Sovereign risk and macroeconomic stability Giancarlo Corsetti, Keith Kuester, Andr´e Meier, Gernot M¨ uller
December, 2011 The views expressed are those of the authors. They do not necessarily coincide with those of the IMF, the Federal Reserve Bank of Philadelphia or the Federal Reserve System.
Sovereign risk and debt level 5-Yearr Sovereign CDS Spread (basis points, as of May 6, 2011)
1800 1600 1400
GRE
Fitted risk premium
1200 1000 800
POR
600
IRL
400 200 AUS
0 0
20
ESP GBR CZE AUT FRA SLV DNK SWE NOR SLO DEU FIN NLD
BEL USA
ITA
40
60 80 100 120 General Government Gross Debt (percent of GDP, forecast for 2011 and 2015)
140
160
Sovereign CDS spreads for advanced economies (bar Japan). End-of-the year ratio of gross debt/GDP (2011: blue dots, 2015: green triangles); IMF WEO April 2011. Other indicators of the fiscal position paint similar picture. Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
1/26
Sovereign and non-financial corporate CDS 350 300
450
CDS Spreads in Low!Spread Euro Area (Basis points) Nonfinancial corporates
250
Sovereigns
200
CDS Spreads in High!Spread Euro Area (Basis points)
400 350
Nonfinancial corporates Sovereigns
300 250 200
150
150
100
100
50
50
0 2008
0
2009
2010
2011
2008
2009
2010
2011
For better comparability, we focus here only on euro area countries. For the figure, we split the sample into two sets of countries: those with relatively high sovereign spreads (Greece, Ireland, Italy, Portugal and Belgium), and those with relatively low spreads. Correlation for low-spread countries: 0.3; for high-spread countries: 0.7. Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
2/26
Today’s talk: the sovereign risk channel Sovereign risk by itself might not affect economic activity (what matters is ex ante real return, not notional return). However, well-documented that private sector risk premia tend to comove with sovereign risk. This is true both for financial firms (banks) and for non-financial corporates. It is true from small and medium size firms, but also for large one, operating in different markets and in principle able to obtain financial means outside the Home country of their headquarters. Why? Different reasons: from balance sheet effects from falling debt prices, to jurisdiction risks in general, linked to anticipated higher taxation, days of work lost because of strikes and protests, anticipation of falling demand because of contractionary policies and the like. !
What are the implication of the spillovers from sovereign risk on private borrowing costs, especially at the ZLB, for stabilisation policy?
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
3/26
The sovereign risk channel !
tends to exacerbate the effects of negative cyclical shocks, unless the central bank manages to offset it by lowering policy rates or using other measures;
!
tends to delay the exit from the zero lower bound;
!
makes the economy (more) vulnerable to belief-driven equilibria;
!
undermines key stabilisation prescriptions (e.g. regarding the desirability of systematic anti-cyclical fiscal policy);
!
affects the magnitude and even the size of fiscal multipliers, making them quite sensitive to markets’ beliefs about the duration of a recession, and the elasticity of the risk premium to the fiscal outlook.
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
4/26
Related literature Model sovereign default and interest rate spillover: Arellano 2008, Mendoza and Yue 2010 Fiscal limit and sovereign risk premium: Bi 2010, Bi/Leeper 2010 or Juessen/Linnemann/Schabert 2011 Policy-regime switching: Davig/Leeper 2011 Country risk and fiscal consolidations: Erceg/Linde 2010 Note: our analysis purely positive (don’t model default)
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
5/26
Remainder of the talk Model and calibration. Analytical results: transmission, determinacy and multipliers. Dynamic simulations (ZLB endogenous, distortionary taxes, shifts in risk premium). Conclusions.
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
6/26
New Keynesian model with sovereign-risk channel Variant of Curdia and Woodford 2009 !
Heterogeneity in non-financial private sector
!
Costly financial intermediation drives spread between borrowing and lending rate
Fiscal policy !
