EUROPEAN ECONOMY EUROPEAN COMMISSION DIRECTORATE-GENERAL FOR ECONOMIC AND FINANCIAL AFFAIRS

Size: px
Start display at page:

Download "EUROPEAN ECONOMY EUROPEAN COMMISSION DIRECTORATE-GENERAL FOR ECONOMIC AND FINANCIAL AFFAIRS"

Transcription

1 EUROEAN ECONOMY EUROEAN COMMISSION DIRECORAE-GENERA FOR ECONOMIC AND FINANCIA AFFAIRS ECONOMIC AERS ISSN hp://europa.eu.n/comm/economy_fnance N January 5 An esmaed new keynesan dynamc sochasc general equlbrum model of he Euro area by Marco Rao**, Werner Röger*, Jan n Veld* and Rccardo Grard** *Drecorae-General for Economc and Fnancal Affars ** Jon Research Cenre

2 Economc apers are wren by he Saff of he Drecorae-General for Economc and Fnancal Affars, or by expers workng n assocaon wh hem. he "apers" are nended o ncrease awareness of he echncal work beng done by he saff and o seek commens and suggesons for furher analyses. Vews expressed represen exclusvely he posons of he auhor and do no necessarly correspond o hose of he European Commsson. Commens and enqures should be addressed o he: European Commsson Drecorae-General for Economc and Fnancal Affars ublcaons BU - -/8 B - 49 Brussels, Belgum ECFIN/7878/4-EN ISBN KC-AI-4--EN-C European Communes, 5

3 able of conens Inroducon... 3 he DSGE Model Esmaon Mehodology Solvng he model wh lnear approxmaons Maxmum lkelhood esmaon and nference Bayesan esmaon and nference Implemenaon: MCMC (Meropols-Hasngs Model comparson... 4 Esmaon ror dsrbuons arameer esmaes and shocks denfed VAR comparson Whch srucural shocks drve he euro economy?... Esmaed mpulse responses of srucural shocks Conclusons References... 4 Annex:... 4 A. Resuls from poseror maxmzaon... 4 A. oseror smulaon:... 4

4 Inroducon In recen years a new consensus has emerged n macroeconomcs n general and n model buldng n parcular, he so called New Keynesan aradgm (NKM. hs paradgm s well esablshed as can be seen from he promnen reamen n recen exbooks (see, for example Obsfeld/Rogoff and leraure surveys (se for example Clarda, Gal Gerler. In a sense he NKM paradgm combnes elemens from he RBC leraure wh more radonal Keynesan deas. radonal Keynesan models suffered from underdeveloped mcrofoundaons and a lack of long erm facors nfluencng he economy whch made hem subec o he ucas crque. I also dd no have a coheren heorecal explanaon for he sluggsh behavour of prces assumed. On he oher hand, RBC modellers bul her models up from he acons of opmsng economc agens whose choces are made whn specfed consrans. he New Classcal vew of RBC modellers saw busness cycles as largely he resul of shocks o producvy and preferences, and downurns as merely he opmal adusmen of he economy o such dsurbances. NKM models correc he RBC models by nroducng frcons n goods, labour and fnancal markes n order o provde a beer f wh acual daa, bu a he same me res o model he frcons explcly as consrans faced by households and frms. hs allows combnng opmal behavour wh rgdes n a way whch avods he ucas crque. he QUES model has been se up n he spr of a NKM model, wh a srong emphass on heorecal conssency of he behavoural equaons. However a he me when QUES II was nroduced he esmaon echnology for DSGE models was no suffcenly developed o allow for rgorous esmaon and esng of hese models. arge pars of hese models needed o be calbraed. Followng recen developmens n Bayesan esmaon echnques (see, e.g., Geweke 999 and Schorfhede, has become possble o esmae hese ype of models. Smes and Wouers (3 have been he frs o esmae such a model for he Euro area. hey followed Crsano, Echenbaum, and Evans ( and desgned a DSGE model for he Euro area feaurng prce and wage sckness, paral ndexaon of prces and salares, exernal capal formaon, varable capal ulsaon rae and sochasc shocks o each srucural equaon of he resulng model. Smes and Wouers (3 show ha he curren generaon of New-Keynesan DSGE models s suffcenly rch o capure he me-seres properes of he daa, as long as a suffcen number of srucural shocks s consdered. In parcular, s able o mach he degree of emprcal perssence found n he euro area daa for nflaon and wages que well. hs paper apples Bayesan esmaon echnques o a me seres daa se of he euro area and presens esmaes of a DSGE model. he purpose of hs paper s no o esmae he curren verson of he QUES model drecly wh hese mehods bu raher o esmae a prooype new generaon New-Keynesan DSGE model. hs model can hen serve as a benchmark for an esmaon of a QUES specfcaon. In fac n some dmensons he QUES model may need o be adused o come closer o a DSGE model. One of he common feaures beween he QUES II model and he esmaed New-Keynesan DSGE model presened n hs paper are ha boh, n he long run, closely resemble he sandard neoclasscal growh model. All behavoural relaons are derved from dynamc opmsaon problems of households and frms, wh opmsaon subec o echnologcal consrans, budge consrans and/or nsuonal consrans, ofen capured as adusmen - 3 -

5 coss. hs leads o a descrpon of economc behavour ha s a mxure of backward and forward lookng behavour. he man dfferences wh he QUES II model are n he specfcaon of consumpon. Whle consumpon n he QUES model s based on a permanen ncome model for fnely-lved households, as popularsed by Blanchard (984, n hs DSGE model n conras, consumpon s derved from neremporal opmsaon for nfnely lved households, lke n all oher curren DSGE models. However, consumpon s modelled more backward lookng by allowng for hab perssence (.e. consumpon decsons oday depend parally on he prevous paern of consumpon. On he oher hand, QUES allowed for lqudy consrans, a feaure sll mssng n he consumpon framework here. he dfferences n he nvesmen specfcaon are only of mnor mporance, wh a sronger emphass on adusmen coss here. he modellng of he labour marke s subsanally dfferen. Here, lke n oher DSGE models, hs s based on a neoclasscal labour supply wh monopoly power for workers. One of he dsngushng feaures of he QUES model s he labour marke specfcaon derved from a heorecal search model based on he work by ssardes (99. he esmaed model as presened here s sll ncomplee snce reas he Euro area as a closed economy. he closed economy seng was chosen because we frs waned o concenrae on he man aggregaes consumpon and nvesmen as well as on prces and wages and her neracons. However, addng a rade secor would be among our frs prores for furher exensons of hs model. he model wll hen nclude a more explc modellng of rade frcons whn he framework of convex adusmen coss, whch would dsngush from he QUES model, where rade s modelled hrough an ad-hoc specfcaon of adusmen lags n quanes and prces. An mporan reason for esmang a DSGE model was also o be able o compare esmaon resuls wh he exsng leraure and o make sure ha he esmaon yelds resuls whch are conssen wh he resuls obaned wh smlar specfcaons and smlar daases. he man goals of hs exercse are: Demonsrae ha models derved from economc heory can f Euro area daa, provded one allows for suffcen nsuonal resrcons. We compare he predcve performance of he esmaed DSGE model wh ha of a VAR model esmaed over he same euro area daa se. (We nend o conduc a smlar exercse for he US economy o see wheher nsuonal consrans play he same role here. Idenfy he man srucural shocks hng he Euro area economy n a heorecally conssen way. An advanage of an explc srucural model s he fac ha resduals can be gven a srucural nerpreaon,.e. we can denfy shocks whch orgnae from consumpon, echnology, labour supply, labour demand, nvesmen and fscal and moneary polcy. hs may be of added value n ryng o undersand he naure of he curren economc suaon. For example, he model denfes a declnng rend n governmen spendng, whch s reversed n recen years, a declne n prce mark-ups, reflecng ncreased compeve pressures, a rend ncrease n oal facor producvy n he 98s, followed by a declne n he lae 99s and a rend ncrease n labour supply, reflecng a declnng NAIRU. 3 rovde he ypcal response of he economy o he ndvdual shocks n he form of mpulse responses, lke n VAR sudes, wh he addonal benef ha confdence nervals can be provded o show he uncerany surroundng hese responses

6 he oulne of he paper s as follows. Secon presens he model. In secon 3 dscusses he esmaon mehodology. Snce esmaon of hese models s non sandard a farly comprehensve explanaon wll be provded. Secon 4 presens he esmaon resuls. In order o provde a more nuve undersandng for he qualy of he f of hs model, a comparson wh a smple VAR model s gven. Secon 5 presens and nerpres he srucural shocks denfed by he model esmaes and secon descrbes he dynamc adusmen of he euro area economy o srucural shocks. he DSGE Model Households: he household secor decdes abou consumpon and asse accumulaon (ncludng fxed capal. Each household supples a specfc varey of labour n a monopolscally compeve fashon,.e. he household secor ses he wage gven he demand curve for labour. When makng decsons he household also faces adusmen coss for changng wages. hese adusmen coss are borne by he household (see budge consran. he household maxmses a uly funcon subec o a budge consran. he agrangan of hs maxmsaon problem s gven by ( Max U = = C λ β β C ( U ( C + V ( + Z( M / M + B + R V + R W γ w + w w M B V + AX he household maxmses a uly funcon over consumpon, lesure and real money balances. Followng he recen leraure we allow for hab perssence n consumpon. hs s an mporan modfcaon w. r.. he curren consumpon specfcaon n QUES whch was based enrely on a pure lfe cycle model. he curren verson allows for lagged adusmen of consumpon and we choose a logarhmc specfcaon C (a U C ε log( C habc ( = C where ε denoes a sochasc preference shock for consumpon n perod. hs specfcaon yelds he followng expresson for he margnal uly of consumpon (b C U C, = ε. ( C habc - 5 -

7 he consumpon ndex s self an aggregae over dfferen goods whch are mperfec subsues. he preferences of households are expressed by a CES uly funcon τ τ (c C C τ = d wh τ where τ measures he nverse of he me varyng elascy of demand of households for consumpon goods of ype. he erm τ s gven by (d τ = τ + τ ( Y Ypo + ε τ where τ ε s a auocorrelaed shock o he demand elascy. For labour supply we use a CES uly funcon (e where gven by ω κ s a possbly auocorrelaed labour supply shock. he margnal uly of lesure s κ V ( = ε ( wh κ >, ε κ (f V = ε ω(., he household decdes abou consumpon, asse accumulaon and he supply of labour (or more correcly abou wages and real money holdngs. he frs order condons of he household (FOCs wh respec o consumpon and fnancal wealh are gven by he followng equaons: U C (3a => U C, λ = C U (3b => λ + λ+ β = B R U M ζ (3c => Y R = M + he labour supply decson s slghly more complex, snce s assumed ha workers have a ceran marke power n he labour marke, because hey offer servces, whch are mperfec subsues o servces offered by oher workers. ha means aggregae labour demand of frms s a compose of labour suppled by ndvdual workers. oal employmen n producon s characersed by a CES funcon Wh an neres rae rule as specfed below, an opmaly condon for money would only deermne he desred money holdngs of he household secor whou any furher consequence for he res of he economy. For ha reason any furher dscusson on money demand s dropped here. - -

