Manajemen Bisnis
Transcription
Manajemen Bisnis
Modelling Financial Time Series and Its Application as an Alert System by: Dr. Heri Kuswanto, M.Si Department of Statistics- Laboratory of Economic, Financial Statistics and Actuarial Science Institut Teknologi Sepuluh Nopember (ITS)- Indonesia Presented at the Research Methodology Seminar- Prince Songkla University-Thailand 12 May 2016 www.its.ac.id Brief. Introduction Education: • 2010 • 2009 Germany • 2005 • 2003 Appointments: • 2016 • 2016 • 2011-2015 • 2013- now :Postdoctoral Research Associate at the Laval University Canada :PhD at the Institute of Statistics, School of Economics and Management- Leibniz Hannover Univ., :MSc. in Statistics- ITS Indonesia :BSc. in Statistics-ITS Indonesia : Head of Postgraduate Study Program in Statistics-ITS Indonesia : Coord. of Climate Change-Expertise at the Research Center for Earth, Disaster and Climate Change : Deputy Head of the International Relation and Coopertaion Office-ITS : Lecturer and Researcher at the Department of Statistics-ITS Research Interest : Financial Econometrics, Time series forecasting, (Extreme) weather and climate events, Risk Modelling www.its.ac.id Outline of the Presentation 1. Financial Time Series Models : Definition and scope 2. Developments in Financial Time Series Modelling 3. Some Applications for Indonesia’s case • Alert System for Crisis Management Protocol • Market Surveillence System www.its.ac.id What Experts says? http://www.stern.nyu.edu/rengle/ www.its.ac.id Another expert www3.stat.sinica.edu.tw/statistica/J17N1/editorial_2.pdf Journal of Business and Economic Statistics, Journal of Financial Econometrics, Journal of Applied Econometrics and the Econometrics Journal Etc. www.its.ac.id Scope of Financial Time Series Financial Theory Time Series www.its.ac.id Financial time series concerns with Theory and Practice of Asset Valuation over time What modelling financial time series for? • • • • monitor price behaviour to understand the probable development of the prices in the future Many traders deal with the risks associated with changes in prices. Forecasts of future standard deviations can provide up-to-date indications of risk, which might be used to avoid unacceptable risks perhaps by hedging. Aas and Dimakos, Statistical modelling of financial time series: An Introduction, Norwegian Computing Center 2014. www.its.ac.id Stylize Fact about statistical properties of financial time series • Excess Volatility • Heavy Tails • Absence of autocorrelation in returns • Volatility clustering • Volatility autocorrelation www.its.ac.id Scope of financial time series modelling • Mean response mean equation ARMA • Conditional variance price of stock depends on volatility of stock returns with high excess kurtosis and serially uncorrelated GARCH The development on financial time series modelling grows extremely fast new facts in the data www.its.ac.id Assumption of Classical financial time series models (log returns) • Normality assumption www.its.ac.id Acta oeconomica pragensia 9: (4), 2001. FINANCIAL TIME SERIES AND THEIR FEATURES Josef ARLT, Markéta ARLTOVÁ* Linearity assumption www.its.ac.id Data availability www.its.ac.id Data Analysis Support www.its.ac.id Available package in R www.its.ac.id Applications www.its.ac.id Study on The Factors Influencing the Determination of Level and Indicators of Crisis Management Protocol (CMP) in Indonesia Joint work with Prof. Nur Iriawan (ITS), Dr Suhartono (ITS), Dr. Brodjol SU (ITS) and The Ministry of Finance-Indonesia www.its.ac.id Introduction : background www.its.ac.id Why CMP is necessary? Global crisis in Europe and US gives significant impact to the stability of financial system in Asia. Crisis Management Protocol (CMP) is required to minimize the impact of global financial crisis To give direction towards actions has to be carried out when there is movement on the domestic financial market as a result of crisis in global financial market. As an early warning system towards possibility of crisis in domestic financial market To recommend a standard procedure the to be carried out by the management of SUN* (Government Securities/GS) to set up policy to encounter crisis of GS market. *Surat Utang Negara (SUN) is a securities in the form of debt certificate issued by the Government of Indonesia in the form of State Bonds with 12 (twelve) months tenor and regular coupon payment with a minimum transaction of Rp. 250,000,000,-.Type of SUN: Fixed Rate (FR Series), Variable Rate (VR Series) Determination of CMP level needs variabes which can be a trigger significantly Econometrics Modeling www.its.ac.id Reseach Methodology: Data (daily series) 1. 2. 3. 4. 