PRESS RELEASE

DATE 21 September 2007 SVERIGES RIKSBANK

SE-103 37 Stockholm
NO. 38 (Brunkebergstorg 11)

CONTACT Press Service, tel. +46-8-787 02 00 Tel +46 8 787 00 00

Fax +46 8 21 05 31
registratorn@riksbank.se
www.riksbank.se

What have economists learned about

monetary policy over the past 50 years?

Deputy Governor Lars E. O. Svensson gave a speech today at the conference

“Monetary Policy over Fifty Years” in Frankfurt am Main, on the occasion of the
50th anniversary of the Deutsche Bundesbank. In his speech Mr Svensson gave a
personal account of his view on the development of monetary policy research
over the past 50 years”.

“In economic research, 50 years is a long time. I will actually start in 1967, with
Milton Friedman’s presidential address at the meeting of the American Economic
Association, so I will only cover about 40 years. I will give a very personal view of
what economists have learned since Friedman’s address that is most relevant for
practical monetary policy. I will hence be selective, eclectic, and possibly contro-
versial,” began Mr Svensson.

Friedman’s “The role of monetary policy” – a classic

“The title of Friedman’s address was ‘The role of monetary policy.’ It remains a
classic in the development of a monetary-policy framework. He discussed what
monetary policy cannot do, what it can do, and how monetary policy should be
conducted. Regarding what monetary policy cannot do, he noted that it cannot
in the long run control real variables such as unemployment and GDP; it can only
control nominal variables, such as the exchange rates, the price level, or mone-
tary aggregates. These insights were not obvious at the time, but they are now
part of the conventional wisdom,” continued Mr Svensson.
“Regarding what monetary policy can do, Friedman emphasized three things:
First, monetary policy can avoid being a major source of disturbance; it should
avoid major mistakes. Second, monetary policy can provide a stable background
for the economy, preferably by achieving price stability. Finally, monetary policy
can contribute to offsetting major disturbances in the economic system arising
from other sources. On the last point, Friedman emphasized the danger of being
overly ambitious: ‘[T]he potentiality of monetary policy in offsetting other forces
making for instability is far more limited than is commonly believed. ... In this

1 [5]
PRESS RELEASE NO. 38

area particularly the best is likely to be the enemy of the good. Experience sug-
gests that the path of wisdom is to use monetary policy explicitly to offset other
disturbances only when they offer a clear and present danger’,” said Mr Svens-
son.

“Target money – too difficult to target the price level”

“Regarding how monetary policy should be conducted, Friedman stated two re-
quirements: The first requirement is that central banks should only target vari-
ables that they can control, such as the exchange rate, the price level, or a mone-
tary aggregate. Friedman considered an exchange-rate target unsuitable for the
U.S., since it implies adapting to the average of whatever policies monetary au-
thorities in the rest of the world adopt. He stated that targeting the price level
would in principle be best, but emphasized that control of the price level was too
imperfect, with policy actions having uncertain effects on the price level with long
and variable lags. Therefore, he recommended targeting a monetary aggregate,
since central banks have better control over money and lags are shorter. The sec-
ond requirement is that central banks should avoid sharp swings in policy. There-
fore he recommended central banks to achieve a steady but moderate rate of
growth of a specified monetary aggregate – which rate of growth and particular
aggregate is not so important as long as they are explicitly stated and adopted.
Friedman said: ‘That is the most that we can ask from monetary policy at our pre-
sent state of knowledge’,” continued Mr Svensson, emphasizing Friedman’s
qualification “at our present state of knowledge”.