Government defaults on debt with a certain probability, increasing in the level of debt
!
Ex-post: lump-sum transfer payments neutralize redistribution due to default
!
Ex-ante: Sovereign risk increases costs of financial intermediation
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
7/26
Households With probability 1 − δ new type is drawn: τt ∈ {b, s }; probabilities πs and πb → borrowers/savers in equilibrium Complete financial markets, but infrequent state-contingent payoffs: borrowing and lending via financial intermediaries ! " −1 −1 ∞ ψτ (ξ τ )στ [ct (i )]1−στ t 1+ ν max E0 ∑ (et β ) − ht ( i ) 1+ν 1 − στ−1 t =0 Intertemporal optimality conditions # %& 1 + itd $ (1 − δ)πb λbt+1 + [ δ + (1 − δ)πs ] λst +1 et λst = βEt et +1 Πt +1 # %& 1 + itb $ (1 − δ)πs λst +1 + [ δ + (1 − δ)πb ] λbt+1 et λbt = βEt et +1 Πt +1 # %& ( 1 − ϑt +1 )(1 + itg ) $ et λst = βEt et +1 (1 − δ)πb λbt+1 + [ δ + (1 − δ)πs ] λst +1 Πt +1
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
8/26
Financial intermediaries Perfectly competitive, risk neutral intermediaries take on deposits and provide loans (no equity) Spread between borrowing and lending rates
(1 + ωt )(1 + itd ) = 1 + itb Costs of intermediation due to loan losses/“fraud” χt bt Profit maximization subject to financing constraint dt > bt (1 + ωt ) gives optimality condition ω t = χt Our assumption (“jurisdiction risk”) χt = χ ψ
Introduction
Model
Calibration
'
1 + itg 1 + itd
Analytical results
(αψ
−1
Simulations
Conclusions
9/26
Firms Continuum of monopolistically competitive firms j ∈ [0, 1] with technology yt (j ) = h (j )1/φ Calvo price setting, FOC Kt Ft
=
Kt
=
Ft
=
' ∗ (1+ θ ( φ −1) Pt Pt !' " ' (φ ( Πt +1 θφ yt p + αβEt λt et µ φwt Kt +1 zt Π " !' ( Πt +1 ( θ −1 ) Ft + 1 λt et yt + αβEt Π
Law of motion for prices (inflation) 1−α
Introduction
Model
Calibration
'
Πt Π
( θ −1
= (1 − α )
Analytical results
' ∗ (1− θ Pt Pt
Simulations
Conclusions
10/26
Fiscal policy Evolution of government debt btg = (1 − ϑt )
btg−1 (1 + itg−1 ) Tg Tc + gt − t − τtw wt ht + t Πt Pt Pt
Transfers neutralize consequences of default ex post:
Ttc Pt
= ϑt
btg−1 (1+itg−1 ) Πt
Spending exogenous; tax revenues trt = τtw wt ht + Ttg /Pt : trt − tr = φT ,y (yt − y ) + φT ,bg (btg−1 − b g )
!
Actual default is neutral (by assumption), but probability of default is not
Stochastically determined “fiscal limit” ( ) ' g 1 bt ϑdef g g ; ϑ = ; α , β pt = Fbeta t b 0 4y b g,max b
with prob pt with prob 1 − pt
Spending exogenous, tax revenues respond to economic activity Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
11/26
Monetary policy and equilibrium Interest rate rule log (1 + itd,∗ )
=
log (1 + i d ) + φΠ log (Πt /Π) − φω log ((1 + ωt )/(1 + ω ))
itd
=
max{itd,∗ , 0}
Good market clearing yt =
Introduction
Model
* 1 0
ct (i )di + gt + Ξ bt = πb ctb + πs cts + gt + Ξ bt
Calibration
Analytical results
Simulations
Conclusions
12/26
Calibration Most parameters standard. If country-specific, use U.S. as the baseline: !
Output semi-elasticity of tax revenues (OECD): φT ,y = 0.34.
