Risk is the one of the core concept of finance industry. The subprime mortgage crisis that happened in the United Stats since the end of 2007 queried the ability and the methods of the more efficient risk management. The traditional risk management that using variance to measure risk has been unable to play a part in present capital market. In other words, basing on a risk management sentiment, the occurrence of crisis required new approaches capable of precisely evaluating the systematic risk under the background that the crisis accelerated volatility of the financial markets.

Measures of risk that take into consideration of huge losses need also be used to measure the market risk of firms. Theoretically, firm characteristics express their potentially in estimating the risk. On the other hand, risk measures are generally obtained by quantile-based approaches, and among these the most commonly applied is the Value-at-Risk called VaR. This method refers to market fluctuations that the portfolio of a financial security or the highest possible loss (Jorion 1997). Hence VaR model is one of the most available approaches for measuring risk which serves as a valid method to determine risk in financial market. It has become the worldwide standard for anticipating risk in all kinds of financial institutions.

Generally speaking, the reliance on VaR crossing the financial market has been supported by a number of literatures. For example, in the study by Dias(2013), she explores the role of market capitalization, one of the firms’ variables, in the estimation of VaR. Dias observes that taking market capitalization into account produces better VaR estimates especially during the periods of crisis. This finding support the assumption that the firms variables will play a part in measuring market risk via the VaR method.

The VaR forecasting performance of this study is analysed for the firms’ characteristics in determining the market risk. Additionally, the research will observe how VaR measures are affected during the period since the occurrence of crises started. For this purpose, We analyse the numerical monthly data on each of these variables for a set of 150 firms randomly chosen from three stock markets: Japan, UK and US in the period dating from 1 January 2000 to 1 October 2014. The study will specifically consider the difference via considering respectively the estimation of VaR before crisis periods, after crisis periods, and the full sample hence ignoring crises. The term of crisis is determined from 1 December 2007 to 1 June 2009. The study identify firms’ characteristics associated with the market risk by considering average firm return, market capitalization (stock market value), firm’s region, and price-to-book ratio.

The rest of the essay is organized as follows: the next section will present the models and methodology. Then we apply the graphic analysis of the fluctuation of mean return in section 2. In section 3 regression analysis of the relationships between the firms variables and Var model. The final section will summarizes the main outcomes of the study.

(1) Models and methodology
Nowadays, VaR, also called value at risk, has been regarded by quiet a few financial managers as the most valid and simplest approach to anticipate losses. VaR is the maximum loss basing on a certain probability that the financial assets may face in a period of future time. In the field of financial risk management, VaR is a widely applied tool for risk measuring of the risk of loss on a portfolio, which is composed by financial securities. VaR has been treated as a fundamental element for measuring risk because of its abstract simplicity and accuracy of estimating risk at a rational computational cost.

In the market, according to a certain confidence, the VaR value is the largest possible loss statistically in stock markets. While the practical losses could exceed VaR value, by the sampling distribution theory, the possibility of the