基于先行指标的经济周期转折点预测及影响因素分析

发布时间：2019-07-05 07:18

【摘要】：经济周期用来表示经济运行的周期性经济收缩的过程中循环出现的经济扩张周期的一种现象,我们通常把经济周期分为复苏、繁荣、衰退和萧条四个阶段,经济周期可以分为理论研究和实证研究两个方面的问题,理论研究的目的在于通过建立在一定假设基础上的理论模型,探寻经济周期的形成原因,并根据模型的分析结果指出经济运行的变动趋势；而实证研究的主要任务在于描述经济周期,根据各种可以利用的经济指标对经济运行的状态给出较准确的测量。本文即是对经济周期的实证研究。自从2007-2009年爆发全球经济危机以来,有关于经济周期波动转折点预测的问题又引起了广泛关注。经济周期转折点的确定对于家庭、企业和投资者来说都是非常重要的问题,因为其当前对消费、生产和投资的决策,都依赖于他们对未来经济前景的期望。对于中央银行的银行家和政策制定者来说,对经济状况的准确和可靠的预测,有利于实施适当和先发制人的政策。本文首先确定了基准指标,即国家统计局公布的一致合成指数,并使用B-B方法对基准指标确定转折点,然后在对国外的先行指标体系进行概述的基础上,利用时差相关分析和峰谷对应法等多种计量方法,对本文所使用的先行指标认真进行了大量的筛选工作,从不同的领域筛选出了12个先行指标,分别是：粗钢产量增速、化肥产量增速、汽车产量增速、生铁产量增速、水泥产量增速、M2月末数同比增速、固定资产投资完成额增速、固定资产投资本年新开工项目个数增速、出口额累计同比增速、逆转CPI、股票成交量增速和商品房销售额增速。本文所使用的数据时期为1999.1—2013.5。根据文章实证研究的需要,我们最终选定了12个先行指标,采用的主成分分析法是基于相关矩阵求解因子载荷,同时计算出与其相对应的贡献率、累积贡献率和特征值,以使因子载荷矩阵得到更明确的经济含义,采用最大方差法对初始因子载荷阵进行正交旋转。最终提取出四个因子,分别为：主要上游产品产量因子、流动性和投资因子、抑制需求因子和财富效应因子。之后本文基于四个因子及不同形式的probit模型方法,以统计局公布的一致合成指数为基准,对经济状况的转折点进行了分析与预测。本文主要对静态概率模型和动态概率模型两种不同形式的模型进行预测效果的比较,当使用单变量静态probit模型进行估计时,发现其预测结果并不可靠,所以进一步构建多变量的模型。虽然两种不同形式的probit模型所包含的解释变量是相同的,但每个解释变量的滞后期是不同的,我们根据最高的McFadden R2,来选择在不同模型中每个解释变量的最优滞后阶数。样本内预测结果表明：使用先行指标所提取的四个因子可以对我国的经济周期波动作出一定程度的预测,而且动态概率模型比静态概率模型的表现要好,我们还分别设置了0.5和0.25的概率门限值,结果发现,当使用更低的概率门限值时,提高了对衰退月的预测,但同时也增加了错误的衰退预警信号。对于样本外的预测,我们分别使用平均绝对误差(MAE)和均方根误差(RMSE)进行度量比较,结果表明：我们使用四个因子所建立的预测模型是合理、有效的。动态概率模型的样本外预测的结果也要优于静态概率模型。而且使用不同的概率门限值来预测衰退月的百分比,发现使用0.25的概率门限值比使用0.5的概率门限值可以更精确的对衰退月进行样本外预测。最后根据FAVAR模型对经济形式的影响因素进行了分析,实证结果发现：(1)在当期给主要上游产品产量一个单位的正向冲击,会对我国经济状况产生一个正向的响应,这种促进作用大概可以持续一年左右的时间；(2)流动性和投资因素对经济形势的冲击也是正向的,当增加货币供给量或者增加固定资产投资,都会对经济有一个促进作用,而且这种促进作用比主要上游产品产量的促进作用持续时间要长,约持续两年左右的时间；(3)抑制需求因素对经济会产生抑制作用,因此需求因素对经济是产生促进作用的,所以增加需求可以促进经济发展；(4)财富效应对我国经济有一个正向的冲击作用。但总体看来,对经济冲击影响最大的是流动性和投资因素,即货币供应量和投资的变化。本文的主要创新之处主要有以下三点：(1)本文对先行指标提取因子作为相应的解释变量进行实证分析。(2)本文使用了多变量的不同形式的probit模型进行预测,并对预测结果进行了比较。(3)本文使用了比较前沿的FAVAR模型实证分析了四个因子对经济状况的冲击作用。
[Abstract]:The economic cycle is used to represent a phenomenon of the cycle of economic expansion in the process of the cyclical economic contraction of economic operation, and we usually divide the economic cycle into four stages of recovery, prosperity, recession and depression, The economic cycle can be divided into two aspects: the theoretical research and the empirical study. The purpose of the theoretical study is to explore the reason of the formation of the economic cycle by establishing a theoretical model based on a certain hypothesis, and to point out the change trend of the economic operation according to the analysis result of the model; The main task of the empirical study is to describe the economic cycle and give a more accurate measurement of the state of the economic operation according to the economic indicators that can be used. This paper is an empirical study of the economic cycle. Since the outbreak of the global economic crisis in 2007-2009, there has been a wide range of problems related to the prediction of the turning point of the economic cycle Note: The determination of the turning point of the economic cycle is a very important issue for families, businesses and investors, as their current decisions on consumption, production and investment depend on their future economic outlook hope. For bankers and policymakers of the central bank, an accurate and reliable prediction of the economic situation is conducive to the implementation of appropriate and pre-emptive administration The paper first defines the benchmark index, that is, the agreement synthetic index published by the national statistical office, and uses the B-B method to determine the turning point for the reference index, and then, based on the outline of the foreign advanced index system, the paper makes use of the time difference correlation analysis and the peak-to-valley correspondence method. In this paper, we have carried out a lot of screening work on the leading indicators used in this paper, and 12 leading indicators were selected from different fields, namely, the growth of crude steel, the increase of the output of chemical fertilizer, the growth of the output of the automobile, the increase of the output of pig iron, the growth of the cement output, and the number of the end of the month of the month of the month of the month of the month. The increase of the growth rate, the completion of fixed assets investment, the increase of the number of new start-up projects in the fixed assets investment in the current year, the cumulative year-on-year growth of the export volume, the reversal of the CPI, the growth of the stock volume and the sales of the commercial The data period used herein is 1999.1 to 201. 