The Factors of Growth of Small Family Businesses. A Robust Estimation of the Behavioural Consistency in Panel Data Models

85 Volume 25 | Numbe 01 | 2016 PRAGUE ECONOMIC PAPERS THE FACTORS OF GROWTH OF SMALL FAMILY BUSINESSES – AROBUST ESTIMATION OF THE BEHAVIOURAL CONSISTENCY IN PANEL DATA MODELS Vladimí Benáček, E a1Michalíko á* Abs ac The pape quan i ies he ole o ac o s associa ed wi h he g ow h (o decline) o mic o and small businesses in Eu opean economies. The g ow h is ela ed o he le els o employmen and alue added in en e p ises, as well as, en ins i u ional a iables. We es he da a o consis ency o beha iou al pa e ns in a ious coun ies and g adually emo e ou lying obse a ions ha can lead o e oneous conclusions when using he classic es ima o s; his is aqui e unique app oach in panel da a analysis. In he i s pa o his pape we ou line ahighly obus me hod o es ima ion based on ixed e ec s and leas immed squa es (LTS). In i s second pa we apply his me hod on he panel da a o 28 coun ies in 2002–2008 es ing o he hypo hesis ha mic o and small businesses in Eu ope use di e en s a egies o hei g ow h. We un ase ies o econome ic es s whe e we eg ess employmen and o al ne p oduc ion in mic o and small businesses on h ee economic ac o s: g oss capi al e u ns, labou cos gaps in small ela i e o la ge en e p ises and GDP pe capi a. In addi ion, we es he ole o 10 ins i u ional ac o s in he g ow h o amily businesses. Keywo ds: amily business, obus es ima o , LTS, ixed e ec s JEL Classi ica ion: C01, C23, C51, C82, F21, F40 1. In oduc ion As a consequence o he wo ldwide inancial and economic c isis, he e is a enewed ising in e es in he pe o mance o small and amily businesses which o many esea che s and poli icians p esen a c ucial ehicle ha d i es upwa d mo emen s in bo h employmen and compe i ion le els h oughou he wo ld’s economies. This esea ch is a ollow-up o he analysis o Benáček, Michalíko á (2010), in which we assessed he ole o economic and ins i u ional ac o s on he ise and decline o amily businesses and applied hem o an analysis o da a conce ning mic o- and small businesses in 28 coun ies o Eu ope by means o panel da a o he yea s 2002–2008. We disco e ed ha his speci ic da a on mic o- and small businesses in so many di e en coun ies did no demons a e a homogenous pa e n o beha iou in i ms ha di e no only in sizes, bu also as ega ds he a ious ins i u ional se ups ha also change o e ime. In his esea ch *1 Vladimí Benáček, Cha les Uni e si y in P ague, FSV-Ins i u e o Economic S udies, P ague (benacek @ s .cuni.cz); E a Michalíko á, Facul y o Business and Managemen , B no Uni e si y o Technology, B no (michaliko a@ olny.cz); The au ho s a e g a e ul o he i nancial suppo o he G an Agency o he Czech Republic No. P402/12/0982 and Resea ch P ojec o VUT in B no No. FP-S-15-2787 and Resea ch P ojec o Ins i u ional De eloping P ojec o FP VUT in B no, No. IS 2014-04. DOI: 10.18267/j.pep.538 86 Volume 25 | Numbe 01 | 2016PRAGUE ECONOMIC PAPERS we ha e es ed he po en ial o such a he e ogenei y in he beha iou o small amily businesses in a ious coun ies ha may be demons a ed by sepa a ing he o iginal panel da a in o wo subpopula ions. Hence, we ha e concen a ed ou analysis on he echniques o obus es ima ion. In his pape , we apply a obus e sion o he classic wi hin-g oup es ima o on da a o wo g oups o amily businesses. In Sec ion 2, we desc ibe and apply a obus e sion o he classic wi hin-g oup es ima o on da a o he wo g oups o amily businesses. In Sec ion 3, we desc ibe he ole o amily businesses a he p esen ime. In Sec ion 4, we apply a obus e sion o he wi hin-g oup es ima o on economic da a ela ing o amily businesses and we examine how employmen and ne p oduc ion in amily businesses depend on wo ela i e indica o s ep esen ing bene i s and cos s in small (o mic o) and la ge en e p ises. Addi ional explana o y a iables include he GDP pe capi a and en ins i u ional a iables. Sec ion 5 summa izes ou indings. 