Monetary targeting failed, but inflation targeting has worked fine

“Monetary targeting was tried in several countries during the 1970s and 1980s,
but it constantly failed and was therefore abandoned. In contrast, when inflation
targeting, the current best practice in monetary policy, was introduced in the
1990s in New Zealand, Canada, the U.K., Sweden, Finland, Australia, and later in
many other countries, it worked fine. The great exception to the failure of mone-
tary targeting appears to be the great performance in Germany by the Bundes-
bank, which kept inflation low and stable when inflation became high and vari-
able in other countries,” said Mr Svensson.
“However, closer scrutiny and many studies of the Bundesbank’s monetary policy
have revealed that the Bundesbank was actually a secret inflation targeter:
Whenever there was a conflict between achieving the money-growth target and
the inflation target (which was called ‘unavoidable inflation’, ‘price norm’, or
‘medium-term price assumption’), the Bundesbank consistently gave priority to
the inflation target and willingly missed its money-growth target. Thus, the
Bundesbank’s great and admirable performance actually demonstrates the suc-
cess of inflation targeting, not of monetary targeting – although the success of
rather nontransparent inflation targeting,” added Mr Svensson.
“Given the evidence, Friedman actually changed his view. Some years ago I had
the opportunity to ask him personally at a conference at the San Francisco Fed
whether the success of inflation targeting would make him revise his recommen-
dation to target money instead of targeting inflation and the price level directly.
He answered that he had revised his recommendation. In an interview with the
Financial Times in June 2003, Friedman also conceded that targeting money had
not been a success,” continued Mr Svensson.

2 [5]
PRESS RELEASE NO. 38

“Inflation targeting is in practice always flexible inflation targeting. That is, it

aims to stabilize not only inflation around an inflation target but also the real
economy. Furthermore, because inflation and resource utilization respond with
considerable lags to monetary-policy actions, it is necessary to rely on forecasts.
Flexible inflation targeting then boils down to what I have called ‘forecast target-
ing.’ That is, it consists of choosing and implementing an interest-rate path such
that the resulting forecasts of inflation and measures of resource utilization ‘look
good’. ‘Looking good’ means that inflation approaches the inflation target and
resource utilization approaches a normal level at a suitable pace,” continued Mr
Svensson.

Better knowledge about the transmission mechanism

“Why has inflation targeting worked so well? I believe central banks now have a
better knowledge about the transmission mechanism of monetary policy than 40
years ago. This allows them to produce usable forecasts of inflation and resource
utilization conditional on alternative interest-rate paths, the forecasts that are a
necessary requirement for forecast targeting,” said Mr Svensson.
“The conventional wisdom about the transmission mechanism with aggregate
demand and aggregate supply/Phillips curves is still relevant, but it has been
much refined with more transmission channels and, in particular, much better
understanding of the role of expectations. It is now generally acknowledged that
monetary policy works mainly through the private-sector expectations of future
interest rates and future inflation that central-bank actions and statements give
rise to. Those expectations matter much more than the current interest rate. That
is, monetary policy is ‘the management of expectations,’ as Michael Woodford
has expressed it. A few central banks, namely the Reserve Bank of New Zealand,
Norges Bank, the Riksbank, and Sedlabanki Islands, now choose and publish
what can be interpreted as an optimal interest-rate path or plan, as a more effec-
tive and transparent implementation of policy. This practice will gradually become
the norm, I believe. For instance, the Czech National Bank has announced that it
will start doing this from 2008,” said Mr Svensson.
“Theoretical research on the transmission mechanism has been paralleled by bet-
ter empirical methods, including Kalman filtering, vector auto regression, and
Bayesian estimation methods. The current research frontier of monetary policy
seems to be the development at several central banks of empirical Bayesian dy-
namic stochastic general equilibrium (DSGE) models for policy analysis, forecast-
ing, and simulation. The Riksbank already has such a model, RAMSES, in opera-
tional use,” continued Mr Svensson.

Monetary aggregates matter little for monetary policy

“We have also learned, I believe, that monetary aggregates matter little, or even
not at all, for monetary policy. Credit aggregates may matter, though. But eco-
nomic theory has not found any separate transmission channel from money to
prices. Empirically, a large amount of research has confirmed that there is little or
no information in money about future inflation beyond other explanatory vari-
ables, both for high- and low-frequency fluctuations,” said Mr Svensson.
“Furthermore, I believe that Friedman’s statement that ‘inflation is always and
everywhere a monetary phenomenon’ is often misunderstood. It refers to a cor-

3 [5]
PRESS RELEASE NO. 38

relation between endogenous variables, inflation and money growth, but it says
nothing about causality, that is, which variable determines the other. The direc-
tion of causality is instead determined by the monetary-policy regime. More pre-
cisely, in general equilibrium, endogenous variables are determined by exogenous
variables. Which variables are exogenous and endogenous is determined by the
monetary-policy regime. Strict monetary targeting would make money growth
effectively exogenous. Then inflation becomes endogenous and is determined by
money growth. Strict inflation targeting would make inflation effectively exoge-
nous. Then money growth becomes endogenous and is determined by inflation.
A fixed exchange rate makes both inflation and money growth endogenous vari-
ables. Then both variables are determined by other, exogenous domestic and for-
eign variables, including foreign monetary policy,” said Mr Svensson.

