n the wood and other sectors. Because some of the nvestments n food and wood producton were foregn, t has been suggested that these sectors are the most productve and technologcally advanced. Output n the food sector comprses a wde range of commodtes, ncludng gran mllng, dary products, cannng and preservaton of meat, frut and vegetables, bakery and confectonery, beverages, food preservatves and anmal feed (Njkam and Cockburn, 007). An mportant advantage of the data set s that t enables one to test for neffcency usng truly mcroeconomc data. In fact, t has been found that emprcal tests whch rely on mcroeconomc data provde clearer evdence of neffcency than studes that make use of more aggregate data, snce there s a loss of nformaton n the aggregaton process (Schmdt and Lovell, 1979). Appendx 1 shows the dstrbuton by sze, the sector of actvty and the ages of the frms. The greater proporton of medum sze frms are 0 years old and above. Generally, there are more medum sze frms n the sample, followed by large frms. 3. Analytcal Framework and Model Specfcaton The specfcaton of the stochastc fronter model s a producton functon wth an error term ncorporatng two components: the output-based unobservable techncal neffcency factor u, and a symmetrc component v, capturng random varatons across producton unts and random shocks that are external to ts control. Followng Farrell (1957), Agner and Chu (1968) the model s specfed as; Y f( X, ) e ; 1,,..., N 1 Where Y represents the potental output level on the fronter for frm, a gven technology f (.), s a ( 1 k) X vector of nputs and other explanatory varables assocated wth the th frm. β s a ( k 1) vector of unknown parameters. The error term e s composed of two ndependent elements,.e., e v u, wth the v term beng a random (stochastc) error, whch s assocated wth random factors not under the control of the frm. It s assumed to be ndependently and dentcally dstrbuted as N(0, ), where v v stands for the varance of Publshed by Scedu Press 174 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 stochastc dsturbance v. If ndustres acheve ther maxmum output, then they would be techncally effcent and t means that u 0. u s assocated wth the techncal neffcency of the th frm and defned by the truncaton (at zero) of the normal dstrbuton N(, ), where z s a ( 1 m) vector of explanatory varables assocated z u wth techncal neffcency of frms; and s an ( m 1). Due to ts ablty to decompose the composte error term nto a techncal neffcency term and a stochastc error term, the stochastc fronter analyss has been used n estmatng techncal effcency. The measure of effcency s gven as the rato of the observed output of the th frm to the potental output defned by the fronter functon and s outlned as: TE y exp( x ) exp( x u ) exp( x ) exp( u ) It s assumed that the fronter has frm effects whch are dstrbuted as a truncated normal random varable, n whch the neffcency effects are drectly nfluenced by a number of varables. Gven the objectves of the study, the emprcal model s specfed as follows: InY 0 1In( K ) In( L ) 3In( H ) 4In( R ) v u where the four categores of nputs used n the study nclude: Captal (K), Labor (L), Human Captal (H) and ntermedate nputs (R). In the study, deprecaton allowances are used to measure the captal nput (Lundvall et al., 00). Labor nput (L) s measured as total number of hours worked n the frm. Human captal captures the specfc mpact of human qualfcatons and ntermedate nputs varable s measured as the expendtures on nputs (raw materals and supplementary materals such as sold and lqud fuel, electrcty and water costs) adjusted for stock changes. Knowng that frms are techncally neffcent mght not be useful unless the sources of the neffcency are dentfed. Therefore, the second stage of the analyss nvestgates the frm specfc attrbutes that nfluence ts techncal effcency. The neffcency functon s specfed as follows: u frmsze age foregn unon reglabor corrupton taxrates accessfn Mngedu Mng exp 8 0 1 9 3 10 4 where frmsze s a categorcal varable for varous frm szes, age s the age of the frms, foregn s ownershp of the frms (domestc and foregn), unon s the exstence of trade unons n the frms, reglabour s the labor regulatons n the frms, corrupton s the ndex of corrupton, taxrates s the rate of taxes faced by the frm, accessfn s the access to fnance by the frm, Mngedu s the manager s educatonal level (n contnuous years) and Mngexp s the manager s experence. 