Breymann PDF

UNIVARIATE AND MULTIVARIATE ANALYSIS
OF HIGH-FREQUENCY FINANCIAL DATA

Wolfgang Breymann
Applied Zurich University of Applied Science, Switzerland
Meielisalp, 8-12 July 2007
OUTLINE
I. Introduction
II Basic variables and methods
III Stylized facts
IV A universal method for univariate and multivariate
deseasonalization
V Dependence structure analysis and modeling
1st International R/Rmetrics User and Developer Workshop
Meielisalp, Switzerland, 2007
I. INTRODUCTION
General remarks on financial market modelling
Basic properties of financial markets
GENERAL REMARKS ON
FINANCIAL MARKET MODELLING
WHY STUDYING HIGH-FREQUENCY DATA

Because they exist.
For understanding the market.
Intraday information may be important
ANNUALIZED USD/JPY VOLATILITY

FROM DAILY AND HIGH FREQUENCY DATA
annualized volatility [%]
20.0
18.0
16.0
14.0
Jan
Feb
Jan
Feb
price
120
115
110
Top: Annualized
USD/JPY volatility
from daily prices at
7am GMT (circles),
from daily prices at
5pm GMT (full
circles) and from
high-frequency data
(solid lines).
Bottom: USD/JPY
high-frequency prices.
Period:
Jan.Feb. 1999.
time

Because they exist.
More stable environment because of less structural breaks
(Permanence Hypothesis)

Because they exist.
More stable environment because of less structural breaks
(Permanence Hypothesis)
More complex models can be tested
OBSTACLES
Collection, collation, storage, retrival, manipulation of
high-frequency data ist costly.
The statistical apparatus has mainly been developed and
thought for homogeneous time series.
Little work is done for irregularly spaced time series.
However:
Development of computer technology has made data more
easily available.
High frequency data are slowly becoming a fanstastic experimental bench for understanding market microstructure and markets in general.
10
METHODOLOGICAL REMARKS ON THE EMPIRICAL

INVESTIGATION OF HIGH FREQUENCY DATA
Quantity of data: 1 year of high frequency data corresponds to
300 1000 of daily data !
New methods for data analysis needed:
Interpolations, processing, filtering, . . .
Important: Development of statistical methods with as few

assumptions as possible concerning the generating process
(explorative analysis)
Demanding task: Modelling daily and weekly seasonalities, scaling
laws, dependencies, empirical distributions, conditional properties,
...
(Over)hasty modeling: Danger to overlook important features
11
METHODOLOGICAL REMARKS (CONT.)

Three step procedure, which has been proved successful in hard
science:
Explore the data to discover their fundamental statistical
properties with a minimum set of assumptions.
(Finding the stylized facts.)
Use the empirical facts to formulate adequate models.
(Finding the stylized facts.)
Verify whether the models reproduce the stylized facts in
a satisfactory (quantitative) way and make predictions of
future behavior (as to price movements, risk, etc.).
12
APPLICATIONS
(Beyond the purely scientific approach)
Integrate the results into practical applications such as
trading models, derivative pricing asset allocation, or risk
management algorithms.
13
TIME SCALES AND SAMPLE SIZES

Time scale
Sample Size
Effect
Example
Price formation
process
14 years Olsen tickby-tick data
1 min
1.5 107
10 min
500 000
1 hour
80 000
intra day effects

(seasonality)
1 day
15000
breakdown of Permanence Hypothesis
1 week
5000
1 month
2000
cotton prices 18801940 (Mandelbrot, 1963)
1 year
700
commodity prices from England and Holland

(Froot et al., 1995)
S&P500 daily
19281991
14
TYPES OF MODELS
Time scale
Sample Size
1 min
1.5 107
10 min
500 000
1 hour
80 000
1 day
15000
1 week
5000
1 month
2000
1 year
700
Type of model
Market Microstructure Models
Time Series and Financial Models
Macroeconomic models
Different models focus on different aspect of the markets.

With high-frequency data it should be possible to reconcile these
types of models.
15
INTERRELATING DIFFERENT TIME SCALES

High-frequency data opens the way for studying financial markets
at very different time scales.
Some empirical properties are similar at different time scales
leading to fractal behaviors. This may result in so-called scaling
properties.
Behaviors observed for daily or weekly data gain an additional
weight when also observed with high significance for intra-day
data.
16
FOREIGN EXCHANGE MARKETS: BASIC FACTS

Data available in high frequency over long sampling periods
Market active round the clock during working days
Data irregularly spaced, with gaps during weekends and nights.
Large liquidity
Symmetric because both exchanged assets are currencies
THEREFORE:
New facts often found first in FX spot data

FX studies served as role model for study of other highfrequency markets
17
AVAILABILITY OF DATA
OTC markets
Data are collected and provided by data providers
(Reuters, Boomberg, etc.)
Prices are quoted prices, no transaction prices
No volume data available
Data quality often not very good
18
FOREIGN EXCHANGE MARKETS: VOLUME

The largest financial market
Daily turnover (spot and forwards):
Year
Billion USD
1992
832
(more than total non-gold reserves
(USD 555.6 billion) in 1992)
1995
1190
1998
1500
90% of volume attributed to intraday interbank market

19
FOREIGN EXCHANGE MARKETS: NO. OF TICKS

Rate
Period
#Ticks
Daily
[million]
freq.
EUR/USD
Jan 1999Jun 2001
10.9
17 440
GBP/USD
Jan 1987Jun 2001
9.8
2 703
USD/CHF
Jan 1987Jun 2001
10.0
2 759
USD/JPY
Jan 1987Jun 2001
12.7
3 503
EUR/JPY
Jan 1999Jun 2001
3.5
5 600
EUR/GBP
Jan 1999Jun 2001
3.3
5 280
USD/DEM
Jan 1987Dec 1998
18.4
5 660
USD/FRF
Jan 1987Dec 1998
3.7
1 138
DEM/JPY
Oct 1992Dec 1998
1.3
400
Number of quotes of entire FX market ca. 275 000 ticks/day.

20
COMPLEXITY OF THE FX MARKET

Homogeneity of market agents disappear
Reason: Interaction of market participant with different
objectives
geographical locations
institutional constraints
time horizons
risk profiles
Heterogeneous Market Hypothesis
21
FOREIGN EXCHANGE MARKETS: STRUCTURE

Two dominating parts:
FX spot market
FX forward market
Repartition was 50:50 in 1992 and 40:60 in 1998
FX spot quotes also used for FX forward market
Smaller but growing group of FX derivatives
22
FOREIGN EXCHANGE MARKETS: ACTIVITY

No business hour limitations, trading is practically continuous
Global market
Activity pattern composed of three continental components:
East Asia (Tokio)

(Australia, Hong Kong, India, Indonesia, Japan, South Korea,
Malysia, New Zealand, Singapore)
Europe (London)
(Austria, Bahrain, Belgium, Germany, Denmark, Finland,
France, Great Britain, Greece, Ireland, Italy, Sweden, Switzerland, Turkey, United Arab Emirates)
Americal (New York)
(Argentina, Canada, Mexico, United States)
23
FOREIGN EXCHANGE MARKETS: TRADING

Electronic trading through automated, electronic order-matching
systems:
High quality data
Transaction prices and volumes available
Use of electronic trading growing
Direct OTC market between banks (and brokers):
Major institutions quote bid and ask offers (quoted price)

Transmission to customers in real time by data provider
Deals negociated over telephone
The only source for large high-frequency samples
24
FOREIGN EXCHANGE MARKETS: ACTORS

Type of actors:
Nonfinancial corporations
(mutual funds, pension funds, insurance companies)
Institutional investors
Hedge funds
Central banks
Technology-driven shift from long-term buy and hold investment

to short-term profit-making transactions
Increasing volume of short-term, intraday transactions
HOWEVER: Central banks can afford large open positions and
can have significant impact on the market in the long run
25
VARIABLES
Most important variable under study:
Logarithmic middle price:
p
1
x(ti = (log pask (ti ) + log pbid (ti )) = log pbid (ti ) pask (ti )
2
where ti is the time stamp.
Advantages: Logarithmic prices depend only on the relative
price movement and not on its absolute level.
Returns are defined with respect to a given time horizon t.
The logarithmic return r(t; t) at time t wrt t is defined as
r(t; t) = x(t) x(t t).
4.4
4.8
5.1
FX LOG. PRICES AND LOG. RETURNS
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
0.10
0.06
1987
1975
1980
1985
1990
USD/JPY log. prices and log. returns.
1995
2000
1999
27
VOLATILITY: OVERVIEW
Importance
Different types
Definition
Scaling and annualization
28
VOLATILITY: IMPORTANCE
Option pricing.
Risk management.
LeBaron Effect:
Market reaction depends on volatility.
(Blake LeBaron, Journal of Applied Econometrics,
1992; Journal of Business, 1992.)
29
TYPES OF VOLATILITY
Realized (or historical) volatility
Computed from historical data
Model volatility
Virtual variable in theoretical models such as
GARCH or stochastic volatility models
Implied volatility
A volatility forecast computed from market prices
of derivatives.
30
VOLATILITY: DEFINITION
Realized (or historical) volatility is defined for
regularly spaced returns as
v(ti ) = v(t, n, p; ti ) =
where
1
n
n
X
i=1
|r(t; tin+j )|
! p1
t is the return interval (time horizon of return)

and
nt is the size of the total sample.
Mathematically, realized volatility is p-norm.
Most frequent choices for p are 1 or 2.
Parameters t and n may vary over large ranges.
31
VOLATILITY: SCALING AND ANNUALIZATION

v(t = 10 min, n, p; ti ) gives volatility for regularly spaced
10 min time series.
For convenience one may want volatility in scaled form:
r
tscaled
vscaled =
v
t
Most popular choice for tscaled is 1 year
annualized volatility:
r
1year
vann. =
v
t
Practitioners often use annualized volatility in percent.
(For FX spot prices of major currencies about 10%.)
32
0.05
0.25
ANNUALIZED USD/CHF VOLATILITY
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
0.000
0.020
1987
USD/CHF daily volatility computed from 5 min. data (top)

and daily data (bottom).
33
VOLATILITY: CHOICE OF PARAMETER p

Larger values of p give more weight to the tails.
If p too large, expectation of volatility may no longer
exist.
Therefore, p should be smaller than tail index .
If volatility correlations are considered, p should be
smaller than /2.
For high-frequency FX spot: 3.5 p = 1.
34
0.02
0.18
p-DEPENDENCE OF REALIZED VOLATILITY
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
0.1
0.5
0.05
0.30
1987
Annulaized USD/CHF volatility with t = 5 min, n = 288, and p = 1

