Macro Models Unit
20 August 20041
Models, Forecasting and the Transmission
Mechanism of Monetary Policy
Presentation to the Portfolio and Select Committee on Finance
Cape Town20 August 2004
Macro Models Unit
20 August 20042
Contents
• Why do central banks need forecasts• General remarks about models and
forecasting• Models used by the SARB• Communicating the forecast• The monetary policy transmission
mechanism• Concluding remarks
Macro Models Unit
20 August 20043
Why Are Policy Makers Concerned With the Future?
• Monetary policy works with a considerable time lag
• Long time lags from changes in interest rates to output and from output to inflation
• It differs across countries and over time
• Policy makers wish to know what will happen if they change the central bank’s interest rate
• Our knowledge about the MPTM is limited
Macro Models Unit
20 August 20044
Monetary policy and forecasts
• Regardless of the monetary policy framework, a policymaker must have a view of the future because of the existence of transmission lags
“Implicit in any monetary policy action or inaction is an expectation of how the future will unfold, that is, a forecast. There is no way to avoid making a forecast, explicitly or implicitly.” Alan Greenspan, 1994
Macro Models Unit
20 August 20046
Central Banks & models
• When sufficient data are available, central banks should make use of econometric techniques and of models:– To understand and quantify how the
economy and monetary policy works (i.e. the transmission mechanism)
– Monitor where the economy stands and make short-term projections
– Make longer-term projections to determine impact of monetary policy
CCBS Handbook in Central Banking, #3
Macro Models Unit
20 August 20047
Benefits of Models
• Provide a simple framework with which to analyse and quantify short to medium-term macroeconomic developments
• Assists in the construction of forecasts in a systematic and consistent manner
• Useful for alternative policy simulations
Macro Models Unit
20 August 20048
Requirements for Models (1)
• To capture structural changes in the economy
• Not necessarily to supply precise forecasts, but at least a good indication of directional effects of policy options
• To correspond to linkages in the economy and to track the dynamics in the model easily
Macro Models Unit
20 August 20049
Requirements for Models (2)
• To follow a pluralistic approach: not one model for all occasions, but a suite of models
• To follow a pragmatic approach: supplement model output with surveys, value judgements, etc.
• To organise thoughts about policy and the economy, but their limitations must always be kept in mind.
Macro Models Unit
20 August 200410
Difficulties in modelling individual behaviour
• A market economy is steered by decisions of millions of consumers and workers, and hundreds of thousands of managers, investors and entrepreneurs
• A gigantic task to take account of the different circumstances in which each finds him/herself and the different choices and incentives they face
Macro Models Unit
20 August 200411
Modelling with and without theory
• Econometricians learn from observing past behaviour of the economy as a whole, rather than individual behaviour
• Econometrics is based on probability theory – used to test the validity of economic theory
and– also to look for relationships with no
underlying rationale in economic behaviour
Macro Models Unit
20 August 200412
Steps in model building
• Literature study• Gather data and estimate
individual equations• Model validation
– Tests for system dynamics & stability– Test forecasting ability and accuracy:
A lengthy process
Macro Models Unit
20 August 200413
A modern approach to modellingEconomic theoryEconomic theory
Generalempirical model
Generalempirical model
DataData
Are the dataI(0) or I(1) ?Are the dataI(0) or I(1) ?
Are there cointegrating relationships ?
Are there cointegrating relationships ?
I(0) I(1)
YesYes NoNo
Data transformationData transformation
Estimation: - model in levels - model in first difference - ECM - VAR - VECM
Estimation: - model in levels - model in first difference - ECM - VAR - VECM
Specification testing and diagnostic checkingSpecification testing and diagnostic checking
Is the model statistically adequate ?Is the model statistically adequate ?
