What determines government spending multipliers - PDF document

FinancialcrisisWeakpublicnancesPegSpending 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -0.5 0 0.5 1 1.5 2 Output 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Consumption 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -6 -4 -2 0 2 4 Investment 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 Netexports 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 Exchangerate 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 Ination 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 Interestrate 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Figure11:ResultsforsamplewithoutUnitedStates.Notes:quantitiesmeasuredinoutputunits;solidline:pointestimateforbaselinescenario(shadedarea:onestandarderror);dasheddottedline(dashedlines:onestandarderror):deviationfrombaselineunderspecicconditionindicatedontopofcolumn.45 FinancialcrisisWeakpublicnancesPegSpending 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -0.5 0 0.5 1 1.5 2 Output 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Consumption 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -6 -4 -2 0 2 4 Investment 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 Netexports 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 Exchangerate 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 Ination 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 Interestrate 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Figure10:Resultsforsamplewithout2007–2008.Notes:quantitiesmeasuredinoutputunits;solidline:pointestimateforbaselinescenario(shadedarea:onestandarderror);dasheddottedline(dashedlines:onestandarderror):deviationfrombaselineunderspecicconditionindicatedontopofcolumn.44 FinancialcrisisWeakpublicnancesPegSpending 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -0.5 0 0.5 1 1.5 2 Output 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Consumption 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -6 -4 -2 0 2 4 Investment 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 Netexports 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 Exchangerate 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 Ination 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 Interestrate 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Figure9:ResultswithoutCLIinrststep.Notes:seeFigure543 FinancialcrisisWeakpublicnancesPegSpending 0 2 4 6 -1 0 1 2 0 2 4 6 -1 0 1 2 0 2 4 6 -1 0 1 2 Output 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Consumption 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 Investment 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 Netexports 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 Exchangerate 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 Ination 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 Interestrate 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Figure8:Resultsforrst-stepspecicationwhichincludescontemporaneousvalueofcrisisdummy.Notes:seeFigure5.42 FinancialcrisisWeakpublicnancesPegSpending 0 2 4 6 0 1 2 3 4 0 2 4 6 0 1 2 3 4 0 2 4 6 0 1 2 3 4 Output 0 2 4 6 -4 -2 0 2 4 6 8 0 2 4 6 -4 -2 0 2 4 6 8 0 2 4 6 -4 -2 0 2 4 6 8 Consumption 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 Investment 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 Netexports 0 2 4 6 -10 0 10 20 30 0 2 4 6 -10 0 10 20 30 0 2 4 6 -10 0 10 20 30 Exchangerate 0 2 4 6 -50 -40 -30 -20 -10 0 0 2 4 6 -50 -40 -30 -20 -10 0 0 2 4 6 -50 -40 -30 -20 -10 0 Ination 0 2 4 6 0 5 10 0 2 4 6 0 5 10 0 2 4 6 0 5 10 Interestrate 0 2 4 6 -10 -5 0 5 10 15 0 2 4 6 -10 -5 0 5 10 15 0 2 4 6 -10 -5 0 5 10 15 Figure7:Resultsfordifferencespecication.Notes:responsesdisplaycumulativeeffectofgovernmentspendinginnovationformodelestimatedinrstdifferences;quantitiesmeasuredinoutputunits;solidline:pointestimateforbaselinescenario(shadedarea:onestandarderror);dasheddottedline(dashedlines:onestandarderror):deviationfrombaselineunderspecicconditionindicatedontopofcolumn.41 FinancialcrisisVeryweakpublicnancesPegSpending 0 2 4 6 -1 0 1 2 0 2 4 6 -1 0 1 2 0 2 4 6 -1 0 1 2 Output 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Consumption 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 Investment 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 Netexports 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 Exchangerate 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 Ination 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 Interestrate 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Figure6:Resultsforalternativedenitionofweakpublicnances(governmentdebt120percentofGDPand/orlaggednetborrowingpercentofGDP).Notes:seeFigure5.40 FinancialcrisisWeakpublicnancesPegSpending 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -0.5 0 0.5 1 1.5 2 Output 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Consumption 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -6 -4 -2 0 2 4 Investment 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 0 2 4 6 -5 0 5 Netexports 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 0 2 4 6 -10 -5 0 5 Exchangerate 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 0 2 4 6 -30 -20 -10 0 Ination 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 Interestrate 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -4 -2 0 2 4 6 Figure5:Resultsfornarrowdenitionofnancialcrisis.Notes:quantitiesmeasuredinoutputunits;solidline:pointestimateforbaselinescenario(shadedarea:onestandarderror);dasheddottedline(dashedlines:onestandarderror):deviationfrombaselineunderthespecicconditionindicatedontopofcolumn.39 Austria, 1978-2008Australia, 1992Austria, 2008 38 1/ All statistics refer to final sample described in Table 1. 37 !"#"$%"$&'()*+),-".'/01'02'#"*'3)#+&)'10("*$4"$&'30$564#&+0$706$&*8965&*)-+):;=&#x-400;;:;?&#x-400;@;&#x-400;@A;=B;;;@@:;;;;:;　?A;;;䀀;;;=DE;?;;FE;=&#x-400;DFE;　@?FE;=?;FE;;ABFE;;;AFE;;G　F965&*+)　?BBCCC:;@;GCCC:;?=;CC;=B@;;=;;;;A;;;　;;;@HE;&#x-400;GHFE;&#x-400;HAFE;　BDFE;&#x-400;BBFE;;HAFE;;??FI"-1+64;@@BCCC;　AB;;@H:;&#x-400;;?:;;@B;;&#x-400;B:;;B　CCC;;;;AA;　　E;&#x-400;AGFE;&#x-400;G@FE;　@BFE;&#x-400;HHFE;;AGFE;;;AFE;;　BF7)$)%);A怀HCCC:;;;B;&#x-400;A;:;;HA:;&#x-400;B&#x-400;C:;;　H;;怀@;;;&#x-200;:;;=;;;;;AD;?DE;&#x-400;A=FE;&#x-400;H=FE;　G　FE;&#x-400;A@FE;;GGFE;;　;FE;;&#x-400;&#x-400;FE;;&#x-400;HFE;;B?F!"$4)*J　;;GCCC:;;DB:;;=B:;&#x-400;&#x-400;&#x-400;;;GH:;;　@:;;&#x-400;;:;;;@;;;;AA;AGE;&#x-400;AAFE;=;@FE;=　;FE;=&#x-400;HFE;;DDFE;;　GFE;;;AFE;;　GFK+$-)$%　;H;CC:;B?=;B?&#x-400;C:;;GG:;;=;;;&#x-400;=:;;;;;;D=;;;;AA;GBE;?H?FE;?G&#x-400;FE;==;FE;=BAFE;;DBFE;;　　FE;;&#x-400;AFE;　　@FK*)$3";H怀;CCC;=GG;;D@;;A=:;;DA:;;;D:;;;;:;G&#x;400;C;;;　;;;B?E;=;;FE;&#x-400;D?FE;==　FE;=HGFE;&#x-400;B=FE;;　　FE;;;GFE;;?HFL*"-)$%;G;ACCC:;;G@;;;=:;;;D;BHHCC:;;　　:;;BHC:;　DDCC;;;　;;;B=E;&#x-400;D&#x-400;FE;&#x-400;G&#x-400;FE;=??FE;=?DFE;&#x-400;H=FE;;=DFE;;=BFE;;DGFL&)-8　;AACCC:;=?@;=AA;;怀?:;;;D:;;&#x-400;@:;;;H:;G;C;;;;AA;A@E;=;DFE;=;;FE;==@FE;=@DFE;&#x-400;;@FE;;;AFE;;&#x-400;;FE;;BHFM)#)$;H=;CCC;=;@:;@䀀ACCC;H;=CCC;;@A:;;;A:;;&#x-400;DCC:;;;=:;;　;;;;　;;;?BE;&#x-400;&#x-400;BFE;&#x-400;=AFE;　@@FE;&#x-400;HDFE;;@GFE;;;AFE;;;GFE;;&#x-400;=FE;;@DFN"&O"*-)$%5;GDBCCC:;=&#x-400;;:;　@B:;;@;;　　B:;;　A:;;=H;;;;AA;;DE;&#x-400;DAFE;==&#x-400;FE;=DHFE;?=&#x-400;FE;=&#x-400;DFE;;　?FE;;B=FN0*P)8　;怀@CCC:;?;@C;&#x-400;BG:;;B?;;?@;;　BC:;;　　;;;　;;;;GE;&#x-400;H@FE;&#x-400;@&#x-400;FE;=　=FE;&#x-400;H?FE;&#x-400;B&#x-400;FE;;;DFE;;?@FQ0*&61)-:;;G@;　BD　&#x-400;A=CC:;?DG:;&#x-400;D;;;=B:;　?H;;;;AA;?@E;?@&#x-400;FE;=AHFE;BGBFE;@@;FE;=&#x-400;@FE;;　HFE;　ADFR#)+$;@??CC;　H　;B@D:;??H;;　D;;;?:;;@BCCC:;;A;CC;;;　;;;　AE;==GFE;==?FE;=A@FE;=ABFE;&#x-400;?HFE;;;DFE;;&#x-400;HFE;;?GFRP"%"$;GHDCCC;;H@:;䀀@B:;&#x-400;HD;;@B:;;　BC:;;;G;;??C:;;B=;;;;AA;BBE;=;?FE;=;&#x-400;FE;　G@FE;&#x-400;GHFE;;G;FE;;;DFE;;&#x-400;;FE;;&#x-400;DFE;;?DFST;AD;CCC:;&#x-400;BG;;@;;怀HD:;&#x-400;HD;;;@;;怀怀:;;&#x-400;G;;;@;;;;AA;=DE;&#x-400;D&#x-400;FE;&#x-400;DDFE;=@=FE;=B@FE;&#x-400;HHFE;;　;FE;;&#x-400;@FE;;&#x-400;DFE;;H　FSR9;AADCCC:;=@GC;=B;CC:;=DHCC:;&#x-400;@@CC;;=?CCC:;;;?:;;HACC;;;;AA;B=E;&#x-400;=AFE;&#x-400;=HFE;　;HFE;&#x-400;&#x-400;GFE;;G&#x-400;FE;;;@FE;;;HFE;;=DFU"1*"550*K:&"5&'02'V0+$&'5+1$+2+3)$3"'E#:()-6"F9*"--)$0:I0$%'&"5&'02')6&030**"-)&+0$'E#:()-6"FU'5W6)*"%7/L'E:　F/)11"%'#"1'%6448I"1+$$+$1:02 : #"*+0%'10(&'%",&怀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x-400;;'#"*3"$&'ECFZ'@'#"*3"$&'ECCFZ'0*'&#x-400;'#"*3"$&'ECCCF'-"("-'9*"--)$0:I0$%'&"5&'&"5&5'20*')6&030**"-)&+0$'02'*"5+%6)-5'6#'&0'&O*""'-)15'6$%"*'&O"'$6--'O8#0&O"5+5'02'$0')6&030**"-)&+0$`),-"'?'U"56-&5'02'K+*5&:R&"#'U"1*"55+0$' X[0("*$4"$&'5#"$%+$1'E: F[0("*$4"$&'5#"$%+$1'E:=F[!Q'E: F[!Q'E:=F/)11"%'2+$'3*+5+5'%6448I)%'2+53)-'&+4"5'%6448 36 Government spendingLog of real per capita government consumptionOECD Economic Outlook Database: volume of final government consumption expenditure (CGV); OECD Analytic Database: population size (POP).OECD Economic Outlook Database: value of gross domestic product (GDP), GDP deflator (PGDP); OECD Analytic Database: population size (POP).OECD Main Economic Indicators database: CLI amplitude-adjusted; normalized by subracting 100, and Primary source: IMF World Economic Outlook: General government gross debt (GGD), nominal GDP (NGDP); where unavailable: OECD Analytic Database: General government gross financial liabilities as Financial crisis dummyTakes on value of 1 during financial crises, and 0 Reinhart and Rogoff (2008, pp. 65 ff.) and Reinhart and Rogoff (2010): the relevant crisis episodes, broadly defined, are (severe financial crises in bold): Australia 1989-92; Bad fiscal times dummyTakes on value of 1 when lagged public debt OECD Analytic Database: Government net lending as a percentage of GDP (NLGQ); see also entry for Ilzetzki, Reinhart, and Rogoff (2008): categories 1-8 of the authors' fine classification scheme are defined as "peg"; category 14 ("freely falling") coded as "not available"; within-year changes in regime category also lead to coding as "not available"; time series updated for 2008 (identical classification as Private consumptionLog per capita real private consumptionOECD Economic Outlook Database: volume of final private consumption expenditure (CPV); OECD Analytic Database: population size (POP).OECD Economic Outlook Database: volume of private total fixed capital formation (IPV); OECD Analytic Database: population size (POP).IMF World Economic Outlook: exports of goods and services at current prices (NX), imports of goods CPI-based real effective exchange rate (in percent)OECD Main Economic Indicators Database (CCRETT01.IXOB).IMF International Financial Statistics: money market rate. A OECD Main Economic Indicators Database (CPALTT01.GY). 35 Tablesandadditionalgures AustraliaAustria1975, 1977-1992, 1995-20081980-1991, 1998-20081975-90, 1992, 1994-20081978-1989,1997-20082/ Final sample determined by availability of all regressors included in the most 1/ Initial sample comprises the same set of OECD countries considered in Perotti (1999), with maximum time period 1975-2008, depending on data availability. Missing observations for individual years within a given time series accounted for by within-year changes in the 34 C.ReinhartandK.Rogoff.Bankingcrises:Anequalopportunitymenace,2008.NBERWorkingPaper14587.L.A.Ricci,G.M.Milesi-Ferretti,andJ.Lee.Realexchangeratesandfundamentals:Across-countryperspective.IMFWorkingPaper08/13,2008.J.RotembergandM.Woodford.Oligopolisticpricingandtheeffectsofaggregatedemandoneconomicactivity.JournalofPoliticalEconomy,100(6):1153–1207,1992.C.Sims.CommentonGlennRudebusch's“Domeasuresofmonetarypolicyinavarmakesense?”