Deflation and output across sectors: results for the Czech Republic

94 2018, XXI, 4 Ekonomie DOI: 10.15240/ ul/001/2018-4-007 In oduc ion The s udy o de l a ion seems o be gaining e e mo e impo ance. Cen al banks in mos ad anced economies, including he Eu opean Cen al Bank and he Czech Na ional Bank, ha e obse ed CPI in l a ion unning below hei a ge s o e en in ou igh de l a iona y e i o y. These cen al banks a gue ha de l a ion should be a oided a all cos s and employ ex ao dina y policies such as quan i a i e easing, o eign exchange in e en ions o nega i e nominal in e es a es o i gh agains i . As hese policies ha e no always led o highe economic g ow h and highe in l a ion, he e ha e been calls o e en mo e ex ao dina y measu es. Among hose is a p oposal o abolish cash money in o de o allow banks o ca y o e he nega i e in e es a es mo e e i cien ly on hei deposi o s and s oke in l a ion (see e.g. Bui e , (2009) o Rogo (2014)). The e o e, a lo is a s ake in knowing whe he de l a ion (especially mild de l a ion) should be a oided by all means. The e a e subs an ial p oblems wi h s udies ha ha e ied o assess he ela ionship be ween economic g ow h and he p ice le el. Fi s , all o hem ha e used agg ega e da a o coun ies such as GDP and CPI. While agg ega e a iables p o ide in o ma ion abou he economy as a whole, hey canno e eal he po en ial a ia ion in ou pu and p ices ha akes place ‘inside’ he economy. As a esul , we know e y li le abou whe he i ms and sec o s wi hin he economy ace de l a ion and whe e i comes om. We can also ha dly assess whe he de l a ion- ecession heo ies hold. Second, episodes o de l a ion in he agg ega e CPI o in he GDP de l a o ha e been sca ce in he pas decades. As a esul , s udies on de l a ion ha e had o ely on his o ical da a, o en be o e Wo ld Wa I. Wha e e hese s udies show, hei conclusions can be c i icized as i ele an , because hey a e based on ou da ed obse a ions. In his pape , we ake a di e en app oach. Ins ead o using agg ega e da a on GDP and in l a ion, we ocus on sec o da a on p oduc ion and p ices. Speci i cally, we use da a om he Czech S a is ical O i ce on p oduc ion, g oss alue added and p ices in sec o s o he Czech economy om 1993 o 2015. This da ase has h ee ad an ages: (1) i is ich in in o ma ion on i ms’ ou pu and inpu s, (2) i p o ides ecen obse a ions, and (3) i con ains nume ous episodes o sec o de l a ion, which would on he mac oeconomic le el be concealed unde he agg ega e CPI o de l a o numbe s. We hink his app oach is no el: we do no know o ano he s udy ha would analyze de l a ion and g ow h using sec o da a. We i nd ha , con a y o common wisdom, de l a iona y p essu es in he Czech economy ha e been coming om sec o s wi h inc easing ou pu and inc easing g oss alue added, no om he de e io a ing ones. This shows ha de l a ion was mos o en g ow h-d i en, ep esen ing ising p oduc i i y. This ex p oceeds as ollows. In Sec ion 1, we p esen he cu en s a e o esea ch on de l a ion and show i s main d awbacks. In Sec ion 2, we p esen ou sec o da a om na ional accoun s. We i s show some desc ip i e s a is ics in Sec ion 3. We hen pe o m eg ession analysis in Sec ion 4, whe e we use he i xed e ec s model and Gene al Me hod o Momen s es ima ion o eg ess he g ow h in p oduc ion and g oss alue added on he g ow h o p ices plus con ol a iables. The las sec ion concludes he pape . 