Catchng up or fallng behnd n Eastern European agrculture the case of mlk producton Lukas Cechura, Aaron Grau 2, Henrch Hockmann 2, Inna Levkovych 2, Zdenka Kroupova Czech Unversy of Lfe Scences Prague Faculty of Economcs and Management, Department of Economcs Kamycka 29 Prague, Czech Republc e-mal: cechura@pef.czu.cz 2 IAMO Agrcultural Markets, Marketng and World Agrcultural Trade Theodor-Leser-Str.2 Halle, Germany e-mal: hockmann@amo.de Paper prepared for presentaton for the 42 nd EAAE Semnar Growng Success? Agrculture and rural development n an enlarged EU May 29-30, 204 Corvnus Unversy of Budapest Budapest, Hungary Copyrght 204 by Lukas Cechura, Aaron Grau, Henrch Hockmann, Inna Levkovych, Zdenka Kroupova. All rghts reserved. Readers may make verbatm copes of ths document for non-commercal purposes by any means, provded that ths copyrght notce appears on all such copes.
Abstract: The paper deals wh the analyss of catchng up and fallng behnd processes n European agrcultural sector usng a stochastc fronter multple output dstance functon for 24 EU Member States n the perod 2004 20. The metafronter estmates reveal that there are consderable productvy dfferences n mlk producton across the EU at the NUTS 2 level. Productvy s the hghest n the Old Member States, especally n those regons located n the Northwest of the EU. The lowest productvy s observed n Eastern Europe. The same structure as for TFP was found for TFP development. Moreover, the results for techncal change suggest that farm szes are not optmal n many regons n Central and Eastern Europe from a dynamc perspectve. The comparatve analyss suggest that n the New compared to the Old Member States fewer farms could benef from the movement on the fronter. Moreover, there are no sgns that poor performng farms are catchng up to the best performng farms n the regons/countres. Key words: Productvy, effcency, New Member States, metafronter analyss, SFA. JEL Classfcaton: D 24, O 2, P 27 INTRODUCTION The recent accesson of the New Member States (NMS) to the European Unon (EU) has been a mammoth task, accompaned by many sde dstortons for themselves as well as for the old member states. Nevertheless, the economc theory of ntegraton suggests that deeper ntegraton among new and old member states would net-benef mutually ther domestc economes. Steps taken towards a deeper ntegraton have been the abolshment of border controls, the creaton of common, congruent polces regardng competon, state ad, and food safety and envronmental standards as well as regulatons, and the establshment of an ntegrated capal market. These polcy actons tred to set the foundatons for prospery va enhancng cross-border actves. One of the man challenges of the ntegraton process has been hherto the adopton of the CAP for the NMS. A successful ntroducton of the CAP s of hgh sgnfcance to the entre ntegraton process, snce has been the EU s man fnancal reallocaton channel of the past sx decades and the economes of the NMS have been orentated largely towards the prmary sector. Several attempts to study these effects have already been carred out (Bakucs et al. 200, Csák and Jámbor 200 and 203). However, the researchers predomnantly focused on one or a related group of countres (Latruffe et al. 202a and b, Cechrua 202). Moreover, f the comparson was done among countres, was usually based on country specfc model estmates whch do not provde conclusve results or can be msleadng respectvely. In ths paper we wll concentrate on the agrcultural sector, n partcularly on the mlk producton, and nvestgate whether the ntegraton processes have had the benefs mentoned above. We wll measure the benefs n terms of TFP development and the mpact of techncal change n agrculture. In partcular, we wll address the followng research questons: How has regonal agrculture benefed from the adopton of nnovatons? Is there an ndcaton that regonal scarcy of resources was the source of techncal progress? Is there an ndcaton that techncal change (and other sources of adustment) have led to a convergence of the regons n terms of TFP?
What are the causes of the regonal convergence or dvergence? How dd farm structure, market structure, and cred constrants affect the adopton of nnovatons? Gven the sgnfcant dualy of farm structures n some countres can the same patterns for large agrcultural companes and small farmers be observed or can we dentfy dosyncratc developments? The structure of the paper s as follows. Chapter 2 contans the theoretcal framework and presents the estmaton strategy. Chapter 3 descrbes the data set. Chapter 4 presents descrptve analyss of dary producton. Chapter contans estmaton results, compares and dscusses the estmated technology, technologcal change, trends n techncal effcency and TFP developments, and provdes the results of the metafronter analyss. Chapter 6 contans a dscusson of the results and concludng remarks, ncludng polcy recommendatons. 2 THEORETICAL FRAMEWORK AND ESTIMATION STRATEGY The theoretcal background s gven by neoclasscal producton economcs, especally productvy and effcency analyss (Fred et al., 2008). The research questons wll be dealt wh () estmaton of country specfc multple output dstance functon for farms specalzng on mlk producton usng the FADN database. (2) Based on the estmated parameters the effcent output level wll be calculated. These wll be used n a metafronter approach to determne the TFP level and development. In order to produce coherent results, all models (the country specfc models n () as well as the metaproducton models n (2)) wll make use of the same procedure: The models are formulated as output dstance functons wh three outputs and fve nputs. In all models s consdered explcly that agrcultural producton possbles are affected by frm heterogeney, whch affects the level as well as on the shape of the producton possbles. 2. Multple output dstance functon We assume that producton possbles can be well approxmated by the output dstance functon ntroduced by Shephard (970): D O (x, y) = mn θ: ( y ) P(x), () θ where y stands for the output vector, y ϵ R M +, and x denotes the nput vector, x ϵ R K +, P(x) represents the output set, such as: P(x) = {y R + M : x can produce y}. (2) As provded by Coell et al. (200), D O (x,y) exhbs the followng propertes. It s nondecreasng, posvely lnearly homogenous and convex n y, and decreasng and convex n x. Moreover, holds that D O (x,y) f y ϵ P(x) and D O (x,y) = f y ϵ Isoq P(x). In our applcaton we use a translog functonal form, snce s flexble and provdes a good approxmaton of the producton process. In addon, perms the mposon of homogeney (Coell and Perelman, 996). The translog output dstance functon for 3 outputs and nputs, as s the case n our emprcal applcaton, s: D O = α 0 + 3 m= α m ln y m + 3 3 α 2 m= n= mnlny m lny n + k= β k ln x k + + β 2 k= l= kllnx k lnx l + γ 2 n= kmlnx k lny m 3 k= (3) where subscrpts, wh =,2,,N, and t, wh t =,,T, refer to a certan producer and tme (year), respectvely. α, β, and γ are vectors of parameters to be estmated. Output dstance functon s homogenous of degree n outputs. Ths requres:
3 m= α m =, 3 n= α mn = 0, for m =,, 3, and (4) 3 m= γ km = 0, for k =,,. Symmetry restrctons are as follows: α mn = α nm, and β kl = β kl. () Followng Lovell at al. (994) the homogeney s mposed by choosng one output and dvdng the other outputs by. Thus, we get: lnd O lny = α 0 + 3 m=2 α m ln y m + 3 3 α 2 m=2 n=2 mnlny m lny n + k= β k ln where y m = y m y. + β 2 k= l= kllnx k lnx l + γ 2 k= m=2 kmlnx k lny m If we ntroduce statstcal nose, v, and assocate lnd O wh neffcency term, u = -lnd O, we get a stochastc fronter multple output dstance functon: lny = α 0 + 3 m=2 α m ln y m + 3 3 α 2 m=2 n=2 mnlny m lny n + k= β k ln x k + β 2 k= l= kllnx k lnx l + γ, (7) 3 2 k= m=2 kmlnx k lny m + u + v 2 + 2 where we assume that v ~ N(0, σ v ), u ~ N (0, σ s ), and they are dstrbuted ndependently of each other and of the regressors (Kumbhakar and Lovell, 2000). Productvy fnds s expresson n the shape of (7), and thus n the parameter vectors (α, β, γ). Snce the coeffcents depend on the qualy of the ndvdual nputs and nput qualy s determned by the embedded knowledge,.e. human capal for labour, technologcal knowledge for capal, and embedded nnovaton n materals (Barro and Sala-I-Martn, 99), technology mproves over tme due to technologcal progress and learnng by dong. Ths wll not only nduce shfts n the output dstance functon but wll also affect the productvy of ndvdual nputs. Moreover, can be assumed that the varous mprovements n qualy have rather dfferent drect and ndrect effects on the ndvdual nputs. However, due to lmatons n data avalably, the mpacts of the varous mprovements cannot be estmated separately. Instead, s commonly assumed that a trend varable (t) can be ncorporated whch captures the ont effects n nput qualy mprovements. Proceedng ths way, the resultng functon s: lny = α 0 + 3 m=2 α m ln y m + 3 3 α 2 m=2 n=2 mnlny m lny n + k= β k ln + β 2 k= l= kllnx k lnx l + γ 2 k= m=2 kmlnx k lny m +δ t t + δ 2 ttt 2 + 3 m=2 α mt tlny m + k β kt tln x k + u + v Here δ t and δ tt captures the global effect whle the α mt and β kt measure the bas of techncal change. Stochastc fronter output dstance functon n (8) wll play a central role n our emprcal applcaton. 2.2 Heterogeney n technology Heterogeney n technology s captured usng a fxed management model. Álvarez et al. (2003 and 2004) specfed the fxed management model as a specal case of random parameters model n the followng form. The technology s gven by the consderaton of a 3 3 x k x k (6) (8)
frm specfc factor (m *) whch enters the dstance functon n the same way as techncal change (8): lny = α 0 + 3 m=2 α m ln y m + 3 3 α 2 m=2 n=2 mnlny m lny n + k= β k ln + β 2 k= l= kllnx k lnx l + γ 2 k= m=2 kmlnx k lny m +δ t t + δ 2 ttt 2 + 3 m=2 α mt tlny m + k β kt tln x k +α m m + α 2 mmm 2 + δ tm m t + β km m lnx k + u + v Wh m * the frm uses s productve capaces optmally. However, the realsaton of the frm specfc factor may be suboptmal (m ). The dfference between the real (m ) and optmal level of m determnes techncal effcency. Techncal effcency, TE, wh 0 < TE <, captures devatons from the maxmum achevable output. Usng the defnon of the dstance functon n (9) provdes: lnte = u = lny m lny m 0 Substutng (9) nto (0) and rearrangng terms provds: lnte = γ 0 + γ t + γ x ln x, () t 2 2 where γ β ( m m ) + β ( m m ) γ 0 = m mm ( m m ) ( m m ) t = βtm γ x = β xm Thus, techncal effcency conssts of three components: 2 3 k= () tme nvarant frm specfc effect pure heterogeney γ 0, () nteracton of m * wh tme technologcal change γ t, () nteracton of m * wh nputs scale effect γ x. The model (9) cannot be estmated by maxmum lkelhood snce m * s not observable. Álvarez et al. (2004) propose a maxmum smulated lkelhood approach where m * s m ~ N 0,. smulated by several draws for the standard normal dstrbuton, e. g. ( ) The m * can be fted accordng to (Álvarez et al. 2004): [ y, X, δ] Eˆ m * * ( y y, x, t, m ; α,β, γ,δ) x k (9) (0) R ˆ m, r f m, r R r= =, (3) R ˆ * * f ( y y m, x, t, m ; α,β, γ,δ) R r= where R denotes the number of repetons fˆ denotes the value of the rght term n (9). Moreover, once the m * are determned, the u can be estmated wh Jondrow at al. (982) formula:
[ ε m ] E u ( ( ε m ) λ / σ ) ( ( ε m ) λ / σ ) ( ε m ) σλ φ λ, =, (4) 2 ( + λ ) Φ σ σ u 2 2 2 where λ =, σ = σ u + σ v, ε = v + u and φ and Φ denote the densy and dstrbuton σ v of the standard normal. The software NLOGIT.0 was used n the applcaton. 2.3 TFP calculaton and decomposon Total factor productvy s calculated n the form of the Törnqvst-Thel ndex (TTI) (see, e.g. Čechura and Hockmann, 200). The TTI exactly determnes the changes n producton resultng from nput adustments f a functon has the translog form (for proof see Dewert, 976). Furthermore, Caves et al. (982) present a TTI extenson for multlateral consstent comparsons. The ndex s constructed as the devaton from the sample means. The nput ndex for varable returns to scale (VRS), or constant returns to scale (CRS), respectvely, s gven by: lnι VRS resp. lnι CRS K = 2 = = 2 K = ( ζ + ζ ), 0 K wh ζ =, 0 ζ, 0 ln x, ln x ln f ζ =, 0 + K ζ = ζ 0 ln x + * * ( y, x, t, m ; α,β, γ,δ) m, ln x ζ ln x ζ ln x, + K =, 0 ζ ζ ln x 0, ln x, ζ K =, 0 ζ, 0 ln x, (). (6) A bar over a varable specfes the arhmetc mean over all observatons. That s, the output ndex and the effcency ndex are defned as: lnψ = ln y ln y and lnυ = lnte lnte. (7) Snce TFP s a combnaton of scale effect, techncal effcency, technologcal change and management effect, the requred ndces are defned as: * * ln f [( )( ) ] ( y m, x, t, m ; α,β, γ,δ) ζ + ζ t t + ζ t ζ t, wh ζ = lnτ = t t t t t, (8) 2 t ln µ = m0 2 ln f wh ζ = m [( ζ + ζ )( m m ) + ζ m ζ m ], m * * ( y, x, t, m ; α,β, γ,δ) m m Usng these defnons, TFP and s breakdown s gven by: m m. (9) lntfp = lnψ lnι CRS = lnι + lnυ SE TE + lnτ + ln µ, TCH MAN wh lnι = lnι VRS lnι CRS. (20)
Changes n TFP can be expressed eher as a rato (on the mean) of the output and nput ndex (for CRS) or as a multplcaton of TFP components,.e., scale effect (SE), techncal effcency effect (TE), technologcal change effect (TCH) and management effect (MAN). 2.4 Metafronter analyss The metafronter analyss wll be conducted usng the same model specfcaton as for the ndvdual countres. We wll calculate the effcent output based on the parameter estmates of country multple output dstance functon and wll use them n the estmaton of stochastc metafronter multple output dstance functon. We use agan 3 outputs and nputs. Moreover, we wll employ agan the Fxed Management model to capture the heterogeney. The estmated metadstance wll allow a coherent comparson of TFP levels among the EU member countres. 3 DATA The panel data set s drawn from the FADN database provded by the European Commsson. The data set contans data on 24 EU member countres (Cyprus, Malta and Luxemburg were excluded) and covers the perod from 2004 to 20 except for Austra (200 20), Bulgara and Rumana (2008 20). The analyss focuses on mlk producton and uses the followng data: y mlk producton, y 2 other anmal producton, y 3 plant producton, x labour, x 2 land, x 3 capal, x 4 specfc materal and x other materal. Labour s represented by the total labour measured n AWU. Land s the total utlsed land. Capal s a sum of contract work and deprecaton. Specfc materal creates cost on feed for grazng lvestock. Outputs as well as nputs (except for labour and land) are deflated by country prce ndexes on each ndvdual output and nput (200 = 00). The country prce ndexes are taken from the EUROSTAT database. The multple output dstance functon s estmated only for specalzed producers. The specalzaton s defned as at least 0 % share mlk producton on total anmal producton. Snce not all nformaton can be found n the database, only those producers havng non-zero and posve values are used for the varable of nterest. Moreover, we reected producers wh less than fve observatons to decrease the problem wh entry and ex of the producers from the database. The country sample descrptve statstcs are provded n the Appendx Table A. 4 DISTRIBUTION OF DAIRY PRODUCTION Fgure provdes nformaton on the mportance of mlk producton for agrcultural holdngs n form of s shares n total agrcultural ncome. The hghest share of mlk producton n total agrcultural producton s observed n the area rangng from the Scandnavan and Baltc countres to Northern France. Here mlk producton accounted for more than 30% of agrcultural gross producton. The regons n the South of Europe specalzaton towards mlk producton s less pronounced. The same holds for the new EU member states.
