最优利率规则：一般理论与应用

【作者】 郭凯；

【导师】 艾洪德；

【作者基本信息】 东北财经大学 ， 金融学， 2007， 博士

【摘要】 利率规则理论是以短期利率作为货币政策工具而发展起来的一套新的理论，它体现了货币经济学家在货币政策领域内的新的尝试和努力。在利率规则理论中，货币经济学家试图解决两个问题：非预期的利率规则冲击对非政策经济变量的短期影响是什么，以及一个好的利率规则应当具有怎样的性质。利率规则将短期利率作为非政策经济变量的内生反应函数，使得货币经济学家可以在一般均衡模型中探讨这两个问题。然而，由于利率规则从一开始就是出于一种偏好的设定，因此，讨论的结果并不符合稳健性和科学性的要求，这就削弱了理论预言的可靠性。在某种意义上，利率规则理论若想成为一门真正的“科学的艺术”，它就必须为内生利率规则寻找到一个坚固的微观基础。本文在一个改进的一般线性理性预期模型的框架内，在完全时间一致性标准下，对Giannoni和Woodford(2002)发展起来的最优利率规则理论进行了新的解释。最优利率规则理论是对传统利率规则理论的突破和发展，既延续了利率规则理论的科学发展方向，又将货币政策宏观分析建立在更加牢固、规范和科学的微观基础之上。如果利率规则理论会因此而成为一门“科学的艺术”，那么最优利率规则理论就是这门“科学的艺术”的核心。本文建立的完全时间一致性标准下的最优利率规则理论主要是围绕着最优利率规则的内在属性、规则形式和外在特征这三个核心理论命题依次展开的。从内在属性出发，我们提出了最优利率规则的三条界定标准，即完全时间一致性标准、确定性标准和稳健性标准；从规则形式出发，我们考察了Taylor规则形式(或工具规则形式)和非Taylor的目标规则形式；从外在特征出发，我们分别探讨了不同情形下的完全时间一致性工具规则和完全时间一致性目标规则的普遍最优外在特征。具体地，除导言和结论外，本文的正文将分四章来分别展开论述。第2章对利率规则理论的发展进行了综述。在这一章，利率规则理论的经验文献、理论文献以及政策评价文献均被纳入到一个统一的逻辑思维框架体系内，目的并不是要简单地罗列这一理论的事实和结论，而是要表明这些事实和结论的最终目的都是要对最优利率规则形式作出量的描述和质的规定，但由于分析方法的限制，这些描述和规定或者产生不确定性均衡，或者不稳键，或者即使是确定性的和稳健的也不可避免利率规则是事先设定的缺陷，从而为利率规则理论科学发展方向的指出和最优利率规则内在判别准则的提出奠定了基础。第3～4章构成了理论篇。在第3章，我们提出了最优利率规则的三条界定标准，即完全时间一致性标准、确定性标准和稳健性标准，这三条标准均以方程或不等式的形式给出，它们构成了最优利率规则的一般线性理性预期模型的非经济约束条件。三条界定标准使最优利率规则具有一个好的利率规则的内在属性，为进一步推导最优利率规则的一般形式和具体形式提供了一个额外的筛选机制，从而缩小最优利率规则可供选择的范围。第4章对最优利率规则的一般线性理性预期模型进行了改进，在此基础上分析了最优理性预期均衡的完全时间一致性动态特征及完全时间一致性稳健最优均衡的存在性，并进一步探寻了完全时间一致性稳健最优工具规则和完全时间一致性稳健最优目标规则的一般化形式。完全时间一致性稳健最优利率规则不仅具有完全时间一致性、确定性和稳健性的内在属性，而且还具有诸如理性预期、惯性、最小惯性等最优外在特征。尽管完全时间一致性稳健最优目标规则的惯性程度低于完全时间一致性稳健最优工具规则的惯性程度，但这并不意味着哪种规则形式比另一种规则形式更优。实际上，两种规则形式可以相互等价转化，将二者相结合的既不失“弹性”又不失“可操作性”的完全时间一致性稳健最优利率规则组合才是最优利率规则的最理想选择。第5章构成了应用篇。在这一章，我们分别对Giannoni和Woodford(2002)提出的最优利率规则的各种可能的情形进行了改进。我们发现，即使对这些应用模型进行了改进，完全时间一致性稳健最优工具规则仍然具有惯性、最小惯性、超惯性、直接性、隐含性、有限平均预期期限等外在特征，完全时间一致性稳健最优目标规则则具有惯性、有限平均预期期限、通货膨胀预期目标倾向以及权衡等外在特征。通过考察完全时间一致性最优利率规则的平均预期期限的变化轨迹，我们发现，无论是完全时间一致性工具规则或是完全时间一致性目标规则，平均预期期限均不超过5期，这就为在实行预期工具规则和预期目标规则时主张将目标变量的预期期限定为5期以上甚至更远的观点提供了不利证据。更多还原