1. Cu en Resea ch 1.1 Empi ical S udies The e a e h ee main lines o easoning why de l a ion is hough by many o be ha m ul o economic g ow h. Fi s , he expec a ion o alling p ices may delay spending by consume s DEFLATION AND OUTPUT ACROSS SECTORS: RESULTS FOR THE CZECH REPUBLIC Pa el Ryska, Pe Sklenář EM_4_2018.indd 94EM_4_2018.indd 94 28.11.2018 13:12:5728.11.2018 13:12:57 95 4, XXI, 2018 Economics and businesses. Fi ms see hei e enues all and i hey canno adjus wages and o he cos s acco dingly, hei ma gins sh ink and hey ha e o lay o wo ke s. Second, o a gi en nominal in e es a e, de l a ion aises he eal in e es a e, which in u n migh de e in es men . Thi d, i he e is signi i can indeb edness in he economy, de l a ion may cause bank up cies as i inc eases he eal alue o deb . Se e al s udies ha e a emp ed o assess empi ically whe he he alleged link be ween de l a ion and ecession holds. Bo do and Redish (2003) es ic hei a en ion o he Uni ed S a es and Canada in he pe iod 1870- 1913, which was ich in episodes o de l a ion. They i nd ha p ices did no a ec ou pu . Bo do, Lane and Redish (2004) add Ge many o he sample and each a simila conclusion. O he ecen s udies use long da ase s ha span mo e han a cen u y and con ain many coun ies. A keson and Kehoe (2004) conclude om hei eg ession o ou pu g ow h on g ow h o p ices ha in he whole sample, he co ela ion is close o ze o. An excep ion is he subsample o he G ea Dep ession in he 1930s, whe e he link was posi i e. Bo io and Fila do (2004) as well as Bo io e al. (2015) gene ally con i m his esul . By con as , Gue e o and Pa ke (2006) lean o he conclusion ha de l a ion is bad o g ow h, i nding an economically small bu s a is ically signi i can link. Acco ding o hem, lagged de l a ion nega i ely a ec s ou pu . Benhabib and Spiegel (2009) i nd a non-linea ela ionship be ween de l a ion and ecession. In ecen yea s, some esea che s ha e ocused on he link be ween expec ed in l a ion (o de l a ion) and consump ion. In e es ingly, hey ha e p oduced s a kly di e en esul s. Bachmann, Be g and Sims (2015), using ecen da a on he US economy, e use he hypo hesis ha he expec a ion o de l a ion leads o lowe p esen consump ion. By con as , Ichiue and Nishiguchi (2014) suppo his hypo hesis using su ey da a on mode n Japan. O e all, he exis ing empi ical wo k on de la ion is a he limi ed – which is su p ising gi en i s impo ance o mone a y policy – and inconclusi e. As we explain below, he e is a majo p oblem o applicabili y o he majo i y o hese s udies o mode n-day mone a y policy. 1.2 ‘Good’ o ‘Bad’ De l a ion? Two lines o hough can be aced in cu en mains eam mac oeconomics. The i s g oup conside s de l a ion decisi ely ha m ul and s udies ways o a oid i . He e we can include in l uen ial mac oeconomis s K ugman (1998), Be nanke (2002) o S ensson (2003). The second g oup akes a mo e de ailed look a de l a ion. Bo do and Redish (2003) coined he e ms ‘good’ and ‘bad’ de l a ion. De l a ion is o he good ype i is associa ed wi h ising ou pu . In pa icula , good de l a ion occu s when i ms in es o dec ease uni cos s and inc ease ou pu . As a esul , i he economy ope a es wi h a cons an money supply, consume s buy an inc eased ou pu wi h he same amoun o money, so he p ice le el dec eases. This is a synonym o an ou wa d shi in he agg ega e supply cu e. This si ua ion is hough o ha e been ypical o he p e-Wo ld Wa I pe iod. By con as , bad de l a ion is associa ed wi h alling ou pu . In his case, de l a ion esul s om dec easing nominal demand, which canno be immedia ely passed on o lowe p ices o inpu s. Fi