Fgure : Share of dary gross producton n total agrcultural gross producton, NUTS2 level Source: own calculaton The specalsaton s measured consderng all farms n the data set. However, does not provde nformaton about the ntensy of mlk producton. Fgure 2 gves nformaton on the ntensy, measured by the mlk delvery per cow on NUTS2 level. The hghest mlk producton per cow s observed n the Northwest of the EU, especally Denmark, Fnland and the Netherland. Moreover, the old member states have a hgher ntensy as the NMS. Exceptons are the Czech Republc and some regons n Hungary and Slovaka, whch reach ntensy levels comparable wh the medum performng regons n the old member states. However, mlk producton per cow s only a poor ndcator of the productvy of dary producton, snce s a partal ndcator not consderng the other factors of producton. In order to provde a more comprehensve measure these have to be taken nto account. Ths leads to the calculatons of total factor productvy as shown n the next chapter.
Fgure 2: Productvy of dary gross producton (mlk yeld per cow), NUTS2 level Source: own calculaton ESTIMATION RESULTS. Metafronter analyss Table provdes the parameter estmates of the stochastc metafronter model for mlk producton usng an output dstance functon. Almost all parameters are sgnfcant even at % sgnfcance level. Ths also holds for the maory of the other fted parameters. As far as theoretcal consstency s concerned, the estmated model mples that the estmaton should nher the propertes of an output dstance functon. Accordng to Coell et al. (200), the output dstance functon should be non-decreasng, posvely lnearly homogenous and convex n outputs, as well as decreasng and quas convex n nputs. That s, the monotoncy requrements for outputs mply: βy 2 > 0, βy 3 > 0 and βy 2 + βy 3 < ; and for nputs: βx q < 0 for q=, 2, 3, 4,. Table shows that these condons are met. Moreover, convexy n nputs requres βx qq +βx q 2 βx q > 0 for q =, 2, 3, 4,. Ths condon holds evaluated on the sample mean. Snce all varables are normalsed n logarhm by ther sample mean, the frst-order parameters of outputs represent the shares of outputs y 2 and y 3 n the total output. Snce we analysed farms specalzed n mlk producton s natural that the shares of plant producton and other anmal producton s relatvely small (βy 3 = 0.272 and βy 2 = 0.0726). The parameters of nputs can be nterpreted as elastces of producton on the sample mean. The hghest producton elastcy exsts for materal nputs (x 4 and x ) and the lowest for capal (x 3 ). The estmates further show that heterogeney s an mportant determnant of dary producton n the EU. It contrbutes posvely to the producton and the mpact s
acceleratng. The ncrease n heterogeney (suably of dary producton) has a posve mpact on producton elastces of materal nputs and a negatve one on labour, land, and capal. The effect of technologcal change on techncal effcency s negatve wh ncreasng management. The sum of producton elastces s 0.889. Thus, decreasng returns to scale were estmated for the EU member countres. The estmates reveal that the scale effcency wll have a sgnfcant mpact on productvy change, evaluated on the sample mean. However, ths also holds for most ndvdual EU member countres (see next secton). Table : Parameter estmates metafronter for dary Means for random parameters Coeffcent on unobservable fxed management Varable Coef. SE P [ z >Z*] Varable Coef. SE P [ z >Z*] Const. -0.6 0.000 0.0000 Alpha_m -0.38 0.0006 0.0000 Tme -0.0076 0.0002 0.0000 Tme -0.002 0.0002 0.0000 X -0.072 0.00 0.0000 X -0.0607 0.000 0.0000 X2-0.398 0.0008 0.0000 X2-0.0386 0.0006 0.0000 X3-0.069 0.0008 0.0000 X3-0.0082 0.0008 0.0000 X4-0.32 0.0006 0.0000 X4 0.087 0.0006 0.0000 X -0.2893 0.00 0.0000 X 0.04 0.000 0.0000 Alpha_mm -0.07 0.0007 0.0000 Varable Coef. SE P [ z >Z*] Varable Coef. SE P [ z >Z*] TT -0.0009 0.0002 0.0000 X3-0.0039 0.00 0.0003 Y2 0.0726 0.000 0.0000 X4 0.0489 0.000 0.0000 Y3 0.272 0.0004 0.0000 X 0.0232 0.00 0.0000 Y2T -0.00 0.0002 0.0000 X23-0.0090 0.0007 0.0000 Y3T 0.000 0.000 0.000 X24 0.0344 0.0007 0.0000 Y22 0.030 0.000 0.0000 X2 0.0020 0.0009 0.0297 Y33 0.079 0.0003 0.0000 X34 0.0220 0.0008 0.0000 Y23-0.0024 0.0003 0.0000 X3 0.0 0.0009 0.0000 XT 0.003 0.0003 0.0000 X4 0.0097 0.0009 0.0000 X2T 0.0029 0.0002 0.0000 Y2X -0.0066 0.0008 0.0000 X3T -0.00 0.0002 0.0000 Y2X2 0.0068 0.0006 0.0000 X4T -0.0034 0.0002 0.0000 Y2X3 0.0024 0.0006 0.0002 XT 0.0029 0.0003 0.0000 Y2X4 0.0062 0.000 0.0000 X -0.0444 0.002 0.0000 Y2X -0.0099 0.0008 0.0000 X22-0.0230 0.00 0.0000 Y3X -0.006 0.0006 0.0000 X33-0.080 0.000 0.0000 Y3X2 0.0083 0.000 0.0000 X44-0.22 0.0007 0.0000 Y3X3-0.0049 0.000 0.0000 X -0.0723 0.006 0.0000 Y3X4 0.0049 0.0003 0.0000 X2-0.0033 0.004 0.07 Y3X -0.0022 0.0006 0.0006 Sgma 0.48 0.0004 0.0000 Lambda.226 0.036 0.0000 Note: ***, **, * denotes sgnfcance at the %, %, and 0% level, respectvely Source: own calculaton Fgure 3 shows the dstrbuton of total factor productvy for specalzed mlk producers n the EU at the NUTS 2 level. The hghest TFP s observed n Northern Central Europe (Denmark, Belgum, Germany), Northern Italy and France, and some regons n Span. In Eastern Europe above average TFP s only observed n some Polsh regons. In general, the
average TFP s Eastern Europe farly lackng behnd the EU average wh the lowest TFP levels observed n Romana and Bulgara. Moreover, most regons n the old memebr states experenced an above average TFP growth. In Central and Eastern Europe ths holds only for the Czech Republc, Hungary and Slovaka. The Baltc countres, Poland, Slovena, as well as Bulgara and Romana had an below average TFP growth rate. Gven that ths group of countres had also below average TFP levels, can be concluded that TFP n mlk producton was more and more fallng behnd those of the old member states. Only n the three earler mentoned countres there are sgns of a catchng up process. A TFP only ltle below the EU average was accompagned wh an above average TFP development. Fgure 3: TFP dfferences among regons, NUTS2 level Source: own calculaton.2 Country multple output dstance functon estmates and TFP calculatons Tables 2, 3, and 4 provde parameter estmates of the multple output dstance functon (relaton 0) for 23 EU member countres (the multple output dstance functon for Greece could not be estmated due to the low number of observatons). Instead of dscussng each country estmate separately, we wll evaluate and compare the results for all member countres together. Ths strategy helps to understand better the common and ndvdual specfcs of mlk producton n EU member countres as far as technology, effcency, and productvy are concerned. We start wh the dscusson of the frst order parameters and economes of scale (Table 2). Then we verfy the sgnfcance of heterogeney n producton structure. In partcular, we evaluate the parameters on unobservable fxed management (Table 3). Fnally, we concentrate on technologcal change and based technologcal change (Table 4).
