Download Main objective of the research - Jedenaste Warsztaty Doktorskie

How Forward-looking are Central Banks?
Some Evidence from their Forecasts*
Michał Brzoza-Brzezina, Jacek Kotłowski, Agata Miśkowiec
Summary prepared for „Warsztaty doktorskie z zakresu ekonometrii i statystyki”
* This is a preliminary summary of the paper. Please do not recirculate.
April 30, 2011
Key words: Taylor rule, forward-looking monetary policy, feedback horizon
JEL classification: C25, E52, E58
Main objective of the research
The main aim of this paper is to determine whether central banks adopt forward-looking approach in their
conduct of monetary policy, and if so - what length of horizon they take into account.
Introduction and related literature
According to both central banks and academics, efficient monetary policy should be forward-looking. Such
monetary policy is able to better stabilize output and inflation fluctuations (see inter alia Rudebusch (1998),
Bernanke et al. (1999), Orphanides (2001), Clarida et al. (2001), Mishkin (2007)). This observation has even
led to suggestions that central banks should explicitly adjust interest rates to deviations of forecasted output
and inflation from targets (Svensson (1997)).
Central banks have claimed for many years that they conduct monetary policy in a forward-looking way. In
order to convince the public that this is the case they publish inflation reports, minutes of decision-makers
Strona 1 z 7
meetings and, in particular, macroeconomic projections (including inflation, GDP and sometimes even
interest rate path). Especially publishing inflation forecasts - which often are de facto intermediate targets for
monetary policy - has become a common practice (Fracasso et al. 2003). Mishkin and Schmidt-Hebbel
(2001) listed central banks that failed to fulfill that unwritten rule. Today, each bank from their list publishes
an inflation forecast in order to increase transparency and be in line with the best practices.
Things become less clear when it comes to decide how forward-looking central banks should be and how far
they look indeed into the future.
As to the first question, regarding the optimal targeting horizon, the literature is vast. Several theoretical
studies have been conducted, analyzing the performance of monetary policy rules with various horizons.
Batini and Haldane (1999) estimated the optimal forecast horizon (according to their definition - one that is
securing as good inflation as any other, while at the same time delivering significantly lower output
variability) to be between three and six quarters ahead. They also noticed that the greater the degree of
forward-lookingness by the private sector, the less the compensating need for forward-looking behaviour on
the part of the central bank. At the same time however, Batini and Pearlman (2002) illustrated analytically
using a New Keynesian closed economy model that long inflation feedback horizons (with or without
additional output gap feedbacks) can lead to indeterminacy. The same was later proved for open economy
models (Batini et al. (2004)). However, in spite of the numerous studies in the subject, no consensus over the
question how forward-looking central banks should be has been reached. Commenting on the literature on
optimal policy horizons, it may be worth citing Goodhart (2001), who suggested that the selection of
monetary policy horizons is so model specific that little advance can be made unless such studies apply the
context of monetary policy as found in practice.
In line with this argumentation, several authors tried to determine optimal targeting horizons for specific
economies. For instance, Batini and Nelson (2000) found that for the Bank of England the best feedback
horizon to focus can range from 2 quarters if agents other than the central bank are assumed to be forwardlooking, up to 15 quarters in models with no forward-looking behaviour in the economy. In case of the Czech
Republic, Strasky (2005) observed a trade-off between the inflation and output variability that accompanied
the policy horizon increase. At longer horizons the inflation variability augmented (it was the smallest for
two quarters horizon), while the output gap variability was gradually reduced (reached minimum for nine
quarters horizon). The simulation results suggested the Czech National Bank's optimal targeting horizon was
three quarters.
As to the second question, i.e. how far in practice central banks look into the future while setting interest
rates, there is substantially less evidence. Central banks themselves usually avoid pointing at particular dates
at which they target inflation or output. They prefer to provide this information indirectly, e.g. stating that
they intend to bring inflation to target in the medium term. The medium-term orientation of the European
Strona 2 z 7
Central Bank's monetary policy was emphasized at the start of Stage Three of European Monetary Union. As
explained by Trichet (2003), the medium-term orientation clarifies that there is no fixed time horizon over
which price stability has to be re-established. It gives ECB the possibility to vary the appropriate monetary
policy horizon and adapt it to the nature of a shock the bank responds to. Moreover, this approach is believed
by Trichet to be superior over central banks that adopted a fixed horizon, as it gives more flexibility and
limits scope for short-sighted reactions.
