Impact of Macroprudential Policy Measures on Economic Dynamics: Simulation Using a Financial Macro-econometric Model Koji Nakamura E-mail: kouji.nakamura@boj.or.jp Bank of Japan “The Interaction of Monetary and Macroprudential Policy” Reserve Bank of New Zealand Workshop 22 October, 2014 The views expressed here are those of the authors and should not be ascribed to the Bank of Japan. 1
Presentation Outline 1. Introduction 2. Financial Macro-econometric Model (FMM) 3. Simulations and Results 4. Discussions and Extensions 2
1. Introduction • Motivation • Brief Literature Review 3
Motivation • After the global financial crisis, “financial cycles” attract more attention of policy makers. • There is a hope that “macro-prudential measures” could counter “financial cycles.” • However, we do not have sufficient experiences of macro-prudential measures so far. • We need some “experiments” of macro- prudential measures to check pros and cons. 4
Brief Literature Review • Single macro-prudential measure with a small theoretical model: Cristensen et al. (2011) • Single macro-prudential measure with a small empirical model: Aiyar et al. (2012). • Multiple macro-prudential measures with a small theoretical model: Angelini et al. (2011) and Goodhart et al (2012). • Multiple macro-prudential measures with a large empirical model: This paper. 5
2. Financial Macro-econometric Model (FMM) • Overview of the FMM • Structure • Feedback loop • Banks’ activities • “Expectation channel” • Spending activities • Monetary policy 6
Overview of the FMM • The FMM is used for BOJ’s macro stress testing exercise and is a medium-sized structural model with the detailed financial sector and the macro economic sector. • About 120 banks* are explicitly modeled with actual banks’ data such as capital, loan amounts, and transition probabilities. • The macro economic sector is simpler than other large scale macro models such as FRB/US, but incorporates a feedback loop between the macro economy and the financial sector. * The latest version of FMM includes 373 banks and has been improved for the financial sector. See Kitamura et al. (2014) for the details of the latest version of FMM. 7
Structure <Financial Sector> <Macroeconomic Sector> Credit cost, Lending interest rate, GDP (Corporate capital spending, Capital adequacy ratio, Household expenditure, etc) Bank earnings, Lending volume Corporate earnings, Employee compensation <Expected Growth and Asset Price Factors> Expected growth rate, Stock prices, Land prices 8
Model Performance Evaluation 9
Feedback Loop 10
Second-round Effect 11
Banks’ Activities (1) • Banks’ activities affect the real economy through loan amount and loan interest rate. Bank i’s loan amount to corporate sector = Bank i’s fixed effect + 1.5*expected growth rate - 1.9*(Bank i’s loan interest rate – CPI ) + 0.4*Bank i’s capital ratio gap + 0.3*land price Bank i’s loan interest rate = Bank i’s fixed effect + 0.95*Bank i’s funding rate + 0.01*loan amount gap Bank i’s funding rate = Bank i’s fixed effect + 0.7*policy rate – 0.1*Bank i’s capital ratio gap Land price = -4.00 + 0.16*nominal GDP growth + 1.03*loan amount (-1) + 1.83*CPI Capital ratio gap is the difference between actual capital ratio and its regulatory level. Loan amount gap is the difference between actual loan amount and its level consistent with potential GDP. 12
Banks’ Activities (2) Individual banks’ credit costs are influenced by the developments of • macro economy and borrowers’ financial condition. Bank i’s credit cost Credit cost = Σ m Σ n(transition probability of Bank i’s self-assessment from m to n)* (loss ratio at time of downgrading of Bank i’s self-assessment from m to n)* (exposure of Bank i’s self-assessment of m) Transition probability = Bank i’s fixed effect + α *nominal GDP growth + β *borrowers’ liquid asset-liability ratio*nominal GDP growth + γ *borrowers’ interest coverage ratio*nominal GDP growth The specification of transition probability is revised recently. For more details, see Kitamura et al. (2014). 13
“Expectation Channel” • Okina, Shirakawa, and Shiratsuka (2000) : “the intensified bullish expectations which played an important role behind the large fluctuations in asset prices and the economy.” • In FMM, real expenditures and asset prices are affected by expected growth rate. Expected growth rate = 0.77*potential GDP growth rate + 0.10*actual GDP growth rate Stock price = 9.63*corporate profit + 1.39*expected growth rate + 0.32*U.S. stock price 14