2. Robus Es ima o s and Robus Es ima ion o Panel Da a Models Classic me hods o es ima ion ely hea ily on assump ions, which a e o en no me in p ac ice. Un o una ely, some alues o a iables o en happen o all a away om o he obse a ions in he sample. These disc epan alues migh be he esul o epo ing e o s, di e en me hodologies used by he epo e s o idiosync a icies in he beha iou o obse ed agen s. The isk o incidence o all hese dis u bances is qui e high in panel da a in which he ime and he geog aphic discon inui y may lead o da a inconsis ency. Obse a ions o he la e so a e e e ed o as ou lie s. They a e o en de i ed by e o s and omissions in he da a collec ion p ocess. Howe e , ou lie s can also be gene a ed when epo e s mix up wo o mo e subpopula ions o da a ha ep esen agen s, whose beha iou is mu ually inconsis en . An example o his occu ence is he case when he analys s p esume ha mic o businesses (such as sel -employed pe sons) and businesses up o 50 employees ollow iden ical s a egies o hei g ow h in all s udied coun ies, i espec i e o he a ying ins i u ional a angemen s. These kinds o inconsis encies which occu when ca ying ou obse a ions a e ou main conce n. Small amily businesses a e subjec o speci ic ci cums ances ha inc ease he unce ain y and inconsis ency o hei epo ed da a. Thei accoun ancy need no always be done by p o essionals and hus mo e open o e o s and omissions. Thei ue p oduc ion, employmen and cos s can be igged due o much easie ax e asion. Repo ing o s a is ical o ices is i egula , elying on andom (o en non- ep esen a i e) samples and he eedback on i s accu acy is limi ed as well, a ying by coun ies. Thus, a obus echnique o hei es ima ion is a necessa y and adequa e app oach in o de o a oid he aps and pi alls ha esul om using biased da a. Since obus es ima ion has no been he s anda d echnique o analysis in his kind o panel da a, we will i s desc ibe ou app oach o da a p ocessing in which he cen al issue ocuses on ou lie s. The e m obus es ima o deno es an es ima o ha is no s ongly a ec ed by ou lie s. I means ha he main aim is o i a eg ession o he dominan in e ela ions in he da a and hen disco e he ou lie s o u u e ea men . As a measu e o obus ness we should conside he exis ence o he b eakdown poin o es ima o s. Gene ally speaking, he b eakdown poin o an es ima o is de ined as he smalles ac ion o an ou lying obse a ion ha causes a b eakdown o he es ima o (Rousseeuw, Le oy, 1987). 87 Volume 25 | Numbe 01 | 2016 PRAGUE ECONOMIC PAPERS Many s a is icians belie e ha ou lie s can be iden i ied simply by eye h ough he use o g aphs. Howe e , i is di icul o diagnose ou lie s by eye, especially in he case o panel da a because la ge panels o coun ies, companies, o o he agen s may con ain a ypical obse a ions o g oss e o s subjec o a mul i ude o exogenous a iables. Un o una ely, econome ics is limi ed o a scan amoun o li e a u e desc ibing obus me hods o panel da a1. We will y o ind a obus al e na i e o he wi hin-g oup es ima o 2 which can be a ec ed by he p esence o ou