5 6 7 4 3 Publshed by Scedu Press 175 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 Source: Authors Fgure 1. Conceptual model of manufacturng frms techncal effcency The producton fronter specfed n equaton (3) and the neffcency models defned by equaton (4) above were jontly estmated by the maxmum-lkelhood (ML) method usng FRONTIER 4.1 (Coell, 1996). The FRONTIER software uses a three-step estmaton method to obtan the fnal maxmum lkelhood estmates. Frst, estmates of the parameters are obtaned by OLS. The two-phrase grd search for s conducted n the second step wth estmates set to the OLS values and other parameters set to zero. The thrd step nvolves an teratve procedure, usng the Davdon-Fletcher-Powell Quas-Newton method to obtan fnal maxmum-lkelhood estmates wth the values selected n the grd search as startng values. 4. Results and Dscusson 4.1 Hypothess Testng In estmatng the producton technology for the overall sample and fve sectors of Cameroon s manufacturng frms, the Cobb-Douglas and trans-log producton functons are specfed for the emprcal analyss. Based on Coell and Battese (1996), varous tests of hypotheses of the parameter n the fronter producton functon and the neffcency models are performed usng the generalzed lkelhood rato test statstc, defned by the negatve of twce the logarthm of the lkelhood rato as approxmately the dstrbuton wth degree of freedom equal to the dfference of the estmated parameters between the two nested hypotheses. Coell and Battese (1996) defned the generalzed lkelhood rato statstcs as: where ( H 0 ) alternatve H ) [log( L( H )) log( L( H ))] 0 1 L and L H ) denote the values of the lkelhood functon under the null ) ( 1 ( 1 ( 0 5 H and the hypotheses, respectvely. If the null hypothess s accepted, then the lkelhood rato test statstc has an approxmately Ch-square or a mxed Ch-square dstrbuton wth degrees of freedom equal to the dfference Publshed by Scedu Press 176 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 between the number of parameters n the unrestrcted and restrcted models. Two tests are performed; Frstly on the functonal form, the form of producton functon encompasses the Cobb-Douglas form (Cobb-Douglas s nested n the trans-log form). So the test of preference for one form over the other can be undertaken by analyzng the sgnfcance of the cross terms n the trans-log form. Secondly, as concerns the neffcency effects model, the null hypothess s tested as: H :... 0, whch specfes that the techncal neffcency effects are not present n the model, 0 0 1 10 that s, manufacturng frms n Cameroon are effcent and have no room for effcency growth. Table 1. Test of hypothess for techncal effcency Food Textle and Metal and Processng Wood Garments Machnery Electroncs Overall sample Crtcal value H for all Cobb Douglas functon 0 0 1...4 4.47* 17.58 17.0 16.79 16.54 36.3* 17.67 H :... 0 ( No neffcency effects ) 0 0 1 10.63* 19.65* 16.91* 15.57* 1.85* 3.9* 10.37 Notes: * denotes cases where the null hypothess s rejected. Ths happens when the calculated value exceeds the crtcal value. Crtcal values are obtaned from Kodde and Palm (198) and are at 5% level of sgnfcance The results from Table 1 show that the Cobb-Douglas producton functon s accepted for four sectors (Wood and furnture, Textle and Garments, Metals and Machnery, and Electroncs), except for the food processng and the overall sample gven the assumpton of the trans-log producton functon. Therefore, the Cobb-Douglas functon s specfed for the four sectors whereas the trans-log specfcaton s adopted for the food processng and the overall sample. The null hypothess of no techncal effcency effects s rejected for all the sectors ncludng the overall sample. Therefore, there are neffcency effects n all the frms n the sample. Ths mples that, the tradtonal average (OLS) functon s not sutable for estmatng the results of the paper. Hence, the maxmum lkelhood estmaton (MLE) method s appled. 