1 The Economic Impac of Cloud Compuing on Business Creaion, Employmen and Oupu in Europe An applicaion of he Endogenous Marke Srucures Approach o a GPT innovaion Federico Ero ABSTRACT Cloud compuing is a new general purpose Inerne-based echnology hrough which informaion is sored in servers and provided as a service and on-demand o cliens. Adoping he endogenous marke srucures approach o macroeconomics, we analyze he economic impac of he gradual inroducion of cloud compuing and we emphasize is role in fosering business creaion and compeiion hanks o he reducion of he fixed coss of enry in ICT capial. Our calculaions based on a DSGE model show a significaive impac for he European Union wih he creaion of a few hundred housands new SMEs and a significan conribuion o growh. Governmens could enhance hese benefis by subsidizing he adopion of cloud compuing soluions. JEL CODES L, E32. KEYWORDS I. Inroducion Endogenous marke srucure, Cloud compuing, Uiliy compuing, General purpose echnology, Firms Enry, Business Cycle The inroducion of a general purpose echnology can provide a fundamenal conribuion o promoe growh and compeiion, 2 and i can help he economy o recover from a severe downurn as he curren one. In his aricle we employ he endogenous marke srucures approach (Ero, 2004, 2007a) 3 o sudy he economic impac of an innovaion in he hardware-sofware field which is going o have a profound effec on he marke srucure of many secors and on he global macroeconomic performance in he nex years. This innovaion is associaed wih cloud compuing, he new fronier of he Inerne era, a echnology hrough which informaion will be sored in servers and provided on line as a service o cliens in a pay-as-you-go manner. Firms will be able o adop his service on demand, so as o 2009 / 2 Review of Business and Economics
2 80 Federico Ero Review of Business and Economics 2009 / 2 avoid large up-fron coss (ha are currenly necessary for hardware and sofware equipmen) and spend in ICT according o heir producion necessiies see Dubey and Wagle (2007) and Armbrus e al. (2009) for early reviews of he opic. This will have a large impac on he cos srucure and hrough i on he producion possibiliies of all firms, especially small and medium size enerprises (SMEs). Our focus will be mainly on he heoreical impac of his innovaion on he creaion of new firms and producs, on he empirical evidence abou he impac of is inroducion for he European economy and on he implicaions for policies supporing cloud compuing adopion. This allows us o apply he heory of endogenous marke srucures o examine a macroeconomic experimen ha can be hardly approached wihin he neoclassical or New-Keynesian frameworks. In he nex Secion II we will describe he naure of he general purpose echnology represened by cloud compuing, is genesis and is likely impac on he economy and on he srucure of many markes. As we will see, he inroducion of cloud compuing is going o reduce drasically he fixed coss of enry and producion, urning par of hem ino variable coss relaed o he producion necessiies. This will have a posiive impac on enry and compeiion in all secors where fixed ICT spending is crucial. The posiive associaion beween ICT innovaions and compeiion is well known, and policymakers recognize ha i may work in boh direcions: on one side compeiive secors adop ICT innovaions earlier and become more producive, on he oher side ICT adopion enhances compeiion. For insance, he e-business Wach of he European Commission (2008) noices ha while i seems obvious ha increasing levels of compeiion can push companies o adop and use ICT, he opposie migh well also be he case. In fac, ICT and he usage of he inerne have drasically impaced on cerain secors such as banking and reshaped he compeiive scenario (p. 42). In Secion III we will describe our approach o he esimae of he economic impac of cloud compuing. We will adop a sandard macroeconomic model augmened wih endogenous marke srucures and will simulae he impac of a gradual reducion of he fixed coss of enry. The experimen will be based on a dynamic sochasic general equilibrium model calibraed on he European economy and ranslaed in empirical resuls on he basis of Eurosa daa. In his sense, his paper provides a simple applicaion of he endogenous marke srucures approach o macroeconomics. However is resuls should be only seen as preliminary and empaive: furher work is needed o improve he mach beween our heoreical foundaions and he empirical exercise. Secion IV will presen he resuls. Saring from conservaive assumpions on he cos reducion process, we obain ha he diffusion of cloud compuing will provide a posiive conribuion o he annual growh rae (up o a few decimal poins), conribuing o creae abou a million new jobs hrough he developmen of a few hundred housand new SMEs in he whole EU-27. We will presen he resuls on business creaion for each counry and for each macro-area. The driving
