Financial Engineering

Stylized Facts

Linda Ponta

1
 Financial Markets
• In financial markets, securities worth over
20,000 billion euros are processed daily!
• There are huge economic and scientific
interests in the studing and modeling
financial markets.
• Economists and mathematicians have
been studying financial markets for at least
100 years
• Over the last 20 years huge masses of
computerized data have been accumulated 2
Arbitrage
• There is arbitrage when you can buy and
sell the same good, without providing added
value, making a profit

• Arbitrage opportunities appear all the time,

but tend to disappear due to the law of
supply and demand

3
Arbitrage example
New York Stock exchange
Arbitrage
Step 1: buy 1 barrel in New York
Step 2: sell the barile in London

1 Oil Barrel earning = 0.37 USD

71.27 USD

London stock exchange

1 Oil Barrel
Exchange rate 39.19 GBP
1 USD = 0.54699 GBP
1 GBP = 1.82820 USD 4
71.65 USD
Efficient market theory
• A market in which information propagates
instantaneously and in which the
possibilities of arbitrage immediately cancel
is called an efficient market
• If an operator found a winning strategy,
others would soon follow and the strategy
will loose its value
• In other words, there is no strategy capable
of yielding above-average profits with
certainty

5
Consequences of efficient
markets
• One consequence of the theory is that all operators
are perfectly informed and rational, and therefore
equal to each other
• If prices change, this is due to the effect of news
coming from outside
• The efficient market separates past behavior of prices
from their future behavior

?
7
However...
• Anyone with some experience of financial
markets knows that the operators are far
from operating in the same way
• Furthermore, there are price dynamics
that clearly cannot be explained only with
the arrival of external news
• The price series of all real markets show
the so-called stylized facts, which are not
obtainable with the theory of efficient
markets 8
The “Stylized facts”
1. For price time series, it is not possible to
reject the hypothesis of "random walk"
2. The returns calculated on intervals from a
few seconds to about a month have a
power distribution (fat tails)
3. Returns are characterized by volatility
"clusters" over time, i.e. periods of relative
"calm" alternate with periods of great
variation
9
The path of the drunk
◼ An example
of a random
walk is the
path of the
drunk

10
Random walk
200

180

160

140

120

100

80

60 11
0 200 400 600 800 1000
tempo
Theoretical Excursus: main
stages

1900 1955 1964

Bachelier Samuelson Osborne

Random Walk Formalization of

Model Interest of the random walk
economists process -
for variation of
1905 Bachelier's stock prices
Einstein doctoral follows
dissertation Brownian motion

Model of the
random motion
of molecules,
known as
Brownian
Motion
13
 Stock prices
90 General Motors
Fannie Mae
80 Pfizer, Inc.

70

60
prices($)

50

40

30

20
02/01/01 07/08/01 19/03/02 21/10/02 28/05/03 12/09/03 14
days(02-Jan-2001 - 19-Sep-2003)
Returns distribution

• For a very wide variety of markets, the

distribution of returns up to 1-2 months is
a symmetrical, non-Gaussian "bell" with
fat tails

The tails of the distribution can be

approximated with a Pareto distribution
F (x) = 1 – a x-
15
 Excursus teorico:
tappe principali

1900 1955
Samuelson 1963 1964 1976 1982
Bachelier Osborne
Mandelbrot Fama Engle

Random Walk
Formalization of
Model Interest of
Pareto Stable the random walk ARCH-GARCH
economists
Distributions process - Process
for
variation of
1905 Bachelier's
stock prices Efficient Market
Einstein doctoral
follows Hypothesis
dissertation
Brownian motion

Model of the
random motion
of molecules,
known as
Brownian
Motion
16
Standard&Poors 500 Index
1600

1400

1200

1000
USD
800 Crisis 1987

600

400

Internet bubble
200

0 17
20-Oct-82 1-Oct-86 14-Sep-90 29-Aug-94 13-Aug-98 8-Aug-02 22-Oct-04
 “Fat Tails” in one day returns of S&P
500 index
60
S&P500

50
Gaussian
Frequency

40

30

20

10

0
-0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 18

Returns
Pareto-Levy and Gaussian
distributions

19
 Volatility clustering
Returns alternate periods of calm and turbulence

0.1

0.05
One-day returns S&P500

-0.05

-0.1

-0.15
Calm period

-0.2
Turbolence period

-0.25 20
20-Oct-82 1-Oct-86 14-Sep-90 29-Aug-94 13-Aug-98 8-Aug-02 22-Oct-04
Auto-correlation of one-day returns
S&P500
1.2

1
Autocorrelation function
Returns
0.8
ABS(Returns)
0.6
Squared Returns

0.4

0.2

-0.2 21
0 10 20 30 40 50 60
Lags
 S&P 500
Composition

USD

◼ Market are
aggregate of
financial assets
◼ Open question
– How to model
and to validate
a multiassets
artificial stock
market?

trading days (2.1.2001 – 3.2.2003)

22
Historical background

1988
1999
Stock & Watson
Plerou, Stanley
et. ali

Common trends
and
cointegration Universal and
Nonuniversal
Properties of Cross
Correlations

23
Multivariate stylized facts
Cross-Correlation Matrix:
The eigenvalues of the cross-correlation matrix almost follow a
random matrix distribution. The outliers, i.e., the largest
eigenvalues, correspond to the financial sectors. The largest
eigenvalue is the market

Variance-Covariance Matrix:
Only a reduced number assets in a market are independent
integrated processes. The common trend are the eigenvectors
derived by principal component analysis of the variance-
covariance matrix. The largest eigenvalue is the market

24
Eigenvalues PDF of the Cross-
Correlation matrix of returns

Random Matrix Theory

Outliers

25
S&P 500 Sectors and
Eigenvectors 0.02
S&P500 close values (02 Jan 2001-- 19 Sep 2003)

Consumer Discretionary
0.018

Telecommunicatons Services
0.016 Materials

Financials-Consumer Staples
0.014
Consumer Staples

Energy
0.012
Financials

Consumer Discretionary
PDF

0.01

Information Technology
0.008

0.006

0.004
Market

0.002

0
1 2
10 10

Cross correlation matrix OutLiers Eigenvalues

26
PCA Eigenvector of prices variance-
covariance matrix in S&P 500
0

-5

-10
ADF t-statistic

-15 Integrated Processes

-20

-25

-30

-35
1% critical value
5% critical value
-40 -2 -1 0 1 2 3 4
10 10 10 10 10 10 10 27
Eigenvalues
Statistical properties of returns

The last thirty years of The returns on financial assets

research have been follow a random walk trend
dominated by the Efficient and the distribution that
Market Hypothesis describes them is normal

however…

… it is an approximation

Identification of statistical distributions that Development of market models that justify

are better able to capture the the characteristics identified
characteristics of financial markets 28
The new frontiers of finance
In the last thirty years ….

Global coin market(1973)

Explosion of derivatives markets

Interdisciplinary research,
which makes use of the
Financial market crisis (1987, 2008…)
knowledge gained in
engineering, physics,
Relevance of electronics and information technology mathematics, economics
and finance
powerful computers

Considerable amount of historical data

29