|
BIS/CEPR
Conference On 26/27
January 1998, the Bank for International Settlements (BIS) and CEPR
organized a conference on ‘Asset Prices and Monetary Policy’. The
conference, which was hosted by the BIS in Basel, was organized by Renato
Filosa (BIS) and Francesco
Giavazzi (IGIER, Università Bocconi, Milano, and CEPR). Six papers
were presented and the conference ended with a panel discussion on some
of the key issues. In his paper on ‘Open-economy Inflation Targeting’, Lars Svensson (Institute for International Studies, Stockholm, and CEPR) extended the analysis of inflation targeting to a small open economy with forward-looking aggregate supply and demand. This allowed Svensson to discuss the role of the various exchange-rate channels in inflation targeting, the use of a monetary conditions index (MCI), the choice between CPI and domestic inflation targeting, and the appropriateness of simple instrument rules such as the closed-economy Taylor rule. Svensson found that strict CPI inflation targeting resulted in considerable real exchange-rate variability, owing to heavy reliance upon the direct exchange-rate channel. Flexible CPI targeting appeared as an attractive alternative when the policy objective was to stabilize both inflation and real variables. As the Taylor rule excluded any direct response to the exchange rate or foreign disturbances, it was less attractive in this open economy set-up. Finally, Svensson found some support for the notion of an MCI, although the index derived in his model differed from the commonly used concept. Charles Freedman (Bank of Canada) observed that separate analysis of
the effects of introducing forward-looking behaviour and an
exchange-rate channel on the appropriateness of various forms of
inflation targeting would be useful. On the question of which price
index should be targeted, he noted that central banks also had good
practical reasons for using the CPI (rather than, for example, the GDP
deflator). Freedman considered that inclusion of the long real rate in
the aggregate demand function was an empirical issue. For example, in
Canada and the United Kingdom, empirical research had shown that it was
mostly the short-term rate that was important for spending decisions.
Finally, Freedman thought that the way inflation targeting was
implemented in practice – whereby the policy rate responds gradually
to an inflation forecast six to eight quarters out – was very much
akin to what Svensson calls flexible CPI targeting. Jürgen von Hagen (ZEI, Universität Bonn, and CEPR) raised three
issues. First, he questioned the statement that inflation targeting was
characterized by a high degree of transparency. In particular, monetary
targeting as used by the Bundesbank was more precise about the inflation
objective (2%), but was less precise about the time horizon. Most
inflation-targeting regimes, on the other hand, were specific about the
horizon but allowed for a wider inflation band. He also questioned
whether inflation reports made the inflation-targeting regimes very
transparent, as these reports did not specify the weights attached to
the different pieces of information that had a bearing on inflation.
Second, von Hagen wondered what the micro foundations were for the
shocks in the model. He asked why changes in the exchange rate had
immediate effects, whereas changes in domestic costs had only lagged
effects. Third, he argued that a Taylor rule with domestic inflation did
almost as well as flexible domestic inflation targeting. Carlo Favero (IGIER, Università Bocconi, Milano, and CEPR) observed that in the VAR literature there was not much evidence in favour of a direct exchange-rate effect on inflation. Creon Butler (Bank of England) asked about the impact of the noisiness of the exchange rate on its role in CPI inflation targeting. Mark Gertler (New York University) thought the framework of the paper was not suitable for dealing with the concept of transparency. Moreover, if transparency was related to the concept of credibility, then equilibria with commitment should also be looked at. In the same vein, Paul Söderlind (Stockholms Universitet and CEPR) observed that it was hard to compare the performance of a Taylor rule, which assumes commitment, with the discretionary rules considered in the paper. Svensson agreed that accountability and transparency were difficult to discuss within his model as there was full transparency. Replying to von Hagen, he added that posting the inflation forecast as a summary statistic of the relevant information concerning future inflation was an effective way of increasing transparency as it was not so easy to doctor a forecast. Michael Woodford (Princeton University) replied to von Hagen that a money-demand equation could easily be added to the model without altering the equilibrium. There was a nominal anchor as long as the reaction function included a nominal variable. Ignazio Visco (OECD, Paris) wondered how nominal income targeting would compare with flexible CPI targeting, and Alexander Swoboda (Graduate Institute of International Studies, Geneva) questioned how robust the results were with respect to differences in the loss function and model parameters. In his paper
with Julio Rotemberg (Harvard
University) on ‘Interest-rate Rules in an Estimated Sticky Price