(top), p = 2 (middle), and p = 3 (bottom).
35
VOLATILITY: CHOICE OF PARAMETER t

Theoretical considerations (diffusion theory) suggest to
make t as small as the available data allow.
However, for too small values of t realized volatility
becomes biased (Andersen et al., 2000).
Without bias correction, t 15 min . . . 2 hr optimal.
With bias correction, t 5 min possible
(Corsi et al., 2001).
36
IRREGULAR TIME SERIES OPERATORS

Suitably defined linear operators provide an alternative
way to analyse high-frequency data.
Dacorogna et al.: An Introduction to High-Frequency
Finance, chap. 3.3.
Main idea: Define linear operators
Z t
[z](t) =
dt (t t ) z(t )
which transforms one irregularly spaced time series into

another one.
For computational efficiency, kernels are built out of exponential moving averages of different order.
37
III. STYLIZED FACTS

Price formation (market microstructure effects)
Distributional properties of returns
Scaling laws and volatility autocorrelation
Seasonality and exogenous impact
38
SF: PRICE FORMATION EFFECTS

Negative 1st order autocorrelation of returns
Discreteness of quoted spreads
Short-term triangular arbitrage
39
SHORT RANGE CORRELATIONS OF RETURNS

At highest frequencies return display negative
autocorrelation.
Correlation disappears after about 5 minutes.
40
0.0
0.10
0.15
ACF
0.05
10
15
20
25
30
Lag [min.]
Autocorrelations of USD/DEM 1 min. returns (sample: 19871998).
41

At highest frequencies return display negative
autocorrelation.
Correlation disappears after about 5 minutes.
This is a sign of the price formation process.
Reasons:
Autocorrelation reflects different opinion of traders

Quotes of market makers are skewed depending on
their objectives
42
DISCRETENESS OF QUOTES SPREAD

Quoted spread have discrete values. There are major
peaks at 5 and 10 basis points.
Over the year, maximum shifted from 10 bps to 5 bps.
Traded spreads computed from transaction data are
smoothly distributed over all possible values.
43
8000
4000
0
counts
12000
HISTOGRAM OF BIDASK SPREAD
10
15
20
Bidask spread [bps]
Histogram of bidask spread of USD/CHF FX spot rates during

January 1993.
44
SHORT-TERM TRIANGULAR ARBITRAGE

On the very short term, the major rates have predictive
power over the cross rates.
The reason is that it takes some time before trend
reversals in the major rates are reflected in the cross
rates.
This gives rise to very short term triangular arbitrage.
45
SF: DISTRIBUTIONAL PROPERTIES OF

UNIVARIATES RETURNS
Non-gaussian return distributions

Decreasing heavy tails
Characterization of tails
Application to risk measurement
46
THE FIRST MOMENTS

Mean:
=
Variance:
2 =
Skewness:
s=
Kurtosis:
=
(x a) dF
(x a)2 dF
(x a)3 dF
3
4
(x
a)
dF
47
Empirical Moments of return Distributions
fx rate
time interval
USD/DEM
USD/JPY
mean
variance
skewness
kurtosis
30 minutes
1.40 106
7.53 107
0.60
46.10
6 hours
1.68 105
8.42 106
0.27
11.75
24 hours
6.62 105
3.48 105
0.12
6.04
1 week
4.65 104
2.41 104
0.17
4.24
30 minutes
2.20 106
7.16 107
0.05
24.02
6 hours
2.65 105
7.98 106
0.17
11.64
24 hours
1.06 104
3.13 105
0.16
7.06
1 week
7.57 104
2.22 104
0.23
4.29
The four first moments of the return distribution at different time intervals for the major
currencies against the USD. The estimations are performed on samples from January 1, 1987
to June 30, 1996.
48
fx rate
time interval
GBP/USD
30 minutes
2.33 107
6 hours
USD/CHF
USD/FRF
mean
variance
skewness
kurtosis
6.56 107
0.14
21.49
2.79 106
7.77 106
0.42
10.89
24 hours
9.86 106
3.13 105
0.34
7.50
1 week
8.53 105
2.38 104
0.84
7.32
30 minutes
1.53 106
9.20 107
0.52
66.76
6 hours
1.82 105
1.04 105
0.12
10.09
24 hours
7.17 105
4.24 105
0.04
5.68
1 week
5.06 104
2.85 104
0.07
3.90
30 minutes
1.28 106
6.13 107
0.27
28.01
6 hours
1.54 105
7.59 106
0.21
9.89
24 hours
6.04 105
3.10 105
0.09
5.92
1 week
4.25 104
2.19 104
0.19
4.32
49
CUMULATIVE DISTRIBUTIONS OF
USD/JPY RETURNS
0.999
0.998
0.995
0.990
The cumulative distributions for 10 minutes, 1

day and 1 week analysis
interval of the USD/JPY
returns shown against the
Gaussian probability on the
y-axis. On the x-axis the
returns normalized to their
mean absolute value are
shown.
0.980
0.950
cumulative frequency
0.900
0.800
0.500
0.200
0.100
0.050
0.020
0.010
0.005
0.002
0.001
-10.0 -8.0
-6.0
-4.0
-2.0
0.0
2.0
4.0
6.0
8.0
10.0
normalized return
50
CHARACTERIZATION OF THE TAILS

Classification of distributions
Characterisation of fat-tailed distributions
Empirical estimation of tail behavior
Importance of the sample size
How heavy are the tails of financial instruments?
Aggregation properties
51
CLASSIFICATION OF DISTRIBUTIONS
Thin-tailed distributions:
Cumulative distribution function declines

exponentially in the tails
All moments are finite
Fat-tailed distributions:
Cumulative distribution function asymptotically

declines as a power with exponent
Bounded distributions without tails
52
MOMENTS OF FAT-TAILED DISTRIBUTIONS

The density of fat-tailed distributions asymptotically behave as

1
1
(x) = a (x x0 )
+ o (x x0 )
Moments:
E[X k ] =
+
C +a
(x x0 )+k1 dx
o (x x0 )
only exist for k < .
Tail index stable under aggregation.
53
HOW HEAVY ARE TAILS OF

FINANCIAL DISTRIBUTIONS?
Mandelbrot: Returns follow stable distributions.
Implications:
Tail indices between 0 and 2
Variance does not exist.
Risk has often be associated with variance
(e.g. Markowitz portfolio theory)
Existence of variance central for all areas of finance
Try to decide this question empirically
54
EMPIRICAL ESTIMATION OF TAIL BEHAVIOR

Let X1 , X2 , . . . , Xn a sample of n independent n observations
with unknown probability distribution function F .
Let X(1) X(1) . . . X(n) the descending order statistics.
The so-called Hill-estimator (Hill, 1975)
H
n,m
m1
1 X
=
ln X(i) ln X(m)
m 1 i=1
with m > 1 is a consistent estimator of = 1/.
55
ln(m)
10
TAIL BEHAVIOR OF STUDENT-4 DISTRIBUTION
ln(X(1))ln(X(m))
Sample size: 100000, average computed over 50 simulations.

56
HILL-ESTIMATOR: PROPERTIES
(n,m ) m is asymptotically normally distributed.

H
n,m
biased for finite sample size.
H
depends on m.
n,m
No easy way to determine best value of m
57
EMPIRICAL RESULTS
Estimated by a subsample bootstrap method
Confidence ranges are standard errors times 1.96
(would correspond to 95% confidence for normally-distributed
errors)
Computation of standard error by jackknife method:
Data sample modified in 10 different ways by removing

one-tenth of the total sample
Tail index computed separately for every modified sample
58
TAIL INDICES FOR MAIN FX RATES

Rate
30m
1 hr
2 hr
6h
1 day
USD DEM
3.18 0.42
3.24 0.57
3.57 0.90
4.19 1.82
5.70 4.39
USD JPY
3.19 0.48
3.65 0.79
3.80 1.08
4.40 2.13
4.42 2.98
GBP USD
3.58 0.53
3.55 0.65
3.72 1.00
4.58 2.34
5.23 3.77
USD CHF
3.46 0.49
3.67 0.77
3.70 1.09
4.13 1.77
5.65 4.21
USD FRF
3.43 0.52
3.67 0.84
3.54 0.97
4.27 1.94
5.60 4.25
USD ITL
3.36 0.45
3.08 0.49
3.27 0.79
3.57 1.35
4.18 2.44
USD NLG
3.55 0.57
3.43 0.62
3.36 0.92
4.34 1.95
6.29 4.96
XAU USD
4.47 1.15
3.96 1.13
4.36 1.82
4.13 2.22
4.40 2.98
XAG USD
5.37 1.55
4.73 1.93
3.70 1.52
3.45 1.35
3.46 1.97
59
TAIL INDICES FOR SOME CROSS RATES

Rate
30m
1 hr
2 hr
6h
DEM JPY
4.17 1.13
4.22 1.48
5.06 1.40
4.73 2.19
GBP DEM
3.63 0.46
4.09 1.98
4.78 1.60
3.22 0.72
GBP JPY
3.93 1.16
4.48 1.20
4.67 1.94
5.60 2.56
DEM CHF
3.76 0.79
3.64 0.71
4.02 1.52
6.02 2.91
GBP FRF
3.30 0.41
3.42 0.97
3.80 1.34
3.48 1.75
FRF DEM
2.56 0.34
2.41 0.14
2.36 0.27
3.66 1.17
DEM ITL
2.93 1.01
2.60 0.66
3.17 1.28
2.76 1.49
DEM NLG
2.45 0.20
2.19 0.13
3.14 0.95
3.24 0.87
FRF ITL
2.89 0.34
2.73 0.49
2.56 0.41
2.34 0.66
60
RESULTS: SUMMARY
For 30 min returns all rates against USD as wells as nearly all cross
rates outside the EMS have tail indices around 3.5 (between 3.1
and 3.9).
Gold and silver have tail indices above 4.
These markets differ from the FX market, with very high
volatilities in the 80s and much lower volatilities in the 90s
(structural change).
EMS rates show lower tail indices around 2.7:
Reduced volatility induced by the EMS rules is at the cost of a

larger probability of extreme events for realigning the system.
Argument against the credibility of systems like the EMS.
61
RESULTS: SUMMARY (CONT.)
Tail index reflects:
Institutional setup
The way different agents interact
They are an empirical measure of market regulation and efficiency

Tail index large:
Free interactions between agents with different time horizons
Low degree of regulation

Smooth adjustment to external shocks
Tail index small:
High degree of regulation

Abrupt adjustment to external shocks
62
Tail indices between 2 and 4

2nd moments of return distributions finite
4th moments of return distributions usually diverge
Absolute returns are preferred for computation of
volatility autocorrelation.
63

Return distributions are fat-tailed with tail index > 2 and
therefore non-stable.
Invariance under aggregation satisfied up to 2 . . . 6 hours.
For larger time horizons tail behavior can no longer be observed
because:
Transition to tail behavior occurs at increasingly larger
(relative) returns.
Sample size decreases.
Detailed studies have shown that 18 years of daily data are not
enough to estimated the tail index reliably.
64
0.00001
0.00100
0.10000
AGGREGATION OF t-3.5 DISTRIBUTION

