当前，高杠杆被普遍认为是导致金融危机发生的重要诱因。国内外多认为中国杠杆率过高，系统性金融风险不断加大，有研究甚至认为我国很快将爆发金融危机，因此应该大力推进“去杠杆”。也有学者认为我国有许多特殊国情，即使杠杆率较其他国家为高，也不会导致太大的经济和金融风险，应避免政策反应过度。对这个问题的判断，在很大程度上将影响今后几年的宏观政策的走势和结构性改革的方向与步伐。 本文应用定量研究来考察杠杆率与金融危机概率之间的关系，采用跨国数据建立包括杠杆率在内的诸多经济变量预测金融危机概率的多组模型。本文的核心解释变量是杠杆率，采用两类指标进行衡量，即M2与GDP的比值以及信贷总量与GDP的比值。对比以往文献，本文创新性地同时将杠杆率和杠杆率的变动幅度作为核心解释变量同时引入模型中。实证分析从三方面入手，研究杠杆率与银行危机及IMF援助危机、杠杆率与货币危机以及杠杆率与基于综合考察的金融危机之间的关系。针对不同数据类型，采用Logit模型与横截面回归进行建模，并针对因变量、核心变量和其他解释变量的测量误差进行稳健性检验，以及考虑了遗漏变量。 本文的实证分析显示：（1）杠杆率是影响一个国家发生金融危机的重要因素，但它不是唯一的决定危机概率的因素。除了杠杆率之外，储蓄率、资本项目管制程度、外汇储备、外债水平、实际利率等都会影响危机概率。（2）对中国来说，虽然杠杆率较高，但较高的储蓄率和外汇储备及较低的外债水平帮助降低了危机概率。（3）杠杆率的上升幅度比杠杆率本身对预测概率的影响更大。本文的政策含义在于，相对比杠杆率的水平值而言，应更加关注杠杆率在短时间内的过快上升所可能带来的金融危机概率的显著上升。

At present, high leverage is generally considered to be the leading cause of financial crisis. It is believed that China's leverage ratio is too high and the systemic financial risks are increasing. There are studies even arguing that China will soon break out the financial crisis. Therefore, the "de leveraging" should be vigorously promoted. Some scholars believe that there are many special national conditions in our country. Even if the leverage ratio is higher than other countries, it will not lead to too much economic and financial risks, and the overreaction of policy should be avoided. The judgement of this problem will largely affect the trend of macro policy in the next few years and the direction of structural reform. This paper uses quantitative research to investigate the relationship between leverage ratio and the probability of financial crisis, and uses cross-country level data to establish a multigroup model for predicting the probability of financial crisis, including a number of economic variables. The core explanatory variable of this paper is leverage ratio, which is measured by two indicators, that is, the ratio of M2 to GDP and the ratio of total credit to GDP. Compared with the previous literature, this paper innovatively introduces the rate of leverage and leverage as a core explanatory variable in the model. This article defines three kinds of crises, bank crisis and IMF aid crisis, currency crises, and financial crisis based on comprehensive investigation. The Logit model and cross section regression are used to model the different data types, and the robustness test for the measurement errors of the dependent variables, the core variables and other explanatory variables is carried out, what’s more, the missing variables are considered. The empirical analysis of this paper shows that: (1) Leverage is an important factor affecting a country's financial crisis, but it is not the only factor in determining the probability of a crisis. In addition to the leverage ratio, the saving rate, the degree of capital account control, the foreign exchange reserve, the foreign debt level and the real interest rate will all affect the probability of the crisis. (2) For China, although the leverage ratio is high, the higher savings rate, foreign exchange reserve and the lower foreign debt level help to reduce the crisis probability. (3) The leverage ratio has greater impact on prediction probability than leverage itself. The policy implication of this paper is that, relative to the level of leverage ratio, we should pay more attention to the significant increase in the probability of the financial crisis that may be brought by the rapid rise of leverage ratio in a short time.