8 (4a θ θ θ θ d = wh θ > where he parameer θ deermnes he degree of subsuably beween labour suppled by ndvdual households. Correspondng o he CES aggregaor here exss a wage ndex (4b W θ θ = W hs echnology yelds a labour demand equaon as perceved by household (4c W = W θ In a monopolsc labour marke he elascy of subsuon beween dfferen ypes of labour s mporan for deermnng he mark-up of wages over he equlbrum wage. hs elascy s defned by (4d W W = θ W θ W = θ W. Now he wage seng rule can be derved akng dervaves of he agrangan w.r.. wages. Usng symmery: W = W and neglecng second order erms allows us o wre (5a U W W ( θ W W => V = λ γ wπ + λ+ βγ wπ +, θ where rule (5b W π s he growh rae of nomnal wages. hs can be reformulaed as a wage seng π w = C w W w ( mup + + γ w ω( R π wh κ w mup = θ where wage nflaon s deermned by he gap beween he reservaon wage and he real wage adused for a wage mark up. he forward lookng naure of wage seng s refleced by he forward wage nflaon erm. hs formulaon generalses he neoclasscal labour supply model along wo dmensons. Frs, by nroducng convex wage adusmen coss ( γ w >, workers wan o smooh wage adusmens, akng no accoun curren and fuure expeced labour marke condons. Second, because workers offer servces whch are mperfec subsues o servces offered by oher workers, hey can demand wages whch are above her reservaon wage. he reservaon wage s he margnal value of lesure, dvded by he Noce n he lmng case of perfec subsuably ( lm θ, he mark up approaches zero

9 margnal uly of consumpon. ha means for a gven uly of lesure he reservaon wage ncreases wh a declne n he margnal uly of consumpon ha an addonal un of labour can buy. In esmang he wage rule wo furher generalsaons have been nroduced. Some search heorec generalsaons of he neoclasscal wage rule sugges rules where wages are ndexed o boh he reservaon wage and he margnal value produc of labour wh wegh bg reflecng he barganng srengh of workers (see, for example, Sh e al. (999. In order o allow for backward lookng behavour s assumed ha only a fracon sfw of workers form raonal expecaons of fuure wages, whle he remanng workers follow a smple rule of humb where expecaons are deermned by pas nflaon. hese wo modfcaons lead o he followng wage equaon (5c w π κ w w ( sfwπ + ( sfw C Y w W = ( bg + bgα η ( mup + + π γ w ω( R sfw Frms: here are N frms ndexed by. Because goods produced by ndvdual frms are mperfec subsues, frms are monopolscally compeve n he goods marke and face a demand funcon for goods gven by τ ( Y = s ( C + G + I Oupu s produced wh a Cobb Douglas producon funcon (7 Y = ( ucap K ( U α α wh capal and labour as npus. Frms can also decde abou he degree of capacy U ulsaon he level of echnology s subec o random echnology shocks ( ε and follows he auoregressve process U U U (8 log( U = ρ log( U + ( ρ log( U + ε. he obecve of he frm s o maxmse he presen dscouned value of s cash flows. Dynamc consderaons ener he problem of he frm because frm faces quadrac coss of changng capal, employmen and prces. Fnally frms mus also choose he opmal level of capacy ulsaon. (9-8 -

10 [ ] [ ] [ ] CA K K I K d q U K ucap Y d ucap ad ad I K ad adc I W Y d Max V ( ( ( ( (, ( ( (. = = δ η α α where = + + = l l l l rp r d s he dscoun facor, whch consss of he shor erm neres rae and a rsk premum (rp. he rsk premum can be subec o random shocks and generaed by he followng auoregressve process ( rp rp rp rp rp rp ε ρ ρ + + = ( For adusmen coss we choose he followng convex funconal forms ( (, ( /, ( ( I I K K W I K I I K ad wh ad W ad + + = = = + = γ ε γ π π γ γ ε * ( * ( ( ucap ucap a ucap ucap a ucap ad CA + = he frm deermnes labour npu, he capal sock, capacy ulsaon and prces opmally n each perod gven he echnologcal and admnsrave consrans as well as demand condons. he frs order condons are gven by: (a ( ( ( W W R Y V ε γ γ η α + = + => + (b ( ( = + + => + I I K q I R I K I I V γ ε γ (c ( ( (( ( + = + + => q rp r q ucap a ucap a a a a K Y K V δ η α (d ucap K Y ucap a a a ucap V η α ( ( (( = + => - 9 -

11 V τ (e => η = τ + τ Y Ypo + ε + γ [ βπ π ] ( ( Y Frms equae he margnal produc of labour, ne of adusmen coss, o wage coss. Wage coss nclude a sochasc wage cos shock. hs should be seen as shocks o admnsrave burdens relaed o curren employmen. As can be seen from he lef hand sde of equaon (a, he convex par of he adusmen cos funcon penalses n cos erms acceleraons and deceleraons of changes n employmen. Equaons (b-d only deermne he opmal capal sock and opmal capacy ulsaon. he frm equaes he margnal produc of capal o he renal prce of capal, adused for capal coss. he frm also equaes he margnal produc of capal servces (K*ucap o he margnal cos of capacy ulsaon. Equaon (e defnes he mark up facor as a funcon of he elascy of subsuon and changes n nflaon. We follow Smes and Wouers and allow for addonal backward lookng elemens by assumng ha a fracon (-sfp of frms keep prces fxed a he - level. hs leads o he followng specfcaon: τ (e η ( τ + τ ( Y Ypo + ε + γ [ β ( sfp * π + ( sfp π π ] = + + sfp Governmen secor: he governmen secor and fscal polcy s reaed n a raher rudmenary fashon. he share of governmen purchases (3a G / Y = gs + ε G flucuaes sysemacally wh he busness cycle accordng o he followng rule Y (3b gs = Y Ypo where Y g g ( measures he degree of auomac sablsaon of governmen expendure. Dscreonary fscal acon s characersed by he varable whch s allowed o be auocorrelaed process. Implcly s assumed ha governmen expendure s fnanced, by lump sum axes. G ε Cenral bank polcy rule (neres rae rule: Moneary polcy s modelled va he followng aylor rule, whch allows for some smoohness of he neres rae response o he nflaon and oupu gap (4a nom = lag * nom π M ( π π + ( lag * ( Ex. R + π + Y M ( Y Ypo ( Y + π M Ypo ( π π + + ε M Y M ( Y Ypo + - -

12 M he erm ε capures random dscreonary shocks o moneary polcy and π s a me varyng nflaon arge, specfed as follows π π π (4b π = ρ π + ( ρ π + ε. π ε s an..d. shock o he nflaon arge. I s assumed ha boh fscal and moneary auhores base her polces on a concep of poenal oupu whch s a smooh funcon of pas oupu (5 Ypo = ρ ypoypo + ( ρ ypo Y 3 Esmaon Mehodology We presen he frs aemp o apply a Bayesan esmaon approach o brng he model drecly o he daa. hs approach has been dscussed by many auhors n he leraure n he las few years (e.g. Schorfhede, ubk and Schorfhede, 3, Smes and Wouers, 3. Schemacally, he mehod consss of he followng seps: he non-lnear DSGE model s solved va a lnear approxmaon: a lnear raonal expecaon sysem s obaned ha mus be obanng a sandard lnear model n sae space form; he sae-space approxmaon of he orgnal non-lnear model allows he denfcaon of a lkelhood funcon (va e.g. Kalman recursons and a subsequen nference based on (maxmum lkelhood esmaon, ec.; usually heorecal model mply few, well defned shocks; unforunaely hs ofen mples sngulares n he deermnaon of he lkelhood (n he Kalman fler he number of shocks mus a leas be as large as he number of observables, mplyng he nroducon of addonal srucural shocks and/or measuremen errors; lkelhood-based nference presens a seres of ssues: specfcally he lack of denfcaon (global: mulple maxma; local: over-parameersaon,.e. he maxmum s gven by a complex muldmensonal combnaon/neracon srucure raher hen by a sngle pon n he parameer space; he Bayesan analyss s performed: pror dsrbuons for model parameers have o be defned, represenng he pror belefs of he analys on her plausble values, whch, n combnaon wh he lkelhood funcon, allows o oban he poseror dsrbuon; he Bayesan nference needs he use of sochasc smulaons, specfcally Markov Chan Mone Carlo (MCMC echnques, allowng o oban samples from he poseror on pdf of he model parameers and subsequenly o make an nference n whch he parameer uncerany and he shape of he lkelhood are aken no accoun; he model s fnally compared o an emprcal model; n he leraure hs s usually a VAR model. - -

13 From he compuaonal pon of vew, he lnear approxmaon and he soluon of he obaned RE can be done auomacally usng he DYNARE program (Jullard, 99, 3, whch apples he generalsed Schur decomposon soluon mehod (Klen,. DYNARE s a sofware for he smulaon of DSGE models, freely avalable and oally open source. resenly, an esmaon module s mplemened on DYNARE, o nclude he mos recen developmens n Bayesan esmaon macro-economc models n an exremely effcen and easy way. DYNARE s also exremely flexble, and allows o easly ncorporang problem specfc mehodologcal ssues or cusomsaons. 3. Solvng he model wh lnear approxmaons e a model be defned and frs order condons denfed. hs can be expressed as: E f y, y, y, ε ; θ = ( E E { ( + } { ε } = { ε ε '} = Σ where y s vecor of endogenous varables, ε s he vecor of exogenous sochasc shocks, θ s he vecor of parameers and E s he expecaon operaor. he non-lnear model s solved va a lnear approxmaon around he deermnsc seady sae y such ha f ( y, y, y,; θ =. A lnear raonal expecaon (RE sysem s obaned, wh forward lookng componens + (7 A E yˆ + + A yˆ + A yˆ + Bε =, where yˆ = y y he sysem s solved for he reduced form sae equaon n s predeermned varables (Blanchard and Kahn, 98; generalsed Schur form, Klen,. An observaon equaon s also added o lnk he observed varables (8 y yˆ = M ( y( θ + yˆ + η = G( θ yˆ E( η η ' = V ( θ E( ε ε ' = Q( θ + H ( θ ε y o he predeermned ones, obanng: where η s he measuremen error, f any. he sysem marces G, H, V and Q and he seady sae vecor y (θ are funcons of he vecor of srucural parameers θ of he orgnal model. Vecor θ ncludes he nose parameers Σ. he sae space represenaon (8 allows use of Kalman flerng for he compuaon of he log-lkelhood, and a subsequen nference based on (maxmum lkelhood esmaon, ec.. In he orgnal model specfcaon (, well-defned (and relavely few shocks are usually presen. If he number of shocks s smaller han he number of observed varables, sngulares n he Kalman fler wll be presen,.e. he probably dsrbuon of he observables p( Y θ (he lkelhood can be degenerae. hs mples he nroducon of addonal shocks unl he sysem becomes non-sngular, ncludng eher measuremen errors η (as e.g. n Ireland, 4, who also models he measuremen error as a VAR( process or addonal srucural shocks n he sae equaon (as e.g. n Smes and Wouers, 3. Rgorously, n such cases, as clearly saed by Schorfhede (, he evaluaon approach here appled wll lead o an assessmen of he modfed model raher han he orgnal one. In such cases, he parameer vecor θ wll be augmened for he addonal nose erms and, f any, also for he VAR coeffcens n he measuremen errors as n Ireland (4. In hs paper we follow he Smes and Wouers approach and nroduce a suffcen number of srucural shocks n he sae equaons. - -