+++++++
+
+ ++
+
+
+
+
+
+
+
+
++
+
+
+
+
+
++
+
+
+++
+
+
++
++++
++
+
+++
+
+
++
++
+++++
+++++++++
++++++++
+ +
+
+
+
+ +

+ +
+
+
+
+
+ + + ++
+ ++++++++++++
+ +++
+
+
+

15
10
10
15
x/[Stddev]
Density of t-3.5 distribution (black squares), sum of 4 t-3.5 distributed

random variables (+) and sum of 10 t-3.5 distributed random variables.
Note the transition from normal to tail behavior.
65

MULTIVARIATE RETURNS:
DEPENDENCE STRUCTURE
Covariance matrix describes only

linear part of the dependence
66
RETURN SCATTER PLOTS
0 01
00
0 01
0 02
0 01
0 02
0 02
0 01
00
0 01
0 02
0 02
0 01
00
0 01
USD DEM
USD DEM
8 Hours re urns
12 Hours re urns
1 Day re urns
0 01
USD DEM
0 03
0 02
00
0 02
0 04
0 04
0 02
00
USD JPY
USD JPY
0 02
0 03
0 01
0 01
0 02
0 04
USD DEM
0 01
0 03
00
USD JPY
0 01
0 02
0 01
0 02
0 03
USD JPY
00
0 01
0 04
0 02
4 Hours re urns
0 02
0 01
00
USD JPY
.
.
..
. ..
..
... . ... .. .. . .
.
.
. . ........... .. .. ..... . .
. .. .................................................................. . . ....... .
.
............................................................................... .. . .. .
.
................................................................................. . . .
........
.......
.................................... . .
......
.....
......
.........
....
......
...
....................................
...
..
...
..........
..
..........
.....
......
..
.....
...
.....
.................
...
..
....
.........
.........
..
....
. .. .
..........
....
....
.....
..
...
..
.....
..
..
.........
..
....
....
...
.......
...
..
.......
...
...
...........
...
..
...
......
..........
......
..
......
...
........
..
...
........
........
....
..
..
...
........
.......
.....
....
.......
..........
..
..
...
.........
....
...............................
...
........................ . . . . .
.. . .. ...............................................................
......
......
.............
.......
....
..
....
....
...
..................
.
.
..
.............
..
......
..
.
.
.
..
.
.
..
.
.
.
.
...
.
.
.
..
.
.
.
..
.
.
.
.
..
.
.
.
.
.
.
..
..
.
.
.
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . ... . .............................................................. . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
.. .
.
. ................................................................................... . . .
. .... . .. . ....... .... ... ........ . . .
.
.
.
.
.
.
.. . .. .. .. . . .
. .
.. .
. .. . .
.
0 02
0 01
2 Hours re urns
USD JPY
0 02
1 Hour re urns
0 04
0 02
00
0 02
0 04
USD DEM
0 04
0 02
00
0 02
0 04
USD DEM
67

MULTIVARIATE RETURNS:
DEPENDENCE STRUCTURE
Covariance matrix describes only

linear part of the dependence
Before description of the full
dependences structure,
deseasonalization necessary
68
S.F.: VOLATILITY AUTOCORRELATION

AND SEASONALITIES
Pronounced daily and weekly seasonalities in
autocorrelation function of absolute returns
Slowly decaying autocorrelation function of absolute
returns (after deseasonalization)
AUTOCORRELATION FUNCTIONS
OF ABSOLUTE RETURNS
0.2
0.1
0.0
0.0
0.1
ACF
0.2
0.3
USD/DEM and USD/JPY
0.3
USD/DEM
200
400
600
800
1000
200
600
800
1000
800
1000
0.2
0.1
0.0
0.0
0.1
ACF
0.2
0.3
USD/JPY
0.3
USD/JPY and USD/DEM
400
1000
800
600
Lag
400
200
200
400
Lag
600
0.02
0.02
HOURLY RETURNS
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
0.010
0.010
1987
Jan
Feb
Mar
Apr
May
Jun Jul
1990
Aug
Sep
Oct
Nov
Dec
Jan
1991
71
INTRADAY AND INTRAWEEK

VOLATILITY PATTERNS
There are many different patterns observed.
Extreme cases:
OTC markets as the FX spot market

Localized, exchange-traded markets as equity indices.
The patterns may change when the institutional setting is

changing (example: change of the opening hours of a market).
THE WEEKLY VOLATILITY PATTERN
5e04
1e04
5 min volatility
9e04
USD/DEM
Days of week
5e04
3e04
1e04
5 min volatility
USD/JPY
Days of week
73
0.2
0.15
0.15
EUR/USD
0.2
0.1
0.05
EUR/USD
SEASONAL VOLATILITY PATTERN FOR FX SPOT RATES
0.1
0.05
0
Sat Sun
0.2
0.15
0.15
USD/JPY
0.2
0.1
0.05
12 15 18 21 24
12 15 18 21 24
USD/CHF
0
Mon Tue Wed Thu Fri
0.1
0.05
0
0
12 15 18 21 24
Full, red line: summer time. Dashed, green line: winter time.
74
SP500 future (qt)
0.5
0.4
0.3
0.2
0.1
0
0
12
15
18
21
24
3500
3000
2500
2000
1500
1000
500
0
IR future DEM (liffe), 3m
SEASONAL VOLATILITY PATTERN FOR

EXCHANGE-TRADED INSTRUMENTS
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0
12
15
18
21
24
12000
Nikkei225
GDAXI
10000
8000
6000
4000
2000
0
0
12 15 18 21 24
12 15 18 21 24
75
LONG RANGE VOLATILITY AUTOCORRELATIONS
After deseasonalization:
Autocorrelation function of absolute returns displays
long-time dependence.
ACF OF DESEASONALISED HOURLY ABS. RETURNS
0.2
0.1
0.0
0.0
0.1
ACF
0.2
0.3
USD/DEM and USD/JPY
0.3
USD/DEM
200
400
600
800
1000
200
600
800
1000
800
1000
0.2
0.1
0.0
0.0
0.1
ACF
0.2
0.3
USD/JPY
0.3
USD/JPY and USD/DEM
400
1000
800
600
Lag
400
200
200
400
Lag
600
77
S.F.: SCALING LAWS

Empirical scaling behavior of first moments of returns
An application of scaling behavior analysis
Empirical scaling behavior of arbitrary moments
True vs. apparent scaling
Anomalous scaling behavior can be explained by multi-fractal
properties.
78
SCALING LAW
Dependence of mean absolute return on the size of the time

interval on which the return is observed.
Empirically, we find a power law for mean absolute returns:
|r[t]| = C (t)
t is the interval size and D the drift exponent (a constant).

The drift exponent is around D = 0.58 for freely floating FX
rates, with a confidence of typically 0.01.
A Gaussian i. i. d. random walk would have D = 0.5.
79
SCALING LAW GRAPH

1 year
log(mean absolute return)
-2
-4
1 day
-6
Scaling law:
D
T
hr[T ]i = 11230
days)
10 min
-8
Mean absolute change of the

logarithmic middle price as a
function of the time interval
(in seconds). for USD/DEM
spot rates (Period: Feb 1986
to Sep 1993, ca. 10 million
quotes).
Vertical bars indicate the
total error (stochastic and
bid-ask uncertainty).
10
12
14
16
size of analysis time interval: log( interval in seconds )
18
with drift exponent

D = 0.59.
80
DRIFT EXPONENT VERSUS TIME
Drift exponent D
(a) USD rates
(b) EMS rates
0.70
0.70
0.65
0.65
0.60
0.60
0.55
0.55
0.50
0.50
0.45
0.45
0.40
0.40
0.35
0.35
0.30
87
88
89
90
91
92
93
0.30
94
87
88
Years
89
90
91
92
93
94
Years
Yearly estimation of the drift exponent of the scaling law for USD rates
(a) on the left (DEM (circle), FRF (triangle), JPY (star)) and (b) EMS
rates against the DEM on the right (ITL (box) and FRF (diamond)).
The drift exponents reflects the institutional framework of FX rates.
81
GENERALIZED SCALING BEHAVIOR
Dependence of the moments of absolute returns on the size of the

time interval on which the return is observed.
Empirically, we find a power law for mean absolute returns:
|r[t]|n = C (t)
t is the interval size and n the scaling exponent.

The scaling exponent for absolute returns is around 1 = 0.58 for
freely floating FX rates
The scaling exponent for squared returns is around 1.
82
10
0
Log moment
15
Scaling of Moments
10
12
Log time horizon
Scaling behavior of the absolute moments of returns for USD/DEM

spot rates. (Feb 1986 to Sep 1998, more than 10 million quotes).
83
SCALING EXPONENTS VS. ORDER
1.0
0.5
Exponent
1.5
Scaling Exponents
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Order
CURRENCIES: (): USD/DEM; (+): USD/CHF; (x): GBP/USD;

(triangle): DEM/FRF; (diamond, red): SCM model.
84
SF: INFORMATION FLOW

FROM LONG TO SHORT HORIZONS
Heterogeneous objects.
Not only the market markers but many participants.
Different people with different interests, different geographical
locations, and different time horizons.
Time scale (frequency of interventions) is the aspect for which the
heterogeneity becomes the most obvious.
Information flow from long to short time scales.
85
CATEGORIES OF MARKET PARTICIPANT

Intraday dealer
Fund manager
International Companies
Central banks
They influence the market on different time scales.
86
A NON-TRADITIONAL VIEW OF A FINANCIAL MARKET
Long-term traders
Market Makers
Medium-term traders
Time Horizons
(price changes)
News hit the market at the center. If they are important the price changes will hit different
market components and feedback effects will happen in the center. Price changes and time
horizons are related through the scaling law |x| tD .
87
AN ASSYMMETRIC INFORMATION FLOW

Coarse volatility predicts fine volatility better
than the other way around:
3

Q
3

Q

QQ
s
88
INFORMATION FLOW ASYMMETRY, THE HARCH EFFECT

Information Flow for the Volatility
12
Future Granularity
10
0
0
10
12
Past Granularity
-0.3 -0.2 -0.1 -0.0 0.1
0.2
0.3
Linear Correlation
I(n, n ) = Corr([T,
T
T
T
T
](t),
[T,
](t
+
T
))
Corr([T,
](t
+
T
),
[T,
](t))
2n
2n
2n
2 n
89
IV. DESEASONALIZATION OF
HIGH-FREQUENCY FINANCIAL DATA
Preliminary with daily data
Approaches for deseasonalizing univariate data
Criteria for deseasonalization methods
Adaptable method for deseasonalizing multivariate highfrequency financial time series
Demonstration of a software tool for deseasonalization
Applications
1.4
1.8
2.2
FX PRICES FOR USD/DEM AND USD/JPY
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
90
130
170
1987
91
DESEASONALIZATION OF FINANCIAL DATA