3 Economic Impac of Cloud Compuing on Business Creaion, Employmen and Oupu 8 mechanism behind he posiive conribuion of cloud compuing works hrough he incenives o creae new firms, and in paricular SMEs. One of he main obsacles o enry in new markes is represened by he high up-fron coss of enry, ofen associaed wih physical and ICT capial spending. Cloud compuing allows poenial enrans o save in he fixed coss associaed wih hardware/sofware adopion and wih general ICT invesmen, and urns par of hese coss ino variable coss. This reduces he consrains on enry and promoes business creaion. The imporance of such a mechanism is well known a he policy level, especially in Europe, where SMEs play a crucial role in he producion srucure. Again he e-business Wach of he European Commission (2008) emphasizes his aspec clearly: SMEs form significan indusry segmens in he EU and accoun for he majoriy share in EU employmen. Thus, hey require specific policy aenion. While heir srengh lies in he flexibiliy wih which hey can adjus o changing marke condiions, heir small size makes hem less able o face high up-fron coss.(p. 53). Our analysis will emphasize a mechanism of propagaion of he effecs of cloud compuing which depends on he endogenous marke srucures. Through business creaion, he adopion of cloud compuing is going o enhance compeiion in each secor and o increase producion and lower mark ups. This will have a posiive impac on consumpion so as o conribue o he recovery of he EU economy. Mos of all, par of hese effecs are going o be posiively relaed o he speed of adopion of he new echnology. Our resuls are only preliminary and fuure research should ry o improve he calibraion of a heoreical model or he esimaion of an empirical model. Moreover, alernaive mehodologies would be helpful in crosschecking he validiy of our resuls. Beside our quaniaive resuls, our main conribuion relies in he descripion of a mechanism hrough which cloud compuing is likely o creae a posiive effec on GDP, employmen and business creaion. We need o noice ha our approach neglecs oher posiive effecs exered by he inroducion of cloud compuing, mainly he creaion of new and mulilaeral nework effecs and he posiive exernaliies due o energy savings, 4 whose consideraion would be subjec o excessive uncerainy. Therefore, we can look a our esimaes of he impac of cloud compuing on he economy as conservaive esimaes. Secion V discusses policy implicaions. Since a large par of he posiive effecs of cloud compuing are posiively relaed o is speed of adopion, our invesigaion suggess ha policymakers should promoe as much as possible a rapid adopion. For insance, governmens could finance, up o a limi, he variable coss of compuing for all he firms ha decide o adop a cloud compuing soluion. These policies may be easily uned o opimize he process of adopion of he new echnology and o srenghen he propagaion of is benefis wihin he counry. In a conex as he European one, smaller counries would be able o obain larger gains from similar policies a leas in he iniial phase, because hey would easily arac foreign invesmens from he larger counries. In a period of increasing limis o oher forms 2009 / 2 Review of Business and Economics
4 82 Federico Ero Review of Business and Economics 2009 / 2 of fiscal compeiion, especially in he inegraed marke, a policy of subsidizaion of cloud compuing could generae subsanial capial flows oward smaller counries wih good general infrasrucures (hink of Mala or Luxembourg in he E.U.). The aricle draws he conclusions in Secion VI. The Appendix conains an advanced reamen of he analysis. II. Feaures and Implicaions of Cloud Compuing The erm cloud compuing refers o an Inerne-based echnology hrough which informaion is sored in servers and provided as a service (Sofware as a Service, or SaaS) and on-demand o cliens (from he clouds indeed). Is impac will be specacular on boh consumers and firms. On one side, consumers will be able o access all of heir documens and daa from any device (he personal lapop, he mobile phone, an Inerne Poin..), as hey already do for services, and o exploi impressive compuaional capabiliies. 5 On he oher side, firms will be able o ren compuing power (boh hardware and sofware) and sorage from a service provider and o pay on demand, as hey already do for oher inpus as energy and elecriciy: ha is why we alk of uiliy compuing. The former applicaion will affec our lifesyles, bu he laer will have a profound impac in erms of cos reducions on he sofware indusry. According o Armbrus e al. (2009) his impac will be similar o he one ha semiconducor foundries had on he hardware indusry. 6 Moreover, cloud compuing will also exer a fundamenal impac on he cos srucure of all he indusries using hardware and sofware, and herefore i will have an indirec bu crucial impac on heir marke srucures. In preparaion o his new scenario, many hardware and sofware companies are invesing o creae new plaforms able o arac cusomers on he clouds. Cloud plaforms provide services o creae applicaions in compeiion or in alernaive o on-premises plaforms, he radiional plaforms based on an operaing sysem as a foundaion, on a group of infrasrucure services and on a se of packaged and cusom applicaions. The crucial difference beween he wo plaforms is ha, while on-premises plaforms are designed o suppor consumer-scale or enerprise-scale applicaions, cloud plaforms can poenially suppor muliple users a a wider scale, namely a Inerne scale. Cloud compuing has been seen as a sep in he commodiizaion of IT invesmens (Carr, 2003), as he oucome of an evoluion oward a uiliy business model in which compuing capabiliies are provided as a service (Rappa, 2004), as he core elemen of he era of Web 2.0, in which Inerne is used as a sofware plaform (O Reilly, 2005), or simply as an applicaion of he generaiviy power of he Inerne (Zirain, 2007).
