Problemstatement
Sincelastdecadesof20thc e n t u r y ,theworldhasexperiencedtheunexampledevolution o f advancedtechnology- intensivem a n u f a c t u r i n g industriess u c h aspharmaceuticals,computers,telecomm unications,precisionengineering,or aircraft.T h o s e high- technology(HT)industrieshavecontributedconsiderablyinpromotinghumanbeings’ healthand longevity,ex te nd in g theabilityofcommunication ,an d improvin gtheknowledgeaccessibility(HamburgInstituteforEconomicResearch[ H W W
The High Technology (HT) industries are widely perceived as key drivers of significant economic growth, promising high-value and high-wage employment opportunities From a microeconomic standpoint, HT firms are known to invest heavily in research and development (R&D) and innovation, which facilitates the creation of new products, increases market share, enhances resource productivity, and generates positive social returns that benefit other sectors Recent global value-added figures indicate a clear upward trend in the growth of the HT sector, particularly in the dynamic Asia region.
World North America European Union Asia
Thepowerofnationsisalsobelievedtonotinfluencedbyheavyindustrieslike steelsbuttheroleisnowplayedbyHTmanufacturingindustriesandknowledge- b a s e d services,w h i c h meanst h a t t h e n a t i o n a l autonomyc a n b e improvedb yd e v e l o p i n g t h e s e i n d u s t r i e s ( H W W A e t al.,1 9 9 6 ).Indeed,theagingi n d u s t r i a l economyhasbowedouttogivewaytothepromisingknowledge- basedandtechnology- intensiveeconomy.Duetothatimportance,HTmanufacturingindustriesa r e thetargetofindustrial policiesinmanycountriesandregions,includingVietnam.H T manufacturingindustrieshave beenpaidmoreattentioninVietnamrecentlywithmanyhigh- techFDIprojectsbuiltup(MoIT&UNIDO,2011),togetherwithnewLawsa n d D e c i s i o n s approvedt o facilitatethescienceandtechnologyactivities.Interestingly,evenduri ngthetimeofglobalcrisis(2008-
9 3 million,w h i l e l o w - t e c h a n d medium- techsectorsexperiencedareductioninexports(MoIT&UNIDO,2011).Indeed,du ringtheperiod2000to2009,overhalfoftotalexportsareHTproducts
A n n u al gr o w th ra te
China India Indonesia Japan Malaysia Philippines Singapore South Korea Taiwan Vietnam Thailand
-10% Proportion in World Market share (2012)
(MoIT&UNIDO,2011).AlthoughVietnameseHTmanufacturesaccountforonlyasmallpro portion ofthe worldmarketshare,itsannualgrowthrateshows apotentialofdevelopmentforthissector(seeFigure2).Moreover,thegovernmentexpectsH
2 1 o f Sciencea n d E n g i n e e r i n g Indicators2014(N at ional ScienceBoard,2014)
Vietnam's manufacturing sector is primarily focused on assembly lines rather than research and development (R&D) and innovation A significant reliance on imported components for high-tech (HT) products may impact the performance of Vietnamese HT firms This raises the question of whether these industries perform as well as their outward appearance suggests To address this, productivity must be carefully examined, as it is a crucial indicator of a firm's performance and survival Empirical research often measures total factor productivity (TFP) to assess overall productivity and its growth Additionally, studies explore the factors driving TFP growth, with theoretical literature suggesting that advancements in technology are a primary driver for HT firms in Vietnam.
& Kalirajan, 2 0 0 5 ).However,i f governmentsonlyfocusonattractinginvestmentstoenh ancetechnologicalprogresso f HTsector,theymayignorethecontributionofotheri mportantsourcessuchase f f e c t s fromchangesinscaleofproduction(Hamit-
Haggar,2011;Kim&Han,2001).B e s i d e s , empiricalstudiesshowevidencethatfirmscan alsoobtainhigherTFPgainsiftheyapplybestpracticemethodsofthegiventechnology,suchfirms areconsidered“technicallye f f i c i e n t ” ( Kalirajan,O b w o n a , & Z h a o , 1 9 9 6 ).Int h i s c i r c u m s t a n c e , technologicalprogressmaybeabsent;instead,effectsfromimprovingt echnicale f f i c i e n c y arethekeysourcecontributing toTFPgrowth.Thus,suchc omponentssh ould betakenintoaccountwhenmodellingtheproductionfunctionandmeasur ingT F P TheywillprovidemorecomprehensiveinsightsofHTsector’sstatusforpolicymake rsintakingHTdevelopmentpoliciesinconsideration.