.InternationalEconomicReview,39:933–941,1998.A.Sutherland.Fiscalcrisisandaggregatedemand:Canhighpublicdebtreversetheeffectsofscalpolicy?JournalofPublicEconomics,65:147–162,1997.A.Tagkalakis.Theeffectsofscalpolicyonconsumptioninrecessionsandexpansions.JournalofPublicEconomics,92:14861508,2008.M.Woodford.Simpleanalyticsofthegovernmentexpendituremultiplier.AmericanEconomicJournal:Macroeconomics,3:1–35,2011.33 S.KimandN.Roubini.Twindecitortwindivergence?Fiscalpolicy,currentaccount,andrealexchangerateintheU.S.JournalofInternationalEconomics,74(2):362–383,2008.R.Kollmann.Governmentpurchasesandtherealexchangerate.OpenEconomiesReview,21:49–64,2010.P.R.LaneandR.Perotti.Theimportanceofcompositionofscalpolicy:Evidencefromdifferentexchangerateregimes.JournalofPublicEconomics,87:2253–2279,2003.L.LinnemannandA.Schabert.Fiscalpolicyandthenewneoclassicalsynthesis.JournalofMoney,CreditandBanking,35(6):911–929,2003.T.MonacelliandR.Perotti.Fiscalpolicy,therealexchangerateandtradedgoods.EconomicJournal120:437–461,2010.A.MountfordandH.Uhlig.Whataretheeffectsofscalpolicyshocks?JournalofAppliedEconometrics,24(6):960–992,2009.W.NeweyandD.McFadden.Largesampleestimationandhypothesistesting.InR.EngleandD.McFadden,editors,HandbookofEconometrics,volume4,chapter36,pages2111–2245.Elsevier,1994.M.ObstfeldandK.Rogoff.Exchangeratedynamicsredux.JournalofPoliticalEconomy,103:624–660,1995.R.Perotti.Fiscalpolicyingoodtimesandbad.QuarterlyJournalofEconomics,114(4):1399–1436,November1999.R.Perotti.EstimatingtheeffectsofscalpolicyinOECDcountries.mimeo,2004.R.Perotti.Insearchofthetransmissionmechanismofscalpolicy.NBERWorkingpaper13143,2007.V.A.Ramey.Identifyinggovernmentspendingshocks:It`sallinthetiming.QuarterlyJournalofEconomics,126(1):1–50,2011.V.A.RameyandM.D.Shapiro.Costlycapitalreallocationandtheeffectsofgovernmentspending.CarnegieRochesterConferenceonPublicPolicy,48:154–194,1998.M.O.Ravn,S.Schmitt-Groh´e,andM.Uribe.Consumption,governmentspending,andtherealexchangerate.JournalofMonetaryEconomics,59:215–234,2012.C.Reinhart.Thistimeisdifferentchartbook:Countryhistoriesondebt,default,andnancialcrisis.NBERWorkingPaper15815,2010.32 W.Edelberg,M.Eichenbaum,andJ.D.M.Fischer.Understandingtheeffectsofashocktogovernmentpurchases.ReviewofEconomicDynamics,2:166–206,1999.G.B.EggertssonandM.Woodford.Thezerointerest-rateboundandoptimalmonetarypolicy.BrookingsPapersonEconomicActivity,1:139–211,2003.Z.Enders,G.J.M¨uller,andA.Scholl.Howdoscalandtechnologyshocksaffectrealexchangerates?NewevidencefortheU.S.JournalofInternationalEconomics,83:53–69,2011.C.ErcegandJ.Lind´e.Isthereascalfreelunchinaliquiditytrap?CEPRDiscussionPaperNo.7624,2010.C.Erceg,C.Gust,andD.L´opez-Salido.Thetransmissionofdomesticshocksinopeneconomies.InJ.GaliandM.Gerlter,editors,InternationalDimensionsofMonetaryPolicy.UniversityofChicagoPress,2010.J.Fern´andez-Villaverde.Fiscalpolicyinamodelwithnancialfrictions.AmericanEconomicReview:Papers&Proceedings,100:35–40,May2010.J.Gal´andR.Perotti.FiscalpolicyandmonetaryintegrationinEurope.EconomicPolicy,37:534–572,2003.J.Gal´,J.D.L´opez-Salido,andJ.Vall´es.Understandingtheeffectsofgovernmentspendingonconsumption.JournaloftheEuropeanEconomicAssociation,pages227–270,2007.F.GiavazziandM.Pagano.Canseverescalcontractionsbeexpansionary?TalesoftwosmallEuropeancountries.NBERMacroeconomicsAnnual,5:75–111,1990.F.Giavazzi,T.Jappelli,andM.Pagano.Searchingfornon-lineareffectsofscalpolicy:Evidencefromindustrialanddevelopingcountries.EuropeanEconomicReview,44:1259–1289,2000.R.E.Hall.ByHowMuchDoesGDPRiseiftheGovernmerntBuysMoreOutput?BrookingsPapersonEconomicActivity,(2):183–231,2009.J.D.Hamilton.TimeSeriesAnalysis.PrincetonUniversityPress,1994.E.O.Ilzetzki,C.Reinhart,andK.Rogoff.Exchangeratearrangementsenteringthe21stcentury:Whichanchorwillhold?,2009.Mimeo.E.O.Ilzetzki,E.Mendoza,andC.Vegh.Howbig(small?)arescalmultipliers?,2011.mimeo.InternationalMonetaryFund.Fromrecessiontorecovery:Howsoonandhowstrong?WorldEconomicOutlook,April:103–138,2009.31 O.J.BlanchardandR.Perotti.Anempiricalcharacterizationofthedynamiceffectsofchangesingovernmentspendingandtaxesonoutput.QuarterlyJournalofEconomics,117(4):1329–1368,November2002.B.BornandG.J.M¨uller.Governmentspendinginquarterlyandannualtimeseries.JournalofMoney,CreditandBanking,44(2-3):507–517,2012.C.Burnside,M.Eichenbaum,andJ.D.Fisher.Fiscalshocksandtheirconsequences.JournalofEconomicTheory,115:89117,2004.F.CanovaandE.Pappa.Pricedifferentialsinmonetaryunions:Theroleofscalshocks.EconomicJournal,117:713–737,March2007.M.B.Canzoneri,R.E.Cumby,andB.Diba.Newviewonthetransatlantictransmissionofscalpolicyandmacroeconomicpolicycoordination.InM.Buti,editor,MonetaryandFiscalPoliciesinEMU,pages283–311.CambridgeUniversityPress,2003.L.Christiano,M.Eichenbaum,andC.Evans.Theeffectsofmonetarypolicyshocks:Evidencefromtheowoffunds.ReviewofEconomicsandStatistics,78(1):16–34,1996.L.Christiano,M.Eichenbaum,andS.Rebelo.Whenisthegovernmentspendingmultiplierlarge?JournalofPoliticalEconomy,119(1):78–121,2011.J.Cimadomo.Real-timedataandscalpolicyanalysis:asurveyoftheliterature.WorkingPapers11-25,FederalReserveBankofPhiladelphia,2011.J.F.Cogan,T.Cwik,J.B.Taylor,andV.Wieland.NewKeynesianversusoldKeynesiangovernmentspendingmultipliers.JournalofEconomicDynamicsandControl,34,2010.G.CorsettiandG.J.M¨uller.Twindecits:Squaringtheory,evidenceandcommonsense.EconomicPolicy,48:598–638,2006.G.Corsetti,K.Kuester,andG.J.M¨uller.Pegs,oatsandthetransmissionofscalpolicy.CEPRDiscussionpaper8180,2011a.G.Corsetti,A.Meier,andG.J.M¨uller.Fiscalstimuluswithspendingreversals.ReviewofEconomicsandStatistics,forthcoming,2011b.G.Corsetti,K.Kuester,A.Meier,andG.J.M¨uller.Sovereignrisk,scalpolicy,andmacroeconomicstability.IMFWorkingpaper12/33,2012.R.Dornbusch.Exchangerateeconomics:Wheredowestand.BrookingsPapersonEconomicActivity,1:143–185,1980.30 References.Acconcia,G.Corsetti,andS.Simonelli.Maaandpublicspending:Evidenceofthescalmultiplierfromaquasi-experiment.mimeo,2011.A.AfonsoandJ.Jalles.Assessingscalepisodes.TechnicaluniversityofLisbonWorkingPaper15/2011,2011.A.AuerbachandY.Gorodnichenko.Measuringtheoutputresponsestoscalpolicy.AmericanEconomicJournal:EconomicPolicy,forthcoming,2010.A.AuerbachandY.Gorodnichenko.Fiscalmultipliersinrecessionandexpansion.NBERWorkingpaper17447,2011.D.Backus,P.Kehoe,andF.Kydland.Dynamicsofthetradebalanceandthetermsoftrade:theJ-curve?AmericanEconomicReview,84(1):84–103,March1994.E.Baldacci,S.Gupta,andC.Mulas-Granados.Howeffectiveisscalpolicyresponseinsystemicbankingcrises?IMFWorkingPaperWP/09/160,2001.R.BarroandC.J.Redlick.Macroeconomiceffectsfromgovernmentpurchasesandtaxes.QuarterlyJournalofEconomics,126:51–102,2011.M.BaxterandR.G.King.Fiscalpolicyingeneralequilibrium.AmericanEconomicReview,83(3):315–334,June1993.R.Beetsma,M.Giuliodori,andF.Klaasen.Tradespill-oversofscalpolicyintheEuropeanUnion:apanelanalysis.EconomicPolicy,48:640–687,2006.R.Beetsma,M.Giuliodori,andF.Klaasen.TheeffectsofpublicspendingshocksontradebalancesintheEuropeanUnion.JournaloftheEuropeanEconomicAssociation,6(2-3):414–423,2008.R.Beetsma,M.Giuliodori,andF.Klaassen.Temporalaggregationandsvaridentication,withanapplicationtoscalpolicy.EconomicsLetters,105:253–255,2009.A.S.B´en´etrixandP.R.Lane.Fiscalshocksandtherealexchangerate,2009.IIISDiscussionPaper286.G.BertolaandA.Drazen.Triggerpointsandbudgetcuts:Explainingtheeffectsofscalausterity.AmericanEconomicReview,83(1):11–26,1993.F.O.Bilbiie,A.Meier,andG.J.M¨uller.WhataccountsforthechangesinU.S.scalpolicytransmission?JournalofMoney,Credit,andBanking,40(7):1439–1469,2008.29 6Conclusionconomictheorysuggeststhatthetransmissionofscalpoliciesmayvaryacrosseconomicenvironments—thereisnosinglemultiplierdescribingtheeffectivenessofscalstimulus.However,empiricalevidenceonthedeterminantsofscaltransmissionhasremainedpatchy.Inthispaperwetakeasteptowardasystematicempiricalanalysisofhowdifferenteconomicconditionsaffectthetransmissionofgovernmentspendingshocks.Specically,weproposeatwo-stageproceduretoestimateimpulseresponsefunctionsforeconomiesthatdifferintermsoftheirexchangerateregime,thehealthoftheirpublicnances,andthestateoftheirnancialsystems.Whileourndingsconrmstandardestimatesofaveragescalmultipliers,theyalsosuggestthattheseaveragesmasksubstantialdifferencesacrosseconomicenvironments.Twomainresultsstandout.First,wendthattherealexchangerateresponsetoaspendingshockvariessystematicallywiththeexchangerateregime.Thisunderscorestheimportanceofinteractionsbetweenscalandmonetarypolicy—aninterestingavenueforfutureresearch,especiallyinanopeneconomycontext.Ourndingalsoshedsnewlightonthedebateaboutscaltransmissionviainternationalprices.Specically,severalrecentstudieshavedocumentedexchangeratedepreciationafteraspendingincrease,contrarytothepredictionofstandardtheoreticalmodels.ThesestudiestendtofocusonsamplesincludingAustralia,theUnitedStates,andtheUnitedKingdom,i.e.,countrieswithoatingexchangerateregimes,whereasotherstudiesndingrealappreciationtypicallyconsidersamplesdominatedbyeuroareacountries.Ouranalysisindicatesthatthesedifferencesmaybesystematicallyrelatedtodistinctexchangerateregimes.Asecondndingwithdirectpolicyimplicationsconcernsthemarkedincreaseinscalmultipliersduringtimesofnancialcrisis.Ontheonehand,thismaybetakenasevidenceinsupportofscalstimulusduringtimesofacutenancialstress.Ontheotherhand,ourempiricalresultsalsoshowthatmanycountrieshavehistoricallycutbackgovernmentspendingduringnancialcrises,presumablyoutofconcernoverdebtsustainability.Inthissense,alargeconditionalmultiplierprovidesastarkwarningaboutthecostsofnancialturmoilandanargumentinfavorofbuildingupscalbuffersinnormaltimessoastoavoidscalretrenchmentwhenitismostpainful.28 inspendingdecidedinanticipationofcyclicalmovementscouldbemistakenforpolicyinnovations.However,thereisalsoapotentialerrorworkingintheoppositedirection:supposethatspendinghasnoeffectonoutput(themultiplieriszero),butpolicymakersbelieveittobeaneffectivestabilizationinstrument.Inthiscase,includingtheCLIintherststepmayinduceaspuriouscorrelationbetweenscalpolicyandoutput,evenifthereisnone.39Thesignofsuchabiasinthemultiplierwoulddependonthecyclicalpropertiesofspending,forwhichtheevidenceismixed.Inanycase,Figure9suggeststhatourmainresultsarenotsensitivetoomittingtheCLIfromtherststep,withtheexceptionoftheresponseofGDPanditscomponentsunderapeg.Aswasthecaseforthemorerestrictivedenitionofhighdebt(Figure5),scalmultipliersappeartobesmallerunderapeg,whiletheresponseofthetradebalanceandtheexchangerateremainsimilartotheevidenceinFigure2.Next,weconsiderresultsforalternativetimedimensionsofoursample.First,welimitthesampletoendin2006inordertoexcludeobservationspertainingtotheglobalnancialcrisis(Figure10).Indeed,theresponsetimeofscalpolicyappearstohavebeenunusuallyshortduringthislatestepisode,potentiallyjeopardizingouridenticationstrategy.Inaseparate(unreported)robustnesscheck,weextendoursampletotheearly1970s,thusincludinganumberofobservationsfromthelatestagesoftheBrettonWoodsperiod.Neitheroftheseexperimentsgeneratesappreciabledifferencesrelativetoourresultsforthebaselinesample.Thus,theobservationsforthersttwoyearsoftheglobalnancialcrisisdonotseemtobeinuentialindrivingthedifferentmultipliersweestimate.Bythesametoken,thetransmissionofscalpolicydoesnotappeartobespecictotheinternationalmonetarysystemthatemergedafterthedemiseofBrettonWoods.Weconcludeoursensitivityanalysiswithoneadditionalvariationofthebaselinesetup.ResultsshowninFigure11arebasedonestimateswheretheUnitedStatesareexcludedfromthesample.ThisaddressesconcernsthattheUnitedStatesareapossibleoutlierwithrespecttotheeffectsofscalpolicy.Intheevent,wendthatresultsarebroadlyunchangedrelativetoourbaselinesetup. 