!
Price rigidities: θ = 0.9.
!
Labor supply: ω = 1/1.9 (Hall, 2009). Steady-state output normalized to unity.
!
Share of government spending: 19 percent.
!
Monetary policy: φΠ = 1.5. Steady-state inflation: 2 percent p.a. Steady-state nominal rate Rt is 4.6 percent ann.
!
For dynamic simulations: use steady-state gross debt to GDP ratio of 60 percent, but allow for higher initial levels of debt.
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
13/26
Calibration of default probability and spillover !
Fix haircut in case of default to d = 0.5 (Panizza et al., JEL 2009). (Not important for results that follow).
!
Pin down distribution of “fiscal limit” by targeting cross-section evidence (April 2011) on risk premia (CDS spread) on public debt max = 2.56. (shown before): αb = 3.70, β b = 0.54, and b
!
Spillover: set χψ = 0.55 (Harijes 2011).
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
14/26
The sovereign risk channel in action: analytical results !
!
As in Christiano/Eichenbaum/Rebelo (2010) and Woodford (2011), we assume that policy rates fall initially to the zero lower bound, and persist there with probability µ in each period. With probability (1-µ), monetary policy returns to the unconstrained Taylor rule, remaining unconstrained thereafter. the shock to the discount factor shock + e follows the same Markov structure. To provide analytical insight on the transmission, we simplify the model at first, assuming that the sovereign risk premium is a function of expected deficits, instead of debt.
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
15/26
What difference the RS-channel make? NKPC standard (all in deviations from SS) + t = βEt Π + t +1 + κy y˜t − κg g˜ t , Π
(1)
Euler equation/IS curve with interest rate spread , + t + 1 + φω 1 +t + + y˜t − g˜t = Et y˜t +1 − Et g˜ t +1 − σ¯ +itd − Et Π et
(2)
Monetary policy (possibly constrained)
+ t − φω 1 +itd = max{φπ Π +t , − log(1 + i d )}
(3)
Assume for now: default probability depends on expected primary deficit +t = ξEt (g˜ t +1 − φT ,y y˜t +1 ) (4) 1 Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
16/26
Proposition 1: determinacy region shrinks In the ZLB phase, posit government spending is exogenous; then the economy has a unique bounded equilibrium iff (a) µ(1 + ξχ2) < 1/( βµ) and
(b) (1 − βµ)(1 − µ(1 + ξχ2)) > µ2κy
Determinacy region the smaller the weaker the fiscal position (the larger ξ).
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
17/26
Self-fulfilling expectations At ZLB, monetary policy cannot respond to adverse shift in expectations. Sovereign risk can increase the indeterminacy problem. Say, agents expect lower output for some non-fundamental reason. Lower output means higher fiscal deficit. Higher deficit means higher spreads, which, in turn, depresses output—thus validating initial expectations.
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
18/26
Proposition 2: Determinacy regions enlarges with pro-cyclical spending Assume now that government spending follow a systematic rule. In the ZLB phase,procyclical spending rule (g˜ t = ϕy˜t , with ϕ > 0) — that implements cuts in recessions – may help anchoring expectations. However, as apparent from graph to follow, procyclical spending cuts are no panacea...
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
19/26
Determinacy region (grey) ZLB binds for an average of ... 7 qtrs
ϕ
6 qtrs 1
1
1
0.8
0.8
0.8
0.6
0.6
0.4 0.2
0.6
0.4
Procyclical response
0.4
Procyclical response
0.2
0 −0.4
0
−0.2
Countercyclical response
−0.2
Countercyclical response
−0.4 −0.6
−0.6
−0.8
−0.8
−0.8
−1
−1
0.1
ξ
0.2
0.3
0
0.1
ξ
0.2
Countercyclical response
−0.4
−0.6
0
Procyclical response
0.2
0
−0.2
8 qtrs
0.3
−1
0
0.1
ξ
0.2
0.3
Spending rule: g˜t = ϕy˜t Assuming 8 quarter average duration of ZLB-period: ξ = 0.05 for 110% debt-to-GDP ratio and ξ = 0.08 for 140% debt-to-GDP ratio. Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
20/26
Proposition 3: State-dependent Multiplier Effects Adjust government spending while ZLB binds (back to steady state afterwards). Focus on parameterizations that imply determinacy. The government spending multiplier is positive iff
(1 − µ ) −
µκ > µξ σ˜ 1 − βµ
!