5 Economic Impac of Cloud Compuing on Business Creaion, Employmen and Oupu 83 The inroducion of cloud compuing is going o be gradual. Currenly we are only in a phase of preparaion wih a few pioneers offering services ha can be regarded as belonging o cloud compuing. Meanwhile, many large high-ech companies are building huge daa cenres loaded wih hundreds of housands servers o be made available for cusomer needs in he near fuure. 7 Amazon has been he firs mover in he field, providing access o half a million developers by way of Amazon Web Services (iniially developed for inernal purposes). Through his cloud compuing service, any small firm can sar a webbased business on is compuer sysem, add exra virual machines when needed and shu hem down when here is no demand: for his reason he uiliy is called Elasic Cloud Compuing. 8 For insance, Animoo, an applicaion ha produces videos from user-seleced phoos and music, has been a successful business of his kind. When Animoo was launched on he leading social nework Facebook, i was forced by exponenially increasing demand o bring he number of machines used on he Amazon Web Services from 50 o 3500 wihin hree days, somehing ha would have been impossible wihou relying on a cloud plaform. Google is also invesing huge funds in daa cenres. 9 Already nowadays Google provides word processing and spreadshee applicaions online, while sofware and daa are sored on he servers. Google App engine allows sofware developers o wrie applicaions ha can be run for free on Google s servers. Even he search engine of Google or is mapping service can offer cloud applicaion services: for insance, when Google Maps were launched, programmers easily found ou how o use heir maps wih oher informaion o provide new services for insance he locaion of houses from he renal and sales lisings of Craiglis. Microsof has sared laer bu wih huge invesmens in he creaion of new daa cenres. In he fall of 2008, he leading sofware company has inroduced a cloud plaform called Windows Azure, currenly available only in a preview version. Azure is able o provide a number of new echnologies: a Windows-based environmen in he cloud o sore daa in Microsof daa cenres and o run applicaions; an infrasrucure for boh on-premises and cloud applicaions (hrough.net Services); a cloud based daabase (hrough SQL Daa Services, which can be used from differen users and differen locaions); and an applicaion ool o access Live Services which allows o synchronize and consanly updae daa across sysems joined ino a mesh (for insance all he personal devices as he PC, he office s compuer, he mobile phone and so on). Moreover, Windows Azure provides a browser-accessible poral for cusomers: hese can creae a hosing accoun o run applicaions or a sorage accoun o sore daa in he cloud, and hey can be charged hrough subscripions, per-use fees or oher mehods. 0 Oher sofware and hardware companies have been acively invesing in cloud compuing (3Tera and Saleforce.com are paricularly acive). Social neworks have moved in he same direcion urning ino social plaforms for consumer based applicaions, wih Facebook in he fron road. Yahoo! is developing server farms as 2009 / 2 Review of Business and Economics
6 84 Federico Ero Review of Business and Economics 2009 / 2 well. Oracle has inroduced a cloud based version of is daabase program and has bough Sun Microsysems o prepare furher expansion in he field. The bale for he clouds beween hese companies is going o reshape he ICT marke srucure as PC disribuion did in he 80s. Bu according o he Economis (2008): cloud compuing is unlikely o bring abou quie such a dramaic shif. In essence, wha i does is ake he idea of disribued compuing a sep farher. Sill, i will add a couple of layers o he IT sack. One is made up of he cloud providers, such as Amazon and Google. The oher is sofware ha helps firms o urn heir IT infrasrucure ino heir own cloud, known as a virual operaing sysem for daa cenres Will his prospecive plaform war produce a dominan company in he mould of IBM or Microsof ha is able o exrac more han is fair share of he profis? Probably no, because i will be relaively easy o swich beween vendors... Nor is i likely ha one firm will manage o build a global cloud monopoly. Alhough here are imporan economies of scale in building a nework of daa cenres, he compuing needs of companies and consumers vary oo widely for one size o fi all. Mos imporan, he need of creaing nework effecs in he developmen of a cloud plaform will keep low he margins for a while and will maximize he speed of diffusion of cloud compuing beween firms a he global level. Therefore, in he long run, we expec a raher compeiive siuaion on he supply side of cloud compuing. In fron of hese rapid evoluion, i is crucial o undersand he economic impac of he inroducion of his general purpose echnology. For sure, he diffusion of cloud compuing is going o creae a solid and pervasive impac on he global economy. The firs and mos relevan benefi is associaed wih a generalized reducion of he fixed coss of enry and producion, in erms of shifing fixed capial expendiure in ICT ino operaive coss depending on he size of demand and producion. This conribues o reduce he barriers o enry especially for SMEs, as infrasrucure is owned by he provider, i does no need o be purchased for one-ime or infrequen inensive compuing asks, and i generaes quick scalabiliy and growh. The consequences on he endogenous srucure of he markes wih larges cos savings will be wide, wih enry of new SMEs, a reducion of he mark ups, and an increase in average and oal producion. Anoher imporan benefi is associaed wih he creaion of mulidimensional nework effecs due o he new possibiliies of produc creaion in he clouds, ha is beween companies exploiing in differen ways he poenialiies of cloud compuing hrough he same plaform or differen ones. This is relaed o anoher new possibiliy, he rapid adopion of changes: i is no uncommon, ha applicaions in he clouds are modified on a daily base (o accommodae new requiremens, or enable new economic venues), which is impossible wih on-premise soluions. I is imporan o noice ha he aggregae role of hese nework effecs can be relevan bu i is exremely difficul o measure. Finally, cloud compuing is going o inroduce he possibiliy of a) sharing resources (and coss) among a large pool of users, b) allowing for cenralizaion of