14 3. Maxmum lkelhood esmaon and nference he sae space form of he lnear approxmaon obaned can be subec o a classcal lkelhood analyss. he sysem can be fed o a Kalman fler and he lkelhood funcon p( Y θ can be compued n a sandard way (Kalman, 9, Kalman and Bucy, 9 or, n he case of non-saonary models, wh exac nal Kalman flerng (Koopman, 997. hs allows performng maxmum lkelhood esmaon, usng a numercal opmsaon roune, obanng: ϑˆ = arg max p( Y θ M k ϑ R where Y s he nformaon se gven by a seres of observaons Y wh =,...,. Gven he parcular naure of he sae space model fed o he M opmsaon, n whch all marces of coeffcens are funcons of he srucural parameers θ : A = A(θ, B = B(θ, C = C(θ, D = D(θ, he algorhm for opmsaon also mples ha, for each parameer ral, he whole procedure prevously presened (log-lnearsaon, soluon of he RE model, mplemenaon of he Kalman fler and compuaon of he lkelhood value mus be repeaed. he lkelhood-based nference can presen furher problems, specfcally regardng he lack of denfcaon. global: he lkelhood funcon may have mulple maxma; local: he lkelhood funcon does no have a unque maxmum n he neghbourhoods of some θ *. o beer explan he laer case, n such suaons here exs many combnaons of model parameers ha provde he same lkelhood value,.e. he maxmum s no gven by a sngle pon n he parameer space, bu by a complex muldmensonal srucure. In some dscplnes hs s referred as over-parameersaon,.e. here are many parameer values or model specfcaons ha are compable wh he same emprcal evdence. hs also means ha he number of parameers o esmae s oo large. A rval remedy o can be o fx some parameers (n some cases mos of hem! and maxmse wh respec o he remanng ones, even f hs soluon can be regarded as arbrary. hs also mples ha he maxmsaon s compuaonally more dffcul han for sandard sae space models. Moreover, also he represenaon and summary of resuls s dffcul. For example, M nference s ofen accompaned by asympoc heory o provde confdence nervals, samplng dsrbuon of he M esmaes, ec. Bu wha f he maxmum s no unque? Moreover, he lack of denfcaon ofen leads o ll-condoned covarance marces (.e. he Hessan marx s ofen nearly sngular. akng no accoun parameer unceranes (or n oher words he shape of he lkelhood funcon can be herefore a very dffcul problem. All hese ssues call for a Bayesan approach, n whch pror nformaon s combned wh he lkelhood and whch s especally useful n problems wh many parameers and few observaons. he use of prors s very naural, snce economss have srong belefs abou plausble values of srucural parameers; parameers have a well-defned nerpreaon and have a bounded doman. From he compuaonal pon of vew, he use of a pror makes he opmsaon algorhm more sable, namely because curvaure s nroduced n he obecve funcon. Maxmsaon of he poseror s hence (relavely easer han he maxmsaon of he lkelhood. Moreover, parameer unceranes and he shape of he lkelhood (or beer of he poseror dsrbuon are reaed naurally by applyng sochasc smulaon approaches

15 he prce o pay s ha Bayesan mehods are exremely compuaonally nensve. We descrbe he Bayesan roue n he nex secon. 3.3 Bayesan esmaon and nference Roughly speakng, Bayesan nference s based on pullng he maxmum lkelhood esmaes oward values hough as plausble a pror. From he Bayes heorem, he poseror dsrbuon s obaned as (9 p( θ Y ( ( = p Y θ p θ p( Y θ p( θ p( Y θ p( θ, ( θ Y p( θ summarsng all our nformaon (pror and lkelhood abou he parameer vecor θ. he lkelhood funcon s any funcon ( θ Y p( Y θ. Knowng he poseror dsrbuon, allows mplemenng he Bayesan nference. In general, he obecve of Bayesan nference can be expressed as E[ g( θ Y ] where g (θ s a funcon of neres (a forecas, he vecor of model parameers self, ec. and E[ g( θ Y g( θ p ( θ Y dθ ] = g( θ p( θ Y dθ = = * p ( θ Y dθ * * g( θ ( θ Y ( θ Y p( θ dθ p( θ dθ where p ( θ Y p( θ Y p( θ ( θ Y s any poseror densy kernel for θ. For example, for a quadrac loss funcon, he pon esmae of model parameers s gven by he poseror mean: ˆ θ = θp( θ Y dθ. he problem n Bayesan nference s ha he negrals nvolved have almos never an analycal soluon and need a numercal approach, specfcally hrough sochasc smulaon. he key sraegy s o generae draws of θ from he poseror dsrbuon dscussed n he nex secon Implemenaon: MCMC (Meropols-Hasngs p( θ Y. hs s he key concep of Mone Carlo smulaon s as follows. Assume a vecor of random varables θ wh a on pdf π (θ. If we can draw an..d. sample θ, θ,..., θ n from π (θ, we can approxmae he negrals by dscree sums: n ( g = / n g( θ E( g( θ = g( θ π ( θ dθ almos surely as n. = If he varance σ of g (θ s fne, hen ( n( g( θ E( g( θ ~ N(, σ provdes an esmaon error. he generc dsrbuon π (θ can be he poseror dsrbuon p( θ Y and hence Mone Carlo smulaon can be appled o solve he Bayesan nference problem. he Mone Carlo approxmaons can be hen used o compue predcons, mpulse response funcons, ec

16 he almos surely n equaon ( above means ha convergence s subec o some regulary condons of he funcon g (θ ; specfcally absolue convergence of he negral mus be sasfed, see Geweke (999. he requred sample from he poseror dsrbuon s a mulvarae sample. hs s no an easy problem and has been he subec of a huge amoun of leraure o fnd echnques for hs samplng problem: from accepance samplng, mporance samplng, o Markov Chan Mone Carlo approaches (Gbbs sampler and Meropols-Hasngs algorhm. he laer approach s probably he bes suable for he problem a hand and s he one whch s appled n all he recen leraure on Bayesan analyss of DSGE models. e us frs defne an m-saes Markov process x. We denoe he possble saes of x by S = { s,..., s m } and defne he ranson probables p pm = M O M pm pmm where s he probably of movng from sae o sae. e w( = [ w (,..., w ( ] an p m vecor of probables of x beng n sae n perod, hen he correspondng probables for perod + are w ( + = w( he Markov chan has an equlbrum dsrbuon f here exs a dsrbuon π such ha π = π ; A Markov chan s reversble f probably of s he same as : π p = π p. A chan ha s reversble has an equlbrum dsrbuon and o sample from he equlbrum dsrbuon one can sar he chan from any w( unl seles down o he equlbrum dsrbuon. A Markov chan s no d, snce he sample s serally correlaed. he dea of he Meropols algorhm s o consruc he ranson marx from an easy ranson marx Q (e.g. correspondng o a mulvarae normal dsrbuon, such ha has he desred equlbrum dsrbuon π (.e. he poseror dsrbuon. hs because we are no able o draw from he poseror dsrbuon (correspondng o π n our case, bu we are able o draw sample form a normal dsrbuon (correspondng o usng Q. Of course, only he Q ranson s no suffcen o assure convergence o π, so we have o add an addonal rule for he ranson o one sae o anoher. Suppose a eraon we are n sae and based on Q we draw a proposed sae acceped (or a probably. We defne a probably α ha he proposed sae s ha he new sae s reeced and we say n. o defne s α he probably α, we use he obecve dsrbuon π as follows: α = mn[, π / π ] and he resulng chan s reversble and has equlbrum dsrbuon π. In our specfc problem, he Meropols Hasngs algorhm s mplemened as follows: s s m - 5 -

17 Condonal on daa Y and a se of parameer values θ, he Kalman Fler s used o evaluae he log-poseror densy up o a consan: p( θ ( θ Y p( θ ; ~ Wh a numercal opmsaon roune he mode θ of he poseror densy can be esmaed, and he nverse Hessan Σ ~ a he mode compued; 3 Implemen he Random Walk Meropols Algorhm: a Draw a canddae parameer vecor ϑ from a umpng dsrbuon J s ( ϑ θ, wh ( s ~ J s ~ N( θ, c Σ [he Q ranson defned above ( s b he ump from θ s acceped ( θ = ϑ wh probably reeced ( θ ( s = θ ( s ( ϑ Y r = ( θ Y oherwse, wh p( ϑ ( s ( s (s p( θ (s Y ( s α s, s=mn(r, and he seres of draws { θ } s serally correlaed (no d bu, afer a burn n perod, converges o he desred poseror dsrbuon (e.g., draw, samples and reec he frs,. he speed of convergence s a crcal ssue of MCMC mehods. here no general rule of creron ha can assure ha he chan has converged. here are a number of nformal echnques o assess convergence, such as: n s plo / n s g( θ as a funcon of n s ; s= ( s sar he Markov-Chn a over-dspersed (.e. exreme values of θ and check wheher dfferen runs of he chan sele o he same dsrbuon; more general mehods, whch combne n a more rgorous way he wo above emprcal deas, such as he poenal scale reducon facor (SRF and s mulvarae exenson (Brooks and Gelman, 998; mplemened n DYNARE. Roughly speakng, hs es ams a verfyng ha he samples obaned wh a number of parallel chans are drawn from he same dsrbuon. When converged, he chan sasfes a weak low of large numbers,.e. he approxmaons ( and ( apply for he Markov chan, whch can hen be sued for he Bayesan nference. 3.4 Model comparson In he Bayesan framework, models are compared and ranked accordng o he negraed lkelhood (or margnal daa densy. Havng a se of models =,,M, he poseror wegh of he -h model s ( w ( Y = p ( Y θ p( θ dθ Θ and, f he models have equal pror probables, he poseror probably on model s w / w. As usual, he compuaon of hs negral s unfeasble analycally n mos cases, bu can be esmaed usng a sample from he poseror dsrbuon. Specfcally, he margnal daa densy of he DSGE model s here approxmaed wh Geweke's (999 modfed harmonc mean esmaor (mplemened n DYNARE. In he presen repor we make a prelmnary comparson wh a VAR( model, usng RMSE s. Recenly, Sms (3 provded a general dscusson abou pfalls of Bayesan model comparson mehods, hghlghng several ways hey end o msbehave. In hs vew, here s no pon n showng a prelmnary Bayesan comparson, comparng, e.g., he - -

18 margnal daa densy obaned here wh he daa densy of a VAR( where he prors are defned wh a ranng se. As dscussed by Sms, such knd of comparson could be oally arbrary and meanngless. A full Bayesan comparson s beng mplemened, ryng o carefully address he ssues rased by Sms. 4 Esmaon 4. ror dsrbuons Srucural shocks. Afer n nally seng prors o nv-gamma, he radonal prors of sandard errors n Bayesan analyss, we preferred o se up fla prors n a relavely large nerval of values, reflecng more clearly our pror gnorance abou possble values of shocks. Above all n vew of a complee Bayesan comparson wh oher models (VARs, hs assumpon mgh be revsed by consderng, e.g., a ranng se. hs because oo large a pror range mgh unduly penalse he presen model, by gvng oo low wegh o he lkelhood (a oally unnformave pror n he range [-nf, nf] would gve a unformly zero wegh o any lkelhood value, mplyng he reecon of any model; see Sms, 3, for a full dscusson on hese maers. δ Concernng he shock o me-varyng δ (, we se a much smaller range, snce we do no ε wan δ o absorb whaever s mssed by he res of he model, bu we us allow he mnmum shock necessary o reconsruc he deprecaon pah. able.a rors srucural shocks Frms: F shock Deprecaon shock Dsrb. Mn Max U ε unform.e-. δ ε unform.e-. rp Rsk premum shock ε unform.e-. τ Mark-up shock ε unform.e-. W Wage cos shock ε unform.e-. Households: C Consumpon reference shock ε unform.e-. abour supply shock ε unform.e-. olcy: G Governmen expendure shock ε unform.e-. π Inflaon arge shock ε unform.e-. M Ineres rae shock ε unform.e