High frequency financial data present strong seasonalities
Main periodicities: Daily and weekly
Seasonalities cover more subtle statistical properties
Affected by Daylight Saving Time (DST)
Holiday effects
Change of the seasonal pattern over time
92
AUTOCORRELATION OF
1.0
DAILY ABSOLUTE RETURNS
0.6
0.4
0.2
ACF
0.8
0.0
10
20
30
Lag [days]
Autocorrelation of daily absolute returns for USD/DEM FX spot rates,

weekends included.
93
AUTOCORRELATION OF
1.0
DAILY ABSOLUTE RETURNS
0.6
0.4
0.2
ACF
0.8
0.0
10
20
30
Lag [days]
Autocorrelation of daily absolute returns for USD/DEM FX spot rates,

weekends dropped.
94
DESEASONALIZATION OF DAILY DATA
10
Disregarding weekends implicitly consists in a time transformation

on so-called business time:

Business time [days]
10
12
14
Physical time [days]
Disregarding public holidays breaks weekly periodicity, thus blurring

remaining weekly seasonalities
95
APPROACHES FOR UNIVARIATE

DESEASONALIZATION
Two different groups:
Time transformation methods

Weighting methods
Both are based on a daily or weekly conditional mean of

volality.
96
CRITERIA FOR DESEASONALIZATION METHODS
Criteria
Aggregation of returns
Connection to foundamental properties of markets
Extension to multivariate
case
Time transformation
Volatility weighting
through quadratic variation connected to stochastic processes

not clear
purely phenomenological
straight forward
97
AGGREGATION OF RETURNS
Returns of price series can be aggregated:
r[t1 ](t) + r[t2 ](t + t1 ) = r[t1 + t2 ](t)
Its a consequence of the fact that returns are differences
of logarithmic middle prices.
Aggregation property remains valid after an arbitrary
time transformation.
It is lost if the returns are weighted by a time-dependent
weight function.
98
RELATION TO GENERATING PROCESS
Price series can be modeled by stochastic processes with (semi-)

martingale properties.
The process is driven by trading (or market) activity, which can be
view as the directing process.
The idea: activity make the (intrinsic or economic) time advance
(first models: Clark (1973); Tauchen and Pitts (1983)).
Because of the diffusion law, volatility should be intimately related
to the activity-based time (through quadratic variation).
Deseasonalization through time transformation is thus a concept
naturally rooted in the theory of stochastic processes.
99
EXTENSION TO MULTIVARIATE SERIES

Synchronicity of events of different coordinates of the series
should be conserved.
Since time is treated in the same way for all coordinates,
requirement fulfilled by weighting methods.
Activity-based time scales are different for different coordinates.
Time transformation to individual activity-times destroys
synchronicity.
Therefore this sort of time transformation is not adequate for
multivariate desasonalisation.
100
ADAPTABLE DESEASONALIZATION OF
HIGH-FREQUENCY MULTIVARIATE
FINANCIAL TIME SERIES
Modelling requirement
The integrated volatility
Computing the seasonal volatility
Relative market activity
Holiday tracking
101
MODELLING REQUIREMENTS
The flexibility of modelling arbitrary patterns that display

abrupt volatility changes ( Japanese lunch break)
Taking into account slow temporal changes in the habits
of the market participants, institutional changes, etc.
Keeping track of Daylight Saving Time (DST) to take
into account the one hour displacement between DST and
nonDST periods
The modelling of the geographical decomposition of market activity to take into account local public holidays and
other irregularities
102
THE INTEGRATED VOLATILITY

Vt2
Integrated squared volatility:

= t t (vt [])
(Discrete approximation of quadratic variation).
Volatility wrt horizon T :
Vt2 [T ]
Vt2
2
VtT
Deseasonalised returns:
t tT
.
xt [T ] = q
Vt2 [T ]
Pn1
i=0
(vti []) = (vt [T ])
= 5 minutes: elementary time step; n = T /

Aggregation property:
xt [T ] =
xtT2 [T1 ]
2
VtT
[T1 ] + xt [T2 ]
2
q
Vt2 [T ]
Vt2 [T2 ]
103
COMPUTING THE SEASONAL VOLATILITY

Decomposition of the volatility:

2
2
(d)
vt [] = at v []
with relative market activity factor at and

Volatility averaged over DST period d conditional to
the time in the week, = t mod (1 week):
Nd

2
X
1
(rti + [])2 .
v(d) [] =
Nd i=1
104
DAYLIGHT SAVING TIME (DST)
DST periods exist in the American and the European

market components.
DST periods of different market components are in
general different.
To simplify modeling we work with one approximate
DST period for both the American and the European
component.
105
LIST OF SOME DST PERIODS

DST Periods
UK
US
Averaged
27.03.199425.09.1994
03.04.199430.10.1994
03.04.199409.10.1994
25.09.199426.03.1995
30.10.199402.04.1995
09.10.199402.04.1995
26.03.199524.09.1995
02.04.199529.10.1995
02.04.199508.10.1995
24.09.199531.03.1996
29.10.199507.04.1996
08.10.199507.04.1996
31.03.199627.10.1996
07.04.199627.10.1996
07.04.199627.10.1996
27.10.199630.03.1997
27.10.199606.04.1997
27.10.199606.04.1997
30.03.199726.10.1997
06.04.199726.10.1997
06.04.199726.10.1997
26.10.199729.03.1998
26.10.199705.04.1998
26.10.199705.04.1998
29.03.199825.10.1998
05.04.199825.10.1998
05.04.199825.10.1998
UK (left), US (middle), and approximate (right) DST periods.

THE WEEKLY VOLATILITY PATTERN
5e04
1e04
5 min volatility
9e04
USD/DEM
Days of week
5e04
3e04
1e04
5 min volatility
USD/JPY
Days of week
107
0.2
0.15
0.15
EUR/USD
0.2
0.1
0.05
EUR/USD
SEASONAL VOLATILITY PATTERN FOR FX SPOT RATES
0.1
0.05
0
Sat Sun
0.2
0.15
0.15
USD/JPY
0.2
0.1
0.05
12 15 18 21 24
12 15 18 21 24
USD/CHF
0
Mon Tue Wed Thu Fri
0.1
0.05
0
0
12 15 18 21 24
Full, red line: summer time. Dashed, green line: winter time.
108
RELATIVE MARKET ACTIVITY
Total relative activity:

a(t) =
n
X
i=1
where
ai (t) = s0 +
n
X
hi (t) si (t),
i=1
s0
ai (t) =
+ hi (t) si (t)
n
and
si (t), i = 1 . . . n determines the momentary share of the ith
market component, s0 being a constant factor, and
hi (t), i = 1 . . . n is the holiday factor of market component i,
which assumes values between 0 and 1.
109
MARKET SHARE FACTORS

s0 accounts for market components which are not
explicitly modelled.
Normalization condition:
n
X
si = 1.
i=0
110
REGIONAL MARKET COMPONENTS

Tick frequency not reliable as indicator for market
activity.
Therefore separation of market components relies on
daily opening periods.
Share factors given as
w0
w +
wi oi (ti )
0
i=1
si (t) =
wi oi (ti )
wi oi (ti )
w0 +
if i = 0
otherwise,
i=1
111
with smoothed step functions
oi (t) =
where
(o)
ti
(o)
1 + exp
n
n
(o)
i (t
(o)
ti
+ )
1
(c)
(c)
1 + exp i (ti t)
(c)
(ti ) are (approximate) market opening (closing) times and

(c)
(i ) determine the slopes of the corresponding edges.
By construction the si (t) fulfil the required normalization condition

Pn
ti .
i=0 si (ti ) = 1
112
HOLIDAYS FACTORS
During public holidays or other exceptional days the holiday factor of the market component concerned is set to non-negative
value strictly smaller than 1:
h() if t [T () , T () ]
i
i,e
i,b
hi (t) =
1
otherwise.
where
i relates to a specific component and labels the public

holidays,
()
hi
()
is the holiday activity factor, and

()
Ti,b (Ti,b ) is the beginning (end) of holiday of component i.

113
HOLIDAYS FACTORS (CONT.)
()
For public holidays, hi
()
= 0, Ti,b = 00:00:00 and
()
Ti,e = 24:00:00
()
For fuzzy holidays, hi > 0.

(Example: days between Christmas and New Year.)
()
()
For partial holidays, Ti,b > 00:00:00 or Ti,e < 24:00:00.

(Example: Christmas Eve.)
114
PUBLIC HOLIDAYS IN 1999

Date
Country and Name
Date
Country and Name
01.01.1999
31.05.1999
U: Memorial Day
05.07.1999
U: Independance Day Obs.
15.01.1999
J: New Years Day

E: New Years Day
U: New Years Day
J: Adults Day
06.09.1999
U: Labor Day
18.01.1999
U: MLK Day
15.09.1999
J: Respect F. T. Aged Day
11.02.1999
23.09.1999
J: Autumnal Equinox Day
11.10.1999
15.02.1999
J: National Founding Day

U: Lincolns Birthday
U: Presidents Day
22.03.1999
J: Vernal Equinox Day
03.11.1999
J: Health-Sports Day
U: Columbus Day
J: Culture Day
02.04.1999
E: Good Friday
11.11.1999
U: Veterans Day
05.04.1999
E: Easter Monday
23.11.1999
J: Labor Thanksgiving Day
29.04.1999
J: Greenery Day
25.11.1999
U: Thanksgiving Day
03.05.1999
J: Constitution Day
23.12.1999
J: Emperors Birthday
04.05.1999
J: National Holiday
31.12.1999
J: New YearsEve(Bank Hol.)
05.05.1999
J: Childrens Day
115
FUZZY AND PARTIAL HOLIDAYS