7 Economic Impac of Cloud Compuing on Business Creaion, Employmen and Oupu 85 infrasrucures in areas wih lower coss, and c) allowing for peak-load capaciy increases (generaing efficiency improvemens for sysems ha are ofen only 0-20% uilized). These feaures will lead o addiional savings in energy and o greaer environmenal susainabiliy, whose measure, however, is again subjec o large uncerainy. 2 A recen sudy of he Inernaional Daa Corporaion (2008) has examined he role of IT cloud services across five major produc segmens represening almos wo-hirds of oal enerprise IT spending (excluding PCs): business applicaions (SaaS), infrasrucure sofware, applicaion developmen & deploymen sofware, servers and sorage. Ou of he $ 383 billion ha firms have spen in 2008 for hese IT services only $ 6.2 billion (4%) could be classified as cloud services. In 202 he oal figure was expeced o be a $ 494 billion and he cloud par a $ 42 billion, which would correspond o 9% of cusomer spending, bu also o a large par of he growh in IT spending. The majoriy of cloud spending is and will remain allocaed o business applicaions, wih a relaive increase of invesmen in daa sorage. Even if he relaive size of IT cloud services may remain limied in he nex few years, i is desined o increase and o have a relevan macroeconomic impac, especially in erms of creaion of new SMEs and of employmen. In imes of global crisis, his could be an imporan conribuion o promoe he recovery and o foser growh. Cloud plaforms and new daa cenres are creaing a new level of infrasrucures ha global developers can exploi, especially SMEs ha are so common in Europe. This will open new invesmen and business opporuniies currenly blocked by he need of massive up-fron invesmens. The new plaforms will enable differen business models, including pay-as-you-go subscripions for compuing, sorage, and/or IT managemen funcions, which will allow small firms o scale up or down o mee he demand needs. As he Economis (2008) claims, he inerne disruped he music business; Google disruped he media; cloud-based companies could become disrupers in oher inefficien indusries. The macroeconomic impac of he diffusion of his new general purpose echnology may be quie large, as i happened for he inroducion of he Inerne in he 90s. 3 III. Evaluaing he Economic Impac of Cloud Compuing: Mehodology and Daa Our approach in he evaluaion of he impac of he diffusion of a new general purpose echnology as cloud compuing is based on macroeconomic heory and macroeconomic daa. We emphasize he effec ha his innovaion has on he cos srucure of he firms invesing in ICT and consequenly he incenives o creae 2009 / 2 Review of Business and Economics
8 86 Federico Ero Review of Business and Economics 2009 / 2 and expand new business, on he marke srucure and on he level of compeiion in heir secors, and ulimaely he induced effecs for aggregae producion, employmen and oher macroeconomic variables. Our mehodology is based on a dynamic sochasic general equilibrium (DSGE) calibraed model augmened wih endogenous marke srucures in line wih recen developmens in he macroeconomic lieraure (see Ero, 2009, for a survey). This model is perurbed wih a realisic srucural change o he cos srucure, wih he purpose o sudy he shor and long erm reacions of he economy. Therefore, our mehodology is based on a solid heoreical framework and provides resuls ha can be easily replicaed by economiss. However, i has some limiaions ha we need o poin ou. Firs of all, while he mehodology is useful o esimae he aggregae impac of a shock on he macroeconomy, i is less so a he microeconomic level. Second, he experimen we presen is highly speculaive because he naure of he cos shif (due o he inroducion and diffusion of cloud compuing) can only be conjecured (i will depend on many fuure macroeconomic and policy facors), and also because we are in a momen of high macroeconomic uncerainy. For his reason, we will focus on he ne expeced impac of cloud compuing on he economy, meaning he expeced addiional impac above and beyond he cyclical behavior of he macroeconomic variables (associaed wih any oher moivaions). Moreover, we will presen esimaes for differen scenarios of slow and rapid adopion of he new echnology ha should cover he range of possible oucomes wih a good approximaion. Third and las, our approach neglecs some of he posiive effecs exered by he inroducion of cloud compuing, mainly he creaion of new nework effecs and he posiive exernaliies due o energy savings. Their consideraion would be subjec o excessive uncerainy, bu because of his we can look a our esimaes of he impac of cloud compuing on he economy as conservaive esimaes. In he res of his secion we briefly describe our baseline model, our daa sources and he experimen associaed wih he inroducion and diffusion of cloud compuing. Furher analyical deails on he model are provided in he Appendix. The resuls of he numerical exercise are presened in he nex secion. For our esimaes, we develop a DSGE model calibraed on he EU economy in he radiional way. 4 The infinie horizon model accouns for he dynamics of oupu, consumpion, working hours (as an endogenous facor of producion), and accumulaion of ICT capial (as a reproducible facor of producion) 5 in a sandard way. The producion of final goods in differen secors derives from Cobb-Douglas funcions wih consan reurns o scale in labor and in he socks of hardware and sofware, which joinly deermine he sock of ICT capial. The aggregae producion funcion reads as: Y AL H S ( ) ( )( ) ()
9 Economic Impac of Cloud Compuing on Business Creaion, Employmen and Oupu 87 where Y is oupu, A is oal facor produciviy, which is assumed exogenous in our analysis (is growh would no change our qualiaive resuls), L are oal labor hours, H is he aggregae sock of hardware and S is he aggregae sock of sofware, while (,] 0is he labor share, whose realisic value is around 2/3, and [,] 0represens he elasiciy of ICT capial o he sock of hardware, wih a realisic value beween 3/4 and 9/0 (hardware represens he main share of ICT spending compared o sofware). Noice ha physical capial of a differen naure could be added wihou loss of generaliy, bu he deails of his exension are beyond our curren ineress. Consumer preferences are logarihmic in he consumpion index C wih a consan elasiciy of labor supply: / L U E log C, 0 / 0 where L is employmen of he represenaive agen and is he discoun facor. These assumpions are he radiional ones and deliver sandard consumpion and labor supply funcions. We assume for simpliciy homogenous goods wihin each secor. However, since our focus is on he impac of shocks on he marke srucure and hrough ha on he economy, we augmened he model wih endogenous marke srucures as recenly inroduced by Ghironi and Meliz (2005), Bilbiie e al. (2007, 2008,a,b), Ero (2007a), Ero and Colciago (2007) and Colciago and Ero (2008). 6 New firms can be creaed wih an iniial fixed invesmen: since his is expressed in erms of he final good, i implicily requires he use of labor, hardware and sofware. Once acive in a secor, each firm compees wih a number of rivals in he choice of he producion level. In he Courno-Nash equilibrium, he aggregae producion is inversely relaed o he equilibrium mark up and price and posiively relaed o he number of firms (which affecs mark ups in urn). The number of compeiors in each secor N, and herefore he concenraion of each secor is endogenized hrough he process of business creaion, which operaes by equaing in each period he sock marke value of a firm o he enry cos. Enry and ICT capial accumulaion occur gradually over ime because of he business creaion/ desrucion process and of he invesmen/depreciaion process. Joinly, hey deermine he marke srucure and is dynamic evoluion, in erms of how many firms are acive in each secor, how much each firm produces and which one is he equilibrium mark up in each secor and in each period. 7 We assume homogenous goods in each secor, which implies he following mark up: N N (2) (3) 2009 / 2 Review of Business and Economics