lying obse a ions. The b eakdown poin is he measu e o obus ness and he leas immed squa es is he es ima o wi h high b eakdown poin . We will ocus on he second applicable 50% b eakdown poin es ima o – he leas immed squa es – LTS (Rousseeuw, 1983). We will desc ibe a high b eakdown poin es ima o o he ixed e ec s panel da a model based on LTS as an es ima ion p ocedu e, which is less sensi i e o he p esence o abe an obse a ions. We conside he ollowing o m o he ixed e ec s linea panel da a model: yi = αi + xi ΄ β + εi , 1, , ; 1, ,iN T (1) whe e i deno es he c oss-sec ion dimension (numbe o coun ies) and deno es he ime- se ies dimension (numbe o yea s). xi is a column ec o o explana o y a iables wi h dimension K x 1 while β is a K x 1 ec o o eg ession pa ame e s. αi is he unobse able ime-in a ian indi idual ixed e ec s and εi is he e o e ms o dis u bance e ms, unco ela ed h ough ime and h ough c oss-sec ions. The classical wi hin-g oup es ima o s o ixed e ec panel da a models is based on cen e ing wi hin e e y ime-se ies: 1 1 ˆ T i i i yy y T   1 1 ˆ T i i i xx x T  , and hen he basic o m o he ixed e ec s panel da a models, desc ibed in (1), can be exp essed as: ˆ ˆˆ , i i i yx    1, , ; 1, ,iN T (2) whe e 1 1 ˆ T i i i T     , and he ixed e ec s αi ha e disappea ed om he model (2) by he cen e ing ope a ion. Then we can eg ess ˆi yon ˆi x by OLS and we will ge wi hin- g oup es ima o deno ed by ˆWG  . O cou se, ixed e ec s pa ame e s can be es ima ed as well (Bal agi, 1998). Cen e ing has a c ucial ad an age because i eno mously educes he numbe o pa ame e s. In o de o ge a obus e sion o his es ima o we ha e o cen e he ime se ies obus ly and hen a obus eg ession will be applied o he cen e ed da a. The ime-se ies mus be cen e ed by emo ing he median ins ead o mean because he mean is la gely dis o ed by ou lie s since he median is known o be min-max obus (Hube , 1981). We will ge : () i i i yymedy , () () () jj i i i xx medx  whe e 1, , ; 1, ,iN T and j = 1, ..., K. x(j)i deno es he j- h explana o y a iable measu ed a ime in he i- h ime-se ies. The numbe o pa ame e s is educed as in he 1 Kalina (2012) desc ibed obus e sion o ins umen al a iables wi h applica ion o economic da a. 2 Since ou panel con ain all coun ies o in e es , he i xed e ec s model is mo e app op ia e han a andom e ec s models o ou da ase . 88 Volume 25 | Numbe 01 | 2016PRAGUE ECONOMIC PAPERS case o de-meaning. This implies ha compu a ion ime o obus eg ession algo i hm emains easible (B ama i and C oux, 2004). The e o e, we can un a obus es ima o (and eg ess i y on i x  o iden i y he ou lie s). Fo his pu pose we will apply he LTS es ima o on cen e ed da a. LTS es ima o is de ined as ˆLTS  which minimizes he sum o he smalles h squa ed esiduals: 2 () 1 ˆa g min ( ) h LTS k k i k yx       whe e 22 2 2 (1) (2) ( ) ( ) ()() () () kk kk kk k kk NT yx yx yx yx             a e o de ed squa ed esiduals (Rousseeuw, 1983). The alue 1 ≤ h ≤ NT is a imming alue. As men ioned be o e, his es ima o has a b eakdown poin a aining 50%. A de aul choice can be h=[3NT/4] o h=[4NT/5], making i possible o cope wi h up o 25% o ou lie s (o 20%, espec i ely) o we can selec h su icien ly small o each an accep able coe icien o de e mina ion o he model. The ad an ages o he LTS es ima o a e i s p ope ies: he LTS es ima o in i s basic e sion is eg ession, scale, and a ine equi a ian (Rousseeuw, 1983). Howe e , due o he nonlinea i y o he cen e ing ans o ma ion by he median βLTS is only scale equi a ian (B ama i and C oux, 2004). We can use his algo i hm di ec ly: cen e ing he da a by median, using leas immed squa es and disco e ing he ou lie s. Then we can wo k wi h he es o he da a and eg ess dependen a iable on o he eg esso s (Benáček, Víšek, 2000; Ve a di, Wagne , 2010). Howe e , i can also be employed in a di e en way by using ou lie s only as a diagnos ic ool o ecognize “suspicious” beha iou o an agen . In o he wo ds, we can d op ou whole g oups o agen s ( i ms, coun ies, e c.) whe e mos o he obse a ions a e ea ma ked as ou lie s and wo k wi h he es o obse a ions (Michalíko á, Galeo i, 2010). In his pape we will iden i y he ou lie s in a cen e ed model, sepa a e hem, and hen use he LTS on he es o da a. This echnique makes i possible o ecognize ou lie s which a e no able o be de ec ed by eye o by means o adi ional eg ession diagnos ics. Once, we ha e sepa a ed he obse a ions (conside ed o be ou lie s), we can moni o i his subpopula ion o da a is subjec o ce ain sys emic egula i y. We may, o example, be p ima ily in e es ed i a g oup o coun ies beha e in an idiosync a ic way. Secondly, we may be wa ching i he emo al o ou lie s b ings some imp o emen in he es ima ed eg ession model. Fo example, we may moni o he dec ease in he esidual sum o squa es o he inc ease in he coe icien o de e mina ion. Fu he mo e we may moni o he s abili y o es ima ed eg ession coe icien s in he case o inc easing h. Las , bu no leas , we wonde i p- alues o es ima ed eg esso s a e imp o ing as he ou lie s a e d opped ou om he model. 3. The Fac o s o G ow h o Family Businesses 3.1 Family business and small and medium-sized en e p ises In he ea ly 1990s, amily-led en e p ises we e supposed o ge a new boos as p o-ma ke o ces iumphed. This was an e o in judgemen . Au hen ic small-scale amily busi- nesses we e o en squeezed ou o he ma ke a ena by he apid de elopmen by su i - ing, o me s a e-owned en e p ises which we e con e ed o co po a ions owned o mally by housands o mino s ock-owne s and a na ow class o inside s wi h dominan s akes 89 Volume 25 | Numbe 01 | 2016 PRAGUE ECONOMIC PAPERS (Benáček, 2006). The pa alleled opening-up o egional and in e na ional make s due o globalisa ion o e ed new windows o oppo uni y o la ge en e p ises domina ed by manage s. In he la e 1990s, he loodga es o expansiona y mone a y policies opened up and go e nmen deb g ew. Simul aneously, en ep eneu ship in he majo i y o ad anced capi alis coun ies led by la ge inancial ins i u ions u ned ei he o asse s, whose p ices could ise in a icious ci cle o supply and demand, o o an alignmen wi h public admini- s a o s, in which achie ing social e iciency was an objec i e ha could be sac i iced, which was a mo e simila o he de elopmen in pos -Communis coun ies. Bo h bubbles inally bu s , and subsequen ly d o e he economies in bo h de eloped and pos -Commu- nis coun ies in o a las ing ecession (Lawson, Zimko á, 2009). Rising axes, as a conse- quence o in e en ions, disc imina ed agains small amily businesses. The expec a ion is ha a u na ound om he p esen ecession should come om an inc ease in domes ic agg ega e spending and employmen in small and medium-sized en e p ises (SME) domi- na ed by amily businesses, which in almos e e y coun y ha e been he main sou ce o employmen and job c ea ion, bu no he engine o spending dynamics. The main objec- i e o his pape is o add ess he ques ion: which economic and ins i u ional ac o s a e associa ed wi h he de elopmen and g ow h o amily businesses? A