Modeling General-Equilibrium Macroeconomic Stress Scenarios in MATLAB

Transcription

1 Modeling General-Equilibrium Macroeconomic Stress Scenarios in MATLAB Jaromír Beneš International Monetary Fund MATLAB Computational Finance Conference May 2015

2 Macroprudential Policy Is What? Narrow versus broader scope Broad policy objectives Minimize incidence of balance sheet crises Limit disruptions to key financial services Key elements in macroprudential analysis Tail-risk distress scenarios Feedback between balance sheets and real economy Possibility of severe nonlinearities Models to support macroprudential policy 2

3 Plan of Presentation Policymaking Perspective Modeling Perspective Simulation Experiment Programming Perspective 3

4 POLICYMAKING PERSPECTIVE 4

5 Macroprudential Policy Exercise Assess risks Design macro(- economic/- financial) stress scenario(s) Slow- burn (low- frequency) shocks and risks Unlikely yet plausible scenarios with large impact Evaluate impact of scenario(s) Resilience of sectoral balance sheets Feedback between financial and real sectors Communicate with policy makers Consider possible policy responses Communicate to public Regular Financial Stability Reports 5

6 General Equilibrium Model Framework Empirical estimates (PD, LGD, EAD, etc) CCA Real economy and monetary policy analysis Structural models Spreadsheet scenarios Financial stability analysis Time series analysis Sectoral analysis 6

7 Role of General Equilibrium Models Integrate all pieces of information Balance sheet consistency across sectors and time Facilitate internal communication (explicit assumptions, most critical assumptions) Make process accountable Make external communication transparent, credible Model-based scenario analysis, not accurate probabilistic predictions 7

8 Limitations Fundamental uncertainty Nonlinear feedback Corridor stability Estimation and backtesting difficult in crisis modeling Distress episodes are few and far apart Each has a different cause Need to evaluate international evidence è Common variety of macro models not well-suited 8

9 MODELING PERSPECTIVE 9

10 Macro Model with Credit Risk Real economy Relatively standard (DSGE) structure Monetary economy Optimizing agents with finite (short) planning horizons Mixed expectations Financial sector Credit risk on loan books Asymptotic single factor risk model Advanced IRB to model regulatory capital constraints 10

11 Structure of Model Real economy Banking sector Producers Exporters Households Loan portfolio Other assets Resident deposits Nonresid deposits Bank capital Rest of world Exports and imports 11

12 Real economy Banking sector Producers Exporters Households Loan portfolio Other assets Resident deposits Nonresid deposits Bank capital Exports and imports 12

13 Demand for Deposits Real economy Banking sector Producers Exporters Loan Δ t = φ " # P C,t C t + P I,t portfolio I t +κ P H,t H t Households Δ t = R t 1 D t 1 R L,t 1 L t 1 + NL t Other assets Resident deposits $ % Nonresid deposits Bank capital Exports and imports 13

14 Real economy Banking sector Producers Exporters Households Loan portfolio Other assets Resident deposits Nonresid deposits Bank capital Exports and imports 14

15 Individual Bank Loans Real economy i R L,t i L t i > k exp u t+1 ( ) P i H,t+1H Banking sector t default Producers Exporters i R L,t Households # ) % + % 1 λ Φ + $ % * log R i L,t Loan portfolio i L t Other assets # $% k P H,t H t i E t log P H,t+1 σ Resident deposits Nonresid deposits P H,t Bank capital &,& '(.( * ( = R t. -' ( Exports and imports 15

16 Real economy Banking sector Producers Exporters Households Loan portfolio Other assets Resident deposits Nonresid deposits Bank capital Exports and imports 16

17 Bank Capital Regulation Real economy Banking sector h t i Producers # Φ 1 i ( q) % pd t = Φ Exporters % $ ( ) Households Φ 1 q 1 ρ ( ) ρ Loan portfolio & ( ( ' Other assets Resident deposits Nonresid deposits Bank capital E t i i L t 1 R i L,t R t # i $ 1 λh t ( q) % & Exports and imports 17

18 PROGRAMMING PERSPECTIVE 18

19 IRIS Toolbox 60+ classes, 30+ packages, 2,300+ functions toolbox.com Structural modeling (DSGE) MV time series analysis Time series and database management Reporting Documentation 19

20 Model Related Classes @theparser 20

21 Building Model Equations Two types of equations Behavioral rules Stock-flow and other identities Behavioral rules Optimizing principles Rules of thumb Empirical equations Stock-flow identities Logical structure of the model 21

22 Transactions Flow Matrix Market clearing Transaction flows Capital account Current account Households Producers Exporters Banks Central Bank Rest of world Σ Consumption PC C PC C 0 Wage bill W N W NY W NX 0 Imports PM MY PM MX PM M 0 Exports PX X PX X 0 Distrib surplus producers Π Π 0 Distrib surplus banks Γ Γ 0 Distrib surplus CB Ω Ω 0 Deposit interest RD(0) D(0) RD(0) D(0) 0 Loan interest RL(0) L(0) RL(0) L(0) 0 Loan loss L(0)[1+RL(0)] UL L(0)[1+RL(0)] UL 0 CB liquidity surplus interest R(0) B(0) R(0) B(0) 0 Net foreign liabs interest RF(0) F(0) RF(0) F(0) 0 Net acquisition of housing PH ΔH 0 Chng in deposits ΔD ΔD 0 Chng in loans ΔL ΔL 0 Chng in CB liquidity surplus ΔB ΔB 0 Chng in foreign liabs ΔF ΔF 0 Σ Budget constraints 22

23 Net Worth Matrix Delegated agents Change in net worth Transaction chng Reval Households Producers Exporters Banks Central Bank Rest of world Opening net worth VH0 0 0 VB(0) VC(0) VR(0) Net acquisition of housing PH ΔH Chng in deposits ΔD ΔD Chng in loans ΔL ΔL Chng in CB liquidity surplus ΔB ΔB Chng in foreign liabs ΔF ΔF Revaluation of housing ΔPH H(0) Revaluation of net foreign liabs J F(0) Closing net worth VH 0 0 VB VC VR Laws of motion for net worth (equity) 23

24 Stock-Flow Builder 24

25 Stock-Flow Builder 25

26 Stock-Flow Builder 26

27 Behavioral Equations (Regexp Parser) 27

28 SIMULATION EXPERIMENT 28

29 House Price Bubble and Burst Irrational expectations House price growth Consumption, real economy Easier cheaper credit Risk perceptions down

30 House Price Bubble and Burst Irrational expectations Sequence of shocks Burst of the bubble Unexpected event House prices go down to fundamentals Painful deleveraging in both real and financial sector Re-simulate with loan-to-value caps Inequality constraint in households decision Complementary slackness translated into min( ) 30