10 88 Federico Ero Review of Business and Economics 2009 / 2 which is decreasing in he number of firms. The laer follows he equaion of moion: N ( ) N N E, where N E, is he number of new firms and is he rae of business desrucion. The real value of a firm V is he presen discouned value of is fuure expeced profis, or in recursive form: V ( ) E V ( N ) r where ( N ) is he profi funcion depending on he number of firms and r is he ineres rae. Endogenous enry of new firms equaes his value o he fixed cos of enry in each period. Noice ha a reducion of he enry cos due o a reducion in he fixed invesmen in ICT capial promoes enry, and wih i compeiion, which reduces he mark up and increases aggregae producion, employmen and consumpion. This mechanism is a he core of our analysis. Our experimen is focused on Europe. Therefore, all our daa derive from official EU saisics (Eurosa), mainly for he number of firms, which is basically equivalen o he number of small and medium size enerprises (SMEs), employmen and gross domesic produc. In paricular, we used daa for mos of he EU member counries and Norway for which we had complee daa. Moreover, we focused on a few aggregae secors for which we have deailed and comparable EU saisics: Manufacuring Wholesale and reail rade (WRT) Hoels and resaurans (HR) Transpor sorage and Communicaion (TSC) Real esae rening and business aciviies (REB) These aggregae secors cover he majoriy of firms in erms of number (more han 7 million firms) and a large par of employmen for he European counries (more han 3 million workers), and include all he secors where he effecs emphasized in our analysis are relevan, namely manufacuring and service secors, where he use of ICT capial and he role of enry coss and compeiion effecs are more relevan. We ignored oher aggregae privae secors (as elecriciy, gas and waer supply) and he public secor, where we believe ha our mechanisms are eiher weaker or absen, or secors where comparable daa were no available (as par of he financial secor). All he resuls are based on he numerical simulaion of a calibraed model (see he Appendix for deails). However, counry specific heerogeneiy and secorial differences were aken in consideraion on he basis of saisics on (4)
11 Economic Impac of Cloud Compuing on Business Creaion, Employmen and Oupu 89 he labor marke and he enry/compeiive condiions a he level of EU counries and heir aggregae secors. A main aspec of our empirical exercise concerns he naure of he shock affecing he economy. The inroducion of cloud compuing allows firms from all secors o reduce fixed coss in ICT and urn par of hem ino variable coss. In our analysis we focus on he reducion in he fixed cos associaed wih he inroducion of cloud compuing. The increase in he marginal cos of producion is endogenous and depends on he echnological choices of he firms, which decide how much hardware and sofware (or, generally speaking, ICT capial) o use according o heir producion necessiies, and on he endogenous renal rae of ICT capial. In general, his depends on he marke srucure of he hardware and sofware secors, which we will assume o be perfecly compeiive. One may augmen he analysis wih an increase in he uniary cos of producion (a reducion of A) associaed wih he cloud compuing echnology, bu here we will ignore his aspec and assume ha his echnology reduces he fixed coss of producion only. 8 The reducion in fixed ICT spending is gradually exploied by he new and exising firms, and he speed of adopion of he new possibiliy remains an unknown variable for us. I will depend on a number of macroeconomic and microeconomic facors and on policymaking as well. Moreover, i will be characerized by imporan sraegic complemenariies: adopion for a single firm is crucial if many firms are expeced o adop i, bu i is no if hey are no, which means ha muliple equilibria could emerge (wih slow or rapid adopion and wih limied or deep adopion). For his reason we will adop a reduced form model of adopion based on a gradual reducion of he coss of enry and we will differeniae he simulaions for cases of slow and rapid adopion o evaluae he range of resuls. The general specificaion of he process of cos reducion ha we assume in he simulaions works as follows. Define as he fixed cos of enry in a secor, which consrains business creaion a ime. The fixed cos of enry a ime is given by: ~ ( ) where (,) 0. The above formula for he fuure cos is a weighed average of he curren cos and a long run cos ~. The dynamic pah of depends on wo parameers. The firs parameer, ~, is he seady sae level of he fixed cos, which is going o be lower han he iniial one. The second parameer,, represens he speed of diffusion or adopion of he new echnology. Changes in hese wo parameers allow us o depic scenarios wih slow or rapid adopion and o paramerize he absolue size of he shock. I urns ou ha changes of he speed of adopion wihin a realisic range, from fas adopion ( ) o slow adopion ( ), have a limied impac on he aggregae effecs (because of he forward looking naure of he firms and agens behavior), while he long run size of he cos reducion is crucial / 2 Review of Business and Economics
12 90 Federico Ero Review of Business and Economics 2009 / 2 A key facor for he impac of cloud compuing is he size of fixed cos savings. The business lieraure emphasizes large savings. Dubey and Wagle (2007) conjecure large reducions in he cos of ownership for ypical business services. 9 Carr (2003) suggess ha abou half of capial expendiure of modern firms is ICT relaed. While his maybe rue in a number of secors and for advanced companies, we prefer o adop a more conservaive assumpion for our macroeconomic invesigaion. One of he bes reviews of he sae of ICT in Europe is provided by he e- Business Wach of he European Commission. The 2006 e-business Repor provides a comprehensive survey of ICT adopion and spending, showing ha 5% of oal coss is spen in ICT. Figure shows he variabiliy of his figure across a few key secors on which European Commission (2007) has focused is analysis, while Figure 2 shows he lile variabiliy beween firms of differen sizes. Since only par of he oal cos corresponds o fixed coss of producion, he average ICT budge mus be more han 5% of he oal fixed coss of producion. Of course, only a par of ICT spending represens fixed coss, and only a par of i will be cu even afer he adopion of cloud compuing in alernaive o a fully inernal soluion. For his reason, we decided o adop a conservaive assumpion and o consider a range of reducion in he fixed coss in he long run beween % and 5%. Our main purpose is o show ha even such a limied echnological change Figure. Average share of he ICT budge as % of oal coss (by secor). Figure 2. Average share of he ICT budge as % of oal coss.