19 able.b rors shock perssence arameer Dsrbuon Mean S. dev. Suppor Frms: U F ρ bea.9.4 [ ] δ Deprecaon ρ bea.9.4 [ ] rp Rsk premum ρ bea.5. [ ] W Wage coss ρ bea.5. [ ] Households: abour supply olcy: Governmen expendure ρ bea.5. [ ] G ρ bea.9.4 [ ] π Inflaon arge ρ bea.5. [ ] Model parameers. he model parameers o be esmaed have a srucural economc nerpreaon and are herefore resrced o le n ceran nervals dcaed by economc heory or mpled by long run consrans. he followng ranges have been chosen for he ndvdual coeffcens: able rors model parameers arameer Dsrbuon Mean S. dev. Suppor Frms: Deprecaon rae δ bea.5.5 [.] Capacy ulsaon a bea.5.8 [.] Adusmen cos, capal γ K bea 5 5 [ 3] Adusmen cos, nv. γ I bea 3 [ ] Adusmen cos, labour γ bea 5 5 [ 3] Adusmen cos, prce γ bea 5 5 [ 3] Adusmen cos, wage γ bea 5 5 [ 3] W Mark-up, cyclcal τ bea -..3 [-. ] Share of fwd lookng prce seers sfp bea..5 [.5 ] Households: Hab perssence hab bea..5 [.9] abour supply elas. κ gamma.5.4 [ Inf] abour supply cons ω gamma..5 [ Inf] Barganng srengh bg bea [.75] Wage mark-up θ gamma.8 [ Inf] Share of fwd lookng wage seers sfw bea.8. [.5 ] - 8 -

20 olcy: Y Fscal response o ygap G bea.4 [- ] Ineres rae smoohng lag bea.8. [ ] π Ineres rae response, M bea..9 [.4] Y Ineres rae response, M bea..45 [.] π Ineres rae response, M bea.5.3 [.5 ] Y Ineres rae response, M bea.3. [.5] Smoohness, rend GD ρ YO bea.9.5 [.7 ] Noe: he followng parameers were fxed: oupu elascy of labour α =.594, dscoun facor β =.989, neres elascy money demand ζ = -.4, Mark-up level τ =.; he remanng parameers are deermned by seady sae consrans. a = ( r ss + r p + δ ( γ Iδ + s parameer of capacy ulsaon α α A = ss K ss echnology consan. r = ( τ ( α δ / I /( γ δ + δ r Rsk premum. ss I ss We denfed bea or gamma pror dsrbuons for model parameers 3. he pror Y specfcaon of sfp and M requred a parcular aenon, whereby we had o gve lower wegh o values ha were preferred by he lkelhood, bu ha mpled unreasonable dynamcal behavour n he mpulse responses. So, we se asymmerc dsrbuons ha Y prvleged he lower par of range for sfp and hgher par for M. hs mpled only a slghly worse f, bu a much beer model behavour n erms of heorecal consderaons. hs knd of approach s legmae n a Bayesan framework, and dsngushes from a plan consraned opmsaon, whch would be consdered much more arbrary. he fac ha we gve a smaller (bu non-zero! pror probably o some poron of he parameer ranges, always gves he possbly o he lkelhood o overrde hs assumpon, f he daa srongly suppors hypoheses abou such values ha were unlkely a pror. Moreover, hs approach does no rule ou possble msspecfcaon or he reecon of he presen model wh respec o compeng ones. In he laer case, he negraed lkelhood of he presen model would be penalsed wh respec o a compeng one ha provded more agreemen beween pror assumpons and lkelhood shape. he plos of he pror dsrbuons are gven below. he model was esmaed usng he followng egh seres as observaons: Y, I, C, K,, π, W, nom. 3 lease noe ha he suppor of bea dsrbuons mgh be larger han he ranges specfed n Secon, bu means and he sandard devaons are se n such a manner ha pror probably s larger han zero only n he accepable range

21 Fgure ror dsrbuons σ(ε C σ(ε δ σ(ε τ σ(ε G σ(ε π e σ(ε.e-5.e-5.e-5 σ(ε M.5..5 σ(ε U.5..5 σ(ε rp.5..5 σ(ε W a.5.5 barg 8 4 δ Y G..4. γ K

22 Fgure (con d ror dsrbuons. γ I γ γ γ W.5 hab lag ρ π κ ω ρ U ρ δ ρ G ρ ρ rp ρ W sfp 3 sfw 8 τ 3 π M 4 Y M θ π M Y M ρ YO arameer esmaes and shocks denfed he poseror esmaon followed he mehodology of Secon. Frs he mode of he poseror s esmaed usng a non-lnear opmsaon roune (values repored n he Annex. - -

23 hen, a sample from he poseror dsrbuon s obaned wh he Meropols algorhm usng he nverse Hessan a he poseror mode as he covarance marx of he umpng dsrbuon. he scale coeffcen was se o.5, allowng a good accepaon rae (5%. We ran 4 parallel Markov chans. Snce he refnemen of he convergence ess proceeded slowly by ncreasng he lengh of he chans, we decded o updae he covarance marx of he umpng dsrbuon accordng o he las poron (3% of he chans based on he nverse Hessan. hs allowed us o oban good convergence ess of 4 new chans (of 4, runs each based on he updaed covarance marx. Fgure Convergence es Meropols MCMC Mulvarae SRF de(σ x whn chan beween chan x 4 hs Fgure shows he convergence ess for Meropols MCMC. Upper panel shows he mulvarae poenal scale reducon facor, whch should be near o a convergence. ower panel shows he deermnan of he beween chans and whn chans covarance marces of he Mone Carlo sample. Afer dscardng he nal 7% of runs, we could proceed o he Bayesan nference. he followng fgures show he esmaed margnal poseror dsrbuons (black lnes, compared o prors (grey lnes and he pon esmae of he mulvarae mode (vercal dashed lnes. I s neresng o noe ha for some parameers he maxmum of he margnal dsrbuon s shfed wh respec o he mode of he mulvarae dsrbuon (n parcular γ K and γ I. hs mples ha such a local maxmum s n a very narrow regon wh almos zero mass, relaed o very specfc parameer combnaons. o gve an dea of hs, s neresng o - -

24 noe ha he log-poseror a he mode s abou 38, whle he Markov chans evolve n a range [37 379].e. almos log-pons lower wh respec o he mode. In spe of hs que hgh dfference n level, even mposng a sarng pon very near o he mode, he evoluon of he Markov chan evolved smlarly o he ones shown here, mplyng ha such a local opmum s locaed n a regon so small o mply an almos zero probably for a chan o fall here. In he Annex we also repor he values of he poseror mean wh confdence bands for he esmaed parameers. he logarhm of he margnal lkelhood for hs model s abou 33. Fnally, Fgure 4 shows he -perod ahead predcons of he model for he man model varables, ncludng he deprecaon rae δ and governmen expendure G. Dashed lnes are observaons; connuous lnes are model predcons. On he whole, he model fs he daa remarkably well. One pon ha s parcularly noeworhy s ha he model over-predcs M nom n he las years (coupled wh loose. ε Fgure 3 ror and poseror dsrbuons sandard errors srucural shocks and parameers σ(ε C σ(ε δ σ(ε τ σ(ε G σ(ε π e4.e4 5.e σ(ε e-5.e-5 9.e-5 σ(ε M σ(ε U.. σ(ε rp 4.. σ(ε W a 4 3 barg 3 δ 5 5 Y G γ K

25 Fgure 3 (con d ror and poseror dsrbuons sandard errors srucural shocks and parameers..5 γ I γ.5..5 γ γ W 8 4 hab lag 3 ρ π 3 κ 3 ω ρ U ρ δ.5 ρ G 3 ρ ρ rp.8.9 ρ W sfp sfw τ π M Y M θ π M Y M ρ YO

26 Fgure 4 -sep ahead predcon. C. δ. G.4 I nom 8 K..3 π W / Y VAR comparson If we compare he RMSE s of he DSGE model compued a he poseror mean, wh he RMSE s of a VAR( model wh he same 8 observed seres : Y, I, C, K,, π, W, nom. able 3 RMSE comparson wh VAR RMSE's VAR model (pos. mean C e- 8.58e- I 5.558e e- nom.4e-.8e- K e e e-7 π 4.573e- 9.93e- W/.994e e-5 Y.94e-5.547e-5 RMSE s of he DSGE are hgher bu of he same order of magnude han he VAR, excep for K, where he VAR performs much worse (RMSE s fory mes larger

27 5 Whch srucural shocks drve he euro economy? One of he maor advanages of he modellng approach used here s ha sysem esmaon of he model yelds, besdes he poseror dsrbuon of he model parameers, srucural shocks whch have an unambguous nerpreaon and whch can help us undersand he curren economc suaon. he srucural shocks denfed n he esmaon of hs model are shocks o households, frms, governmen and fscal polcy. Households are affeced by shocks o preferences for consumpon and labour supply. Frms are h by shocks o echnology, markups, adusmen coss for labour and capal and a rsk premum shock. Several aspecs of he esmaed shocks (fgure 5 and mpled unobserved varables (fgure are worh hghlghng. Demand Shocks he consumpon preference shock C ε appears slghly negave a he end of he esmaon sample. hs suggess lower preferences for consumers spendng and may be a reflecon of savngs uncerany concernng fuure pensons and ax lables. Noce, however, he sze of C he shock s no exraordnary large, gven he flucuaons of ε over he enre sample perod. Invesmen s h by wo auonomous shocks, a rsk premum shock and a shock o adusmen coss. he laer are unmporan and no furher consdered. he rsk premum rp shock ( ε does no show any parcular rend and appears o behave normally n recen years. hs suggess ha nvesmen flucuaons are explaned by fundamenals. he G smoohed auo-correlaed fscal polcy shock z( dsplays a urnaround n -. he fall n hs auo-correlaed shock shows clearly he fscal consoldaon perod sarng n he lae 98s, bu he declnng rend n governmen spendng s reversed n he early s. hs refues he vew ha fscal polcy has been overly resraned by he SG and been less counercyclcal over he las years. Supply shock Our Kalman fler esmaon allows o decompose observed oal facor producvy no a capacy ulsaon and a rue F componen (denoed as U. Accordng o hese esmaes U shows a rend ncrease n he second half of he 98s, a movemen along he rend n he 9s, bu a sharp declne n he lae 99s- early s. hus he fall n F n he recen pas s largely a srucural phenomenon and no he resul of a lack n demand, snce capacy ulsaon (UCA shows a normal cyclcal behavour n recen years. Noce, F s also one of he maor drvng forces of nvesmen and s herefore one of he fundamenal facors for he slowdown of nvesmen. rends and flucuaons n mark-ups are mporan measures for he supply poenal of he euro area economy. Accordng o hese esmaes he mark-up has declned on average snce he early 99s (η=-mark-up from around o 8%. hs could reflec ncreased compeve pressure due o goods marke reforms (nernal marke programme bu also ncreased pressure from global compeon. abour marke shocks W he model denfes a labour supply ( ε and a labour demand shock ( ε. he rend ncrease n labour supply (or n model erms a rend declne n he preference for lesure (Z( ε afer 995 s conssen wh he observaon of ncreased labour force parcpaon and a declnng NAIRU n he Euro area. Noce, however, he shock o labour supply has a pronounced cyclcal paern, whch suggess ha he smple wage rule used here does no ε - -

