شفافیت بانک مرکزی : آیا مهم است؟

We study transparency of monetary policy in a dynamic stochastic general equilibrium model. The economy consists of many industries and experiences both supply and demand shocks. The central bank has private information regarding these shocks and releases its forecasts of shocks under the transparent regime. For a certain class of preferences social welfare does not depend on the degree of transparency and the policy that keeps the wedge between the marginal rate of substitution and the marginal product of labor constant across the states is shown to be optimal. However, in general the opaque regime welfare dominates the transparent regime.

As central banks have been moving towards more transparent policies over the last two decades, some of them started publishing their own forecasts on the future state of the economy. Projections of future growth rate of GDP and inflation rate are but two examples of such forecasts. These changes in the practice of central banking resulted in considerable growth in the literature. The standard approach in the literature on central bank transparency uses a steady-state log-linear approximation of the economy. The central bank is assumed to maximize an ad hoc welfare function, which is usually quadratic in inflation and employment. In this paper, we study the issue of transparency of monetary policy based on an explicit dynamic general equilibrium modeling approach without having to rely on linear or quadratic approximations. Our model, a variant of the standard DSGE model with sticky prices (e.g., Woodford, 2003), consists of a representative household, a continuum of monopolistically competitive firms belonging to several industries, and a central bank who chooses monetary policy action to maximize welfare of the representative household. It is assumed that the central bank has private information about supply and demand shocks in the economy.2 The aspect of transparency we are interested in has to do with release of central bank projections of the future state of the economy. In the model, the central bank can operate under two monetary policy regimes, transparent and opaque. In the transparent regime, it announces its forecasts of the shocks. We assume that there is no strategic attempt to manipulate the public's beliefs and thus the forecasts are truthful. In the opaque regime, the central bank does not release its forecasts. We first establish that the Friedman rule of zero interest rate holds in both regimes even though the central bank pursues an activist monetary policy. We next demonstrate for two special cases (when (i) the economy consists of only one industry, or (ii) when preferences are Leontief with respect to goods from different industries) that under the opaque regime, the central bank can replicate the outcome of the optimal policy under the transparent regime. In other words, the welfare level under the opaque regime is at least as high as under the transparent regime. This is due to the fact that the set of allocations available to the Ramsey planner under the opaque regime includes all allocations available under the transparent regime. We next establish that for a special class of preferences (when the utility function is logarithmic in consumption or linear in leisure), the two regimes deliver the same level of welfare. This is due to the fact that the perfect-forecast allocation is optimal for both regimes; for this allocation, the wedge between the marginal rate of substitution between labor and consumption and marginal product of labor is constant across times and states.3 For preferences outside this class we are able to show that the welfare level is higher under the opaque regime as long as the central bank's forecasts are sufficiently accurate. The optimal allocation exhibits varying wedges across states. The presence of multiple industries makes the analysis more difficult than in the single-industry case because it creates nonlinearity in expectations in the implementability condition of the Ramsey problem. A numerical analysis supports our findings. It indicates that under the opaque regime, the wedges in the optimal allocation vary across states, and this yields higher welfare than the constant-wedge allocation. The standard New Keynesian model exhibits two kinds of distortion: (i) a real distortion: monopolistically competitive firms, having some monopoly power, charge a price higher than their marginal cost; this price markup (over marginal cost) induces an inefficiently low level of output; and (ii) a nominal distortion associated with stickiness of prices: the inability of firms to adjust prices when hit by supply or demand shocks. The role of monetary policy is to deal with these distortions. The markup in the economy is the ratio of price to the firms' marginal cost. It may be viewed as a distortionary tax: a higher markup raises the implicit tax on labor. With sticky prices, expansionary monetary policy leads to an increase in the marginal cost and thus to a fall in the markup and the corresponding distortion (e.g., Goodfriend & King, 1997). The sticky price distortion stems from the inability of firms to react to shocks. Let us suppose for a moment that the forecasts are perfect. Under the transparent regime, firms would know what the shocks will be in the next period and thus set their prices accordingly. In this case, even though the level of price stickiness has not changed, the prices themselves are set as if they were flexible because of the firms' knowledge of the future shocks. Therefore, with perfect forecasts, transparency completely eliminates the sticky-price distortion. Yet the sticky price distortion is still present under opacity. The same intuition works for the case when the forecasts are not perfect. However, the welfare effects of the markup distortion work in the opposite direction: the markup distortion under transparency turns out to be worse than that under opacity. Under the transparent regime, the firms' markups (the ratio of price over marginal cost) turn out to be the same and do not depend on the value of the shocks. Under the opaque regime, on the other hand, the markups change with the shock values, and this is welfare improving. Thus, our result stating welfare superiority of the opaque regime can be viewed as follows: when switching from the transparent regime to the opaque one, the effect mitigating the markup distortion associated with monopolistic competition is stronger than the effect pronouncing the sticky price distortion. Finding