The importance of the institutional framework

“We have also learned the importance of the institutional framework for a stable
and successful monetary policy. The Bundesbank has been a starting point for
later institutional developments. Good and stable monetary policy is now seen as
resting on three pillars: (1) a mandate, with priority to price stability but also with
some weight on real stabilization, (2) independence, to avoid short-term political
interference and to give the central bank the possibility of achieving its mandate,
and (3) accountability, which improves with transparency, creates the incentive to
the central bank to achieve its mandate, and provides democratic control of a
powerful institution. The importance of credibility and transparency is also much
better understood these days, and a good institutional framework contributes to
the credibility of the monetary-policy regime,” continued Mr Svensson.

Taylor rules are robust but often overemphasized and misunderstood

“Taylor rules, where the interest rate is set mechanically as a simple linear func-
tion of current inflation and output, are frequently referred to in current mone-
tary-policy analysis. I believe that they are often overemphasized and misunder-
stood. In many research papers, monetary policy is modelled as if the central
bank is committed to follow a Taylor rule. Taylor rules are often treated as a
structural equation. But no central bank has made such a commitment, and infla-
tion-targeting central banks respond to much more information than current in-
flation and output. The empirical fit of Taylor rules is modest: I am not aware of
any estimation that has a higher R-square for interest-rate changes than about
2/3, meaning that 1/3 of the variance of interest-rate changes is explained by
other things than the Taylor rule. Taylor rules are actually empirical and simplified
reduced forms, not structural equations. They are not optimal, and they lack mi-
crofoundations,” said Mr Svensson.
“But Taylor rules are quite robust, in the sense that a Taylor rule almost never
works very badly in reasonable models. Why is that? The reason is that an opti-
mal instrument rule responds to all the determinants of the forecasts of the target
variables (inflation and resource utilization). Current inflation and output are im-
portant determinants of future inflation and output. Therefore, responding only
to current inflation and output is not that bad and actually quite a robust policy.
If the central bank knows little about the economy and the transmission mecha-
nism but can at least observe current inflation and output, it can do much worse
than following a Taylor rule. But today’s inflation-targeting central banks have

4 [5]
PRESS RELEASE NO. 38

much more information, know more about the transmission mechanism, and
therefore can do better than the Taylor rule. Consequently they respond to more
information and deviate quite a bit from the Taylor rule,” said Mr Svensson.

What do we not know?

“What do we not know, and where should we try to make progress in future re-
search? I believe that it is desirable to do flexible inflation targeting more explic-
itly. This requires more work on measures of resource stabilization and potential
output and raises difficult but important conceptual and empirical issues. I also
believe that flexible inflation targeting would benefit from more use of explicit
loss functions to evaluate alternative policy options and determine the optimal
one. Norges Bank seems ahead of other central banks in this regard. As every
maker of monetary policy knows, model uncertainty is always present and some-
times very large. We could certainly benefit from better techniques to incorporate
model uncertainty in forecasts. It would also be desirable to have better model-
ling of financial markets and the determination of yield curves and exchange
rates, the credit channel, and labor markets in the existing empirical DSGE models
of the transmission mechanism,” continued Mr Svensson.
“Regarding the ongoing discussion about the role of asset prices in monetary pol-
icy, I believe we know enough to state that asset prices should not be targets of
monetary policy. As long as their development is not a threat to financial stability
and the payment system, they are relevant for monetary policy only as indicator
variables, that is, only to the extent that they contain some information about the
future target variables (inflation and resource utilization). However, if credit or
asset-price developments indicate threats to financial stability or the payment sys-
tem, this may impose restrictions on the normal conduct of monetary policy and
also require special actions,” concluded Mr Svensson.

5 [5]