摘 要 考察经济周期波动的事实可以从经济史和时间序列分析两个层面入手，我们称前者为描述性事实，称后者为特征事实。描述性事实围绕特定事件展开，主要用语言文字对经济周期波动现象进行描述。虽然也会使用一些统计数据，但其结论主要是定性的，关注具体事件和细节。特征事实主要是指利用特定的方法和技巧对宏观经济时间序列进行转换和对比，从而发现的具有一定普遍性的经验规律。它的结果是定量的，关注经济周期波动在一个较长时段内的整体性质。中国的经济周期波动研究面临这样一个问题，一类经济学家很少借鉴西方主流宏观经济学的研究成果，其研究不够系统、不够深入；另一类经济学家则不注意考察历史，缺乏对中国经济周期波动现象的全面而细致的把握。本文的研究从两个层面入手，就是力图避免这种割裂，从而为进一步的理论研究提供一些基础性的指导。 从描述性事实的角度来看，我国的经济周期波动表现出十分明显的阶段性，各阶段都有其鲜明特征。造成这种状况的原因主要有两个：第一，建国前后我国基本上是一个封闭的农业经济，现代化和工业化是一项长期任务，经济发展过程的阶段性导致了经济周期波动的阶段性；第二，建国后我国迅速建立了单一公有制集中计划经济体制，这种体制的弊端很快就显现出来，此后，对经济体制进行修正和改革的各种尝试不断进行，尤其是在1978年以后进行的市场导向的经济体制改革，使经济结构和经济运行方式发生了根本变化，因此，体制转轨的阶段性造成了经济周期波动的阶段性。 在经济周期波动研究中，如何从原始时间序列中分解出波动成分一直是一个重要问题。在本文的特征事实研究中，我们选择在RBC理论研究中大量使用的滤波方法。随着RBC理论的产生和发展，先后出现了三种有影响的滤波方法，即HP滤波、BK滤波和CF滤波。由于这些滤波的目的都是要分离出原始时间序列中特定频段的成分，因此被称为频率选择滤波。在详细分析了各个滤波器的机理和特点之后，我们发现很难从理论上辩明其优劣，因此，我们同时利用三种滤波器进行研究，所幸它们的结果差异很小。 从特征事实的角度来看，改革开放前后我国宏观经济时间序列的行为差异很大，我们一共总结出十三条重要变化。这些变化中的大多数大体符合基本的经济学常识和直觉，体现了我国经济市场化、货币化和工业化程度的提高，比如：实体经济和价格的波幅变化、国家财政收支的周期性质的变化、货币供给的周期性质的变化、与英、美经济的同步性的变化等，而这些情况又都和经济发展和体制转轨这两大因素密切相关。然而，有些变化却显得比较怪异和反常，比如：价格和通货膨胀周期性质的变化、总就业的周期性质的变化、实际利率周期行为的变化等，这类变化很难用一般原理简单进行解释，需要进行更深入细致的分析。另外，我们还用中国数据检验了两个经典宏观经济关系：菲利普斯曲线关系和通货膨胀—货币增长关系。我们利用CF滤波的研究发现，改革开放前后这两个关系的表现有很大不同。改革开放前比较反常，而改革开放后的数据基本上证实了这两个关系，即失业和通货膨胀之间的替换关系只是一种短期现象，在长期这种关系不存在；货币增长和通货膨胀在长期高度相关，而在短期这种相关性并不显著。 改革开放前后的这种明显变异，给我国经济周期波动的理论研究造成了很大困难。如何在理论的内部一致性和外部一致性之间达致一个合理的折衷，建立能够解释中国现实的经济周期模型，是我们今后面临的一个重要课题。 关键词：经济周期；经验研究；描述性事实；特征事实；RBC理论；频率选择滤波 JEL分类：B30, B41, C22, E30, N15 Abstract The facts of business cycle can be studied by two ways. One is economic history, another is time series analysis. We call the former description facts, and the latter stylized facts. The description facts mostly describe the business cycle phenomenon in language. Their conclusion is qualitative and based on the events and details which had really happened, while some statistic data are always cited. The stylized facts are empirical laws which are found through applying some transforms and contrasts to the macroeconomic time series. Their conclusion is quantitative and paying more attention to the uniform characters of business cycle in a long time period. Chinese business cycle researchers can be divided into two types. Some researchers make less use of the achievements of the mainstream macroeconomics, so their works lack universality and profundity. On the other side, some researchers think little of the economic history, and their works are short of all-sided and detailed understanding to real world. Our research tries to avoid these two biases through investigating two kinds of facts, thereby providing some basic guidance for the theoretical study of Chinese business cycle. From the viewpoint of the description facts, Chinese business cycle can be divided into a few obvious stages, and each stage has its sharp-cut peculiarity. There are two good reasons why this condition exists. First, China economy had been closed and agriculture dominating when the PRC was founded in 1949. Modernization and industrialization are two long-range goals, hence the stages of economic development results in the stages of business cycles. Second, after the foundation of the PRC, the centralized planned economic institution was established rapidly, which was based on unitary public ownership. Some modifications and reforms had been carried on continuously because the abuse of this institution appeared soon. After the market-oriented economic reform started in 1978, the economic structure and operation had changed hugely. Therefore the stages of institutional change shaped the stages of business cycles.How to decompose the original time series and get the fluctuation components is always a crucial problem in business cycle research. In our stylized facts analysis, we chose the filters which were widely used in RBC program. With the emergence and development of RBC theory, three influential filters were designed to detrend the data successively, which are HP filter, BK filter and CF filter. They can