Nevertheless,thereareveryfewpapersanalyzingthestatusofTFPchangeofV i e t n a m e s e H T manufacturingindustriesas wellas itsdecomposition.The study ofNguyen,Pham,Nguyen,andNguyen(2012),whichcanbetheonly papertouchingt h a t fieldofTFPgrowth’sdecompositionforVietnamesemanufacturingsec toruntiln ow,isnotfocusedonHTmanufacturingindustries.Otherstudies,ifconducted ina n a l y s i s o f H T s e c t o r , s t o p a t measuringT F P (Newman& N a r c i s o ,
2 0 0 9 ),o r investigateonlyonesourceofTFPchange,namelytechnicalefficiencywithanalysesonitsvario usdeterminants(firmsize,firmlocation,ownership…)
(Le&Harvie,2 0 1 0 ;Nguyen,Giang,&Bach,2007).Obviously,theliteratureofem piricalr e s e a r c h e s onTFPofVietnamesehigh- techmanufacturingindustriesanditssourcesofchangeisratherpoor.
Thus,withlongertimespan(2000-2012)andnarrowerresearchobject(high- tech industries),besidesestimatingdeterminantsoftechnicalinefficiency, thispaperattemptst o measureT F P growtho f V i e t n a m e s e H T manufacturersa s w e l l a s i t s decomposition.Theresultsofthestudymayprovidesomeinformationtounderstandt h e performanceofVietnameseHTsectorandbehelpfulforHTsectordevelopmentp o l i c i e s
Researchobjectivesandhypotheses
Scopeofstudy
Theunbalancedpaneldatainthisresearchincludes5822observationsof2403Vietname sehigh- techmanufacturingfirmsthrough13yearsfrom2000to2012.Theselectedsectorincludesfiv esub-industries:
Firmsi n t h e s a m p l e includevarioussizesf r o m smalltol a r g e , differento w n e r s h i p s f r o m s t a t e o w n e d , foreignowned,t o privateo w n e d , w i t h t h e i r he adquarterslocatednationwideinsixregionsofVietnam.
Structureofthesis
Chapter2presentstheliteratureofproductivityandefficiencymeasurementan ddecomposition.Startingwitht h e d e f i n i t i o n s o f keyc o n c e p t s s u c h a s high- technology,productivity,andefficiency,variousapproaches dealingwiththeproduc tivitymeasurementa r e thenreviewed.Moreover,d i f f e r e n t modelso f productivitydec ompositionandefficiencyestimationarealsodiscussedw i t h advantagesanddisad vantagesofeachown.
Chapter4 d e s c r i b e s t h e s p e c i f i c r e s e a r c h methodology,i n w h i c h t h e parametricapproachandregressiontechniqueareexpressedindetails.Thischaptera l s o d iscussesthesevenhypothesesmentionedinthesecondpartofchapterIntroductionmo reclearlywiththetestingmethods.
Chapter5 p r e s e n t s t h e empiricalr e s u l t s i n t w o p a r t s , namelydescrip tivestatisticsofthedata andresultsoftheregression.Basedonempiricalevidencefromeconometricmodels,theinf erenceandanalysisisthendrawnanddiscussedaboutproductivityandefficiencyof Vietnamesehigh-techsector.
Chapter6concludesmainfindingsofthestudyaswellaspolicyandmanagerialim plicationsstemmedfromt h e r e s u l t s p r e s e n t e d i n C h a p t e r 5 Thisc h a p t e r a l s o p o i n t o u t limitationso f t h e t h e s i s a n d t h e n r e f e r t o d i r e c t i o n s f o r researchesinthefuture.
Thischapterprovidessomedefinitionsofkeyconceptssuchastotalf ac t o r productivityandkindsofefficiency.Inaddition,variousapproachesmeasuringandd e c o m p o s i n g productivitychangea r e a l s o d i s c u s s e d i n t h i s c h a p t e r E s p e c i a l l y , stochasticproductionfrontieranalysis(SPF)isthemainfocusofthischapter.
