基于累积和模型的中国上市公司财务危机预警研究

发布时间：2018-06-03 13:06

本文选题：财务危机 + 累积和模型　；参考：《浙江工商大学》2017年硕士论文

【摘要】：近年来,上市公司由于财务危机而陷入经营困境,最终被特殊处理的例子已经屡见不鲜。财务危机的出现存在多方面的原因,内部原因有经营效率低下、盈利能力孱弱、为不良关联方担保、公司内部治理结构不完善、资本结构不合理、企业家诚信低下和现金流短缺等问题。外部原因有经济环境变化、政策和国家法规挑战和汇率变动等原因。在大多数情况下,造成上市公司出现财务危机的原因不是单一因素造成的,而是多个因素一起作用的结果。国内外的许多学者都对财务危机预警进行了研究,并建立了相应的预警模型。但是,以往大多数预警模型如多变量判别模型和条件概率模型都是以公司的静态横截面数据作为研究对象,而没有考虑财务指标所具有的时间序列特点。事实上,企业财务危机的发生是逐步演变的,传统的财务预警模型无法反映财务状况的动态特征。并且,传统模型一般采用单期年度会计数据,使用单期年度会计数据无法反映以往公司财务状况的相关信息和演变趋势。并且,使用单期年度数据容易将以往财务表现良好而仅仅单期财务状况不良的公司判断为危机公司。基于以上原因,本文选择能将企业财务状况进行"累积"的累积和模型(CUSUM Model)。本文选取上市公司中的制造业公司,选取2014-2016年的上市公司的季度数据作为研究对象,界定因为财务原因被实施特殊处理的公司为财务危机公司样本。当危机公司和财务健康公司样本的比例相同时,有可能会高估模型的预测能力。为避免高估预测正确率,选择危机公司和财务健康公司的比例为1:3。由于累积和模型为动态模型,本文选择ST公司被特殊处理前20个季度数据,而配比的健康公司选取同样的时间跨度。根据偿债能力、盈利能力、营运能力、成长能力、现金能力和资本结构的角度选取备选财务指标,使用Mann-Whieney U检验和相关性检验对财务指标进行了筛选,以最能反映危机公司和财务健康公司的流动比率、净资产收益率、总资产增长率和营运资本周转率这四个指标为研究变量。首先对其进行财务指标的数据平稳性性检验,然后进行累积和模型建模。为了进一步验证累积和模型的准确性,本文使用在财务预警研究中运用广泛的Logistic模型进行对照。使用同一训练样本进行Logistic模型的建模,同时累积和模型和Logistic模型两个模型均使用检验样本对模型进行准确性检验,以保证模型的适用性。将两个模型样本被ST前T-1季度、T-3季度、T-5季度、T-7季度、T-9季度、T-11季度的预测率进行对比,以及运用ROC曲线对两模型效果进行对比分析。实证结果显示,累积和模型考虑了公司财务状况的演进的动态过程,具有良好的预测能力,能提早预测财务状况的异常。Logistic近期预测能力比累积和模型更高,而远期预测能力累积和模型更优秀。
[Abstract]:In recent years, it is common for listed companies to get into trouble because of financial crisis. There are many reasons for the emergence of financial crisis, such as low operating efficiency, weak profitability, guarantee for bad related parties, imperfect internal governance structure and unreasonable capital structure. Entrepreneur integrity low and cash flow shortage and other problems. External reasons include changes in the economic environment, policy and national regulatory challenges and exchange rate changes. In most cases, the financial crisis of listed companies is caused not by a single factor, but by a combination of several factors. Many scholars at home and abroad have carried on the research on the financial crisis early warning, and have established the corresponding early warning model. However most of the early warning models such as multivariate discriminant model and conditional probability model take the static cross-sectional data of the company as the research object without considering the time series characteristics of the financial indicators. In fact, the occurrence of enterprise financial crisis is gradually evolving, the traditional financial early-warning model can not reflect the dynamic characteristics of the financial situation. In addition, the traditional model generally uses single-period annual accounting data, and the use of single-period annual accounting data can not reflect the related information and evolution trend of the past company's financial situation. Moreover, it is easy to judge a company with good financial performance but only a poor financial position as a crisis company by using single period annual data. Based on the above reasons, this paper chooses the CUSUM Model which can "accumulate" the enterprise's financial situation. This paper selects the manufacturing companies in the listed companies, selects the quarterly data of the listed companies from 2014-2016 as the research object, and defines the companies with special treatment for financial reasons as the financial crisis companies sample. When the ratio of crisis companies to financial health companies is the same, it is possible to overestimate the predictive power of the model. To avoid overestimating the accuracy of forecasts, the ratio of crisis companies to financial health companies was 1: 3. Because the cumulative sum model is a dynamic model, the first 20 quarters of St company are specially processed, while the matched health company chooses the same time span. According to the perspectives of solvency, profitability, operating capacity, growth capacity, cash ability and capital structure, the alternative financial indicators are selected, and Mann-Whieney U test and correlation test are used to screen the financial indicators. The four indexes, which can best reflect the liquidity ratio of crisis companies and financial health companies, the rate of return on net assets, the growth rate of total assets and the turnover rate of working capital are taken as the study variables. Firstly, the data stability of the financial index is tested, and then the cumulative sum model is built. In order to further verify the accuracy of the cumulative sum model, this paper uses the widely used Logistic model in financial early warning research for comparison. The Logistic model is modeled with the same training sample, and both the cumulative model and the Logistic model are used to verify the veracity of the model in order to ensure the applicability of the model. The two model samples were compared with the forecast rate of the T-7 quarter and the T-9 quarter of the T-5 quarter, and the results of the two models were compared and analyzed by using the ROC curve in the first quarter of the St T-1 quarter and the T-3 quarter of the T-5 quarter. The empirical results show that the cumulative sum model takes into account the dynamic process of the evolution of the company's financial situation and has a good forecasting ability. The short-term forecasting ability of Logistic is higher than that of the cumulative sum model. And the long-term forecast ability accumulates and the model is better.
【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】：F275;F832.51

【相似文献】