ms see hei p o i abili y dec ease and cu p oduc ion. This was mos p obably one o he cha ac e is ics o he G ea Dep ession o he 1930s. O he wo ks adhe ing o his dis inc ion be ween good and bad de l a ion include Beckwo h (2007), Bo do and Fila do (2005) and o a ce ain ex en also Bo io e al. (2015). The e ms good and bad de l a ion do no desc ibe causali y – hey only desc ibe co ela ions. Fo example, bad de l a ion does no necessa ily mean ha de l a ion causes alling ou pu . Dec easing p ices may only be a symp om o alling demand, no he oo cause i sel . Howe e , he e ms may be use ul in p ac ice because policymake s y o de e mine he ype and decide whe he hey should o should no coun e ac i . I de l a ion esul s om dec easing nominal demand, cen al banks conside i undesi able and p e e o o se i by easing policy. On he con a y, i i esul s om cheape p oduc ion wi hou di ec link o demand, cen al banks may be mo e willing o le such de l a ion un i s cou se. The e o e, i we i nd ha mos de l a ion going on in he economy is o he good ype, i may ha e a di ec policy implica ion. Fo hese pu poses, he exis ing empi ical esea ch has a clea disad an age. All he s udies men ioned in Sec ion 1.1 use only agg ega e mac oeconomic a iables, mos o en he GDP and CPI in l a ion. This causes a ade-o : he agg ega e annual da a do no EM_4_2018.indd 95EM_4_2018.indd 95 28.11.2018 13:12:5828.11.2018 13:12:58 96 2018, XXI, 4 Ekonomie show almos any yea s wi h de l a ion in he las decades o mos de eloped economies (Fo he Czech Republic, oo, he agg ega e p ice de l a o shows only one yea wi h de l a ion in he pe iod 1993-2015.). When esea ches wan o use agg ega e da a and lea n mo e abou de l a ion, hey ha e o use p e-Wo ld Wa I da a which a e ich in obse a ions o de l a ion. Bu hen he esul s a e based on old obse a ions which may ha e li le ele ance o oday because mone a y egimes ha e changed, he e a e s a kly di e en le els o deb , he e is mo e i nancial in e media ion, e c. By con as , using sec o da a, we ha e many episodes o mode n-day de l a ion and we ci cum en he ade-o . We a e hus be e equipped o make judgmen s abou a po en ial link be ween p ices and ou pu and o assess pa icula de l a ion- ecession heo ies. (Below we use he e ms in l a ion and de l a ion also o he change o sec o p ices. Some may insis ha in l a ion and de l a ion deno e only agg ega e p ice mo emen s, bu no sec o p ice mo emen s. Howe e , all agg ega e numbe s ha e hei sou ces in hei componen s, and i is p ecisely hese sou ces ha we analyse below. Thus we use in l a ion and de l a ion in he b oade sense.) 2. Da a We use da a om he Czech S a is ical O i ce (2017). I has comple e da a on 86 sec o s o he Czech economy as de i ned in na ional accoun s. The de i ni ion o each indus y is based on NACE classi i ca ion and alues a e published in he s a is ics o Na ional Accoun s. The da a spans om 1993 o 2015. As we use g ow h in a iables a he han le els, one ime pe iod d ops ou , so we ha e 22 ime pe iods. Tha gi es a o al o 1892 obse a ions. The a iables a e ou pu p ices (implici p ice de l a o ), ou pu , g oss alue added (GVA), employmen and in e media e inpu s. Ou pu , GVA, employmen and in e media e inpu s a e a ailable bo h in nominal and eal e ms, bu in his a icle we use hem in eal e ms (i.e., in cons an p ices) as we a e in e es ed in assessing he e ec o p ices on eal ou pu o i ms. An impo an ea u e o he NACE classi i ca ion is ha i does no co e only sec o s p oducing o i nal