Concepts
Total factorproductivity(TFP)
Productivityofafirmimpliestheratioofoutputsoverinputsinproduction( Co elli,Rao,O’Donnell,andBattese(2005).Inotherwords,itshowshowwelltheoutputs canbeproducedfromgivenamountsofinputs.Productivityisoften usedtocompareperformancebetweenfirmsorindustries:thelargertheratiois,thebett ert h e firm(orindustry)performs.Incasetherearemultipleoutputsandmultipleinputsinvol vingtheproduction,partialproductivitymeasures,whichonlytakeonefactorofpro ductionintoaccount,maybeselectedtoestimatetosimplifytheestimationp r o c e s s
Productivity can be measured through various partial metrics such as labor, land, and fuel productivity However, Total Factor Productivity (TFP) is a more comprehensive measure that considers all production factors, defined as the ratio of aggregate output to aggregate input (Coelli et al., 2005) TFP is preferred over partial measures because the latter can misrepresent a firm's performance due to the omission of key influencing factors For empirical calculations, Multi-factor Productivity (MFP) is a more accurate term, although TFP and MFP are often used interchangeably in research, including this thesis Over time, TFP typically shows positive changes, which are crucial for both the survival of firms from a micro perspective and for economic growth from a macro perspective in the long run.
Technicalchangeor Technologicalprogress(TP)
Accordingto neo- classicaleconomists,duetothelawofdiminishingreturns,t h e firmcannotincreaseitsoutp utlevelsforeverifitkeepsaccumulatingfactorsofp r o d u c t i o n , giventhecurrenttec hnology(Sharma,Sylwester,& Margono,2 0 0 7 ).T h u s , whenafirmisobservedtoincre aseitsTFPinthelong-run,theyarguethattheonlyreasonforTFPgrowthisthat thefirmhasadoptedmoreadvancedtechnology,i m p l y i n g thatthereistechnologicalprog ress(TP).
InSolow(1957)’smodel,positivetechnicalchange(orTPasinsomereviews),whi ch isexogenousand unexplainableby themodel,is theonlysourceoflong- rungrowthofpercapitaincome.Graphically,TPisexpressedastheupwardshift oftheproductionfrontier.Inotherwords,withthepresenceofTP,afirmcanincreaseits p o t e n t i a l productivityb e y o n d previousl i m i t s ( s e eFigure3 f o rillustration).
However,arguingagainstSolow(1957)’s,later studies ofother authorshaveprovedt h a t n o t o n l y T P i s t h e mains o u r c e o f T F P growth,t h e impr ovemento f technicalefficiency,theexploitationofscaleeconomiesorallocativeefficiencya lsodriveTFPgrowth(Coellietal.,2005).
Technicalefficiency(TE)andTechnicalefficiencychange(TEC)
Aproducerisconsideredastechnicallyefficient“ifandonlyifitisimpossiblet o pr o d u ce moreofanyoutput withoutproducinglessofsomeotheroutput orusingmoreofsomeinput”(Koopmans,citedinKumbhakar&Lovell,2000).Despiteth ep o p u l a r i t y ofSolow(1957)’s,thismodelhasacriticalweaknesswhenassumingthatt h e firmsareoperatingwithfullefficiency,i.e.thefirmsareoperatingalongwiththep ro d u ct io n frontier(seeFigure4).IfignoringthepotentialcontributionofefficiencychangestoTFPgrowth ,theestimateofproductivitymaybebiasedandmisleading(Hamit-
Nishimizu and Page (1982) were pioneers in linking efficiency change to productivity growth, highlighting that the assumption of full efficiency is often unrealistic due to many firms operating below their production potential Kim and Han (2001) emphasized that improvements in technical efficiency (TE) can drive total factor productivity (TFP) growth for firms that are not fully utilizing existing technology, often due to organizational constraints The literature further supports that positive changes in technical efficiency (TEC) can lead to progressive TFP changes, while negative TEC can result in regressive TFP shifts, as evidenced by Nguyen et al.