NoNo YesYes
Tests of economic hypothesesTests of economic hypotheses
Using the model for predictions and policyUsing the model for predictions and policy
Macro Models Unit
20 August 200414
Hold-out Periodor
Out-of-sampleex-post forecast
Hold-out Periodor
Out-of-sampleex-post forecast
Ex-ante forecastEx-ante forecastEstimation period
orIn-sample forecast
Estimation periodor
In-sample forecast
t1980t1980 t2000t2000 t2004t2004
Root mean square errors (RMSE’s)Root mean square errors (RMSE’s)
tt
YY
Model evaluation
Macro Models Unit
20 August 200416
Structural and Atheoretical models developed and used by the SARB
• Core model• Small-scale model• Phillips-curve model • Vector auto-regressive (VAR) model• Auto-regressive integrated moving-average (ARIMA)
model• Indicator models• Disaggregated inflation model
Assist in thinking about a wide range of issues in a structured and quantified way and fosters debate around modelling, forecasting and monetary policy issues
Macro Models Unit
20 August 200417
Core model
• Keep relatively small: 63 equations of which 25 are structural equations
• Focus on inflation and other key economic variables impacting on inflation
• Economic theory determines long-run relationship between variables
• No long-run trade-off between inflation and output
• Short-run dynamics to explain short-term fluctuations in variables
Macro Models Unit
20 August 200418
Small-scale model
• Highly aggregated: 5 equations
• Easier to experiment with different
behavioural equations
• This type of model is used
primarily by some central banks
• Limitations in use of model
Macro Models Unit
20 August 200419
Phillips-curve model
• Single equation model
• Useful to describe the determinants of inflation
and for inflation forecasting
• Original concept relates to wage inflation and
unemployment, but modern versions relate to
price inflation and output gap
• Cross-check forecasts derived from core model
Macro Models Unit
20 August 200420
Vector auto-regressive (VAR) model
• Dynamic interaction between a set of variables
• Do not require strong theoretical assumptions -
model is rather based on actual trends in data
• Useful over short time horizons
• Primarily used for short-term forecasting (up to 4
periods)
• Impulse responses describe reaction of variables on
exogenous shocks
Macro Models Unit
20 August 200421
Auto-regressive integrated moving-average (ARIMA) model
• Single equation model
• Models a single variable in terms of an auto-regressive component and a moving-average component
• Easy to estimate and solve the model
• No strong theoretical assumptions
• Useful over shorter time horizons
• Disadvantage : ARIMA models cannot predict turning points easily as they are based on historical figures (or trends of the past)
Macro Models Unit
20 August 200422
Indicator models
• Utilised to identify early indications of
sources of inflationary pressure
• Not used to forecast inflation
• Effectiveness decreases as forecast
period is prolonged
• Currently 19 univariate models - monitor
the monetary sector, foreign sector,
labour market and domestic demand
Macro Models Unit
20 August 200423
Disaggregated inflation model
• Model the components of CPIX
independently
• Uses monthly data
• Possible to identify sources of inflation
• Components are modelled mainly as a
function of unit labour costs, import
prices and the output gap
Macro Models Unit
20 August 200424
Forecasting process in the Bank
• Intensive process spanning 6 weeks, 6 times a year : from information collection and analysis to final discussion at the MPC meeting
• Collection and analysis of all relevant information on international and domestic economy
• Initial assumptions by technical staff
• Comments and suggestions by senior staff
• Final assumptions and specification of alternative scenarios by members of the MPC
• Preparation of forecasts with all the models and cross checking
• Preparation of final MPC document
• Discuss forecasts, risks and uncertainties at MPC meeting
• Publish the inflation forecast in the MPR twice a year
Macro Models Unit
20 August 200425
New dataNew dataHighFrequency dataHighFrequency data SARB expertsSARB experts
OECDWorld Bank
IMFOther SA
OECDWorld Bank
IMFOther SA
Data updateData update
Modelling TeamModelling Team
Final forecastlocked in and
alternative scenarios
Final forecastlocked in and
alternative scenarios
Formulation ofassumptions
Formulation ofassumptions
Add-factors updateAdd-factors update
Forecast reviewsForecast reviewsSARB expertsSenior staff
MPC
SARB expertsSenior staff
MPC
MPCMPC
Schematic illustration of the forecasting process in the Bank
Macro Models Unit
20 August 200426
Role of forecasts
“ In an ever-changing economy, no single model can possibly assimilate in a comprehensive way all the factors that matter for policy. Forming judgements about those factors, and their implications for policy, is the job of the Committee, not something that can be abdicated to models or even modellers. But economic models are
indispensable tools in that process.” - Bank of England
Macro Models Unit
20 August 200427
SARB models and forecasts only one consideration in policy formulation
Core model
Forecast
Policy
Assumptionsand judgements
Other models
Other issues andpolicy judgements
Macro Models Unit
20 August 200429
Fan chart
• Many central banks use a fan chart in presenting the forecast for inflation
• A point forecast does not have much chance of matching actual outcome
• Used to convey a more accurate representation of the assessment of medium-term inflationary pressures
• Helps to focus the discussion on uncertainty & risks around the forecast
• Fan indicates probabilities, NOT upper & lower bounds for inflation
Macro Models Unit
20 August 200430
Requirements for fan chart
• The following inputs are required to construct the fan chart:– the outlook for inflation - based on the
models– the degree of uncertainty - determines the
width of the bands– the degree of skewness - upside or
downside risk
• Requires mode and median of
distribution
Macro Models Unit
20 August 200434
“The tail wags the dog. By gently touching a tiny tail, Alan Greenspan wags the mammoth dog, the great American economy. Isn’t that remarkable? The federal funds rate is the shortest of all interest rates, remote from the rates on assets and debts by which businesses and households finance real investment and consumption expenditures counted in GDP. Why does monetary policy work? How? It’s a mystery, fully understood by neither central bankers nor economists.” James Tobin
The mystery of the MPTM
Macro Models Unit
20 August 200435
The transmission mechanism
• Describes how monetary policy affects output and inflation
• When the official interest rate is changed: • By how much is inflation affected, and when? • Is output affected in the short-run?• Through which channels does this happen?