i m is conside ed o be a amily business i a membe o one o mo e amilies is i s con olling owne ; hus, implying a manage ial commi men owa d he businesses’ o e all pe o mance. I was gene ally belie ed ha e en hough SME could p o ide he majo i y o jobs, hei ole in he p og ess o economies was only o subsidia y impo ance (Schumpe e , 1942). Fo a long ime, he dominan p esump ion was ha employmen in small businesses was nega i ely ela ed o GDP pe capi a, causing a bias owa d la ge en e p ises (Lucas, 1978; Acs e al., 1994). We hink ha hese p esump ions should be e-conside ed because hey a e no consis en wi h empi ical obse a ions. Acs, Aud e sch (1988) eached he conclusion ha inno a ions we e nega i ely ela ed o concen a ion and ha inno a ion inc eased wi h he esea ch and de elopmen (R&D) expendi u es a a less han p opo ional a e. Symeonidis (1996) concluded his ex ensi e su ey o empi ical li e a u e on he alleged ad an ages o la ge o e small i ms wi h he inding ha “li e a u e su ey sugges s ha he e seems o be li le empi ical suppo o he iew ha la ge i m size o high concen a ion a e ac o s gene ally conduci e o a highe le el o inno a i e ac i i y” (p. 33). The ou b eak o he wo ld inancial and economic c isis in 2008 b ough a new wa e o a en ion o ac s e u ing he alidi y o he so-called Schumpe e ian hypo hesis ega ding he demise o small en ep eneu ship (Schumpe e , 1942: pp.134–143). Mic o and small businesses (i.e. MB and SB) co e 98.7% o all EU en e p ises. In addi ion, app oxima ely 50% o MB in he EU a e o med by he sel -employed. The e o e, in he es o his s udy we shall use mic o and small businesses as a p oxy ca ego y o amily businesses. We will hus dis inguish be ween wo ypes o amily businesses (deno ed FB): hose anging in size om sel -employed indi iduals o en e p ises wi h 10 employees (i.e. MB) and en e p ises wi h 10 o 50 employees (i.e. SB). I is necessa y o no e ha we will wo k wi hin he non- inancial p i a e sec o only, hus we will analyse an incomple e pa o he na ional economies o Eu ope. The wo ld economic c isis slashed he EU expo s o goods and se ices om p e ious annual g ow h a es close o 6% o a me e 1.6% in 2008 and a decline o 14% in 2009 (Eu os a , 2010). This se e ely damaged he us in he g ow h leade ship o la ge businesses. Go e nmen de ici spending compensa ed pa ially o he missing expo s, 90 Volume 25 | Numbe 01 | 2016PRAGUE ECONOMIC PAPERS bu he e was no o he segmen o economy capable o illing he looming gap in bo h agg ega e demand and e iciency. Wi h he excep ion o Poland, he p i a e sec o was no able o ac as an agen o sus ained g ow h. Ne e heless, SME ha e sa ed many Eu opean economies om d as ic alls in employmen . The expec ed mild economic eco e y o he GDP g ow h o 1.6% in he EU-27 in 2011, d i en mainly by expo s, will equi e ha a complemen a y esou ce is s a ed up o subs i u e o he ading and ine icien go e nmen de ici spending. We p edic such a esou ce o exis due o he e i al o au hen ic en ep eneu ship ha used o be ep esen ed by FB. Tha e i al should ac ually be aced back o 1948–1965, when in e nally d i en de elopmen in FB was s ill dominan and had no ye been c owded ou by globalized businesses. 