(2012)andKimandHan(2001),aftermeasuringanddecomposingT F P change,d r e w a c o n c l u s i o n a b o u t t h e positivec o n t r i b u t i o n o f T E C i n t o T F P growth,w h e r e a s f i n d i n g s o f KimandShafi'i( 2 0 0 9 ) a n d Hamit-
ScaleeconomiesandScalechange effects(SCE)
According to the theoretical background reviewed by Coelli et al (2005) and Kumbhakar and Lovell (2000), a firm exploits scale economies when its production function is tangent to the production frontier, marking the point of optimal scale and maximum productivity Even when a firm operates at technical efficiency, it can enhance productivity through scale change effects (SCE) Specifically, when the production function exhibits increasing returns to scale (IRS), SCE positively contributes to total factor productivity (TFP) growth, whereas decreasing returns to scale (DRS) negatively impact TFP growth In cases where constant returns to scale (CRS) are present, there are no scale effects on TFP improvement or decline.
1 9 9 4 , KimandHan(2001)showthatthealmostscalecomponentsarenegativeorclosetozer o,whichresultsinadecreaseinTFPgrowth.Inotherwords,Koreanmanufacturerswereo peratingatDRSorCRSduringtheperiodofstudy.KimandShafi'i(2009)whenestima tingTFPgrowthfor thecaseofMalaysianproducersalsoconfirmedthatSCEinfluencesignificantlyon theoverallproductivity;however,theimpactisdifferentacrossindustries.
Figure3:Productionfrontier,Technologicalprogress,Technicalefficiency,a n d opt imalScaleofproduction
F’0:Productionfrontierattime0F’1:Pro ductionfrontierattime1x:input y:output A:thefirmhastechnicalefficiencyattime0B:thefi rmhastechnicalinefficiencyattime1C:theoptim alscaleattime0
Allocativeefficiency(AE)
Allocative efficiency is a crucial component of Total Factor Productivity (TFP) growth, focusing on the optimal selection of input mixes, such as capital and labor, to minimize production costs and enhance firm productivity According to Schmidt and Lovell (1979), allocative inefficiency arises when the marginal cost of an input does not equal its marginal revenue product, resulting in an inefficient production process A firm achieves allocative efficiency by selecting the appropriate proportions of inputs, which can lead to reduced production costs and increased output levels while keeping input levels constant, or by decreasing input levels with fixed output levels Graphically, firms can be either technically efficient, where the output lies on the efficient isoquant, or allocatively efficient, where the output value aligns with the isocost line The combination of allocative and technical efficiency results in economic efficiency, defined as the point where the output value intersects the isoquant curve and the isocost line.
Haggar(2011).Ontheotherhand,estimationofKimandHan(2001)forKoreanfirmsexpre ssedthatAEhadanegativeimpactonTFPgrowth.Ino t h e r words,therewereanineffici entallocationofinputsinproduction.Theyalsoimpliedt h a t t h e degreeo f c a p i t a l marketd i s t o r t i o n mightb e t h e c a u s e o f AEdifferenceacrossindustries:The allocativeinefficiencywasmoreclearlyobservedinindustriessupportedbythegovernment.H owever,allocativee f f i c i e n c y i s a l m o s t unlikelytobeestimatedinempirical studieswhenthereis theunavailabilityofdatao n costsandprices(Sharmaetal.,2007).
Theisocostline,representingthecombinationsofinputsthat minimizethecosts x1,x2: inputq: output R:ThefirmisallocativeefficientbutnottechnicallyefficientQ : Thefirmi stechnicallyefficientbutnotallocativeefficient.Q’:Thefirmisbothtechn icallyandallocativeefficient, indicatingthatitreachesfulleconomicefficiency.
ApproachestomeasureanddecomposeTFPgrowth
Primalordualapproachwithproduction,cost,orprofitfunction
From neo- classicalperspective,TFPgrowthcanbemeasuredastheresidualf a c t o r bydeductingi nputgrowth(laborandcapital)fromoutputgrowth,expressedinthefamousmodelofSol ow(1957),knownas“growthaccounting”.TheoriginalSolowmodelisalsocalledthe“pr imalresidual”approach.Studiesadoptingprimala p p r o a c h oftenuseproductfunction s tomeasureTFP.The productionfunctionof afirmisconsideredthetechnologicalpossibilitiesofthatfirmtoproduceanou tputusingsomeamountsofinputs.
Aproductionfunctionshouldhaveseveralpropertiess u c h asnon- negativity,weakessentiality,monotonicity,andconcavity(Coellietal.,2 0 0 5 ).