• “To be successful in conducting monetary policy, the monetary authorities must have an accurate assessment of the timing and effect of their policies on the economy, thus requiring an understanding of the mechanisms through which monetary policy effects the economy” (Mishkin, 1995).
Macro Models Unit
20 August 200436
Difficult to estimate
• The lags with which monetary policy acts are long and variable (Friedman)
• Our knowledge of the transmission mechanism is imperfect.
• Estimating the transmission mechanism is difficult. (Mahadeva and Sinclair, 2001)– Lack of reliable data– Unforeseeable changes in the structure of
the economy– Separating cause from effect in the data
Macro Models Unit
20 August 200437
E
DEPTH AND BREADTH OF FINANCIAL MARKETS LOW HIGH
Impo
rtan
ce o
f in
tere
st r
ate
in tr
ansm
issi
on
mec
hani
smHIGH
Macro Models Unit
20 August 200438
Officialrate
Market rates
Asset prices
Exchange rate
Domestic demand
Net externaldemand
Totaldemand
Domesticinflationarypressure
Importedinflation
InflationExpectations/confidence
The transmission mechanism
Bank of England
Core model consistent with this view of the transmission mechanism
Macro Models Unit
20 August 200439
SARB diagram
• The SARB view is similar to other central banks’ view of the MPTM
• The starting point is again the official interest rate changes and then illustrate in more detail the key influences and channels running to the rest of the economy and ultimately to the rate of inflation.
• Can create wrongly the impression that one can isolate the impact through these channels
• Both diagrams essentially say the same thing• Main impact between 12 to 24 months, in line with
international experience
Macro Models Unit
20 August 200440
rates
Inflation rate
of goods and
&
of goods and
Net exports
Relativeprices
Nominalexchange
Import prices
Wages
Inflation
Other asset prices
Expenditure&
Investment
Net wealth
of goods and
Money
&
Demand & supplyof goods and
services
Repurchase rate
Other interest
rates
Inflation rate
Demand & supplyof goods and
services
Expenditure&
Investment
Demand & supplyof goods and
services
Net exports
Relativeprices
Nominalexchangerate
Import prices
Wages
Inflationexpectations
Other asset prices
(equity, land, property)
Expenditure&
Investment
Net wealth
Demand & supply of goods and
services
Money and credit
Expenditure&
Investment
Demand & supplyof goods and
services
Monetary Policy Transmission Mechanism
Current
Account
Capitalmovements
Change in
reserves
Wages
Inflation
expectations
Macro Models Unit
20 August 200441
SARB practice
• Baseline forecast with fixed repo rate• Alternative scenarios
– with “market expectations” repo rate– with Taylor-type policy reaction function– with alternative specifications by MPC– with alternative assumptions on ‘risk’
variables
Macro Models Unit
20 August 200443
Finally
• SARB follows inflation targeting framework with a Monetary Policy Committee setting official interest rates
• Model estimation procedures in line with modern approach to modelling
• Model development process in line with international practice• Core model reflects the monetary policy transmission
mechanism• Coefficients and impulse response functions are broadly in line
with research findings of SARB and others• Forecasting process covers the various conventions of setting
policy rates in models and the use of alternative scenarios• Forecast is a combination of staff and MPC inputs• Inflation forecast is published, using the fan chart to indicate
uncertainty inherent in forecasting