13 Economic Impac of Cloud Compuing on Business Creaion, Employmen and Oupu 9 due o cloud compuing will deliver subsanial effecs a he macroeconomic level. Needless o say, larger shocks will be associaed wih wider effecs. IV. Evaluaing he Economic Impac of Cloud Compuing: Resuls for EU Counries In his secion we repor he resuls of our experimen on he inroducion and diffusion of cloud compuing in he European economy. We focus on he impac on GDP, business creaion and employmen in he shor erm, ha is afer one year, and in he medium erm, ha is afer 5 years. Two scenarios are considered: slow adopion corresponds o he case of a % percen slow reducion ( )inhe fixed coss of enry and rapid adopion o a 5% rapid reducion ( ) in he fixed coss furher deails are in he Appendix. Table. Addiional oupu variaion in Europe. Shor Term Medium Term Slow Fas Slow Fas Oupu The conribuion o GDP growh can be hardly differeniaed beween counries and secors, herefore we simply summarize our average esimaes o he European counries. As shown in Table, he range is beween 0.05% in he shor run under slow adopion and 0.3% in he medium run under fas adopion of cloud compuing. These resuls are derived in erms of percenage variaion from a seady sae value in a saionary model, bu since growh is expeced o be abou zero in 200, hese percenage deviaions can be approximaely inerpreed in erms of addiional conribuion o he growh rae. Given he conservaive assumpions on he size of he shock, hese are remarkable conribuions o GDP growh, and hey have a direc counerpar in he effecs on employmen. One should ake he esimaes on he impac on employmen wih care. Even if we ook in consideraion counry specific facors relaed o he labor marke condiions, our basic simulaions emphasize he impac in erms of hours worked, whose ranslaion in new jobs depends on a number of insiuional and srucural feaures of he labor markes and heir counry-specific regulaion. Keeping his in mind, we found ha he inroducion of cloud compuing could creae, on average, abou a million addiional jobs in Europe. Abou wo hirds of job creaion is expeced o occur in he six larges counries (Unied Kingdom, Germany, France, Po / 2 Review of Business and Economics
14 92 Federico Ero Review of Business and Economics 2009 / 2 land, Ialy and Spain), bu each counry could enjoy a emporary increase in he work force. Of course his increase is going o vanish over ime because he srucural feaures of he economy lead employmen oward is naural level, which is affeced only in a small measure from he reducion of he fixed coss. However, he shor run impac can be quie srong and, in a period of crisis as he one depiced for he forhcoming years, i can conribue o limi he increase of he unemploymen rae in a subsanial way. Our esimaes of he reducion of he unemploymen rae in he European counries due o he inroducion of cloud compuing are around 0.5% in he shor run and 0.2-3% in he medium run. 20 Before enering in furher deails, i is worhwhile o skech he mechanism emphasized in our model. The gradual inroducion of cloud compuing reduces he fixed coss needed o ener in each secor and increases he incenives o ener. This increases curren and fuure compeiion in each marke and ends o reduce he mark ups and increase producion. The associaed increase in labor demand induces an upward pressure on wages ha induces workers o work more (or new agens o ener in he labor force). The curren and expeced increase in oupu affecs consumpion/savings behavior. In he shor run, he demand of new business creaion requires and increase in savings, which may induce a emporary negaive impac on consumpion. However, in he medium and long run he posiive impac on oupu leads o an increase in consumpion oward a higher seady sae level. Of course, a faser adopion exers a large impac on business creaion and herefore on oupu and employmen as well. Given his overview of he main resuls in erms of GDP and employmen, i is now ime o presen our full resuls in erms of esimaes of new business creaion for each counry and each one of he aggregae secors we ook in consideraion: manufacuring, wholesale and reail rade, hoels and resaurans, ranspor sorage and communicaion and finally, real esae rening and business aciviies. Figures 3-4 provide our esimaes on he creaion of new SMEs. The larges impac is expeced o occur in he aggregae secors of wholesale and reail rade (plus 56 housand firms in he medium run under fas adopion) and of real esae and oher business aciviies (plus 44 housand new SMEs). While he reader can easily look a he resuls divided by aggregae secors in each counry, we will no commen on he emerging differences across secors because we do no wan o overemphasize he limied predicive power of our exercise a his level of deail and also because we did no find subsanial heerogeneiy in he resuls in erms of percenage conribuions. Neverheless, our empirical exercise shows a srong impac on he creaion of new SMEs, in he magniude of a few hundred of housand in he whole EU (again his is addiional o a normal siuaion wihou he inroducion of cloud compuing). Noice ha he effec is permanen and ends o increase over ime: he creaion of new SMEs is no going o vanish, bu i is going o remain over ime wih a permanen impac on he srucure of he economy. Moreover, he effec is deeper
