GARCH Models and Asymmetric GARCH models

Slides:



Advertisements
Similar presentations
MIT Plan for the Session Questions? Complete some random topics Lecture on Design of Dynamic Systems (Signal / Response Systems) Recitation on HW#5?
Advertisements

Multivariate Cointegartion
Cointegration and Error Correction Models
Multiple Regression.
Dummy Dependent variable Models
Autocorrelation and Heteroskedasticity
Introduction Describe what panel data is and the reasons for using it in this format Assess the importance of fixed and random effects Examine the Hausman.
Volatility in Financial Time Series
Regression Analysis.
Things to do in Lecture 1 Outline basic concepts of causality
Applied Econometrics Second edition
1 Regression as Moment Structure. 2 Regression Equation Y =  X + v Observable Variables Y z = X Moment matrix  YY  YX  =  YX  XX Moment structure.
Brief introduction on Logistic Regression
Lecture 8 (Ch14) Advanced Panel Data Method
Xtreg and xtmixed: recap We have the standard regression model (here with only one x): but think that the data are clustered, and that the intercept (c.
Primbs, MS&E 345, Spring The Analysis of Volatility.
Forecasting JY Le Boudec 1. Contents 1.What is forecasting ? 2.Linear Regression 3.Avoiding Overfitting 4.Differencing 5.ARMA models 6.Sparse ARMA models.
CHAPTER 3 ECONOMETRICS x x x x x Chapter 2: Estimating the parameters of a linear regression model. Y i = b 1 + b 2 X i + e i Using OLS Chapter 3: Testing.
NEW MODELS FOR HIGH AND LOW FREQUENCY VOLATILITY Robert Engle NYU Salomon Center Derivatives Research Project Derivatives Research Project.
Specific to General Modelling The traditional approach to econometrics modelling was as follows: 1.Start with an equation based on economic theory. 2.Estimate.
Econ Prof. Buckles1 Multiple Regression Analysis y =  0 +  1 x 1 +  2 x  k x k + u 4. Further Issues.
Introduction to Logistic Regression. Simple linear regression Table 1 Age and systolic blood pressure (SBP) among 33 adult women.
Modelling volatility and correlation
12.3 Correcting for Serial Correlation w/ Strictly Exogenous Regressors The following autocorrelation correction requires all our regressors to be strictly.
Chapter 11 Multiple Regression.
Lecture 5 Curve fitting by iterative approaches MARINE QB III MARINE QB III Modelling Aquatic Rates In Natural Ecosystems BIOL471 © 2001 School of Biological.
Dealing with Heteroscedasticity In some cases an appropriate scaling of the data is the best way to deal with heteroscedasticity. For example, in the model.
Psych 524 Andrew Ainsworth Data Screening 2. Transformation allows for the correction of non-normality caused by skewness, kurtosis, or other problems.
Business Statistics - QBM117 Statistical inference for regression.
12 Autocorrelation Serial Correlation exists when errors are correlated across periods -One source of serial correlation is misspecification of the model.
Time-Varying Volatility and ARCH Models
Temperature correction of energy consumption time series Sumit Rahman, Methodology Advisory Service, Office for National Statistics.
Multiple Regression. In the previous section, we examined simple regression, which has just one independent variable on the right side of the equation.
2-1 MGMG 522 : Session #2 Learning to Use Regression Analysis & The Classical Model (Ch. 3 & 4)
ALISON BOWLING THE GENERAL LINEAR MODEL. ALTERNATIVE EXPRESSION OF THE MODEL.
Slide 1 DSCI 5340: Predictive Modeling and Business Forecasting Spring 2013 – Dr. Nick Evangelopoulos Lecture 8: Estimation & Diagnostic Checking in Box-Jenkins.
9-1 MGMG 522 : Session #9 Binary Regression (Ch. 13)
Estimation Kline Chapter 7 (skip , appendices)
Maximum Likelihood Estimation Methods of Economic Investigation Lecture 17.
ELEC 303 – Random Signals Lecture 18 – Classical Statistical Inference, Dr. Farinaz Koushanfar ECE Dept., Rice University Nov 4, 2010.
© Department of Statistics 2012 STATS 330 Lecture 20: Slide 1 Stats 330: Lecture 20.
Lecture 4: Statistics Review II Date: 9/5/02  Hypothesis tests: power  Estimation: likelihood, moment estimation, least square  Statistical properties.
Business Statistics: A Decision-Making Approach, 6e © 2005 Prentice-Hall, Inc. Chap 13-1 Introduction to Regression Analysis Regression analysis is used.
Robert Engle UCSD and NYU and Robert F. Engle, Econometric Services DYNAMIC CONDITIONAL CORRELATIONS.
NEW FRONTIERS FOR ARCH MODELS Prepared for Conference on Volatility Modeling and Forecasting Perth, Australia, September 2001 Robert Engle UCSD and NYU.
STAT 497 LECTURE NOTE 9 DIAGNOSTIC CHECKS 1. After identifying and estimating a time series model, the goodness-of-fit of the model and validity of the.
8-1 MGMG 522 : Session #8 Heteroskedasticity (Ch. 10)
Estimation Kline Chapter 7 (skip , appendices)
Logistic Regression Saed Sayad 1www.ismartsoft.com.
ALISON BOWLING MAXIMUM LIKELIHOOD. GENERAL LINEAR MODEL.
Review of statistical modeling and probability theory Alan Moses ML4bio.
ESTIMATION METHODS We know how to calculate confidence intervals for estimates of  and  2 Now, we need procedures to calculate  and  2, themselves.
Stochastic Error Functions I: Another Composed Error Lecture X.
Computacion Inteligente Least-Square Methods for System Identification.
The Probit Model Alexander Spermann University of Freiburg SS 2008.
MODELING VOLATILITY BY ARCH- GARCH MODELS 1. VARIANCE A time series is said to be heteroscedastic, if its variance changes over time, otherwise it is.
Analysis of financial data Anders Lundquist Spring 2010.
Statistics 350 Lecture 2. Today Last Day: Section Today: Section 1.6 Homework #1: Chapter 1 Problems (page 33-38): 2, 5, 6, 7, 22, 26, 33, 34,
Instructor: R. Makoto 1richard makoto UZ Econ313 Lecture notes.
Chapter 4. The Normality Assumption: CLassical Normal Linear Regression Model (CNLRM)
The Probit Model Alexander Spermann University of Freiburg SoSe 2009
Chapter 4: Basic Estimation Techniques
Estimating Volatilities and Correlations
Generalized regression techniques
Ch8 Time Series Modeling
Chapter 6: Autoregressive Integrated Moving Average (ARIMA) Models
ASSET PRICE VOLATILITY: THE ARCH AND GARCH MODELS
The regression model in matrix form
The Regression Model Suppose we wish to estimate the parameters of the following relationship: A common method is to choose parameters to minimise the.
Mathematical Foundations of BME Reza Shadmehr
Presentation transcript:

GARCH Models and Asymmetric GARCH models

VECM (Review) Cointegrating Eq: Error Correction: D(R1) D(R10)  (0.07657) [-12.8046] C  0.603495 Error Correction: D(R1) D(R10) CointEq1 -0.029996  0.015287  (0.01783)  (0.01140) [-1.68255] [ 1.34155] D(R1(-1))  0.273219 -0.028276  (0.06803)  (0.04348) [ 4.01619] [-0.65026] D(R1(-2)) -0.087596  0.025434  (0.06772)  (0.04328) [-1.29358] [ 0.58761] D(R10(-1))  0.370337  0.425735  (0.10747)  (0.06869) [ 3.44593] [ 6.19757] D(R10(-2)) -0.263587 -0.266142  (0.10796)  (0.06901) [-2.44152] [-3.85675] C -0.000459

Structure of today’s session Brief re-cap of last lecture Estimation of ARCH models Maximum likelihood estimation The Glosten Jaganathan and Runkle model which introduces asymmetric adjustment into the model. Revision

Problems: What value of q should be chosen? The number of lags, q, might be large. The non-negativity constraints might be violated. Generalised ARCH (GARCH) models: Allows the conditional variance, σt2 , to depend its own lags as well as lagged squared residuals. GARCH(1,1) The interpretation here is that the current fitted variance is a weighted function of a long term average value (α0), volatility during the previous period (α1ut-12), and the fitted variance from the model during the previous period (βσt-12).