the mentioned trade-off between the two distortionary effects in the context of central bank transparency is one of the contributions of this paper. Let us discuss the intuition behind our results in the case when the utility function is linear in leisure or logarithmic in consumption. Consider the limiting case when the central bank receives a perfect signal about the shock. In the transparent regime, the central bank releases its forecast of the shocks, and producers set their prices taking into account this information. This is equivalent to having perfectly flexible prices. In this case, keeping wedges constant across states is shown to be an optimal policy. When the regime is opaque, price-setting firms make their decisions based only on their own knowledge of the economy. The central bank moves after the prices are set, and the optimal monetary policy actively responds to both technology and preference shocks. Despite the result mentioned earlier that the central bank can obtain a richer set of outcomes under the opaque regime, we demonstrate that it cannot do better than the optimal allocation attained under the transparent regime. The policy of keeping the wedges constant across states is still optimal. When the utility function is linear in leisure, the marginal utility of leisure does not depend on consumption or leisure, and the constant wedge allocation satisfies the optimality conditions for both the transparent and opaque regimes. When the utility function is logarithmic in consumption, aggregate employment is constant across states, and the constant wedge allocation again satisfies the optimality conditions for both the transparent and opaque regimes. Putting it differently, for this class of preferences the trade-off between the two kinds of distortions discussed earlier disappears. In practice, a growing number of central banks release their forecasts of inflation and other macroeconomic variables. For example, the Federal Reserve in the U.S. publishes its forecasts of GDP growth, unemployment rate and inflation four times a year; they look three years into the future and include the range of views of the members of the Federal Open Market Committee. The Bank of Canada publishes point estimates of inflation rate and the growth rate of GDP. The Bank of England and Swedish Riksbank publish inflation and GDP fan charts, which are probabilistic distributions over possible future values of inflation and GDP. In our model, the central bank may release a one-period-ahead probabilistic forecast of supply and demand shocks. Both households and price-setting firms take this into account when making their decisions. Despite increasing transparency of central banks and its public support this development is not without its doubters. In Morris and Shin (2002), an individual agent's payoff depends on both fundamental values and, crucially, on the expectations of other agents. If the central bank releases a sufficiently clear signal, it can act as a coordinating point that can distract agents from their private information and fundamentals, which reduces social welfare. Dale, Orphanides and Österholm (2011), in a theoretical model, demonstrate the possibility of the private sector's overreaction to central bank communication if it cannot learn the precision of the central bank's information. Our paper may be viewed as another word of caution. Given that all advanced economy central banks publish macroeconomic forecasts of some sort, one possible interpretation of our results would be a call for less detailed forecasts (e.g., publishing just point estimates instead of fan charts in the case of the Bank of England or Swedish Riksbank, or, in the case of the Fed, perhaps making the forecast horizon of individual FOMC members shorter than three years). Another interpretation has to do with forecasts of policy instrument rates. Several central banks (the Reserve Bank of New Zealand, the Riksbank, the Norges Bank, and, starting in 2012, the Federal Reserve) publish forecasts of their own policy rates. This remains a controversial issue, however (a good discussion of arguments for and against publishing instrument-rate paths is given in Svensson, 2009). In our model, all the results would still apply if the central bank forecasts involve only one shock (say, the supply shock) and its policy rate (the money growth rate) instead of both supply and demand shocks. This is true because in equilibrium, the knowledge of both shocks pins down the policy rate, and similarly, the knowledge of one shock and the policy rate pins down the information on the other shock. With this in mind, the results of this paper suggest that publishing policy-rate paths may be welfare reducing, which may explain why only a small number of central banks have adopted this policy. An important assumption made in our analysis is that the central bank truthfully publishes its forecasts. Allowing for manipulation of forecasts would, of course, make the model more realistic. Nonetheless, allowing game-theoretic issues in a general equilibrium model would considerably complicate our task. Instead, let us argue that this is a reasonable simplification. Theoretically, under rational expectations untruthful central bank forecasts will be understood by the public as such and thus either ignored or appropriately filtered. This inevitably affects the nature of equilibrium (e.g., Amador and Weill, 2010 and Moscarini, 2007). In a learning environment (Evans & Honkapohja, 2001), a bias in central bank forecasts will eventually be learned by agents. Empirically, if central banks successfully used their forecasts as a tool to manipulate the public's expectations, we would observe an increased bias between central bank forecasts and forecasts of the private sector. However, a number of empirical papers have found the opposite. Rafferty and Tomljanovich (2002) find that transparency improves forecast accuracy for the U.S. Treasury yield curve. Middeldorp (2011) examines the impact of increased transparency in a group of 24 countries and finds improved accuracy of private sector forecasts of interest rates. (Note that increased forecast accuracy does not necessarily imply improved welfare. In our model, transparency leads to better predictability but lower welfare.) A related issue is the role of judgment in producing forecasts. Even though this may be viewed as strategic manipulation we tend to think of it as part of forecasting that is complementary to formal modeling. Let us quote the views of two prominent central