39ethankMikeWoodfordforraisingthisissue.27 impactofscalshocksunderapeg.Asapparentfromtherightcolumn,thedifferenceintheimpactresponsesofoutputandinvestmentacrosscurrencyregimesislargerthaninFigure2.Figure6reportsimpulseresponsesforthecaseofastricterdenitionofweakpublicnances,withtherelevantthresholdsfordebtandthedecitraisedto120percentand7percentofGDP,respectively.Inthiscase,theresultsconditionalonnancialcrisisandweakpublicnancesarenotsubstantiallyaffected.Themostapparentchangeconcernsthecurrencyregime.AsshownbytherightcolumnofFigure6,ourmainresultregardingoutputandinvestmentisreversed:thepointestimatesoftheirimpactresponsesarenowlowerunderapeg,relativetoaregimeofexibleexchangerates.Incontrast,theresponsesofexternalvariables,i.e.,therealexchangerateandthetradebalance,aresimilar,ifnotstronger,thaninFigure2.Therealexchangerateappreciatesonimpact,andthetradedecitworsens.Figure7displaysresultsforthemodelspeciedindifferences,ratherthanlevels.Toallowacomparisonwiththeresultsobtainedabove,weconsiderthecumulativeeffectofagovernmentspendingshock.Overall,theresultsfromthelevelspecicationturnouttobequiterobustwithrespecttothisalternativespecication.Unexpectedspendingincreasesareagainfoundtobeconsiderablymoreeffectiveduringnancialcrisesthaninnormaltimes.Theoutputandconsumptionmultipliersremainaround2,investmentrisesovertime,netexportsdeteriorate,andtherealexchangeratedepreciates.Similarly,Figure7conrmsourpreviousconclusionsregardingtheeffectsofscalpolicyconditionalonweakpublicnancesandanexchangeratepeg.Ifanything,resultsunderapeg,especiallythoseconcerningtheoutputmultiplier,aremoreclearlyinlinewiththereceivedwisdom.Figure8reportsimpulseresponsesforaspecicationwhereweincludethecontemporaneousratherthanthelaggedvalueofthecrisisdummyintherststep.Thisamountstoassumingthat,duringcrises,thegovernmentcanconsiderablyshortendecisionandimplementationlagsregardingitsplannedconsumptionexpenditure,allowingforasystematicwithin-yearresponseofgovernmentspendingtoanancialcrisis.Whilethishypothesisishardlysupportedbythedata,ourexercisecanshedlightontheextenttowhichourmainndingsmaybedrivenbyashortcominginouridenticationstrategyspecictocrisisepisodes.Asapparentfromthegure,resultsarequitesimilartothoseobtainedunderthestandardspecication,reportedinFigures2through4.38InFigure9,inturn,wedisplaytheresultsobtainedifweexcludethelaggedvaluesoftheOECD'sCompositeLeadingIndicatorfromtherststep.Theinclusionofthisvariableinourmodelismeanttoaccountforanyadvanceinformationaboutcyclicaldevelopmentsintheeconomywhichthegovernmentmayuseinpreparingitsnextbudget.Omittingitwouldmeanthatsomechanges 38relatedconcernisthatourcrisisdummymaybetoocrudeameasureofnancialdislocations.Assomecrisesaredeeperthanothers,andsomecrisisyearsinagivencrisisepisodeareworsethanothers,includingthecrisisdummymaynotpreventourestimatedshocksfromcontainingsomesystematicscalresponsetothecrisis.Toassessthisissue,weestimatedthemodelexcludingallthosecountriesforwhichweestimateanegativecrisiscoefcientintherststep.Resultsfromthisexercise(notshownduetospaceconstraints)donotbearoutthisconcern.26 inducedpolicymakerstoincreaseexhaustivegovernmentspending.Asshownabove,thecrisisdummyinourrst-stepregressiontakesonanegativesignforseveralcountries,probablyreectingconcernsaboutnancingorscalsustainability.Indeed,itseemsplausiblethatscalsustainabilityisacrucialpreconditionforobtaininghighpositivemultipliersattimesofnancialcrisis(asdiscussedinCorsettietal.2012).Inthatsense,ourndingsalsocontainawarningthatanancialcrisiscanbeevenmoredamagingifitforcesthegovernmenttoretrenchspendinginthemidstofthedownturn.Thisclearlyunderscoresthecaseforpreservingandstrengtheningscalbuffersingoodtimes.Analcaveatisinorder:thenumberofobservationsofnancialcrisesisnecessarilylimitedinoursample,andalthoughwefollowtheclassicationofReinhartandRogoff(2008),notallofthesenancialcrisisepisodesareidenticalorevensimilarintheirnatureanddepth.Nevertheless,asweshowinthefollowing,ourmainresultsappearquiterobusttochangesinthedenitionsofourdummies,includinganarrowerdenitionofnancialcrisis,andthespecicationofourempiricalmodel.5.3SensitivityanalysisWeexploretherobustnessofourresultswithrespecttoalternativedenitionsofthedummyvariablesandmodelspecications.ResultsareshowninFigures5through11intheappendix,whereeachcolumndisplaysresultsforoneofthreepossibledeparturesfromthebaselinescenario,i.e.nancialcrisis,weakpublicnances,andanexchangeratepeg.StartingwithFigure5,wedisplayresultsobtainedunderanarrowerdenitionofnancialcrisis,includingonlytheso-called“bigve”bankingcrises(Finland1991–94,Japan1992–97,Norway1988–93,Spain1978–85,andSweden1991–94)plustheglobalnancialcrisisstartingin2007.Clearly,anarrowerdenitionofnancialcrisishastheadvantageofselectingcriseswithacomparablylargemagnitude,albeitatthepriceofreducingthenumberofobservationsconsiderably.Ineachcolumnwealsoreportresultsforthebaselinescenario(nopeg,lowdebt,nocrisis).Generally,ourndingsarenotmuchalteredrelativetothoseobtainedunderthewiderdenitionofnancialcrisis.Theimpulseresponsesconditionalonanancialcrisisareshownbythepanelsintheleftcolumn.WhiletheoverallresultsofFigure4areconrmed,therearethreenotabledifferences.First,governmentspendingitselfappearstobelesspersistent.Second,whiletheimpactresponseforoutputandconsumptionremainscloseto2,overallactivityrevertstothetrendmorequickly.Third,theresponsesofinationandtherealexchangeratearestronger,matchedbyasharperriseintheshort-terminterestrate.Anarrowerdenitionofthecrisisdummydoesnotaffectmultipliersconditionalonhighdebt,showninthesecondcolumnofFigure5.37Bycontrast,itappearstostrengthenourresultsconcerningthe 37ote,however,thatthepuzzlingresultofadelayedriseinconsumptiondisappearsinthisspecication.25 GovernmentspendingOutputConsumptionInvestment 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -5 0 5 NetexportsRealexchangerateInationInterestrate 0 2 4 6 -10 -5 0 5 0 2 4 6 -30 -20 -10 0 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 6 Figure4:Impulseresponsestogovernmentspendingshock:baselinescenariovsnancialcrisis.Notes:seeFigure2.expansion(aboveandbeyondwhatisexplainedbythesystematicresponseofspendingtocyclicalconditions)duringtimesofnancialcrisis.Ourresults,showninFigure4,suggestthattheresponseofconsumptionandoutputtoascalexpansionispositiveandlargeonceweconditionontheoccurrenceofanancialcrisis:consumptionandoutputriseabouttwiceasmuchastheinitialincreaseingovernmentspending.Correspondingly,thetradebalancedeterioratessignicantlyandpersistently.Theresponseofinvestmentisinitiallymuted,butappearstostrengthenovertime.36Duringanancialcrisis,aspendingexpansionappearstobeassociatedwithmoreinationandcurrencydepreciationthaninthebaselinescenario.Mostimportantly,itpromptsasharp,albeittemporary,riseininterestrates.Inlightofthestrongdepreciationresponseoftheexchangerate,thesepatternsmaypointtoapolicymodelwherebythecentralbank,evenattimesofnancialcrisis,specicallytargetscurrencystability.Ontheonehand,ourndingssupporttheargumentthathigherscaloutlaysareparticularlyeffectiveasastabilizationtoolinanancialcrisis,perhapsbecauseagentsareliquidity-constrained.Infact,therecentpolicydebateintheUnitedStateshasfeaturedrepeatedclaimsthatscalmultiplierscouldbeaslargeastwoundercurrentconditions.Ourestimatesareconsistentwithamultiplierofthismagnitude,eventhoughourresultsfor“normaltimes”donotsuggestlargescalmultipliersatall.Ontheotherhand,itisworthstressingthatnotallhistoricalnancialcrisesinoursamplehave 36earenotawareofempiricalstudiesattemptingtoquantifytheimpactofnancialcrisesonscalpolicytransmission.Onarelatedtheme,however,Tagkalakis(2008)ndsthatrecessions,morebroadlydened,tendtoraisethegovernmentspendingmultiplieronconsumption.Ilzetzkietal.(2011)andAuerbachandGorodnichenko(2010,2011)alsondstrongeroutputmultipliersduringrecessions.24 GovernmentspendingOutputConsumptionInvestment 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -5 0 5 NetexportsRealexchangerateInationInterestrate 0 2 4 6 -10 -5 0 5 0 2 4 6 -30 -20 -10 0 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 6 Figure3:Impulseresponsestogovernmentspendingshock:baselinescenariovssituationofweakpublicnances.Notes:seeFigure2.dashedlinesinFigure3showtheeffectsofgovernmentspendingshocksforsuchaneconomy,whilethesolidlinesdisplaytheresponsesunderthebaselinescenarioofsoundpublicnances.Wendthat,inthecaseofweakpublicnances,theimpactresponseofoutputandinvestmentislowerrelativetothebaselinescenario.Contrastingwithourtheoreticalpriors,however,thereisnodifferenceintheimpactresponseofconsumption.Moreover,theresponseofconsumptionturnspositiveovertime.Meanwhile,inationandtheinterestratefollowsimilarpatternsasinthebaselinescenariowithsoundpublicnances,buttheirresponsesaremorepronounced.Boththeinitialdeclineininationandthesubsequentpeakarelarger,andthemonetarystanceislooserthroughout.Conversely,theexchangeratedoesnotrespondonimpactanddepreciatesbylessovertime.Asdifferencesrelativetothebaselinearenonethelessmoderate,withtheexceptionofthepuzzlingdelayedincreaseofconsumption,ourresultsarelikelytoreectthefactthathighpublicdebtandlaggeddecitsarenotsufcientlypreciseindicatorsforthekindofscalstressthatwewouldexpecttoaffectthescaltransmissionmechanism.Japan'slongexperienceoflivingwithveryhighandrisingpublicdebtandyetverylownancingcoststestiestotheempiricalchallengefacingresearchinthisarea.35Financialcrisis.Thelastsetofresultspertainstothedistinctionbetweennormaltimesandtimesofnancialcrisis.Thepossibilitythatanancialcrisisaffectsscalpolicymakingisalreadycapturedinourrst-stagespecication.Theanalysisinthissectionfocusesontheeffectofagivenscal 35heanalysisofCorsettietal.(2012)alsosuggeststhathighdebthasadecisiveeffectonscaltransmissiononlyifmonetarypolicyisconstrained.23 accommodativeunderaxedexchangerate:theshortraterisesonimpactandremainspositiveforsomeextendedperiod.Correspondingly,wendthattherealexchangerateappreciatesunderapeg.Itisworthstressinginthisregardthatthetypicaltextbooknotionofmonetarypolicybeingmoreaccommodativeunderapegisnotageneralpredictionofstandardopeneconomymodels,whetherclassic(Dornbusch1980)ormoremodern(Corsettietal.2011a).Moreover,standardanalysisoftenproceedsundertheimplicitassumptionthatthecurrentpegisperfectlycredible—anassumptionthatisunlikelytobesatisedinpractice.Withimperfectcredibility(assuggestedbythehistoricalrecordcontainedinoursample),agovernmentspendingexpansionmaygeneratetensionsinthecurrencymarket,promptingthecentralbanktodefendthecurrencywithanincreaseintheinterestrate.ThemonetaryresponsetotheidentiedgovernmentspendingshocksalsomarksakeydifferencebetweenourempiricalresultsandthoseofIlzetzkietal.(2011).Thelatterstudy,usingadifferentmethodologyandadifferentsample,ndstheresponseofthepolicyratetobelessaccommodativeunderaexibleexchangeratethanunderapeg.Consistentwiththatnding,Ilzetzkietal.(2011)reportthattherealexchangeratetendstostrengthenunderaoat,notablycomparedtotheresponseunderapeg.Thepatternwedetectinoursampleistheopposite.Giventheimportanceofscalandmonetaryinteractionsthesedifferencesclearlysuggestacaseforfurtheranalysis.Nonetheless,ourresultsregardingtherealexchangerateshedlightontheseeminglycontradictoryconclusionsofearlierempiricalstudiesthat,likeours,focusonOECDcountries.Someofthesestudiessuggestarealdepreciation,somearealappreciation,inresponsetopositivegovernmentspendingshocks.Muchofthisdivergenceintheliteraturecouldbeexplainedinlightofourresultthatthesignoftherealexchangerateresponseappearstodependontheexchangerateregime.34DifferentexchangerateregimescouldalsoaccountforthemixedevidenceacrossVARstudiesasregardstheimpactresponseofnetexports,pointingtointerestingdirectionsforfutureresearchintotheinteractionofscalandmonetarypolicy.Inthiscontext,wenotethatourndingsforthebehaviorofexternalvariablesunderapeg—realappreciationandatradedecit—arequiterobusttoalternativespecicationsofourmodel,asdiscussedbelow.Weakpublicnances.Wenowconsideraseconddimensionwhichmaybecriticalforhowgovernmentspendingaffectstheeconomy,i.e.,thehealthofpublicnances.SimilartoPerotti(1999),wedenecountriesashavingweakpublicnanceswhenbeginning-of-periodgrossgovernmentdebtexceeds100percentofGDPand/orlaggednetgovernmentborrowingexceeds6percentofGDP.The 34nadditiontoBeetsmaetal.