Regulation and Policy Coordination in Normal

Regulation and Policy Coordination
in Normal and Crisis regimes
Joe Pearlman
City University
Participants and Deliverables
City (Pearlman, Melina) - D9.1, 9.3, 9.7
UvA (Hommes) - D9.4
CERGE (Slobodyan) – D9.2
CEP (Ragot, Iliopoulos) – D9.5, 9.6
• D9.1, D9.4 and D9.2 are being summarized at
this meeting
Background to Macroprudential
• Ayah El-Said has written a paper “On The
Impact of MacroPrudential Policy: Lessons
From Emerging Markets”.
• Emerging markets employed macroprudential
tools for at least two decades to pursue
financial stability and reduce systemic risk,
with monetary policy pursuing price stability.
• She has examined this in the context of a
structural VAR.
Countries of Interest
• Brazil
• Turkey
Czech Republic
South Korea
South Africa
Saudi Arabia
Monthly Data
1990 Onwards
• Uses the SVAR specification
• Y : a vector of 5 endogenous variables with monthly logarithms of indicators
of economic activity, prices, credit, monetary policy, and macroprudential
• Economic activity: unemployment/industrial production
• Prices: CPI/core CPI/Housing Specific CPI
• Monetary Policy: Policy Rate (ordered last, as it is fast-moving)
• Macroprudential Tools
loan to value ratios,
minimum and total regulatory capital,
Required reserve ratios, effective required reserves
provisioning values (buffer of banks’ own funds from retained profits)
Main Results
• LTV ratios are useful for managing price of housing –
significant mainly in Latin America
• Increase in provisioning curbs credit growth
• Regulatory capital does not have much effect
• Required or effective reserves show some effects on
• Macroprudential tightening tends to be associated with
currency depreciation which is in line with previous
• Macroprudential policies not always applied
countercyclically. In the UAE, the measures were used
to lower credit in bad times.
9.1 A stylized model of European monetary union for
analysing coordination games for monetary and macroprudential policy (Cantore, Levine, Melina, Pearlman)
• Initially use a model that includes two nontraded sectors, one traded sector, then estimate
using German and peripheral EU data.
• Then investigate a Nash game in simple rules for
monetary and macroprudential policymakers
• Currently this work is following Quint and
Rabanal (2014), with one or two modifications.
Description of Model
• Two-country, two-sector, two agent general equilibrium model of a single
currency area.
• Two types of goods, durables and non-durables, produced under
monopolistic competition and nominal rigidities (Calvo or Rotemberg)
• Non-durables are traded, durable goods are non-tradable.
• In each country, there are two types of agents, savers and borrowers, with
different discount factor and habit formation parameters. Both agents
consume non-durable goods and purchase durable goods to increase their
housing stock.
• Borrowers are more impatient than savers, which motivates credit
• To introduce credit frictions borrowers are hit by idiosyncratic quality
shocks to their housing stock, which affects the value of collateral that
they can use to borrow against.
• BGG then applies to residential investment: shocks to the valuation of
housing affect the balance sheets of borrowers, which affect the default
rate on mortgages and the lending-deposit spread.
Description of Model (cont)
• Domestic financial intermediaries take deposits from savers,
grant loans to borrowers, and issue bonds.
• International financial intermediaries trade these bonds
across countries to channel funds from one country to the
• Thus excess credit demand in one region can be met by
funding coming from elsewhere.
• International financial intermediaries charge a risk premium
dependent on the net foreign asset position of the country.
Modifications to Quint and Rabanal
• We firstly use a non-separable utility function.
This is consistent with
– balanced growth
– the observed relationship for interest rate in the
Euler equation (Collard and Dellas, 2012)
– the increase in consumption in response to an
increase in government spending (Bilbiie, 2009)
• Secondly we correct what appears to be an error in Quint and
• In their paper their macroprudential instrument is used to regulate
the ratio of borrowings/savings for impatient and patient agents.
The instrument reacts to either credit growth or credit/GDP relative
to steady state.
• This implies that borrowing could be > saving half the time!
• The ratio of mortgage lending to deposits for all banks in the
Bankscope database for 1990-2012 is about 0.5; this will represent
our base value therefore for the instrument.
• In addition we allow for the instrument to depend on GDP growth
as in Lambertini et al (2013), as recommended by Goodhart, and
following new BoE practice.
• The welfare in Quint and Rabanal uses a 2nd
order approximation, but we use the actual
nonlinear form (unlikely though to make much
• We follow Quint and Rabanal and evaluate the
effects of non-coordination of monetary and
macroprudential policies for a two bloc model
of the Euro-area.
• No results as yet.
WP11.2/3: Robust Policy Rules
(Amisano, Levine, McAdam, Pearlman)
• We evaluate competing models by Predictive
Density Forecasts
• Then use a sample of draws generated by
MCMC techniques to design robust simple
rules whose average welfare is maximized
across the sample.
• The latter technique is very similar to that of
Batini et al (2006) and Levine et al (2012)
Novelty of the Work
• The basic idea is that of Geweke and Amisano (2012):
• Models M1,…,MM: generate a series of 1-step ahead predictive
densities for y, and define the weighted sum of predictive
• A linear pool of the predictive densities is then created and the
optimal weights are derived by maximizing that pool of (log)
predictive densities:
• Maximizing over the weights usually does not lead to just
one model being vastly preferred over others
Modification to Geweke and Amisano
• Since our focus is robust rules, in order to correctly compare
models, we need to have the same rule (represented by
coefficients r), so we need to solve
• So the algorithm is as follows:
• Fix the rule, and optimize for r and wi using Bayesian
maximum likelihood (less time consuming than MCMC)
• Estimate each model Mj with optimal rule r using MCMC.
• Sample and design the optimal robust rule.
• All steps are in place, apart from the optimal search for r.
Which models will be used?
• Clearly there is a vast range of possibilities
• We will restrict ourselves to a small number of
different types of financial frictions, and also
address internal and external, and deep habit.
• Then evaluate the robust simple rules.

similar documents