.2. Parameter estmates Table 2 provdes selected estmated parameters of the output dstance functon,.e. frst order parameters on outputs and nputs. Almost all parameters are sgnfcant, even at % sgnfcance level. Ths also holds for the maory of the other fted parameters. Moreover, the monotoncy and convexy condons holds for all countres at the sample mean. Table 2: Frst order parameters of the multple output dstance functons mlk producton EU member country Other Anmal producton Plant producton Labour Land Capal Specfc materal Other materal y2 y3 x x2 x3 x4 x Austra Coeff. 0.69 0.06-0.0436-0.2823-0.49-0.288-0.7 Belgum Coeff. 0.0863 0.093-0.0776-0.939-0.0439-0.2227-0.2800 Germany Coeff. 0.29 0.288-0.2-0.2373-0.0986-0.2648-0.3063 Denmark Coeff. 0.0370 0.26-0.0749-0.77-0.09-0.23-0.2892 *** *** *** *** * *** *** Span Coeff. 0.034 0.0849-0.639-0.0390-0.032-0.4243-0.82 Fnland Coeff. 0.030 0.0907-0.0979-0.93-0.0933-0.2809-0.23 France Coeff. 0.0864 0.32-0.0924-0.9-0.8-0.29-0.30 Great Bran Coeff. 0.0697 0.86-0.083-0.0827-0.0676-0.48-0.377 Ireland Coeff. 0.394 0.062-0.0960-0.220-0.074-0.263-0.302 Italy Coeff. 0.0906 0.408-0.07-0.689-0.0897-0.2-0.00 Netherlands Coeff. 0.0293 0.07-0.0873-0.380-0.0864-0.330-0.684 Portugal Coeff. 0.07 0.726-0.40-0.06-0.020-0.482-0.20 Sweden Coeff. 0.0223 0.3663-0.02-0.2230-0.0438-0.3808-0.2602 Bulgara Coeff. 0.476 0.3648-0.932-0.222-0.0909-0.3779-0.2466 Czech Republc Coeff. 0.090 0.4207-0.467-0.80-0.034-0.2836-0.34 *** *** *** *** ** *** *** Estona Coeff. 0.080 0.3-0.3-0.282-0.0800-0.4832-0.2333 Hungary Coeff. 0.082 0.340-0.49-0.0706-0.00-0.364-0.3732 Lhuana Coeff. 0.0 0.492-0.03-0.202-0.0896-0.393-0.3249 Latva Coeff. 0.63 0.4493-0.069-0.0292-0.09-0.4290-0.3269 *** *** *** *** *** *** Poland Coeff. 0.0946 0.322-0.08-0.2439-0.32-0.233-0.372 Romana Coeff. 0.829 0.4990-0.074-0.3282-0.07-0.26-0.284 *** *** *** *** ** *** *** Slovena Coeff. 0.269 0.268-0.0880-0.2877-0.3-0.3994-0.2422 Slovaka Coeff. 0.060 0.3043-0.333-0.2020-0.070-0.20-0.2270 Note: ***, **, * denotes sgnfcance at the %, %, and 0% levels respectvely. Source: own calculaton RTS -0.9008-0.88 -.090 -.000-0.8229-0.9003-0.9262 -.0227-0.9360 -.0236-0.9907-0.903 -.002 -.0308-0.9693 -.0760-0.9729 -.0404 -.0029 -.070-0.8979 -.30 -.0246
The frst order parameters of outputs (y 2 and y 3 ) pont out the producton structure dfferences among EU member countres. Snce we analysed farms specalzed n mlk producton wh the share of mlk producton n total anmal producton exceedng 0%, the share of other anmal producton n total output s lower than 0% for all analyzed countres. Specalzed mlk farms wh a hgher share of other anmal producton can be found n Romana, Austra, and Lhuana, where the parameter of y 2 exceeds 0.. Agrcultural companes n Romana can be characterzed also by the hghest share of plant producton, almost 0%, pontng to the hgh producton dversfcaton on Romanan farms as well as to a hgh proporton of own feed producton. The share of plant producton s hgher than 40 % also n the Czech Republc, Italy, Lhuana, and Latva. On the other hand, farms n Austra, Span, Fnland, and the Netherlands are hghly specalzed n anmal producton. The share of plant producton n ther total output s lower than 0 %. The producton elastces of the ndvdual countres have some common patterns. The elastces for materals nputs (specfc and other materals) have the hghest values and the elastces for capal the lowest. However, some exceptons can be found. In the case of Slovaka, surprsngly labour has the hghest elastcy. Ths suggests low capal ntensy n dary cows breedng n Slovaka. Romana s another excepton, where the prevalng pasture breedng leads to the hgh mpact of land on mlk producton. On the other hand, land has the lowest mpact n Span where land elastcy s -0.04. As far as economes of scale are concerned they are slghtly devatng from. Constant returns to scale were estmated (the sum of the elastces s about one) for the average farm n Germany, Great Bran, Hungary, Italy, Latva, the Netherlands, Sweden, and Slovaka. On the contrary, the mpact of scale effcency on productvy change can only be assumed n other EU member states, where the returns to scale are eher ncreasng (Bulgara, Denmark, Estona, Lhuana, Poland, and Slovena) or decreasng (Austra, Belgum, the Czech Republc, Span, Fnland, France, Ireland, Portugal, Romana). Ths suggests that n most countres the average farm operates almost at optmal (statc) farm sze. Ths mples that optmal farm sze s a functon of many dfferent determnants as heterogeneous as the condons n the countres. Table 3 provdes the parameter estmates on unobservable management. The coeffcents on unobservable management are hghly sgnfcant n the maory of cases. Ths suggests that the estmated relatonshp s approprately approxmated by chosen specfcaton and that heterogeney among farms s an mportant characterstc for mlk specalzed producers n EU member states. The effect of unobservable management s posve (α m < 0) and predomnantly deceleratng (α mm > 0). The mpact of unobservable management on producton elastces dffers sgnfcantly among the analysed countres. In general, better suably for mlk producton goes hand n hand wh a more productve use of n specfc materals (fodder) (β mx4 > 0). On the other hand, suably for mlk producton leads to the decrease n labour, land, and capal elastcy (β mx < 0, β mx2 < 0, and β mx3 < 0).