However, some central banks commit (in an explicit or implicit way) to specific targeting horizons. For
example, the Reserve Bank of New Zealand (after serious problems with hitting the inflation target in 1998
and 1999) has modified its inflation-targeting regime to lengthen the horizon over which it tries to achieve its
inflation target to 6-8 quarters (Drew and Orr (1999)). The Bank of England has the policy horizon of two
years, which means that the Monetary Policy Committee tends to put weight on the inflation projection
around two years ahead (King (1997)).1
Even less is said about the horizon at which monetary authorities care about output fluctuations. As to the
academic literature, it is often limited to the statement that in case of a nonzero weight put on output
fluctuations, a longer policy horizon should be implemented in order to avoid major output fluctuations
(Mishkin (2002)).
In empirical papers gauging optimal policy horizons, focus tends to be on inflation forecasts horizons, while
output values are limited to current ones (like in the framework presented by Rudebusch and Svensson
(1998)). In particular, to our best knowledge, no paper has tried to exploit the evidence implicit in central
banks' own macroeconomic forecasts. This study tries to fill this gap.
Data and model
We collect data on macroeconomic projections of three important central banks - the Bank of England (BoE),
Schweizerische Nationalbank (Swiss National Bank, SNB) and Sveriges Riksbank (SR) and use it to recover
the horizon at which these banks look at future inflation and output. The choice of central banks is
determined by the availability of a sufficient number of projections based on the assumption of constant
interest rates during the projection horizon (conditional forecasts). These seem best suited to conduct our
study, since they show explicitly the consequences of not changing the interest rate. If following such a
forecast the central bank changes the interest rate it is relatively easy, from the econometric point of view, to
1
However, some economists argue, that in current circumstances there might be a case for considering a longer-term
policy horizon. One reason for this would be that quantitative easing may have longer lags before it reaches maximum
effect than do changes in bank rate (and part of the reason for focusing around two years ahead is that it is believed to
be the point of maximum policy effect). It might also be useful to reconsider the merits of looking to inflation prospects
beyond the normal forecast horizon, to ensure any future risks to economic stability are taken fully into account (see
Barker (2010)).
Strona 3 z 7
figure out which forecast horizon triggered this decision.
Following Orphanides critics (2001), only real time data is employed. Model uses information available at
the time the forecasts are made, moreover - only from central banks' own projections. Yet - as opposed to the
BoE or SR - the Swiss National Bank does not publish its own projection of GDP. Therefore, VAR model
with three variables (GDP growth, inflation and unemployment rate) was used to estimate the projected value
of the output gap.
For the estimation we used an ordered probit model. This econometric approach has been widely used in the
literature to estimate monetary policy rules. Even though it was employed in a model describing Fed's
discrete interest rate policy in the 50s of the twentieth century (see Rosett (1959)), it became commonly
employed not until a paper by Eichengreen et al. (1985). Currently, it is often used in papers which address
the Taylor's reaction function. The method is based on the fact that the movements of the central bank
reference interest rates take the form of a multiple of some value (typically 0.25 percentage point). As a
result, it is the sign of interest rate change, not its magnitude, that is examined. In our framework the
dependent variable in the Taylor reaction function - change of the bank rate - can take values (-1, 0, 1).
We assume that the central banks' decisions about interest rates depend on the variables presented in the form
of an anticipatory Taylor function presented below:
∆BRt = f (E(πt+k |Ωt ) − πt ) + g(E(xt+l |Ωt )),
where ∆BRt - central bank reference rate change; πt+k - inflation in time t + k (in %); πt - inflation target in
time t (in %); xt+l output gap in time t + l (i.e. deviation of GDP growth from the potential growth rate, in %);
k and l - forecast horizons; Ω - information available in time t; f, g – Gaussian functions.
One important technical issue that must be discussed here, is the way we include future explanatory variables
into the model. Central bank forecasts extend over several quarters, which means that in practice we have to
deal with a large amount of explanatory variables: two for each forecast horizon (one for inflation and one
for output). Given the limited data set and the potential high correlation of forecasts at different horizons, the
approach of treating inflation and output at each horizon as a separate explanatory variable seemed
unrealistic.
We decided to assume a specific functional form that would combine forecasts at all horizons into one value
for inflation and one for output and estimate the parameters of this function. Our choice of functional form
was a Gaussian function.
In our view this function is well suited to describe realistic ways of aggregating forecasts at various horizons.
