هم ارزی اقتصادسنجی کلان، نا هنجاری اقتصاد خرد، و طراحی سیاست های پولی

Macroeconometric equivalence means that estimates of DSGE models using first-order approximations to equilibrium conditions fail to distinguish between alternative preference/technology configurations. Microeconomic dissonance means that the underlying microeconomic differences between ostensibly equivalent models become important when optimal monetary policy is derived. The relevance of these concepts is established by analysis of optimal monetary policy using a small-scale New Keynesian model. Microeconomic and financial datasets are promising tools with which to overcome the equivalence/dissonance problem.

A significant dilemma for monetary policy advice and model selection arises from the coexistence of two phenomena: macroeconometric equivalence and microeconomic dissonance. The term macroeconometric equivalence describes a situation where approaches based on estimating first-order approximations of model equilibrium conditions on aggregate time series data do not reveal definitively the economy's underlying preference/technology structure. For some positive-economics applications—for example, determining the degree of forward-looking behavior in pricing or spending decisions—the equivalence need not pose major problems. The first-order properties of the model may be sufficient for answering many positive-economics issues, and no harm may arise from taking two models to be interchangeable if their first-order dynamics are isomorphic. Normative applications, however, raise more concern. Results regarding optimal monetary policy do depend on the objective functions and production functions in the underlying nonlinear economy. Models that are equivalent when loglinearized therefore need not be equivalent in what they imply for optimal monetary policy—i.e., for the optimal steady-state inflation rate and the characteristics of efficient policy in the stochastic economy. Microeconomic dissonance refers to case where two models whose structural equations are first-order equivalent yield different optimal monetary policies. This study considers several strategies for resolving the dilemma posed by the equivalence/dissonance dichotomy, and offers conclusions about which strategy should be followed. The macroeconometric equivalence/microeconomic dissonance issue has received little attention in the modern monetary policy literature. While King and Wolman (1996), for example, provide a detailed analysis of the sensitivity of the optimal inflation rate to different parameter assumptions in their dynamic general equilibrium model, they do not consider the sensitivity of the optimal-policy analysis to assuming a different (but first-order equivalent) price-setting specification. This is despite the fact that economists have been well aware of the tendency for macroeconometric equivalence to arise between models that are far apart in their basic assumptions about private sector behavior. The notion that different rational expectations models may deliver the same linearized dynamics is of long standing: Sargent (1976) noted that two different structural models can deliver the same reduced form even when only one model imposes the natural rate restriction, while Taylor (1997), among others, noted that certain sticky-price and Lucas-style imperfect-information models deliver similar aggregate supply relationships. Likewise, instances of microeconomic dissonance, while less prevalent and less appreciated, underpinned such early contributions to the New Keynesian literature as Caplin and Spulber (1987) and Ball and Romer (1990). Caplin and Spulber produced a case where price stickiness at the microeconomic level magnifies the welfare costs of inflation but produces identical monetary-neutrality results to those of a flexible-price model. Ball and Romer provided an example of two preference specifications which, while equivalent in their implications for the degree of aggregate output volatility, lead to substantially different welfare costs from that volatility. But the taking-off of New Keynesian models in the last fifteen years has not been associated with a major reaffirmation of the dissonance warning. The modern New Keynesian literature has typically proceeded under the assumption that observationally equivalent models do deliver similar policy prescriptions. Our conjecture is that this conclusion has been prevalent until now because it followed from the study of the best-known instance of macroeconometric equivalence in the New Keynesian literature: that of the Rotemberg (1982) and Calvo (1983) price-setting specifications.1 Rotemberg and Calvo price schemes have very different microfoundations: in the Rotemberg setup, all firms vary prices each period as a continuous function of marginal cost; in the Calvo setup, a fraction of firms is selected randomly to adjust prices each period, the remaining fraction being prohibited to adjust, so price adjustment at the individual-firm level is very abrupt rather than continuous. Yet the two price adjustment specifications deliver equivalent aggregate Phillips curves (Rotemberg, 1987 and Roberts, 1995). There is therefore macroeconometric equivalence and, given the different model underpinnings, the potential for microeconomic dissonance. But the optimal policies implied by the Calvo and Rotemberg alternatives are not, in fact, very different quantitatively.2 This influential equivalence result is therefore probably responsible for the widespread impression that microeconomic dissonance is not an important phenomenon in modern New Keynesian modeling. The objects of this study are to dispel this impression and offer strategies to resolve the resulting dilemma for policymaking and modeling. Our examples of equivalence do not simply draw on the existing literature; nevertheless, and unlike the aforementioned early New Keynesian contributions, the focus is on the standard, modern New Keynesian model consisting of the forward-looking IS and Phillips curves. This focus establishes that important equivalence and dissonance results emerge even with this widely used benchmark model. This model is, in addition, essentially a restricted and stripped-down version of the dynamic stochastic general equilibrium (DSGE) models estimated in such studies as Christiano et al. (2005), Smets and Wouters, 2003 and Smets and Wouters, 2005 and Levin et al. (2005). As these medium-scale