PMI指数的复制：决定变量、路径分析和指数预测

发布时间：2018-06-12 03:47

本文选题：PMI指数 + 定量数据　；参考：《浙江工商大学》2017年硕士论文

【摘要】：国际金融危机过后,虽然全球经济逐步回升,但是危机深层次的影响依然存在,再加上政治等非经济因素的影响逐步加深,使得我国的发展面临了许多的不确定性和挑战性。为了科学、及时的监测经济发展状态,我国构建了采购经理——PMI指数体系,采用非定量的问卷调查数据,于每月月初计算并对外公布上一个月的PMI指数数值,来综合的反映宏观经济发展态势。在此背景下,本文基于前人的研究基础,采用PMI(-1)(滞后一期PMI指数值)、超额准备金率、法定准备金率、进口金额、出口金额、平均汇率、流通于银行体系以外的现钞M0等36个客观定量数据作为PMI指数体系的可能影响变量,同时,采用Li and Racine(2004)提出了混合数据下变量剔除的非参数方法,来确定与PMI指数存在相关关系的变量,复制出PMI指数的决定变量,并进一步,构建半参数时变系数的完全模型、路径模型,对所筛选出的变量进行了联合效应、个体效应的分析,最后,将模型推广为预测模型,对PMI指数进行了预测,从而将PMI指数由描述性统计指数推向推断性统计指数,填补了已有文献在这方面的空白。具体所做工作和得到的结论如下:首先,通过混合数据下变量剔除的非参方法进行变量的相关性和线性性的选择,模拟出PMI指数的决定变量,发现与PMI指数存在线性关系的变量为:PMI(-1)、出口金额(CE)、工业增加值(GZ);存在非线性关系的变量为:股票成交金额(GE)、公共财政收入(GR)、公共财政收支差额(GGCE)、税收收入(SR)、平均汇率(PL)、活期存款利率(HL)。其次,通过半参数时变系数的完全模型和路径模型的实证分析,发现加入6个非线性变量后,它们的联合作用会使得模型对于PMI指数有更加明显的解释能力。同时,发现各个非线性变量对PMI指数的影响各不相同,对PMI指数的拟合存在正向影响的非线性变量为:GE、GR、GGCE,没有存在负向影响的变量,其中SR、PL、HL对PMI指数拟合结果的正负向影响不明显。最后,本文构建了变异系数,通过比较,发现各个非线性变量对PMI指数波动影响大小依次是:GGCE、GE、HL、PL、GR、SR(剔除后)。最后,通过半参数时变系数的预测模型的实证分析,不仅说明了非参变量选择所筛选出的变量在未来经济运行中依然能够解释PMI指数,而且,提供了一个整体预测效果较好的预测模型,为企业、金融机构和政府等提供了判断经济形势、制定发展计划的有力依据。
[Abstract]:After the international financial crisis, although the global economy is rising gradually, the deep influence of the crisis still exists, and the influence of non-economic factors such as politics is deepening gradually, which makes the development of our country face a lot of uncertainty and challenge. In order to monitor the state of economic development in a scientific and timely manner, China has constructed a purchasing manager PMI index system, which uses non-quantitative questionnaire data to calculate and publish the PMI index value of the previous month at the beginning of each month. To reflect the macroeconomic development situation. In this context, based on the previous research basis, this paper adopts PMI-1N (PMI-1U), the excess reserve ratio, the legal reserve ratio, the import amount, the export amount, the average exchange rate. 36 objective quantitative data, such as cash M0, which are circulating outside the banking system, are regarded as possible influential variables in the PMI-index system. At the same time, a non-parametric method for the elimination of variables under mixed data is proposed by using Li and Racine 2004). To determine the variables related to PMI index, duplicate the determinant variables of PMI index, and further, construct the complete model of semi-parametric time-varying coefficient, path model, and carry on the joint effect to the selected variables. Finally, the model is extended to predict the PMI index, thus the PMI index is pushed from descriptive statistical index to inferential statistical index, which fills the gap in the previous literature. The specific work and conclusions are as follows: firstly, the determinant variables of PMI index are simulated by selecting the correlation and linearity of variables by the non-parametric method of variable elimination under mixed data. It was found that the variables with linear relationship with PMI index were: PMI-1 / 1, export / export value / value added / industrial / industrial value added / GZN, and the nonlinear relationships were as follows: stock transaction value / stock turnover, public finance revenue / expenditure / GGCEC / GGCEA, tax revenue / tax / tax revenue / expenditure balance The exchange rate is high, and the demand deposit rate is high. Secondly, through the empirical analysis of complete model and path model of semi-parametric time-varying coefficient, it is found that the combined action of six nonlinear variables will make the model have a more obvious ability to explain PMI index. At the same time, it is found that the influence of each nonlinear variable on PMI index is different. The nonlinear variable with positive influence on PMI index fitting is the one with no negative effect. The positive and negative effects of SRL PL HL on PMI index fitting results are not obvious. Finally, the coefficient of variation is constructed, and by comparison, it is found that the influence of each nonlinear variable on the fluctuation of PMI index is in turn: 1. Finally, through the empirical analysis of the semi-parametric time-varying coefficient prediction model, it not only shows that the variables selected by the non-parametric variables can still explain the PMI index in the future economic operation, but also, A better forecasting model is provided, which provides a powerful basis for enterprises, financial institutions and governments to judge the economic situation and formulate development plans.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F224;F124

【参考文献】