Value-at-risk has been broadly used in practice; however, it has some weaknesses. The most serious shortcoming is that it neglects risk exceeding the VaR value. Conditional Value-at-Risk (CVaR), the expected value of risk beyond the VaR, taking the whole loss distribution into account, considering the frequency and size of extreme events at the same time, will serve as a more suitable risk measure. In this article we investigate the differences from using VaR and CVaR as the risk measure. We conduct empirical study on the returns of eight industry indexes and the loss ratios of nine non-life insurance product lines. Some discoveries can be concluded as follows: 1. For the equity returns of industry indexes, the rankings of VaR and CVaR display divergent order in the long-term study period. It reveals that the most risky industry we suppose originally based on using VaR as the risk measure varies when we using CVaR as the risk measure instead. Furthermore, many loss distributions of industry indexes exhibit fat-tailed characteristic; this effect can be observed more obviously under CVaR computation. 2. In the short-run empirical analysis in returns of industry indexes, Based on the daily data, we find out the rankings of VaR and CVaR are similar, but if we convert daily computing frequency to weekly one, the rankings turn into different again. 3. Regarding the loss ratios of non-life insurance product lines, the differences “CVaR-VaR” based on historical simulation vary widely among the product lines. It implies that the loss distributions of non-life insurance may not be fat-tailed but have an extreme event, which produces huge loss with small probability. If we only use VaR to measure the risk, we would overlook this event. We suggest using CVaR as the risk measure, at least as a supplementary tool besides VaR to establish more comprehensive risk management.