3.2 Fac o s a ou able o ad e se o amily business de elopmen We will y o es a hypo hesis ha he de elopmen o SB could ha e deepe mic oeco- nomic ounda ions. We aced hem o wage and p o i s uc u es, and o he compe i ion wi h la ge en e p ises which p essed “ inge compe i o s” o espond wi h s a egies idio- sync a ic o smallness ha allowed hem o wi hs and he compe i i e ace. The ollowing heo e ical assump ions will be used as guidelines o hypo heses in ou empi ical es s: a) The objec i e unc ion o en ep eneu s is p o i maximiza ion. The maximiza ion o g oss capi al e u ns pe alue added (KR/VA), whe e capi al K is de ined by educing o al labou compensa ion (W) om he ne income o en e p ises (VA)3, is s ill a plausible c i e ion because i ep esen s a social e iciency o capi al alloca ed among businesses o a ious scales. We could se up a hypo hesis ha coun ies wi h highe KR/VA in any g oup o FB could also see he s onge de elopmen o FB. I he space o K=VA-W inc eases (e.g. as a esul o inno a ion o lowe ansac ion cos s), i will induce he en ep eneu s o expand hei employmen in o de o bols e he sales and ne ou pu . This will esul in an inc ease o labou income W and a aise in he wage a es pe labou W/L. Ne e heless, a e y high KR/VA may also imply a sho age o capi al (unde capi aliza ion and/o oo expensi e capi al). Because o his high capi al e u ns could ac as an impedimen o FB g ow h, i.e. KR/VA could be nega i ely ela ed o g ow h in employmen . b) FB de elopmen is no au onomous in isola ion wi hin hei own SME ca ego ies because an FB’s ela i e pe o mance is-a- is la ge businesses (LB) ma e s signi ican ly. Small FB compe e wi h LB o limi ed na ionally a ailable economic esou ces. We will es whe he (lowe ) wages pe wo ke in FB ela ed o (highe ) wages pe wo ke in LB a e associa ed wi h highe g ow h in FB. Thus, we can aise a hypo hesis o empi ical es ing o FB de elopmen assuming ha LFB is a nega i e unc ion o ela i e wage a es (WFB/LFB)/(WLB/LLB). I is an ou come o an assump ion ha LB and FB di e in hei mic o- echnologies, which a e d i en by di e en ela i e ac o p ices, i.e. di e en a ios o wage a es pe capi al en al a es. Thus he isoquan s in FB end o be capi al-sa ing; while in LB hey a e labou -sa ing, which in he end makes he o me ne job c ea o s. c) Ano he hypo hesis abou he de e mining ac o s o g ow h in FB ha we will es conce ns he deg ee o gene al economic de elopmen ep esen ed by GDP pe capi a. We canno hen e i y whe he ising p ospe i y is a ac o ha enhances o es ains he de elopmen o FB. 3 Ne income (i.e. he alue added) o en e p ises is de i ned as di e ence be ween sales S and ma e ial inpu s M. 91 Volume 25 | Numbe 01 | 2016 PRAGUE ECONOMIC PAPERS d) Con empo a y economic hinking s esses he impo ance o ins i u ions, as adminis a i e bodies ha de ine he “ ules o he game” o incen i es whose pu pose is o educe unce ain ies and ansac ion cos s in business in e ac ion (S igli z, 1998). Na ional ins i u ions a e impo an ac o s ha may ha e bo h posi i e and nega i e impac s on businesses o di e en sizes. Thus, h ee economic indica o s ela ed o he in e nal a es o g oss capi al e u ns (KRFB/VAFB), ela i e wages a es (WFB/LFB)/(WLB/LLB), and GDP pe capi a, plus en ins i u ional indica o s ha e been selec ed as causal ac o s ela ed o he g ow h o FB, i.e. he MB and SB. 4. Es ima ion o Reg ession Models 4.1 The e iew o a iables and models o empi ical es ing In his chap e we will empi ically es he ex en o which he g ow h in FB in 28 Eu o- pean coun ies was in luenced du ing 2002–2008 by he h ee economic ac o s desc ibed abo e and by he isks o bene i s associa ed wi h en coun y-speci ic and ime-speci ic socio-poli ical ins i u ions. Small Business Ac Fac shee s (Eu os a and DG En e p ises and Indus y); GDP s a is ics o he Wo ld Bank; and Da abase on he