Semivariance Significance
Baishi Wu, 2/13/08
Outline Motivation Background Math Data Preparation Altria Group/Phillip Morris (MO) Plots Apple (APPL) Plots Summary Statistics Future
Introduction Used Paper by Barndorff-Nielsen, Kinnebrock,
and Shephard (2008) “Measuring downside risk – realized semivariance” as the model
Examine new realized semivariance and bipower downward variation statistics to test for jumps in this model, ought to focus on squared negative jumps
Also did a focus on only positive jumps and computed z-scores for the following as well
The separation of RS from RV is supposed to beat out the prediction mechanism used solely on GARCH memory
Equations Realized Volatility (RV)
Bipower Variance (BV)
Equations Realized Semivariance (RS)
Running an “if” loop to only take values of the returns if they are less than zero in order to solely decreases
Bipower Downard Variance (BPDV) BPDV = RS – (1/2)BV
if r(i,j) <= 0 RS(1, j) = sum(r(:,j).^2); BPDV(1,j) = RS(1,j) -.5*BV(1,j); else RS(1, j) = 0; BPDV(1, j) = 0; end
Equations Tri-Power Quarticity
Relative Jump
Equations Max Version z-Statistic (Tri-Power)
Take one sided significance at .999 level, or z = 3.09
Data Collected at five minute intervals Rewrote code so that the first data point
collected is the fifth entry for that day while the last data point is the last entry of the day (as there are exactly 385)
Two stocks are being analyzed, notably for their differences for the results in the analysis as they respond uniquely to the downward variance analysis
Altria Group is sampled between 1997-2008 (2669)
Apple is sampled between 1997-2000 (676)
Altria Group (Phillip Morris)
Realized Volatility, Bipower Variance
Realized Variance, Z-Scores
Semivariance, Bipower Downward Variance
Realized Semivariance, Z-Scores
Upward Variance, BPUV
Realized Upvariance, Z-Scores
Apple Computers
Realized Volatility, Bipower Variance
Realized Variance, Z-Scores
Semivariance, Bipower Downward Variance
Realized Semivariance, Z-Scores
Upward Variance, BPUV
Realized Upvariance, Z-Scores
Summary Statistics
MO AAPL
Mean Std Mean Std
RV 3.32E-04 0.0017 0.0012 0.0017
upRV 2.24E-04 0.0017 0.0007 0.0008
RS 1.74E-04 3.24E-04 0.0007 0.0017
BV 2.65E-04 4.37E-04 0.001 0.001
BPUV 1.39E-04 0.0016 0.0004 0.0004
BPDV 9.74E-05 1.90E-04 0.0004 0.0013
Jumps 0.37% 0.15%
Jumps Down 0.00% 52.66%
Jumps Up 57.00% 0.59%
Questions Problems with the code? Is there something that
I’m not doing correctly with measuring downside risk
Why the difference in the two stocks’ characteristics?
Improvements in the Tri-Power or Max z-statistic that explain the drastic differences in z-scores that you see?
Verified decreases in mean and standard deviation for the one-directional jumping (is this just because values have been replaced by zeros?)
Extend to GARCH model analysis…?
Additional Extensions Determining Tri-Power Quarticity for only
semivariance Using a larger sample of stocks to view effects
of trimming the data Effect of noise on data