Table 3: Parameters on unobservable heterogeney mlk producton EU country Austra Belgum Germany Denmark Span Fnland France Great Bran Ireland Italy Netherlands Portugal Sweden α m Tme Labour Land Capal Specfc Other materal materal α mm t x x2 x3 x4 x Coeff. -0.266 0.0039-0.0038 0.2042 0.0778 0.027 0.07 0.73 Coeff. -0.24-0.0043 0.030 0.00-0.0078 0.0229-0.044 0.028 *** *** *** *** *** *** Coeff. -0.2077-0.00 0.003-0.096 0.022 0.042-0.0392 0.039 Coeff. -0.3-0.002-0.0-0.04-0.0034 0.0374 0.034 0.029 *** ** ** *** *** *** Coeff. -0.226 0.00-0.0776-0.02-0.004 0.9 0.074-0.0337 *** *** *** *** ** *** *** *** Coeff. -0.092 0.0064-0.0962-0.033-0.0243-0.0839 0.0002 0.3363 Coeff. -0.202-0.0036-0.0278-0.027 0.02 0.0487-0.03 0.07 *** Coeff. -0.668 0.0043-0.06-0.0344 0.0083 0.0389 0.0067 0.009 *** *** ** *** *** *** Coeff. -0.2727-0.00-0.0627 0.0489-0.044 0.0393-0.349 0.2374 *** Coeff. -0.283 0.004-0.067-0.086 0.0063 0.37-0.0073-0.0369 *** *** *** *** * *** *** *** Coeff. -0.0784 0.0004-0.078 0.04-0.0474-0.0762-0.09 0.2693 Coeff. -0.0868-0.000-0.09-0.82-0.024 0.39-0.29 0.39 *** Coeff. -0.2234 0.0002 0.04-0.042 0.083 0.007-0.76 0.94 *** *** *** *** *** *** Bulgara Coeff. -0.623 0.027-0.02-0.066-0.063 0.096 0.093-0.096 *** *** ** ** *** ** *** *** Czech Coeff. -0.0466-0.0033 0.08-0.2099 0.0344 0.009 0. 0.3494 Republc *** *** *** *** *** ** *** *** Estona Coeff. -0.49 0.0008 0.09-0.0793-0.0062 0.08 0.02 0.046 *** * *** *** *** Hungary Coeff. -0.3-0.00-0.0033 0.076-0.06-0.2 0.4 0.0999 *** *** *** *** *** *** Lhuana Coeff. -0.093 0.0076 0.0074 0.084 0.0343-0.067-0.0489 0.28 Latva Coeff. -0.082 0.06-0.007 0.0204-0.002 0.0476-0.0342 0.047 *** *** *** *** Poland Coeff. -0.20-0.0037-0.034-0.039 0.00 0.0227 0.042 0.009 *** *** ** *** *** *** *** Romana Coeff. -0.26-0.006-0.047-0.0386-0.000 0.0243 0.0379-0.0608 *** * *** *** *** *** *** Slovena Coeff. -0.824-0.0049-0.02-0.0864-0.002 0.072 0.0497-0.027 *** ** *** *** *** *** Slovaka Coeff. -0.39-0.02-0.206-0.048-0.0237 0.072 0.088 0.089 ** Note: ***, **, * denotes sgnfcance at the %, %, and 0% levels respectvely; LNO Low Number of Observatons Source: own calculaton
Gven the assumpton regardng the unobservable component, m * ~ N(0,), the value of the constant term can be regarded as the standard devaton of the determnant n the sample. In general, there s no pronounced dfference betwen the average value of ths determnant between Old and New Member States. Ths shows that the country groups have smlar dstrbutons or the suably of mlk producton, however, on a dfferent levels (see Fgure 3). Moreover, there s ndcaton that n countres whch have a hgh TFP, lke Denmark and Netherlands, the suablly of mlk produkton s relatvely homogenously dstrbuted. The same effect s observarble for the New Member States (e.g. Czech Republc). Table 4: Technologcal change and based technologcal change mlk producton EU country t tt x*t x2*t x3*t x4*t x*t Austra Coeff. -0.0070-0.0080-0.0090 0.0026-0.020-0.008 0.004 *** *** *** ** *** * Belgum Coeff. -0.0043 0.0003 0.0038 0.063 0.0024-0.008-0.062 *** *** *** *** Germany Coeff. -0.032-0.00-0.00-0.008-0.007-0.0034 0.0 *** ** *** * *** *** Denmark Coeff. -0.043 0.007 0.049 0.0002 0.086-0.0207-0.03 *** *** *** *** *** ** Span Coeff. 0.0038 0.002 0.037 0.009 0.002-0.0279-0.0022 *** ** *** * *** *** Fnland Coeff. -0.08 0.0098 0.002 0.0044-0.0092-0.0 0.0033 *** *** *** *** France Coeff. -0.000-0.0047 0.0024-0.0026-0.0037-0.004 0.0076 *** *** ** ** *** *** *** Great Bran Coeff. -0.00-0.007-0.024 0.002-0.0032-0.002 0.02 *** *** *** Ireland Coeff. -0.063-0.009 0.0044 0.007-0.0023 0.0048-0.08 *** *** *** Italy Coeff. -0.0248 0.00-0.0078 0.00 0.0067-0.008 0.0006 *** *** *** *** *** *** Netherlands Coeff. -0.0078-0.0046-0.00-0.0004-0.0030 0.0070-0.0084 *** *** *** ** Portugal Coeff. -0.063 0.0022-0.009-0.0020 0.0028-0.0039 0.0038 *** * Sweden Coeff. -0.009-0.0049-0.0037-0.0077-0.0079 0.0049 0.024 *** *** ** *** *** Coeff. -0.006-0.0029 0.0244-0.080-0.0042 0.008-0.0063 Bulgara ** *** Czech Republc Coeff. -0.0209-0.0006 0.0063 0.0007-0.0073-0.0028 0.002 *** ** *** Estona Coeff. -0.0038 0.002 0.000 0.03-0.002-0.0039-0.06 * ** * * Hungary Coeff. 0.000-0.008 0.0086 0.027-0.008-0.0088-0.003 * Lhuana Coeff. 0.002-0.008-0.0024 0.0084 0.06-0.0020-0.04 *** Latva Coeff. 0.0063 0.0070 0.0039 0.0039-0.06 0.0062 0.007 ** Poland Coeff. 0.0-0.02 0.003-0.0043 0.008-0.0027-0.0038 *** *** *** *** ** ** Romana Coeff. -0.0389-0.027 0.0004-0.000 0.000-0.0047-0.0072 *** *** Slovena Coeff. -0.008-0.0094 0.0078-0.08 0.008-0.0043 0.0072 *** *** ** Slovaka Coeff. -0.0296-0.0003-0.06-0.0090-0.026 0.0040 0.0347 *** *** *** *** Note: ***, **, * denotes sgnfcance at the %, %, and 0% levels respectvely Source: own calculaton Table 4 provdes the parameter estmates on technologcal change and based technologcal change. The mpact s sgnfcant at 0 % level for almost all countres. It s sgnfcantly posve n most of the Old Member States (β t < 0) whle n the New Member States deteroraton of producton possbles domnates (β t > 0). A sgnfcant posve mpact of
