1
Asset Price Bubbles and the Conduct of Monetary Policy
Forthcoming in The New Monetary Policy Consensus, edited by Arestis, Baddeley and McCombie, Edward Elgar 2005
Asset Price Bubbles and the Conduct of Monetary Policy
Under the New Monetary Policy Consensus
Nigel F.B. Allington and J. S.L. McCombie
INTRODUCTION
With the stock market boom in Japan in the late 1980s and its subsequent fall in the 1990s, as well as the US stock market boom after 1995 and fall from 1998 until early 2000, attention has focussed anew on the appropriate role for central banks and monetary policy in managing asset prices generally, but stock prices specifically.[1] When US growth approached zero in 2001 economists wanted to know whether or not earlier intervention might have tempered the boom and softened the impact of its collapse on the economy. What is not disputed is that asset markets are forward looking, as central bankers should be, so that the monetary authorities, from observing the behaviour of asset markets, can garner much useful information such as expectations about inflation and the real rate of interest that can help frame monetary policy. The relatively new markets for futures and options also provide information for the effective operation of monetary policy.
In this chapter, we consider the broad question of whether or not central banks should take
into account, either explicitly or implicitly, the rate of asset price inflation, especially when
this is much more rapid than the growth of the CPI, in the determination of the nominal
interest rate. We shall largely focus on stock market prices and confine our attention to the
advanced countries. (Emerging markets raise additional considerations, such as poor
prudential financial regulation). This immediately raises both normative and positive
issues. First, should the central bank attempt to influence asset prices through the use of
interest rates and then if it should, in what circumstances? Secondly, if the central bank
should intervene, how effective is the use of interest rates in achieving this? If the answer
to the latter is that it is not very effective, then should other instruments be used and if so, what instruments? This merges into the wider question as to whether interest rates are effective in controlling inflation, per se, through influencing aggregate demand, as the “new monetary policy consensus” suggests.
We begin by briefly considering what is the so-called new monetary consensus and inflation targeting. We next review the evidence as to whether or not the crash of an asset price bubble has any serious adverse effects on the real economy. This leads on to a consideration of two competing theories as to the causes asset bubbles, namely the efficiency markets hypothesis and the behavioural finance approach. The “new consensus” macroeconomic model discussed below and used by Meyer (2001), Arestis and Sawyer (2003), inter alios, does not include assets in its specification and so we next consider their role in the transmission mechanism. If the central bank should take cognisance of the rate of increase of asset prices, should it do so by explicitly including them in the inflationary target, thus removing any element of discretion? This is examined with reference to both the theoretical arguments and the empirical evidence. A number of simulation models that have also attempted to answer this question are assessed. Finally we look at the issue as to the relationship between interest rate changes and asset prices and report some new empirical results.
WHAT IS THE NEW MONETARY CONSENSUS?
In an influential article, Meyer (2001) has noted that a new consensus has emerged over the way macroeconomic policy should be pursued by independent central banks. This has encapsulated the essential features of the earlier monetarist orthodoxy, while removing the empirically irrelevant assumption that the central bank is able directly to control the money supply (however defined). The consensus is that central banks should target, either explicitly or implicitly, inflation using exclusively the interest rate. In effect, as Meyer makes clear, the consensus model “treats ‘the’ short-term interest rate as an index of overall financial conditions, assuming that long-term interest rates, equity prices, and the exchange rate all move in a stable and predictable way with changes in the policy rate”. Provided the inflation rate is kept to with in a narrowly specified band, in the UK two percent plus or minus one percentage point, this would also automatically lead to full employment and growth at the maximum potential level (Fischer 1996). It thus retains the monetarist assumption that what prevents the economy from being at full employment is both the signal extraction problem caused by excessive variability of inflation and the problems caused by a high rate of inflation.
Meyer (2001) presents a useful, stripped down, three-equation model that encapsulates the essentials of the theory behind the new consensus approach. The first equation is an IS curve with the output gap a function of it’s expected and lagged values and the real interest rate.
Y = aY + bE(Y) – c[R - E (p)] + u1 (1)
Yg is the output gap, R is the nominal interest rate, p is the rate of inflation, and u1 are stochastic shocks. E is the expectations operator.
The second equation is a Phillips curve that allows for sticky prices in the short run and flexible prices in the long run. The former gives scope for policy intervention, while the later incorporates the monetarist’s long-run neutrality condition.
p = d(Y) + wp + wE (p) + u2 (2)
where u2 is again a stochastic shock and w + w = 1.
The model is closed by a Taylor rule or monetary policy rule where the interest rate is a function of the equilibrium rate, the output gap and the divergence of the actual from the target inflation rate.
R = r+ E (p) + fY + g (p - p) (3)
where Y is the output gap, R the nominal interest rate; r the equilibrium real interest rate; p the inflation rate; p the inflation target; and x and z are stochastic shocks. All of the coefficients are positive. The money supply is not included as it is merely a “residual”. [2],[3]
The central banks can be given either a hierarchical or a joint mandate. The first is to achieve stability with some other subordinate objectives such as high employment, fast growth and possibly exchange rate stability. This is typified by the mandate of the UK’s Monetary Policy Committee. The second type of mandate is where price stability and high (or full) employment are given equal importance and an example of this is the Federal Reserves “partial” inflation-targeting regime. In practice, this distinction is more apparent than real, as no central bank can completely ignore output stabilisation. Moreover the assumption that ensuring price stability occurs will also “promote maximum sustainable growth” removes the Tinbergen problem that the number of instruments should be no less than the number of targets. Fiscal policy, although the budget deficit/surplus plays a role as an automatic stabiliser, is largely dispensed with.
