Samac, Martina (2017) VaR i ES u vrednovanju valutnog rizika : završni rad. Other, Faculty of Science > Department of Mathematics.

PDF
Language: Croatian Download (722kB)  Preview 
Abstract
Risk can be defined as exposure to the uncertainty and the possibility of incurring losses due to negative deviations from expected outcomes. The exposure in the context of insurance and banking arises from any transaction or business decisions containing uncertainty of results of operations. Institution recognizes risk taking as a potential instrument for generating revenue, with overall risks not endangering its existence. Market risks are defined as potential losses that external variables have on the assets, liabilities and offbalance sheet positions of the institutions, which are caused by price or negative developments in the financial markets. Similarly, foreign exchange risk is defined as the risk of loss arising from changes in currency exchange rate and / or changes in gold prices. When taking risk it is necessary to establish an adequate system of further risk management which includes both quantitative and qualitative approaches to manage and control risk. Many financial institutions use various statistical methods to manage and control risk. The most common method for managing market risk used by financial institutions today is the Value at Risk method. VaR concept developed in the late 80s of the last century after the great stock market crash in 1987. This was the first great financial crisis that standard statistical models could not predict and questioned the basic tenets of the previous risk management system which resulted in the development of new statistical methods that adequately quantify possible losses in the future. Although VaR methodology was developed primarily for managing market risks today is applied on the integrated approach in measuring credit and market risks together, and more recently on other types of risk such as liquidity and operational risk. VaR method calculates the expected maximum permissible loss over a period of time within the statistically defined acceptance areas (certain probability). The development of VaR risk measurement system clearly diversified three main methods of measuring VaR: historical simulation, variance / covariance method and Monte Carlo simulation. The historical simulation represents a nonparametric method of VaR estimation which predicts the risk in the near future using data from the recent past. Variance / covariance method of VaR calculation is based on the assumption that the distribution corresponds to a return of the theoretical distribution, such as the normal distribution taking into account the correlation between investment instruments. Applying these assumptions market risk VaR is calculated based on two basic parameters: the mean gains / losses (or rate of return) of the observed portfolio and standard deviation of the observed data. Monte Carlo simulation is very similar to the historical simulation, with the distinction that hypothetical changes in market factors are not exercised on the basis of past observed changes in market factors, but is randomly taken from the statistical distribution that adequately represents the actual statistical properties of changes in market factors. Although VaR methods are very popular and widely applicable, they carry certain deficiencies, primarily failure to satisfy the condition of subadditivity and the fact that VaR method ignores large, potentially catastrophic losses in the tail of the distribution. Therefore, many investors, including banks started using a new measure of risk called Expected shortfall method. EC method is designed to measure the risk of extreme losses and represents a kind of upgrade of VaR methodology because it calculates the total amount of losses when an adverse event occurs and the loss is greater than the calculated VaR. Unlike VaR, ES method quantifies the risk in the tail of the distribution and satisfies the condition of subadditivity. The use of VaR and ES methods in the evaluation of foreign exchange risk of all banks in the Republic of Croatia has shown that both, VaR and EC method give results with equally satisfactory precision and stability, but due to the condition of subadditivity, priority is given to ES method. Also, research has shown that historical simulation and method that assumes a normal distribution of data are both equally stabile and precise, while Monte Carlo method is less stable and precise due to the assumption of a uniform distribution of data. However, following the global financial crisis, the market in the Republic of Croatia is no longer so deeply developed and the question is whether the costs of developing such method for managing foreign exchange risk of banks are profitable.
Item Type:  Thesis (Other) 

Supervisor:  Huzak, Miljenko 
Date:  2017 
Number of Pages:  68 
Subjects:  NATURAL SCIENCES > Mathematics 
Additional Information:  Završni rad na poslijediplomskom specijalističkom studiju aktuarske matematike 
Divisions:  Faculty of Science > Department of Mathematics 
Depositing User:  Iva Prah 
Date Deposited:  04 May 2017 08:49 
Last Modified:  04 May 2017 08:53 
URI:  http://digre.pmf.unizg.hr/id/eprint/5488 
Actions (login required)
View Item 