Date
Country and Name
Time Period
15.09.1999
whole day
0.3
hline 24.12.1999
E: Christmas Eve
whole day
0.5
U: Christmas Eve
00:00:0012:00:00
0.5
U: Christmas Eve
12:00:0024:00:00
0.1
15.09.2000
whole day
0.3
31.12.1999
E: New Years Eve
whole day
0.5
U: New Years Eve
00:00:0012:00:00
0.5
U: New Years Eve
12:00:0024:00:00
0.1
Factor
116
WEEKENDS
During weekends large returns occur quite often.
117
WEEKEND RETURNS
0e+00 4e04 8e04
USD/DEM weekend squared weekend returns
1987
1989
1991
1993
1995
1997
1999
0.00002
0.00010
USD/JPY squared weekend return
1987
1989
1991
1993
1995
1997
1999
118
WEEKENDS (CONT.)
Jumps cannot be accounted for by mean weekend

activity.
Without special treatment they deteriorate the
deseasonalisation result
Cutting out returns destroys consistency of return series.
Therefore define weekend volatility:

r

v (w) [] = rt(end) [Tw ]
,
w
Tw
(end)
with weekend length Tw = tw
(start)
tw
119
AUTOCORRELATION FUNCTIONS
OF ABSOLUTE RETURNS
0.6
0.5
0.5
0.3
0.2
0.1
0
0.4
XDAX
0.4
USD/JPY FX spot
0.6
0.3
0.2
0.1
0
-0.1
0 1.4 2.8 4.2 5.6 7 8.4 9.811.212.6 14

time [days]
0 1.4 2.8 4.2 5.6 7 8.4 9.811.212.6 14

time [days]
ACF OF DESEASONALISED HOURLY ABS. RETURNS
0.2
0.1
0.0
0.0
0.1
ACF
0.2
0.3
USD/DEM and USD/JPY
0.3
USD/DEM
200
400
600
800
1000
200
600
800
1000
800
1000
0.2
0.1
0.0
0.0
0.1
ACF
0.2
0.3
USD/JPY
0.3
USD/JPY and USD/DEM
400
1000
800
600
Lag
400
200
200
400
Lag
600
121
V. DEPENDENCE STRUCTURES
122
MOTIVATION
The Goal:
Studying the dependence structure across time scales
Why?
The change of the behavior as a function of the time horizon

contains important information
Improves extrapolation from small to large time horizons
Requires:
Characterising dependence for horizons from minutes to months
Here:
Restriction to high-frequency region (1 hour 1 day)
Pecularities of high-frequency data are taken into account
123
QQ-PLOTS FOR USD/DEM
10
5
0
5
4 hou s e DEM
10
5
0
5
10
2 hou s e DEM
5
0
5
10
1 hou e DEM
10
.
.
.
..
...
.
...
....
.
..
.
.
..
.
.
........................
......
.....
.............
...
........
.....
.....
.....
..
....
..
...
.............
..........
.............
......................
.
..
.
.
.
..
...
...
....
..
Quan es o S anda d No ma
4
4
6 4 2
1 day e DEM
2
6
2 0
12 hou s e DEM
5
0
5
10
8 hou s e DEM
124
5
0
5
10
4 hou s e JPY
10
...
..
.
.
...
...
.....
.
.
.
..
.
..
.
.
..
..
.
.................................
.......
....
......
.....
..
...
.....
.........
.....
...
...
...
...............
........................
.......
.
.
...
..
....
.
..
2 hou s e JPY
5
0
5
10
1 hou e JPY
10
QQ-PLOTS FOR USD/JPY
4
1 day e JPY
4
2
0
12 hou s e JPY
5
0
5
10
8 hou s e JPY
125
DEPENDENCE STRUCTURE MODELLING

FOR THE WHOLE DATASET
Exploratory analysis (scatter plots)
Families of copulas across time scales
Tail coefficient estimates
Goodness of fit and ellipticality test
126
SCATTER PLOTS OF DESEASONALISED RETURNS

1 Hour re urns
.
.
. . .. . . .
. . .. . . . .
... . . .. ... ..... .
.. . . ................................................. . ... .
.
.
.
.
.
.
. . . . .. ................................................................... ..... .
.. .................................................................
............................................... . ..
..............................
. ..............................
......
.............
...................... . . . .
.
...............
.................................
.....
......
....
....
..
....
...........
....
....
..........
.....................
..
......
....
.
..
..
........
. . ............................................................
..
.
.
..
.
.................................................... . .
.
..
..
..
.
.
.
.
..
.
.
..
.
..
.
.
.
...
.
.............
...........
...
...............
...
..........
....
...
.....
...
...
.....
...
..
.....
........
.......
..
..
....
...
..
.............
...
......... . ........
..
...
.....
.....
......
........
........
..
..
...
...
....
...
........
....
.............
....
.................... .....
...
....
.....
....
....
..
......
...........
..
........
..
.......
.
.
.
.............
..
.
.
.
.
.
.
.
.
..
..
.
.
.
..
..
..
.........................................................................
.
..
.
.
.
..
.
.
.
.
..
..
.
.
.
.
..
..
.
.
.
..
.
..
..
.
.
..
.
.
.
.
.
.
.
.
.
.
.
.
.
........................................... . . .
.........................
.. .......................................................................
................................................................... .... . .
.
. .... .........................................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.......................................................... . . .
.
.................... ....... . .
. ..
.
.
.
.. ............ ...... .... .... . .
.
.
. . ..
.. ..
. . .. . . .
.
.
.
.
2 Hours re urns
4 Hours re urns
USD JPY
USD JPY
USD DEM
8 Hours re urns
12 Hours re urns
1 Day re urns
USD DEM
USD JPY
USD DEM
USD JPY
USD DEM
USD JPY
USD JPY
USD DEM
USD DEM
127
COPULA: DEFINITION
Given bivariate cumulative distribution function FX (x1 , x2 )
The copula of FX is defined as
CF (u1 , u2 ) =
1
1
FX FX
(u
),
F
1
X1 (v2 )
1
Standardized description of the dependence structure

The copula density is uniform if and only if X1 and X2 are
independent
128
4 Hours returns
........ ... ...... . ..... ......... . .. .. .... ....... ..............................................................................................

. . ... .... . . . ..... ...... ... ......... ... ....... ................................... .......................................
. ......................... ................................................................................ ............................................................................................................................................................................
.. ................... ......... ..................... ......... ...................................................................... ......................................
.. .
..................
. ..... . . . .. . .... .. . .. ..... .......... .... ............
.... .......................................................................................................... .........................................................................................................................................................................
...... .................................................................................................. ........................................................................................................................................
...... .. .... ......... ................................ .. ..............
..................... ........ .
.... ............................
....................................................................................................................................................................................................................................................................................
...... .... ...... .................................................................. ................ ............................. ..........................................................
........................ .......................... ............................. ........... ........................ ......................................................
.. ..... ........ ......................................................................................................................................................................................................................................................
............................................... ......
........................................................................... ................................. ..................................... . ..................
..
....... ................ .......... .................................. ...
............................................................................................................................................
... ... .... .. .... . ... ... .... ..
....... ........................................................................................ ...........................................................................................................................
..... ......... .................. ... ............. .............. ............................................................... ........... ... ....
............................................................................................................................................................................................................ ....................................................................
.. .. ... . ... .. . ..... . ..... . . ...... . . .... . ....... . . . .. ... . ... .
.......................................................................................................................................................... ................................................................................................................................. .....
...................................................................................................................................... ............................................................... ......................................................
........... .... ................. ...... ....... ....... ........ .... .. ......... .. ... .. . .. . . ..
................................................................................................................................... ................. ........................................ ..................................... .
............................................................................................... ......... ..... ............................ ...... .. .......
................................................................................................................................................................................................................ .............................. .......
........................... ...................... ..... .... .... . ... .. .... ... . .. . ..... . .. .
.................................................................................................... ................... . ........................ ....... ........ ...... .. ....
.. .... . . ..... . . . .. . . . .. . ... . ... ....... .. .....................

. ............... ..... .... .............................................. ..............................................................................................................................................................
.
. ...............................................
. .. .
....... ......... .................................................... ....... .................................................................................................
..............................................
. . .. . . . ..
. . .. . .
. ........................ ..... ................ .......... . ..........................................................................................................
. ......... .................................................... ................................... .........................................................................................................................................................................
...................... ................................................ . ..................................................................................................................
. .. . . . ....... ..... .. ............... . ..... ........... ................................. . ..
...............................................................................................................................................................................................................................................................................................
..
. ... . .. . . .. . ..
. . . . . ... . ... . .
... ......... ............................................................................................................................................................................................................................................................................
..................... .. .............................................................................................................................................................................. ............ .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .... . . ... . ...... ...... ... ........ .... ............................ . ....
.......... ......... . .
... .. .. .................................................................................................................... ..................... .
. ...... .................................................................................................... ................................................................................. ................ .. ...
.
.
.
.
.
. .. . .
.
.
.
.
.
.
.
.
.
.
.
. .. .
.. .
.
... ......... .... ............................................................................................ ........................................... ... ..............
.. ..... .. ........................................................................................................................................................ ... . ..
... ............................................................................................................................................................................................................................................................................................
...................................................................... .......... ...................................................... ............. ........ . .....
................................................................................................................................................................................ ........... ................ .......
. .
. .
.
. .
.
............................................................................................................................................................................................................................................................... ..... ...
............................................................................................... ............................................................. . ..
..................................................................................................................................................................................... ....................................................... ....
.................................................................................................................................................................................... .... ...................... ..........................
............................................................ ...................... . . ... . .... .. .. ......... .. ..... . ... .....
............................................................................................................................................ ...... .......................................... .... ................. .. ....... ...
. .. .
. ..
.................. ... . .. .. . .
.
. .. .. .. . .. ...... . . . .. . . .. .... . .. . .... ..... ......... ...............

