Movements in asset prices (such as equity or housing prices or the exchange rate), can have significant effects on real economic activity. Yet the consensus view of monetary policy is that central banks should set interest rates in response to forecast inflation, and (possibly) the output gap, but that they should not react directly to movements in asset prices. The argument is usually that asset prices are too volatile to be of much use in determining policy; that misalignments in asset prices are almost impossible to identify, let alone correct; and that systematically reacting to asset prices may prove to be destabilizing.

The second of CEPR's Geneva Reports on the World Economy, produced in conjunction with the International Centre for Monetary and Banking Studies (ICMB), reviews the arguments on these issues and presents new analysis and evidence. The Report takes issue with the consensus view, concluding that central banks can improve macroeconomic performance (defined in terms of minimizing the variability of output and inflation) by reacting systematically to asset prices, in addition to their reaction to the inflation forecast and output gap. Of course, it can be argued that some asset prices (e.g. the exchange rate) already appear in central banks' policy functions as they often have a direct effect on the banks' inflation forecast. However, the Report's recommendations go beyond this, concluding that a central bank should react to asset prices themselves, not just the effects of asset prices.

Theory

The Reports notes that asset price bubbles create distortions in investment and consumption, leading to extreme rises and then falls in both output and inflation. Modestly raising and lowering interest rates as asset prices rise and fall above and below what are estimated to be warranted levels helps to smooth these fluctuations, thus reducing the possibility of an asset price bubble forming in the first place. The intuition behind this result is based on two simple arguments.

The first is an application of the classic Poole analysis, which states that a central bank should 'lean against the wind' of significant asset price movements if these disturbances originate in the asset markets themselves, say, for example, because of 'irrational exuberance'. Such a policy should attenuate the disturbance's influence on the real sector of the economy. In contrast, if the disturbance originates in the real sector, asset prices should be allowed to adjust accordingly. The second argument is explicitly intertemporal. It is based on the notion that when significant asset price misalignments occur, they help create instability in inflation and/or employment that may be exacerbated when the misalignment is eventually eliminated. A pre-emptive policy approach will tend to limit the build-up of such misalignments and the size of the eventual correction, thereby lowering the medium-term variability of inflation and output. The Report argues that such a policy is desirable in general, even if it means a temporary departure from the short-term inflation target.

The Report then examines whether these intuitive arguments are robust, by conducting extensive simulations using two models incorporating sophisticated treatments of asset markets and realistic assumptions about the dynamic effects of policies and disturbances. The first of these is a generalized version of a standard dynamic new-Keynesian model used by Bernanke and Gertler in their recent influential study of how central banks should react to equity price bubbles. Bernanke and Gertler conclude that policy should not respond to movements in asset prices except insofar as asset prices signal changes in expected inflation. By contrast, the Report finds that in the vast majority of cases it is strongly advisable for interest rates to respond to equity prices.

The disparity between the Report's conclusions and Bernanke and Gertler's study appears to stem from the wider range of policy responses that the Report considers. Specifically, the Report allows for the inclusion of an output gap in the policy rule (making the rule more like a standard Taylor rule); the introduction of an objective function for the central bank so that optimal policy rules can be calculated; the inclusion of interest rate smoothing in the policy rule; variations in the extent to which agents are backward looking; and a reduction in the degree of leverage that stems from the knowledge that the authorities will react to a bubble. The authors of the Report conclude that macroeconomic stability is well served if monetary policy reacts in part to asset price misalignments; the reaction is not necessarily large – but it should nevertheless be there.

The second model the Report considers explicitly examines the question of exchange rate misalignments. In closed economy models, economists have argued that optimal policies are versions of inflation targets and Taylor rules. In an extension of such models, a paper by Ball finds that inflation targeting is sub-optimal in an open economy, in that the optimal policy targets 'long-run' inflation – i.e. inflation that is adjusted for the temporary effects of exchange rate fluctuations. These are important results, but some have questioned them on the grounds that the equation for the exchange rate in Ball's model is unconventional: it does not incorporate the uncovered interest parity condition (a cornerstone of most theoretical models of the exchange rate), and exchange rate expectations do not play a role in affecting the current exchange rate. The Report re-examines this issue with a model that includes both these aspects of exchange rate determination and has previously been used by the Bank of England to analyse optimal inflation forecast horizons.

Specifically, a conventional small open-economy model is shocked each quarter by unobserved random disturbances in aggregate demand, aggregate supply and capital flows. The results of the simulations suggest that the optimal policy is to respond to exchange rate fluctuations when these fluctuations arise from portfolio shocks. This response is above and beyond the direct effect that the exchange rate movements have on the inflation forecast. If, instead, exchange rate fluctuations stem from aggregate demand shocks, then the optimal policy ignores the exchange rate altogether. Hence whether a central bank should respond to a shock crucially depends on the type of shock.

A commonly held view among economists is that the UK's experience of shadowing the DM during the late 1980s was a mistake as it allowed inflationary imbalances to build up. How might this experience be viewed in the light of the models discussed above? First, at least a part of what happened then was a significant shock to aggregate demand associated with financial liberalization. Hence the experience can be viewed as consistent with the model, which suggests that the only effective policy is to 'lean into' portfolio shocks. Second, the open-economy model only considers one asset price (i.e. the exchange rate). Yet the first model illustrated that there may well be a case for including other asset prices (i.e. stock prices, the housing market) in the bank's reaction function. It is worth recalling that in the late 1980s the UK's housing market appeared overvalued, with a price-earnings ratio higher than at any other time during 1970-2000. A monetary policy rule that took all asset markets into account would surely have reacted to the overvalued housing market in the 1980s.