consump ion, bu also sec o s uppe in he p oduc ion chain – i.e., sec o s p oducing capi al goods and in e media e inpu s. As a esul , i ms’ ou pu does no mean only consume goods bough by consume s, bu also capi al goods bough by o he i ms. Fig. 1: P oduc ion and GVA g ow h agains change o p ices Sou ce: Czech S a is ical O i ce (2017), own compu a ions EM_4_2018.indd 96EM_4_2018.indd 96 28.11.2018 13:12:5828.11.2018 13:12:58 97 4, XXI, 2018 Economics 3. Desc ip i e S a is ics In ou sample, posi i e p ice g ow h p e ails and accoun s o 80% o all obse a ions (see Fig. 1). A he same ime, app oxima ely h ee i hs o obse a ions ha e posi i e g ow h o p oduc ion and g oss alue added. As a esul , oughly hal o obse a ions lie in he op- igh quad an in Fig. 1. In his sec ion we ocus on basic s a is ical ela ionships be ween he sign o p ice change (in l a ion/de l a ion) and he sign and g ow h a e o p oduc ion o GVA. In pa icula , we ocus on whe he declines in p oduc ion a e associa ed wi h de l a ion. In Tab. 1, we epo he compa ison o g ow h in p oduc ion and g oss alue added unde inc easing e sus dec easing p ices. We also isualize he compa isons wi h densi y g aphs in Fig. 2. Ou o obse a ions wi h p ice de l a ion, 69.7% eco ded posi i e g ow h in eal p oduc ion, while he co esponding pe cen age o obse a ions wi h in l a ion is only 58.8%. This is also e l ec ed in he a e ages: he a e age g ow h a e o p oduc ion du ing episodes o p ice de l a ion is 5.47%, which is sizably highe han he a e age g ow h wi h in l a ion (2.85%). The e o e, his p elimina y obse a ion con adic s he no ion ha de l a ion is linked o subpa ou pu g ow h. To check his i nding, we also an a o mal es o equali y o means. As seen in Tab. 2, he - es ejec s he null hypo hesis ha he a e age p oduc ion g ow h unde in l a ion is equal o ha unde de l a ion, he eby con i ming he i nding. In con as , he s anda d de ia ions o g ow h a es a e simila , which is also suppo ed by he es ing o equali y o a iances. We pe o med he same analysis also o eal g oss alue added (see Tab. 1). He e he be e pe o mance unde de l a ion is e en mo e p onounced: 77.8% o de l a iona y obse a ions epo posi i e g ow h in g oss alue added, while o in l a iona y obse a ions he pe cen age is only 52.3%. The a e age g ow h o g oss alue added is 14.1% wi h de l a ion, while only 0.68% wi h in l a ion. The di e ence is also con i med by he - es . The only di e ence wi h g oss alue added as opposed o p oduc ion is he highe s anda d de ia ion wi h de l a ion han wi h in l a ion. The eason o hese a he con incing i ndings may be he p e alence o ‘good de l a ion’ in ou sample: he sec o s ha epo ed p oduc p ice de l a ion could be p ecisely he ones ha in es ed mos in p oduc ion and he e o e enabled cheape and g ea e p oduc ion. Anyway, his p elimina y look uns agains he claim ha de l a ion ha ms p ospec s o g ow h. In he Appendix, we e e se ou pe spec i e and ask he ques ion how much g ow h in p ices he e is unde he opposi e scena ios o ise and all in p oduc ion and GVA. All da a In l a ion De l a ion Obse a ions 1,892 1,509 383 P oduc ion obs. wi h p oduc ion inc ease 61.00% 58.80% 69.70% obs. wi h p oduc ion dec ease 39.00% 41.20% 30.30% Mean g ow h 3.38 2.85 5.47 S anda d de ia ion o g ow h 12.89 12.71 13.39 G oss alue added obs. wi h GVA inc ease 57.50% 52.30% 77.80% obs. wi h GVA dec ease 42.50% 47.70% 22.20% Mean g ow h 3.40 0.68 14.10 S anda d de ia ion o g ow h 34.86 25.74 57.05 Sou ce: Czech S