هدفگذاری تورم، یادگیری و نوسانات توبینز کیو در اقتصادهای کوچک باز

This paper examines the welfare implications of managing asset-price with consumer-price inflation targeting by monetary authorities who have to learn the laws of motion for both inflation rates. The central bank can reduce the volatility of consumption as well as improve welfare more effectively if it adopts state-contingent Taylor rules aimed at inflation and Q-growth targets in this learning environment. However, under perfect model certainty, pure inflation targeting dominates combined consumer and asset-price inflation targeting.

Many countries now practice inflation targeting, but that has not immunized economies from experiencing asset-price volatility (for example, in the form of exchange rate instability in Australia or share-market bubbles in the United States). The practice of controlling changes in goods prices is taken for granted by many central banks, but there is no consensus about the management of asset-price inflation, except in the sense that it is not desirable for asset prices to be too high or too volatile. At the World Economic Forum in Davos in 2003, Lawrence Summers suggested that policy makers should use other tools, such as margin lending requirements or public jawboning, to combat asset-price inflation. He compared raising interest rates to combat asset-price inflation to a preemptive attack, and stated ‘it takes enormous hubris to think you know when the right moment has come to start a war (Summers, 2003, p. 1).’ Recent research shows that central bankers should not target asset prices (e.g., Bernanke and Gertler, 1999 and Bernanke and Gertler, 2001; Gilchrist and Leahy, 2002) for a closed economy study. However, Cecchetti et al. (2002) have argued that central banks should ‘react to asset price misalignments’. In essence, they show that when disturbances are nominal, reacting to close misalignment gaps significantly improves macroeconomic performance. Smets (1997) has also stressed that the proper response of monetary policy to asset-price inflation depends on the source of the asset-price movements. If productivity changes are the driving force, accommodation is called for, and real interest rates should remain unchanged. However, if the source is due to non-fundamental shocks in the equity market, in the form of bullish predictions about productivity, then monetary policy should raise interest rates. In contrast to previous studies we evaluate monetary policy in a small open-economy framework, and in particular we are concerned with investment in a resource-rich small open economy subjected to the vagaries of international terms-of-trade shocks. Detken and Smets (2004) have shown that high cost asset-price booms are as common in small open economies subject to fundamental terms-of-trade shocks as they are in relatively closed economies driven by fundamental productivity shocks. We also highlight learning on the part of the central bank. For a small open economy subject to terms-of-trade movements, learning behavior on the part of the policy authority is an appropriate assumption, since movements of the terms-of-trade are determined in international markets far removed from the influence of domestic policy actions. In this context, central banks are more likely to be engaged in learning behavior.1 The economy we study has an export sector and an imported manufactured goods sector. The terms-of-trade are driven by movements in the commodity export price relative to the price of manufactured goods. The volatility of this relative price in turn affects share prices and investment in the booming (or declining) export sector. In this paper, we consider the rate of growth of Tobin's Q, first introduced by Tobin (1969), as a potential target variable for monetary policy. Our reasoning is that Q-growth would be small when the growth in the market valuation of capital assets corresponds roughly with the growth of replacement costs. Since asset prices (in the market value) are a lot less sticky than goods prices (in the replacement cost), the presence of high Q-growth would be indicative of misalignment of market value and replacement cost, in other words an indication of an ‘excessive’ change in the share price. Thus monitoring and targeting Q-growth may be viewed as a proxy policy for monitoring and targeting asset-price inflation, but with the advantage that the asset price is evaluated relative to a benchmark (the replacement cost). The focus on Q is also influenced by Brainard and Tobin (1977), who argued that Q plays an important role in the transmission of monetary policy both directly via the capital investment decision of enterprises and indirectly via consumption decisions. Thus volatility of Q has implications for inflation and growth. Large swings in Q can lead to systematic over-investment, and in the open-economy context, over-borrowing and serious capital account deficits. This paper is concerned with the thought experiment: what happens to consumption, inflation and welfare if the central bank also monitors Q? In particular, we will generate the welfare implications of adopting a stance of monetary policy which includes targeting consumer price inflation as well as changes in Q. We assume that the policy makers have to learn about the nature of the shock as well as the underlying laws of motion of Q-growth and price inflation, subject to uncertainty about the underlying model. We first examine the performance of Taylor rules in a no-learning context when the central bank knows the underlying true model. Two types of Taylor rules are examined – optimal Taylor rules and rules with pre-set Taylor coefficients. We show that, in a no learning environment, there is no case for including asset-price inflation as a target for monetary policy – the welfare differences are minor. In this case, since the underlying true inflation process depends, in part, on asset-price inflation, there is no need to target asset-price inflation as well as goods-price inflation. We then present the implications for two monetary policy scenarios with learning behavior. First, we consider standard Taylor rules for inflation targeting with and without reacting to Q-growth and then we examine state-contingent Taylor rules where monetary policy is more cautious. In this case, policy makers react to price inflation or Q-growth only when their forecasts cross-critical thresholds; otherwise they refrain from taking action by raising or lowering interest rates, except in a worst-case scenario. This approach is similar to a ‘worst-case’ robust control approach to monetary policy design, put forward by Rustem et al. (2005). They show that under uncertainty this approach leads to more moderate policy responses and represents a form of ‘cautionary monetary policy’ advocated by Brainard (1967), who argued that the degree of policy activism should vary inversely with the extent of uncertainty about policy effectiveness (Rustem et al., 2005, p. 15).2 To anticipate the result, we show in this paper that learning is the key assumption for justifying asset-price inflation targeting but that the presence of forecast errors implies that the Q-growth target should only be incorporated in a state-contingent Taylor rule. The paper is organized as follows. The model is described in Section 2, and the solution algorithm is presented in Section 3. Section 4 contains the simulation results for the alternative policy frameworks with and without learning. Concluding remarks are in Section 5.

This paper examined the effect of incorporating the rate of growth of Tobin's Q as an additional target to inflation for monetary policy in a learning environment. Our simulation results show that, in a learning environment, adding Q-growth in a linear Taylor case is welfare-reducing, but adding Q-growth in the state contingent case is welfare-improving. The intuition is that there are errors associated with forecasts, and the addition of an extra target compounds the forecast errors (especially since inflation and Q-growth are positively correlated). In the linear Taylor rule case, the interest rate is reacting continuously to information (measured with errors) whereas in the state-contingent case, the interest rate is only reacting to information about inflation and Q-growth when they are outside the upper and lower bounds. In the learning environment with state-contingent Taylor rule, adding information about asset-price inflation helps because it improves the central bank's ability to forecast inflation. When the central bank does not know the true model, including asset-price growth brings into the policy process forward-looking information which then improves the effectiveness of monetary policy. Under no learning, or perfect model certainty, the introduction of a Q-growth target in addition to CPI inflation is, on average, welfare-reducing, albeit negligible. In this paper, we have assumed that the driving force for Q-growth comes from fundamentals, both in the underlying model and in the learning process. Given that the central bank has to learn the laws of motion of Q-growth as well as inflation, and set policy on the basis of longer-term laws of motion of these variables, it seems reasonable to start with Q driven solely by fundamentals. We leave to further research an examination of the robustness of our results to the incorporation of bubbles and other non-fundamental asset-price shocks.10 Finally, we note that our time-varying state-contingent interest-rate rules, coming from uncertainty about the true laws of motion of consumer and asset-price inflation dynamics, generated by a non-linear stochastic model, is a step away from the design of a non-linear interest-rate rule, in which the laws of motion are approximated by non-linear approximation methods. It may be that non-linear policy rules may show even more beneficial effects from a cautionary monetary policy aimed at asset price as well as consumer price inflation.