Model’, Michael Woodford (Princeton University) used a small model of the US
economy to discuss three topics. First, what must be the constraints on
policy in order for a rational expectations equilibrium to be
determinate? Second, how would different rules affect the variability of
inflation, output and interest rates? Third, how would these rules
perform in terms of expected utility? The paper’s main finding was
that low and stable inflation, together with stable interest rates,
could be achieved by letting the funds rate respond positively to
inflation, while also responding – with a coefficient greater than one
– to the lagged funds rate itself. Such a rule did almost as well as
the optimal rule, in the sense of maximizing the expected utility of the
representative household. Hans Genberg (Graduate Institute of International Studies, Geneva)
compared the Woodford-Rotemberg approach with the traditional Poole
analysis, in which the interest-rate reaction function would depend on
the weights in the loss function and the relative variance of the
shocks. The advantage of Woodford and Rotemberg’s analysis was that
their model was less sensitive to the Lucas critique, although this
often came at the cost of oversimplification. Turning to the calibration
of the model, Genberg wondered whether the results were sensitive to
alternative VAR specifications. He was also worried that the optimal
rules described were dangerously close to the unstable regions. He
suggested looking at alternative theoretical structures, such as RBC
models and open-economy extensions. Marvin Goodfriend (Federal Reserve Bank of Richmond) observed that,
by advocating a policy rule whereby the central bank promised
ever-greater interest rate increases in response to deviations of
inflation from target, Rotemberg and Woodford leaned a little too
heavily in the direction of perfect credibility. The idea of a delayed
monetary-policy response contrasted sharply with the recent emphasis on
pre-emptive policy. On a theoretical level, the promise of ever-greater
interest rate increases was not observable, and Goodfriend doubted
whether it was realistic to assume that such out-of-equilibrium
possibilities were able to constrain the equilibrium. Lars Svensson
pointed to a difference in modelling strategy between his and
Woodford’s paper. Woodford had good micro foundations, but no inertia.
To fit the data, the inertia was built into the shocks, whereas he (Svensson)
had obtained inertia through partial-adjustment mechanisms. Might this
explain why there was more scope for pre-emptive monetary policy in his
model? And were rules like the one proposed incentive-compatible? Arturo
Estrella (Federal Reserve Bank of New York) asked whether, given the
wide confidence bands in the VARs, Woodford had done any sensitivity
analysis with respect to this uncertainty. Charles Freedman said optimal
rules, like the one suggested by Woodford, would always depend on the
structure of the model and the nature of the historical shocks. In this
sense, a forward-looking rule was more robust as it could more easily
take such changes into account. In reply, Woodford acknowledged the need
to consider incomplete credibility but did not agree that the policy
rule was unobservable. The public observed variations in the Fed funds
rate and could estimate the feedback rule. Parameter uncertainty was
important, but the authors had experimented with different parameters
which had not thrown the model into the unstable regime. As a general
point, Woodford said that, in terms of instability, forward-looking
rules were much more sensitive to specific parameters than
backward-looking rules. In ‘Why do
Interest Rates Predict Macro Outcomes?’, Arturo
Estrella (Federal Reserve Bank of New York) used a small theoretical
rational-expectations model to analyse the role of monetary policy in
determining the empirical relationships between, on the one hand, the
term structure of real and nominal interest rates and, on the other
hand, future real output and inflation. Estrella found that a policy
reaction function, whereby the central bank’s policy rate responded
gradually to output and inflation, could explain the empirical
regularities. These empirical relationships were not structural,
however, and alternative monetary-policy regimes could lead to very
different outcomes. For example, under a regime of strict inflation
targeting, interest rates became uninformative with regard to future
output and inflation. Peter Kugler (Universität Basel) focused on the model’s stability
under various reaction functions, and particularly the need for a strong
policy response to output. There was potential for unstable dynamics in
the output equation, which could be avoided only if real interest rates
rose strongly to stabilize output. Kugler conjectured that one reason
for this strong result might be that inflation expectations did not
enter the aggregate supply curve; introducing them might have a
stabilizing impact. Axel Weber (Universität Bonn and CEPR) was critical of Estrella’s
model in four respects: the microfoundations of the reduced-form
equations should have been worked out to provide a theoretical
foundation for the inertia; in view of the arbitrary lag structure, it
would have been useful to fit the model to the data; the assumption that
