Approaching the property of shares’ volatility by a use of ARCH effect has significant influence from the point of view of risk measurement. In the case of shares, which are characterized by the existence of the ARCH effect, the use of the ARCH-type models results in changes in VaR even by 49%. Models with the additional conditional variance function can easily detect current shares’ behaviour and reflect it in the risk level, thus it differentiate shares in terms of risk more. Thus, an ARCH effect is worthy of use for shares with strong ARCH effect and it does not make the result for shares without ARCH effect worse.
The smallest VaR is usually indicated by the variance-covariance method. The Monte Carlo simulation method and the historical simulation method are characterized by the biggest VaRs. No matter if VaRs for banks indicated as the most/medium/the least risky shares differ much or not, the rankings of banking sector shares (from the most to the least risky) may differ significantly. Specifically, the historical simulation method gives rankings that differ the most from results calculated with the use of other methods. Additionally, the historical simulation method is the most fluctuating one.
Similarity of the rankings resulted by variance-covariance and the Monte Carlo simulation methods, possibility of using ARCH effect in these methods, and the fact that the historical simulation method produces the most instable results, indicate the variance-covariance and the Monte Carlo simulation
methods as more believable. In addition to this, because of the model describing a mechanism of return rates on the market, Monte Carlo simulation method can detect current market behavior better. Ability of ARCH type models to detect current volatility of shares enhance to implement the ARCH effect to risk measurement. Taking into consideration all the conclusions above, the Monte Carlo simulation with the ARCH effect is recommended to use in practice.
Length of the observation period significantly influences the results. It is not only connected with the length of the observation period but also with the behavior of return rates of shares in these periods. Thus, more current observation periods are more informative because it considers actual market behaviour. On the other hand, longer time horizon assures that there is more data for estimation. The choice of how long the observation period should be up to the person who conducts an analysis. If the market was stable, a longer observation period may be used. But if trends changed, it should be approached by the use of an appropriate time horizon which describes actual market behaviour. In the research, the current tendency is well represented by a 2 month observation period, which is characterized by the biggest decrease of shares’ values. Thus, the use of this time horizon gives the highest VaRs.
Inappropriate specification of the method, such as not enough replications in the Monte Carlo simulation method, may result in overestimation or underestimation of risk, which may result in an unprofitable investment strategy.
Diversification of risk, by an investment in a portfolio of shares, is profitable even if shares that a portfolio consists of are from the same economy sector.
Although the levels of VaRs differ depending on the method applied, the length of an observation period, a significance level, etc., clear tendencies may be noticed. These tendencies allow to indicate the most and the least risky banking sector shares. An investor may assume that shares indicated, by the presented methods, as the least risky ones are actually safe. Another solution is to assume that shares indicated by the presented methods as the most risky ones are unprofitable and to avoid investments in these shares. It may be a starting point for further investor’s research, thanks to which the most risky shares may be excluded from further analysis.