a is ical O i ce (2017), own compu a ions Tab. 1: G ow h o p oduc ion and GVA unde in l a ion and de l a ion EM_4_2018.indd 97EM_4_2018.indd 97 28.11.2018 13:12:5828.11.2018 13:12:58 98 2018, XXI, 4 Ekonomie This complemen s ou esul he e ha slowe p ice g ow h o ou igh de l a ion is accompanied by quicke ou pu g ow h. 4. Reg ession Analysis 4.1 Time A e ages As a i s look on co ela ions, we pe o med a simple eg ession o ime a e ages. We eg essed a e age annual ou pu g ow h in he 86 sec o s o he Czech economy o e 1993- 2015 on he co esponding a e age annual p ice g ow h, as seen in Tab. 3. In wha ollows below, we deno e x he annual pe cen age change in a iable X. Tha is, p is he g ow h in p ices, q in eal ou pu (p oduc ion) and g a in eal g oss alue added. We es ed he esul s in Tab. 3 o he p esence o he e oscedas ici y and whene e i was de ec ed we used he Whi e’s he e oscedas ici y-co ec ed co a iance ma ix. Tes s a is ic p- alue P oduc ion - es o equali y o means 3.44*** <0.001 F- es o equali y o a iances 1.11 0.19 G oss alue added - es o equali y o means 4.49*** <0.001 F- es o equali y o a iances 4.91*** <0.001 Sou ce: Czech S a is ical O i ce (2017), own compu a ions No e: We use he Welch - es , i.e., a s anda d unpai ed and wo-sided - es o de e mine whe he he means o p odu- c ion g ow h unde in l a ion and de l a ion a e equal o each o he . The null hypo hesis is ha he wo means a e equal. To es he equali y o a iances, we use an F- es , whe e he null hypo hesis is ha he a io o he a iances o he samples is equal o 1. No e: The sign *** deno es signi i cance a 1% le el. Tab. 2: Tes s o equali y o means and a iances Fig. 2: Densi y o p oduc ion and GVA g ow h unde de l a ion and in l a ion Sou ce: Czech S a is ical O i ce (2017), own compu a ions EM_4_2018.indd 98EM_4_2018.indd 98 28.11.2018 13:12:5828.11.2018 13:12:58 99 4, XXI, 2018 Economics The eg ession coe i cien on p (pe cen age g ow h in p ices) is -1.142, signi i can a 1%, sugges ing ha slowe p ice g ow h has been associa ed wi h as e ou pu g ow h. As depic ed in Fig. 3, he e is one dis inc ou lie on op le . This sec o is he manu ac u ing o compu e , elec onic and op ical p oduc s. In e es ingly, i is bo h he sec o wi h he as es g ow h a e o p oduc ion and he only sec o wi h a e age annual de l a ion o e 1993- 2015. We also an he eg ession wi hou his ou lie , bu he nega i e coe i cien and high s a is ical signi i cance emain (see Tab. 3). We ied lea ing ou up o 10 mos ex eme obse a ions, bu he nega i e and s a is ically signi i can coe i cien o p emained. We can obse e he gene al pa e n o he sec o s in he Czech economy om Fig. 3: while All da a Wi hou ou lie Coe i cien p- alue Coe i cien p- alue Reg essing q on: In e cep 7.082*** <0.001 -6.118*** <0.001 p -1.142*** <0.001 -0.919*** <0.001 Obse a ions 86 85 Adj. R2 0.237 0.167 Reg essing g a on: In e cep 8.996*** <0.001 8.824*** <0.001 p -1.998*** <0.001 -1.819*** <0.001 Obse a ions 86 85 Adj. R2 0.380 0.323 Sou ce: Czech S a is ical O i ce (2017), own compu a ions No e: The sign *** deno es signi i cance a 1% le el. Tab. 3: Time a e ages: Reg ession o a e age annual g ow h in ou pu and GVA on in l a ion Fig. 3: Reg essing ime a e ages: (a) Ou pu g ow h on in l a ion, (b) GVA g ow h on in l a ion (all da a) Sou ce: Czech S a is ical O i ce (2017), own compu a ions EM_4_2018.indd 99EM_4_2018.indd 99 28.11.2018 13:12:5828.11.2018 13:12:58 100 2018, XXI, 4 Ekonomie he con ac ing sec o s had bo h low and high in l a