(2008)andB´en´etrixandLane(2009),anumberofstudieshavedocumentedrealappreciationincountries/regionswithaxedexchangerate.IntheiranalysisofU.S.statesandEMUmembercountries,CanovaandPappa(2007)ndthatgovernmentspendingshocksraisethepricelevelrelativetothepricelevelintherestoftheunion.RelyingonastaticregressionmodelestimatedondataforOECDcountries,LaneandPerotti(2003)ndthatanincreaseinthegovernmentwagebillappreciatesthenominalexchangerate.Ricci,Milesi-Ferretti,andLee(2008)alsodocumentapositiveassociationbetweengovernmentconsumptionandtherealexchangerateforapanelof48countries.22 GovernmentspendingOutputConsumptionInvestment 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -4 -2 0 2 4 6 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -5 0 5 NetexportsRealexchangerateInationInterestrate 0 2 4 6 -10 -5 0 5 0 2 4 6 -30 -20 -10 0 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 6 Figure2:Impulseresponsestogovernmentspendingshock:baselinescenariovspeg.Notes:quantitiesmeasuredinoutputunits;solidline:pointestimateforbaselinescenario(shadedarea:onestandarderror);dasheddottedline(dashedlines:onestandarderror):deviationfrombaselineunderexchangeratepeg.sampleperiod.Ilzetzkietal.(2011)alsondnosignicantoutputeffectofgovernmentspendingunderexibleexchangerates.Exchangeratepeg.Accordingtoconventionalwisdom,scalpolicytransmissionvariessystem-aticallyacrossexchangerateregimes.Togetherwithresultsforthebaselinescenario,Figure2alsodisplaystheestimatedimpulseresponsesandcondenceintervalsforthecaseofacurrencypeg(assumingsoundpublicnancesandnonancialcrisis);thesearedenotedbythedashed-dottedanddashedlines,respectively.FromFigure2itisapparentthattheoutputmultiplierispositiveandlargerthanintheeconomywithaexibleexchangerate,consistentwiththetextbookMundell-Flemingmodel.Inthisdimension,ourndingsmatchthosebyIlzetzkietal.(2011),whoalsoreportthattheoutputeffectsofgovernmentspendingshocksaresignicantandsizeableunderxedexchangerates.Furtherconrmingconventionalwisdom,netexportsareseentodecrease,whiletherealexchangerateappreciates.33However,therearealsosomestrikingdifferencesacrosscurrencyregimeswhichdonotsquarewellwiththetextbookMundell-Flemingmodel.First,notethattheimpactresponsesofconsumptionandinvestmentarenotappreciablydifferentunderapegrelativetothecaseofamoreexibleexchangerate.Second,andmoreimportant,whiletheresponseofinationexhibitsthesamepatternacrossbothregimes(withaninitialdropfollowedbyarise),monetarypolicyappearstobemuchless 33otethatweclassifyasanexchangeratepegallcurrencyregimesuptoandincludingcategory8inIlzetzkietal.(2009),thatis,a“defactocrawlingbandthatisnarrowerthanorequalto+/-2percent;”evenanominalpeg,therefore,allowsforsomelimitedexchangerateexibility.21 thusestimateequation(2)forourentirenalsample,includingallrelevantinteractionterms.Figures2to4representtheresultsgraphically.Foreachdimensionoftheeconomicenvironment,wereportimpulseresponsesforallvariablesofinterest,comparingdynamicsrelativetoacommonbaselinescenario.Althoughthisfocusesthediscussionononlyoneconditioningfactoratatime,itisworthstressingthattheresultspertaintoacomprehensivespecicationthatsimultaneouslycontrolsforallthreedimensionsweareinterestedin.Thisisimportantinsofarasvariationsalongthesedimensionsarenotnecessarilyorthogonaltoeachother.Forthepurposesofourcharts,weconditiononanidenticalenvironmentinthetwodimensionsthatarenotthefocusoftherespectivechart.Baselinescenario:exibleexchangerate,soundpublicnances,nonancialcrisis.InFigures2through4,wereportresultsforthemarginaleffectsofagovernmentspendingshockforabaselinescenariodenedbyaneconomythatdoesnotpegitscurrency,ischaracterizedbyrelativelysoundpublicnances(debtbelow100percentofGDPandlaggednetborrowingbelow6percentofGDP),anddoesnotexperienceanancialcrisis.Ineachgure,theimpulseresponsesforthisbaselinescenarioaredenotedbyasolidline,thecondenceintervalsbyshadedareas.Wendtheresponseofoutputtobeessentiallyzerointhiscase;thereismildcrowdingoutofconsumption,andpronouncedcrowdingoutofprivateinvestment.Comparedtotheunconditionalndingsdiscussedabove,theseresultssupportthenotionthat,inaneconomywithexibleexchangeratesinnormaltimes,themacroeconomiceffectsofagovernmentspendingexpansionaregenerallyweak.IncontrasttotheunconditionalndingsshowninFigure1,theresponseofnetexportsisnolongernegative.Thiscontradictsconventionalwisdom,butismoreinlinewiththeevidencereportedfortheUnitedStatesbyKimandRoubini(2008).Observefurtherthatinationfallsonimpact,butrisesvisiblyovertime.Still,monetarypolicyremainsquiteaccommodative:theresponseoftheshort-terminterestrateisactuallynegativebothonimpactandforsomequartersthereafter.Correspondingly,therealexchangeratedepreciationismorepronouncedandpersistentthaninFigure1,sharpeningthecontrastwiththeconventionalwisdom.Asdiscussedabove,thisndingmatchesagrowingbodyofevidence,butcallsintoquestionthetransmissionmechanismofmanytheoreticalmodels.Specically,conventionalwisdomholdsthatgovernmentspendingcausesrealappreciationandthusdecreasesnetexports,implyingamutedoveralloutputresponse.Ourempiricalresultsareatoddswithsuchaview.Whileoutputisessentiallyat,netexportsdonotdeteriorate,andtherealexchangeratedepreciates.Fromatheoreticalpointofview,itisimportanttostressthattheconventionalwisdomreliesonspecicassumptionsaboutthestanceofmonetarypolicythataretypicallyleftimplicit.Atthesametime,severalrecentVARstudiesonthebasisofU.S.timeserieshavealsoreportedamutedorevennegativeresponseofoutput;seeMountfordandUhlig(2009)ortheresultsinRamey(2011)andPerotti(2004)foramorerecent20 thenotionof“twindecits,”denedasaconditionalpositivecorrelationbetweenbudgetandcurrentaccountdecits.FortheUnitedStates,however,KimandRoubini(2008)ndthat,inresponsetoscalshocks,thetradebalanceandthegovernmentbudgetactuallymoveinoppositedirections,aphenomenontheauthorsdub“twindivergence.”SubsequentworkbyCorsettiandM¨uller(2006)andMonacelliandPerotti(2010)extendstheanalysistocountriesotherthantheUnitedStates.Overall,itappearsthatexpansionaryscalmeasuresworsenthetradebalance,thoughnotstronglyso.Inthisregard,ourestimationmethodologymayhelpshedlightonwhetherthereisanysystematicpatternlinking“twindecits”or“twindivergence”toparticularfeaturesoftheeconomy.ThesecondrowofFigure1alsodisplaystheresponsesofination,theshort-termnominalinterestrate,andtherealexchangerate.Inationdropsonimpact,butthenrisestemporarilyabovebaseline.Theshort-termdeclineininationafteranincreaseofgovernmentspendingissurprisinginlightofpredictionsofstandardtheoreticalmodels.However,ithasbeendocumentedundervariousalternativeidenticationschemes;seePerotti(2004)andMountfordandUhlig(2009).Theshort-terminterestrate,inturn,risesonimpactandthenmovesintandemwithination.Finally,therealeffectiveexchangerateappreciatesslightlyonimpact,butthendepreciatesovertime.Whiletheinitialappreciationconformswellwithconventionalwisdom,itcontrastswiththeevidencefromafewrecentstudieswhichdocumentafallintherealexchangerateafterariseingovernmentspending.30However,thestudiesreachingthisconclusiontypicallyfocusoncountriesfeaturingexibleexchangerateregimes.Incontrast,studiesfocusingoneuroareacountriestendtodocumentrealappreciationinresponsetopositivespendingshocks.31Thesecontrastingndingssuggestthatthedynamicresponseoftherealexchangeratemaydiffersystematicallyacrosscurrencyregimes.Ourmethodologyisagainwellsuitedtoexaminethisissuemoreclosely.Indeed,thepremiseofourstudyisthatndingsbasedonsimpleunconditionalmodelsmayconceallargedifferencesintheresponsetoscalshocksacrosseconomicenvironments.32Withthisinmind,weturnnexttotheresultsforourfullsecond-stepspecication,whichaccountssimultaneouslyforseveraldimensionsoftheeconomicenvironment.5.2AccountingfortheeconomicenvironmentThethreedimensionsoftheeconomicenvironmentwhichwearguearelikelytohaveabearingongovernmentspendingmultipliersincludetheexchangerateregime,thestateofpublicnances,andthehealthofthenancialsector.WhilethedenitionsofthesecharacteristicsareprovidedinTable2,Table5givesasummaryofwhichcountry-yearobservationstundereachofthesecategories.We 30imandRoubini(2008)andEndersetal.(2011),forinstance,reportadepreciationoftheREERfortheUnitedStates.MonacelliandPerotti(2010)ndadepreciationforAustralia,theUnitedStates,andtheUnitedKingdom.31See,forexample,Beetsmaetal.(2008)andB´en´etrixandLane(2009).32FurthersupportforthispointcomesfromthendingofPerotti(2004)thattheeffectsofgovernmentspendingontheeconomytendtobemoremutedafter1980intheUnitedStates.19 GovernmentspendingOutputConsumptionInvestment 0 2 4 6 -0.5 0 0.5 1 1.5 2 0 2 4 6 -2 -1 0 1 2 0 2 4 6 -1 -0.5 0 0.5 1 0 2 4 6 -2 -1 0 1 NetexportsRealexchangerateInationInterestrate 0 2 4 6 -2 -1 0 1 0 2 4 6 -6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 0 2 4 6 -4 -2 0 2 4 6 Figure1:Impulseresponsestogovernmentspendingshock:unconditionaleffects.Notes:quantitiesmeasuredinoutputunits;solidline:pointestimate(shadedarea:onestandarderror).asgovernmentspending,theresponsesrepresentestimatesofthegovernmentspendingmultipliersonoutputandconsumption,respectively.Ourresultsfortheaveragemultipliersarewellwithintherangeofresultsreportedintheliterature.29AsrecentlysummarizedbyHall(2009),“empiricalworkusingsimpleregressionsormoreelaborateVARsndsoutputmultipliersintherangefrom0.5to1.0,withafewexceptions,andconsumptionmultipliersintherangefromsomewhatnegativeto0.5.”Asregardstheconsumptionmultiplier,forinstance,studiesadoptingtheBlanchard-Perottiidenticationtypicallyreportapositiveestimate,althoughmorerecentevidencebasedonthisapproachsuggestsadeclineinthemultiplierovertime;seePerotti(2004)andBilbiieetal.(2008).StudiesdrawingontheRamey-Shapiroapproach,inturn,oftenreportanegativemultiplierforconsumption,buttheeffectistypicallyquitecontainedinquantitativeterms.Lastly,whileusingdifferentidenticationstrategies,neitherMountfordandUhlig(2009)norBarroandRedlick(2011)orHall(2009)ndasignicantresponseofconsumptiontoadecit-nancedgovernmentspendingshockoranincreaseindefensespending,respectively.TurningtothelastpaneloftherstrowofFigure1,wendthatgovernmentspendingreducesprivateinvestment.Afallininvestmentisaclear-cutresultdocumentedinvariousstudies,includingBlanchardandPerotti(2002),MountfordandUhlig(2009),andRamey(2011).Theestimatedresponseofthetradebalancesuggeststhatapositivespendingshockalsotriggersadeclineinnetexports.Thisagainsquareswellwithconventionalwisdomandearlierempiricalstudies,supporting 29otethatwemeasuretheabsoluteresponseofthevariableofinterestinpercentofoutput,givenaninitialincreaseofgovernmentspendingbyonepercentofGDP.Analternativemeasureofthemultiplierdividesthecumulativeresponseofthevariableofinterestbythecumulativeresponseofgovernmentspending,asthelatterrespondsendogenouslytotheshock.Forourresultsbelow,weobtainverysimilarresults,notablyregardingtheeffectoftheeconomicenvironment,ifwecomputecumulativemultipliers.18 