28 properly accoun for he dynamc adusmen of wages over he busness cycle. On he oher W hand, he upward rend of Z( ε reflecs he ncrease n non-wage labour coss over he sample. Ineresngly hs rend has sopped n he lae 99s, possbly reflecng a success of varous labour marke reform measures nended o reduce regulaory burdens for frms relaed o employmen. Moneary polcy shocks M he moneary polcy shock ε s negave for he years afer. hs would sugges a looser moneary sance han suggesed by he esmaed aylor rule. hs could be lnked o an underesmaon of he declne n he nflaon obecve π, whch shows a clear rend declne, bu may neverheless underesmae he acual declne n he moneary polcy s nflaon obecve. hs s one aspec ha may need furher aenon n fuure exensons of hs model. Fgure 5 Esmaed smoohed shocks a he poseror mean.5 ε C -4 x ε δ. ε τ x -3 z(ε G x -3 ε π z(ε x ε M ε U z(ε rp z(ε W Noe: f shocks are auo-correlaed, he auo-correlaed funcon s ploed (z(ε - 7 -

29 Fgure Unobserved varables.94 η.3 q. r U π ucap Ypo

30 Esmaed mpulse responses of srucural shocks In hs secon, we presen he esmaed mpulse responses of he nne srucural shocks n our model. he mpulse response are generaed on he bass of he reduced form represenaon of he model (polcy and reacon funcons - see annex. hs s parcularly smple, snce he reduced form s a lnear model (formally equvalen o a mulvarae ARMA. hey depc he responses for he endogenous varable followng a one-perod shock o each of he srucural shocks (whch are n mos cases auo-correlaed, each for a 5 year ( perods horzon. A full Bayesan IRF analyss s here presened, pckng, samples ou of he full Mone Carlo sample and compung IRF s for each of hem. Fnally, he mean pah (sold lnes and he confdence band (dashed lnes can be obaned, as shown n he Fgures. Fgure 7 Consumpon preference shock C ε. Y(% vs ε C I(% vs ε C.8 C(% vs ε C x -3 K(% vs ε C x -3 (% vs ε C M/(% vs ε C -. 8 x -5 π vs ε C -4 5 x -3 W/(% vs ε C nom(% vs ε C C Fgure 7 presens he esmaed effec of a consumpon preference shock (. hs shock s a combned shock affecng consumpon and lesure choce and has a drec mpac on consumpon and labour supply (hrough λ. he effec of hs preference shock s o rase consumpon by. percen, and employmen by.7 per cen (n he second perod afer he shock. he boos o demand rases nflaon and nomnal neres raes rse, bu he presence of adusmen coss lms he exen of he prce rse. he shock leads o crowdng ou of nvesmen and a decumulaon of capal. ε - 9 -

31 δ Fgure 8 Shock o deprecaon rae ( ε. Y(% vs ε δ I(% vs ε δ. C(% vs ε δ K(% vs ε δ (% vs ε δ -.5. M/(% vs ε δ x -4 π vs ε δ -.. W/(% vs ε δ nom(% vs ε δ δ(% vs ε δ Fgure 8 depcs a shock o he deprecaon rae of.5 per cen on mpac. hs mples an ncrease n he cos of capal of. percenage pons. hs shock s hghly perssen, n fac s has no dsappeared afer perods, due o he large esmaed auocorrelaon erm n he deprecaon rae. he ncrease n he cos of capal leads o a declne n nvesmen and he capal sock, lower real raes, hgher consumpon, hgher real wages and lower employmen. Noe however, he large confdence bands, whch sugges a large margn of uncerany surroundng hs ype of shock

32 τ Fgure 9 Shock o prce mark-up ( ε 5 x -3 Y(% vs ε τ. I(% vs ε τ 5 x -3 C(% vs ε τ x -3 K(% vs ε τ x -3 (% vs ε τ x -3 M/(% vs ε τ -5 3 x -4 π vs ε τ -.5. W/(% vs ε τ nom(% vs ε τ Fgure 9 shows ha, followng a posve prce mark-up shock, here s a ump ncrease n nflaon and nvesmen, oupu and consumpon declne. As oupu and nvesmen fall, labour demand s also lower and employmen falls. As he shock s ransory, he effecs fade away and have dsappeared afer 3 years

33 Fgure Shock o governmen spendng ( ε G.3 Y(% vs ε G I(% vs ε G C(% vs ε G K(% vs ε G (% vs ε G.3 M/(% vs ε G 8 x -4 π vs ε G. W/(% vs ε G nom(% vs ε G 4 3 Fgure shows he mpulse response o a governmen spendng shock. hs shock s a perssen shock wh an auocorrelaon of.98, bu he overall effec on GD s shor-lved. he mpac mulpler s small, no larger han., and he ncrease n spendng leads o crowdng-ou of consumpon, and n parcular of nvesmen, whch falls by.4 per cen a s peak. hs crowdng-ou of nvesmen leads o a reducon n he capal sock. he effec on oupu s emporary and over a longer horzon, oupu becomes negave. he small mulpler for a perssen fscal spendng shock s n lne wh resuls of he QUES model, n whch permanen fscal expansons have much smaller oupu effecs han ransory shocks. he ump n nflaon s surprsng and seems a odds wh emprcal regulares. I appears ha he peak nflaon response s already reached n he second quarer. hs hgh responsveness of nflaon s parly due o he hgh esmae of he forward nflaon erm sfp. Emprcal sudes employng VARs show generally posve oupu effecs of an ncrease n governmen spendng, bu ofen hese esmaes are surrounded by large confdence nervals. ero ( fnds for mos counres posve oupu effecs afer an ncrease n spendng, bu n he pos 98 sample, hese posve effecs are small and shor-lved. he effec on nvesmen s n accordance wh many emprcal sudes whch fnd he sronges crowdngou of spendng shocks for nvesmen (Alessna e al.,

34 Fgure Negave shock o employmen ε. Y(% vs ε. I(% vs ε. C(% vs ε. K(% vs ε (% vs ε -. M/(% ε x -4 π vs ε W/(% vs ε nom(% vs ε -.5. x -3 Z(ε vsε Fgure shows he mpulse response of a posve shock o lesure (a negave employmen shock. hs s a hghly auo-correlaed shock wh a hgh perssence of.99, as s clear from Z( ε. Employmen falls, and he reducon n labour supply has a negave mpac on nvesmen and oupu. As consumers ancpae lower ncomes, consumpon also falls

35 Fgure roducvy shock U ε.3 Y(% vs ε U.8 I(% vs ε U.3 C(% vs ε U.5 K(% ε U (% vs ε U. -. M/(% vs ε U x -4 π vs ε U W/(% vs ε U nom(% vs ε U -.. x -3 U vs ε U Fgure plos he esmaed effec of a producvy shock n he model. Oupu, nvesmen and consumpon rse, whle prces fall. However, he fall n nflaon s moderaed by he presence of prce adusmen coss. Employmen falls on mpac, bu over me labour supply ncreases as real wages are hgher. Moneary polcy reacs by lowerng nomnal neres raes, bu moneary polcy s no accommodang enough o preven prces fallng. I s mporan o noe ha hs producvy shock s no a permanen supply shock, bu fades away gradually

Graduate Macroeconomics 2 Problem set 5. - Solutions

Graduate Macroeconomics 2 Problem set 5. - Solutions Graduae Macroeconomcs 2 Problem se. - Soluons Queson 1 To answer hs queson we need he frms frs order condons and he equaon ha deermnes he number of frms n equlbrum. The frms frs order condons are: F K

More information

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany Herarchcal Markov Normal Mxure models wh Applcaons o Fnancal Asse Reurns Appendx: Proofs of Theorems and Condonal Poseror Dsrbuons John Geweke a and Gann Amsano b a Deparmens of Economcs and Sascs, Unversy

More information

V.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS

V.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS R&RATA # Vol.) 8, March FURTHER AALYSIS OF COFIDECE ITERVALS FOR LARGE CLIET/SERVER COMPUTER ETWORKS Vyacheslav Abramov School of Mahemacal Scences, Monash Unversy, Buldng 8, Level 4, Clayon Campus, Wellngon

More information

( ) () we define the interaction representation by the unitary transformation () = ()

( ) () we define the interaction representation by the unitary transformation () = () Hgher Order Perurbaon Theory Mchael Fowler 3/7/6 The neracon Represenaon Recall ha n he frs par of hs course sequence, we dscussed he chrödnger and Hesenberg represenaons of quanum mechancs here n he chrödnger

More information

Fall 2010 Graduate Course on Dynamic Learning

Fall 2010 Graduate Course on Dynamic Learning Fall 200 Graduae Course on Dynamc Learnng Chaper 4: Parcle Flers Sepember 27, 200 Byoung-Tak Zhang School of Compuer Scence and Engneerng & Cognve Scence and Bran Scence Programs Seoul aonal Unversy hp://b.snu.ac.kr/~bzhang/

More information

UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 2017 EXAMINATION

UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 2017 EXAMINATION INTERNATIONAL TRADE T. J. KEHOE UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 27 EXAMINATION Please answer wo of he hree quesons. You can consul class noes, workng papers, and arcles whle you are workng on he

More information

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4 CS434a/54a: Paern Recognon Prof. Olga Veksler Lecure 4 Oulne Normal Random Varable Properes Dscrmnan funcons Why Normal Random Varables? Analycally racable Works well when observaon comes form a corruped

More information

Lecture 6: Learning for Control (Generalised Linear Regression)

Lecture 6: Learning for Control (Generalised Linear Regression) Lecure 6: Learnng for Conrol (Generalsed Lnear Regresson) Conens: Lnear Mehods for Regresson Leas Squares, Gauss Markov heorem Recursve Leas Squares Lecure 6: RLSC - Prof. Sehu Vjayakumar Lnear Regresson

More information

Department of Economics University of Toronto

Department of Economics University of Toronto Deparmen of Economcs Unversy of Torono ECO408F M.A. Economercs Lecure Noes on Heeroskedascy Heeroskedascy o Ths lecure nvolves lookng a modfcaons we need o make o deal wh he regresson model when some of

More information

Dynamic Team Decision Theory. EECS 558 Project Shrutivandana Sharma and David Shuman December 10, 2005

Dynamic Team Decision Theory. EECS 558 Project Shrutivandana Sharma and David Shuman December 10, 2005 Dynamc Team Decson Theory EECS 558 Proec Shruvandana Sharma and Davd Shuman December 0, 005 Oulne Inroducon o Team Decson Theory Decomposon of he Dynamc Team Decson Problem Equvalence of Sac and Dynamc

More information

Variants of Pegasos. December 11, 2009

Variants of Pegasos. December 11, 2009 Inroducon Varans of Pegasos SooWoong Ryu bshboy@sanford.edu December, 009 Youngsoo Cho yc344@sanford.edu Developng a new SVM algorhm s ongong research opc. Among many exng SVM algorhms, we wll focus on

More information

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!") i+1,q - [(!