bond-market participants knew the reaction function of the central bank
was questionable; and the implications of non-linearities in the
Phillips curve had not been addressed. For Lars Svensson, Estrella’s general conclusion that the empirical relationships between the slope of the yield curve and future macro variables were not structural but endogenous had important policy implications. A regime of strict inflation targeting, under which the yield curve gap had no predictive power for output, was an extreme interpretation of actual inflation targeting regimes. These regimes could be better described by flexible inflation targeting, whereby output stabilization also entered the central bank’s objective function. Such a regime would probably look similar to the one Estrella had found to be compatible with the empirical regularities. Ignazio Angeloni (Banca d’Italia) asked whether the author had looked into the consequences of allowing for risk aversion. The behaviour of a risk premium might be important in explaining some of the stylized facts. Ignazio Visco wondered about open-economy extensions. Several participants suggested that the paper’s results be tested empirically by comparing the predictive content of the term structure across different policy regimes. Michael Woodford wondered why the results of his and Estrella’s stability analyses had differed. Estrella
replied that his paper was intended to show that the predictive content
of the slope of the term structure depended on the monetary-policy
regime. Although it would be more elegant to do this in a fully
specified model with microfoundations and model-consistent expectations,
this main result would not be much affected. Similarly, allowing for a
risk premium in the term structure would enrich the model, but was not
necessary to explain the relationship between the term spread and the
first moments of the economy. He agreed, however, that an obvious next
step was to test the predictions of the analysis across different
regimes. Creon Butler (Bank of England) presented ‘Assessing Market Views
on Monetary Policy: The Use of Implied Risk Neutral Probability
Distributions’, which was written with Haydn Davies. According to
Butler, policy-makers are interested in assessing financial-market
expectations of future policy rates for two reasons. First, it may
provide information as to whether the markets expect a planned policy
move or not. Second, it can provide a check on the central bank’s
credibility. In his paper, Butler presented a method for estimating time
series of implied risk-neutral probabilities, based on prices of
exchange-traded options and futures contracts. He then examined whether
the estimated higher moments from these probability distributions could
improve the predictive power of forward rates. For short-term sterling
interest rates, the results suggested that the mode of the risk-neutral
density function was a better predictor of future short-term interest
rates than the forward rate, and that adding in the standard deviation
and kurtosis improved the fit further. A likely interpretation of the
latter results was that the standard deviation and kurtosis captured a
variable risk premium. Butler also illustrated the potential use of the
adjusted forward rate for policy purposes. Pierre Sicsic (Banque de France and CEPR) offered four observations:
(1) most of the results in the paper could be shown by using implied
volatilities directly in the prediction regressions; (2) the use of
linear interpolation to filter out the time-to-maturity effect was
debatable, since – Sicsic conjectured – this effect should not be
linear; (3) the finding of stationarity depended on the inclusion of a
trend, but no trend appeared to be included in the prediction
regressions; and (4) there was a serious errors-in-variables problem, as
the Melick-Thomas method used in the paper often led to big shifts in
parameters that were hard to interpret. Paul Söderlind thought it might
be easier to interpret the results if the analysis was performed within
an appropriate asset-pricing theory in which the risk premium could be
modelled. Very similar results could be obtained by adding the level of
interest rates into the usual prediction regression. He also pointed to
a small-sample problem: the overlapping structure of the data series
meant there were really only 13 independent observations. Robert McCauley (BIS) suggested that there might be institutional
reasons why the difference between the forward and realized short rates
was related to volatility in the United Kingdom, but not in Germany. One
reason was that, in the United Kingdom, more floating-rate notes were
issued and thus the demand for hedging against interest-rate risk was
greater. Alternatively, there could be differences in the respective
central banks’ operating procedures. Ignazio Angeloni wondered whether
the skewness effect in Germany could be related to the effect of US
monetary policy in 1994 and to the question about decoupling. Henri
Pagès (BIS) wondered why the authors had not made use of prices of
options on swaps. Butler’s response was that work at the Bank of
England had started to look at asset-pricing models (such as the
consumption CAPM model) to explicitly model risk premia. The main
problem with using options on swaps was how to deal with credit risk.