all be regarded as frequency selective filter, because of their ability to remove the components of all the other frequencies while leaving the desired components intact. After analyzing the mechanisms and peculiarities of these filters, we found it’s difficult to select the best one in theory. Thus we use all the three filters in our study, and what made us so lucky is that the differences are not so much.From the viewpoint of the stylized facts, the behaviors of Chinese macroeconomic time series are very different between pre and post reform and opening up. We summarize thirteen important changes in all. Most of them exhibit the promotion of the degrees of marketization, monetization and industrialization of China economy, which are consistent with the economics common sense and intuition, for example, the change of real economy and price’s amplitudes, the changes of the fiscal revenue and expenditure’s business cycle property, the changes of the money supply’s business cycle property and the change of the co-movement with UK and US etc. Furthermore, these changes are all correlated with economic development and institutional change. On the contrary, some changes seem to be relatively erratic and abnormal, for example, the changes of price and inflation’s business cycle properties, the change of aggregate employment’s business cycle property and the change of real interest rate’s business cycle property etc. It is not so easy to explain such changes in general economic theories simply, and some profounder and more prudent research must be carried on. In addition, the Phillips curve relationship and the inflation-money growth relationship are tested by China data in this paper. Using the CF filter, we found that the behaviors of the two relationships are very different while regarding the reform and opening up as the dividing line. In pre reform and opening up duration the behaviors are abnormal. However, these relationships are verified in post reform and opening up duration: the trade-off between the unemployment and inflation is only a short-run phenomenon, and it does not come into existence in the long-run; money growth and inflation is correlated significantly in the long-run, while this correlation is not significant in the short-run.The break between pre and post reform and opening up makes it very difficult to explain Chinese business cycle in theory. How to get a reasonable trade-off between theoretical internal consistency and external consistency, and built a business cycle model which can explain the realities of China economy, is a very important task in future. Key words: Business cycle; Empirical study; Description facts; Stylized facts; RBC theory; frequency selective filter JEL Classification: B30, B41, C22, E30, N15 目录 第一章 导 论 1 第二章 经济周期波动研究的方法论 4 一、科学哲学理论简述 4 1、理性主义与经验主义 4 2、怀疑论、本质主义和工具主义 5 3、证伪主义和新异端 6 4、方法论个人主义与方法论整体主义 8 二、几个重要概念 9 1、事实 9 2、理论 9 3、事实和理论 10 4、解释 11 三、RBC理论的方法论评介 12 1、RBC理论概述 12 2、理性主义的典范 12 3、“判决性实验”？ 13 4、建模策略 13 四、经济周期波动研究中的三种事实 14 1、经济周期波动研究的目的和任务 14 2、经济周期波动研究中的三种事实 15 3、事实的客观性 17 4、描述性事实的重要性 18 第三章 经济周期理论发展简史 19 一、19世纪的百年：历史性和因果性的说明 19 二、20世纪初叶：动态分析和时间序列分析 20 三、凯恩斯革命：两个Adelman的标准 21 四、经济周期理论研究的复兴：RBC研究纲领 23 第四章 中国经济周期波动事实Ⅰ：描述性事实 26 一、初始条件的形成（1949年～1957年） 26 1、国民经济恢复时期：1949年～1952年 26 2、“一五”计划：1953年～1957年 27 3、小结 28 二、集中计划体制下的20年（1958年～1978年） 28 1、“大跃进”：1958年～1960年 28 2、国民经济调整：1961年～1965年 29 3、“文化大革命”：1966年～1976年 31 4、拨乱反正和经济恢复：1977年～1978年 32 5、小结 33 三、经济体制改革的初步探索（1979年～1991年） 33 1、改革初期的通货膨胀和失业：1979年～1983年 34 2、未能完全到位的紧缩：1984年～1986年 35 3、挤兑、抢购风潮和保守思想回潮：1987年～1991年 35 4、小结 36 四、经济体制改革的全面推进和经济周期波动的新阶段（1992年～2006年） 37 1、实现“软着陆”：1992年～1996年 37 2、通货紧缩与扩大内需：1997年～2002年 38 3、局部过热：2003年～2006年 39 4、小结 39 五、结论 40 第五章 中国经济周期波动事实Ⅱ：特征事实 41 一、引言 41 二、滤波方法的评介和选择 43 1、HP滤波器 44 2、BK滤波器 45 3、CF滤波器 46 4、评介与选择 47 三、数据的获得和初步处理 49 四、中国经济波动的特征事实 49 1、支出法GDP 51 2、就业 51 3、价格和工资 52 4、财政收支 53 5、金融部门 54 6、国外部门 55 7、小结 55 五、两个经典宏观经济关系在中国的检验 55 1、菲利普斯曲线关系在中国的检验 56 2、通货膨胀－货币增长关系在在中国的检验 58 第六章 总 结 81 参考文献 83 后记 87