Fiscal spillovers in the Euro area

Transcription

1 Fiscal spillovers in the Euro area Fabio Canova, EUI and CEPR Matteo Ciccarelli, ECB Pietro Dallari, UPF November 23

2 Introduction The nancial crises have put scal policy back in the spotlight of academic research (e.g. Hall, 29, Woodford, 29, Canova and Pappa, 2; Auerbach and Gorodnichenko, 22). Renewed debate on the size of the scal multiplier (e.g. IMF, 22; Mertens and Ravn 22; Blanchard and Leigh, 23). Important academic discussion about the eects anticipated vs. unanticipated scal shocks (e.g. Ramey (2), Leeper et al (22)). Policy debate regarding the Euro area eects of Mediterranean countries. scal contractions in

3 Theory: eects of scal contractions may depend on the state of the economy, level of debt, monetary policy, condence of private agents, etc. Empirics: still far from clear what are the consequences of scal policy actions in the macroeconomy, the channels of domestic and international transmission, and the eects of coordinated scal retrenchment in a situation of high debt, nancial weakness and low condence.

4 We want to provide empirical evidence on the spillovers of scal shocks in the Euro area. - Construct a (historical) measure of the scal stance. - Explore the domestic eects of unexpected expenditure contractions and their international transmission. - Measure (domestic and international) output multipliers. - Investigate the eects of coordinated vs. idiosyncratic changes. - Study the eects of expenditure reducing vs. decit reducing policies. - Compare pre and post crisis transmission.

5 Related literature - Auerbach and Gorodnichenko (23), Faccini et al. (22), Corsetti and Mueller (22): Open economy multipliers. - Beetsma and Giuliodori (2): Multipliers in the EU. - Benassy-Quere and Cimodamo (26): Beggar-thy-neighbor eects. - Alesina et al. (22); Batini et al. (22); Hall (22): Consequences of scal consolidations. - Corsetti et al. (22), Iltzeski et al. (22), Christiano et al. (2), Caldara and Kemps (22), Erceg and Linde (22), Leeper et al. (2), Woodford (29): What determines the size of the multipliers?

6 Findings ) Contractionary expenditure shocks in the periphery have heterogeneous domestic eects. Trade balance (export) dynamics key to understand cross-country dierences. 2) The eects on debt and long term yields are small or perverse. The signalling eects of contractionary measures in nancial markets limited. 3) International transmission important also before 27. Trade channel are more important than capital movements channel. 4) Decit cuts have larger eects than expenditure cuts, both domestic and internationally.

7 5) Magnitude of the spillovers changes after 27; responses more homogeneous. 6) Virtuous domestic eects disappear in periphery after 27; the core seems to benet more - redistribution shock.

8 2 The empirical model Panel VAR model (e.g. Canova and Ciccarelli, 29): y it = D i (L)Y t + F i (L)W it + e it () e it N (; i ) y it : G ; Y t = (yt ; : : : y Nt ) ; i = ; ::; N (countries); t = ; ::; T (time) W it : M vector of (weakly) exogenous variables. D i;j : G NG for each lag j = ; :::; p; F i;j : G M for each lag j = ; : : : ; q; Variables are demeaned, standardized, year-on-year growth rates.

9 Important features:. Allow for dynamic and static interdependencies. 2. Allow for cross sectional heterogeneity in dynamic relationships and the variance of the error vector. Add realism but curse of dimensionality problem: too many parameters to estimate relative to standard sample sizes.

10 Model (in regression format): Y t = Z t + E t E t N (; ) (2) where Z t = I NG Xt ; X t = (Y t ; Y t 2 ; : : : ; Y t p ; W t ; : : : ; W t q ), = ( ; : : : ; N ) and i are Gk vectors containing, stacked, the G rows of the matrix D i and F i, while Y t and E t are NG vectors. - Curse of dimensionality: is NGk where k = NGp + Mq and is a full NG NG matrix. - If N= countries, G=5 variables, p=2 lags, q= (no exogenous variables), is 5 vector and has (NG (NG ))=2 = 25 free parameters!

