Financial Frictions and Credit Spreads. Ke Pang Pierre L. Siklos Wilfrid Laurier University

Transcription

1 Financial Frictions and Credit Spreads Ke Pang Pierre L. Siklos Wilfrid Laurier University 1

2 Modelling Challenges Considerable work underway to incorporate a role for financial frictions the financial crisis highlighted this as a weakness of DSGE models Understood here to represent a cost that gives rise to a spreadbetween borrowing and lending rates. A wedge exists (i.e., due to asymmetric information) between borrowers and savers, and there are intermediation costs financial institutions must absorb 2

3 Selected Spreads: view I 10 8 Diffe erential in percent (%) I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II year versus 3 month 3 month verus fed funds libor versus OIS prime rate versus yield on M2 Taylor & Williams 2009 staple of many Term structure studies May be representative of theoretical spread In this study 3

4 Selected Spreads: view II Percent (%) Percent (%) I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II year vs 3months fed funds vs 3 month libor-ois prime rate vs M2 yield 4

5 Modelling Challenges The canonical model (Woodford2003) has been seen as not ideally suited to handling to capital market imperfections In general, the weaknesses of the New Keynesian paradigm are well-known (Goodhart2008, Tovar 2008, Chari et.al. 2009) But...its either the best we have or it may be more fruitful to repair it rather than discard it completely Provides a disciplined way of thinking about interactions of key macroeconomic variables 5

6 Focus of the Study Spreads play a central role in the transmissions mechanism Bernanke and Gertler(1989, 1995) The economy is interest sensitive, that is, there exists a credit channel May operate through balance sheets or bank lending behaviour We don t take a stand on one versus the other although focus is on the latter in this study Walsh (2009)...factors that generate movements in spreads, or the degree to which these movements reflect inefficient fluctuations that call for policy responses still eludes us In particular why are spreads subject to sharp movementsand why can they be so volatile? Do they really matter (in a crisis): NO Chari. et.al. 2008); YES (Cohen-Cole e.al.2008) 6

7 Overview of the Approach of the Paper Credit frictions model of Curdia& Woodford(2009, 2009a) is starting point NOTE: has changed in its various incarnations The model is adapted to the concerns of this study, namely attempting to replicate movements and volatility in spreads Agents are heterogeneous, intermediation is inefficient or costly Actual U.S. time series are used for exogenous factors (e.g., TFP shocks, government spending) We try to replicate movements in selected spreads We explore the impact of two types of monetary policies QUANTITATIVE EASING: varying the amount of aggregate reserves to influence the spread between the fed funds rate and the interest rate on reserves (liabilities of the Fed s balance sheet) CREDIT EASING: debt-financed fiscal policy (asset composition of the Fed s balance sheet) 7

8 MODEL: Households I 2 types of households bmore impatient than s b borrows, s saves Remain the same type form one period to the next with prob. [δ, 1-δ] δ Borrowing is done ONLY via intermediary One period contracts (riskless) + households can insure against various risks Necessary because 1. heterogeneity of households; 2. credit frictions; 3. risk sharing Represents a key source of financial frictions Lifetime utility function t= 0 β t τt( i) τt( i) { U [ ct()] i V [ ht()]} i Household i s(net) wealth A i B i i B i i D i T i d b int t() = [ 1()] + (1 1) [ 1()] t + t + t (1 + t 1) + t () + t() Deposit and borrowing rates (riskless) Budget constraint g B () i = A() i Pc () i + Wh() i + D () i + T () i t t t t t t t t 8

9 MODEL: Households II B t (i) is the budget constraint Lifetime utility is maximized subject to A t (i)& B t (i) Euler equation governs labour supply w λ = V [ h ] τ t ( i) τ t ( i) τ t ( i) t t h t Optimal consumption for borrower (b), saver (s) 1 + i λ β {[ δ (1 δ) π ] λ (1 δ) π λ } b b t b s t = + b t+ 1 + s t+ 1 Π t i λ β {(1 δ) π λ [ δ (1 δ) π ] λ } d s t b s t = b t s t+ 1 Πt+ 1 inflation Probability of being type bor s 9