【Abstract】 Interest-rate rules theory, which views short-term interest rate as monetary policy instrument, reflects the new ideas and efforts of monetarists engaged in monetary policy. Two problems have to be resolved in this theory, that is, "what effects does an interest rate policy shock have on non-policy variables?" and "what is characteristic of a good interest-rate rule?" A typical interest-rate rule means that short-term interest rate is an endogenous feedback function of non-policy variables, so that the problems can be analyzed in the framework of general equilibrium models. However, the results of discussion do not satisfy the criterion of robustness and science, because of interest rate rules being predetermined by individual preferences of monetarists. As a result, the reliability of the conclusions deduced from the assumption that interest rate is an endogenous feedback function of non-policy variables is impaired. In a sense, until endogenous interest-rate rules find out a firmer micro-foundation for themselves, ’art of science’ does not fall on the shoulders of interest-rate rules theory.This paper, in the case of an improved general linear rational-expectations model and under full time consistency criterion, gives a new explanation to the optimal interest-rules theory developed by Giannoni and Woodford (2002) . The optimal interest-rules theory is the breakthrough and development of the traditional one. Not only does it succeed to the same way as the latter strives for being ’an art of science’, but also bases macroeconomic analysis of monetary policy on a firmer, more normative and more scientific micro-foundation. As a consequence, the dreams of making interest-rate rules theory evolve into a true ’art of science’ maybe come true, and then the optimal interest-rules theory developed here does be the kernel of this whole ’art of science’. Three core questions are raised and answered in turn in this thesis, namely intrinsic properties, alternative forms and extrinsic aspects of full time consistent optimal interest-rate rules, which offer an outline for the new theory. The first question leads us to bringing forward three criteria used to define essentially an optimal interest-rate rule. The second question provides an opportunity for us to distinguish between Taylor rules (or instrument rules) and non-Taylor targeting rules. And the third question guides logically our attention to seeking what is characteristic of a full time consistent robustly optimal instrument rule or full time consistent a robustly optimal targeting rule.Specifically, besides the introduction and summary, the text divides naturally into four chapters.Chapter 2 gives a survey to interest-rate rules theory. In this chapter, relevant empirical literatures, theoretical literatures and policy evaluation literatures are all touched on in a systematic logical way. The important facts and conclusions laid out here either are quantitative descriptions or are qualitative definitions of optimal interest-rate rules. However, these descriptions and definitions undoubtedly result in indeterminate equilibriums or non-robust interest-rate rules, or even if neither is the case, the disadvantage that interest-rate rules are simply came up with by preferences of monetary analysts is unavoidable. Thus, following the significations of ’art of science’ and the criteria of optimality, traditional interest-rate rules are not scientific and not optimal. It is the inferiority that lays a foundation for the following chapters.Chapter 3～4 constitute improvement on a general theory of optimal interest-rate rules. In chapter 3, we propose three criteria of optimality that are full time consistency criterion, determinacy criterion and robustness criterion. These three criteria amount to different equations or inequalities, corresponding to non-economic restrictions as a part of the general linear rational-expectations model constructed in next chapter. Therefore, criteria of optimality characteristic of a good interest-rate rule serve as an additional filter and more efficient an approach to identification of optimal forms of policy rules.Improvement on the general linear rational-expectations model is presented in chapter 3. Also, characterization of full time consistent dynamics of optimal equilibrium and proof of existence of full time consistent optimal interest-rate rules are accomplished in this chapter. Furthermore, in the framework of the linear rational-expectations model, we discuss general forms of full time consistent robustly optimal instrument rules and full time consistent robustly optimal targeting rules. We find that whatever alternative forms optimal policy rules take, they conform to three criteria of optimality, and some extrinsic aspects such as rational expectation, inertia and minimal inertia are always characteristic of them. As far as inertia is concerned, a full time consistent robustly optimal targeting rule is inertial to a lower degree than a full time consistent robustly optimal instrument rule, but hardly can we make certain that one policy rule is superior to the other. As a matter of fact, each policy rule has its strong point, for example, instrument rules have the advantage of ’flexibility’ and targeting rules gain from ’operationality’. A full time consistent robustly optimal instrument rule can be interpreted equivalently as a full time consistent robustly optimal targeting rule, which indicates that full time consistent robustly optimal interest-rate rules combination containing both policy rules is optimal.Chapter 5 constitute applications of optimal interest-rate rules theory. In the chapter, we improved four cases brought forward by Giannoni and Woodford (2002)。Under the circumstances, besides inertia, other apparent features such as super-inertia, directness, minimal inertia, implicitness and limited mean distance of expectation are characteristic of a full time consistent robustly optimal instrument rule, while the features such as limited mean distance of expectation, inflation-forecast target bias and trade-off are characteristic of a full time consistent robustly optimal targeting rule. Additionally, the locus of the numerical values of mean distance of expectation is reviewed, and we come to a conclusion that mean distance of expectation is no more than five quarters. Thus we find little justification for a policy that give primary attention to the targeting variables forecast at a horizon five quarters longer in the future, the mean future horizon of inflation forecast is equal to two years for instance, as is true of the inflation-forecast targeting currently practiced at the Bank of England.更多还原