monetary policy, banking and systemic risk in open economies

Monetary policy, banking and systemic risk in open economies

Jaromir Benes (IMF)Andrea Gerali (Banco d’Italia)

David Vavra (Czech National Bank)

Plan of presentation

• Motivation• Features added• Prototypical SOE model• Policy experiments

Motivation

• A simple (DSGE) model framework with interactions between real and banking sectors

• Provide dynamic and macro consistency in systemic risk, early warning, or contagion exercises

• Integrated approach to monetary cum macro-prudential policies

• Evaluate policy options under various constraints– Shock to bank capitalisation– Currency mismatches– Maturity mismatches– Booms and busts in asset prices

Motivation

• Reminiscences of 20 years ago– Monetary policy paradigm not established in the early 1990s

as it is now– Model-based frameworks popped up (BoC, RNBZ) with many

ad-hoc features that became justified by proper theory only later on

• Our work tries to incorporate some of the important links between real economy and banking (and monetary policy and macro-prudential policy) taking a few shortcuts to keep the framework operational, e.g.– no explicit debt/loan contracts– cost function increasing in banks’ leverage

Features added

• Banks as agents with their own net worth– Bank capital subject to regulation– Bank capitalisation affects lending rates and volumes– Fresh capital not trivial to raise

• Banks bear (some of the) aggregate macro risk: non-performing loans– Different from most of the current literature. Bernanke,

Gertler & Gilchrist 1998 accelerator assumes debt contract contingent upon macro outcomes

– Bank capital subject to losses

Features added

• Simple housing– fixed supply of houses– house prices subject to bubbles

• Multi-period loans– banks issue multi-period loans, refinance themselves short– hence exposed to maturity mismatches

• ...hence multiple balance-sheet effects– currency mismatch risk– loan-to-value ratio affects premium–maturity mismatch risk

Design of the real sector

Consumption

Imports

Local production

Exports

Intermediates

Design of the real sector• Simple but flexible to get aggregate elasticities right• Roundabout production function

Pro-cyclical real marginal cost, no explicit labour market

• Imports both directly consumed (Leontieff) and used as inputs (Cobb-Douglas)– Helps to flexibly calibrate the aggregate elasticity of import

demand and exchange rate pass-through to the CPI

• Price-elastic export demand with export prices subject to costs of deviating from world prices– A wide range of assumptions about responses in export prices

and export volumes

• Simple housing (fixed supply of houses) => LTV

Design of the banking sector

Loans to consumers

Foreign funds

Bank capital (equity, net

worth)

Design of the banking sector

• Consumers net debtors at all time, foreign borrowing intermediated through banks

• Banks combine foreign funds and their own net worth (bank capital, equity) to make loans

• Bank capital is made indispensable by introducing a cost function increasing in leverage:– Regulatory costs– Reputational costs– Smooth cost function rather than an inequality constraint

analogy with inventory stock-out models

Design of the banking sector

• Banks extend multi-period loans• Multi-period loans can be handled easily on the

consumers’ side• …but to keep the problem tractable on the banks’

side, we in fact split the bank into its “wholesale” and “retail” branches that take the other’s decisions as given

• This split is (for ease of this exposition) not presented here

Banks

(For ease of notation here: all assets and liabilities except F denominated in local currency.)

• Balance sheet• Gross earnings

tt ttL B F E

1, 1 1 1 1 1 1

1 1

*

1·

: 1

/

t

t

St L tt tt tt S

tt

t

t

gV R L R

f E

R B F

E L

Non-performing loans Banks’ costs

increasing in leverage

Banks

• Banks must follow a fixed dividend policy

• This is to give bank capital non-trivial role– capital not easy to raise fresh capital– consumers (owners) cannot simply pour money into banks

to re-capitalise them– shock to capital (leverage) costly for the banks

dividends ·new equity (1 )·

tt

tt

dVE

V

What does the cost function do?

• Prevents banks from going infinitely leveraged—return on equity diminishes in leverage

• Affects marginal cost of lending => lending rates

• After a hypothetical shock to bank capital:– the total costs increases…– …but the marginal costs increase more still– so does retail lending rate– lending volumes drops in response

, /L tt ttR R f e

, , 1: ( / )1 t

ttL

E t L tt ttE tR R R R f Lg E

Non-performing loans

• NPLs are still repaid by the consumers, but the repayments never reach the bank

• NPLs are an ad-hoc function of some macro variables• We experiment with NPL functions decreasing in– loan-to-value ratios (=used in simulations here)– loan-to-current-income ratios

• Non-linear function with a “threshold”• Must be, though, sigmoid (flattens for very large

values) – otherwise the simulation would explode

Non-performing loans

Multi-period loans

• Model the effects of the existence of multi-period (nominal) loans, not portfolio/term-structure choice

• Introduce a “geometric” loan– infinite number of geometrically decaying instalments– instalments cannot be re-negotiated at a later time

• Why geometric?– everything can be expressed recursively– average maturity (Macaulay’s duration) can be calibrated

using just one parameter– no new state variables needed to mimic very long terms

Multi-period loans

• Average maturity (duration) imposed, not determined endogenously or optimally.

• Consumer ex-ante intertemporal choice not affected (up to first order) by multi-period loans: Euler equation still has the underlying one-period rate in it.

• Ex post, duration of loans matters to the extent the economy is hit by unforeseen shocks (e.g. large increases in short-term rates make consumers better off if they go long).

Simulation experiments

• Expose the country to a premium shock• Full dollarisation of the banking sector and the loans• NPLs function of loan-to-value ratio• Simulate two policy regimes– IT with a flexible exchange rate– An exchange rate peg

• Simulate two magnitudes of the shock– a “small” shock (100 bp)– a “large” shock (1,000 bp) going beyond the threshold of the

NPL function, resulting in sizeable balance sheet effects

100 bp country premium shock

1,000 bp country premium shock

Comments on simulations

• First, notice the non-linearities from the NPL function– Large shock simulation is more than just 10x the small shock

simul (the shock is 10x bigger, in a linear world the simulations would be identical, just the magnitudes would multiply), and variables have different profiles

– Output loss under IT is closer to output loss under peg in large shock simul (than in small shock simul) – this the balance sheet effect. A large depreciation (plus drops in house prices) raises the loan-to-value ratio significantly, and triggers defaults (NPLs)

• Second, let’s turn to the large shock simulation. Banks run huge losses in the first period (unexpected NPLs were not reflected in the lending rate setting)

Comments on simulations

• How do the banks react to losses and drops in bank capital?

• Their total costs increase, but even more so the marginal costs. The banks raise the lending rates significantly.

• This has two main implications:– The banks start making profits, and cumulate bank capital

again (recall profits are the only source of new capital).– Demand for loans drops.

Comments on simulations

• In real world, the banks would not lift the lending rates so much (above 50 % PA in simulations – not shown in the graphs), but would combine lending rate increases with credit rationing.

• However, with credit rationing, the shadow value of loans would increase exactly so as to depress demand sufficiently to restore equilibrium at the rationed levels.

• Whether the drop in loans is because of credit rationing or high market rates is therefore irrelevant.

Comments on simulations

• Differences in the two policy regimes:– An IT central bank can “transform” an interest rate shock

to an exchange rate shock (by cutting the rates). The exchange rate shock is more favourable to the real economy, because re-directs demand from foreign goods towards local goods, whereas interest rates shocks depress overall demand.

– On the other hand, flexible exchange rates can trigger large valuation effects, seeing the households default on their debt, and the banks run losses.