5. Yield of GS benchmark series response variable IDX Composite CDS (Credit Default Swap) 5Y dan 10Y Exchange Rate (IDR/USD) Foreign Asset This research is done by analyzing three differents data periods, i.e. starting from 2008 (data1), 2009 (data2) and September 2011 (data3) www.its.ac.id Econometric models • Transfer function : used to know the factors/ secondary indicators which significantly influence the yield movement, as well as the time delay • ARIMAX : used to model the mean level and identify outliers • ARMA-GARCH: used to model the movement of mean and volatility of primary and secondary indicators as the basis for determination of CMP level www.its.ac.id Basic Statistical models • ARIMA Model (Wei, 2006) • The Box-Jenkins approach typically comprises four parts: - Identification of the model - Estimation, usually OLS - Diagnostic checking (mostly for autocorrelation) - Forecasting www.its.ac.id Autoregressive Conditional Heteroskedasticity (ARCH)-Engle 1982 - used to model the volatility - case of nonhomogoneous variance of the residual model GARCH (Bollerslev(1986)) q 2 t i i 1 www.its.ac.id p 2 t i j j 1 2 t j ARIMAX : Determination of CMP level for the primary indicators • • This study recommends to use sigma of 0.07 as the threshod for primary indicator, obtained from the standard error estimate of 5Y tenor with alpha 0.1% Compared to the other values, threshold 0.07 listed as the hihgest among the models, which means also that the sensitivity is good. www.its.ac.id No. Alpha SIGMA limit 1. 1% 2.57583 2. 0.5% 2.80703 3. 0.1% 3.29053 4. 0.01% 3.89059 5. 0.27% 3 6. 0.00633% 4 Schwart, 1931 ARIMAX Models www.its.ac.id Defined levels (using 3sigma, 4.5sigma and 6 sigma): o o o o • On primary indicators: Level normal Level alert Level pre-crisis Level crisis : < 21 bps : 22 bps – 31.5 bps : 32 bps – 42 bps : > 42 bps The levels above are significantly different with the levels specified by Ministry. Furthermore, validation to data 1 shows that the ARIMAX levels are very reasonable. www.its.ac.id Backtesting for levels developed by data3 (Sept 2011) www.its.ac.id Secondary indicators Level Indicator Alert Pre-Crisis Crisis Decreasing Foreign Asset 1.2% per day or 2.463T IDR per day 1,8% per day or 3.6945T IDR per day 2.4 % per day or 4.926T IDR per day Decreasing IDX Index 3.228% per day or 123.63 per day 4.842% per day or 185.445 per day 6.456% per day or 247.26 per day Decreasing Exchange Rate 1.287% per day or 94.548 per day 1,9305% per day or 141.822 per day 2.574% per day or 189.096 per day Decreasing CDS 10Y 9.414% per day or 20.814 per day 14.121% per day or 31.221 per day 18.828% per day or 41.628 per day Decreasing CDS 5Y 9.945% per day or 16.473 per day 14.917% per day or 24.709 per day 19.89% per day or 32.946 per day www.its.ac.id Indicator Yield IDX Composite Exchange Rate Foreign Asset www.its.ac.id Weight Indicato r 0.7 0.3 Level of the weight normal 0.25 1 1 0.493 0.38 0.127 1 1 1 alert 0.25 2 pre crisis 0.25 3 crisis 0.25 4 2 2 2 3 3 3 4 4 4 Composite Index Indicator Yield IDX Composite Exchange Rate Foreign Asset Index Yield IDX Composite Exchange Rate Foreign Asset Index Yield IDX Composite Exchange Rate Foreign Asset Index level normal normal normal normal 0.25 alert alert alert alert 0.5 precrisis preciriss precrisis precrisis 0.75 www.its.ac.id level normal normal normal alert 0.2595 alert alert alert pre crisis 0.509525 preciriss precrisis precrisis crisis 0.759525 level normal normal alert alert 0.288025 alert alert precrisis precrisis 0.538 preciriss precrisis crisis crisis 0.788025 level normal alert alert alert 0.325 alert Precrisis Precrisis Precrisis 0.575 precresis crisis crisis crisis 0.825 level alert normal normal normal 0.425 precrisis alert alert alert 0.675 crisis precrisis precrisis precrisis 0.925 level alert alert normal normal 0.461975 precrisis precrisis alert alert 0.711975 crisis crisis precrisis precrisis 0.961975 level alert alert alert Normal 0.490475 Precrisis Precrisis Precrisis Alert 0.740475 Crisis Crisis Crisis Precrisis 0.990475 Final Composite Index of CMP Condition I II III IV www.its.ac.id Index 0.25 ≤ index < 0.325 0.325 ≤ index < 0.575 0.575 ≤ index < 0.825 Index ≥ 0.825 level normal alert Pre-crisis Crisis Original Series Percentage of change volatility www.its.ac.id Development of Market Surveillence System Indonesia Securities Investment Protecton Funds (Indonesia SIPF) www.its.ac.id Goals of Market Surveillence 1. 2. 3. 4. To ensure that the members run ther activities normally To detect any possibility of market abuse, To protect investors, To ensure that the market is normal, efficient and transparent to reduce systematic risk Alert system has to be able to detect unusual events www.its.ac.id Component of the Survival System 1. Feature to perfomr the investor profile using some relevant statistics Reporting How : Benchmarking, intensive discussion 2. Models which are able to perform and predict the fluctuation of changing in assets Alert system How: Alert system to detect the mean level and volatility, if they are still on a normal postion, or alert. www.its.ac.id Illustration of Alert System Financial time series model is used to costruct the band www.its.ac.id ARIMA and GARCH • ARIMA ARIMAX (ARIMA with outlier) Introduced by Box & Jenkins (1976): Identification, estimation, diagnostic checking, forecasting AutoARIMA dari package “forecast” di R Minimize user specification • Outlier Detection Package “tsoutliers” to detect Innovative Outlier (IO), Additive Outlier (AO), Level Shift (LS) dan Temporary Change (TC) www.its.ac.id Thank you www.its.ac.id