Butler repeated that substituting the mode for the forward rate did a
lot of the work, thereby suggesting that the ‘peso problem’ may be
an important issue. But more work clearly needed to be done to make this
information useful in policy-making. Mark Gertler (New York University) presented ‘Monetary Policy
Rules in Practice: Some International Evidence’, which was co-authored
by Richard Clarida and Jordi Galí. Their paper reported estimates of
monetary-policy reaction functions for two sets of countries: the G-3
(Germany, Japan and the United States) and the E-3 (France, Italy and
the United Kingdom). The paper found that, since 1979, the behaviour of
each of the G-3 central banks could be described by a forward-looking
version of the Taylor rule with interest-rate smoothing. As for the E-3,
even prior to the emergence of the ‘hard ERM’, their central banks
were heavily influenced by German monetary policy. Using the
Bundesbank’s policy rule as a benchmark, the authors found that, at
the time of the collapse, interest rates in each of the E-3 countries
were much higher than domestic macroeconomic conditions warranted. Otmar Issing (Deutsche Bundesbank) argued that the use of a rule
facilitated communication with the public, and had the potential for
anchoring expectations. The policy rule estimated in the paper appeared
to be a nice ex post
description of the G-3 central banks’ interest-rate policy, but what
was important was the ex ante
perception that a rule existed. What form the rule should take depended
on the structural features of the economy. The Bundesbank had pursued
monetary targeting for 24 years for a number of reasons: the long lags
and uncertainties in the transmission mechanism had made the medium-term
focus afforded by a money-growth target appropriate; monetary aggregates
had clear advantages in terms of transparency and accountability; there
was quite a lot of evidence on the stability of the demand for German
M3; and the trend growth of money had clear predictive power for future
inflation. A clear indication of the Bundesbank’s resolve to pursue a
policy oriented towards price stability – as soon as it was released
from its task of pursuing exchange-rate stability in 1973 – was that
the special Lombard rate had risen to 13% in 1974. Issing agreed that
one of the secrets of successful central banking was to be
forward-looking, i.e. focusing on the medium-term trend of inflation. In
the paper, however, this had been translated into a one-year-ahead
inflation forecast. Given that the transmission lag from policy to
inflation was at least two years (and probably longer), he found the
evidence not completely convincing in this respect. Issing further
thought that the fact that monetary variables were found insignificant
in the estimated reaction function did not imply that monetary targeting
was not followed – the targets were not pursued for their own sake,
but because of their relationship with future inflation. Exactly for
that reason, the estimation methodology did not allow for discrimination
between inflation targeting and monetary-growth targeting. Ignazio
Visco applauded the ingenious estimation methodology, which consisted of
using actual future inflation as a measure of expected inflation and
using instrumental variables to estimate the feedback parameters. He had
several queries, however. First, a little more sensitivity analysis –
for example, with respect to the specification of the output gap –
would have been useful. Second, in some cases, the appropriate policy
rate would be the repo rate, rather than the overnight rate, and he
wondered why the speed of adjustment towards the target rate was so
slow. Third, he questioned the role played by the output gap and asked
whether some form of lexicographic ordering could explain the results.