11 Factor structure for coecients (shrinkage prior): = u u N(; 2 I) (3) where ; 2 ; 3 ; 4 are NGk N, NGk G; NGk M; NGk p known matrices; - : movements in the coecient vector which are country-specic; - 2 : movements which are variable-specic; - 3 : movements which are specic to predetermined variables; - 4 : movements which are specic to the lags.

12 Kroneker structure for the variance: = P V, where - P is NxN matrix: average (across variables) cross country spillovers. - V is GxG matrix: average (across country) cross variable spillovers.

13 Advantages: - Instead of NGk coecients and NG(NG )=2 covariance elements, estimate + N + G + M + p factors and (N(N ) + G(G ))=2 covariance parameters. Note: reparametrized model (using (3) into (2)) is Y t = 4X j= Z jt j + v t (4) where Z jt = Z t j, v t = E t + Z t N(; t ); and t = (I + 2 Z t Z t). - SUR model with particular heteroschedastic structure. - No a-priori restrictions on the nature of interdependencies (as e.g. in GVARs or spatial VARs).

14 - Factors on the coecients rather than the variables (as in factor models). - Del Negro and Schorfheide (2): cross sectional shrinkage prior (as opposed to time series shrinkage prior as in Minnesota prior). - Can have time variations in the coecients, see Canova and Ciccarelli (29).

15 Fiscal stance indicators An indicator of the scal stance in country i is NF I jit = Z it i + Z 2jt 2j An aggregate scal stance indicator is given by RF I jt = NX i= Z it i + Z 2jt 2j - Z 2jt 2j could be either the decit or the debt variable. - Gives historical record: the stance loose if indicator is positive; the stance is tight if indicator is negative.

16 - Could transform it into a leading indicator - to forecast if future stance will be loose or tight. - Note: eects of exogenous variables netted out - focus on regional and national specic uctuations.

17 Estimation Easy to estimate (4). Can use both classical and Bayesian methods. Prior assumptions: R ; Q ; s 2 estimated from the data. Posterior proportional Y t j t j =2 N( ; R ) W (NG; Q ) (5) 2 G(:5; :5s 2 ) 2 A exp 4 2 X t (Y t Z t ) ( t ) (Y t Z t ) 5 (6) Use Gibbs sampling (with a Metropolis step for 2 ) to obtain sample from the posterior. 3

18 Inference Identication: Recursive restrictions on, i.e. domestically, government expenditure is assumed to be predetermined relative to other domestic variables; internationally periphery predetermined to the core. To be changed in future versions. Plan to use both sign and (relative) magnitude restrictions. Multipliers: computed as in Auerbach and Gorodnichenko (22).

19 3 The data Seven countries (87 percent of the area GDP): Greece, Italy, Spain and Portugal (Periphery); France, Germany and the Netherlands (Core) Eight endogenous variables: real government consumption expenditure (G), total government revenues (T), total gross government debt (D), real gross domestic product (Y), real total private consumption (C), real total xed investment (I), capital account (CA), and -years bond yields (LR). Six exogenous variables (short term rate, GDP and CPI in US and EA) plus G forecasts by OECD - to control for predictable components, see Ramey (2). Sample: 99: to 2:4.

20 4 The Results - Fiscal Indicators - Domestic Transmission pre crisis - Multipliers pre crisis - International Transmission pre crisis - Post crisis transmission/multipliers - Government vs decit cuts

21 Aggregate debt based fiscal stance

22 Country specific debt based fiscal stance.5.5 Greece Italy Portugal S pain France 95 5 Germ any 95 5 Netherlands 95 5

23 Aggregate deficit based fiscal stance

24 Country specific deficit based fiscal stance 2 2 Greece Italy Portugal S pain France Germ any Netherlands

25 - Fiscal Indicators - Domestic Transmission pre crisis - Multipliers pre crisis - International Transmission pre crisis - Post crisis transmission/multipliers - Government vs decit cuts

26 NX I C Y G Greece Italy Portug al Spain

27 CA R Debt G Greece Italy Portugal S pain

28 Large cross country heterogeneities Dynamics of the trade balance crucial to understand domestic eects Eect on long term interest rates initially perverse.