10 MODEL: Financial Intermediaries Perfectly competitive Intermediation costs are non-linear (convex function) Interest rates are given, determine supply of loans to maximize profits Leads to a functional form that describes spread Intermediation costs create a spread & changes, NOT increased risk Technolog[ies] d = b + Φ( b) t t t Spread Real deposit, real credit Equilibri[a] d = b + Φ( b b) t t t b d 1 + i = (1 + ω )(1 + i ) t t t ω = Φ ( b) Benchmark/Modified t t ω = Φ ( b b) t t 10

11 MODEL: Firms & Government Firms A single good Perfectly competitive price takers Isoelasticproduction function (subject to a TFP (i.e., productivity) shock [ TFP is exogenous] Government Budget is balanced every period [spending and transfers are exogenously given] 11

12 MODEL: Monetary Policy A Taylor type rule Contemporaneous The policy rate is the deposit rate CB makes optimal policy projections that asymptotically approaches the s.s. (Svensson& Tetlow 2005) Model closed with 2 market clearing conditions Policy rule i π y Π Y = ı, γ π, γy 0 Π Y d d t t t γ Goods & labour markets b s Y = π c + π c + G + Φ( b) t b t s t t t h = π h + π h b s t b t s t γ 12

13 Evolution of b Aggregate over all borrowers A i di δp b i δπ D δ π A B t b int t() = t 1 t 1(1 + t 1) + b t + (1 ) b t Aggregate budget constraints b b g Pb = A() i di + π ( Pc Wh D T ) t t B t b t t t t t t t Substitution yields debt dynamics: A = P [ d (1 + i ) b (1 + i )] + D d b int t t 1 t 1 t 1 t 1 t 1 t { c b, c s, h b, h s, b, Y, h} t t t t t t t QUANTITIES b s b s b = π π [( c c ) w ( h h )] π Φ( b) t b s t t t t t b t d (1 + it 1) [ bt 1 πb ( bt 1 ) πsbt 1ωt 1 ] Πt { Zt, gt, τt} EXOGENOUS δ + + Φ + d { i, Π, ω, w } t t t t PRICES 13

14 Calibration -Baseline η= = 51.6 (Curdia-Woodford) δ=0.9 π b = π s =0.5 β/i d = 4% φ s =1, φ b /h for both types the same in s.s. ω=2% in s.s. Debt/GDP=80% in s.s. Y,Z=1 in s.s. Conventional TR coeffs 1 1 τ τ σ τ τ τ θ ( ct) τ t σ 1 1 τ σ U ( c ) =, > 0 τ τ 1+ ν τ τ ϕ ( ht ) V ( ht ) =, ν ν Φ ( b) = ϕb η, η > 1 t t 14

15 Calibration Sensitivity Analysis I eta phi phib thetab thetas beta

16 { Z, g, τ } t t t TFP from Chen et.al. (2008) 16

17 Results Solution is numerical to non-linear equations Allow 200 periods [years] to converge (happens much faster Implies 1600 equations What TFP? Role of exogenous drivers Simulated spreads: what they look like Model assessment: a bird s eye view Impact of unconventional monetary policies 17

18 Which TFP? TFP_Chen TFP_FM TFP_FMU TFP_FMLP

19 The role of Specific Shocks I 19

20 The Role of Specific Shocks II 20

21 Simulation I: Benchmark.04 Long -term Govt bond less 3 month Tbill :1 1965:1 1970:1 1975:1 1980:1 1985:1 1990:1 1995:1 2000:1 2005:1 Simulated Actual 21

22 Simple test I Variable Coeff. Std error Z-statistic p-value C omega Eta =

23 Simple test Ia Variable Coeff. Std error t-statistic p-value C omega Eta = 2 23

24 Simulation II: Benchmark 3 month less fed funds rate Simulated Actual 24

25 Simple test II Variable Coefficient Std Error t-statistic p-value C Omega Eta =

26 Simulation III: Benchmark Prime rate less M2 yield :1 1965:1 1970:1 1975:1 1980:1 1985:1 1990:1 1995:1 2000:1 2005:1 Simulated Actual 26

