Tesis ini meneliti validitas pengukuran risiko nilai tukar IDR dengan USD, JPY dan SGD pada nilai ekstrim. Analisis ini dilakukan karena permodelan VaR dengan cara tradisional berdasarkan distribusi normal seperti metode Historical Simulation. Risk Matrics, EWMA dan GARCH, sering gagal dalam mencakup probabilitas untuk nilai ekstrim dari pergerakan valuta asing yang tidak diharapkan dan menghasilkan analisis risiko dengan error yang tinggi. VaR pada nilai ekstrim dianalisis untuk periode 1997-1998 dan 2003-2004 dengan metode Historical Simulation, GARCH dan Generalized Extreme Value Distribution. Hasil uji validitas menunjukkan bahwa metode Generalized Extreme Value Distribution valid untuk mengukur risiko nilai tukar saat nilai ekstrim dengan confidence level 99%, sedangkan metode tradisional berdasarkan distribusi normal tidak valid untuk confidence level 99%.

---

The purpose of this thesis is to reserch the validity of currency risk of IDR to USD, JPY, and SGD at extreme value. This analysis is carried out due to the failure of VaR model in traditional method based on normal distribution like Historical Simulation, Risk Matrics, EWMA, and GARCH to cover the probability at extreme value of un-expected movement in foreign exchange market and produce risk analysis with high errors. VaR at extreme value is calculated for two periods, 1997-1998 and 2003-2004 with methods of Historical Simulations, GARCH, and Generalized Extreme Value Distribution.

The result of validity test shows that Generalized Extreme Value Distribution is valid to measure currency risk at extreme with confidence level 99%, while traditional method with normal distribution are not valid for confidence level 99%.