. .. ....... . ...................... ................. .. .........................................................................................................................................................................
. . . . .. .... ... . . ... ... .
...
. . ............ ... . ..
........... .. ............. . ......... ...................................................................................................................................................................................................................
.. . ..... . ... . .... . .... .... . . .... . ........... ...... ... ..... ...........
.......
....... .. .........................................................................................................................................................................................................................................................
......... ........................................................................................................................................................................................................................................................................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .. . .... . . . . . . . . .. . . .. . .. .. . .. ..
................ .................................................................................................................................................................................................................................................................
.
.. .. .. . . . ... . . .. . ... . .. .. .
.... .........................................................................................................................................................................................................................................................................
. ...... ...................................................................................................................... ........................................................ .....
... ... ..................................................................... ........................................................ ................................................
. . ....... .... ...... ........ .. . . . .. .........................................................................................
.. ......... ............ .................................................................................
....... ..... ....... ..
.............. ..........................................................................................................................................................................................................................................
........ ..................................................................................................................................... ..... ....... .......... ...
.... .............................................................. ................. .. ..................................... ....... ....... .... ... . .
........................ ............................................................................................. ................... ................. ...
..................................................................................................................................................................................................................................... ....................... ......
........................................................................................................................................................................................................................................................................ .....
..................................................................................................................................................................................................................................... ............. ...
.
....... .. ..... .. . .... .. . .... . . ... ... . .
............................................................................................................................................................................... .................................................................................. ..
.. . ....... . . . . . .... .. . .. . ..
.. . . .. .. . . . . . . .
................................................................................................................................................................................................................................................... ..... ..
............................................ .. ..... .... .. . ... .. ............ . ... . . . .
................................................................................................................................. .................. ........... ............ . .......... ..... ...... .
............................................ ....... ....... .. . . ... . .. ........... .. ... ...... . ... .. .. ... . .
USD/JPY
2 Hours returns
USD/JPY
1 Hour returns
USD/DEM
USD/DEM
8 Hours returns
12 Hours returns
1 Day returns
. . .. ...... .
...
. . . . ..
. . . .. . .. .............. ......................................
......
. ... . . .. . . . . . .. . ..... .... ...... ....... ...............................................
. .... . ..... . ............ ............................ ......... ...............................................................................................................................................................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.. . .
. . .
...
.
.
.
....... ......... ..... ...................... ........................................................................................................................................................................................................
.. . . . . .. . . .. ... . . .. . .... ....... .. ......... . . . ...........
.. . .. ..... ....... ............................................................................................................................................................................................................................................
. ............................. ......... ............................................................... ..............................................................
. ....
.
.
..
.... ....... . ... ... ................................................................................................................................... .........................
. .. ... . .......................................................................................................................................................................................................................................................................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
....... ......
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.. .
.
.
.
.... .........................................................................................................................................................................................................................................................................
. .. .. ... .. . . . . ......... . . .. . . . . .... . . .
.. .. .................................................................................................................................................................................................................................... .....................
. .................................... .................................................................................................................................................................................. ............ ....
. ........... ................................ .... ............... . ............. ........... .............. .
................................................................................................................................................................................................................................................ ................... .....
........ .................. ........................................................................................................... ... ... ...................................
. ...................................................................................................... ............... .................... ..... ... ....... .. ......... ..
................................................................................................................................................................................................................................................................. ...... .
.
.
. . . .. . ... . . ... . .. . .
..................................................................................................................................................................................................................................... ...................... ........ . . ..
.................................................................. .................... ...................... ............................ ....... ...... ..... . .. . . .
........................................................................................................................ ... . ... . ...... ... . .. .. ... .
................................................................................................................................................................ .................... ............. ........... ... ..... .......
..........................
.................................................... .................... ........ . ... ..... . .. .. . .. ..
................................................................................................................................... .............................. ............ . .......... .. ..... ..... ..... . .
.. .
.............. ....... .......... .... . . ... . .. . . . ... .. . . . .
.. ..
. .. . . ...... ... . . ... . ...............................................
. ... .. .... ..... ... ... . .. ... ........ ........................................................... ......................................................................
. . . .... .... . ... .......... ................... .. .............................................................................................
.. . . ...... .. .. ......... .. ........ ..........................................................................................................................
. . .. .
. .. . ... . . .. . . . ... .. . . ... . . ..
... ... ........................ . .................................................................................................................................................................................
.... ... ..... ................ ....... ..... ............... ........................................... ........... ....................................... ..
.
.
.
.
.
.
. .. . ... .. .. . .. .... ...... . ............ ....................... ...........................................................................................................................
.
. . .
.. . .. .
.
.... ..... ..................... ..................................................................................................................................................................................
.. .. .. . ..... . . . ... ................ ...... . ....................... ... ... ......... ............. .... .
.. ............ .... ........... . ..... ..... .................................. .... .......................... .............. .........
. . ... . ........................................................ ........................................ .......................................... .
. . .. . .. . ...... ..... .. ......... ............. .. .... ... .. . .... . . .. .
. ................................................................................... .................................................................................... ...................................... ..
......................................................................................................................................................................................... ......... ... .......
.. . ...... . ................................................ ................................... ..... ...... .. ...... .. . .. ..
. ..... . . . . .... . . . ... ..... .... .... . . . .... . . . . . ..
....................................................................................................................................................................................... ........ ............ . .
. ..... ...... ... ............. ......... ..................... .. ... ...... . . ... . . . .. ... .
. .......................................................................................................................................................................................... ............... .. .
.... ........................................................ ....... ............. .. .................. ... .... . .... .... .. . . . .
. ..... ...... ...... ...... ... .............. ...... ..... . . . ...... .. .... . ... .
................................................................................................................................................. .......................... ... ............. .... ... .. .
. . . . ..... .... .. .. . . . ... .. . .. .
...................................................................................................................................... .... ..................... ............. ...... .... .
............................................................................................................................. ............... . .. .. . ...... . . .. . .. ..... . . ...
.
.
.
. ..
......................... ...................... . .......... ... .. .. . .. . . .. . . .
. . .
.
.... . .... .. .. .. .. .. .... .. .......... ..................
. . . ... . .. . ..... .. ....... ... ... . .........................................................................
. . .. .. . ............ ..... ........ ............ ......... ... .......... .......................................................
. . . .. ..... . ... .. ..... .. .... ........ ........ .. ....... ........ ............ ......................
. . . . . .. . .... ........ ...... . .... ........... ............ ......... ........................................
. . . . . .. . .
. ...... ... .... .. .. ............. .. ..... ...... ............................................... ....... ............................. ........
. . . . . . . .. . .. . . .. .... . . .. .. .. ......... ... .
.
. . . .... . .. .. ........... ............................... ............ ......................................................... .... .
. . . . ........... .. . ........................ ........ ............ ........................................ .............. ....
.... ... . .. ..... ............ ... . ..... ......... ............... . ..... .. . .
. . . . . . ...... .. ..... . ... . .. . .. .... . .
.... ............ .................... .................... .................................. ..... .. ....................... . ....... ..
.
. . . .. ... ............................. ...................... ............................................. ........ .... . . .
. . . ... .. .. . . . ..... ... ... .... . . ... . .. . .. . .. . . .
. . ............. ..................................... ................. .......... .................. .................................. . .. .. .
... ..... ..................... ......................... .............. .. ..... .. ..... ... .... .... ... . .... .. .
.
. ... . . . . .. . . . ... . . . .. . .. . .
.
..... ....................... .................................. .. .............................................. . ..... . .... . .. .. . ..
.. .......... ............... . ........... ... .. .......... ... .... .. ........ .. ........ . .. . . .
....... . ........... .. ....................... ....... .... ........ ... . . . .. ... . . . . . .
. . . . . . . .. . . .
.
..................................................... ...... ....... ........... .......... . .. .. . .... . ... .... . . .
. ................................................................... ........................ ... ..... ... . . ..... .. .. . .... ... ...
. ...
........................... ... . ...... . ..... . .. .... .. . . .. . .... ..
.................. ..... ..................... .. . . . ..... . .... .. ...... . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
. .
..................... ....... ... . . . . .. .. .. . . . . . . . . . . .
USD/DEM
USD/DEM
USD/JPY
USD/DEM
USD/JPY
USD/JPY
USD/JPY
COPULA DENSITY OF DESEASONALISED RETURNS
USD/DEM
129
PSEUDO OBSERVATIONS WITH NORMAL MARGINS

4 Hours re urns
USD JPY
2 Hours re urns
USD JPY
. .
..... .
.
... . . . . . . .................... .
.
. . . . . . .. .................................. ................ .... .
..
... . . .. .. .. ........ .....
. . .
. .. ... . .......... ................................................................................................
. . .... ................................................................................................................................................................................................................................. ...
. ...... ... .............................................................................................................................................................................. . .
.
................................... . .
. .... .............................................................................................................
. ........... ..................................................................................................
................................................................................................................................... .
.. ... . ..
............
.......................................................................
. . .. ...............................................................................................................................
............................................................................................................ ..
......................
......
..............................
. ...... ............................................
............................................
.....................
.
.....
. . ... ....................................................................
.....................
............................................................................................... ..
......
.....................
.....
...
.........
................
.......
...........
.....................
.................................................................................
... ......................................................
......................... .. . ..
.....
..........
.......................................
.....
........
........
........
.................
.......
....
........
.........
..........
............................
........................
..
.....
.... .........................................................................
......
................................................
.
.
.
..
.
.
.
.
.
.
.
..
.
........................................... ... . .
.
.
..
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
....................................................................................................................................................................................................................................................................................... . ...
.
. .. .. ............................................................................................................................................................................. ..... .
....... .. .. . .. ........... .. .. . . .
..... ........................................................
........................................................................................................................................................................................................................................................................... . .. .
.....................................
.. ......................................................................................................................................................................................................................................................................................... .. . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. ... ................................................................................................................................................................... ....... . . .. . .
. .................................................................... ......... ........................... .. ....... .. . .. . .
.............. . .... . . . .
. . ......
.
.
.
.
.
.
.
.
.
.
. ...... . .. ..
..
. . .. .
.. .
USD DEM
8 Hours re urns
12 Hours re urns
1 Day re urns
USD DEM
USD JPY
USD DEM
USD JPY
USD DEM
USD JPY
USD JPY
1 Hour re urns
USD DEM
USD DEM
130
FAMILIES OF COPULAS
Gaussian copula for correlation :

CGa =
1 (x)
1 (y)
ds dt
(s2 2 s t + t2 )
p
exp
2
2(1 2 )
2 (1 )
t copula for degrees of freedom and correlation :

t
C,
=
t1
(x)
t1
(y)
ds dt
p
2 (1 2 )
1+
(s 2 s t + t )
(1 2 )
(+1)
2
131
FAMILIES OF COPULAS (CONT.)
Gumbel copula:
CGu
n
o
= exp ( log x)1/ + ( log y)1/
Clayton copula :
CCl
n
o
= max ( log x)1/ + ( log y)1/ , 0
Frank copula:
CF r =
1
(e
log 1 +
1)(e
e 1
1)
132
0.0
0.2
0.4
0.6
Clayton(1.1) copula
0.8
1.0
0.6
0.8
1.0
0.8
0.6
0.4
0.2
1.0
0.0
0.2
02
04
06
08
0.6
0.8
1.0
08
10
Gumbel(1.54) copula
... .... .. .... .. ........... .. ......................... ...... . ...................................................................................................