4. Techncal Effcency Analyss Table reports the estmates of the Cobb-Douglas producton functons for four sectors (Wood and furnture, Textle and Garments, Metals and Machnery, and Electroncs) and the trans-log estmates for the food processng sector and the overall sample. The varables of the producton functon dsplay the expected postve sgns. The coeffcents are generally sgnfcant at the conventonal statstcal level although the coeffcent of expendture on raw materal s not sgnfcant for the two sectors. The results show that the elastcty of output wth respect to labor domnates over captal. Smlar results are obtaned by Chapelle and Plane (005) among Ivoran manufacturng sectors. Ths ndcates that for specfc polcy formulaton n addressng low productvty, there s a possblty of ncreasng the number of hours worked n the frms n Cameroon. More so, an ncrease n total annual deprecaton (K) and average educatonal attanment (H) wll sgnfcantly and postvely ncrease the frms output. Ths shows that techncal effcency and output should ncrease wth ncrease n the average educatonal attanment of the workers snce educaton and captal replacement were expected to be postvely correlated wth techncal effcency. In Table, the negatvty of the generalzed log lkelhood rato shows the presence of the neffcency term across all the sectors. Publshed by Scedu Press 177 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 Table. Cobb-Douglas and trans-log stochastc fronter estmaton of techncally effcency Varable Food Processng Wood & furnture Textle & Garments Metals and Machnery Electroncs Overall Sample Constant 0.603 0.379 1.184 1.743*** 0.046 1.431 (0.35) (0.17) (0.44) (4.6) (0.03) (1.38) Loglabor(L) 0.748*** 0.645*** 0.660*** 0.056 0.909*** 0.651*** (7.16) (5.31) (4.31) (0.37) (10.11) (1.79) Logcaptal(K) 0.013 0.013 0.165* -0.9** 0.089* 0.03 (0.9) (0.19) (1.61) (-.8) (1.54) (1.09) Loghumcap(H) 0.33 0.49* -0.505 0.677-0.871** -0.056 (1.13) (1.47) (-0.71) (1.14) (-.1) (-0.34) Logntermedate(R) 0.9*** 0.405*** 0.71* 0.49*** 0.168** 0.348*** (3.48) (3.91) (1.83) (3.69) (.13) (8.4) (1/)log(k*K) 0.51 0.016 (0.39) (0.14) (1/)log(L*L) 0.115*** 0.341*** (6.01) (8.3) (1/)log(H*H) 0.333 0.07** (0.13) (.0) (1/)log(R*R) 0.19** 0.115** (.4) (.35) log(k*l) 0.71*** 0.39** (7.3) (1.98) log(k*h) -0.017-0.07 (-0.5) (-0.34) log(k*r) -0.013* 0.99** (-1.40) (.47) log(l*h) 0.4** -0.36 (.79) (-0.97) log(l*r) 0.19** 0.4** (.33) (.79) log(h*l) 0.196* -0.034 (1.9) (-0.39) Sgma-squared 1.11 1.37 3.8 1.81 0.614 1.59 Lambda 0.017 0.007 0.008 0.013 0.015 0.006 No. Obs. 71 55 41 39 37 319 Wald Ch 344.43 19.91 71.43 35.46 67.16 919.53 Prob>Ch 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** Mean TE 0.74 0.653 0.555 0.498 0.631 0.619 Log-lkelhood -104.3-86.714-8.57-66.897-43.484-56.94 Notes: ***, **,* show sgnfcance level at 1%, 5% and 10% respectvely. Values n parenthess are the z-values. Sgma squared ( ). Source: Authors s v Publshed by Scedu Press 178 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 4.3 MLE of Stochastc Fronter Model Accountng for Heteroskedastcty Problems wth effcency estmaton can arse when the varance of the dependent varable vares across the data, known as heteroscedastcty. Heteroscedastcty affects standard errors, and thus the determnatons of sgnfcance of a gven varable. Standard tests for heteroscedastcty followng a lnear regresson are not avalable for fronter maxmum lkelhood estmaton. However, the Cobb-Douglas functon shows that the frms are usng labor, captal, human captal and ntermedate nputs n the producton process wth constant returns-to-scale technology, but