15 Economic Impac of Cloud Compuing on Business Creaion, Employmen and Oupu 93 M WRT HR TSC RES Toal Counries Slow Fas Slow Fas Slow Fas Slow Fas Slow Fas Slow Fas Belgium Bulgaria na na Czech Republic Denmark Germany Esonia Ireland Greece Spain France Ialy Lavia Lihuania Hungary Neherlands Ausria Poland Porugal Romania Slovenia Slovakia Finland Sweden Unied Kingdom Norway EU 25 counies Creaion of SMEs firms in he shor erm Legenda: Slow sands for slow adopion of he new echnology, whie Fas sands for fas adopion Secors: M (manufacuring), WRT (Wholesale arid Reail Trade), HR (Hoels and Resaurans) TSC (Transpors Sorage and Communicaions), REB (Real Esae and Business Aciviies). Figure 3. The impac of cloud compuing on business creaion in he shor run / 2 Review of Business and Economics
16 94 Federico Ero Review of Business and Economics 2009 / 2 M WRT HR TSC REB Toal Counries Slow Fas Slow Fas Slow Fas Slow Fas Slow Fas Slow Fas Belgium Bulgaria na na Czech Republic Denmark Germany Esonia Ireland Greece Spain France Ialy Lavia Lihuania Hungary Neherlands Ausria Poland Porugal Romania Slovenia Slovakia Finland Sweden Unied Kingdom Norway EU 25 Counries Creaion of SMEs firms in he shor erm Legenda: Slow sands for slow adopion of he new echnology, whie Fas sands for fas adopion Secors: M (manufacuring), WRT (Wholesale arid Reail Trade), HR (Hoels and Resaurans) TSC (Transpors Sorage and Communicaions), REB (Real Esae and Business Aciviies). Figure 4. The impac of cloud compuing on business creaion in he medium run.
17 Economic Impac of Cloud Compuing on Business Creaion, Employmen and Oupu 95 in counries where he diffusion of SMEs is paricularly srong or where ICT adopion has been generally rapid. In absolue erms, Ialy is esimaed o have he larges impac in erms of new business (wih 8 housand new SMEs in he medium run under fas adopion), followed by Spain (plus 55 housand), France (48 housand), Germany (39 housand), Unied Kingdom (35 housand) and Poland (32 housand). We have also examined he impac on employmen in each counry wih a disincion beween aggregae secors. In absolue erms, he larges impac is expeced for he manufacuring secor and also for he secor under he label hoels and resaurans, and his is no surprising given he high number of workers in hese aggregae secors. Overall, he impac on employmen is more limied compared o he impac on business creaion for a simple reason. One of he main advanages of cloud compuing is an induced change in he marke srucure of many secors, wih he creaion of more firms and an increase in he level of compeiiveness (associaed wih a reducion in prices as well). This change in he marke srucure associaed wih larger efficiency induces a re-allocaion of jobs ha does no increase by much he number of workers. Anyway, also in his case we are alking abou a few hundreds of addiional workers (or a corresponding lower number of unemployed agens) a he European level. Noice ha our simulaion emphasizes a slow reducion of he ne impac on employmen in he medium run compared o he shor run: his is normal because he absolue impac on he labor force ends o vanish in he long run. According o our esimaes, Unied Kingdom is going o exhibi he larger impac in erms of new workers (wih 240 housand new workers in he shor run under fas adopion), followed by Germany (60 housand), France (00 housand), Poland (94 housand), Ialy (76 housand) and Spain (69 housand). Overall, he resuls counry by counry are in par affeced by differences in labor marke condiions, ha end o affec he abiliy of he economy o reac o a posiive change hrough job creaion, and in he regulaory framework and in he compeiive condiions of he goods markes, ha creae he condiions for quick business creaion. Wihou overemphasizing our quaniaive resuls, we noice ha, in percenage erms, counries as Unied Kingdom, Finland, Poland, Czech Republic and Slovakia are expeced o perform beer han average, while counries as Hungary, Ialy, Ausria and Spain are expeced o perform more poorly han average. V. Policy Implicaions Cloud compuing is a new general purpose Inerne-based echnology hrough which informaion is sored in servers and provided as a service and on-demand o cliens. In his paper we analyzed he economic impac of is gradual inroducion 2009 / 2 Review of Business and Economics
18 96 Federico Ero Review of Business and Economics 2009 / 2 in he nex years and we emphasized is role in fosering business creaion and compeiion in all he markes hanks o he reducion of he fixed coss of enry in ICT capial. Our calculaions for European counries show a significaive medium erm impac of he diffusion of cloud compuing on enry of SMEs, employmen and growh. Saring from conservaive assumpions on he cos reducion process, he analysis shows ha he diffusion of cloud compuing will provide (in a medium erm range of five years) a posiive impulse o he annual growh rae, conribuing o creae abou a million new jobs hrough he developmen of a few hundred housand new SMEs in he counries under invesigaion. Par of he posiive effecs of cloud compuing are going o be posiively relaed o he speed of adopion of he new echnology. For his reason, our invesigaion suggess ha policymakers should promoe as much as possible a rapid adopion of cloud compuing. Concree possibiliies include fiscal incenives and a specific promoion of cloud compuing in paricular dynamic secors. For insance, governmens could finance, up o a limi, he variable coss of compuing for all he (domesic and foreign) firms ha decide o adop a cloud compuing soluion. 