techncal change occurred especally n those countres, whch are catchng up (the Czech Republc and Slovaka). The based technologcal change s pronounced for almost all analysed countres, except for Hungary and Romana. However, dstnct dfferences n the drecton of based technologcal change can be observed. The labour-savng technologcal change can be found n Bulgara, the Czech Republc, Denmark, Span, and France, and labour-usng technologcal change n Austra, Great Bran, Italy, and Slovaka. The based technologcal change s land-savng n Belgum, Estona, and Span, and land-usng n Bulgara, Germany, France, Poland, Sweden, and Slovena. The capal-usng based technologcal change s pronounced n most EU member countres. It s capal-savng only n Denmark, Span, Italy, Lhuana, and Poland. The estmates of the capal elastcy together wh the drecton of the based technologcal change suggest that mlk producers do not face capal market mperfectons, what allows them to upgrade ther producton technology and makes them more competve on the European Common market. Table : Techncal effcency mlk producton EU country σ λ Statstcal characterstcs of techncal effcency Mean Std.Dev Mn. Max st Decle 0th Decle Austra 0.49*** 0.8607*** 0.887 0.0709 0.9 0.9909 0.7828 0.93 Belgum 0.484***.8004*** 0.90 0.049 0.494 0.9882 0.847 0.926 Germany 0.4***.287*** 0.946 0.0367 0.244 0.9839 0.8700 0.923 Denmark 0.76***.929*** 0.9223 0.0420 0.63 0.9842 0.8696 0.9634 Span 0.2248***.368*** 0.8702 0.064 0.378 0.9760 0.799 0.929 Fnland 0.426***.420*** 0.8690 0.0723 0.60 0.9840 0.7630 0.942 France 0.33***.8293*** 0.929 0.046 0.449 0.9879 0.869 0.976 Great Bran 0.329***.33*** 0.9208 0.0346 0.642 0.9800 0.877 0.96 Ireland 0.26***.678*** 0.908 0.007 0.6 0.983 0.8490 0.960 Italy 0.2320*** 0.860*** 0.8900 0.038 0.66 0.9734 0.8 0.9228 Netherlands 0.00***.0222*** 0.9268 0.0367 0.7084 0.994 0.8780 0.9634 Portugal 0.748*** 0.8438*** 0.898 0.0394 0.6890 0.9746 0.843 0.938 Sweden 0.66***.790*** 0.8870 0.08 0.49 0.979 0.804 0.9467 Bulgara 0.368*** 2.998*** 0.7886 0.080 0.2662 0.9466 0.6343 0.90 Czech Republc 0.340***.0643*** 0.8904 0.03 0.88 0.9802 0.88 0.9482 Estona 0.2046***.772*** 0.8704 0.0640 0.427 0.9743 0.7824 0.9379 Hungary 0.20***.7663*** 0.860 0.063 0.93 0.966 0.7706 0.9343 Lhuana 0.796*** 0.7746** 0.8984 0.0379 0.724 0.9673 0.8487 0.938 Latva 0.2336*** 0.0000 - - - - - - Poland 0.2066***.476*** 0.8776 0.039 0.4274 0.974 0.8064 0.9333 Romana 0.2232*** 0.7307*** 0.9030 0.024 0.78 0.990 0.8732 0.9292 Slovena 0.2247*** 2.208*** 0.877 0.0768 0.3264 0.979 0.7602 0.9344 Slovaka 0.294*** 3.86*** 0.8304 0.0998 0.3086 0.976 0.6949 0.9364 Note: ***, **, * denotes sgnfcance at the %, %, and 0% levels respectvely Source: own calculaton.2.2 Techncal effcency and TFP Table provdes the estmates of parameter σ, λ, and the statstcal characterstcs of techncal effcency. The parameter σ provdes nformaton about the ont varaton of u and v. λ s
the relaton between the varance of u and v. Thus, the parameter ndcates the sgnfcance of TE n the resdual varaton. A value smaller than one suggests that varaton n u s less pronounced than varaton n the random component v. Snce λ s hghly sgnfcant n all EU member countres except for Latva and n maory of countres hgher than one, the estmates ndcate that effcency dfferences among mlk producers are mportant reasons for varaton n producton. The countres λ are on average lower n the Old than the New Member States. Ths suggests that frm performance n mlk producton n accesson countres s more heterogeneous. Ths concluson s renforced when lookng n more detal at the dstrbuton of neffcences. The st decle of farmers n most New Member States have effcency scores whch were lower than n the other EU member states. At the same tme the best decle of farmer n the Old Member States reaches hgher effcency values than n NMS, so more farms are lackng behnd and fewer farms are at the natonal fronters n the New Member States. Fnally, the Spearman s rank correlatons of techncal effcency (Table A n the appendx) ponts out that leapfroggng n techncal effcency appears to be a common phenomenon n maory of member countres. However, the Spearman s rank correlaton for TFP suggests that the order of mlk producers s stable over tme. That s, leapfroggng can be excluded as far as TFP development s consdered. Structural change seems to occur n such a way that the most successful producers strengthen ther posons. Producers wh poor performance wll not be able to catch up wh the developments of the sector leaders, and therefore are expected to fall more and more behnd. 6 CONCLUSIONS Ths secton summares the results and wll dscuss them n lght of the research questons asked n the ntroducton, n detal: How has regonal agrculture benefed from the adopton of nnovatons? Is there an ndcaton that regonal scarcy of resources was the source of techncal progress? Is there an ndcaton that techncal change (and other sources of adustment) have led to a convergence of the regons n terms of TFP? What are the causes of the regonal convergence or dvergence? How dd farm structure, market structure, and cred constrants affect the adopton of nnovatons? Gven the sgnfcant dualy of farm structures n some countres can the same patterns for large agrcultural companes and small farmers be observed or can we dentfy dosyncratc developments? The metafronter estmates reveal that there are consderable productvy dfferences n mlk producton across the EU at the NUTS 2 level. Productvy s the hghest n the Old Memeber States, especally n those regons located n the Northwest of the EU. The lowest productcy generaly could be observed n Eastern Europe. The same strucure as for TFP we found for TFP development. We found an above average ncrease n the Old and a below average development n the New Member States. Ths mples that posve economc effects expected from the economc ntegraton have not been realzed yet. Moreover, only few regons manly located n Slowaka, the Czech Republc, and Hungary could keep pace wh the developments n the Old Member States and thus are catchng up. However, most regons n Eastern Europe were fallng more and more behnd durng the frst years of EU acesson. The determnants for ths development were further nvestgated by analysng the sources of TFP development usng the natonal producton fronters. In partcular, we nvestgated how the scale, techncal change, and techncal effcency contrbuted to TFP levels and growth.