... . ... .. ... ... . ... .............................. ..............................................................................................................................................................
.. . ... .... ... . . .. . ... .... .......... .... ......... ............. ...........................
. ..... ... ...... ............................................................................................................................................................................................................................................
.......... ..............................................................................................................................................................................................................................
.. .. ... ............................................................................................................................................................................................................................................
.. .. ... ... ........... ....... ....................... ........................................ ...........................................................
..... ................. .............. ...................... ........................................................ ....................................................................
. .. .. .
.... . .. ............ .. ... ... ......... .. ... .. .............. . .. .
....................................... .............................................................................................................................................................................................................................................
.... ....... . . ............................................................................................ ................................. .............................
.......................................................................................................................................................................................................................................
..... ... .... ....................................................................... ....................................................................................... .......... .
.. .......................................................................................................................................................... ........ .................
... . . .
. . . . . . . . . .. . . . .
............................................................................................................................................................................................................................................................................... ......
........................................................................................................................... ......... ......................... ........ ... ....
.. .
. .
. .
.
............................................................................................................................................................................................................................................................................ ...................
... .... .... .... ... . . .. .. . .. ... . .. ..... . ... . . . . ..
...................................................................................................................................................................................................................................................... .................... ..
....................................................................................................................................................................... .......... ..... ............ ..... ...
................................................................................................................................................................................................................................... ............... .. . ..
............................................ ..... ... ...................... ............ ..... .. ... .... .......... .. . ...
............................................................................................................................................................................................... .................................. .. . ....... ..... .. .
.. .. ...... . . .. .. .... . .. .
.. . .
............................................................................................................................................................................................................................. ... ... ... ......... .. .... .
........................................................... ....... .............. ...... ................................. . .... . .. . ....... . ....
.......................................... ....... . ...... ...... ... . .... .. .. . . .. . .
00
0.4
10
F ank 3 5 copu a
10
10
08
06
02
04
e 2
. . ..
. . . .. . ..
. .. ... ........................................................................................................................................................................................................................................................
. .. .. . ...........................................................................................................................................................................................
... ........... .......................................... ...... ...................................................................................... ................................
. . . . ..................................................................................................................................................
. .. .. . . .
.
.
. . . ..
. . .................. .......................................................................................................................................................................................................................................
. .. .... .. ............. .... .............................................................................................................................................................................................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . .. ...... ........ ......................................................................................................................................................
.... ..... ................. ..................... ...................................................................................................................................................................
... ...... ... ... ..... . ...... ..... ........................................ ....................... .............. ..........
........ ................................. ..........................................................................................................................................................................................................................
...................... ...............................................................................................................................................................................................................................
. . .... . . . ............... ...... . ......... .. ............. ..... ....... ..................... ........ ...
... ... ............................ .............................................................................. ................. ........... ......
.. ............ .. .. ....................... ............................................................................. .....................................
.......................................................................................................................................................................................................................................................................................
... .................................................................................................................... .. ................... . ..................
...................................................................................................................................................................................................................................................................................
............................................................................................................................................................................................................................................................... .......
........................................ ..................... ............... ... .......................... .. .... ......... ... .... ...
. ...................................................................................................................................................................................................................................... ..... .. .............................
......................................................................................................................... .......... ....... .. ........ . ....... . ........
............................................................................................................. ................................ ......... ... . ........ ............ .. ..... . ...
. .
......
.................................................................................. ............ .... ... .. . . . ... . .. . . ... . .
.................................................................................................. ................... ........ . ..... .. .. . .. ... ... ... .. . ... . . .
... . .. ..
0.0
0.4
t copula (df: 4; cor: 0.5)
00
0.0
0.2
0.4
0.6
0.8
1.0
Normal copula (cor: 0.5)
0.2
0.0
1.0
0.6
0.4
0.2
1.0
08
0.8
06
0.6
04
0.4
. . .. . ... ... . . ... ............ .. ...........................................

.. ...
. . . . . .................. .......... . .... ...... .... ..................................................................................................................................................
... .. .. . .. .. . . . .. . . . . .... ..
.. .
... ............... ...................................................... ........ .....................................................................................................................................................................................
. . .. . ....... ......... ......... . ...................................................................
.. ...........................................................................................................................................................................................................................................................................
. . ... . .. . ....... ... . ...... . ..... ............ ............ .. ...................................... ....
............... ..................................................... ............................................................................................................................................................................................................
. .. . .. ... .. ..
... .. .
.
.
.
.... ..... ....... ... ................... ... ........................................................................................ ................ .
......................................... ..............................................................................................................................................................................................................................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
................ ............. .... ............... .................................................................................................................................................... .... ..
......... .......................... ................................................................................................................................... .............................. ........
................................................................................................................................................................................................................................................................... .........................
............................................................. .................................................. ............................ .................................. .... ..
.
. .
. .
.
.. . .. . .
... .................................................................................................................................................. ........................................................... ... .
................................................................................................................................. ................................. ........ .. .. .
.............................................................................................................................................................................................................................................................. ........... .. ..
.............................................................. ....................................................................................... ............ ....................... . .... . .
......................................................................................................................................................................................... ............................ .
................................................................................................................................................................................................................................... .................................. .
......................................................................................................... .................................................................... .................... .............. . ......
............................................................................................................................................................................................................... ..... .....
.. . . ...... ........ ...... . . ...... . ..
..............................................................................................
............... .... .. ....................................... ................................ .... ..... .... ..... .. .... .... ... ..
........................................................................................................................................................... ........... ... . ..... ...... ..
02
0.2
....... . ... ........ ............ . . ..... . .. . ..... .... .. ... ... ... ........... ......................................
... ........... . . . .. . .. . . . ....................... ...................................................................................
....... .. .... ............. ..........................................................................................................................................................................................................
. . ...
............................................................ ......... ............................................................................................................................................
..................................
....... ...................................................................................................................................................................................................................................
...... . ... .... ................ ................................................................................................................. .............................
.... . .... . .. .. . .. ...... .. ........ ........ .... ... ............................. .
..... ................................................................................................................................................................................................................................................................................
... ... ..... ............................................... ..................................................................................................................
.. .................... ..... ..........................................................................................................................................................
. .. . . ... .. ...... .. ............. . . . .... ....... ........ ..... ... ..
....... ............................................................................................................................................................................................................................................................ ..
.. .............................................................................................................................................................. .
.................................................................................................................................................................................................................................................................................
. .. ..... .. . ... ....... . ........... . . ... ... .. . ..... .............. ...... ... .. ... . ..
............................................................................................................................................................................................................................................................................................
.........................................................................................................................................................................................................................................................................................
. .............................................................................................................................................................................................................. ............................. ..................
.. .. . . .. .......... . .... .......... ... . ... .. .. . . . . .. . .
............................................................................................................................................................................................................................................................................................
. ............................................................................ ............................................................. .............. .................... .. ..
....................... ............................. ..................... .... ...................................... ...... .. ..... .. ... ..
........................................... ..................................................... ...................... ....... ........ ........... . .
................................................................................................................................................................................................................................................ ...... ................
........................................................................................................................................ ... .......... ............ . ..... ..... ... ...
.......................................................................................... ............... ... ........... . . .. .......... . .. . ........
................................. ...... ........ ... .... ..... .. .. ... . ..... ...... ... ......... ...........
00
0.0
0.8
. .. . . . .. . .. .... ..... ....... ............. .............

.
. ........ .. ............ .................................................................................................................................................................................................................
.. ..... .. .......... .. .......................................................................................................................................................................................................................................
.
.
.
.
.
.
. .. . .... ..... ....... .. .............. ............... ......... ....... ............. ..
....... .. ........ ............................. . ...................................... .......................................................... ................................................................................................
. .. . . . ......... .... .. ...... ....... .. .. .... . .. . .. .... .. . . ..
. ....... .. ....... ..................................................................................................................................................................................................................................................
...................................................................................................................................................................................................................................................
... .... .. ... ......... ... ........ ... ........ ...................... .............. . .. .. ..
........ ...................................................................................................................................................................................................................................................................................
. .. . ... . .. . . ...
. .. . .. . .. . .. . .
.......... .......................................................................................................................................................................................................................................................................................
...................................................................................................................................................................................................................................... .......
. ............. ..........................................................................................................................................................................................................
... . .... ........... ............................. .. ................. .............. ...... .............. .. ..... .. .
....................................................................................................................................................................................................................................................................................
............... ................... .............. .. ....... ...... ...................... ............... ....... ................ . .
............................................................................................................................................................................. ......... .. ... ..
..........................................................................................................................................................................................................................................................................................................
........................................................................................................................................................ ........... ................. .............. ...... ....
.............................................................................. ................................................. ... .......... .................... ............ ... .
.................................................................................................................................................... ................... ...... ....... ..... ...........
.. ... . . . .. . . .
. .. ... .
.............................................................................................................................................................................................................................................. ......... ..... ..
....................................................................................................................................................................................................... ............ . .....
............................................................................................................................................................................ ......................... ....... .... ... . .
.....................
....................................................................... ............... ... . .. ... . . . .
0.0
0.0
0.2
0.4
0.6
0.8
1.0
COPULA DENSITIES FOR SELECTED COPULAS
00
02
04
06
ndependence copu a
133
. ..
. .
. . . . .. .
. .. .... . ....... . . .. ... .
. . ....... ......... .... .. ... . .
.
..
. . . .... . .. .......... ........................................................................... ...... .... ...
.
.
. .... . . ... .... ... ... .
. .. . .................................................................................................................................................................................... .
... .. ....
........................ ..... . .
. . ...... ..........................................................................................................................................................
.................... .. . .
. ..... ..............................................................................................................................................................
............................................................................................................ . .. ...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.................................................................................................. . ... .
. .. .................................................................................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.. . ...... ..................................................................................................................................................................................................... .
...................................................... .. .. .
. ... ........ ......................................................................................................
. .. .. . . .. ............................................................................................ ..... .
. . . ... .......................................................................................................................................................................................................................................................................... ... .
....................................................................................................................................... .
.
.
.
.
.
.
.
.
.. . .........................................................................................
. .. . ... .
. . . ......................................................................................... . .......... . .. .. . .
.. ......... ................................... . .... .. . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . .. .. ............... .. ................... .... .. ..
. . . .. . . . . . .
.
.. . . . . . .. ... . ... . . .
..
.
.
Clayton(1.1) copula, N(0,1) mrg.
.
4
2
0
..
Gumbel(1.54) copula, N(0,1) mrg.
..
. . . .
... .... ... . .. . ..
.
. . . .. .
.. . . ......... ... ........ ........... . .... . . . .
. . ........................................ .............. . ..
.
.. .. ......................................................................................................................................... . .
.. ...................................................................................... .. .
.
. ..... .. .............................................. ..
..
. . .... .......................................................................................................................................................................................................... .......... . .
... .. ................................................................................................................
..................................................... .
..................
. . ........................................................................................................
.
.
.
. ........................................................................................................................................................................................................................ ... .... .
..
....................................................................................... .. . . . . .
.........................
............................................................................
. ......................................................................................................................................................................................................... ...
.. .....................
.................................................................................................................. .. .
......................................................
.
. .. ............................................................................................................................................ . .. . .
.. .......................................................................................... ... ... .
.. .................................................................... ... ....... .....
.. ............................................................................................................... .......... ... . . .
. . ..................................... ..... ... ..... . .
. . ............ . . ..
..................... ... .. ..
......... .. ... .. . ..
.
. .
.. .
.
2
t copula (df: 4; cor: 0.5), N(0,1) mrg.
2
4
.
. . ..
. ... . . . . .
. .... . .. .... ....................... .. .
. . ..
.
. . .. . ... .. .............................................................. .
.. . ... ..... ... ................................ . .
.
. . . ... . ...................................................................................................................................................
.
. ....... .. . .. . .
. .. .......................................................................................................................................................................................................... .
. . .. .. .....................
............................................................... .
.. .........................................................................................................................................................................................................................................................
.. . ... ....
. . . ...................................................................................
.
. .......................................................................................................................................................... ...
.
... . ...............................................................
.. . .
.
.
. .. ............................................................................................................. .. . ..
. .. ... ...................................................................................................................................................................................................................................................... ... .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .............................................................................................................. .. .
.. ... . ................ ....... .................. .... . .. .
. . .......................................................................................................................................................................................................... ..... .
........ ...................................................................................................... ........... ....... ... .
.
.
. .. .......... ........................................ .... . ...
... . .......... ........................................................ . . ..
.
. . ... . . . .... .. .. . . .
. .. . . ......... ... . .. .... .. .
.
... . .... .
.
.
.
Bivariate Normal (cor: 0.5)
qno m e 2
2
0
2
.
4
..
4
2
0
2
4
.
.
. . . ... . . ..
..
... . .. ........ ....
.
.
.
.
.
.
.
.
... .... ... ... ........... .... ... . . .
. . . . . . ... ................. ..... ........ .. . .
. .... . ...... ................................................................................................................................ . .
. . .................................................................................................................... ...... . .
.
. . . . . . ...... ...... ..... ... .... ......... . . .
. . . ................................................................................................................................................................................................................................ ..... ....
.. . . ................ ... ......................................................................................... ...
. . ..... ..................................................................................................
........................................................................................ .
... ... ..........................................................................
... .. .... .. ..
...... . ....... .... .................................................................. . . .
.
. ..................................................................................................................................................................................................................................................................... . ...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... . ........................................................................................................................................................................................ .
... .....................................................................
........................................................................................ . . .
. .. .......................................................................................................................................................................................................... .... . . .
... . ........ ............................................................................................................... ........ .. ... . . ...
.
. ... .... . . .
. . ..
. ................................... ... .. .. .... .... .. .... .... . ...
.
.......... ... ...................... ... ... .. ... . .. .. . . . ... . .
.. . .... ...
.
.
. . . . .. .
.
.
. .... . . .
.
. .
.
.
.
. .
. .. ....... . . ... . . .
. .......... .. ... ..... .... .
. . .... ..... ........... ...........
.
.
.
.
.
.
.
.
.
.
.
.
.
. . .... ........ .. ......... .................... . .. . ... . .
. .. ... ....... ... .......... ... .. . ... .. .
. ....... . ...................................................................................................................................................... .
.. ....... ................................................................................................................................................................................................. ....... ...
. . ........ ......................................................................................................................................................... ......... . .
. . . ..................................................................................................................................................................................................................................................................... .......... . .
................... .. .. . ... . . .
. . . . . .. ..............................................................
.
.
.
. .. ................................................................................................
............................................................................................................ . . .
.
.... ............... ............................................................................................................................. . ...
. .. ... ....................................................................................................................................................................................................................................................................... ..
... . ...................................................................................................... .. . ..
. . ................................................................................................................................................................................................................ ........ ... .
.. ............................................................................................................................................................................................................. ... ....
. . . .. .. ........... ....... ................. ....... .. . .
. . . ........ ............................................................................................. .. .... .
. .. . ............................. ... ..... .... ..
.. .
. .
. .. . . ........ ..... .. . . .... .. .. . .
..
. . .. .
.
COPULA DENSITIES WITH NORMAL MARGINS
F ank 3 5 copu a N 0 1 m g
ndependence copu a N 0 1 m g
134
(4.3)
(4.3)
(4.3)
(4.8)
(5.4)
(5.7)
n
0.05
n
g
n
g
n
g
n
g
n
f
f
g
c
f
t
n
0.10
( AIC(t copula)AIC(copula) ) / n.obs.
0.0
GOODNESS OF FIT FOR DIFFERENT FREQUENCIES
Gumbel
Clayton
Frank
t
Gaussian
c
c
0.15
Ret 1h
Ret 2h
Ret 4h
Ret 8h
Ret 12h
Ret 1d
Frequency of returns
135
TAIL COEFFICIENT ESTIMATES