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!) i+1,q - [(! ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL The frs hng o es n wo-way ANOVA: Is here neracon? "No neracon" means: The man effecs model would f. Ths n urn means: In he neracon plo (wh A on he horzonal

More information

( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model

( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model BGC1: Survval and even hsory analyss Oslo, March-May 212 Monday May 7h and Tuesday May 8h The addve regresson model Ørnulf Borgan Deparmen of Mahemacs Unversy of Oslo Oulne of program: Recapulaon Counng

More information

Advanced Macroeconomics II: Exchange economy

Advanced Macroeconomics II: Exchange economy Advanced Macroeconomcs II: Exchange economy Krzyszof Makarsk 1 Smple deermnsc dynamc model. 1.1 Inroducon Inroducon Smple deermnsc dynamc model. Defnons of equlbrum: Arrow-Debreu Sequenal Recursve Equvalence

More information

Let s treat the problem of the response of a system to an applied external force. Again,

Let s treat the problem of the response of a system to an applied external force. Again, Page 33 QUANTUM LNEAR RESPONSE FUNCTON Le s rea he problem of he response of a sysem o an appled exernal force. Agan, H() H f () A H + V () Exernal agen acng on nernal varable Hamlonan for equlbrum sysem

More information

THE PREDICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS

THE PREDICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS THE PREICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS INTROUCTION The wo dmensonal paral dfferenal equaons of second order can be used for he smulaon of compeve envronmen n busness The arcle presens he

More information

January Examinations 2012

January Examinations 2012 Page of 5 EC79 January Examnaons No. of Pages: 5 No. of Quesons: 8 Subjec ECONOMICS (POSTGRADUATE) Tle of Paper EC79 QUANTITATIVE METHODS FOR BUSINESS AND FINANCE Tme Allowed Two Hours ( hours) Insrucons

More information

Robustness Experiments with Two Variance Components

Robustness Experiments with Two Variance Components Naonal Insue of Sandards and Technology (NIST) Informaon Technology Laboraory (ITL) Sascal Engneerng Dvson (SED) Robusness Expermens wh Two Varance Componens by Ana Ivelsse Avlés avles@ns.gov Conference

More information

Fall 2009 Social Sciences 7418 University of Wisconsin-Madison. Problem Set 2 Answers (4) (6) di = D (10)

Fall 2009 Social Sciences 7418 University of Wisconsin-Madison. Problem Set 2 Answers (4) (6) di = D (10) Publc Affars 974 Menze D. Chnn Fall 2009 Socal Scences 7418 Unversy of Wsconsn-Madson Problem Se 2 Answers Due n lecure on Thursday, November 12. " Box n" your answers o he algebrac quesons. 1. Consder

More information

Linear Response Theory: The connection between QFT and experiments

Linear Response Theory: The connection between QFT and experiments Phys540.nb 39 3 Lnear Response Theory: The connecon beween QFT and expermens 3.1. Basc conceps and deas Q: ow do we measure he conducvy of a meal? A: we frs nroduce a weak elecrc feld E, and hen measure

More information

Lecture 18: The Laplace Transform (See Sections and 14.7 in Boas)

Lecture 18: The Laplace Transform (See Sections and 14.7 in Boas) Lecure 8: The Lalace Transform (See Secons 88- and 47 n Boas) Recall ha our bg-cure goal s he analyss of he dfferenal equaon, ax bx cx F, where we emloy varous exansons for he drvng funcon F deendng on

More information

. The geometric multiplicity is dim[ker( λi. number of linearly independent eigenvectors associated with this eigenvalue.

. The geometric multiplicity is dim[ker( λi. number of linearly independent eigenvectors associated with this eigenvalue. Lnear Algebra Lecure # Noes We connue wh he dscusson of egenvalues, egenvecors, and dagonalzably of marces We wan o know, n parcular wha condons wll assure ha a marx can be dagonalzed and wha he obsrucons

More information

Econ107 Applied Econometrics Topic 5: Specification: Choosing Independent Variables (Studenmund, Chapter 6)

Econ107 Applied Econometrics Topic 5: Specification: Choosing Independent Variables (Studenmund, Chapter 6) Econ7 Appled Economercs Topc 5: Specfcaon: Choosng Independen Varables (Sudenmund, Chaper 6 Specfcaon errors ha we wll deal wh: wrong ndependen varable; wrong funconal form. Ths lecure deals wh wrong ndependen

More information

Volatility Interpolation

Volatility Interpolation Volaly Inerpolaon Prelmnary Verson March 00 Jesper Andreasen and Bran Huge Danse Mares, Copenhagen wan.daddy@danseban.com brno@danseban.com Elecronc copy avalable a: hp://ssrn.com/absrac=69497 Inro Local

More information

FTCS Solution to the Heat Equation

FTCS Solution to the Heat Equation FTCS Soluon o he Hea Equaon ME 448/548 Noes Gerald Reckenwald Porland Sae Unversy Deparmen of Mechancal Engneerng gerry@pdxedu ME 448/548: FTCS Soluon o he Hea Equaon Overvew Use he forward fne d erence

More information

Outline. Probabilistic Model Learning. Probabilistic Model Learning. Probabilistic Model for Time-series Data: Hidden Markov Model

Outline. Probabilistic Model Learning. Probabilistic Model Learning. Probabilistic Model for Time-series Data: Hidden Markov Model Probablsc Model for Tme-seres Daa: Hdden Markov Model Hrosh Mamsuka Bonformacs Cener Kyoo Unversy Oulne Three Problems for probablsc models n machne learnng. Compung lkelhood 2. Learnng 3. Parsng (predcon

More information

2 Aggregate demand in partial equilibrium static framework

2 Aggregate demand in partial equilibrium static framework Unversy of Mnnesoa 8107 Macroeconomc Theory, Sprng 2009, Mn 1 Fabrzo Perr Lecure 1. Aggregaon 1 Inroducon Probably so far n he macro sequence you have deal drecly wh represenave consumers and represenave

More information

Lecture VI Regression

Lecture VI Regression Lecure VI Regresson (Lnear Mehods for Regresson) Conens: Lnear Mehods for Regresson Leas Squares, Gauss Markov heorem Recursve Leas Squares Lecure VI: MLSC - Dr. Sehu Vjayakumar Lnear Regresson Model M

More information

( ) [ ] MAP Decision Rule

( ) [ ] MAP Decision Rule Announcemens Bayes Decson Theory wh Normal Dsrbuons HW0 due oday HW o be assgned soon Proec descrpon posed Bomercs CSE 90 Lecure 4 CSE90, Sprng 04 CSE90, Sprng 04 Key Probables 4 ω class label X feaure

More information

RELATIONSHIP BETWEEN VOLATILITY AND TRADING VOLUME: THE CASE OF HSI STOCK RETURNS DATA

RELATIONSHIP BETWEEN VOLATILITY AND TRADING VOLUME: THE CASE OF HSI STOCK RETURNS DATA RELATIONSHIP BETWEEN VOLATILITY AND TRADING VOLUME: THE CASE OF HSI STOCK RETURNS DATA Mchaela Chocholaá Unversy of Economcs Braslava, Slovaka Inroducon (1) one of he characersc feaures of sock reurns

More information

CS286.2 Lecture 14: Quantum de Finetti Theorems II

CS286.2 Lecture 14: Quantum de Finetti Theorems II CS286.2 Lecure 14: Quanum de Fne Theorems II Scrbe: Mara Okounkova 1 Saemen of he heorem Recall he las saemen of he quanum de Fne heorem from he prevous lecure. Theorem 1 Quanum de Fne). Le ρ Dens C 2

More information

Solution in semi infinite diffusion couples (error function analysis)

Solution in semi infinite diffusion couples (error function analysis) Soluon n sem nfne dffuson couples (error funcon analyss) Le us consder now he sem nfne dffuson couple of wo blocks wh concenraon of and I means ha, n a A- bnary sysem, s bondng beween wo blocks made of

More information

Political Economy of Institutions and Development: Problem Set 2 Due Date: Thursday, March 15, 2019.

Political Economy of Institutions and Development: Problem Set 2 Due Date: Thursday, March 15, 2019. Polcal Economy of Insuons and Developmen: 14.773 Problem Se 2 Due Dae: Thursday, March 15, 2019. Please answer Quesons 1, 2 and 3. Queson 1 Consder an nfne-horzon dynamc game beween wo groups, an ele and

More information

On One Analytic Method of. Constructing Program Controls

On One Analytic Method of. Constructing Program Controls Appled Mahemacal Scences, Vol. 9, 05, no. 8, 409-407 HIKARI Ld, www.m-hkar.com hp://dx.do.org/0.988/ams.05.54349 On One Analyc Mehod of Consrucng Program Conrols A. N. Kvko, S. V. Chsyakov and Yu. E. Balyna

More information

Time-interval analysis of β decay. V. Horvat and J. C. Hardy

Time-interval analysis of β decay. V. Horvat and J. C. Hardy Tme-nerval analyss of β decay V. Horva and J. C. Hardy Work on he even analyss of β decay [1] connued and resuled n he developmen of a novel mehod of bea-decay me-nerval analyss ha produces hghly accurae

More information

. The geometric multiplicity is dim[ker( λi. A )], i.e. the number of linearly independent eigenvectors associated with this eigenvalue.

. The geometric multiplicity is dim[ker( λi. A )], i.e. the number of linearly independent eigenvectors associated with this eigenvalue. Mah E-b Lecure #0 Noes We connue wh he dscusson of egenvalues, egenvecors, and dagonalzably of marces We wan o know, n parcular wha condons wll assure ha a marx can be dagonalzed and wha he obsrucons are

More information

CHAPTER 10: LINEAR DISCRIMINATION

CHAPTER 10: LINEAR DISCRIMINATION CHAPER : LINEAR DISCRIMINAION Dscrmnan-based Classfcaon 3 In classfcaon h K classes (C,C,, C k ) We defned dscrmnan funcon g j (), j=,,,k hen gven an es eample, e chose (predced) s class label as C f g

More information

EUROPEAN ECONOMY EUROPEAN COMMISSION DIRECTORATE-GENERAL FOR ECONOMIC AND FINANCIAL AFFAIRS

EUROPEAN ECONOMY EUROPEAN COMMISSION DIRECTORATE-GENERAL FOR ECONOMIC AND FINANCIAL AFFAIRS EUROPEAN ECONOMY EUROPEAN COMMISSION DIRECTORATE-GENERAL FOR ECONOMIC AND FINANCIAL AFFAIRS ECONOMIC PAPERS ISSN 75-387 hp://ec.europa.eu/economy_fnance/ndex_en.hm N 66 December 6 Fscal polcy n an esmaed

More information

Math 128b Project. Jude Yuen

Math 128b Project. Jude Yuen Mah 8b Proec Jude Yuen . Inroducon Le { Z } be a sequence of observed ndependen vecor varables. If he elemens of Z have a on normal dsrbuon hen { Z } has a mean vecor Z and a varancecovarance marx z. Geomercally

More information

An introduction to Support Vector Machine

An introduction to Support Vector Machine An nroducon o Suppor Vecor Machne 報告者 : 黃立德 References: Smon Haykn, "Neural Neworks: a comprehensve foundaon, second edon, 999, Chaper 2,6 Nello Chrsann, John Shawe-Tayer, An Inroducon o Suppor Vecor Machnes,

More information

THEORETICAL AUTOCORRELATIONS. ) if often denoted by γ. Note that

THEORETICAL AUTOCORRELATIONS. ) if often denoted by γ. Note that THEORETICAL AUTOCORRELATIONS Cov( y, y ) E( y E( y))( y E( y)) ρ = = Var( y) E( y E( y)) =,, L ρ = and Cov( y, y ) s ofen denoed by whle Var( y ) f ofen denoed by γ. Noe ha γ = γ and ρ = ρ and because

More information

Lecture 11 SVM cont

Lecture 11 SVM cont Lecure SVM con. 0 008 Wha we have done so far We have esalshed ha we wan o fnd a lnear decson oundary whose margn s he larges We know how o measure he margn of a lnear decson oundary Tha s: he mnmum geomerc

More information

Technical Appendix for Central Bank Communication and Expectations Stabilization

Technical Appendix for Central Bank Communication and Expectations Stabilization Techncal Appendx for Cenral Bank Communcaon and Expecaons Sablzaon Sefano Eusep Federal Reserve Bank of New York Bruce Preson Columba Unversy and NBER Augus 0, 008 Absrac Ths echncal appendx provdes some