5Theeffectsofgovernmentspendingshockssdetailedintheprevioussections,ourmodelismeanttocapturethedynamicresponseofkeymacroeconomicvariablestoagovernmentspendingshock.Specically,weconsidertheresponsesofeightvariablesofinterest:outputanditscomponents(privateconsumption,privatexedinvestment,andthetradebalance),therealeffectiveexchangerate,CPIination,theshort-termnominalinterestrate,and,ofcourse,governmentspendingitself.Whileoursecond-stepregressionforexpenditureaggregatesisspeciedinloglevels(ortheratiotoGDPinthecaseofthetradebalance),wetransformtheresultssoastoallowasimpleinterpretationintermsofpercentagepointsofGDP.Thebehaviorofeachvariableistracedforsixyearsaftertheimpact.5.1UnconditionaleffectsAlthoughourprimaryobjectiveistoanalyzetheeconomy'sresponsetoaspendingshockconditionalondifferenteconomicenvironments,itisusefultobeginouranalysisbypresentingunconditionalestimates,i.e.,thoseobtainedwithoutregardforanydifferencesacrossenvironments.Intherelevantregressionsforboththerstandthesecondstep,wethusomitalloftheeconomicenvironmentdummieslistedinTable5.26Theunconditionalestimatesnotonlyprovideabenchmarkagainstwhichtoassessthemaincontributionofourpaper,butalsoallowustoverifythatourtwo-stepprocedureproducesaverageresultssimilartothosetypicallyreportedintheliterature.Figure1providesagraphicalrepresentationoftheresultsintermsofimpulseresponsefunctions.Thesolidlinesindicatethepointestimate,whiletheshadedareasrepresentone-standard-deviationcondenceintervals.27Thehorizontalaxismeasuresthetimeaftertheshockinyears,theverticalaxismeasuresdeviationsfromtrend.GDPanditscomponentsaremeasuredinoutputunits,whiletherealexchangerateismeasuredinpercent;theinterestrateandinationaremeasuredinpercentagepoints.WenormalizeimpulseresponsessothattheinitialincreaseingovernmentspendingisonepercentofGDPunderthebaselinescenariodenedbelow.28AsshownintherstrowofFigure1,wendapersistentincreaseingovernmentspendingandasizeableincreaseinaggregateoutputbyabout0.7percentagepointsonimpact,whereasconsumptiondoesnotrespondatall.Astheresponsesofoutputandconsumptionaremeasuredinthesameunits 26oconservespace,wedonotreporttheresultsfromtherst-stageestimationforthismoreparsimoniousspecication.TheresultsareinlinewiththoseshowninTable3.Inparticular,governmentspendingexhibitsnoclearcyclicalpattern,butrespondsnegativelytothelevelofpublicdebt.27Wecomputestandarderrorsbydrawing1,000realizationsofthecoefcientvectorassumingamultivariatenormaldistributionwithvariance-covariancematrixcorrespondingtothatoftheregressioncoefcients;themeanissetequaltothepointestimate.28Typically,thisrequiresconsideringanestimatedshockslightlygreaterthanonepercentofGDP,asthepointestimateoftheresponseofgovernmentspendingtothestandardizedshocktendstobearound0.9only.Thisprobablyreectsamoderateattenuatingbiasinthesecondstepcausedbymeasurementerrorintherststep.InordertoexpressGDPanditscomponentsinoutputunitsweusetheaverageexpendituresharesinoursampleperiod.17 stimuluspoliciesinadvancedcountriesduringthemostrecentcrisis,therelevantcoefcientisestimatedtobenegative,andsometimessignicantlyso,innineoutofthirteencountries.Thisimpliesthatgovernmentspendingwouldslowdown,ratherthanaccelerate,duringnancialcrises(holdingeverythingelsexed).Whileperhapssurprising,thispointstothefactthatwhathasbeenconsidereddesirableduringthelatestglobalnancialcrisisof2008–09,i.e.,disproportionatelystrongcountercyclicalscalstimulus,isnotnecessarilywhatcountrieshavefoundopportuneduringpreviouscrises.Akeyreasonforthismaybeconcernsaboutnancingconstraints,especiallywhenbankingsectorbailoutsandfallingtaxrevenuealreadyputasignicantdentintothepublicnances.Next,weturntotheprimaryoutputofinterestprovidedbytherst-stepregression,i.e.,theestimatedscalpolicyshocks.Althoughourchoiceofanempiricalpolicyruleismotivatedbytheoreticalconsiderationsandpreviouscontributionsintheliterature,itsappropriatenessneedstopassstatisticaltests,too.Specically,iftheresidualsaresupposedtobereliablemeasuresofunanticipatedspendingchanges,oneobviousrequirementisthattheyexhibitnoserialcorrelation.WetestforthispropertyusingArellano-Bondtestsforautoregression(atone,two,andthreelags)foreachofthecountry-specicresidualseriesinoursample.AsthelastcolumnofTable3shows,thenullhypothesisofnoautocorrelationcannotberejectedatconventionallevelsforanyofthecountries.Additionalinformationontheshockseriesretainedforthesecond-stageestimationisprovidedintherightcolumnofTable1andinTable4.Table1describesthecompositionofthenalsample.Itissomewhatreducedfromtheinitialsample,nowcomprising444country-yearobservations.Thisreectsdatagapsforsomeofthevariablesincludedinthesecond-stageregressions.Forexample,wheneveracountry'sexchangerateregimechangedwithinagivenyear,thatobservationisdeletedfromthesample,implyingagapforeachcountry-yearobservationthatrequiresthisdatapointasacontemporaneousorlaggedregressorinthesecondstep.Theshockscontainedinthenalsampleexhibitameanandmedianclosetozeroandastandarddeviationof1.02percentofgovernmentspending.Theminimumandmaximumvalues,rangingfrom-3.57to5.16percentagepoints,arealsoindicatedinTable4.Finally,thecorrelationoftheestimatedshockswiththerawgrowthrateofgovernmentspendingis0.64.Thissuggeststhattherst-stepestimationclearlyremovessomesystematiccomponentofgovernmentspendingchanges,whileproducingashockseriesthatstillbearsaresemblancewiththerawdata,facilitatinganintuitiveinterpretationoftheidentiedscalinnovations.25 25nterpretingtheshocksissubjecttothecaveatstressedbySims(1998).16 Belgium,Canada,Denmark,Finland,France,Ireland,Italy,Japan,Netherlands,Norway,Portugal,Spain,Sweden,theUnitedKingdom,andtheUnitedStates.Table1providesfurtherdetails.ThevariablesusedinourestimationaredetailedinTable2.OurprimarydatasourcesaretheIMFandOECD.Therealexchangerateaswellasmostexpenditureaggregatesareexpressedinlogs;thetradebalanceisexpressedinpercentagepointsofGDP;andinationandthenominalinterestrateareexpressedinpercent.Asregardsourdummies,theclassicationofexchangerateregimesisbasedonIlzetzkietal.(2009),whilethenancialcrisisdatesareprovidedbyReinhartandRogoff(2008)andReinhart(2010).Ourdenitionofweakpublicnancesrequiresgovernmentdebtinexcessof100percentofGDPornetgovernmentborrowingabove6percentofGDP(eachlaggedonce).Thesedenitionsarevariedbelowtoverifytherobustnessofourndings.4Systematicandnon-systematicchangesingovernmentspendingTheprimaryfocusofourstudyisontheeconomy'sresponsetogovernmentspendingshocks,i.e.,changesingovernmentspendingthatarenotsystematicallyrelatedtothestateoftheeconomy.Weestimatetheseresponsesinthesecondstepofourestimationstrategy.Nevertheless,theresultsobtainedfortherststeparealsoofinterestintheirownright,insofarastheycapturethesystematicresponseofgovernmentspendingtothestateoftheeconomy.Forexample,estimatesoftheparametersfortheempiricalscalrulecanshedsomelightonwhetherandhowspendingpolicyrespondstocyclicaldevelopments,ontheonehand,anddebtlevels,ontheother.Table3providesasummaryofresultsfromtherst-stepestimationofthespendingruleforeachcountryincludedinoursample.Afewobservationsstandout.Tostartwith,thetobtainedbyoursimpleempiricalscalruleisveryhigh,reectingtheinclusionofautoregressivetermsinourspecicationforloglevels.Nonetheless,thetremainsquitegoodformostcountriesevenwhenwere-estimatethemodelingrowthrates,aswedoforoneofourrobustnesschecks.Turningtotheparameterestimates,themostgeneralqualitativendingisforgovernmentspendingtorespondnegativelytotheoutstandingstockofpublicdebt.Thecorrespondingcoefcientisestimatedtobenegativeforallbutthreecountries(Austria,Finland,andtheUnitedKingdom),andsignicantlysoforabouthalfofthem.Thisndingalignswellwiththeargumentthatgovernmentspendinghasagreaterroletoplayindebtconsolidationstrategiesthanstandardtheoreticalmodelsassume—apointanalyzedindetailbyCorsettietal.(2011b).Theestimatesofthecoefcientsonlaggedoutputandthecompositeleadingindicatoraresomewhathardertotranslateintoaclearstatementaboutthecyclicalpropertiesofgovernmentspending.Inparticular,therelevantcoefcientsshouldnotberegardedinisolation,astheothertworelatedcoefcientscapturethecyclicalpropertiesofspendingaswell.Anotherresultworthnotingrelatestothesignoftheestimatednancialcrisisdummyforthosecountrieswherenancialcrisesoccurredduringthesampleperiod.Countertotheexperiencefrom15 3.3Thesecondstep:TracingtheeffectsofgovernmentspendingindifferenteconomicenvironmentsInthesecondstep,weusetheestimatedscalshocks(t;i)togaugethedynamicimpactofgovernmentspendingonaggregateoutput,itskeycomponents,aswellasinternationalprices.Webeginthisexercisebydescribingtheeconomy'saverage,orunconditional,responsetoaspendingshock,abstractingfromtheroleofspeciceconomicenvironments.Subsequently,however,weallowtheresponsetobeaffectedbythesetofconditioningfactorsintroducedabove,namelyexchangerateregimes,thestateofpublicnances,andnancialcrises.Accordingly,wespecifythefollowingprototypesecond-stepequation,tobeestimatedinaxed-effectspanelregression:t;itrend;it;i;i;i;it;it;i)+;i;i)+;i;i)+;i;it;i;i;i;it;i(2)wheret;idenotesoneofourmacroeconomicvariablesofinterest(e.g.,consumption);t;iisadummyvariableindicatingacertainfeatureoftheeconomicenvironmentinaparticularyear,suchasacurrencypegoranancialcrisis;andandarethekeyparametersofinterest.Specically,fort;iindexingacurrencypeg,theparameterscapturethedynamiceffect(uptothreeyearsaftertheimpact)ofagovernmentspendingshockineconomieswithaoatingcurrency,whiletheparametersindicatetheadditionalmarginaleffectofthespendingshockunderapeg.Lastly,theparametersaccountforthedirecteffectofthatsameeconomicfeature.Apartfromtherelevantsetofinteractionanddummyterms,wealsoincludefurthercontrolvariablesinourspecications,notablythelaggeddependentvariableandatrend,eachwithcountry-speciccoefcients.Controllingforsuchadditionalvariablesisnotstrictlyrequired:providedthatourrst-stepidenticationstrategydeliversaccurateestimatesofscalpolicyshocks,theseinnovationsareorthogonaltoallothercontemporaneousinformation,thusassuringconsistentsecond-stepestimates.Forallofthespecicationsusingstandardxed-effectspanelestimation,wereportGMM-basedstandarderrorsthatcorrectfortheinclusionofgeneratedregressors,i.e.,thegovernmentspendingshocksestimatedintherststep;seeNeweyandMcFadden(1994).3.4ThedataAsforeshadowedabove,weconsiderannualdata,coveringamaximumperiodfrom1975through2008,althoughforrobustnesswealsoconsiderashortersamplethatexcludestherecentglobalnancialcrisis.Weaimtoincludethesame19OECDcountriesstudiedbyPerotti(1999),butduetodatalimitations(werequireatleast20consecutiveannualobservationstoobtainreliableestimatesforthescalpolicyruleintherststep)windupwithasampleof17countries:Australia,Austria,14 wasasetoftaxrebates,whichwouldnotbeincludedinourconceptofgovernmentspending.24Withthesecaveatsinmind,itisworthnotingthattheBlanchard-Perottiidenticationhasbeenpreviouslyemployedonannualdatabyseveralauthors,includingBeetsma,Giuliodori,andKlaasen(2006)andB´en´etrixandLane(2009).Inpart,thismaysimplyreectpracticalconstraints,asreliablequarterlyscaldataarenotreadilyavailableformorethanahandfulofadvancedeconomies.Inpart,itreectsthesensethatquick-responsescalpolicyisaveryrareexception,andperhapsmostlyfocusedontaxmeasures.Indeed,theabove-mentionedU.S.stimuluspackageswerecloselyrelatedtotheveryexceptionalcircumstancescreatedbytheglobalnancialcrisis.Inordertocapturetheunusualdynamicsofscalpolicyduringsuchexceptionaltimes,weincludealaggednancialcrisisdummyinthespecicationabove.Giventhestartdateofthenancialcrisisin2007,thedummyshouldadequatelycaptureanysystematicscalpolicyresponsetothecrisisduring2008–09,whenthetwoconsecutivestimuluspackageswereagreed.Inaddition,weperformasensitivityanalysiswherewelimitoursampleperiodtoincludedataonlyupto2006.Wendthatourresultsarerobustwithrespecttothisvariationofthesampleperiod.Asidefromtheseconsiderations,thereisanothermoresubstantiveargumentforusingannualdataeveniftheymightattimesgiverisetoendogeneityissuesundertheBlanchard-Perottiidenticationstrategy.Indeed,usingannualdataislikelytoattenuateaseparatepossibleconcernaboutidentication,namelythatidentiedspendingshocksmightbeforeseeable.TheU.S.stimuluspackagesin2008–09againprovideacaseinpoint.ThetaxrebatemeasuresannouncedinJanuary2008,forinstance,wereonlystartingtobeimplementedtowardtheendofthesecondquarterof2008.Treatingthemeasureasanunanticipatedshockinthesecondandthirdquarterwouldthereforebeincorrect,possiblyinducingaseverebiasinestimatesofitseffectontherealeconomy.Thesameistruefortheextraspendinglegislatedintheearly2009stimuluspackage,whichonlystartedcomingonstreamseveralmonthslater.ThisanticipationproblemhasgainedprominencethroughtherecentworkofMountfordandUhlig(2009)andRamey(2011).Althoughtheissueislikelytoaffectscalpolicystudiesingeneral,itisarguablyagreaterconcernforhigh-frequency(suchasquarterly)data.Note,nally,thatpolicyrulessimilarto(1)havebeenconsideredinarangeofrecentquantitativestudiesofscalpolicy.Oneexampleofasingle-estimationapproachlikeoursisGal´andPerotti(2003).However,equation(1)alsomimicsmanyofthegovernmentspendingequationscontainedinVAR-basedstudies,suchasBlanchardandPerotti(2002).Althoughrelativelysimple,theserulesappeartocapturequitewellthemacroeconomicessenceofscalpolicy,thusprovidinguswithusefulmeasuresofscalpolicyinnovations. 