2 Moreover, hey could inroduce business friendly rules for he reamen and movemen of daa beween heir counry and foreign counries. These policies may be sudied in such a way o opimize he process of adopion of he new echnology and o srenghen he propagaion of is benefis wihin he counry. Moreover, in a conex as he European one, smaller counries would be able o obain larger gains from similar policies a leas in he iniial phase, because hey would easily arac foreign invesmens from larger counries. In a period of increasing limis o oher forms of fiscal compeiion (especially in he inegraed marke), a policy of subsidizaion of cloud compuing (wihou discriminaion across firms of differen member counries) could generae subsanial capial flows oward smaller counries wih good general infrasrucures. For insance, early adopion of hese policies by small E.U. counries as Luxembourg or Mala could arac large invesmens by foreign firms (looking for subsidies o adop he new business model) and creae wide effecs in erms of oupu growh and job creaion in hese counries. Of course, inernaional policy compeiion for he subsidizaion of cloud compuing soluions would generae posiive spillovers across counries, and some coordinaion a he E.U. level would be welcome. VI. Conclusion The main conribuion of his paper is an empirical applicaion of he endogenous marke srucures approach o a real world phenomenon, he inroducion of a general purpose echnology which is no direcly augmening oal facor produciviy,
19 Economic Impac of Cloud Compuing on Business Creaion, Employmen and Oupu 97 bu ha is reducing he fixed coss of enry and producion in he economy. The heoreical par of he paper develops an exension of he framework inroduced by Colciago and Ero (2008) and oher relaed works of he recen lieraure on endogenous enry and macroeconomics, assuming ha enry coss are in unis of he final good, and hey follow a deerminisic process depending on he speed of adjusmen oward a long run level. Furher work may generalize his heoreical framework and he cos reducion process. The empirical par of he paper ranslaes he impac of a cos reducing innovaion on he European economy. Fuure work may also improve he calibraion of he heoreical model and he ranslaion of he numerical simulaion in empirical esimaes wih addiional informaion on labor and good markes a he naional and secorial level. Moreover, alernaive mehodologies for he esimae of he impac of cloud compuing would be welcome o crosscheck he validiy of our resuls. Beside our quaniaive resuls, our main conribuion remains in he descripion of he compeiive mechanism hrough which cloud compuing is likely o creae a posiive effec on business creaion, GDP and employmen. Appendix The srucure of he economy simulaed in he experimen is an exension of he model wih endogenous marke srucures developed by Ero and Colciago (2007) and Colciago and Ero (2008) wih invesmen in erms of final good. Consider a represenaive consumer wih uiliy: / L U E log C, 0 / 0 where (,) 0 is he discoun facor, L is labor and C is a consumpion index. Ideally, his index would aggregae consumpion bundles of muliple secors wih a consan elasiciy of subsiuion, bu our assumpions allow us o normalize he number of secors o one wihou loss of generaliy. 22 The represenaive secor is characerized by differen firms i 2,,, N producing he same good in differen varieies. A general formulaion for he consumpion index C of he represenaive secor is based on Dixi and Sigliz (977): C N i C () i where C ()is i he producion of firm i of his secor, and (, ) is he elasiciy of subsiuion beween he goods produced in he secor. The disincion beween dif- (5) (6) 2009 / 2 Review of Business and Economics
20 98 Federico Ero Review of Business and Economics 2009 / 2 feren secors and differen goods wihin a secor allows one o separae limied subsiuabiliy a he aggregaed level, and high subsiuabiliy a he disaggregaed level. When we are in he case of homogenous goods wihin secors, which will be our baseline case. In each period, oal consumpion C is allocaed across he available goods according o he inverse demand funcions: C() i E p( i) N i,2,..., N C () j j where E is oal expendiure. Each good is produced in each period a a consan marginal cos common o all firms. Under he differen forms of compeiion, we obain a symmeric equilibrium mark up depending in general on he kind of compeiion (in prices or quaniies), on he degree of subsiuabiliy beween goods and on he number of firms N. A Courno equilibrium generaes he equilibrium mark up: N ( )( N ) The markup remains posiive for any degree of subsiuabiliy, since even in he case of homogenous goods, we have: lim N N This allow us o consider he effec of sraegic ineracions in an oherwise sandard seup wih perfec subsiue goods (which has been radiionally sudied only under perfec compeiion in he neoclassical radiion). The individual profis in unis of consumpion can be expressed as: ( N ) C ( N) N 2 which are clearly decreasing in he number of firms and in he subsiuabiliy beween goods. In wha follows, we will adop he paramerizaion of Colciago and Ero (2008) and focus on he case of Courno compeiion wih homogenous goods, herefore we will se so ha N /( N ) and ( N ) C / N 2. The resuls can be easily generalized o imperfec subsiuabiliy and oher forms of compeiion as Berrand or Sackelberg compeiion. 23 (7) (8) (9)