bankers on this matter. In Bernanke (2007), Chairman of the Federal Reserve comments: “Another reason for the reliance on judgment in the forecasting process is the practical requirement that the forecast for inflation be consistent with the staff's overall view of the economy ….” And in Svensson (2010), Deputy Governor of Sveriges Riksbank opines: “Projections and monetary-policy decisions cannot rely on models and simple observable data alone. All models are drastic simplifications of the economy. Therefore, judgemental adjustments in both the use of models and the interpretation of their results … are a necessary and essential component in modern monetary policy.” And he illustrates this approach by referring to the Riksbank's policy decision in February 2009: “In the middle of the recent financial crisis and rapidly deteriorating economic situation, the Riksbank posted forecasts quite different from the forecasts generated by the Riksbank's models.” Being aware of some pitfalls of linear-quadratic approximations (e.g., Benigno & Woodford, 2012), this paper uses explicit dynamic general equilibrium modeling approach without having to rely on linear or quadratic approximations. Moreover, it merges the literature on central bank transparency with the literature on the dynamic Ramsey approach to finding optimal policies through corresponding allocations (Adão et al., 2003, Chari et al., 1991 and Lucas and Stokey, 1983). These two features of our analysis allow us to understand the importance of a novel trade-off between two distortions in the economy and obtain, on the one hand, the welfare equivalence result between the transparent and opaque regimes for a particular class of preferences and, on the other, a more precise welfare analysis for other cases when opacity dominates transparency. The paper is organized as follows. Section 2 gives a description of the model. Section 3 characterizes equilibrium, Section 4 characterizes monetary policy, and Section 5 makes welfare comparison between the two regimes. Section 6 reviews the literature and concludes.

Various approaches to modeling central bank transparency have been proposed.7 First, when a central bank's preferences are private information and change stochastically over time, and the policymaker is free to choose the accuracy of monetary control, he may not choose the most effective available control (Cukierman and Meltzer, 1986 and Faust and Svensson, 2001). Second, transparency can make it easier for price-setters to predict the future actions of the central bank. However, this may constrain the policy maker with a low-inflation reputation when more active monetary stabilization policy is desired (Jensen, 2002). Third, when the monetary policy instrument plays a dual role of directly influencing the economy and, in addition, indirectly revealing the central bank's assessment of the economic situation, there is a trade-off between these two channels (Baeriswyl & Cornand, 2010). Fourth, in an environment with learning agents central bank communication may improve economic outcome by helping the economy escape prolonged periods of low economic activity and keeping expectations focused around a more desirable inflation target equilibrium (Eusepi, 2010). Fifth, when the central bank has more accurate information on economic disturbances and the public is uncertain about the central bank preferences, transparency in the form of publication of the central bank forecasts may become either a desirable policy (Geraats, 2005) or, on the contrary, may impair stabilization of current inflation and output (Eijffinger & Tesfaselassie, 2007). Geraat's use of asymmetry with regard to economic shocks is similar to ours. Our finding that transparency may reduce welfare is similar to that in Eijffinger and Tesfaselassie's work. We, however, provide a different intuition for our results. Some recent contributions to the growing literature on central bank transparency include Carboni and Ellison (2011), van der Cruijsen, Eijffinger and Hoogduin (2010), Eusepi and Preston (2010), James and Lawler (2011), Lindner (2009), Rhee and Turdaliev (2010), Sibert (2009), Turdaliev (2010), and Westelius (2009). In this paper, we have examined welfare implications of central bank transparency in the dynamic stochastic general equilibrium framework. The kind of transparency studied in this paper has to do with publication of central bank forecasts on the future state of the economy. We have used the dynamic Ramsey analysis for finding optimal allocations under different policy regimes. We have shown that the optimal monetary policy in both transparent and opaque regimes follow the Friedman rule of zero nominal interest rate. We have found that in choosing between transparency and opacity, the central bank faces a trade-off between the effects associated with two kinds of distortions in our environment: (i) monopolistic competition, and (i) stickiness of prices. In contrast to a large body of literature on central bank transparency where one can find arguments either for or against transparency, one of our main findings is that social welfare does not depend on the level of transparency when preferences are logarithmic in consumption or linear in leisure. It is shown in this case that, under opacity, even though the potential set of equilibrium allocations is larger and includes ones with varying wedges between the marginal rate of substitution and the marginal product of labor, it is still optimal for the central bank to aim for constancy of wedges across states. In other words, the trade-off mentioned above disappears in this case. For preferences outside this class we demonstrate that the trade-off between the two effects is important and the opaque regime is welfare superior to the transparent regime when central bank forecasts are sufficiently accurate. Our numerical analysis confirms that in general the optimal allocation under opacity does not exhibit constant wedges. The assumption of truthful forecasts and lack of any strategic incentives to manipulate the public's belief is a useful simplification. In an interesting study, Tillmann (2011) demonstrates empirically the existence of some strategic motives in submitting individual forecasts of the FOMC members of the Federal Reserve in the U.S.: non-voting members tend to submit more extreme forecasts to influence policy decision. Similar strategic incentives may exist for the central bank as a whole. This issue deserves a more careful analysis and we would like to return to it in future research.