In the absence of sovereign risk channel, multiplier would be strictly positive, regardless of the parameterization (Christiano et al., 2011, and Woodford, 2011).
!
With a SRchannel, multiplier effects can even very different magnitude or even different sign. Multiplier very sensitive to
!
" "
Introduction
expected durations of recession cum ZLB elasticity of risk premia to (dynamic) fiscal outlook
Model
Calibration
Analytical results
Simulations
Conclusions
21/26
Response of current output and deficit to spending cut Output
Introduction
Model
Calibration
Deficit
Analytical results
Simulations
Conclusions
22/26
IV. Dynamic simulations
Consider full model !
Sovereign risk depends on public debt rather than on deficit
!
Large recessionary (discount rate) shock pushes economy at ZLB
!
Exit from ZLB endogenous
Recession may be amplified by sovereign-risk channel !
Tax revenues fall
!
Sovereign-risk increases non-linearly
!
Exit from ZLB considerably delayed at high levels of initial debt
Introduction
Model
Analytical results
Simulations
Conclusion
29/38
Severity of recession: 60 %, 90 % and 110 % initial debt
output (% from ss)
risk-premium (ann. bps) 500
policy rate 3
400
−5
2
300
−10
200
1
100
−15 0
5
10
Introduction
15
20
Model
25
0
5
Analytical results
10
15
20
Simulations
25
0 0
Conclusion
5
10
15
30/38
20
25
Output effect of 2-year spending cut (2 percent of GDP each quarter)
Timing of ZLB exit endogenous
ZLB lasts for 7 quarters
ZLB lasts for 18 quarters
a) Effect on output (% deviation from ss) 2
2
2
1
1
1
0
0
0
−1
−1
−1 −2 0
5
60% debt/GDP 90% debt/GDP 115% debt/GDP 10 15 20 25
Introduction
Model
−2 0
−2 5
Analytical results
10
15
20
Simulations
25
0
5
Conclusion
10
15
20
31/38
25
Conclusions We have shown that the sovereign risk channel: !
tends to exacerbate the effects of negative cyclical shocks, unless the central bank manages to offset it by lowering policy rates;
!
tends to delay the exit from the zero lower bound;
!
makes the economy (more) vulnerable to belief-driven equilibria;
!
undermines key stabilisation prescriptions (e.g. regarding the desirability of systematic anti-cyclical fiscal policy);
!
affects the magnitude and even the size of fiscal multipliers, making them quite sensitive to markets’ beliefs about the duration of a recession, and the elasticity of the risk premium to the fiscal outlook.
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
24/26
Conclusions
Specific implications Effects of spending cuts forced by fiscal crisis in a ZLB environment (or, more generally, when monetary policy is constrained) !
Cuts may help keeping expectations anchored.
!
As negative deflationary effects run against falling spreads, the final outcome may vary substantially depending on:
!
Depth of recession (expected duration of ZLB episode).
!
State of public finances (response of sovereign spread to fiscal stress).
!
If fiscal situation is very bad to start with, however, cuts are less likely to be harmful.
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
25/26
Conclusions Non-standard monetary measure !
Anything that severe or contain the consequences of fiscal stress on the borrowing costs of local firms and households would stem the recessionary implications of sovereign crises.
!
If quantitative interventions could help keeping private borrowing costs low, Greece Spain Italy Portugal (but also Germany and France, you never know) could look more like Wisconsin or Kansas
Introduction
Model
Calibration
Analytical results
Simulations
Conclusions
26/26