the szes of the frms dffer. The sze varaton ntroduces heteroskedastcty nto the dosyncratc error term (Coell, 1995). Stata allows for explct modelng of varables thought to nfluence the varance of both u and v, but an assumpton of a half-normal neffcency error term s requred. Therefore, the parameters of the Cobb-Douglas are estmated takng nto account the heteroskedastc effects. To do ths, a condtonal heteroskedastc half-normal model s used, wth frm sze as an explanatory varable n the varance functon for the dosyncratc error. Table 3 ndcates that the varance of the dosyncratc error term ( ) s not really a functon of frm sze n four of the fve sectors consdered. Heteroscedastcty only occurs n the wood and furnture ndustry. However, when the overall sample s consdered, no strong pattern of heteroscedastcty s apparent. Therefore, the results suggest that heteroscedastcty s not a sgnfcant problem. The Wald ch tests and ts correspondng probablty for all the sectors ndcate that the study fals to reject the hypothess that the frms use constant returns to scale technology. v Table 3. Maxmum Lkelhood Estmaton of Cobb-Douglas and Stochastc fronter models accountng for Heteroscedastcty (Half-normal Maxmum Lkelhood Estmaton) Varable Food Processng Wood & Furnture Textle & Garments Metals & Machnery Electroncs Overall Sample Constant 0.36 -.099-0.804 10.799*** 0.097 1.459* (0.5) (-1.7) (-0.39) (3.48) (0.06) (1.83) Loglabor (L) 0.741*** 0.846*** 0.676*** -0.041 0.90*** 0.643*** (7.11) (9.16) (4.53) (-0.7) (9.73) (1.14) Logcaptal (K) 0.001 0.033 0.6** -0.148 0.089* 0.031 (0.0) (0.54) (.14) (-1.7) (1.55) (1.13) Loghumcap (H) 0.31 0.63-0.89 0.1-0.847-0.071 (1.0) (1.8) (-0.47) (0.30) (-0.0) (-0.4) Lognternputs (R) 0.33*** 0.306*** 0.304** 0.644*** 0.171** 0.354*** (3.78) (3.64) (.48) (3.88) (.18) (8.9) (1/)log(k*K) 0.36 0.704 (0.41) 0.3 (1/)log(L*L) 0.0*** 0.31*** (6.43) 8.059 (1/)log(H*H) 0.085* 0.048 (1.55) 0.88 (1/)log(R*R) 0.100* 0.059* (1.68) 1.76 log(k*l) 0.107* 0.081 (1.7) 1.7 log(k*h) -0.558-0.564 (-0.9) -0.3 log(k*r) 0.779* 0.600* (1.45) 1.69 log(l*h) -0.645-0.05* (-1.35) -1.59 Publshed by Scedu Press 179 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 log(l*r) 0.116** 0.066* (1.96) 1.57 log(h*l) -0.535-0.458 (-0.41) -1.06 In u Frmsze -0.463 1.037*** -1.88-0.575 0.15-0.073 (-1.37) (3.90) (-1.35) (-1.6) (0.8) (-0.57) Constant 1.07* -.009*** 3.815*** 1.617* -0.756 0.614** (1.79) (-3.45) (.73) (1.85) (-0.77) (.5) No. Obs. 71 55 41 39 37 319 Wald Ch 363.54 311.9 103.4 4.17 43.71 93.86 Prob>Ch 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** Log-lkelhood -10.7-79.456-80.993-66.179-43.445-56.773 Note:, and, show levels of sgnfcance at 10%, 5% and 1% respectvely. The values n parenthess are the z-values. 4.4 Determnants of Ineffcency The focus of ths secton s to provde an emprcal analyss of factors that contrbute to techncal neffcency and productvty varablty among manufacturng frms n Cameroon. Therefore, the estmated coeffcents n the neffcency model are presented n Table 4. The analyss of the neffcency model shows that the sgns of the estmated coeffcents n the model have mportant mplcatons on the techncal effcency of manufacturng frms. It should be noted that varables are ncluded as neffcency varables; thus a negatve coeffcent means an ncrease n effcency and a postve effect on frms output. From Table 4, frm sze s negatvely correlated wth frm techncal neffcency effects whch mply a postve effect on productvty. The result conforms to a number of theoretcal arguments. The lterature of early development economcs placed a strong emphass on large frms, whch were consdered as the drvng force of economc growth. Hence, small frms were beng perceved as archac modes of producton. Accordng to Chapelle and Plane (005), large frms wth ther manageral know-how would offer a better organzatonal framework to reduce transacton cost. Hll and Kalrajan (1993) concluded wth respect to Indonesan garment ndustry that large frms beneft from more effcent