ion, booming sec o s ended o ha e lowe in l a ion and hose ha expanded mos quickly – wi h g ow h a e o ou pu abo e 10% – had exclusi ely low in l a ion o ou igh de l a ion. The esul s sugges ha highe ou pu g ow h is linked o slowe p ice g ow h, i.e. low in l a ion o ou igh de l a ion. Howe e , i may no be e iden wha happens o i ms’ p o i s as in l a ion slows o u ns in o de l a ion. We he e o e used g ow h in eal g oss alue added ins ead o eal ou pu g ow h and eg essed i on p ice g ow h. The esul s a e e y simila o hose wi h ou pu g ow h (see Tab. 3 and Fig. 3). O e all, his e idence oo seems o un agains he de l a ion- ecession heo ies. In ou sample, as e ou pu g ow h and p o i g ow h a e associa ed wi h slowe in l a ion o ou igh de l a ion. This could sugges ha highe p oduc ion mos o en esul s om in es men in o lowe -cos p oduc ion, o which consume s eac by pu chasing mo e. 4.2 Panel Da a Two Models Using panel da a na u ally o e s much mo e in o ma ion han ime a e ages because we can make use o he en i e a ia ion o annual obse a ions. Ou da a a e in he o m o balanced panel da a and we use he i xed e ec s model o es ima ion. This model allows unobse ed sec o -speci i c e ec s o ha e any co ela ion wi h he explici eg esso s. We es ima ed he models below using he so-called ‘wi hin es ima ion’. Gene ally, we use wo economic models o es ima ion. The i s is an au o eg essi e dis ibu ed lag (ADL) model which akes a mac oeconomic app oach o p ices and p oduc ion. Jus as s udies ci ed in Sec ion 1.1 eg ess GDP g ow h on p ice g ow h (in l a ion o de l a ion), we eg ess sec o ou pu g ow h on he espec i e g ow h in ou pu p ices o he gi en sec o . This di ec ly add esses he ques ion o co ela ion be ween p ices and ou pu . We also include lagged a iables. Fo example, we es ima e qi = â0 + â1qi -1 + â2 pi + â3 pi -1 + ai + ui (1) whe e qi is eal ou pu g ow h in sec o i and yea , pi g ow h in ou pu p ices in he sec o (bo h in pe cen e ms), ai he sec o -speci i c unobse ed e ec and ui he e o e m. The second model, in con as , akes a mo e mic oeconomic, i m-le el app oach. To p oduce ou pu , i ms mus employ wo ke s and buy in e media e inpu s. The e o e, he model we es ima e is qi = â0 + b1 pi + â 2empi + â3inpi + ai + ui (2) whe e emp is g ow h in labou employed (measu ed in o al hou s wo ked) and inp g ow h in he olume o in e media e inpu s in p oduc ion. The eg esso s include wo usual inpu s in he neoclassical p oduc ion unc ion ( o why we exclude capi al K, see ou discussion below) plus a special e m – he ou pu p ice change p. P oduc i i y, o e i ciency o p oduc ion, is usually accoun ed o as a esidual, i.e., wha emains unexplained in ou pu p oduced i we ake in o accoun changes in labou , capi al and in e media e inpu s. Fi ms ypically in es o ake o he measu es in o de o make p oduc ion mo e e i cien . A e hey do so, hey can p oduce mo e and wi h lowe cos s, so hey can o e hei p oduc s mo e cheaply and each mo e cus ome s. I his is he case, hen he ou pu p ice P ac s as a p oxy o e i ciency. The lowe is he cos o p oduc ion, he lowe is he obse ed p ice P. The e o e, p ice g ow h p is no only a a iable ha we add o i nd ou abou i s co ela ion wi h ou pu g ow h q, bu i also has a conc e e economic in e p e a ion as a p oxy o he e i ciency o p oduc ion. Equa ion 1 abo e con ains he lagged dependen a iable qi -1 among eg esso s. As a esul , he i xed-e ec s es ima o is gene ally no consis en . As Woold idge (2002) shows, he bias alls a a