The analyses suggest that on average n all countres specalzed producers operate at almost constant returns to scale,.e. optmal farm szes from a statc pont of vew. Ths n turn mples that there s no pure techncal defnon of optmal farm sze, rather depends on many mutally related determnants. However, the results for techncal change suggest that farm szes are not optmal n many regons n Central and Eastern Europe from a dynamc prespectve. The mpact of techncal change n Eastern Europe on mlk producton was lower than n the rest of the EU. Moreover, there appears to be a strong correlaton between larger farms and the adopton of techncal changes, snce countres, whose mlk producton s large scaled, were able to generate above average effects of techncal change (Slowaka, the Czech Republc and Hungary). The correlaton of techncal change and sze also mples that techncal change was hghly ndvsble and could be adopted by those farms whch operate beyond a sze-specfc threshold wh mproved machnary, adequate anmal housng etc. or do not face other constrants (cred market mperfectons). Furthermore, the effects, dscussed so far, occur on the fronter of the country specfc producton possbles. The comparatve analyss suggest that despe was explcly accounted for farm heterogenety, mlk producton has a wder more left skewed dstrbuton for techncal effceny n Central Europe than n other regons of the EU. Ths mples that n the New compared to the Old Member States fewer farms could benef from the movement on the fronter. Moreover, snce techncal effency remans relatvely stable n the perod under examnaton there are no sgns that poor performng farms are catchng up to the best performng farms n the regons/countres. The results dscussed so far have mportant mplcatons for the effcency of the CAP. One of the declared targets of EU polcy nterventon s the mprovement of productvy and competveness n European agrculture. The fact that several regons n Eastern Europe are fallng behnd suggests a polcy falure at the EU level. The largest part of EU agrcultural funds s centrally planned at the EU level. However, the redstrbuton does not appear to contrbute posvely to some polcy obectves. From ths follows that s mportant to reallocate funds from pllar to pllar2, where the member states have larger decson-makng power regardng the dstrbuton of funds, as was foreseen by the last CAP reform. Our analyss recommends that the member states should foster the axs competveness n pllar 2 and should, n addon to nvestment ads to farmers, provde some means for related and supportng ndustres of the dary sector. Acknowledgements Ths paper was created whn the proect COMPETE Internatonal comparsons of product supply chans n the agro-food sectors: determnants of ther competveness and performance on EU and nternatonal markets. The proect has receved fundng from the European Unon s Seventh Framework Programme for research, technologcal development and demonstraton under grant agreement no 32029 (www.compete-proect.eu).
References. Álvarez, A., Aras, C., Greene, W. (2003): Fxed Management and tme nvarant techncal effcency n a random coeffcent model. Workng Paper, Department of Economcs, Stern School of Busness, New York Unversy, p. 0. 2. Álvarez, A., Aras, C., Greene, W. (2004): Accountng for unobservables n producton models: management and neffcency. Economc Workng Papers at Centro de Estudos Andaluces E2004/72, Centro de Estudos Andaluces, p. 8. 3. Bakucs, L. Z., Latruffe, L., Fertő, I., Fogaras, J., (200): The mpact of EU accesson on farms' techncal effcency n Hungary, Post-Communst Economes, Taylor & Francs Journals, vol. 22(2), pages 6-7. 4. Barro, R. J. and Sala-I-Martn, X. (99): Economc Growth, McGraw Hll, New York et al.. Caves, D.W., Chrstensen, L.R., Dewert, W.E. (982): Multlateral Comparsons of Output, Input and Productvy usng Superlatve Index Numbers. Economc Journal, 92, pages 73-86. 6. Cechura, L. (202): Techncal effcency and total factor productvy n Czech agrculture, Agrc. Econ. Czech, 8, 4, pages 47 6. 7. Čechura, L., Hockmann, H. (200): Sources of economcal growth n Czech food processng. Prague Economc papers, 200(2), pages 69-82. 8. Coell, T., Perelman, S. (996): Effcency Measurment, Multple-output Technologes and Dstance Fucntons: Wh Applcaton to European Ralways, CREPP 96/0, Unversé de Lége. p. 3. 9. Coell, T.J., Rao, P.D.S., O Donnell, C.J., Battese, G.E. (200): An Introducton to Effcency and Productvy Analyss, 2 nd edon, Sprnger Scence+Busness Mada, USA. 0. Csák, C. and Jámbor, A. (203): Impacts of the EU enlargements on the new member states agrculture, Acta Oeconomca et Informatca,, pages: 3 0. Csák, C. and Jámbor, A. (200): Fve Years of Accesson: Impacts on Agrculture n the NMS. EuroChoces, 9, (2), pages 0 7. 2. Dewert, W. E. (976): Exact and Superlatve Index Numbers. Journal of Econometrcs, 4, pages -4. 3. Jondrow, J. et al. (982): On the Estmaton of Techncal Ineffcency n the Stochastc Fronter Producton Functon Model, Journal of Econometrcs, 9, pages 233-238. 4. Kumbhakar, S.C., Lovell, C.A.K. (2000): Stochastc Fronter Analyss, Cambrdge: Unversy Press, p. 333.. Latruffe, L., Bravo-Ureta, B. E., Morera, V. H., Deseux, Y., Dupraz, P., (202b): Productvy and Subsdes n the European Unon: An Analyss for Dary Farms Usng Input Dstance Fronters, EAAE 202 Conference, Foz do Iguacu, Brazl. 6. Latruffe, L., Fogaras, J., Deseux, Y., (202a): Effcency, productvy and technology comparson for farms n Central and Western Europe: The case of feld crop and dary farmng n Hungary and France, Economc Systems, Elsever, vol. 36(2), pages 264-278.
7. Lovell, C.A.K., Rchardson, S., Travers, P., Wood, L.L. (994): Resources and Functonngs: A New Vew of Inequaly n Australa, Models and Measurement of Welfare and Inequaly, Berln, Sprnger-Verlag. 8. Shepard, R.W. (970): Theory of Cost and Producton Functons, Prnceton, Prnceton Unversy Press.
