« The Use of the Yield Curve as a Leading Indicator | Main | Buteo Jamaicensis Bernakus »

Monday, October 31, 2005

Using Monetary Aggregates to Overcome Model and Data Uncertainty

If you are interested in the finer details of monetary policy, this speech by Professor Hermann Remsperger, Member of the Executive Board of the Deutsche Bundesbank, will be worth reading. The speech is concerned with two types of uncertainty that plague policymakers, model uncertainty and data uncertainty. To make monetary policy as robust as possible against model uncertainty, he advocates using a variety of long-run and short-run models to guide policy and makes particular note of short-run New Keynesian models and more traditional long-run models as part of the analysis. This, along with policymaking by a diverse committee provides some assurance against the probability of catastrophic policy moves from assuming the wrong model. He also remarks on data uncertainty. Two key components in the implementation of the Taylor rule are the output gap and the inflation gap. The remarks focus in on the output gap and note the difficulty in measuring this quantity and evidence of systematic bias in estimates of the output gap. Because of this, he promotes the use of monetary aggregates as a means of coping with both model and data uncertainty and using the deviation of the growth rate of output from potential rather than the level of the output gap. I will keep an open mind on using monetary aggregates, but I am not entirely convinced. Fortunately, the bank has research planned that should help to decide one way or the other:

Money in an uncertain World, by Professor Hermann Remsperger, Deutsche Bundesbank: Ladies and Gentleman, it is a pleasure for me to speak to you this evening on the role of money in an uncertain world. ... All the different types of uncertainty have been picked up by academic literature. ... I will focus on two of them, one is model uncertainty and the other data uncertainty. ... I would like to convey the key message of my talk without any further delay: The analysis of monetary aggregates can play and should play an important role to cope with these two challenges.

Model uncertainty and money As central bankers we need models to structure our thoughts and to estimate the impact of our decisions. However, there is no consensus about which model is ... appropriate... As a consequence, most central banks – including the ECB – use a wide range of models. ... I think that Ben McCallum is perfectly right in that the preferred policy rule should be one that works ‘reasonably well in a variety of plausible quantitative models’. By the way, this approach can also be applied to ... whether monetary policy decisions should be made by a committee or not: The variety of views ... should be larger in a committee. If the committee members agree on a decision that is acceptable for all of the different prevailing assumptions of the transmission mechanism, this could prevent monetary policy decisions which are optimal only in one specific model...

In order to determine ... a robust policy rule, we have to take two decisions: The first is on the set of models which should be taken into account. And the second is how to weigh the results these models present... As concerns the weighting problem, some argue that minimizing the maximum loss across models is appropriate. This would prevent extremely bad results. However, others argue that this method may give too much weight to implausible scenarios. ... The decision about which set of models to include in this decision process and which models to exclude also turns out to be highly important but difficult. The inclusion of an additional model, even if it has a low probability, may strongly influence the resulting decisions.

Concerning the models to be regarded in the decision making process, I think that at least one from the New-Keynesian school ... should be included. These models are designed to capture short-run fluctuations in output and inflation ... Depending on the exact model specification, money can play an explicit role in the determination of output and inflation. However, it usually plays only a passive role... Given this more or less unimportant role for money in New-Keynesian models, some observers already see money leaving the stage of the academic discussion on monetary policy. However, I think that a few qualifications are essential.

First of all, I would argue that there is an implicit key role for money in the basic New Keynesian Model. ...  implementing a certain path of the short-term interest rate is only possible if the central bank can control the corresponding supply of base money. Second, taking New Keynesian Models with a minor role for money on board must not mean that we should leave aside other models with a key role for money. ... Models that capture this long-run relationship between monetary developments and inflation ... are ... the natural complements to models of short-run fluctuations in the toolbox of policy makers. ...

Data uncertainty and money ...I now want to make some comments on the role of money against the background of data uncertainty. One dimension of data uncertainty refers to the problem of measuring well-defined economic aggregates. ... monetary and financial data are far less affected by such measurement problems than data from the real economy. Monetary data are available very quickly. And they are available with a much smaller measurement error than other key variables. Therefore, money may serve as an important indicator of current economic activity. ... Measurement problems become even more severe if certain model variables are unobservable... Important examples include the output gap and the natural rate of interest. ... An empirical quantification of the output gap, for instance, is often based on more or less detailed assumptions about an underlying model structure. If ... policy ... depends crucially on such unobservable variables, errors in assessing the size of those variables may lead to large deviations from optimal policy. ...

Orphanides and others have shown that in the seventies and eighties, estimates of the US output gap have been subject to large and persistent measurement errors. On average, there has been a downward bias of this measure of economic activity which in turn led to a monetary policy of the Fed that - with hindsight - must be judged as too expansionary. As a consequence, the late 1970s and the early 1980s have been characterized by high inflation rates in the USA.

I believe that the Bundesbank managed to prevent this kind of problems. ...the Bundesbank did not assign an important role to the level of the output gap... Instead, we focused on deviations of money growth from target, on deviations of the inflation rate from the price norm and on deviations of the growth rate of real output from the growth rate of potential output. This approach is much less subject to measurement errors than one that uses the level of the output gap. Based on our previous research on “How the Bundesbank really conducted Monetary Policy” we have set up a new project on “Taylor Rules vs. Growth Rate Targeting” ... This follow-up project intends to answer the question whether, and if so why, responding to the inflation gap, to the change in the output gap and to monetary developments instead of putting a strong weight on the level of the output gap can be welfare improving in the presence of data uncertainty. ...

    Posted by on Monday, October 31, 2005 at 12:06 AM in Economics, Methodology, Monetary Policy | Permalink  TrackBack (0)  Comments (2)


    TrackBack URL for this entry:

    Listed below are links to weblogs that reference Using Monetary Aggregates to Overcome Model and Data Uncertainty:


    Feed You can follow this conversation by subscribing to the comment feed for this post.