27 Simple Test III Variable Coeff Std error t-statistic p-value C Omega Eta =

28 Simulations vsfacts I Standard Dev of Inflation ACF (1) H-P filtered PCE inflation Chen TFP = FMTFP = FMLP = FMU =

29 Simulations vsfacts II Statistic Omega(bench) eta=2 Omega(bench) eta = 51.6 Omega(10 yr LESS 3 m) Mean ( 03-09).014( 34-09) Std. Dev AC (1) * AC (2) AC (3) AC (4) AC (5) * 1 st difference 29

30 Varieties of Omegas Be enchmark cases Modified case es OMEGA (benchmark) eta = 2 OMEGA (benchmark) eta = 51.6 OMEGA (modified) eta = 2 OMEGA (modified) eta =

31 Debt Dynamics steady state = B_166 B_216 B_24 B_316 B_366 B_416 B_466 B_516 B_566 B_616 B_666 B_716 B_8 31

32 Convexity steady state = b (modified) eta = 2 b (modified) eta = 51.6 b (benchmark) eta =

33 Inflation PI_CHEN PI_FMLP PI_FM PI_FMU Steady state = 1 33

34 Comparing Inflation PCE inflation Simulated inflation (eta=51.6), TFP from Chen 34

35 Sources of Changes in the Spread η 51.6 ω = η = 51.6 ω ω j ω g -0.06(.70) ω g,ω TFP -0.26(.08) ω τ 0.81 (.00) ω g,ω τ -0.05(.73) ω TFP 0.88 (.00) ω TFP,ω τ 0.46 (.00) IVE Est η = ω = + ω τ + ω g + ω (. 0 0 ) * ( ) * ( ) ( ) * * = 1 %, + = 1 0 % T F P 0.60 (.00) 0.96 (.00) 0.98 (.00) p-values in parenthesis 35

36 Unconventional Monetary Policies Credit easing Central bank does NOT incur the same intermediation costs (passed on in the GVT budget) Quantitative easing An increase in the monetary base (i.e., bank reserves) For C.E. add a new element to b Zero in s.s. b = L + L cb t t t Profit max as before but s.t. Where st.. d = L + Φ( L) t t t Φ ( L) =ϕ t L η t For Q.E. Augment intermediation cb R + d = b + Φ( b) t t t t 36

37 Dynamics of Unconventional MP Credit Easing Quantitative Easing b s b s cb b = π π [( c c ) w ( h h )] π Φ( b L ) t b s t t t t t b t t d (1 + it 1) [ 1 ( cb bt πb bt 1Lt ) πsbt 1ωt 1 ] Πt δ + + Φ + d cb πb(1 + it 1) Lt 1 + ( ωt δ) Π t b s b s b = π π [( c c ) w ( h h )] π Φ( b) t b s t t t t t b t δ + + Φ + d (1 + it 1) cb [ bt 1 πb ( bt 1) πs bt 1 ωt 1 πb Rt 1 ] Πt 37

38 IRFs: CE vsqe.2 Benchmark - credit easing X Benchmark - quantitative easing Modified - credit easing Modified - quantitative easing Change in omega CE:1% increase CB s.scredit, L cb =.01 b cb QE:1% of s.s. Credit R = 0.01b 38

39 Accum. IRFs: CE vsqe Benchmark - credit easing X 10 2 Benchmark - quantitative easing Modified - credit easing Modified - quantiative easing Accumulated change in omega cb R = 0.01 b 39

40 What s next? Simulations Change inflation target Relax perfect substitutability of government debt & deposits Relax the costless CE easing policy assumption Consider other kinds of financial shocks Empirical Many ways to proceed but... 40

41 Conclusions A highly level of convexity is needed to match sharp movements and volatility in the spread a less non-linear intermediation costs function would lead to conditions as in the Great Moderation A challenge is to link this type of phenomenon to how intermediation costs are actually determined Credit easing when its a free lunch to the CB can reduce the spread QE is less effective and actually leads to a rise in the spread. This appears to describe the early days of the crisis in the fall of