management. Thus the larger the sze of a frm, the more labor s avalable for frms operatons therefore ncreasng the effcency of frms. Frm age s also a major determnant of techncal neffcency of manufacturng frms n Cameroon as t reduces the effcency of the frms. Ths s plausble gven that majorty of frms were establshed n the late 1970s (see appendx 1 for mean age of frms). The frms are old and may not be wllng to try new nnovaton and technology due to fnancal constrants. A sgnfcant relatonshp was found between the exstence of trade unon and the techncal neffcency levels of ndvdual frms n the ndustres (except n the metal and machnery and electroncs sub sectors). However, the varable has postve coeffcents for the sgnfcant sub sectors. Ths shows that the varable explanng the exstence of trade unons contrbutes sgnfcantly to techncal neffcency. Publshed by Scedu Press 180 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 Table 4. Ineffcency effect model Varable Food Processng Wood & Furnture Textle & Garments Metals & Machnery Electroncs Overall Sample Frmsze -0.454*** 0.799*** -0.706* 0.80* 0.879** 0.308** (-3.84) (.75) (-1.66) (1.45) (.3) (.47) Frmage 0.076*** 0.03* 0.011 0.047*** 0.059*** 0.053*** (8.16) (1.43) (0.41) (3.9) (.73) (6.76) Foregn -0.491*** -1.706*** -1.506* -0.455 0.751-1.76*** (-8.76) (-3.31) (-1.68) (-0.99) (0.77) (-4.95) Unon 1.16* 0.487*.394*** -0.166-0.391 1.068*** (1.75) (0.75) (.95) (-0.46) (-0.51) (4.67) Reglabor 0.44*** 0.059 0.369* 0.387** 0.0 0.08 (8.6) (0.38) (1.45) (.18) (0.73) (0.34) Corrupton 0.45*** 0.333** 0.108 0.151 0.387* 0.007 (8.6) (.16) (0.8) (1.6) (1.69) (0.10) Taxrates -0.335*** -0.007 0.007-0.35** 0.165 0.0 (-5.40) (-0.03) (0.03) (-.61) (0.6) (0.6) Acessfn -0.48*** -0.136-0.04-0.107 0.600* -0.016 (-8.83) (-0.75) (-0.11) (-0.63) (1.59) (-0.15) Mngedu -0.35*** 0.104 0.77* 0.167-0.094 0.156*** (-5.94) (1.06) (1.78) (-1.) (-0.6) (3.03) Mngexp 0.08*** -0.0003-0.003-0.019-0.00-0.019* (11.05) (-0.0) (-0.07) (-1.04) (-0.68) (-1.73) Constant.061*** 18.686*** 19.987*** 17.64*** 14.87*** 17.745*** (3.5) (6.7) (9.6) (8.71) (4.0) (6.35) Note: ***, **, * shows levels of sgnfcance at 1%, 5% and 10% respectvely. The values n parenthess show the z-statstcs. Another mportant varable whch has an effect n determnng techncal effcency level s the foregn ownershp varable. It s sgnfcant n the food processng and beverages, wood processng, textle and garments as well as n the overall sample. Hence, t ncreases techncal effcency n the sub-sectors. Fnally, the results also show that corrupton plays a sgnfcant role n ncreasng techncal neffcency especally n all the subsectors as ndcated by the postve coeffcent of the varable. 4.5 Mean Techncal Effcency and Ineffcency Scores Table 5 shows the mean techncal neffcency n all the sub sectors and for the overall sample. Techncal effcency s defned as: TE y exp( u ); * where y * y s the producton fronter maxmum output gven the nputs for Publshed by Scedu Press 181 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 each frm. Hence, TE exp( u ). Therefore, n all specfcatons, total average techncal effcency would be: 1 I TE TEˆ 1, for each frm, 1,... I (Coell et al. 005). From the techncal effcency equaton, average I neffcency s calculated as; 1 TE. Table 5. Mean techncal neffcency by sze and sector Sze/Sector Food Processng Wood & Furnture Textle & Garment Metals & Machnery Electroncs Overall Sample Small 0.187 0.10 0.04 0.06 0.194 0.184 SD (0.14) (0.169) (0.155) (0.157) (0.136) (0.15) Medum 0.103 0.181 0.177 0.145 0.113 0.159 SD (0.044) (0.13) (0.19) (0.080) (0.045) (0.157) Large 0.7 0.36 0.41 0.4 0.40 0.36 SD (0.183) (0.188) (0.186) (0.185) (0.175) (0.183) Notes: Values n parenthess are the standard devatons for the mean techncal effcences. 