3.5. According to the need of the empirical study of the article, we have selected 12 leading indicators, and the principal component analysis method is based on the correlation matrix to solve the factor load, and the contribution rate and the cumulative contribution corresponding to it are calculated. The rate and the characteristic value, so that the factor load matrix gets more explicit economic meaning, the initial factor load array is loaded by the maximum variance method. the four factors are finally extracted, namely, the main upstream product yield factor, the liquidity and the investment factor, the demand factor and the money, In this paper, based on the four factors and the method of probit model in different forms, the turning point of the economic situation is put forward based on the consensus synthesis index published by the statistical bureau. In this paper, the prediction results of two models of static probability model and dynamic probability model are compared, and when the single-variable static probit model is used for estimation, it is found that the prediction result is not reliable, so it is further constructed The model of the built-in variable. Although the explanatory variables included in the two different form of probit models are the same, the lag phase of each explanatory variable is different, and we select each explanatory variable in different models according to the highest McFadden R2 The results of the prediction in the samples show that the four factors extracted by using the antecedent index can predict the fluctuation of the economic cycle in China, and the dynamic probability model is better than the static probability model, and we also set the probability gates of 0.5 and 0.25 respectively. Limits, with the result that, when lower probability threshold values are used, the prediction of the recession is increased, but at the same time an error is added The average absolute error (MAE) and the root mean square error (RMSE) are used for the prediction of the outside of the sample. The results show that we use the prediction model established by the four factors It is reasonable and effective. The results of the out-of-sample prediction of the dynamic probability model are also excellent. In the static probability model, and a different probability threshold is used to predict the percentage of the decay month, it is found that the probability threshold value of 0.25 can be more accurate for the recession than the probability threshold value of 0.5 Based on the analysis of the influence factors of the FAVAR model on the economic form, the empirical results show that: (1) The positive impact of a unit in the production of the main upstream product in the current period will affect the economic situation of our country. There is a positive response, which can last about a year or so; (2) the impact of liquidity and investment factors on the economic situation is also positive, and when the amount of money is increased or the investment of fixed assets is increased, it will be The economy has a promoting effect, and the promotion rate is longer than that of the main upstream product, for about two years or so; (3) the inhibition of demand factors has an inhibitory effect on the economy, so the demand factor is Jinan has a stimulating effect, so increasing demand can promote economic development; (4) the wealth effect has a good effect on China's economy A positive impact. But overall, the biggest impact on economic shocks is liquidity and investment, that is, money The main innovation of this paper is the following three points: (1) This paper takes the first index extraction factor as the corresponding (2) In this paper, a multi-variable probit model is used for prediction. The results of the prediction are compared in this paper. (3) The FAVAR model of the comparative front edge is used to analyze the four factors.
【学位授予单位】：东北财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F224;F124.8

【相似文献】