Economic F ee- doms (The He i age Founda ion) we e used as he sou ces o he da a. The obus e sion o he ixed e ec panel da a model will be used o he es ima ion o coe icien s. Dependen a iables LFB i : Employmen in FB (i.e. MB o SB) quan i ied by he numbe o wo ke s in coun y i and yea . VAFB i : The alue o ne ou pu (i.e. he alue added) in MB o SB in coun y i and yea . Economic explana o y a iables KRFB i /VAFB i : G oss capi al e u ns in analysed businesses pe alue added LCFB i /LCLB i : Rela i e a es o ull labou cos s (LC=W/L), i.e. o al labou compensa ion pe wo ke in FB di ided by simila compensa ion in LB GDPi /PCi : GDP pe capi a in pu chasing powe pa i y. Ins i u ional explana o y a iables Reguli : Business eedom ( egula ion) index T adei : T ade eedom ( ade ba ie s) index Mone i : Mone a y eedom (in la ion and p ice con ol) index Go e ni : F eedom om go e nmen (public spending) index Fiscali : Fiscal eedom ( axa ion) index P opRi : P ope y igh s index In es i : In es men eedom (capi al con ols) index Financi : Financial eedom (p i a e banking secu i y) index Co up i : F eedom om co up ion (pe cep ion) index Labou i : Labou eedom index 92 Volume 25 | Numbe 01 | 2016PRAGUE ECONOMIC PAPERS N.B.: Ins i u ional a iables a e he p oxies o economic “ eedoms” anging in hei alues <0, 100>. The highe is he pe cen age index, he mo e libe al and p o-ma ke he local ins i u ional a angemen . The selec ion o 28 coun ies o Eu ope is highly ep esen a i e, co e ing nea ly all o he EU and po en ial accession coun ies (see Table 1). Table 1 | Lis o Coun ies Included in he Analysis ALL Ad anced Eu ope (14) + Eme ging Eu ope (14) Ad anced Eu ope (14) Aus ia, Denma k, Finland, F ance, Ge many, G eece, I eland, I aly, he Ne he lands, No way, Po ugal, Spain, Sweden, he Uni ed Kingdom Eme ging Eu ope (14) Albania, Bulga ia, C oa ia, Cyp us, he Czech Republic, Es onia, Hunga y, La ia, Li huania, Mal a, Poland, Romania, Slo akia, Slo enia Sou ce: Own classi ica ion The i s wo explana o y a iables a e ele an o decision-making in en e p ises. G oss capi al e u ns a e closely ela ed o p o i s and p o i s o m he basis o in es men s in o physical capi al and R&D. High p o i s also mo i a e FB owne s o inc ease he scope o hei p oduc ion and ake ad an age o gains due o economies o scale, which should imply g ow h. The easons o ha ing a high sha e o g oss capi al e u ns on he alue added can be: a) Inc easing labou p oduc i i y wi hou compensa ing wo ke s a a p opo ionally highe wage a e would imply high p o i s; b) Dec easing he ma ginal p oduc o labou by o e s a ing ha is e lec ed in disp opo ionally lowe a e age wages in he en e p ise would imply a high cos o capi al ha bu dens he i m; c) Hi ing and paying labou ou side o icial con ac s, which slashes o al labou cos s. Fo di e en easons ha d i e KR/VA upwa d, we canno be su e whe he his a iable is ela ed o FB g ow h nega i ely o posi i ely. The second a iable LCFB/LCLB es s he ele ance o low ( epo ed) wages and o he gap in FB wage a es ailing behind LB. We can expec o obse e a wide ange o c oss-coun y di e ences in his ela ionship. Wha ma e s is whe he a highe labou cos gap in FB is a d i ing ac o o a es ain upon FB g ow h. Once again we canno be su e a p io i abou he na u e o i s sign. The hi d a iable poin s o a gene al end o de elopmen . Ou only mac oeconomic indica o is subs an ia ed on he heo e ical g ounds ha we e elabo a ed by Lucas (1978), and ollowed by Acs e al. (1994) and To ini (2005). We should expec