1) Small shows frms wth less than 30 employees ) Medum represent frms wth 30 to 100 employees 3) Large represent frms wth over 100 employees. As shown n Table 5, total average techncal neffcency ranges from 10.3% to 4.1% across the fve sectors. For the food processng sector, the average neffcency vares wdely, from 10.3% n medum szed frms to.7% n large frms. Ineffcency n the wood and furnture sector vares across frm szes from 18.1% to 3.6%. Takng the overall sample, neffcency of for-proft vares n the frms across szes from 15.9%% to 3.6%. Thus, the wood and furnture sector s the least effcent amongst the fve sectors, followed by the textle and garments sector. The results also show that the food processng sector s the most effcent sector n the sample. Such a result could be due to the fact that the food processng sector has experenced hgher techncal change than the other sectors n the manufacturng sector, whch could have pushed the producton fronter further for some frms n the sector. It may also be as a result of economes of scale due to the hgh demand for food products. As concerns frm sze, the medum szed frms are found to be most effcent whle large frms are found to be the most neffcent. Although some studes have found a postve relatonshp between techncal effcency and frm sze (Lundvall and Battese, 000; Nrngye et al., 010), the fndngs n ths present study are n conformty wth Bggs et al. (1995) who found an nverted U-shaped relatonshp between frm sze and effcency. They found the sze-effcency relatonshp to be negatve for large frms and postve for small frms, wth the medum-szed frms beng the most effcent. Publshed by Scedu Press 18 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 Table 6. Mean techncal neffcency by ownershp and age for overall sample Varable Ownershp Ineffcency Domestc frms 31.0% Foregn frms 8.77% Frm Age 0 5 years 30.15% 6 10 years 3.07% 11 0 years 7.3% 0 and above 35.97% Notes: 1) Ownershp s measured by number of shares owned n the frm. ) Frm age has been calculated as 009 mnus the year the frm started operatons n Cameroon. Table 6 above shows that foregn owned frms are more effcent than domestc owned frms. Ths mght be explaned by the ssue of transfer technology especally as most of the foregn owned frms n Cameroon export to other countres. Learnng by exportng, n whch experence brngs about mprovements n performance, may be the explanaton for the fndng. Concernng frm age, frms between 6 to 10 years are the most effcent whle the much older frms are the least effcent. Ths mght be explaned by the fact that at the start of the operatons (0 to 5years), frms mght stll be adjustng to cover sunk cost and enter the market where (when) already establshed frms are operatng. More so, n the context of Cameroon, the hgh neffcency of the older frms mght be explaned by the type of technology used n the producton process. Some of the technology s hghly consdered to by archac and out dated. Therefore, older frms operate 35.97% below ther potental fronter producton level wth the gven nputs and producton technology. 5. Concluson The prmary objectve of the study s to analyze the determnants of effcency n manufacturng frms n Cameroon. The model used s that outlned by Battese and Coell (1995) whch determnes the causes of neffcency n the manufacturng sector n Cameroon. The estmates of the stochastc producton fronter wth neffcency effects model ndcate that frms n Cameroon exhbt varous degrees of techncal neffcency for the sample of frms consdered. The results show that frm sze plays an mportant role n explanng techncal effcency n the sub-sector of food processng. However, large frms reduce techncal neffcency levels of frms n all the sub sectors. A sgnfcant relatonshp s found between trade unons exstence and the techncal neffcency levels of ndvdual frms n the ndustres (except n the wood and furnture and metal and furnture sub sectors). The age of frms also play an mportant role n determnng neffcency levels n the ndustry. Ths could be explaned by the fact that most of the older frms were establshed n the post-colonal perods and stll heavly rely on the outdated technology. Another mportant varable whch has an effect n determnng techncal effcency level