GARCH Models The GARCH model is really an ARMA type of process. To show this the squared residual at time t is equal to the conditional variance and a constant term, to give:

Estimating ARCH/GARCH Models In practice the likelihood function is expressed in logs (so that a multiplicative function becomes an additive one). The log-likelihood function (LLF) for an ARCH model with a normally distributed error is given by L: The computer replaces σ2t in the LLF with its ARCH process.

Maximum Likelihood Estimation: Global maximum L Max Local maximum β βMLE Where L is maximised dlogL/dβ = 0. Software use various algorithms for iteration to the global maximum estimate of β.

ML Estimation of ARCH/GARCH models Specify the model and its likelihood function Use OLS regression to get initial estimates (starting values) for β1 , β1 etc. Choose initial estimates for the parameters of the conditional variance function. Eviews (and other software) offers you zeros as starting values for these. In practice it is better to choose small positive values. Specify a convergence criteria (usually the software has a default value for this). Maximise the likelihood by iteration until no further improvement in the model coefficients can be obtained (and the convergence criteria in step 4 is met).

GARCH (p,q): - The current conditional variance depends on q lags of the past squared error and p lags of the past conditional variance. - However higher order models above GARCH(1,1) are rarely used in practice.

Asymmetric GARCH Models Given that all the terms in a GARCH model are squared, there will always be a symmetric response to positive and negative shocks However due to the leveraged nature of most firms, a negative shock should be more damaging than a positive shock and therefore produce greater volatility. There have been two approaches to this stylised fact, the exponential GARCH model and the Glosten, Jagannathan and Runkle models.

Asymmetric Adjustment The asymmetric adjustment is introduced into the model through the use of a dummy variable which takes the value of 0 or 1 It takes the value of 1 if the shock is negative (i.e. <0), and 0 otherwise

Glosten, Jagannathan and Runkle Model (GJR)

GJR Model From the previous slide it is clear that a large positive shock will produce less volatility than a large negative one, assuming the coefficient γ is positive If the final term is significant, according to the t-statistics, then it implies asymmetric responses are an important effect The non-negativity constraint for terms 2 and 4 are now:

GJR Model It is possible to calculate the conditional variance for a positive and negative shock and hence show that these are different. Clearly in the negative shock case the squared error term will have two components, the individual part and the dummied part.

News Impact Curves Both GARCH and GJR models can be used to produce plots of the next period volatility, that arises following both positive and negative shocks. This simply involves substituting values of u(t-1) in the range [-1,+1] into the estimated model, to obtain various values for the conditional variance. The plot for the GARCH model will be symmetric, for the GJR model, negative shocks will be higher than positive shocks.

EGARCH The exponential GARCH or EGARCH model was developed by Nelson (1991)taking the following form:

EGARCH This is the exponential GARCH model and can also be used to explain asymmetries. A further advantage is that the dependent variable is in logs, so the non-negativity constraint is not breached. As with the GJR model, if the following term is negative and significant, then there is evidence of asymmetry (the leverage effect):

GARCH-in-mean This type of model introduces the conditional variance (or standard deviation) into the mean equation. These are often used in asset return equations, where both return and risk are to be considered. If the coefficient on this risk variable is positive and significant, it shows that increased risk leads to a higher return. Similarly it can be introduced into asset parity models, representing the risk premium.

GARCH-in-mean Example The following model uses the return on a bond as the dependent variable:

GARCH-in-mean Example In the previous slide, the positive sign and significant t-statistic indicate that the risk of the bond leads to a higher return. The diagnostic tests are interpreted in the usual way.

Forecasting using GARCH models The GARCH models are useful for forecasting volatility of asset returns, options and other finance series. This is particularly important in options pricing, where volatility is an important input into the pricing of the option. However it is difficult to produce the standard error band for the confidence intervals for the conditional variance forecasts as this requires the variance of the variance.

Forecasting with GARCH Although the GARCH model has the conditional variance of the error term as its dependent variable, this is the same as the conditional variance of the dependent variable in the mean equation. This is the case regardless of what variables are included in the mean equation.

Conclusion The GARCH model is better in general than the ARCH model, as usually a GARCH(1,1) model is sufficient. The GJR model can be used to model asymmetric adjustment, through the use of a dummy variable. This allows negative shocks to have higher conditional variance than positive shocks.