a e 1/T as T g ows, and o a ime dimension high enough, he inconsis ency would be negligible. We hink ha wi h ou T = 22, we do ha e a T high enough o be su e ha mos o he bias disappea s. Howe e , as a check we also pe o m Gene al Me hod o Momen s (GMM) es ima ion in equa ions whe e he e a e lagged dependen a iables. Ou GMM es ima ion ollows A ellano and Bond’s (1991) wo-s ep p ocedu e and we use lags o o de 2 o 5 o he dependen a iable as ins umen al a iables. Finally, in he i xed e ec s models below, each eg ession was es ed o he e oscedas ici y and se ial co ela ion. When ei he o hese was de ec ed, a obus a iance ma ix es ima o was used o co ec he a iances and es s a is ics. EM_4_2018.indd 100EM_4_2018.indd 100 28.11.2018 13:12:5928.11.2018 13:12:59 101 4, XXI, 2018 Economics Resul s: Mac oeconomic App oach In he i s , mac oeconomic app oach, we es ima e he e ec o in l a ion on ou pu g ow h and GVA g ow h in an au o eg essi e dis ibu ed lag model. Tab. 4 shows ha con empo aneous in l a ion p is nega i ely linked wi h ou pu g ow h, con i ming ou i ndings om Sec ion 4.1 abo e. The GMM es ima ion b ings an es ima e o he coe i cien o p e y simila o he i xed e ec s es ima ion (-0.303 s. -0.293) and con i ms i s s a is ical signi i cance. By con as , he coe i cien o p -1 is economically and s a is ically less signi i can , bo h acco ding o i xed e ec s es ima ion and GMM. Simila ly o ou pu , in l a ion is nega i ely linked o he g ow h in g oss alue added o i ms (see he lowe hal o Tab. 4). The only di e ence is ha he magni ude o he coe i cien o p is bigge o g oss alue added han o ou pu as dependen a iable. This is unde s andable since i ms usually ha e ope a ing le e age: a ise in ou pu ( e enues) by 1% causes a ise in g oss p o i o mo e han 1%. Resul s: P oduc ion-Func ion App oach In he second, mo e mic oeconomic app oach, we use ypical inpu s in o p oduc ion unc ion as con ol a iables o assessing he impac o in l a ion on ou pu g ow h. These should be mo e powe ul con ol a iables as hey a e di ec ly linked o ou pu . We es ima e he equa ion wi h g ow h in labou emp ( o al hou s wo ked) and g ow h in he olume o in e media e inpu s inp as con ols. We ha e also added g ow h in g oss capi al s ock k o imi a e mo e he adi ional p oduc ion unc ion Y = (L, K). Because da a on capi al s ock ha e poo e a ailabili y han o he da a, we used an al e na i e sou ce om he OECD (2016). I is bo h na owe (only 57 sec o s) and sho e (1995-2009) han ou main da ase , so i has much ewe obse a ions (798). Resul s om his enla ged eg ession ha includes k a e no ma e ially di e en om hose epo ed in Tab. 5 and we do no epo hem he e. As expec ed, his model shows a much highe i as Adj. R2 is as high as 0.638 in he ou pu eg ession in Tab. 5. And again, in l a ion nega i ely a ec s ou pu g ow h (-0.476) wi h high s a is ical signi i cance. Labou and in e media e inpu s also ha e he expec ed signs and high signi i cance. In he GVA eg ession, in l a ion also has a nega i e coe i cien and he coe i cien is again g ea e in absolu e magni ude. In e es ingly, while he labou inpu has a posi i e and s a is ically signi i can impac on GVA, in e media e inpu s lose hei signi i cance. In e p e a ion The empi ical esul s b ing, in ou iew, wo impo an i ndings. Fixed e ec s GMM Coe i cien p- alue Coe i cien p- alue Reg essing q on: q -1 0.012 0.696 0.042 0.231 p -0.293*** 0.003 -0.303** 0.014 p -1 0.136* 0.052 0.110 0.121 Obse a ions 1,806 1,720 Adj. R2 0.017 - Reg essing