Appendx Table A - Sample descrptve statstcs mlk producton EU member y y2 y3 x x2 x3 x4 x country Mean Std.Dev Mean Std.Dev Mean Std.Dev Mean Std.Dev Mean Std.Dev Mean Std.Dev Mean Std.Dev Mean Std.Dev Cases Austra 36.92 2.9.46 7.83 3.9 9.0.80 0.7 36.6 28.68 20.7 0.68 9.89 8.00 26.39 4. 42 Belgum 00.32 4.6 23.67 24.9 29.2 4.37.80 0.62 60. 34.84 32.00 9.63 28.2 9.9.86 3. 230 Bulgara.39 0.7 9.64 26. 8. 202.28 0.6 30.9 268.08 49.4 6.47 4.27 3.43 8.6 6.96 32.07 2430 Czech Republc 2.62 463. 86.00 23.80 26.47 68.70 40.49 3.90 099.36 9.9 77.4 73. 26.04 234.04 736.44 726.2 2600 Germany 4.8 279.48 2.36 38.22 28.34 07.03 2.72 6.0 04.28 228.6 44.92 79.73 4. 09.3 0.09 223.73 8676 Denmark 393.23 20.8 4.9 70. 03.04 84.24 2.68.49 44.63 8. 36. 9.34 76.03 32.78 84.70 7.4 66 Estona 36.30 264.80 7.4 44.49 68.97 38.00 7.2 2.48 33.82 09.9 37. 8.94 72.4 42. 8.3 67.7 27 Span 2.0 2.24 7.60 20. 2.44 8.8.90 0.94 30.97 38.83 0.76 6.04 7.34 6.42 28.6 34.3 6093 Fnland 02. 73.0 9.07 2.62 7.92.43 2.33 0.92 64.8 39.9 46. 40.0 29.9 23.2 6.48 4.3 280 France 0.76 62.0 26.86 2.62 37.97 3.03 2.0.00 7.2 79.8 2. 34.3 27.49 20.43 83.79.4 98 Great Bran 208.4 63.90 34.27 3.23 27. 0.29 2.63.43 6.9 97.3 43.36 36.24 80.4 67.9 0.82 82.78 3276 Greece 98.8 08.4 2.6 36.3.92 2.62 2.34.9 8.33 3.0 8.90 6.7 60.62 88.37.08 2.67 86 Hungary 374.37 792.68 67.2 230.6 270.90 3.88 22.43 44.4 66.9 78.69 9.2 99.33 22.6 380.68 399.94 30.36 688 Ireland 9.2 6.77 28.84 22.3 8.4 0.97.70 0.7 6.4 32.9 22.29 4.29 29.3 22.87 43.9 26.40 972 Italy 42.7 26.69 20.0 48.4 42.2 9.7 2. 2.24 43.24 74.4 9.84 28.7 84.84 2.8 20.69.0 6748 Lhuana 6.38 68.7 0.8 36.9 49.68 6.29.7 4.63 92.2 48.73 6.04 48. 2.32 72.08 49.4 66.9 79 Latva 9.48 22.73 4.37 0. 48.32 7.29 7. 7.03 237.7 47.79 8.83 43.4 32.7 64.96 63.92 70.46 69 Netherlands 2.79 4.60 20.04 2.97 7.68 76.9.87. 6.7 36.2 3.6 40.96 40.3 30.02 02.94 7.74 27 Poland 29.9 69.3 6.89 8.36.84 77.08 2.48.37 4.38 37.76 8.32 2.9 8.82 22.2 23.94 0.63 30 Portugal.3 46.93.7.67 8.6.30.80 0.74 23.7 20.83 6.4 6.07 23. 24.9 7.39 7.04 879 Romana 9.49 8.8.22 2. 3.04.6 3.84 2.6 78.98 38.86 8.33.79 8. 30.44 27.0 6.84 223 Sweden 30.86 9.98.3 2.93 34.39 44.68 2.8.7 06.47 3.89 47.7 7.3 64. 7.07 88.89 02.64 2388 Slovena 39.22 38.7 7.6 9.30 3.93 8.20 2.43 0.93 24.94 8.60 3.78 0.67 8.87 9.7 8.63 4.88 207 Slovaka 432.4 434.90 48.84 88.88 62.98 622.6 4.88 40.76 83.84 048.62 387.0 380.4 263.94 280.64 784.6 736. 447 Note: y mlk producton (ths. EUR), y2 other anmal producton (ths. EUR), y3 plant producton (ths. EUR), x labour (AWU), x2 land (ha), x3 capal (ths. EUR), x4 specfc materal (ths. EUR) and x other materal (ths. EUR). Source: FADN and own calculatons
Table A2: Spearman s rank correlaton coeffcents of techncal effcency n mlk producton EU country Spearman s rank correlaton coeffcents of techncal effcency 200/2004 2006/200 2007/2006 2008/2007 2009/2008 200/2009 20/200 Austra NA 0.84 0.8299 0.8063 0.8092 0.8270 0.8382 Belgum 0.2746 0.30 0.0804-0.72 0.097 0.00 0.300 Bulgara NA NA NA -0.30-0.043-0.0767 0.380 Czech Republc 0.677 0.649 0.6229 0.600 0.404 0.79 0.622 Germany 0.28 0.94 0.0870-0.0299-0.067 0.483 0.726 Denmark 0.39-0.244 0.4 0.7 0.076 0.690 0.2783 Estona -0.43 0.0824-0.0393-0.004 0.37 0.48 0.092 Span 0.228 0.07 0.0670 0.374 0.03 0.0968 0.784 Fnland 0.6492 0.894 0.642 0.6203 0.64 0.337 0.08 France 0.237 0.300 0.0003-0.0476-0.046-0.0047 0.827 Great Bran 0.0973 0.078 0.0027-0.28 0.3-0.0343 0.244 Greece - - - - - - - Hungary -0.440-0.024 0.0062-0.070-0.277-0.04 0.3697 Ireland 0.338 0.4 0.272 0.0746 0.2369 0.3877 0.2382 Italy 0.0877 0.079 0.229-0.2069 0.0093-0.034 0.29 Lhuana 0.206 0.4243 0.3003 0.4834 0.2822 0.280 0.673 Latva 0.9660 0.996 0.9949 0.9846 0.976 0.9882 0.9888 Netherlands 0.6437 0.643 0.838 0.22 0.403 0.44 0.90 Poland 0.0 0.086 0.0363-0.087-0.003 0.042 0.029 Portugal 0.463 0.834 0.446 0.4362 0.3642 0.27 0.8 Romana NA NA NA -0.976-0.237-0.270-0.2272 Sweden 0.460 0.400 0.243 0.086 0.22 0.280 0.0984 Slovena 0.0278-0.029-0.28 0.0887-0.04 0.2693-0.0430 Slovaka -0.027-0.0806 0.907-0.043-0.66 0.0889 0.2747 Source: own calculaton
Table A3: Spearman s rank correlaton coeffcents of TFP n mlk producton EU country Spearman s rank correlaton coeffcents of TFP 200/2004 2006/200 2007/2006 2008/2007 2009/2008 200/2009 20/200 Austra NA 0.8863 0.909 0.8909 0.9069 0.8892 0.940 Belgum 0.9 0.9472 0.9480 0.94 0.937 0.948 0.930 Bulgara NA NA NA 0.6248 0.666 0.747 0.6368 Czech Republc 0.8700 0.8960 0.8827 0.894 0.8423 0.860 0.8892 Germany 0.979 0.9673 0.9696 0.98 0.963 0.977 0.998 Denmark 0.8437 0.763 0.9090 0.9046 0.860 0.8499 0.937 Estona 0.806 0.8798 0.8684 0.8824 0.934 0.9273 0.946 Span 0.8237 0.823 0.8246 0.8448 0.800 0.79 0.84 Fnland 0.8894 0.899 0.98 0.893 0.9000 0.908 0.8676 France 0.94 0.92 0.940 0.9279 0.934 0.9302 0.9482 Great Bran 0.980 0.974 0.96 0.92 0.9643 0.9349 0.924 Greece - - - - - - - Hungary 0.707 0.8624 0.7602 0.728 0.809 0.8749 0.9072 Ireland 0.960 0.9 0.9394 0.944 0.963 0.960 0.9473 Italy 0.9767 0.9799 0.9804 0.989 0.976 0.9728 0.9776 Lhuana 0.8998 0.999 0.94 0.928 0.898 0.8934 0.937 Latva 0.7383 0.804 0.840 0.8373 0.7692 0.870 0.897 Netherlands 0.90 0.9390 0.927 0.96 0.9444 0.994 0.9392 Poland 0.943 0.938 0.9382 0.9280 0.928 0.9236 0.9346 Portugal 0.976 0.8927 0.8728 0.8349 0.833 0.862 0.8897 Romana NA NA NA 0.933 0.967 0.968 0.9694 Sweden 0.939 0.939 0.8799 0.882 0.8904 0.9043 0.8932 Slovena 0.88 0.8793 0.80 0.877 0.8679 0.9004 0.8928 Slovaka 0.888 0.864 0.97 0.8786 0.7432 0.8372 0.9260 Source: own calculaton