models explain actual U.S. and euro area data well, it is realistic to say that the New Keynesian literature is converging on a DSGE model whose first-order approximation is a good description of macroeconomic data. It has accordingly become imperative to evaluate the differences in policy advice implied by models that are equivalent in their first-order properties, and also to determine the best strategy for discriminating between alternative microeconomic underpinnings of such models. And it deserves emphasis that policy advice cannot typically be determined by the first-order dynamics of these models. True, in some positive-economics applications—for example, estimation of Phillips or IS curves, or estimation of the monetary policy rule over a sample period in which policy has not attempted to maximize household utility—only the loglinear approximation of the model may be needed. But, as noted above, the same is not true for normative applications. Increasingly, it has become standard to draw out the policy implications of a microfounded model by determining optimal monetary policy in that model. Even when studying simple monetary policy rules, it is not unusual to rank these rules according to the extent that they maximize household utility. This involves evaluation of the nonlinear utility function, or of a second- or higher-order approximation of utility. Either way, higher-order properties of the model become relevant, and one cannot draw policy implications immediately from the loglinear representations of the model, which do not adequately identify these nonlinear elements. The equivalence of two underpinnings of the New Keynesian Phillips curve is established below. Each version of the Phillips curve arises from a particular type of strategic complementarity: firm-specific inputs in one case, and a kinked demand structure in the other. Numerical results for optimal steady-state inflation are provided that demonstrate the contrasting policy advice implied by each specification. Analysis of the aggregate demand side then establishes that the standard optimizing IS equation is consistent both with orthodox expected-utility preferences and with Epstein and Zin (1989) preferences, and show that the different rationalizations for the IS curve are associated with distinct dynamic properties of optimal monetary policy. The discussion then turns to an exploration of alternative strategies for dealing with the equivalence/dissonance combination. A potential remedy for the equivalence/dissonance problem may be found in econometric procedures capable of estimating versions of the model based on higher-order approximations (as in Fernández-Villaverde and Rubio-Ramírez, 2007), or in considering alternative macroeconomic data. But our conclusion is that a more promising alternative is to deploy datasets not consisting purely of macroeconomic time series. Microeconomic datasets may be very revealing about economic structure. Studies by Bils and Klenow (2004), Angeloni et al. (2006), and Nakamura and Steinsson (2008) emphasize the value of microeconomic level information in understanding inflation dynamics. The value of these extra datasets in resolving the equivalence/dissonance dilemma does not arise inherently from the fact that they are microeconomic data, but that they constitute a different type of data from standard macroeconomic series, and so help pinpoint parameters not identifiable using macroeconomic data. Resolution of macroeconometric equivalence and microeconomic dissonance need not always involve considering microeconomic data, but will typically involve looking at data beyond macroeconomic time series. Besides microeconomic data, financial data are promising candidates in this connection. Asset price analysis could prove to be similarly revealing about aggregate demand behavior. Taken together, these alternative datasets provide a discipline on the specification of models intended for monetary policy analysis that macroeconomic data often fail to provide, and in so doing help draw out more accurate policy implications of a macroeconomic model. The analysis proceeds as follows. Section 2 describes a prototype New Keynesian model, describing the nonlinear environment, deriving the implied equilibrium conditions, and setting out the central loglinearized equations. Section 3 considers two real rigidities and their implications for the slope of the Phillips curve and optimal policy. Section 4 considers strategies for resolving the equivalence/dissonance dilemma. Section 5 turns the analysis toward the IS equation, focusing on the IS slope parameter, and detailing the different welfare implications of risk-sensitive preferences compared to the standard expected-utility case. Section 6 discusses the deployment of financial data to bring out differences between macroeconometrically equivalent IS curves. Section 7 provides concluding remarks.

The preceding analysis has illustrated the consequences for optimal monetary policy of models which feature macroeconometric equivalence and microeconomic dissonance. Alternative versions of a small New Keynesian model were presented that are isomorphic in their implied linearized macroeconomic dynamics, but whose underlying microeconomic differences return to the surface when welfare in the fully nonlinear model is analyzed. It was shown that optimal monetary policy is sensitive to the specification of economic structure even when the specifications are equivalent in their implications for the slope parameters of the IS and Phillips curves. The welfare differences across specifications are manifested in variations in period-by-period optimal policy in the stochastic economies and in different optimal steady-state inflation rates. Our conclusion is that macroeconomic applications are unlikely to break the equivalence, even those deploying nonlinear estimation methods. But financial data can break the equivalence by determining the nonlinear structure of household preferences and so revealing the underpinnings of the IS equation. Likewise, microeconomic data are revealing about Phillips curve structure because they shed light on aspects of the firm's optimization problem that are aggregated out of macroeconomic data. Together, these alternative datasets serve as an important discipline on monetary policy analysis, as they can deliver empirical rejections of models whose predictions were consistent with the macroeconomic data.