Fourth, he asked whether the estimation results provided evidence that
monetary policy had brought down inflation. Finally, he wondered why the
implicit inflation target for the United States was so high (4%). During
discussion, Pierre Sicsic wondered how to explain the fall in the stress
indicator for France in 1993 (after a strong rise in 1992). Axel Weber
suggested using a threshold model, instead of the partial-adjustment
model, to capture the adjustment of the policy rates to their target.
Michael Woodford questioned the assumption underlying the estimation
methodology, namely that the correlations between available information
and future inflation were not only understood but also constant over the
sample period. The robustness of the results should be tested by using
actual forecasts or rolling regressions. Stefan
Gerlach (BIS and CEPR) raised doubts about the stationarity of the
interest rates and wondered about the impact on the estimation
methodology. Also, would the results concerning the effect of M3 growth
still hold if some low-frequency filter were used to get rid of the
short-term volatility in the M3 series? Ignazio Angeloni thought
inflation differentials should enter the specification for the E-3
countries. William White
(BIS) asked whether it was reasonable to use the Bundesbank’s policy
rule as a benchmark for the United Kingdom, where interest rates had a
more rapid effect. Finally, Charles Freedman noted that the estimated
short-run response to inflation and the output gap was not very strong. ‘A Red
Letter Day?’, presented by Carlo
Favero (IGIER, Università Bocconi, Milano, and CEPR) and
co-authored with Rudi Dornbusch and Francesco Giavazzi, concentrated on
the monetary-policy transmission mechanism in the European economies.
The authors found significant differences across countries in the
response to a common monetary-policy shock. In Germany, France and
Spain, the range of the impact was 1.4–1.5%; in Sweden and Italy, the
impact was much larger; and the effect in the United Kingdom was the
smallest. The authors discussed the sources of these differences and
their implications for a common European monetary policy. Luigi
Spaventa (Università degli Studi di Roma, ‘La Sapienza’, and
CEPR), referring to a BIS study on the transmission mechanism, noted
that in terms of the effects of monetary policy, the relative country
rankings in the paper did not correspond to the received wisdom. For
example, most indicators concerning financial structure would suggest
that monetary policy was most effective in the United Kingdom. José Viñals (Banco de España and CEPR) acknowledged the
authors’ fears that differences in national preferences or national
structures might compromise the European Central Bank’s policy focus
on price stability, and that regional differences could lead to adverse
popular reactions. Policy would only be affected, however, if national
preferences played a role in the Governing Board. Viñals also commented
on the paper’s estimated reaction functions, pointing to the low
estimated feedback parameters on inflation in many of the countries. He
argued further that country-specific factors might create cross-country
differences, and conjectured that a convergence of financial structures
was likely to contribute to a convergence of the transmission
mechanisms. Ignazio Visco
agreed with Spaventa that the estimated output effects were
counter-intuitive. The calculated MCI weight was also doubtful, and the
source of the differences in the relative importance of the
exchange-rate channel in the various European countries was unexplained.
Axel Weber pointed to the potential for both a generated regressor and a
simultaneity problem in the estimated output equations. Charles Freedman
thought that the focus on national differences was misplaced. In Canada,
there were still many regional differences, but these did not prevent
the Bank of Canada from pursuing a Canadian monetary policy. Mark
Gertler argued the need for care in estimating reaction functions over
very short samples. He suggested using IV techniques for resolving the
simultaneity problem in the output equations; and, pointing to the
presence of regional differences in the United States, he suggested that
these divergences might be attributable to many factors other than
monetary policy. Mark Gertler,
Marvin Goodfriend, Otmar Issing and Luigi Spaventa participated in the
concluding panel discussion. The panellists concentrated on two
important questions. First, should monetary policy focus solely on price
stability, or should it aim also at achieving asset-price stability in
order to reduce the risk of financial crisis? Second, do asset prices
provide reliable information about future economic conditions, such as
inflation and real growth, which could be used in the setting of
monetary policy? The views of the panellists have been published
separately in a joint CEPR/BIS publication, details of which are given
below. Asset
Prices and Monetary Policy: Four Views ISBN 1 898128 40 5 xii + 27. £10/$14.95/15 euros Please contact CEPR for details |