More information

Technical Appendix to The Equivalence of Wage and Price Staggering in Monetary Business Cycle Models

Technical Appendix to The Equivalence of Wage and Price Staggering in Monetary Business Cycle Models Techncal Appendx o The Equvalence of Wage and Prce Saggerng n Moneary Busness Cycle Models Rochelle M. Edge Dvson of Research and Sascs Federal Reserve Board Sepember 24, 2 Absrac Ths appendx deals he

More information

Problem 1 / 25 Problem 2 / 15 Problem 3 / 15 Problem 4 / 20 Problem 5 / 25 TOTAL / 100

Problem 1 / 25 Problem 2 / 15 Problem 3 / 15 Problem 4 / 20 Problem 5 / 25 TOTAL / 100 Deparmen of Appled Economcs Johns Hopkns Unversy Economcs 60 Macroeconomc Theory and Polcy Fnal Exam Suggesed Soluons Professor Sanjay Chugh Fall 009 NAME: The Exam has a oal of fve (5) problems and pages

More information

Lecture Notes 4: Consumption 1

Lecture Notes 4: Consumption 1 Leure Noes 4: Consumpon Zhwe Xu (xuzhwe@sju.edu.n) hs noe dsusses households onsumpon hoe. In he nex leure, we wll dsuss rm s nvesmen deson. I s safe o say ha any propagaon mehansm of maroeonom model s

More information

Advanced time-series analysis (University of Lund, Economic History Department)

Advanced time-series analysis (University of Lund, Economic History Department) Advanced me-seres analss (Unvers of Lund, Economc Hsor Dearmen) 3 Jan-3 Februar and 6-3 March Lecure 4 Economerc echnues for saonar seres : Unvarae sochasc models wh Box- Jenns mehodolog, smle forecasng

More information

Ordinary Differential Equations in Neuroscience with Matlab examples. Aim 1- Gain understanding of how to set up and solve ODE s

Ordinary Differential Equations in Neuroscience with Matlab examples. Aim 1- Gain understanding of how to set up and solve ODE s Ordnary Dfferenal Equaons n Neuroscence wh Malab eamples. Am - Gan undersandng of how o se up and solve ODE s Am Undersand how o se up an solve a smple eample of he Hebb rule n D Our goal a end of class

More information

2 Aggregate demand in partial equilibrium static framework

2 Aggregate demand in partial equilibrium static framework Unversy of Mnnesoa 8107 Macroeconomc Theory, Sprng 2012, Mn 1 Fabrzo Perr Lecure 1. Aggregaon 1 Inroducon Probably so far n he macro sequence you have deal drecly wh represenave consumers and represenave

More information

Analysis And Evaluation of Econometric Time Series Models: Dynamic Transfer Function Approach

Analysis And Evaluation of Econometric Time Series Models: Dynamic Transfer Function Approach 1 Appeared n Proceedng of he 62 h Annual Sesson of he SLAAS (2006) pp 96. Analyss And Evaluaon of Economerc Tme Seres Models: Dynamc Transfer Funcon Approach T.M.J.A.COORAY Deparmen of Mahemacs Unversy

More information

2. SPATIALLY LAGGED DEPENDENT VARIABLES

2. SPATIALLY LAGGED DEPENDENT VARIABLES 2. SPATIALLY LAGGED DEPENDENT VARIABLES In hs chaper, we descrbe a sascal model ha ncorporaes spaal dependence explcly by addng a spaally lagged dependen varable y on he rgh-hand sde of he regresson equaon.

More information

J i-1 i. J i i+1. Numerical integration of the diffusion equation (I) Finite difference method. Spatial Discretization. Internal nodes.

J i-1 i. J i i+1. Numerical integration of the diffusion equation (I) Finite difference method. Spatial Discretization. Internal nodes. umercal negraon of he dffuson equaon (I) Fne dfference mehod. Spaal screaon. Inernal nodes. R L V For hermal conducon le s dscree he spaal doman no small fne spans, =,,: Balance of parcles for an nernal

More information

Chapter 9: Factor pricing models. Asset Pricing Zheng Zhenlong

Chapter 9: Factor pricing models. Asset Pricing Zheng Zhenlong Chaper 9: Facor prcng models Asse Prcng Conens Asse Prcng Inroducon CAPM ICAPM Commens on he CAPM and ICAPM APT APT vs. ICAPM Bref nroducon Asse Prcng u β u ( c + 1 ) a + b f + 1 ( c ) Bref nroducon Asse

More information

Machine Learning Linear Regression

Machine Learning Linear Regression Machne Learnng Lnear Regresson Lesson 3 Lnear Regresson Bascs of Regresson Leas Squares esmaon Polynomal Regresson Bass funcons Regresson model Regularzed Regresson Sascal Regresson Mamum Lkelhood (ML)

More information

Clustering (Bishop ch 9)

Clustering (Bishop ch 9) Cluserng (Bshop ch 9) Reference: Daa Mnng by Margare Dunham (a slde source) 1 Cluserng Cluserng s unsupervsed learnng, here are no class labels Wan o fnd groups of smlar nsances Ofen use a dsance measure

More information

Estimation of Cost and. Albert Banal-Estanol

Estimation of Cost and. Albert Banal-Estanol Esmaon of Cos and Producon Funcons ns Movaon: Producon and Cos Funcons Objecve: Fnd shape of producon/cos funcons Evaluae effcency: Increasng reurns, economes of scale Complemenary/subsuably beween npus

More information

Comb Filters. Comb Filters

Comb Filters. Comb Filters The smple flers dscussed so far are characered eher by a sngle passband and/or a sngle sopband There are applcaons where flers wh mulple passbands and sopbands are requred Thecomb fler s an example of

More information

(,,, ) (,,, ). In addition, there are three other consumers, -2, -1, and 0. Consumer -2 has the utility function

(,,, ) (,,, ). In addition, there are three other consumers, -2, -1, and 0. Consumer -2 has the utility function MACROECONOMIC THEORY T J KEHOE ECON 87 SPRING 5 PROBLEM SET # Conder an overlappng generaon economy le ha n queon 5 on problem e n whch conumer lve for perod The uly funcon of he conumer born n perod,

More information

New M-Estimator Objective Function. in Simultaneous Equations Model. (A Comparative Study)

New M-Estimator Objective Function. in Simultaneous Equations Model. (A Comparative Study) Inernaonal Mahemacal Forum, Vol. 8, 3, no., 7 - HIKARI Ld, www.m-hkar.com hp://dx.do.org/.988/mf.3.3488 New M-Esmaor Objecve Funcon n Smulaneous Equaons Model (A Comparave Sudy) Ahmed H. Youssef Professor

More information

Chapter 6: AC Circuits

Chapter 6: AC Circuits Chaper 6: AC Crcus Chaper 6: Oulne Phasors and he AC Seady Sae AC Crcus A sable, lnear crcu operang n he seady sae wh snusodal excaon (.e., snusodal seady sae. Complee response forced response naural response.

More information

Computing Relevance, Similarity: The Vector Space Model

Computing Relevance, Similarity: The Vector Space Model Compung Relevance, Smlary: The Vecor Space Model Based on Larson and Hears s sldes a UC-Bereley hp://.sms.bereley.edu/courses/s0/f00/ aabase Managemen Sysems, R. Ramarshnan ocumen Vecors v ocumens are

More information

CHAPTER 5: MULTIVARIATE METHODS

CHAPTER 5: MULTIVARIATE METHODS CHAPER 5: MULIVARIAE MEHODS Mulvarae Daa 3 Mulple measuremens (sensors) npus/feaures/arbues: -varae N nsances/observaons/eamples Each row s an eample Each column represens a feaure X a b correspons o he

More information

Notes on the stability of dynamic systems and the use of Eigen Values.

Notes on the stability of dynamic systems and the use of Eigen Values. Noes on he sabl of dnamc ssems and he use of Egen Values. Source: Macro II course noes, Dr. Davd Bessler s Tme Seres course noes, zarads (999) Ineremporal Macroeconomcs chaper 4 & Techncal ppend, and Hamlon

More information

CH.3. COMPATIBILITY EQUATIONS. Continuum Mechanics Course (MMC) - ETSECCPB - UPC

CH.3. COMPATIBILITY EQUATIONS. Continuum Mechanics Course (MMC) - ETSECCPB - UPC CH.3. COMPATIBILITY EQUATIONS Connuum Mechancs Course (MMC) - ETSECCPB - UPC Overvew Compably Condons Compably Equaons of a Poenal Vecor Feld Compably Condons for Infnesmal Srans Inegraon of he Infnesmal

More information

Comparison of Supervised & Unsupervised Learning in βs Estimation between Stocks and the S&P500

Comparison of Supervised & Unsupervised Learning in βs Estimation between Stocks and the S&P500 Comparson of Supervsed & Unsupervsed Learnng n βs Esmaon beween Socks and he S&P500 J. We, Y. Hassd, J. Edery, A. Becker, Sanford Unversy T I. INTRODUCTION HE goal of our proec s o analyze he relaonshps

More information

2.1 Constitutive Theory

2.1 Constitutive Theory Secon.. Consuve Theory.. Consuve Equaons Governng Equaons The equaons governng he behavour of maerals are (n he spaal form) dρ v & ρ + ρdv v = + ρ = Conservaon of Mass (..a) d x σ j dv dvσ + b = ρ v& +

More information

Bayes rule for a classification problem INF Discriminant functions for the normal density. Euclidean distance. Mahalanobis distance

Bayes rule for a classification problem INF Discriminant functions for the normal density. Euclidean distance. Mahalanobis distance INF 43 3.. Repeon Anne Solberg (anne@f.uo.no Bayes rule for a classfcaon problem Suppose we have J, =,...J classes. s he class label for a pxel, and x s he observed feaure vecor. We can use Bayes rule

More information

Dual Approximate Dynamic Programming for Large Scale Hydro Valleys

Dual Approximate Dynamic Programming for Large Scale Hydro Valleys Dual Approxmae Dynamc Programmng for Large Scale Hydro Valleys Perre Carpener and Jean-Phlppe Chanceler 1 ENSTA ParsTech and ENPC ParsTech CMM Workshop, January 2016 1 Jon work wh J.-C. Alas, suppored

More information

Lecture Notes 4. Univariate Forecasting and the Time Series Properties of Dynamic Economic Models

Lecture Notes 4. Univariate Forecasting and the Time Series Properties of Dynamic Economic Models Tme Seres Seven N. Durlauf Unversy of Wsconsn Lecure Noes 4. Unvarae Forecasng and he Tme Seres Properes of Dynamc Economc Models Ths se of noes presens does hree hngs. Frs, formulas are developed o descrbe

More information

Appendix H: Rarefaction and extrapolation of Hill numbers for incidence data

Appendix H: Rarefaction and extrapolation of Hill numbers for incidence data Anne Chao Ncholas J Goell C seh lzabeh L ander K Ma Rober K Colwell and Aaron M llson 03 Rarefacon and erapolaon wh ll numbers: a framewor for samplng and esmaon n speces dversy sudes cology Monographs

More information

Relative controllability of nonlinear systems with delays in control

Relative controllability of nonlinear systems with delays in control Relave conrollably o nonlnear sysems wh delays n conrol Jerzy Klamka Insue o Conrol Engneerng, Slesan Techncal Unversy, 44- Glwce, Poland. phone/ax : 48 32 37227, {jklamka}@a.polsl.glwce.pl Keywor: Conrollably.