The cost function is designed to identify the input quantities that minimize costs within the set of all technically feasible input-output combinations, based on current technology In a similar vein, the profit function addresses the problem of maximizing profits given a specific amount of input Utilizing cost or profit frontiers can be particularly advantageous, as they measure economic efficiency in addition to technical efficiency (Coelli et al., 2005) Adopting the cost or profit function represents a "dual approach." For instance, Hsieh (2002) employed the "dual residual" method, which incorporates both quantities and costs of production factors to assess total factor productivity growth in four East Asian countries from 1966 onward.
1 9 9 1 HisestimatesseemtoexplainmoreexactlytheeconomicgrowthoftheseAsiancountri esbecausecostsoffactorsreflectactualmarketconditionsbetterthanquantities.How ever,thedualapproachisdatademanding;itrequirestheinformationo f factorpricesofproduction, whichareoftendifficulttoobtaininVietnameseyoungindustrieslikehigh- tech.Forthatreason,primalapproachusingproductionfunctioni sp r e f e r r e d w h e n s t u d y i n g T F P g r o w t h o f Vietnameseh i g h - t e c h manufacturerb e c a u s e itonlyneedsdataonquantitiesofinputsandoutputs.
Stochasticanddeterministicapproaches
Deterministic production functions fail to account for random events or external factors affecting output, resulting in any deviation from the production frontier being viewed as inefficiency (Coelli et al., 2005) This approach tends to overestimate technical inefficiency (TI), leading to an underestimation of the contribution of technical efficiency changes to total factor productivity (TFP) growth To address this issue, stochastic frontier models introduce a random variable to represent statistical noise, allowing for a more accurate assessment of production, costs, or profits in relation to the frontier (Sharma et al., 2007) Consequently, these models facilitate the calculation of TFP changes and their component sources within a stochastic environment.
Parametricandnon-parametricmethods
TFPimprovements( r e c e s s i o n ) a n d i t s decompositionc a n b e c a l c u l a t e d b ya d o p t i n g parametricornon-parametricmethods.Popularnon- parametricapproachest h a t canbementionedareindexnumbertechniquesanddataen velopmentanalysis(DEA).RegardingindexnumbertechniquessuchasFisher’s(192 2)orTửrnqvist’s( 1936)indices,themainadvantagesarethattheyareeasytocalculateandn eedonlytw o observations( t w o f i r m s / i n d u s t r i e s o r twop e r i o d s o f timeo f t h e samef i r m )
(Kumbhakar&Lovell,2000).However,thosetechniquescannotanswerthequestionabout sou rcesofTFPchange.Tosolvethis,onecanconductDEA,whichhasstrongpointsthatitdoesn otneedthespecificfunctionalformoftheproduction(orcostorprofit)function.Neverthele ss,DEAestimatorscannotseparatetheimpactsofrandomshocksandinefficiencyfromthechangeo fTFPandalsonotapplicablefortimeseriesdataset(Coellietal.,2005).
Parametric approaches address the distributional form of inefficiency terms and statistical noise, along with restrictions on the underlying technology (Coelli et al., 2005) Researchers frequently utilize several parametric methods, including Least Squares (LS) econometric production models and Stochastic Frontiers (SF) Each method has distinct strengths and drawbacks; for example, LS models are straightforward but rely on a deterministic approach, meaning that all output variation not linked to input changes is treated as technical inefficiency (Kumbhakar & Lovell, 2000) In contrast, SF models can distinguish between variation caused by random events and that attributed to technical efficiency.
SFestimatorsincludingexogenousinefficiencydeterminantscanbeobtainedb y u s i n g OrdinaryLeastSquares(OLS) orMaximumlikelihood(ML).SFscan alsob e estimatedintwo- stepprocedure(OLSatfirsttoestimatetheslopeparametersthenM L atsecondtoestimatethei nterceptandthevariancesoferrorterms)orone-steppr ocedure (MLonly).
