Abstract: Because of the extremely unbalanced development of China's regional economy, some provinces have great influence on the overall financial security of China, making it very important for the provincial and national financial security to pay attention to the risks of the provincial financial system and to carry out early warning and supervision. Under the background of further opening up of the financial market, this paper builds the leading early warning indicator system and the financial risk stress index of provincial financial risks in China. Using the provincial monthly sample data of Zhejiang Province from January 2004 to December 2016, this paper conducts an empirical analysis of early warning of provincial financial risks by TAR Vector Autoregressive Model and Ologit Probability Model. Results show that：(1) Inflation rate and export growth rate have significant and negative effects on provincial financial risks, while the new bank loans divided by industrial added value, fixed asset investment growth rate, consumption growth rate and import and export growth rate have significant and positive effects; (2) Financial stress risk of the previous month has a significant and negative effect on that of the current month, which makes the financial stress risk of the current month has some “smoothing” effect; (3) Statistical tests show that Dynamic Logit Early Warning Model is obviously better than Static Logit Early Warning Model in the model's overall significance, goodness of fit, and the accuracy of the prediction; （4） The early warning statistical analysis results of Zhejiang province in the recent 12 months show that strict interval of early warning accuracy is 25%, but the main class interval of early warning accuracy is 50%, and it is relatively better than early warning effect of domestic and foreign models of the same type.

Key words: Ologit model , provincial financial risk , model warning , leading indicator

摘要： 我国区域经济发展程度极不平衡使得部分重点省域对国家总体金融安全影响程度显著而突出，关注我国省域金融系统风险并进行提前预警与监管防范对于省域与国家金融安全尤为重要。本文立足金融市场深入扩大对外开放趋势背景下，构建了我国省域金融风险先行预警指标体系和金融风险压力指数，并以浙江省2004年1月—2016年12月样本数据为对象，通过TAR门限自回归模型和Ologit概率模型对省域金融系统风险状况进行了实证预警分析，研究发现：（1）省域先行预警指标统计检验发现，通货膨胀率与出口增长率对省域金融风险水平呈现显著负相关效应，而新增信贷额/工业增加值、固定资产投资增长率、消费增长率与进出口增长率对省域金融风险水平呈现显著正相关效应；（2）动态概率预警模型检验发现，省域上一月份的金融风险压力值对下一月份的金融风险压力值呈现显著负向影响效应，并使下一月份的金融压力风险起到一定程度的“熨平”作用效应；（3）通过统计检验比较发现，动态Ologit概率预警模型无论是从模型整体显著性与拟合优度，还是预测的准确度都明显优于静态概率预警模型；（4）通过近12个月的样本外数据对模型预警的稳健性检验发现，Ologit概率预警模型金融风险预测准确程度在严格区制上为25%，但大类区制预警准确度为50%，相对优于国内外同类型的模型预警效果。

关键词: Ologit概率模型 , 省域金融风险 , 模型预警 , 先行指标