s the foregn varable. It s sgnfcant n food processng, wood processng, textle and garments as well as n the overall sample. Hence, t ncreases techncal effcency n all the sub-sectors. The results also show that corrupton plays a sgnfcant role n ncreasng techncal neffcency especally n the food processng sector. Fnally, snce an ncrease n age of frms leads to a reducton n effcency levels n manufacturng frms, polces should be adopted to replace the exstng captal n the large frms. References Agner, D., Lovell, K., & P. Schmdt. (1977). Formulaton and Estmaton of Stochastc Fronter Producton Functon Models. Journal of Econometrcs, 6, 1-37. https://do.org/10.1016/0304-4076(77)9005-5 Agner, D.J., & Chu, S.F. (1968). On Estmatng the Industry Producton Functon. Amercan Economc Revew, 58(4), 86 39. Battese G.E., & T. J. Coell. (1995). A model for techncal neffcency effects n a stochastc producton functon for panel data. Emprcal Economcs, 0, 35-33. https://do.org/10.1007/bf010544 Bggs, T., Shah, M., & Srvastava, P. (1995). Technologcal capabltes and learnng n Afrcan enterprses. World Publshed by Scedu Press 183 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 Bank Techncal Paper number 88, Afrca Techncal Department Seres, Washngton, D.C. Chapelle, K., & P. Plane. (005). Techncal Effcency measurement wthn the Manufacturng Sector n Cote d Ivore: A Stochastc Fronter Approach. The Journal of Development Studes, 41(7), 1303-134. https://do.org/10.1080/000380500170964 Coell, T., Prasada Rao, D. S., O'Donnell, C. J., & Battese, G. E. (005). An Introducton to Effcency and Productvty Analyss (nd ed.). New York: Sprnger. Coell, T.J. (1996). A gude to FRONTIER 4.1: A computer program for Fronter producton functon estmaton. Centre for Effcency and Productvty Analyss (CEPA), workng paper 96/07, Department of Economcs, Unversty of New England, Australa. Coell, T.J., & Battese, G. (1996). Identfcaton of factors whch nfluence the techncal neffcency of Indan farmers. Amercan Journal of Agrcultural Economcs, 40(), 103 18. https://do.org/10.1111/j.1467-8489.1996.tb00558.x Farrell, M.J. (1957). The measurement of productve effcency. Journal of the Royal Statstcal Socety, Seres A, 10(3), 53-90. https://do.org/10.307/343100 Hll, H., & K. Kalrajan. (1993). Small Enterprse and Frm-Level Techncal Effcency n Indonesan Garment Industres. Appled Economcs, 5(9), 1137-44. https://do.org/10.1080/00036849300000174 Kodde, D.A., & F.C. Palm. (1986). Wald Crtera for Jontly Testng Equalty and Inequalty Restrctons. Econometrca, 54(5), 143-48. https://do.org/10.307/191331 Lundvall, K., & E. Battese. (000). Frm sze, age and effcency: Evdence from Kenyan Manufacturng frms. Journal of Development Studes, 36(3), 146-163. https://do.org/10.1080/000380008463 Nrngye, A., & E. Luvanda. (010). Determnants of Export Partcpaton n East Afrcan Manufacturng Frms. Current Research Journal of Economc Theory, (), 55-61. Nrngye, A., Luvanda, E., & J. Shtundu. (010). The relatonshp between Frm sze and Techncal Effcency n East Afrca Manufacturng Frms. Journal of Sustanable Development n Afrca, 1(4), 6-36. Njkam, O. (003). Trade reforms and effcency n Cameroon s Manufacturng Industres. AERC Research Paper 133 OECD, 011, Cameroon. Tuns: AfDB/OECD. Njkam, O., & J. Cockburn. (007). Trade lberalzaton and productvty growth: Frm-level evdence from Cameroon. PEP research paper, Canada. Njkam, O., Bamou, E., & C. N donga. (008). The case of the Manufacturng Sector n Cameroon. An AERC Publcaton on Afrcan mperatves n the New World Trade order, Vol., Case studes of Manufacturng and Servces. Oczkowsk, E., & K. Sharma. (005). Determnants of effcency n Least Developed Countres: Further evdence from Napelese manufacturng frms. Journal of Development Studes, 41(4), 617-630. https://do.org/10.1080/0003805000971 RPED. (010). Regonal Program for Enterprse Development (RPED) report, Cameroon, 010. Schmdt, P., & C. Sckles C. (1984). Producton fronters and panel data. Journal of Busness and Economc Statstcs,, 