OLSistheeasiermethodtocalculatethanthelatter.Thereis,however,thedo wnward b i a s i n t h e normalOLSe s t i m a t o r s o f i n t e r c e p t c o e f f i c i e n t s , Modi fiedLeastSquaresorCorrectedLeastSquaresshouldbeappliedtoshiftupthebiase dOLS i n t e r c e p t parameter(Kumbhakar& Lovell,2 0 0 0 ).Meanwhile,despitether e q u i r e m e n t s ofdistributionalformsandcomputationalissuesinreachingconvergenc e,M L i s arguedt o b e t h e b e t t e r c h o i ce w h e n a n a l y z i n g largesample( Coellietal.,2005).
AreviewofalternativeStochasticProductionFrontier(SPF)models
Time-invariantmodels
PittandLee(1981)areconsideredthepioneersinextendingSPFtopaneldata.Allo win g t h e T I varya c r o s s firmsb u t unchangedthroughtime,theyestimatedparametersusing MLandassumed t he T I errorcomponentto follow n or ma l- ha lf normaldistribution,namely:
� � ~𝑁 + (0,𝜎 2 ), (4) where� �� istheactualoutputlevelofthei th producerattimet th , � �� isa1 x K inputvector and�isaKx1parametervectortobeestimated.Afterward,t h i s modeli s generalizedt o n o r m a l - t r u n c a t e d normalc a s e i n
(�=0isthespecialcaseo f t h e d i s t r i b u t i o n f u n c t i o n , w h i c h makesi t becom eP i t t a n d Lee’s( 1981) model).
� � sareproducer-specificintercepts,whichcanbeestimatedbysuppressing thec o n s t a n t t h e n ad di ng dummyvariablesf o r e a c h firms,o r ap pl yi ng th e w i t h i n transformation(runningOLSregressionaftersubstituting� ��by � �� −�̅ 𝑖and soon).
Afterestimatingtheparameters,theestimatorsofthefixed-effects� �c a n be obtainedfrom:
� �𝑟��̂ℎ =max(�̂ 𝑖 ), (9) Estimatesof��o b t a i n e dfromthefixed- effectspaneldatamodelprovetob e consistent(Kumbhakar&Lovell,2000).However,�
�a l s ocapturetheeffectsofo t h e r eventsthatvaryacrossproducersthatmayormaynotthecauseofTI.Forthatreason,SchmidtandSi ckles(1984)alsoproposedtherandom-effectsSPFmodel:
However,t h e assumptiono f T I e f f e c t s t o betime-invariantseemst o b e unrealisticwithlongpaneldatasetswhenfirmscanimprovetheirperformancevialear ning-by-doingovert i m e T h a t i s t h e reasonw h y somea u t h o r s t r y t o s p e c i f y differentfunctionsof� ��to allowfortime-varyingTIeffects.
Time-varyingmodels
Cornwelletal.(1990)replacedthefirm-specificeffects� �in (5)byafunction oftimeallowingparameterstovaryacrossfirms:
ThisspecificationallowstheTItochangethroughtimewhilestillhavingfirm- specificcharacteristics.However,thispatternrequiresNx3parameterstoestimatef i r m effects,whichleadstotheimpactondegreeoffreedom(Belotti,Daidone,Ilardi,
Toreducethenumbersofparameters,alternativeoftime- varyingTIeffectsw e r e p r o p o s e d b y Kumbhakar( 1 9 9 0 ) ,B a t t e s e a n d C o e l l i ( 1 9 9 2 ) ,a n d Leea n d
Kumbhakar(1990): �(�)= [1+exp(��+��(15)B a t t e s e andCoelli(1992):�(�)=exp[−�(�−� � )], 2 )] −1 , (16) LeeandSchmidt(1993): �(�)=θ � , whereθ�i sasetoftimedummy variables.
In( 1 5 ) ,t h e r e a r e o n l y t w o parametersb andc t oestimate.Ita l l o w s0≤
�(�)≤1and�(�)canbeconcaveorconvex,monotonicallyincreasingo r decreasing.However,theestimationandinferenceoftechnicalefficiencychangecan becomplicated.
>0,���d e c r e a s e sastincreases,andviceversa;while�=0meanst ime- invariantTI,i.e.thefirm-specificeffects.Thismodelhas theadvantagethat thereis onlyoneparameter�thatneedstobeestimatedanditiseasytoconcludeaboutt h e techni calefficiencyaswellasitschangeovertime.However,thisspecificationo n l y allowsTEtoincreaseatadecreasingrate(