اندازه گيري ريسك در چهارچوب ارزش در معرض خطر (VaR): كاربردي از مدل هاي GARCH Levy (مطالعه بخش صنايع شيميايي در بورس اوراق بهادار تهران)

Risk Measurement in Value at Risk (VaR): Application of Levy GARCH models (Study of Chemical industries in Tehran Stock Exchange)

ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ اﯾﻨﮑﻪ ﺳﺮﻣﺎﯾﻪﮔﺬاري در ﺑﻮرس اوراق ﺑﻬﺎدار ﺑﺎ رﯾﺴﮏ ﻫﻤﺮاه اﺳﺖ، ﺑﻨﺎﺑﺮاﯾﻦ اﻧﺪازهﮔﯿﺮي آن از ﻣﻬﻢﺗﺮﯾﻦ ﻣﻮﺿﻮﻋﺎت ﻧﺰد ﺳﺮﻣﺎﯾﻪﮔﺬاران ﻣﯽﺑﺎﺷﺪ. ﭘﮋوﻫﺶ ﺣﺎﺿﺮ ﺑﻪ اﻧﺪازهﮔﯿﺮي رﯾﺴﮏ ﺑﺎ ﻣﻌﯿﺎر ﺳﻨﺠﺶ ارزش در ﻣﻌﺮض ﺧﻄﺮ ﻣﯽﭘﺮدازد. در اﯾﻦ ﻣﻄﺎﻟﻌﻪ ارزش در ﻣﻌﺮض رﯾﺴﮏ، ﺑﺎ اﺳﺘﻔﺎده از ﻣﺪلﻫﺎي APARCH ،GARCH و GJR ﺑﺎ ﺗﻮزﯾﻊﻫﺎي ﻧﺮﻣﺎل، ﺗﯽ اﺳﺘﯿﻮدﻧﺖ، ﺗﯽ اﺳﺘﯿﻮدﻧﺖ ﭼﻮﻟﻪ و ﻟﻮي )ﺷﺎﻣﻞ ﺗﻮزﯾﻊ ﻣﻌﮑﻮس ﮔﺎوﺳﯽ ﻧﺮﻣﺎل )NIG( و ﺗﻮزﯾﻊ ﻫﺬﻟﻮﻟﯽ ﺗﻌﻤﯿﻢ ﯾﺎﻓﺘﻪ )GHyp(( ﺗﺨﻤﯿﻦ زده ﺷﺪه اﺳﺖ. در اﯾﻦ ﺗﺤﻘﯿﻖ رﯾﺴﮏ ﺷﺎﺧﺺ ﮐﻞ و ﺷﺎﺧﺺ ﺻﻨﺎﯾﻊ ﺷﯿﻤﯿﺎﯾﯽ ﺑﻮرس اوراق ﺑﻬﺎدار ﺗﻬﺮان در دوره زﻣﺎﻧﯽ 1392/01/05 ﺗﺎ 1398/12/28 ﻣﻮرد ارزﯾﺎﺑﯽ ﻗﺮار ﮔﺮﻓﺘﻪ اﺳﺖ. ﻧﺘﺎﯾﺞ ﺗﺤﻘﯿﻖ ﻧﺸﺎن ﻣﯽدﻫﺪ ﮐﻪ ﻣﺪل ﮔﺎرچ ﺑﺎ ﺗﻮزﯾﻊ ﻟﻮي و ﻣﺪل GJR ﺑﺎ ﺗﻮزﯾﻊ ﻟﻮي و Skewed-t در ﺑﯿﻦ ﺳﺎﯾﺮ ﻣﺪلﻫﺎ از دﻗﺖ ﺑﯿﺸﺘﺮي ﺑﺮﺧﻮردار اﺳﺖ. ﻟﺬا ﭘﯿﺸﻨﻬﺎد ﻣﯽﮔﺮدد ﮐﻪ ﻣﺪﯾﺮان ﭘﺮﺗﻔﻮي در ﺻﻨﺎﯾﻊ ﻣﺨﺘﻠﻒ ﺑﺎ اﺳﺘﻔﺎده از ﻣﺪلﻫﺎي ﻓﻮق ﺑﻪ ﺻﻮرت روزاﻧﻪ، ﻫﻔﺘﮕﯽ و ﯾﺎ ﻣﺎﻫﺎﻧﻪ، ﺣﺪاﮐﺜﺮ زﯾﺎن ﻣﺤﺘﻤﻞ ﭘﺮﺗﻔﻮي ﻣﻮﺟﻮد ﺧﻮد را ﺑﺮآورد ﻧﻤﻮده و اﻗﺪاﻣﺎت ﻣﻘﺘﻀﯽ را ﺑﺮاي ﻣﺼﻮنﺳﺎزي ﭘﺮﺗﻔﻮي از ﭼﻨﯿﻦ زﯾﺎنﻫﺎﯾﯽ اﻧﺠﺎم دﻫﻨﺪ.

Given that investing in the stock market is associated with risk, measuring it is one of the most important issues for investors. The present study measures risk measurement by the measure of risk. In this study, the value at risk, using the GARCH, APARCH and GJR models with normal distributions, T-stents, T-stents, strings and strings, including string distributions; the reverse distribution of normal GIG (NIG) and generalized hyperbolic distribution (GHyp) is estimated. In this study, to measure risk, the efficiency of Tehran Stock Exchange index in chemical industries and total index has been used. The time period in this study includes a seven-year period with a daily frequency during the period of 05/01/1392 to 28/12/1398. The results showed that the Garc models were more accurate with the Levy distribution, and among the Garc models, the GJR model was more accurate, considering the Lou distribution and the Skewed-t distribution used among the other models.