24pendingcutsmightalsobeimplementedswiftlyduringepisodesofintensescalconsolidation,especiallywhengovernmentsareforcedtotightenthebudgetinresponsetofallingoutput.13 Specically,weincludethenormalizedvalueoftheOECD'scompositeleadingindicator(CLI)fromOctoberofthepreviousyearasaproxyforgrowthexpectationsheldaroundthetimeofbudgetformulation.TheCLIisareal-timemeasurewithatrackrecordofpredictingchangesineconomicactivity,especiallycyclicalturningpoints,severalmonthsinadvance.Assuch,itseemswellsuitedtocaptureexpectationsaboutthegrowthoutlookheldbypolicymakersandthepublic.23Inprinciple,ouridentifyingassumptioncouldbeviolatedfortworeasons.First,scalpolicyinmostcountriescontainsnondiscretionarycyclicalelements,orautomaticstabilizers.Forourstudy,however,theseautomaticstabilizersshouldnotposeaproblem,astheyoperateessentiallythrough(tax)revenueandtransferpayments,suchasunemploymentbenets,butnotthroughhigherorloweroutlaysforgovernmentconsumption.Asecondpotentialproblemisdiscretionaryscalpolicyactioninresponsetocontemporaneousoutputdevelopments.Therelevanceofthisconcernobviouslyhingesontheprecisedenitionof“contemporaneous”.BlanchardandPerotti(2002),forinstance,arguethatgovernmentspendingpolicycannottypicallyrespondtooutputshockswithinthesamequarter.Indeed,scalauthoritiesaresubjectnotonlytoconstraintsondataavailabilityaboutreal-timedevelopmentsbutalsotousuallysignicanttimelagsbetweenbudgetformulationandexecution.Whetherornottheseconstraintspreventdiscretionarypolicyresponsesformorethanonequarter,isanopenquestion.Inmostcountriestheofcialtimetableforthebudgetprocesstypicallyfollowsthecalendar,implyingthatthemaindiscretionarymeasurestobeimplementedinanygivenyeararediscussedandadoptedattheendofthepreviousyear,basedoneconomicforecasts.Supplementsthroughouttheyearareneverthelesspossible;seePerotti(2004).Yetitisactuallypossibletotestwhetherannualgovernmentspendingispredeterminedconditionalonbeingpredeterminedatthequarterlyfrequency,atleastforcountriesthatpublishhigh-qualityquarterlyscaldata.BornandM¨uller(2012)performsuchtestsondatafortheUnitedStates,UnitedKingdom,Canada,andAustraliaupto2007Q4.TheirmainresultisthattherestrictionthatgovernmentspendingdoesnotrespondtoothervariablesintheVARwithinayearcannotberejected.Infact,theimpulseresponsesobtainedunderthisrestrictionarevirtuallyidenticaltothoseobtainedundertheconventionalBlanchard-Perottiidenticationassumption.Beetsma,Giuliodori,andKlaassen(2009)performanalternativetestandreachthesameconclusion.Notwithstandingthissystematicevidence,thescalstimuluspackagesadoptedbytheU.S.Congressinearly2008and2009maysuggestthatthetimelagbetweenthearrivalofneweconomicdataandtheimplementationofascalresponsecanbeshortenedtoabout5-8months,atleastunderexceptionalcircumstances.However,itshouldbestressedinthiscontextthattheswiftestelementinU.S.policymakers'responsetotheunfoldingcrisis—inbothdecision-makingandimplementation— 23orourestimation,likemostauthors,werelyonnalratherthanreal-timedata.Giventhesometimessignicantrevisionstoreal-timedata,thisraisesconceptualchallengesfortheinterpretationofpolicyshocks;seeCimadomo(2011).12 ofGDPintheaveragecountryinoursample.Moreimportant,itisheldtocontainvirtuallynoautomaticcyclicalcomponent,facilitatingtheattempttoidentifygovernmentspendingchangesaboveandbeyondsystematicuctuationoverthecycle.Unlikepublicinvestment,governmentconsumptionalsohasnoobviousdirectlinktoprivatesectorproductivity,limitingthenumberofpossiblechannelsthroughwhichscalpolicyaffectstherealeconomy.Weassumethattheprocessofgovernmentspendingisdescribedbyarelativelysimplerulethatrelatesourscalvariableofinterest()toitsownrstandsecondlag,thersttwolagsoflogpercapitaoutput(and),thelaggedvalueofacompositeleadingindicator(cli,whichproxiesdirectlyfortheauthorities'pre-budgetexpectationswithrespecttonext-yeargrowth),andthebeginning-of-perioddebtstock,expressedasashareofGDP().Thespecicationalsoincludesatrendvariableandaconstant.Finally,ourinterestintheconditionaldynamicsofscalpolicymotivatesustoinclude(inmostspecications)asetofdummyvariablescapturingkeyfeaturesoftheeconomicenvironment,i.e.,dummiesindicatinganexchangeratepeg(peg),strainedpublicnances(strain),andanancialcrisis(fc).Notethattheinformationcapturedbyeachofthethreedummiesislaggedbyoneperiod,consistentwithourgeneralidentifyingassumption.Inthecaseofthescalstraindummy,thisisachievedbydeningaperiodofscalstrainasafunctionofhighbeginning-of-perioddebtand/orahighdecitintheprecedingyear.Theresultingequationreadsasfollows:t;itrend;i;i;i;icli;i;ipeg;istraint;icrisis;it;i(1)Therulepositsstableparameters(;;\f;\r;;and)overtimeforeachcountryinthesample,butallowstheparameterstodifferacrosscountries.20Theadditiveshockterm(t;i)ismeanttocaptureunexpecteddiscretionarypolicychanges,whoseimpactontherealeconomyistheultimateobjectofourstudy.Notethatthepolicyrulealsoallowsforthedesirablepropertyofautomaticdebtstabilization,namelywhen21Thekeyassumption,however,relatestothecontemporaneousrelationshipbetweengovernmentspendinganditsdeterminants,notablyoutput.Identicationrequiresthattherebenotwo-waycontemporaneousinterdependence.Thisisachievedbyassumingthatspendingcannotrespondtosimultaneousoutputdevelopments.22Instead,spendingisassumedtorespondtopastgrowthdevelopmentsaswellasexpectationsabouteconomicactivityformedoneperiodinadvance. 20inceequation(1)constitutesaunivariateregression(ratherthanthebynowfrequentlyappliedVARmodel),itisusefultokeepinmindthattheinclusionoflaggedvariablesaddressesconcernsaboutspuriousregressionresults;seechapter18.3ofHamilton(1994).21Corsettietal.(2011b)studytheimpactofdebt-responsivegovernmentspendingonscalmultipliersandprovideevidenceforspendingreversalsinU.S.scaldata.22Thisassumptioncarriesovertotheeffectofeconomicenvironmentvariablesonscalpolicychoices,assignaledbytheuseoflagged-informationdummiesinthepolicyequation.11 strategiesofferssufcientexibility.Irrespectiveofthespecicidenticationscheme,thesimplelinearstructureofstandardVARsseverelyconstrainsanyanalysisofconditionaldynamicsinscalpolicytransmission.ThemostVARstudiesallowforistoexaminedifferencesintransmissionacrossasmallnumberofdistinctsubsetsofthedata,throughappropriatesamplesplits.Ilzetzkietal.(2011),forinstance,estimatepanelVARsfordifferentsubgroupsofcountriesdistinguishedbyincome,thelevelofforeigndebt,theexchangerateregime,openness,andthedegreeofcapacityutilization.17Inordertopreservesufcientlylargedatasets,however,theauthorscannotisolatetheimportanceofmorethanonesuchdimensionatatime.Inaddition,itisdifculttoaccountfortime-varyingattributes,suchaspresenceofanancialcrisis,withintheframeworkdevelopedbyIlzetzkietal.(2011).Hencetheirwork,whilecloselyrelatedandcomplementarytoours,leavesopenthequestionaboutthemarginalimportanceofspeciccountrycharacteristicsforscalpolicytransmission.Atthesametime,thepanelVARsetupimposessignicanthomogeneityonthestructureofscalpolicy-makingacrosscountriesinagivensubsetofthedata.RameyandShapiro's“event”approach,inturn,isconstrainedbytheshortageofepisodeswithclear-cutexogenousscalpolicyshocks,especiallyoncetheanalysisisextendedbeyondtheUnitedStates.18Inthispaper,therefore,wepursueatwo-stageestimationstrategysimilartotheoneproposedbyPerotti(1999).Intherststep,weestimateascalpolicyrulethatismeanttodescribethestatisticalprocessofgovernmentspendingandprovideestimatesofspendingshocks.Thescalpolicyruleweconsiderlinksgovernmentspendingtoimportantmacroeconomicaggregates.ItisthusquitesimilartothestructureembeddedinscalpolicyVARs.Weestimatethesescalpolicyrulesforonecountryatatime,thusallowingforsignicantheterogeneityinnationalpolicymaking.Inthesecondstep,weusetheestimatedpolicyshocksasaregressortotracetheimpactofgovernmentspendingonkeymacroeconomicvariables,includingoutput,privateconsumptionandinvestment,thetradebalance,andtherealeffectiveexchangerate.19Aexiblespecicationischosentoaccountfortheeffectsofspendingshocksindifferenteconomicenvironments,thatis,underpeggedvs.exibleexchangerates,withsoundvs.strainedscalpositions,andduringnormaltimesvs.timesofnancialcrisis.3.2Therststep:IdentifyinggovernmentspendingshocksTherststepconsistsinestimatinganannualtimeseriesofscalpolicyinnovationsforeachcountryinthesample.Asourpolicyvariableofinterest,weconsiderpercapitagovernmentconsumptionexpressedinlogs.Governmentconsumptionissizeable:itaccountsforasignicant21.5percent 17eealsoBeetsma,Giuliodori,andKlaasen(2008)foradistinctionofcountriesbyopennesswithinaEuropeansample.18Morerecently,theliteraturehasexploitedvariationinspendingatthesubnationalleveltoestimatemultiplierscontrollingformonetarypolicyand,whenpossible,budgetadjustmentpolicy,dependingonthebudgetrulesofscalfederalism;seeAcconcia,Corsetti,andSimonelli(2011)andCorsettietal.(2011a)foratheoreticalmodel.19ThisprocedureisasymptoticallyequivalenttothestandardVARapproach;seeChristiano,Eichenbaum,andEvans(1996).10 (2002),severalauthorshavebasedidenticationontheassumptionthatdiscretionarygovernmentspendingissubjecttocertaindecisionand/orimplementationlagsthatpreventpolicymakersfromrespondingtocontemporaneousdevelopments.14Accordingtothisidea,signicantpartsofgovernmentspendingaredeterminedbypastinformationonly.Governmentconsumptionandinvestment,inparticular,arelikelytobeunresponsivetocurrenteconomicconditions,as(unliketransfers)theynormallycontainnoautomaticcyclicalcomponent.15Belowweprovideamoredetaileddiscussionofthisidentifyingassumptionanditsrelationtothefrequencyofscaldata.AnalternativeestimationstrategyissuggestedbyRameyandShapiro(1998),whoconsiderasmallnumberofeventsinpostwarU.S.scalpolicy,includingthemilitarybuild-upfortheKoreanandVietnamwars,thatwerearguablyexogenous(withrespecttoeconomicconditions)andthusprovidenaturalexperimentsfortheeffectofasuddensurgeingovernmentspending.SubsequentstudieshaveusedthisapproachwithinaVARcontext;seeEdelberg,Eichenbaum,andFischer(1999),Burnside,Eichenbaum,andFisher(2004),andRamey(2011).Thelatterstudyalsoconsidersaricherdatasetofmilitaryevents16andanalternativeidenticationstrategyusingforecasterrorscompiledonthebasisofsurveysofprofessionalforecasters.Ramey(2011)highlightsdifferencesbetweenherresultsandthoseobtainedundertheBlanchard-PerottiidenticationschemeandarguesthattheyarelikelytoreectthewrongtimingofshocksundertheBlanchard-Perottiapproach.Specically,thegovernmentspendingshockpickedupbytheeconometricianmaywellhavebeenanticipatedbyeconomicagents.Thus,theadjustmentmayalreadybeunderwaybythetimetheshockisdiagnosed.Finally,MountfordandUhlig(2009)haveputforwardanidenticationschemebasedonsignrestric-tions:governmentspendingshocksareidentiedwithinestimatedVARmodelsbyimposingthesignoftheresponseofcertainvariablesforwhichtheoreticalpredictionsarefairlyuncontroversial.WhileMountfordandUhlig(2009)focusondomesticvariables,Enders,M¨uller,andScholl(2011)derivesignrestrictionsonthebasisofarichlyspeciedopeneconomybusinesscyclemodel,inordertoanalyzetheinternationaltransmissionofgovernmentspendingshocks.Forthepurposesoftheempiricalinterestpursuedinthispaper,noneoftheaboveestimation 14ecentcontributionsincludePerotti(2004,2007)andGal´etal.