GDP تجدید نظر داده شده و آینده نگری سیاست های پولی در سوئیس

This paper analyzes forward-looking rules for Swiss monetary policy in a small structural VAR model consisting of four variables taking into account data revisions for GDP. First, the paper develops an analytical method to analyze the effect of data-revision errors in GDP on the ex-ante or conditional inflation-output-growth volatility trade-off and applies it to Swiss data. Second, the effects of different targets in a forward-looking monetary policy on ex-post or unconditional volatility of inflation and output growth is explored with a simulation exercise. In general, the results suggest that focusing monetary policy on GDP growth instead of inflation may lead to an inefficient policy with both increased medium-term inflation and GDP growth volatility in the presence of GDP data revisions.

After 25 years of monetary targeting, the Swiss National Bank (SNB) adopted a new monetary policy framework at the end of 1999. Severe shocks to the demand for central bank money, especially for large bank notes and for reserves held by commercial banks at the SNB, rendered it impossible to use the medium-term target path for the seasonally adjusted monetary base as a guideline for monetary decisions. Thus, the SNB decided to abandon monetary targeting. The new framework consists of three elements. The first element is an explicit definition of price stability. The SNB regards price stability as achieved if CPI inflation is below 2 percent. The second element consists of the use of an inflation forecast as the main indicator to guide monetary policy decisions. The third element is a target range for three-month Swiss franc Libor as an operational target to implement monetary policy. As in the old concept, maintaining price stability over the medium term remains the main objective of monetary policy also in the new framework. In the new framework, the inflation forecast serves as the main indicator for guiding policy decisions. Although there is no mechanical reaction to the inflation forecast and the inflation forecast is not treated as an intermediate target, the discussion at the board about monetary policy is focused on the inflation forecast. The forecast used in the decision-making process is a consensus forecast that is derived from a series of models and indicators. The SNB recently started to publish studies regarding these models. Jordan and Peytrignet (2001) delivered an introduction to the inflation forecast of the SNB and the models used to derive it. Stalder (2001) presented the large traditional structural macro model of the SNB and Jordan, Kugler, Lenz, and Savioz (2002) provided an overview of the different VAR approaches used at the SNB. For the purpose of analyzing forward-looking policies, a small structural VAR consisting of four variables was developed by Kugler and Jordan (2004) and Kugler and Rich (2002). This research is extended by Kugler, Jordan, Lenz, and Savioz (2005), who developed an analytical method to analyze the ex-ante or conditional medium-term inflation-output-growth volatility trade-off for a forward-looking policy aiming at a convex combination of a medium-term inflation and an output-growth target in a SVAR model. This paper extends this framework in two respects. First, it considers the effects of data-revision errors in GDP on the ex-ante or conditional medium-term inflation-output-growth volatility trade-off. Second, the effects of different targets in a forward-looking monetary policy on ex-post or unconditional volatility of inflation and output growth are explored by a simulation exercise. The foregoing outline of the paper indicates that it is related to a growing literature on the effects of uncertainty about potential output and the output gap on the performance of monetary policy rules. Orphanides, 2000 and Orphanides, 2001 was one of the first, who considered this question using real time data for the U.S. output gap. He concluded that data-revision errors in the output gap were of crucial importance for excessively expansionary U.S. monetary policy in the sixties and seventies and that neglecting such data-revision errors leads in general to policies that are too activist. However, Svensson and Woodford (2000) argue that this result is mainly caused by the fact that in Orphanides’ framework the central bank behaves as if there were no data-revision errors and that the problem disappears when the optimal policy rule is a function of the best estimate of the state variables, in particular of the output gap. Nevertheless, the presence of uncertainty with respect to potential output results in welfare losses and has effects on simple Taylor-like rules in standard macroeconomics models (Smets, 2002). Extensions and quantitative illustrations of these results are found in Ehrmann and Smets (2003), who build a small stochastic general equilibrium model with endogenous persistence calibrated to the euro area. Briefly, their exercise indicates that the data-revision problem leads to substantial welfare losses mainly in the form of high output-gap variability. Moreover, simple Taylor rules appear to work well when an optimal output-gap estimate is used and potential output uncertainty favors the appointment of a conservative (in the sense of Rogoff (1985)) central banker. This paper differs from the literature sketched above in the following ways. First, we do not consider uncertainty with respect to the output gap or potential output, but only with respect to GDP, which is by itself already large as a result of strong revisions of the quarterly national accounts. Note also, that we do not face the problem of obtaining a best estimate of an unobservable variable such as the output gap, since only GDP growth is considered in our model. This approach can be seen as a way to circumvent the problem of measuring the level of potential output as recommended by Orphanides (2000) and others. Second, we do not consider an explicit structural model of the economy, but our analysis is based on the impulse response estimated using a SVAR model. Third, policy makers do not account for uncertainty related to the real-time GDP figures. Thus, we proceed on the assumption that the first-release quarterly GDP figure is the best available at the time the policy decision is taken. The remainder of the paper is organized as follows. Section 2 briefly describes the SVAR model for the analysis of Swiss monetary policy and gives a short account of how the inflation-growth trade-off is determined in the absence of GDP data-revision errors. Data-revision errors are taken into account in Section 3. Section 4 considers the effects of non-equilibrium initial conditions and policy dynamics and Section 5 concludes.