Spending Activities • Household expenditure is affected by expected growth rate through stock prices and bank loan amount. • Capital spending by firms is affected by expected growth rate directly and bank loan amount. Household expenditure = 0.54*labor income + 0.02*stock price + 0.15*loan amount to household – 0.36*loan interest rate Capital spending by firms = 9.0*firm profit + 0.65*expected growth rate – 1.52*(loan interest rate - CPI) + 0.77*loan amount to firms 15
Monetary Policy • Monetary policy is a simplified Taylor rule. R(t) = 0.957*R(t-1) + 0.042*output gap(t) • Since we focus on the average nominal GDP growth and standard deviation in nominal GDP, price/inflation development is exogenously determined. • We could extend the model to treat price/inflation as an endogenous variable. 16
3. Simulations and Results • Simulation procedure • “Bubble economy shock” • Macro-prudential measures • Simulation 1: Fixed duration • Simulation 2: Various policy strength • Simulation 3: Recognition lag • Simulation 4: Use of reference indicators • Assessment of resilience of financial system 17
Simulation Procedure • First, we create “the bubble economy” by producing “shocks” in expected growth rate in order to mimic the actual bubble economy. • Then, we implement counterfactual simulations by using the identified “shocks” and macro-prudential policy measures. • We will look at the average nominal GDP levels and the standard deviations for both baseline and counterfactual cases. 18
“Bubble Economy” • As a “baseline” scenario, we create “bubble economy.” • The expected growth rate is affected by the potential and actual GDP. If the actual GDP is better than the potential, the optimistic expectations are built. Expected growth rate = 0.77*potential GDP growth rate + 0.1*actual GDP growth rate + shock(t) Shock(t) = iid shock(t) + cumulative growth shock(t) Cumulative growth shock(t) = 0.2*(cumulative growth shock(t-1) + (actual GDP growth rate - potential GDP growth rate) )*I(t), where I(t) = { 1 if the economy is in expansion phase, -1 otherwise}. The parameter of cumulative growth shock (0.2) is set so that the simulation recreates the economic fluctuations during the bubble period in Japan. The duration of I(t) is fixed (4 years) to mimic the actual bubble economy. 19
Macro-prudential measures • We use five macro-prudential measures as follows. Time varying capital requirement (countercyclical Capital Buffer, CCB) : Regulatory capital adequacy ratio is raised when credit-to-GDP ratio is above a certain threshold. Credit growth restriction : Restrictions on YoY growth in both household and corporate lending when credit growth is above a certain threshold. Corporate Loan-To-Value (LTV) regulation : Restrictions on YoY growth in corporate lending when corporate LTV ratio is above a certain threshold. Retail LTV regulation : Restrictions on YoY growth in household lending when retail LTV ratio is above a certain threshold. Debt-To-Income (DTI) regulation : Restrictions on YoY growth in household lending when DTI ratio is above a certain threshold. 20
Simulation 1: Fixed duration (1) • First, we run a simulation in which we use the macro- prudential measures with fixed duration (1Y, 2Y, 3Y and 4Y). • For 1Y case, we activate CCB in the first year of the “bubble” economy, then stop it in the second year. For 2Y, we activate CCB for the first 2 years and stop it in the third year (= the policy duration is 2 years). and so forth…For 4 Y case, we activate CCB for an entire “bubble” period. • How much should we increase the level of CCB? We increase the ratio by ¼ σ (=historical standard deviation) as a benchmark case. • The same “strengths” are used for other macro-prudential measures: reductions in bank loan amount growth by ¼ σ for the total loan, the household loan, and the corporate loan. 21
Simulation 1: Fixed duration (2) The baseline and policy simulations are “visualized” as follows. • As shown, the baseline scenario shows “big swings.” • With macro-prudential measures, such “big swings” are repressed, but the • degrees of the policy effects are different. tril. yen 490 480 470 460 Baseline Retail LTV and DTI regulations 450 Corporate LTV regulation 440 Credit growth restrictions Time-varying capital requirement 430 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 years 22
Recommend
More recommend