However, the term “consensus” with respect to monetary policy is somewhat of a misnomer as, given its monetarists foundations, some Keynesians find some serious shortcomings with it (see, for example, Arestis and Sawyer 2002 and 2003).
In the new consensus model the NAIRU is the full employment level of output, and as inflation is a demand phenomenon, raising the rate of interest rate will lower aggregate demand and reduce inflation. The economy will move towards the equilibrium level of output, as the signal extraction problem is reduced. There is thus little room for output to depart substantially from its supply-side equilibrium. The model also has no role for cost-push inflation. Incorporating this in the stochastic term is unsatisfactory if one period’s inflation feeds through, via wage demands, into the next period. In the Keynesian wage-bargaining model, the NAIRU can occur at any level of unemployment depending upon the bargaining power of labour and the firms. In this scenario, if interest rates are raised to reduce inflation, the fall in demand is likely to exacerbate unemployment. There is indeed a Tinbergen problem. Given the various transmission mechanisms through which interest rates act, “the effects of a change of monetary policy may be rather loose, rather than in the precise effect which interest rate has in equation (1) [above]” (Arestis and Sawyer, 2003). Empirically, many studies find the effect of changes of the interest rate on aggregate demand is small, so interest rate changes are a blunt instrument. Given this last point, the use of interest rates may be ineffective in counteracting an exogenous fall in demand so that fiscal policy should have a role to play. But, as we have noted, the role of fiscal policy is minimised in the new consensus model.
What is the evidence about the effectiveness of inflation targeting? Since it has been practiced for ten years, time-series econometric evaluation has become possible, although most studies failed to find evidence that inflation targeting speeded up disinflation and achieved lower inflation rates (see Leiderman and Svensson (1995), Bernanke and Gertler (1999), Carbo, Landerretche Moreno and Schmidt-Hebbel (2001)). Also, importantly, the economic environment of the 1990s was benign with falling world commodity prices so that the success or otherwise of the strategy has not been vigorously tested (Cecchetti and Ehrmann 2000, Arestis and Sawyer, 2003).
More recently, however, Neumann and von Hagen (2002) found positive results although these have subsequently been heavily criticised. It is useful to consider this in a little more detail. Their main conclusions were, first, that after countries adopt inflation targeting, the volatility of inflation, interest rates and output falls to levels that are similar to those in the successful non-inflation targeting countries like Germany and the US. Secondly, the use of a Taylor Rule focuses greater attention on the control of core inflation after inflation targeting has been introduced. Thirdly, VAR evidence demonstrates the importance of inflation shocks as a source of increases in the variance of interest rates post inflation targeting. Hence inflation control assumes greater importance after the introduction of inflation targeting. Finally, inflation responds more favourably to oil price shocks after the introduction of inflation targeting.
While the evidence therefore seemed to be favourable, Mishkin (2002) noted a number of flaws in the analysis. In particular, (a) there is the question whether factors other than inflation targeting are responsible and (b) when they use Taylor Rule evidence to test the validity of their results an inflation coefficient of less than one suggests that the inflation process is unstable.[4] Also, (c) the Taylor Rule estimates do not show that inflation-targeting countries have had a more efficient monetary policy regime. Mishkin claimed (d) that the low estimates are explained by the authors using current inflation rather than forecasts of inflation and that this biases the inflation coefficient downwards (see Orphanides 2001). Furthermore, (e) using non-inflation targeting countries as a benchmark is unsatisfactory because their monetary strategies have not differed significantly from the inflation targeters. Also, (f) the VAR evidence cannot be relied upon because while it does provide useful evidence it cannot encapsulate the underlying structure of the model and the mechanics. Finally, (g) the event studies of oil price rises in 1978 and 1998, using the method of double difference, has arbitrary start dates for the analysis that may undermine the results. Hence in the final analysis their study can only be regarded as suggestive of the benefits of inflation targeting that merely represents convergence towards best practice. The best that can be said is that the evidence overall of the effectiveness of inflation targeting is as yet inconclusive.
WHAT CAUSES ASSET PRICE BOOMS AND BUSTS?
If central banks are to intervene to collapse certain types of asset bubbles in order to prevent serious real effects, they have to be convinced that the latter will materialize in the absence of any action. If, say, a stock market crash occurs without any serious adverse effects on output and unemployment, then the case for intervention, although it still may be there, becomes less pressing.[5] It should be noted though, that asset prices might still convey information about the future state of the economy without them directly affecting this in any causal sense and hence still being of informational value. The problem is that there are very different views as to the causes of asset price bubbles.
The efficient market hypothesis sees the price of an asset as reflecting the best available information concerning the fundamental values of that asset and changes in its price reflecting rational expectations about its future values. Two commonly used indices that have been used to judge whether the value of the stock market is above the fundamental level are the price-dividend and price-earnings ratios.
The purpose of an asset is ultimately to permit consumption smoothing. The standard analysis assumes that arbitrage will equate the risk-free interest rate (such as the yield on Treasury bills) to the expected rate of return on the asset (including any capital gains) less the risk premium. Hence, the price-dividend ratio is given by Pt/Dt = (1+g)/ (i+r-g), where g equals the growth of the dividend, i is the risk free interest rate, andr, is the risk premium. This is a variation of the “Gordon Equation” (Gordon 1962), which is a simple valuation formula widely used in finance. It can be seen that the fundamental asset price will increase as the rate of interest and the risk premium falls, or the expected dividend growth rate rises. If we assume that dividends are a stable fraction of earnings (D = gE), we arrive at an expression for the price-earnings ratio as P/E = l(1+g)/(i+r-g) where empirically for the US the parameter l takes a value of around 0.63 for the last century, falling to 0.52 over 1995-1999.