In this paper, we analyzed forward-looking rules for Swiss monetary policy in a small structural VAR consisting of four variables in the presence of GDP-revision errors. There are two main results of the paper. First, if data-revision errors in GDP are taken into account, there is no longer a convex efficiency frontier between the conditional variance of medium-term inflation and GDP growth: increasing the weight of the output-growth target (1 − α) from 0.58 to 1 increases the conditional variance of both medium-term expected inflation and growth. This result is due to the fact that with data-revision errors, monetary policy reacts too strongly to noisy data if the weight on output-growth targeting becomes too large, as data-revision errors have a strong impact on the growth forecast but not on the inflation forecast. However, if α > 0.82, this effect is slightly out-weighed by a loss of “diversification” brought about by the negative correlation of the policy reaction to data-revision error-induced responses of medium-term inflation and growth. Therefore, a strict medium-term inflation strategy is inefficient, even if the costs of increased volatility of inflation and growth compared to the case α > 0.82 are relatively small. The second result shows that this effect of data-revision error is reinforced when non-equilibrium initial conditions and the consequences of endogenous policy dynamics are taken into account. In fact, in the presence of GDP data-revision errors, policy reactions to the inflation and growth consequences of past policy decisions may destabilize the economy even if the weight on the medium-term GDP growth target is high. In general, the paper indicates that under realistic assumptions, the central bank induces a higher variability of both output growth and inflation by concentrating too strongly on output growth. The existence of data-revision errors for GDP forcefully underlines the limits and the risks of a monetary policy aimed at output stabilization. Even if the central bank only cares about growth stabilization, the weight on this target relative to that on the inflation target should be clearly smaller than 1 or even zero. Thus, our results confirm the concerns about output-growth stabilization first raised by Orphanides (2000), when real-time data problems are taken into account. In addition, our findings are in line with the theoretical analysis of Gaspar and Vestin (2004), who consider a non-optimizing policy maker following reasonable policy precepts in a standard New Keynesian model with uncertainty about the output gap. The results of the paper can also be linked to the literature on time-inconsistency of monetary policy of the type introduced by Barro and Gordon (1983) and Kydland and Prescott (1977).9 In a seminal paper, Rogoff (1985) showed in this context that delegating monetary policy to a conservative central banker, i.e., a central banker who is relatively more concerned about inflation than society as a whole can decrease the variance of both inflation and output growth and thereby improve the welfare of society. In our model, time inconsistency is not a problem. However, we can interpret α as the fraction of inflation hawks in a central bank board. Thus, even if society had strong preferences for output stabilization, i.e., if α is 0, it would be an advantage for society to appoint a conservative board with a α clearly bigger than 0. Such a board may deliver a smaller variability of output growth than a board that reflects exactly the preferences of the public. Consequently, our results support the view that the government should appoint conservative central bankers, when it is politically difficult to motivate a downgrading medium-term GDP-growth targeting by data-revision problems for GDP.