i s sign o be posi i e. The cen al idea behind he choice o ins i u ional a iables is ha ins i u ions a e man-concei ed ac o s ha ha e a wo-p onged impac on businesses: as public goods o as public bads. The depa u e om la gely mac oeconomic o mic oeconomic explana o y a iables ep esen ing incen i es o policy ins umen s, has ecen ly become a s anda d ool o econome ic analysis (Blau 1987; Robson, W en 1999; Da is, Hen ekson 1999). All ou ins i u ional a iables a e based on hei pe cei ed quali ies ha allow ma ke and en ep eneu ial eedom, once he coe icien is posi i e. E en hough we can assume ha mo e libe al economies g ow as e , some s udies o SME ha e e ealed ha e y small businesses a e no posi i ely ela ed o all indica o s o a ee ma ke economy (To ini, 2005). The es consis o ou models ela ed o mic o and small en e p ises, whose spe-ci ica ions a e as ollows: 93 Volume 25 | Numbe 01 | 2016 PRAGUE ECONOMIC PAPERS Lmic o i = α1 KR/VAmic o i + α 2 LCmic o i /LCla ge i + α 3 GDPi /PCi + α x INSTIT a i x+εi Lsmall i = β1 KR/VAsmall i + β 2 LCsmall i /LCla ge i + β 3 GDPi /PCi + β x INSTIT a i x+εi VAmic o i = γ1 KR/VAmic o i + γ 2 LCmic o i /LCla ge i +γ 3 GDPi /PCi + γ x INSTIT a i x+εi VAsmall i = δ1 KR/VAsmall i + δ 2 LCsmall i /LCla ge i + δ 3 GDPi /PCi + δ x INSTIT a i x+εi whe e i = 1, …, 28 a e coun ies, = 2002, …, 2008 a e he obse ed yea s, x = {4, 5, …, 13} indica es he espec i e numbe o ins i u ional a iable 4 h ough 13. 4.2 Commen s on he econome ic esul s In Tables 2 and 3 we epo he esul s o ou eg essions as speci ied abo e4. In each eg ession, we included h ee economic explana o y a iables, as well as, some ele an ins i u ional explana o y a iables. The non-signi ican ins i u ional a iables we e d opped om he model. In he i s column o each eg ession, we epo esul s o classic ixed e ec s model. In he ollowing columns, we epo he esul s o LTS eg ession, applied on he da a cen e ed by median, wi h ega d o a di e en choice o h. Table 2 | Robus Fixed E ec s Reg essions - Models 1 and 2 Model 1 2 Dependen Va iables Lmic o i Lsmall i h% - 95% 85% 75% - 95% 85% 75% Economic KR/VAmic o i -0.080* (0.043) -0.329*** (0.053) -0.210*** (0.047) 0.009 (0.017) KR/VAsmall i -0.164*** (0.021) -0.157*** (0.015) -0.166*** (0.010) -0.005 (0.051) LCmic o/la ge i -0.346*** (0.062) -0.398*** (0.051) -0.318*** (0.039) -0.157*** (0.025) LCsmall/la ge i -0.330*** (0.091) -0.167** (0.074) -0.016*** (0.054) 0.016 (0.049) GDP/PCi 0.509*** (0.039) 0.419*** (0.029) 0.405*** (0.021) 0.377*** (0.018) 0.541*** (0.035) 0.496*** (0.026) 0.407*** (0.003) 0.423*** (0.017) Ins i u ional MONET 0.003** (0.001) 0.0003 (0.001) -0.001* (0.001) -0.003*** (0.001) FINANC 0.001** (0.001) 0.002*** (0.001) 0.0006** (0.0003) 0.0004** (0.002) 0.001** (0.001) 0.001** (0.0004) 0.001** (0.0003) 0.001 (0.001) LABOUR -0.001 (0.001) -0.001 (0.001) -0.002** (0.001) -0.001*** (0.0004) Numbe o obs. 196 187 167 147 196 187 167 147 Adj. R2 0.525 0.603 0.700 0.772 0.634 0.748 0.751 0.837 No es: The alue o h% deno es how many obse a ions we e included in o da a se . * signi ican a 10%; ** sig- ni ican a 5%; *** signi ican a 1%. S anda d e o s a e in b acke s. Fixed e ec s a e no epo ed. Va iance in- la ion ac o does no sugges any p oblems wi h collinea i y in eg essions. Dependen a iables and GDP pe capi a a e in loga i hms. B eusch – Pagan / Cook Weisbe g does no ejec he hypo hesis o homoscedas ici y. Sou ce: Own calcula ion (S a a, Ma lab) 4 All es ima es we e ob ained by S a a and Ma lab