More information

NATIONAL UNIVERSITY OF SINGAPORE PC5202 ADVANCED STATISTICAL MECHANICS. (Semester II: AY ) Time Allowed: 2 Hours

NATIONAL UNIVERSITY OF SINGAPORE PC5202 ADVANCED STATISTICAL MECHANICS. (Semester II: AY ) Time Allowed: 2 Hours NATONAL UNVERSTY OF SNGAPORE PC5 ADVANCED STATSTCAL MECHANCS (Semeser : AY 1-13) Tme Allowed: Hours NSTRUCTONS TO CANDDATES 1. Ths examnaon paper conans 5 quesons and comprses 4 prned pages.. Answer all

More information

Part II CONTINUOUS TIME STOCHASTIC PROCESSES

Part II CONTINUOUS TIME STOCHASTIC PROCESSES Par II CONTINUOUS TIME STOCHASTIC PROCESSES 4 Chaper 4 For an advanced analyss of he properes of he Wener process, see: Revus D and Yor M: Connuous marngales and Brownan Moon Karazas I and Shreve S E:

More information

P R = P 0. The system is shown on the next figure:

P R = P 0. The system is shown on the next figure: TPG460 Reservor Smulaon 08 page of INTRODUCTION TO RESERVOIR SIMULATION Analycal and numercal soluons of smple one-dmensonal, one-phase flow equaons As an nroducon o reservor smulaon, we wll revew he smples

More information

Robust and Accurate Cancer Classification with Gene Expression Profiling

Robust and Accurate Cancer Classification with Gene Expression Profiling Robus and Accurae Cancer Classfcaon wh Gene Expresson Proflng (Compuaonal ysems Bology, 2005) Auhor: Hafeng L, Keshu Zhang, ao Jang Oulne Background LDA (lnear dscrmnan analyss) and small sample sze problem

More information

TSS = SST + SSE An orthogonal partition of the total SS

TSS = SST + SSE An orthogonal partition of the total SS ANOVA: Topc 4. Orhogonal conrass [ST&D p. 183] H 0 : µ 1 = µ =... = µ H 1 : The mean of a leas one reamen group s dfferen To es hs hypohess, a basc ANOVA allocaes he varaon among reamen means (SST) equally

More information

NPTEL Project. Econometric Modelling. Module23: Granger Causality Test. Lecture35: Granger Causality Test. Vinod Gupta School of Management

NPTEL Project. Econometric Modelling. Module23: Granger Causality Test. Lecture35: Granger Causality Test. Vinod Gupta School of Management P age NPTEL Proec Economerc Modellng Vnod Gua School of Managemen Module23: Granger Causaly Tes Lecure35: Granger Causaly Tes Rudra P. Pradhan Vnod Gua School of Managemen Indan Insue of Technology Kharagur,

More information

Single-loop System Reliability-Based Design & Topology Optimization (SRBDO/SRBTO): A Matrix-based System Reliability (MSR) Method

Single-loop System Reliability-Based Design & Topology Optimization (SRBDO/SRBTO): A Matrix-based System Reliability (MSR) Method 10 h US Naonal Congress on Compuaonal Mechancs Columbus, Oho 16-19, 2009 Sngle-loop Sysem Relably-Based Desgn & Topology Opmzaon (SRBDO/SRBTO): A Marx-based Sysem Relably (MSR) Mehod Tam Nguyen, Junho

More information

Knowing What Others Know: Coordination Motives in Information Acquisition Additional Notes

Knowing What Others Know: Coordination Motives in Information Acquisition Additional Notes Knowng Wha Ohers Know: Coordnaon Moves n nformaon Acquson Addonal Noes Chrsan Hellwg Unversy of Calforna, Los Angeles Deparmen of Economcs Laura Veldkamp New York Unversy Sern School of Busness March 1,

More information

F-Tests and Analysis of Variance (ANOVA) in the Simple Linear Regression Model. 1. Introduction

F-Tests and Analysis of Variance (ANOVA) in the Simple Linear Regression Model. 1. Introduction ECOOMICS 35* -- OTE 9 ECO 35* -- OTE 9 F-Tess and Analyss of Varance (AOVA n he Smple Lnear Regresson Model Inroducon The smple lnear regresson model s gven by he followng populaon regresson equaon, or

More information

Existence and Uniqueness Results for Random Impulsive Integro-Differential Equation

Existence and Uniqueness Results for Random Impulsive Integro-Differential Equation Global Journal of Pure and Appled Mahemacs. ISSN 973-768 Volume 4, Number 6 (8), pp. 89-87 Research Inda Publcaons hp://www.rpublcaon.com Exsence and Unqueness Resuls for Random Impulsve Inegro-Dfferenal

More information

Chapter Lagrangian Interpolation

Chapter Lagrangian Interpolation Chaper 5.4 agrangan Inerpolaon Afer readng hs chaper you should be able o:. dere agrangan mehod of nerpolaon. sole problems usng agrangan mehod of nerpolaon and. use agrangan nerpolans o fnd deraes and

More information

Forecasting customer behaviour in a multi-service financial organisation: a profitability perspective

Forecasting customer behaviour in a multi-service financial organisation: a profitability perspective Forecasng cusomer behavour n a mul-servce fnancal organsaon: a profably perspecve A. Audzeyeva, Unversy of Leeds & Naonal Ausrala Group Europe, UK B. Summers, Unversy of Leeds, UK K.R. Schenk-Hoppé, Unversy

More information

Midterm Exam. Thursday, April hour, 15 minutes

Midterm Exam. Thursday, April hour, 15 minutes Economcs of Grow, ECO560 San Francsco Sae Unvers Mcael Bar Sprng 04 Mderm Exam Tursda, prl 0 our, 5 mnues ame: Insrucons. Ts s closed boo, closed noes exam.. o calculaors of an nd are allowed. 3. Sow all

More information

Advanced Machine Learning & Perception

Advanced Machine Learning & Perception Advanced Machne Learnng & Percepon Insrucor: Tony Jebara SVM Feaure & Kernel Selecon SVM Eensons Feaure Selecon (Flerng and Wrappng) SVM Feaure Selecon SVM Kernel Selecon SVM Eensons Classfcaon Feaure/Kernel

More information

5th International Conference on Advanced Design and Manufacturing Engineering (ICADME 2015)

5th International Conference on Advanced Design and Manufacturing Engineering (ICADME 2015) 5h Inernaonal onference on Advanced Desgn and Manufacurng Engneerng (IADME 5 The Falure Rae Expermenal Sudy of Specal N Machne Tool hunshan He, a, *, La Pan,b and Bng Hu 3,c,,3 ollege of Mechancal and

More information

Cubic Bezier Homotopy Function for Solving Exponential Equations

Cubic Bezier Homotopy Function for Solving Exponential Equations Penerb Journal of Advanced Research n Compung and Applcaons ISSN (onlne: 46-97 Vol. 4, No.. Pages -8, 6 omoopy Funcon for Solvng Eponenal Equaons S. S. Raml *,,. Mohamad Nor,a, N. S. Saharzan,b and M.

More information

Lecture 1 The New Keynesian Model of Monetary Policy. Lecturer Campbell Leith University of Glasgow

Lecture 1 The New Keynesian Model of Monetary Policy. Lecturer Campbell Leith University of Glasgow Lecure The New Keynesan Model of Moneary Polcy Lecurer Campbell Leh Unversy of Glasgow The New Keynesan model of moneary polcy s becomng ncreasngly sandard n he analyss of moneary polcy. Ths parcular reamen

More information

Mechanics Physics 151

Mechanics Physics 151 Mechancs Physcs 5 Lecure 9 Hamlonan Equaons of Moon (Chaper 8) Wha We Dd Las Tme Consruced Hamlonan formalsm H ( q, p, ) = q p L( q, q, ) H p = q H q = p H = L Equvalen o Lagrangan formalsm Smpler, bu

More information

WiH Wei He

WiH Wei He Sysem Idenfcaon of onlnear Sae-Space Space Baery odels WH We He wehe@calce.umd.edu Advsor: Dr. Chaochao Chen Deparmen of echancal Engneerng Unversy of aryland, College Par 1 Unversy of aryland Bacground

More information

Genetic Algorithm in Parameter Estimation of Nonlinear Dynamic Systems

Genetic Algorithm in Parameter Estimation of Nonlinear Dynamic Systems Genec Algorhm n Parameer Esmaon of Nonlnear Dynamc Sysems E. Paeraks manos@egnaa.ee.auh.gr V. Perds perds@vergna.eng.auh.gr Ah. ehagas kehagas@egnaa.ee.auh.gr hp://skron.conrol.ee.auh.gr/kehagas/ndex.hm

More information

HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD

HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD Journal of Appled Mahemacs and Compuaonal Mechancs 3, (), 45-5 HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD Sansław Kukla, Urszula Sedlecka Insue of Mahemacs,

More information

3. OVERVIEW OF NUMERICAL METHODS

3. OVERVIEW OF NUMERICAL METHODS 3 OVERVIEW OF NUMERICAL METHODS 3 Inroducory remarks Ths chaper summarzes hose numercal echnques whose knowledge s ndspensable for he undersandng of he dfferen dscree elemen mehods: he Newon-Raphson-mehod,

More information

Bayesian Inference of the GARCH model with Rational Errors

Bayesian Inference of the GARCH model with Rational Errors 0 Inernaonal Conference on Economcs, Busness and Markeng Managemen IPEDR vol.9 (0) (0) IACSIT Press, Sngapore Bayesan Inference of he GARCH model wh Raonal Errors Tesuya Takash + and Tng Tng Chen Hroshma

More information

GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim

GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim Korean J. Mah. 19 (2011), No. 3, pp. 263 272 GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS Youngwoo Ahn and Kae Km Absrac. In he paper [1], an explc correspondence beween ceran

More information

Filtrage particulaire et suivi multi-pistes Carine Hue Jean-Pierre Le Cadre and Patrick Pérez

Filtrage particulaire et suivi multi-pistes Carine Hue Jean-Pierre Le Cadre and Patrick Pérez Chaînes de Markov cachées e flrage parculare 2-22 anver 2002 Flrage parculare e suv mul-pses Carne Hue Jean-Perre Le Cadre and Parck Pérez Conex Applcaons: Sgnal processng: arge rackng bearngs-onl rackng

More information

CHAPTER 2: Supervised Learning

CHAPTER 2: Supervised Learning HATER 2: Supervsed Learnng Learnng a lass from Eamples lass of a famly car redcon: Is car a famly car? Knowledge eracon: Wha do people epec from a famly car? Oupu: osve (+) and negave ( ) eamples Inpu

More information

EEL 6266 Power System Operation and Control. Chapter 5 Unit Commitment

EEL 6266 Power System Operation and Control. Chapter 5 Unit Commitment EEL 6266 Power Sysem Operaon and Conrol Chaper 5 Un Commmen Dynamc programmng chef advanage over enumeraon schemes s he reducon n he dmensonaly of he problem n a src prory order scheme, here are only N

More information

2/20/2013. EE 101 Midterm 2 Review

2/20/2013. EE 101 Midterm 2 Review //3 EE Mderm eew //3 Volage-mplfer Model The npu ressance s he equalen ressance see when lookng no he npu ermnals of he amplfer. o s he oupu ressance. I causes he oupu olage o decrease as he load ressance

More information

Mechanics Physics 151

Mechanics Physics 151 Mechancs Physcs 5 Lecure 9 Hamlonan Equaons of Moon (Chaper 8) Wha We Dd Las Tme Consruced Hamlonan formalsm Hqp (,,) = qp Lqq (,,) H p = q H q = p H L = Equvalen o Lagrangan formalsm Smpler, bu wce as

More information