کاوش تولید بالقوه: سیاست های پولی، اعتبار، و یادگیری بهینه تحت عدم قطعیت

The effective conduct of monetary policy is complicated by uncertainty about the level of potential output. One possible response is for the central bank to “probe,” interpreted here as actively using its policy to learn about the level of potential output. I consider a simple calibrated model in the Canadian context and examine the relationship between credibility and optimal probing. For plausible parameter values, the optimal amount of probing is small and diminishes slightly as credibility rises. Only for unrealistically low levels of credibility or unrealistically large levels of uncertainty or volatility does the optimal policy diverge significantly from a policy that ignores learning.

The effective conduct of monetary policy is complicated by uncertainty. There are many dimensions to this uncertainty: uncertainty about shocks, model parameters, data, and the “correct” model of the economy itself (see Thiessen (1995) and Poole (1998) for discussions). At a practical level, one of the key uncertainties facing policy-makers is the level of output that can be maintained without adding to inflation pressures (referred to as the level of potential output). While policy-makers can continue to refine and improve the measurement of potential output (see Kuttner (1992), Laxton and Tetlow (1992), Butler (1996), St-Amant and van Norden (1997), or Dupasquier et al. (1999) for a discussion of the various ways potential output is measured), to a considerable degree uncertainty about potential output is fundamental. Thus, the challenge for policy-makers is how to deal with this uncertainty. Three possible responses by the central bank to uncertainty about potential output that have been examined analytically are to (i) ignore the uncertainty and follow the “certainty equivalent” policy; (ii) act “conservatively,” by which is meant moving interest rates by less than is implied by the certainty equivalent policy; or (iii) “probe” or experiment, which implies that the central bank actively uses its policy response to learn about the level of potential output. To formalize probing within an economic model, one must understand what it means in terms of the behavior of the central bank. However, there is no consensus on this. One interpretation of probing is that it entails optimal learning, that is, following a more aggressive policy to learn about the parameters of the economy. Probing of this type results in more precise estimates, and therefore smaller policy mistakes in future periods. Building on Wieland’s (1998) analysis of this issue, I consider a simple calibrated model in the Canadian context and examine the relationship between this definition of probing and credibility. Many inflation-targeting central banks have put significant emphasis on attaining credibility for their policy objectives in recent years (see Amano et al. (1999) for details of steps taken by the Bank of Canada, for example). These steps have increased the accountability of the central banks and that, together with the realized inflation record, has enhanced their credibility in the sense that expectations of inflation have become more firmly anchored to the inflation target (see Johnson, 1997 and Johnson, 1998 or Perrier (1998) for evidence of this). The question addressed here is whether an increase in credibility increases the desirability of probing. In other words, should a central bank that has increased its credibility follow a more aggressive policy in order to obtain more precise estimates of the parameters of the economy? I find that, for plausible parameter values, the optimal amount of probing is small and varies little with credibility. It is only for low levels of credibility or unrealistically large amounts of uncertainty or volatility that the optimal policy with probing diverges significantly from a policy that ignores learning. Even then, the optimal amount of probing diminishes as credibility rises. At an intuitive level, the returns to probing decrease as credibility increases because credibility makes learning more difficult. As credibility increases, inflation becomes more firmly anchored to the inflation target; thus the out-turn for inflation is less informative about potential output. To illustrate this with an example, suppose that the central bank is underestimating potential output and, as a result, incorrectly believes that the economy is operating at potential. With low credibility, inflation will lie below the target, allowing the central bank to infer that its estimate of potential was incorrect. At higher levels of credibility, inflation is more firmly anchored to the target, so that inflation provides a weaker signal that potential output is higher than was previously believed. The next section summarizes the literature supporting a conservative monetary policy in the face of uncertainty regarding the economy. Section 3 summarizes articles arguing for a more aggressive policy. An outline of the model is given in Section 4, followed by discussion of the parameter values in Section 5 and results in Section 6. Conclusions follow in Section 7.

Simulations have been conducted here on an artificial economy calibrated to reflect a simple model of the Canadian economy, where probing is interpreted as following a more aggressive policy in order to learn about the parameters of the economy. The optimal amount of probing for a central bank that seeks to target inflation has been shown to be generally small, and to vary little with credibility. Only with low levels of credibility or unrealistically large levels of uncertainty or volatility does the optimal policy with probing diverge by more than one policy increment (25 basis points) from a policy that ignores learning. Even then, for most forms of uncertainty, the optimal amount of probing diminishes as credibility rises. The results also suggest that the optimal amount of probing decreases with credibility because of the positive impact increased credibility plays in reducing output and inflation volatility in the economy. The central bank’s estimated equation (9) effectively equates the inflation gap (inflation less expectations) with the output gap. At higher levels of credibility, these gaps are small on average and increasingly indistinguishable from the shock terms. The information contained in a new observation is small under such circumstances, and the central bank’s estimates do not change very much over time whether the central bank chooses to probe or not. In contrast, at lower levels of credibility, there will generally be significant inflation and output gaps, with larger improvements in the precision of the central bank’s estimates from one period to the next. The informational benefits from probing are therefore greatest at low levels of credibility, resulting in a negative relationship between the optimal amount of probing and the level of credibility enjoyed by the central bank. There are several limitations to this analysis. First, credibility is assumed known by the central bank and is independent of policy. In reality, a monetary authority cannot be sure of the amount of credibility it enjoys, and the act of probing may result in reduced credibility. Second, the scope of uncertainty is this model is limited. The monetary authority is uncertain only about the level of potential output and the slope of the Phillips curve. The reality facing policy-makers is that uncertainty is considerably more pervasive. In this setting, the benefits to probing may be larger, although this remains to be established. Third, the initial estimation errors the central bank makes in estimating β and are independent of each other. In general, these estimates will be negatively correlated so that an overestimate of the slope of the Phillips curve will imply an underestimate of the output gap. Fourth, inflation has no impact on output in this model, and so aside from entering the loss function of the central bank, has no cost to the economy. Therefore, the level of credibility does not influence the policy that is required to attain an output target, although it has a substantial effect on the policy that is required to attain an inflation target. Finally, these results may be sensitive to the interpretation of probing considered. Isard and Laxton (1998) develop an alternative view of probing in which probing only occurs when inflation is low, and credibility is endogenous.