Measurement

Many policy-makers are hostile to the notion of taking action to prevent asset price misalignments because of the difficulties associated with distinguishing between movements in asset prices that are warranted by underlying fundamentals and those that are not. But implementing 'conventional' monetary policy also requires estimates of asset price misalignments since inflation forecasts often depend, in part, on asset prices. In this sense, estimates of asset price misalignments already influence the policy function.

The Report analysed the valuation of the US equity market in 2000, concluding that even under optimistic assumptions about the increase in underlying productivity growth, the equity risk premium is currently towards the lower end of its historical range. Since econometric evidence suggests that this premium is likely to revert towards its mean in the medium term, it is probable that this will occur at least in some cases through an adjustment in equity prices. The Report therefore concludes that measurement difficulties, as real as they are, should not stand in the way of attempting to incorporate estimates of asset price misalignments into the monetary policy-setting process.

In addition, it is probably no more difficult to measure the degree of stock price misalignment than it is to measure the size of the output gap or the equilibrium value of the real interest rate, concepts that many central banks already use in preparing their inflation forecast. Specifically, output gap estimates depend on estimates of underlying productivity growth and the equilibrium equity risk premium. These inputs are also necessary to estimate equity price misalignments.

Practice

The Report's proposal to take asset price developments into account, over and above the effect they have on the inflation forecast, is consistent with the remit given to the Bank of England and other inflation-targeting central banks. One possible way to implement this suggestion would be to adopt an augmented Taylor rule in which asset prices are given a role together with the inflation forecast and some measure of the output gap - this is the type of rule that was shown to be superior to a 'pure' Taylor rule in the context of the Bernanke and Gertler model. As an alternative to actually specifying a policy rule to determine interest rates, a government might instead specify to its central bank that inflation and, perhaps, output deviations should be minimized on average in the future.

In terms of implementation, it is true that the central bank needs to undertake the rather difficult task of estimating the degree of misalignment. But in order to forecast inflation accurately, the central bank needs to do this anyway. The one factor that would make the job of the central bank a little more difficult is that it would have to make a decision about how much weight to attach to the asset price misalignment. Clearly, the answer would be model-specific, but this is also true of inflation forecasts.

Many policy-makers have expressed concern over the potential for moral hazard arising from the perceived asymmetry in their policies. Indeed, these concerns seem to be justified: an informal survey of major fund managers and chief economists, documented in the Report, reveals an almost unanimous belief that the Fed reacts more to a fall in equity prices than it does to a rise. It is entirely possible that this perception has arisen because market price changes are themselves asymmetric. If policy-makers react equally to sudden rises and falls in market prices, but if the markets only exhibit sudden falls, then their behaviour may seem asymmetric. The result of this is that investors may feel insured against big losses and are thus prepared to place one-sided bets, which itself can push equity prices even higher. One advantage of the Report's proposal is that central banks would have a policy rule that is explicitly symmetric, this may help reduce misperceptions among market participants. In addition, the Report examines whether central banks could use alternative policy instruments to influence asset price developments, namely margin requirements and policy signals. It concludes that neither of these alternatives are substitutes for traditional monetary policy in affecting asset prices.

Measuring and Forecasting Inflation

Much of the discussion on the inclusion of asset prices in the measurement of inflation centres on the idea that asset price movements give information about future inflation - i.e. asset prices will increase in anticipation of future price increases. The authors view this issue as entirely empirical: anything that can be used to improve inflation forecasts should be. In the context of this rationale, the weights assigned to asset prices should relate to their relative contribution to the inflation forecast. This necessitates constructing an index of overall inflation that reflects as much as possible the common trend in all prices (i.e. core inflation) and as little as possible the idiosyncratic behaviour of individual prices. The index should be a weighted average of all price changes in the economy, where the weights chosen are inversely proportional to the volatility of the price change of the factor in question.

In order to determine whether asset prices belong in an index based on this approach, the authors use a set of quarterly data for 12 OECD countries, which allows them to calculate the weights of housing and equities in a measure of core inflation for each country. The results are what would have been expected. Since stock prices are so much more volatile than consumer prices, their implied weight is very low, never exceeding 2.5%. But housing prices are less volatile and therefore carry more information about core inflation. Overall, the Report recommends that current inflation measures could benefit from an increased weight on housing, but that the current practice of ignoring equity price changes is justified.

In addition to concluding that (some) asset prices should be included in inflation measures, the Report also concludes that asset prices contain information about future inflation that can be incorporated into inflation forecasts. There exist a large number of empirical studies that show significant relationships between changes in asset prices and future inflation. These relationships are not identical across countries and may even change over time, but by performing out-of-sample forecast comparisons, the authors show that asset prices do provide useful information about future inflation in a number of countries and time periods.

This conclusion is confirmed by simulations carried out on models employed by central banks to prepare forecasts used as inputs in the policy decision process. For example, changes in equity prices have significant effects on both inflation and output in a model used by the Federal Reserve. Similarly, inflation and output are strongly influenced by the exchange rate and the price of housing in the Bank of England's macroeconometric model. Equity prices tend to have a greater impact in the US because of the larger capitalization of the US equity market and the larger share of household wealth represented by stocks, whereas exchange-rate changes are more important in countries where exports and imports make up a greater proportion of GNP. The recognition that asset prices can have strong effects on future inflation implies that central banks have an incentive to forecast asset prices themselves. While the authors recognize that this is not easy, they note that central banks possess a significant informational advantage over their fellow forecasters, in that they have a much better idea of their own reaction function and where interest rates are heading.

Geneva Report on the World Economy No. 2 'Asset Prices and Central Bank Policy' by Stephen G Cecchetti (Ohio State University), Hans Genberg (Graduate Institute of International Studies, Geneva), John Lipsky (Chase Manhattan Bank) and Sushil Wadhwani (Bank of England).