29 Misspecication: single country VARs Government spending shock in Italy 97 Government consumption expenditures Gross government debt Private consumption Total fixed investment Long terminterest rate Net Exports Gross domestic product 5 5 Capital account 5 5

30 - Fiscal Indicators - Domestic Transmission pre crisis - Multipliers pre crisis - International Transmission pre crisis - Post crisis transmission/multipliers - Government vs decit cuts

31 How big are multipliers? - Spilimbergo et al. (29): Average (domestic) multipliers.5, most values above zero but below the mean. - Gechert and Will (22), Andres and Domenech (22): Average (domestic) multipliers (.5,.), depending on instrument and estimation method. - Size depends: (i) nancial frictions; (ii) instrument and level of debt; (iii) nominal and real rigidities; (iv) monetary policy stance; (v) exchange rate regime; (vi) degree of openness; (vii) individual vs. coordinated actions; (viii) credibility of measure (Iltezky, et al (22), Corsetti et al (22), etc.).

32 Short term multipliers (2 quarters) Greece Italy Portugal Spain Common Greece.35* -.3*.4*.3*.26* Italy.75* -.27*.4*.3*.27* Portugal.36* -.32*.9*.3*.5 Spain -.68*.24*.28* -..5 France -2.57* y -.3*.32*.4* -.8* Germany -.44* -.35*.8* * Netherlands.7*.77* -.3* -.6.7* Medium term multipliers (2 quarters) Greece.6.4*.9*.8*.6* Italy -.4.2*.2*.3*.46* Portugal *.*.5*.4 Spain.35.23*.25*.9*.24* France *.24*.28* -.46* Germany *.7*.5* -.4 Netherlands * -.9*..54* *=posterior credible set above, y=posterior credible set above. Sample 99-27

33 - Fiscal Indicators - Domestic Transmission pre crisis - Multipliers pre crisis - International Transmission - Post crisis transmission/multipliers - Government vs decit cuts

34 NX Spread CA Expenditure Impulse in Italy France Germ any.6 Netherlands

35 NX Spread CA Expenditure impulse in Spain France Germ any N etherlands

36 - Fiscal Indicators - Domestic Transmission pre crisis - Multipliers pre crisis - International Transmission crisis - Post crisis transmission/multipliers - Government vs decit cuts

37 NX I C Y G Greece Italy Portugal S pain

38 CA R Debt G Greece Italy Portugal S pain

39 Short term multipliers (2 quarters) Greece Italy Portugal Spain Common Greece.5* * Italy Portugal.56*.59*...4* Spain -.99*. -.22* -.2* -.4* France -.59*.4.6*.25*.55 Germany -.37* -.36*.6* * Netherlands -.67* * * Medium term multipliers (2 quarters) Greece.38* * Italy -.36* Portugal..44*.8..44* Spain -.79* * France *.6. Germany -.43* * Netherlands -.72* * * *=posterior credible set above, y=posterior credible set above. Sample 994-2

40 NX R CA Common shock France Germ any N etherlands

41 - Fiscal Indicators - Domestic Transmission pre crisis - Multipliers pre crisis - International Transmission crisis - Post crisis transmission/multipliers - Government vs decit cuts

42 Short term multipliers, Decit consolidation Greece Italy Portugal Spain Common Greece.32*.8* -.4* * Italy.45* -.73* -.4* * Portugal.3*.76* * Spain -2.3* y.65* -.35* -.29* -.87* y France 2.4* y.74*.56*.3* 3.67* y Germany -.8*. -.49* -.3* -.58* y Netherlands -2.82* y * * y Medium term multipliers, Decit consolidation Greece.86*.36* -.2* -.2.9* Italy * * Portugal * Spain -.39* y.72* -.4* * France.87*.69*.3..78*y Germany -.82*.28* -.54* * Netherlands -.93*y. -.5* *y *=posterior credible set above, y=posterior credible set above. Sample

43

44 5 Still to investigate - Large consolidation shocks (as in Canova and Pappa, 2). - Labor market eects. - Conditional forecasts: (i) debt decline by 5 percent permanently: immediate or gradual. (ii) international eects of particular consolidation shocks. (iii)... - Non-Euro area repercussions.