Ivana Komunjer, "Asymmetric Power Distribution: Theory and Applications to
Risk Measurement", Journal of Applied Econometrics, Vol. 22, No. 5, 2007,
pp. 891-921.

*******************************
*            DATA             *
*******************************

The data were obtained from the Center for Research in Security Prices
(CRSP). These consist of daily prices of two indices, S&P500 and NASDAQ,
one individual security, Microsoft, and one exchange rate, British pound
(BP/USD) expressed in terms of the US dollar, during a period from
01/02/1990 to 05/19/2006. 

All ASCII files are in DOS format. Unix users should use unzip -a.

The file ikdata.zip (86 KB) contains:

IN-SAMPLE RAW DATA SERIES (in-sample period ranges from 01/02/1990 to
12/31/2002 (T observations), and is used for estimation):
S&P500_1990_2002.txt		(S&P500 daily index)
NASDAQ_1990_2002.txt		(NASDAQ daily index)
MSFT_1990_2002.txt		(Microsoft daily prices)
EXUKUS_1990_2002.txt		(BP/USD daily exchange rate)

OUT-OF-SAMPLE RAW DATA SERIES (out-of-sample period from 01/02/2003 to
05/19/2006 (R observations), is used for forecasting):
S&P500_2002_2006.txt		(S&P500 daily index)
NASDAQ_2002_2006.txt		(NASDAQ daily index)
MSFT_2002_2006.txt		(Microsoft daily prices)
EXUKUS_2002_2006.txt		(BP/USD daily exchange rate)

Each of the above TXT files has 4 lines of header with the description
of the data. There are two variables listed in two columns: date (xls
format convention 1900) and price (expressed in US dollar).


*******************************
*          PROGRAMS           *
*******************************

All computations and simulations were performed in Matlab, and the
corresponding m-files are available in the file ikprogs.zip (14 KB).
This folder contains the following files:

APD_ES.M
	Alpha_bar-expected shortfall from a standard APD random variable
with parameters alpha and lambda. X = APDES(ALPHA_BAR,ALPHA,LDA) returns
the ALPHA_BAR-expected shorfall of a standard APD random variable with
shape parameters ALPHA, 0 < ALPHA < 1, and LDA, LDA > 0.

APD_GARCH_LOG_LIK.M
	Constructs the objective function involved in the ML estimation
of the parameters of a GARCH(1,1) process with APD innovations.

DO_APD_MLE.M
	Maximum Likelihood Estimation of the parameters of a GARCH(1,1)
process with APD innovations.

DO_OUT_SAMPLE.M
	Performs the out-of-sample comparison of the one-step-ahead
Expected Shrotfall (ES) forecasts (1%, 5% and 10% ES) obtained from
three competing models: (1) APD-GARCH(1,1) model from Equations
(10)-(11) in Komunjer (2006), (2) asymmetric Student-t-GARCH(1,1) model
from Equation (12) in Komunjer (2006), and (3) GPD-EGARCH(1,1) model
from Equations (13)-(14) in Komunjer (2006).

GPD_EGARCH_LOG_LIK.M
	Constructs the objective function involved in the ML estimation
of the parameters of a GPD-EGARCH(1,1) process.

ML_APD_GARCH.M
	Constructs the ML estimates for a GARCH(1,1) process with APD
innovations.

ML_GPD_EGARCH.M
	Constructs the ML estimates for a GPD-EGARCH(1,1) process.

ML_SKEW_T_GARCH.M
	Constructs the ML estimates for a Skew-t GARCH(1,1) process.

SKEW_T_ES.M
	Constructs the alpha_bar ES of a SKEW-T distribution with pdf as
in Giot and Laurent (2004).

SKEW_T_GARCH_LOG_LIK.M
	Constructs the objective function involved in the ML estimation
of the parameters of a Skew-t GARCH(1,1) process.


*******************************
*            AUTHOR           *
*******************************

If you have any questions or problems, please contact:
Ivana Komunjer
University of California, San Diego
Economics #0508
9500 Gilman Drive
La Jolla, CA 92093 - 0508
USA

e-mail: komunjer {AT] ucsd.edu