By varying the mean parameter we can control the forecast horizon with the highest weight in the central
Strona 4 z 7
banks rule. By varying the standard deviation we can control the weight of the peak horizon relative to the
other horizons. Additionally, this function leaves only two parameters to estimate, which, given our limited
amount of data, is a big advantage.
Results
Our findings are the following. First, all analysed banks are forward looking with respect to inflation. The
BoE and SNB look at the last quarter of their forecasts (respectively 8th and 12th) while the SR takes into
account the middle point of the forecast (4th quarter).
The case of output is more nuanced. Results for Riksbank do not give an unequivocal response as to the
forecast horizon for output employed in the process of interest rate setting. In case of the Bank of England
the forward-looking approach as to output gap is undisputed and the lead amounts to 2 quarters.
Nevertheless, for these analyzed central banks the lead of inflation forecasts is consistently bigger than the
lead of output.
Finally, to our surprise we find little evidence that monetary authorities take into account (albeit with varying
weights) their whole forecasts. In contrary, the evidence speaks mostly for concentrating strongly on one
particular horizon, especially when output is concerned.
The paper will be organised as follows. In section two, after the introduction and related literature, the data
and the econometric approach will be described. In section three the results followed by some critical
remarks will be presented. Conclusions will be given in section four.
Strona 5 z 7
References
Barker, K. (2010): “Monetary Policy - from stability to financial crisis and back?,” speech given at the
National Institute of Economic and Social Research, London, 8 March 2010
Batini, N. and A. Haldane (1999): “Forward-looking rules for monetary policy,” NBER Chapters in:
Monetary policy Rules, pp. 157–202
Batini, N., P. Levine, and J. Pearlman (2004): “Indeterminacy with inflation-forecast-based rules in a twobloc model,” International Finance Discussion Papers, 797
Batini, N., and E. Nelson (2000): “Optimal horizons for inflation targeting,” Bank of England, Working
paper
Batini, N., and J. Pearlman (2002): “Too much too soon: Instability and indeterminacy with forward-looking
rules,” External MPC Unit Discussion Paper, Bank of England, 8
Bernanke, B., T. Labauch, and F. Mishkin (1999): Inflation targeting: lessons from the international
experience. Princeton University Press, Princeton
Clarida, R., J. Gali, and M. Gertler (2001): “Optimal Monetary Policy in Open versus Closed Economies: An
Integrated Approach,” American Economic Review, 91(2), 248–252
Drew, A., and A. Orr (1999): “The Reserve Bank’s Role in the recent business cycle: actions and evolution,”
Reserve Bank of New Zealand Bulletin, 62, No.1
Eichengreen, B., M. Watson, and R. Grossman (1985): “Bank rate policy under the interwar gold standard: A
dynamic probit model,” Economic Journal, 95, 752–745
Fracasso, A., H. Genberg, and C. Wyplosz (2003): How do Central Banks Write? An Evaluation of Inflation
Reports by Inflation Targeting Central Banks. Geneva Reports on the World Economy Special Report 2
Goodhart, C. A. E. (2001): “Monetary transmission lags and the formulation of the policy decision on
interest rates,” Federal Reserve Bank of St Louis Review, (July/August), 165–181
King, M. (1997): “The inflation target five years on,” Bank of England Quarterly Bulletin, 37, 434–42
Strona 6 z 7
Mishkin, F. (2002): “Overview: Rethinking stabilization policy,” Proceedings, Federal Reserve Bank of
Kansas City, pp. 439–446
–
(2007): Monetary policy strategy. The MIT Press, 1st edition
Mishkin, F., and K. Schmidt-Hebbel (2001): “One decade of inflation targeting in the world: What do we
know and what do we need to know?,” Central Bank of Chile, Working Papers, 101
Orphanides, A. (2001): “Monetary Policy Rules Based on Real-Time Data,” American Economic Review,
91, 964–985
Rosett, R. (1959): “A statistical model of friction in economics”, Econometrica, 26, 263–267
Rudebusch, G., and L. Svensson (1998): “Policy rules for inflation targeting,” Seminar Papers 637,
Stockholm University, Institute for International Economic Studies
Strasky, J. (2005): “Optimal forward-looking policy rules in the quarterly projection model of the Czech
National Bank,” Research and Policy Notes, Czech National Bank, 5
Svensson, L. E. (1997): “Inflation forecast targeting: Implementation and monitoring inflation targets,”
European Economic Review, 41, 111–146
Strona 7 z 7