367-374. https://do.org/10.1080/07350015.1984.10509410 Schmdt, P., & C.A.K. Lovell. (1979). Estmatng techncal and allocatve nefcency relatve to stochastc producton and cost fronters. Journal of Econometrcs, 9, 343-366. https://do.org/10.1016/0304-4076(79)90078- Soderlng, L. (1999). Structural polces for nternatonal competveness n manufacturng: the case of Cameroon. Workng paper No. 146, OECD Development Centre. Tamb, E.N. (1984). Agrcultural development polcy and performance n Cameroon, 1960-1980', PhD thess, Unversty of Pttsburgh, USA. Taymaz, E., & G. Saatc, (1997). Techncal change and effcency n Turksh manufacturng Industres. Journal of Productvty Analyss, 8, 461-475. https://do.org/10.103/a:1007796311574 Publshed by Scedu Press 184 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 Appendces Appendx 1. Dstrbuton of frms accordng to sze, age, sector of actvty and by regons n Cameroon Table 7. Dstrbuton of frms accordng to sze, age and sector of actvty Sector of Actvty and szes of frm Age of frm Food Wood Textle Metal Electroncs Non metal others Total [0, 5] (5, 10] (10, 0] (0, +) Total Small (<0) 15 15 11 16 8 14 19 86 4 1 3 38 86 Medum (0-99) 7 6 0 13 16 11 9 19 7 15 38 69 19 Large (100 and above) 9 14 10 10 13 11 0 104 6 13 30 55 104 Total 71 55 41 39 37 6 68 319 17 49 91 16 319 Source: Cameroonan frm level data base, RPED, World Bank. Table 8. Dstrbuton of frms by sze and regon n Cameroon Lttoral (Douala) Centre (Yaounde) West (Bafoussam) Total Small (<0) 38(31) 6(9) (0) 46(40) Medum (0-99) 58(41) 5(15) 4(6) 67(6) Large (100 and above) 47(9) 7(11) 3(7) 57(47) Total 141(101) 18(35) 9(13) 170(149) Source: Cameroonan frm level database, RPED, World Bank. Appendx. Summary statstcs of varables n dfferent sectors Obs. Mean Std. Dev Mn. Max. Food Processng Log of Output 71 0.8808.561 15.4949 6.0846 Log of labor 71 18.8404.1716 14.5856 4.5945 Log of captal 71 17.8965.9496 10.8198 4.5945 Log of human Captal 71 1.14130 0.448 0 1.6094 Log of Intermedate Inputs 71 19.1531.7453 1.49 4.9159 Frm age 71 6.9437 6.944 1 61 Mng exp.(manager experence) 71 16.3944 8.5815 40 Wood and Furnture Log of Output 55 19.566.1665 13.641 5.1053 Log of labor 55 17.6469 1.775 13.0815 3.059 Log of captal 55 17.4585.4471 11.9087 3.363 Log of human Captal 55 1.1955 0.3168 0 1.6094 Log of Intermedate Inputs 55 17.9509 1.1377 11.9184.1096 Frm age 55.8545 14.76 4 61 Mng exp. (Manager experence) 55 18.4546 11.1186 3 50 Textle and Garments Log of Output 41 19.5606.991 14.5087 5.384 Publshed by Scedu Press 185 ISSN 193-403 E-ISSN 193-4031
http://jfr.scedupress.com Internatonal Journal of Fnancal Research Vol. 8, No. 3; 017 Log of labor 41 17.19.3456 13.017 3.059 Log of captal 41 16.3661 3.08 9.9688 3.363 Log of human Captal 41 1.1066 0.4037 0 1.6094 Log of Intermedate Inputs 41 17.589.6484 11.7906.7671 Frm age 41 3.4878 11.485 4 47 Mng exp. (Manager experence) 41 19.3659 8.8396 6 40 Metal and Machnery Log of Output 39 19733 1.887 16.1181 4.599 Log of labor 39 17.3604 1.8581 13.594 1.3609 Log of captal 39 16.7419.385 13.3047 3.11 Log of human Captal 39 1.1470 0.399 0 1.6094 Log of Intermedate Inputs 39 17.6146.1945 14.09.9954 Frm age 39 0.3333 16.177 63 Mng exp. (Manager experence) 39 19.8718 8.7934 3 45 Electroncs Log of Output 37 19.1649.78 14.7318 14.6353 Log of labor 37 17.0 1.9963 13.6171 1.819 Log of captal 37 16.735.3859 10.8198 3.111 Log of human Captal 37 1.5 0.396 0.6931 1.6094 Log of Intermedate Inputs 37 17.9704.41 1.5818 3.719 Frm age 37 3.707 13.84 5 76 Mng exp. (Manager experence) 37 0.6487 8.1454 9 45 Overall Sample Log of Output 319 19.897.4911 13.641 6.0845 Log of labor 319 17.839.0989 13.017 4.5945 Log of captal 319 17.13.6716 9.9688 4.5945 Log of human Captal 319 1.1476 0.44 0 1.6094 Log of Intermedate Inputs 319 18.3764.5805 11.7906 4.9159 Frm age 319 3.5047 15.064 1 76 Mng exp. (Manager experence) 319 18.4075 9.684 50 Source: Author s calculaton usng RPED dataset, World Bank Publshed by Scedu Press 186 ISSN 193-403 E-ISSN 193-4031