(2007),whichfocusondomestic-economyvariables,andCanzoneri,Cumby,andDiba(2003),KimandRoubini(2008),andCorsettiandM¨uller(2006),whichaddresstheinternationaldimension.InanearlycontributionRotembergandWoodford(1992)estimatetheimpulseresponsestoachangeinmilitaryspendingusingaVARmodelonU.S.data.15Onepossibleexceptionistheindexationofgovernmentwages,whichwouldleadtohighernominaloutlaysduringtimesofstrongeconomicactivityandination.Ifbudgetsarexedinnominalterms,realgovernmentspendingwouldfallinthiscase.Previousworkhas,however,foundsuchination-relatedcyclicalitytobeofverylimitedquantitativeimportanceinadvancedeconomies;seePerotti(2004).16Arelatedstrandoftheliteraturehasfocusedsquarelyonthemultiplierfordefensespending,byregressingoutputgrowthonthechangeingovernmentspendingandpossiblysomeadditionalcontrolvariables.Identicationrestsagainontheassumptionthatmilitaryspendingislargelyunresponsivetothestateoftheeconomy;seeBarroandRedlick(2011)andHall(2009)forrecentcontributionsalongtheselines.Animportantcaveatisthatmilitaryexpendituremightrisesystematicallywithcommand-typeinterventionsintheeconomy,thuscausingadownwardbiasintheestimatedmultiplier;seeHall(2009).Moregenerally,itisunclearwhethertheestimatedmacroeconomiceffectsofhighermilitaryexpenditurecanreadilybeextrapolatedtoothertypesofgovernmentspending. period.Alargeweightofthesehouseholdsintheoverallpopulationreinforcestheexpansionaryeffectsofscalpolicy,sincetheirdemandmovesinproportiontoanyincreaseinemploymentandwagesbroughtaboutbyadditionalpublicdemand.10Totheextentthatanancialcrisisraisestheshareofcredit-constrainedagents,themodelimpliesacorrespondingincreaseinthesizeofthescalmultiplier.11Moreover,astheanalysisofCorsettietal.(2011a)shows,creditconstraintshavealargereffectonthemultiplierunderaregimeofxedexchangeratesorinamonetaryunion.Asidefromintensifyingcreditconstraints,nancialcrisescanhaveanotherimportanteffect,asexempliedbythemostrecentexperienceofglobalnancialturmoilsince2007.Specically,anancialcrisismaycausesuchapronouncedrecessionaryimpactthatmonetarypolicywindsupatthezerolowerboundonpolicyrates,impairingthecentralbank'sabilitytofurtherstimulatetheeconomy.Astherecessiontakeshold,aviciouscirclemaysetin:weakdemandcausesrmstocutprices;ifpricingdecisionsarestaggered,fallingpricesgenerateexpectationsoflastingdeation;foragivennominalinterestrate,thistranslatesintohigherrealrates,whichfurtherweakendemand,thusreinforcingthedeationarydynamics;see,forexample,EggertssonandWoodford(2003).12Underthesecircumstances,scalstimuluscanbecomeapowerfultooltohaltthedeationarydynamics,ashighergovernmentspendingisfullyaccommodatedthroughanunchanged(zero)policyrate.Indeed,Christianoetal.(2011)derivescalmultipliersonoutputwhichexceedtwooreventhree;seealsoHall(2009),ErcegandLind´e(2010),orWoodford(2011).133EmpiricalstrategyInthissection,weintroducetheempiricalstrategyweusetoassesstheroleoftheeconomicenvironmentforthetransmissionofscalpolicyshocks.Beforeprovidingdetailsabouteachofthetwostepsrequiredbyourestimationmethod,wediscusshowouridenticationstrategyrelatestoalternativeapproachesintheliterature.Weconcludethissectionwithabriefdescriptionofourdataset.3.1IdenticationissuesMostoftheexistingempiricalworkonscalpolicytransmissionemploysstructuralVARmodelstogaugetheimpactofspendingshocksontheeconomy.FollowingtheleadofBlanchardandPerotti 10ilbiie,Meier,andM¨uller(2008)useasimilarframeworktomatchthetime-seriesevidenceonU.S.scaltransmission,explicitlylinkinghouseholdconsumptiontoassetmarketparticipation.11Inthesamevein,Fern´andez-Villaverde(2010)exploresscaltransmissioninthepresenceofnancialfrictionswhichconstrainrms'investment,therebylinkingpublicdemandtoentrepreneurs'netwealth.12Thesedynamicswouldbelesspronouncedtotheextentthatcentralbankscanfurtherloosenmonetaryconditionsusingunconventionalpolicytools,suchaspurchasesoflong-termassets.Itisdoubtful,however,whethercentralbanksarepreparedandableinpracticetoimpartanydesiredlevelofadditionalstimulusthroughsuchtools.13However,theintroductionofsovereignriskpremiamayaltertheseresults,ashigherpublicdecitsadverselyaffectfundingcostsinthewidereconomy;seeCorsettietal.(2012). comovementaccordswellwiththeneoclassicalaccountofscaltransmission,itwaswidelyperceivedaspuzzlinginthelightof(Keynesian)receivedwisdom.Subsequently,asmallstrandoftheliteraturehasidentiedaspectsofaneconomy'spublicnancesthatcanaltertheprivateconsumptionresponsetoscalmeasures.Startingfromtheobservationthatscalconsolidationsaretypicallyundertakenatexceptionallyhighlevelsofpublicdebt,BertolaandDrazen(1993)proposeaneoclassicalmodelwherethecorrelationbetweenprivateandpublicconsumptionchangesasdebtapproachessomecriticallevelknowntotriggerasharpscalretrenchment.Perotti(1999)derivesasimilarpredictionfromamodelwithaKeynesianbend.Themodeleconomyispopulatednotonlybyunconstrainedhouseholdswhointernalizethegovernmentbudgetconstraint,butalsobysomecredit-constrainedhouseholdswhoconsumetheirentiredisposableincomeineachperiod.Taxesaredistortionary,andtheeconomyisassumedtobeinitiallyawayfromtheoptimaltaxsmoothingpath.Inthismodel,theresponseofaggregatedemandtoscalmeasuresdependsontheinitiallevelofdebt.In“good”times,scalbalancesandprivateconsumptioncomovenegatively;in“bad”times,withhighlevelsofpublicdebt,thecomovementbecomespositive.Intuitively,ifinitialdebtishigh,thedistortionsfromafurtherincreaseintaxratesarelarge,amplifyingthenegativewealtheffectexperiencedbyunconstrainedhouseholdstothepointthatitoutweighsanypositiveeffectofthescalexpansionontheincomeandconsumptionofconstrainedhouseholds.Totakethismodeltothedata,Perotti(1999)adoptsatwo-stepeconometricprocedure,similartotheapproachwepursueinthispaper.Inarststep,heestimatesanear-VARmodelonannualdatafor19OECDcountries,whichheusestoidentifyscalpolicyshocks.Inasecondstep,heregressesconsumptiongrowthonthescalshocksobtainedfromtherstregression,aswellasaninteractiontermofthisshockwithadummycapturing“badtimes.”Thebad-timesdummyisconstructedbasedonthelevelofdebtandthebudgetdecit.ConsistentwithPerotti'stheory,consumptionrespondspositivelytogovernmentspendingincreasesandtaxcutsinnormaltimes,butthesignoftheresponsechangesinbadtimes.2.4FinancialcrisesAdistinguishingfeatureofnancialcrisesisthataccesstocreditbecomesseverelyrestricted.Althoughnotintendedasamodelofscaltransmissionduringnancialcrisis,theworkofGal´etal.(2007)providesausefulstartingpointtothinkaboutthisinteraction.TheauthorsextendthestandardnewKeynesianmodeltoincludeafractionof“constrained”householdswho,asinPerotti(1999),donotparticipateinassetmarketsandthereforeconsumetheirdisposableincomeineach narelatedstudy,Sutherland(1997)focusesontaxpolicyinanoverlappinggenerationsframeworkandndsthattaxcutsmayhavecontractionaryeffectsifdebtishighandcurrentgenerationsofconsumersexpectconsolidationtotakeplaceduringtheirlifetime.Inarecentstudy,AfonsoandJalles(2011)conditiontheeffectsofscalconsolidationsonthelevelofpublicdebt.Theyndthatareductioningovernmentconsumptiontendstoraiseprivateconsumptionifpublicdebtishigh. framework;seeCorsetti,Kuester,andM¨uller(2011a).sidefromtheroleofexchangerateregimesinshapingscalpolicytransmission,thebehavioroftheexchangerateisalsoanimportantbenchmarkforassessingdifferentmacroeconomictheories,muchlikethebehaviorofprivateconsumptionalreadymentionedabove.Moststandardmodelspredictthathighergovernmentspendingappreciatestherealexchangerate.Recently,however,severalcontributionshavequestionedthisresult,identifyingconditionsunderwhichtheoppositeistrue.Kollmann(2010),forinstance,stressesthatanincreaseingovernmentspendingmaydepreciatetherealexchangerateifgovernmentspendingshocksareverypersistentandinternationalnancialmarketsincomplete.Ravn,Schmitt-Groh´e,andUribe(2012)arriveatthesameconclusionbyassumingthatpreferencesofprivatehouseholdsandthegovernmentarecharacterizedby“deephabits.”Underthisassumption,imperfectlycompetitiveproducersnditoptimaltolowermarkupsandpricesintheshortrun,soastolockinhigherpublicdemandforthefuture.Inequilibrium,thepriceofdomesticconsumptionfallsrelativetoforeignconsumption,i.e.,therealexchangeratedepreciates.InCorsetti,Meier,andM¨uller(2011b)wehighlightthattheimpactofshort-runscalstimulus,includingontheexchangerate,cruciallydependsonthemedium-termdebtconsolidationregime.Specically,ifgovernmentspendingcontributestodebtstabilization,higherspendingtodayinducesexpectationsofasystematicreductioninfuturespendingplans.Asaresult,long-termrealinterestratesdonotriseinresponsetotheinitialspendingincrease,andtherealexchangeratedepreciates.Inthisanalysis,theexchangerateresponseisdrivenbytheanticipatedmonetarypolicyresponsetofuturespendingcuts.2.3WeakpublicnancesInaninuentialstudy,GiavazziandPagano(1990)analyzedlarge-scalescalconsolidationsinDenmarkandIrelandduringthe1980s.Theseepisodeswerecharacterizedbyacoincidenceofimprovinggovernmentbalancesandstrongprivateconsumptiongrowth.Whilethispositive eeBackus,Kehoe,andKydland(1994)andMonacelliandPerotti(2010)forananalysiswithintherealbusinesscycleandthenewKeynesianmodel,respectively.IntheKeynesiantextbookexperiment,inturn,governmentspendingraisesdomesticinterestrates,triggeringcapitalinowsandanappreciationofthecurrency.AnexceptionisthemodeldevelopedbyObstfeldandRogoff(1995)whichpredictsthatgovernmentspendingdepreciatesthe(nominal)exchangeratebecauseofscal-monetaryinteractions:householdslowertheirconsumptionand,hence,theirmoneydemandinresponsetoincreasedgovernmentspending.Ifthemoneysupplyisheldconstant,thecurrencymustdepreciateinnominalterms.Asidefromtheexchangerateregime,tradeopennessisanothervariablewithpotentiallysignicantimplicationsforscalpolicytransmission.Specically,awell-knownargumentsuggeststhatgreatertradeintegrationweakenstheeffectivenessofscalstimulus,insofarasitreduceshouseholds'andrms'marginalpropensitytospendondomesticgoods:withalargerfractionofincomespentonimports,moreofthescalstimulus“leaks”abroad.Erceg,Gust,andL´opez-Salido(2010)provideaquantitativeanalysisusingamodernbusinesscyclemodel.Inthepresentpaperwedonotexploretheroleoftradeopenness,becauseinourdatasetmoreopencountries(intermsofimportshares)systematicallytendtobesmalleconomieswithpeggedexchangerates,complicatingtheidenticationofanydistinctopennesseffects.Weleaveananalysisofthisdimensiontofutureworkdrawingonalargerdataset. notonlyhasquantitativelyimportantimplicationsforthesizeofthegovernmentspendingmultiplieronoutput,butitalsoservestodiscriminatebetweentheopposingpredictionsofkeymacroeconomicmodels.Modernbusinesscyclemodels,ofbothneoclassicalandnewKeynesianvarieties,viewprivateconsumptionasgovernedbyintertemporaloptimization.Thisgenerallyimpliesthatprivateconsumptionfallsinresponsetoanincreaseingovernmentspending.Asaresult,outputmultipliersareconsiderablysmallerthanwouldbesuggestedbymoretraditionalKeynesiananalysis,whichpredictsapositiveconsumptionresponse.Forinstance,inaseminalstudybasedonthefrictionlessneoclassicalmodel,BaxterandKing(1993)considervariousspecicationsforhouseholdpreferencesandthedurationofscalstimulus,butndthatimpactmultipliersonoutputhardlyeverexceedunity.Subsequentresearchhasfurtherrenedthetheoreticalanalysisofscaltransmission.However,muchofthisresearchhasbeenconnedtostandardbusinesscyclemodels,abstractingfromtheexchangerateregime,governmentdebt,andnancialfrictions.Wediscussafewnotableexceptionsbelow.2.2PeggedexchangeratesInopeneconomieswithahighdegreeofcapitalmobility,thechoiceoftheexchangerateregimedeterminesthescopeforindependentmonetarypolicy.ThisconsiderationiscentraltoscalpolicyanalysiswithinthetraditionalMundell-Flemingframework.Inthetypicaltextbookexperiment,governmentspendingisineffectiveinstimulatingdomesticeconomicactivityunderexibleexchangerates:assuminganunchangedmonetarystance,ascalexpansioncrowdsoutnetexportsone-for-oneastheexchangerateappreciates.Onlyunderxedexchangeratesdoesscalpolicybecomeaneffectivestabilizationtool,becauseanypressuretowardexchangerateappreciationisimmediatelyoffsetthroughmonetaryexpansion.Althoughthesespecicpredictionsdidnotgounchallenged(seeDornbusch(1980)foranearlycritique),thetraditionalliteratureviewstheexchangerateregimeashavingarst-ordereffectonscaltransmission.TheexchangerateregimecontinuestomatterforscalmultipliersinnewKeynesianbusinesscyclemodels,butthesharppredictionsoftraditionalKeynesiantheorydonotnecessarilygothrough.Onekeyreasonistherichermodelingofhowscalandmonetarypoliciesinteract.Indeed,scalstimuluscanbeeithermoreorlesseffectiveunderaexibleexchangerateinanewKeynesianmodel,dependingonthepreciseassumptionsaboutmonetarypolicyandthemedium-termdebtconsolidation urdiscussionfocusesonthetransmissionofgovernmentspendingshocksassuminglump-sumnancing.Inlinewithmuchoftheliterature,wealsoassumethatgovernmentspendingdoesnotaffectthemarginalutilityofhouseholdsnortheproductivecapacityoftheprivatesector.BaxterandKing(1993)alsoshowthatmultipliersaremuchlargerundertheassumptionthatgovernmentspendingenhancestheproductivecapacityoftheeconomy.Bycontrast,outputmultipliersarenegativeifgovernmentspendingisnancedbydistortionarytaxes,assumingbalancedbudgets.LinnemannandSchabert(2003)provideanearlyanalysiswithinthenewKeynesianbaselinemodel.Cogan,Cwik,Taylor,andWieland(2010),inturn,consideraricherbusinesscyclemodelwithaparticularfocusonquantifyingthemultiplierimpliedbyactualscalpolicymeasuresimplementedintheUnitedStatesafter2009. Asthenumberofobservationsinoursampleislimitedandscalshocksmaybemeasuredwitherror,condenceintervalsare,ingeneral,quitewide,andpointestimatesmustbetakenwithagrainofsalt.Yetourmainconclusions,especiallyregardingtheeffectsofgovernmentspendingshocksduringnancialcrises,appeartoberobustwithrespecttoanumberofvariationsinourempiricalsetup.MostcloselyrelatedtoourworkisindependentresearchbyIlzetzkietal.