Frequency
d.o.f.
correl.
tail dep. coeff.
1 hour
4.339
0.563
0.273
2 hours
4.269
0.585
0.291
4 hours
4.282
0.599
0.299
8 hours
4.833
0.619
0.287
12 hours
5.438
0.623
0.264
1 day
5.712
0.624
0.254
Tail coefficient estimates for the DEM and JPY bivariate

returns for the different time frequencies considered:
= lim P (X2 > F21 ()|X1 > F11 ())
1
136
GOODNESS OF FIT AND ELLIPTICALITY TEST

Time Freq.
Prob. Integral test
Pvalues for ellipticality test
Model
pvalue
original margins
t margins
1 hour
2 hours
4 hours
0.01
0.092
8 hours
0.27
0.231
12 hours
0.19
0.034
0.369
1 day
0.74
0.821
0.675
137
PSEUDO OBSERVATIONS WITH t MARGINS

2 Hours re urns
4 Hours re urns
USD JPY
USD JPY
USD DEM
USD DEM
8 Hours re urns
12 Hours re urns
1 Day re urns
USD DEM
USD JPY
USD DEM
USD JPY
USD JPY
USD JPY
1 Hour re urns
USD DEM
USD DEM
138
TAIL DEPENDENCE
ANALYSIS
Spectral measure estimation
Multivariate excesses modelled by
copulas
139
SPECTRAL MEASURE ESTIMATION

Suppose that the d-dimensional random vector X has a
regularly varying tail with tail index
Limit behaviour of X (vague convergence):
P (k X k> tx, X/ k X k )
v
x P ( ),
P (k X k> t)
with x > 0, t , and random vector on the space
(Sd1 , B(Sd1 )).
Distribution function of is SPECTRAL DISTRIBUTION of X.

Alternatively:
measure and positive sequence (an ), an , such that
nP (a1
n X )
()
for n
140
SPECTRAL MEASURE FOR DIFFERENT HORIZONS
Pi/2
Pi
3Pi/2
0.6
0.4
0.0
Pi/2
Pi
3Pi/2
2Pi
Pi/2
Pi
3Pi/2
Angle
8 Hours returns
12 Hours returns
1 Day returns
Pi
Angle
3Pi/2
2Pi
0.4
0.0
0.2
spectral density estimate
0.4
0.0
0.2
0.4
0.2
Pi/2
2Pi
0.6
Angle
0.6
Angle
0.0
0
0.2
0.6
0.4
0.2
2Pi
0.6
4 Hours returns
0.0
0.2
0.4
0.6
2 Hours returns
0.0
1 Hour returns
Pi/2
Pi
3Pi/2
2Pi
Angle
Pi/2
Pi
3Pi/2
2Pi
Angle
141
MULTIVARIATE EXCESSES MODELLED BY COPULAS

Extreme tail dependence copula relative to a threshold t:
Ct (u, v) = P (U Ft1 (u), V Ft1 (v)|U t, V t)
with conditional distribution function

Ft (u) := P (U u|U t, V t), 0 u 1
Archimedean copulas: cont., strictly decreasing function
: [0, 1] 7 [0, ] with (1) = 0, s.t.
C(u, v) = [1] ((u) + (v))
For sufficiently regular Archimedean copulas limit theorem:

lim+ Ct (u, v) = CClayton (u, v)
t0
142
MULTIVARIATE EXCESSES (CONT:)
Data:
1 hour pseudoreturns of DEM and JPY, (F1n (x1i ), F2n (x2i ))
For several thresholds t for hourly pseudo-returns we modelled
Ct (u, v) = P (U Ft1 (u), V Ft1 (v)|U t, V t)
and
Ct+ (u, v) = P (U Ft1 (u), V Ft1 (v)|U t, V t)
Thresholds:
0.03, 0.05, 0.07, 0.1, 0.2, 0.3, 0.7, 0.8, 0.9, 0.93, 0.95, 0.97
BIVARIATE EXCESSES (DIFFERENT THRESHOLDS)
COPULA DENSITY OF BIVARIATE EXCESSES
145
0.0
-0.05
G
g
C
gC
gC
g
g
C
ttt t
C
c
t
C
g
g
C
g
C
g
C
t
Gumbel
Gumbel(-X)
Clayton
Clayton(-X)
t-copula
g
C
t
G
c
t
g
G
GG
G G
ccc c
G
c
c
G
t ttt
gg g
g
C C
C
C
c
c
C
c
0.6
0.8
-0.15
c
c Gc
G
GG
c
G
G
c
-0.10
( AIC(t-copula)-AIC(copula) ) / n.obs.
0.05
0.10
AIC VALUES FOR DIFFERENT THRESHOLDS
0.03
0.2
0.4
1.2
1.4
1.6
1.8
1.97
Threshold
146
RESULTS DEPENDENCE STRUCTURES

The dependence structure for 2-dim., high-frequency FX return
data has been analysed
Methods used:
Copula modelling
Statistical techniques for extremal clustering
tcopula with successively higher degrees of freedom work best

for the whole dataset
However, tcopulas have not enough structure for the shortest
time horizons
147
RESULTS DEPENDENCE STRUCTURES II

Test for ellipticality only rejected for 1 hour and 2 hours returns
if the margins are transformed to t-distributions with the number
of degrees of freedom adjusted to the result of the copula fit.
With the empirical margins, ellipticality rejected for horizons of
8 hours and shorter
Extreme tails best described by Clayton resp. survival Clayton
copula, as predicted by theory

Breymann PDF

Încărcat de

Informații document

Descriere originală:

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Breymann PDF

Încărcat de

Drepturi de autor:

Formate disponibile

UNIVARIATE AND MULTIVARIATE ANALYSIS

OF HIGH-FREQUENCY FINANCIAL DATA

Meielisalp, 8-12 July 2007

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

WHY STUDYING HIGH-FREQUENCY DATA

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

ANNUALIZED USD/JPY VOLATILITY

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

WHY STUDYING HIGH-FREQUENCY DATA

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

WHY STUDYING HIGH-FREQUENCY DATA

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

Meielisalp, Switzerland, 2007

METHODOLOGICAL REMARKS ON THE EMPIRICAL

Important: Development of statistical methods with as few

Meielisalp, Switzerland, 2007

METHODOLOGICAL REMARKS (CONT.)

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

TIME SCALES AND SAMPLE SIZES

14 years Olsen tickby-tick data

intra day effects

breakdown of Permanence Hypothesis

cotton prices 18801940 (Mandelbrot, 1963)

commodity prices from England and Holland

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

Time Series and Financial Models

Different models focus on different aspect of the markets.

Meielisalp, Switzerland, 2007

INTERRELATING DIFFERENT TIME SCALES

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

FOREIGN EXCHANGE MARKETS: BASIC FACTS

New facts often found first in FX spot data

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

1st International R/Rmetrics User and Developer Workshop

Meielisalp, Switzerland, 2007

FOREIGN EXCHANGE MARKETS: VOLUME

90% of volume attributed to intraday interbank market

Meielisalp, Switzerland, 2007

FOREIGN EXCHANGE MARKETS: NO. OF TICKS

Jan 1999Jun 2001

Jan 1987Jun 2001

Jan 1987Jun 2001

Jan 1987Jun 2001

Jan 1999Jun 2001

Jan 1999Jun 2001

Jan 1987Dec 1998

Jan 1987Dec 1998