Monetary policy rules under model uncertainty

该研究计划概述了一种使用贝叶斯推理来研究货币政策演变的一致方法，并位于应用货币经济学，计算统计学和贝叶斯计量经济学领域的交叉点。主要的研究问题是，是否有重要的证据表明，一种广义的货币政策规则可能会随着时间的推移而变化，这可以解释央行如何决定货币政策立场。研究问题是逐步解决的，可以分为以下4个问题。前三个问题假设在正常时期使用的单一货币政策工具是利率。第四部分通过考虑非常规措施（如量化宽松）重新审视货币政策的决定因素，当利率不能进一步下降到负值区域并触及零利率下限（ZLB）时。1. 假设广义的货币政策规则随时间的推移具有恒定的参数，那么哪些宏观经济和金融变量对短期利率具有最大的解释力？2. 放弃恒定参数的假设并允许时变参数（TVP），是否存在一个通用的货币政策规则，可以最好地描述使用短期利率工具的货币政策行为？3. 考虑到解释变量参数的结构不稳定性和广义货币政策规则模型空间的时变概率，是否存在以下情况的统计证据：一、货币政策由“向后看”转向“向前看”？3.b货币当局调整其政策规则以适应金融稳定任务的更新？4. 假设整个货币政策立场都被影子利率所捕捉，即量化资产购买计划在央行资产负债表扩张时对等效无约束利率指数的影响，那么按照问题1至3所遵循的方法，哪些是最能解释影子利率的决定因素？回答这些问题的结果将大大提高我们对过去货币政策制定真实过程的理解，并将为预测泰勒规则基本面的短期利率和外汇汇率演变提供分析工具。贝叶斯模型平均（BMA）和动态模型平均（DMA）方法的样本内表现及其副产品，即变量选择，将提供足够的证据，证明实际的政策制定过程是否遵循通货膨胀和产出缺口的理论解释框架所描述的规范方法，或者在考虑到所有可用的政策工具的情况下，确实存在无意中考虑或不考虑的其他因素。既有传统的也有非传统的。

The research proposal outlines a consistent methodology to investigate the evolution of monetary policy using Bayesian inference and lies in the intersection of the fields of applied monetary economics, computational statistics and Bayesian econometrics. The main research question is whether there is significant evidence of a generalised type of monetary policy rule which might potentially vary over time that can explain how the monetary policy stance is determined by a central bank. The research question is addressed gradually and can be divided in the following 4 questions. The first 3 questions assume that the single monetary policy instrument that is used under normal times is the interest rates. The 4th re-examines the determinants of monetary policy by taking into consideration unconventional measures such as quantitative easing when interest rates cannot be decreased further down to the negative territory and hit the Zero Lower Bound (ZLB). 1. Assuming a generalised monetary policy rule has constant parameters over time, which are the macroeconomic and financial variables that have the largest explanatory power on the short-term interest rates? 2. Dropping the assumption of constant parameters and allowing for Time Varying Parameters (TVP), is there a generalised monetary policy rule which best describes the conduct of monetary policy using the short-term interest rates instrument? 3. Allowing both for structural instability of the parameters of the explanatory variables and time varying probabilities of the generalised monetary policy rules models' space, is there statistical evidence of when: 3.a Monetary policy changed from backward looking to forward looking? 3.b Monetary authorities adjusted their policy rules to account for the updated mandate on financial stability? 4. Assuming the entire monetary policy stance is captured by the shadow rate, i.e. quantifying the effect of asset purchasing programmes when central banks' balance sheet expansion occurs to an equivalent unconstrained interest rate index, which are the determinants that can best explain the shadow rate along the lines of the methodology followed from questions 1 to 3? The outcome of answering these questions will improve substantially our understanding of the real process of monetary policy making in the past and will provide analytical tools for predicting the evolution of the short-term rates and foreign exchange rates following Taylor rule fundamentals. The in-sample performance of Bayesian Model Averaging (BMA) and Dynamic Model Averaging (DMA) methodologies and their by-product, i.e. variable selection, will provide sufficient evidence of whether the actual policy making process follows a normative approach as described by the theoretical explanatory framework of inflation and output gap or indeed there are further factors that are taken into consideration inadvertently or not, taking into account all the available policy instruments, both conventional and unconventional.