(2011),whoalsoanalyzethetransmissionofscalpolicyacrossdifferenteconomicenvironments.Incontrasttoourstudy,theseauthorsusequarterlytime-seriesdataforbothOECDandemergingmarketcountries,estimatingpanelVARsforgroupsofcountriesdistinguishedbyincomelevel,thesizeofforeigndebt,theexchangerateregime,openness,andthedegreeofcapacityutilization.Theirapproach,however,isolatesonlyonedimensionatatime.Intermsofresults,therearesignicantsimilarities,especiallyregardingtheoutputmultiplieracrossexchangerateregimes,butalsodifferences,notablywithrespecttothedegreeofmonetaryaccommodationandexchangerateappreciationunderapeg.Notwithstandingthesedifferences,bothstudiesclearlyunderscorethecaseforinvestigatingscalpolicytransmissionconditionalontheeconomic,nancial,andpolicyenvironment.Theremainderofthispaperisorganizedasfollows.Section2providesabrieftheoreticaldiscussionofwhythescaltransmissionmechanismmaydifferacrosseconomicenvironments.Section3introducesourtwo-stepestimationapproachindetail.Sections4and5presentthemainresultsfortherstandsecondstep,respectively;andSection6concludes.2FiscalpolicyindifferenteconomicenvironmentsTheveryactiveuseofscalpolicysincethestartoftheglobalnancialcrisis—rsttocounterthedownturnwithscalstimulusand,morerecently,toreininhighdecitsamidbondmarketpressuresintheeuroarea—hasrevivedlongstandingpolicydebatesanddrawnattentiontoanimportanttheoreticalinsight:thereisunlikelytobesuchathingas“the”scalmultiplier.Instead,itseemsreasonabletoexpectthatmultipliersdependoncurrentcircumstancesaswellasunderlyingeconomicstructuresandpolicyregimes(quiteasidefromanyvariationrelatedtothespecicscalmeasureathand).Accordingly,thelikelyimpactofscalpolicycannotbeassessedwithoutproperconsiderationofthekeyfactorscharacterizingtheeconomicenvironmentacrosscountriesandovertime.Inthispaper,weemphasize,inparticular,theroleofexchangerateregimes,thestateofpublicnances,andthehealthofthenancialsector.Beforeassessingtheirrelevanceforscaltransmissionempirically,webrieyreviewinthissectionsomekeytheoreticalcontributionsthatguideourempiricalanalysis.2.1AtheoreticalbenchmarkThetheoreticaldebateonthescaltransmissionmechanismhastraditionallyfocusedontheresponseofprivateconsumptiontoanincreaseingovernmentspending.Indeed,theconsumptionresponse DrawingontheworkbyPerotti(1999),weemployaexibletwo-stagestrategythatallowsustoexploitvariationineconomicconditionsacrossspaceandtimetogaugetheirimpactonscalpolicytransmission.Inarststep,weestimateascalpolicyrulethatismeanttodescribethestatisticalprocessofgovernmentspendingandprovideestimatesofspendingshocks.Theruleweconsiderisverysimilartothestructureembeddedinscalpolicyvectorautoregressions(VARs),linkingourapproachtoanidenticationstrategycommonlyfoundintheliterature.Inasecondstep,weusecontemporaneousandlaggedvaluesoftheestimatedpolicyshockstotracethedynamiceffectsofgovernmentspendingonseveralmacroeconomicvariablesofinterest.Westudytheroleofdifferenteconomicenvironmentsinshapingscaltransmissionbyinteractingtheshockswithdummiesfortheexchangerateregime,thestateofpublicnances,andtheoccurrenceofanancialcrisis.Tosetthestageforouranalysis,weinitiallyestimateourmodelabstractingfromanyoftheseinteractions.Inthiscase,ourtwo-stepprocedureyieldsaseriesofresultsfortheunconditional(oraverage)behaviorandimpactofscalpolicythataccordwellwithearlierndingsfromtheliterature.First,theestimatedspendingrulessuggestthatgovernmentspendingexhibitsnoclearcyclicalpattern,butrespondsnegativelytoweakpublicnances,thuscontributingtodebtstabilization.Second,inresponsetoanunexpectedincreaseingovernmentspending,wendapositive,ifrelativelycontainedincreaseinoutput,almostnoresponseofconsumption,andsomecrowding-outofinvestmentandnetexports.Moreover,thespendingshockpromptsashort-livedrealappreciation,followedbyaweakeningoftherealexchangerate.However,theseunconditionalresultsmaskimportantdifferencesinthetransmissionofscalshocksacrosseconomicenvironments.Onceweturntoourfullyspeciedmodelincludingthreesetsofconditioningvariables,thefollowingpictureemerges.Underaexibleexchangerate,withlowdebtlevelsandnonancialcrisis,wendnoappreciableeffectsofgovernmentspendingshocksexceptoninvestment,whichdeclines,andtherealexchangerate,whichweakensovertime.Relativetothisbaselinescenario,apeggedcurrencyimpliesalargertradedecit,butnorealexchangedepreciationonimpact.Formostspecications,wealsondspending-sidestimulustobemoreeffectiveunderapeg,inlinewithconventionalwisdom.Specically,theresponseofoutputislarger,andinvestmentfallsbyless.Ifpublicdebtorthedecitishigh,theimpactresponsesofoutputandinvestmentarelowerthaninthebaselinescenario.Yetthedifferencesareverymoderate,perhapspointingtothelimitationofourscalmetricsincapturingtheoverallhealthofpublicnances.Amorestrikingsetofndingsemergesforthecaseofscalpolicytransmissionduringtimesofnancialcrisis.Indeed,theresponsesofoutputandconsumptiontoapublicspendingincreasearestronglypositiveduringsuchtimes,implyingascalmultiplierofuptotwo.Inaddition,wendaconsiderablymorepronounceddeclineofnetexportsandtherealexchangerate. hiledifferencesintheresponseofexternalvariablesacrosscurrencyregimesarequiterobust,differencesintheresponseofothermacroeconomicaggregatesaresomewhatmoresensitivetothespecicationofthemodel. 1Introductionhewidespreaduseofscalstimulusmeasurestocountertheglobalnancialcrisisandthemorerecentshifttowardscaltighteninginmanyadvancedeconomieshaverevivedthelongstandingdebateonthesizeofthescalmultiplier.Fromatheoreticalperspective,however,thereisnosuchthingasthemultiplier.Instead,scalmultipliersarelikelytodependonanumberoffactorswhichvarybothacrosscountriesandtime.TraditionalMundell-Fleminganalysispositsthattheeffectivenessofscalstabilizationhingesonnancialdevelopment,capitalmobility,tradeopenness,andtheexchangerateregime.Inaddition,theresponseofprivatedemandtoascalinterventionmayalsodependonthestateofpublicnances.Forinstance,scalexpansionsathighlevelsofdebtcouldplayoutdifferentlyiftheyincreasethelikelihoodofasharpfutureretrenchment.Anotherpotentialdeterminantisthehealthofthenancialsystem,notablytheextenttowhichtheprivatesectorhasaccesstocredit,giventhatbindingliquidityconstraintsgenerallyreinforcetheimpactofscalstimulus.Inasimilarvein,recentquantitativeanalysispredictsexceptionallylargegovernmentspendingmultipliersduringdeeprecessionswhenmonetarypolicyisconstrainedbythezerolowerboundonpolicyrates.Inthispaper,wecarryoutanempiricalexplorationintothedeterminantsofgovernmentspendingmultipliers,bystudyinghowthescaltransmissionmechanismdependsontheeconomicenviron-ment.Intermsofconditioningfactors,wefocusontheexchangerateregime,thelevelofpublicdebtandthedecit,andtheoccurrenceofanancialcrisis.Weconductouranalysisonasampleof17OECDcountriesfortheperiod1975–2008.FortheclassicationofexchangerateregimesandnancialcrisisepisodeswedrawonIlzetzki,Reinhart,andRogoff(2009),ReinhartandRogoff(2008),andReinhart(2010).PriorempiricalworkonscalpolicytransmissionhasmostlyreliedonlineartimeseriesmodelsestimatedonU.S.data.Indeed,anextensiveliteraturehasclariedkeyissuesinidentication,providingalternativestrategiesforquantifyingtheeffectsofscalpolicymeasures;seeBlanchardandPerotti(2002),MountfordandUhlig(2009),andRamey(2011),amongothers.Atthesametime,therearerelativelyfewempiricalstudiesexaminingthedependenceofscalpolicyeffectsoneconomicenvironments,notablyPerotti(1999),Giavazzi,Jappelli,andPagano(2000)and,morerecently,Tagkalakis(2008),Ilzetzki,Mendoza,andVegh(2011),andAuerbachandGorodnichenko(2010,2011).Evenlessempiricalworkhasbeendevotedtothequestionofhowscaltransmissionchangesduringtimesofnancialcrisis. orinsightfultheoreticalanalysesofthelatterthreeaspects,seeBertolaandDrazen(1993),Gal´,L´opez-Salido,andVall´es(2007),andChristiano,Eichenbaum,andRebelo(2011),respectively.Corsetti,Kuester,Meier,andM¨uller(2012)analyzehowdeeprecessionsinteractwithstrainedpublicnancesinshapingscalpolicytransmission.SeeBaldacci,Gupta,andMulas-Granados(2001)andInternationalMonetaryFund(2009)foraninvestigationintohowscalpolicymaymitigatedeeprecessions. © 2012 International Monetary Fund WP/12/ Monetary and Capital Markets Department nt Spending Multipliers? Prepared by Giancarlo Corsetti, Andre Meier, and Gernot J. MüllerAuthorized for distribu rnment spending vary with the economic e nvironment. Using a panel of OECD countries f rom an estimate d spending rule and trace their macroeconomic impact under different c onditions regarding the exchange rate regime f inancial system. The uncondi c onfir m earlier findings. However, conditional responses differ systematically across e xchange rate regimes, as rdeficits occur mainly under c urrency pegs. We also find output and consumption multipliers to be unusually high d uring times of financial crisis. JEL Classification Numbers: E62, E63, F41 Keywords: Multiplier, fiscal policy, financial crisis, government spending, public finances, exchange rate regime giancarlo.corsetti@gmail.com ameier@imf.org gernot.mueller@uni-bonn.de This paper is part of the project “International Dimensions of Fiscal Policy Transmission,” sponsored by the Fondation Banque de France (11 Call for Projects) in cooperation with the Center for Economic Policy Research. An earlier version of the paper was circulated under the title “When, Where, and How Does Government Spending Matter?” The current version was prepared for the Economic Policy Panel in Copenhagen. We thank Philip Lane and three anonymous referees for very useful comments. We also thank Giovanni Callegari, Bianca de Paoli, Robert Hall, Hubert Kempf, Mika Kortelainen, Daniel Leigh, Jesper Linde, Marialuz Moreno Badia, Silvia Sgherri, Mike Woodford and seminar participants at the Banco de Espana, Bank of England, Banque de France, CEPR ESSIM 2010, European Central Bank, the 2010 conference on “Global Dimensions of the Financial Crisis” at the Federal Reserve Bank of New York, and the Federal Reserve Board. Dominik Menno provided excellent research assistance. Corsetti and Müller gratefully acknowledge generous support from the Fondation Banque de France. as representing the views of the IMF. The views expressed in this Working Paper are those of the author(s) and do not necessarily IMF policy. Working Papers descri What Determines Government Spending Multipliers? Giancarlo Corsetti, Andre Meier, and Gernot J. Müller WP/12/150