g a on: g a -1 -0.086*** 0.002 -0.178 0.236 p -1.614*** <0.001 -1.768*** <0.001 p -1 0.614 0.173 0.393 0.617 Obse a ions 1,806 1,720 Adj. R2 0.068 - Sou ce: Czech S a is ical O i ce (2017), own compu a ions No e: The sign * deno es s a is ical signi i cance a 10%, ** a 5% and *** a 1% le el. Tab. 4: ADL: Reg ession o g ow h in ou pu and GVA on in l a ion EM_4_2018.indd 101EM_4_2018.indd 101 28.11.2018 13:12:5928.11.2018 13:12:59 102 2018, XXI, 4 Ekonomie Fi s ly, he nega i e ela ionship be ween g ow h in ou pu and g ow h in p ices sugges s ha obse ed de l a ion has been p edominan ly o he ‘good’ ype. I alling p ices esul ed om alling nominal demand, i ms would no espond by inc easing p oduc ion – on he con a y, hey would cu i , as he ‘bad de l a ion’ hypo hesis goes. Hence, we would obse e a posi i e co ela ion be ween p and q. The al e na i e hypo hesis is ha as i ms in es and imp o e e i ciency, hey manage o cu uni cos s and p ices. This allows hem o sell mo e p oduc s, i.e., each new ma ginal demand h ough lowe p ices. This ‘good de l a ion’ hypo hesis is consis en wi h ou i ndings – we see epea edly a nega i e ela ionship be ween p and q. I could be a gued ha he nega i e ela ionship be ween p and q may in some mo e in l a iona y yea s e l ec a he he compa ison o high e sus low in l a ion han in l a ion e sus de l a ion. Tha may be ue. The shi o he agg ega e demand cu e – o example, h ough cen al bank easing – may mo e he whole mass o obse a ions o he igh in Fig. 3, b inging many o hem ou o he de l a iona y sec ion. Symme ically, lack o agg ega e nominal demand may push hem o he le , which was he case o ins ance in 2009 when he economy as a whole had a 1.4% de l a ion (measu ed as ou pu de l a o ). Howe e , a he han ocusing on he posi ion o he mass o obse a ions, we a e in e es ed in he slope o he eg ession line – ha is, in he shape o he mass. I s nega i e slope sugges s ha he endency o cu p ices o a leas ha e slowe p ice inc eases han he a e age is associa ed wi h g owing i ms, no declining ones. O e all, we a e no claiming ha ‘bad de l a ion’ does no exis o ha i canno be ha m ul. We a e only showing e idence ha he e is much mo e ‘good de l a ion’ han ‘bad de l a ion’ in ou sample. Secondly, using g oss alue added in addi ion o ou pu , we ha e shown ha lowe p ices a e no only associa ed wi h g ea e ou pu , bu also wi h g ea e g oss p o i o i ms. Ou pu in i sel is no he i ms’ goal, bu p o i is. I i ms in es o inc ease p oduc ion and cu p ices, hey also succeed in inc easing p o i s, ou esul s show. This i nding is impo an because one line o easoning (s a ing wi h Fishe , 1933) is ha de l a ion o oo low in l a ion ul ima ely e odes i ms’ p o i s and leads hem o bank up cies. Ou esul s show o he wise, again p o iding suppo o he case ha de l a ion in ou sample s ems om i ms’ own ini ia i es, no om nega i e shocks ha would squeeze p o i s. As poin ed ou in Sec ion 2, he sec o s used a e no only hose ha p oduce consume Fixed e ec s Coe i cien p- alue Reg essing q on: p -0.476*** <0.001 emp 0.280*** <0.001 inp 0.582*** <0.001 Obse a ions 1,700 Adj. R2 0.638 Reg essing g a on: p -1.663*** <0.001 emp 0.889*** <0.001 inp 0.062 0.637 Obse a ions 1,700 Adj. R2 0.112 Sou ce: Czech S a is ical O i ce (2017), own compu a ions No e: The sign *** deno es signi i cance a 1% le el. Tab. 5: P oduc ion unc ion: Reg ession o g ow h in ou pu and GVA on in l a ion EM_4_2018.indd 102EM_4_2018.indd 102 28.11.2018 13:12:5928.11.2018 13:12:59