Sie sind auf Seite 1von 50

Intraday Market Momentum : A functional

Econometric approach

By Idriss TSAFACK
UdeM

June 12, 2019

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approachJune 12, 2019 1 / 50
Introduction

Motivation
There is a large literature documenting the succes of momentum strategy in the
financial market (Jegadeesh and Titman (1993), Griffin, Ji and Martin(2003),
Grundy and Martin, (2001), Moskowitz et al. (2012) ).

2 Categories momentum strategy:

Times Series Momentum.

Cross-sectional Momentum.

Figure: Momentum Strategy

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approachJune 12, 2019 2 / 50
Introduction

Motivation

The research focused most of the time on very low frequency data (Monthly or
longer).

Limitations : lower significance of the performance statistics (lower sharp


ratio, High trading cost) and it underperform in the aftermath of financial crisis
(Chan (2003)).

Only a few findings about the intraday momentum (Gao et al. (2018), Zhang et
al. (2019), Chu et al. (2019)).

Returns are usually observed as scalar =⇒ the information about the dynamics
is not used.

Opportunity : the whole dynamics of the return may contain useful informations
for return prediction.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approachJune 12, 2019 3 / 50
Introduction

This Paper

Objective : Analyze how the one-day-ahead 5 minutes S&P 500 index curve is
predicted by the shape of the previous day with a fully functional autoregressive
model.

Advantages:

Use the forecasts to tactically adjust the trend identification and the market
timing for portfolio exposure within a trading day.

Take into account the dynamics between time (t-1) and (t) on very short
timeframes.

Opportunity to analyze the return predictability more broadly (How the first 5
minutes return today contribute to predict the return at any given time for the next
day).

Opportunity to introduce other tools such as derivative, acceleration of the return


curve.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approachJune 12, 2019 4 / 50
Introduction

Contribution

Main Challenge : High dimensionality issues =⇒ instability of the estimated


parameters.

Suggest 4 estimations approaches to get more stable estimators : FPCA, FPLS,


FLF, FT .

Provide the consistency and asymptotic normality results of our estimations.

Comparison based on simulations and real data application.

Analyze the economic significance (Utility gain analysis) based on the identified
momentum strategy.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approachJune 12, 2019 5 / 50
Introduction

Results

Based on simulation, the FLF method outperform the others in terms of


estimation of the parameter.

The 4 estimation methods tend to display the same predictive performance and
FPLS sometimes outperforms the others.

The FPLS tend to catch a remarkable R2 of 8% in the first and last hours of the
trading session, that is almost 4 times the one obtained by Gao et al.(2018) and
twice the one obtained by Zhang et al. (2019).

FPCA and FT method can reach an R2 of 5% and 6% respectively in the


morning and 2.5% in the ending hour of a trading session.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approachJune 12, 2019 6 / 50
Introduction

The literature

Market Intraday Momentum: Heston et al. (2010), Gao et al. (2018), Zhang et
al. (2019), Chu et al. (2019)).

Functional Data Analysis : Besse & Cardot (2000); Ramsay and Silverman
(1997 - 2010), Ferrati & Vieu (2006); Benatia & al.(2017).

Functional Autoregressive Regression : Bosq (2000); Antoniadis and


Sapatinas (2003); Kargin and Onatski (2008); Hyndman and Shang (2009); Aue
et al. (2015), Kokoszka & Zhang (2012), Shang (2017, 2018).

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approachJune 12, 2019 7 / 50
Introduction

Table of contents

1 Introduction

2 The Model

3 Asymptotic results

4 Simulation results

5 Empirical data analysis

6 Conclusion

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approachJune 12, 2019 8 / 50
The Model

The Model

Let Pi (tj ) be the 5 minutes closing price of the S&P 500 future at time tj , on a
given day i, with j = 1, ..., 273 .

The cumulative intraday return (CIDR) is defined as

Ri (tj ) = 100 ∗ [ln(Pi (tj )) − ln(Pi (t1 ))] (1)

And the continuous curves are constructed by

X(t) = Ri (tj ) with tj ∈ (5(j − 1), 5j]. (2)

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approachJune 12, 2019 9 / 50
The Model

The Model
-3
price evolution within the year 2017 Univariate Instant return Functional Time series
×10
2700 1.5 1.5

2650
1

2600 1

0.5
2550

Intraday cumulative return


S&P 500 Price level

2500 0.5

Instant return
0

2450

-0.5
2400 0

2350
-1

2300 -0.5

-1.5
2250

2200 -1 -2
Jan 01, 2017 Jan 01, 2017 0 0.5 1
Time of the day Time of the day Time of the day

Figure: Intraday cumulative returns of the S&P 500 index in 2017

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 10 / 50
The Model

The Model

We consider a fully functional Autoregressive model of order one (FAR(1)). The


predictor and response variable are both functions.
Z 1
Xi+1 (t) = Ψ(s, t)Xi (s)ds + i+1 (t), i = 1, ..., N, (3)
0

Xi (t) is the cumulative return function at time t of the day i.


i represents a day for the given sample (Xi ; i = 1,...,N ).
t ∈ (5(j − 1), 5j] represents the 5 minutes discretization within a trading day.
The intervals are normalized in [0, 1].
We are interested in the estimation of the autoregressive operator Ψ and the
forecast of the next day shape.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 11 / 50
The Model

The Model
Functional Time series (day n) Functional Time series (day n+1)
1.5 1.5

1 1

0.5 0.5
Intraday cumulative return

Intraday cumulative return


0 0

-0.5 -0.5

-1 -1

-1.5 -1.5

-2 -2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Time of the day (9:30 AM - 4:00 PM) Time of the day (9:30 AM - 4:00 PM)

Figure: Functional Predictor and Functional response

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 12 / 50
The Model

The model

X = (Xi ; i = 1, ..., N ) a sequence of zero-mean functional random variables


defined on the same probability space (Ω, F, P ).

H = L2 ([0, 1]) the space of all square-integrable functions mapping from [0,1] to
R.

H is a Hilbert-space endowed with an inner product < ., . > and a norm ||.||,
Z 1
which are respectively defined as follows: < f, g >= f (t)g(t)dt and
0
 Z 1 1/2
||f || = f2 .
0

ε = (εi ; i = 1, ..., N ) is a sequence of functional strong white noise ( see


Kokoszka et al.(2016)). (We can also consider the martingale difference
configuration)

We assume that there exists an integer k0 ≥ 1, such that ||Ψk0 ||HS < 1.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 13 / 50
The Model

The model

By premultiplying equation (3) by Xi (u) and take the expectation


on both sides we obtain:
Z 1
E[Xi (u)Xi+1 (t)] = Ψ(s, t)E[Xi (u)Xi (s)]ds + E[Xi (u)i+1 (t)]
0

Since that E[Xi (u)i+1 (t)] = 0, we obtain D = Ψ(K).


We can rewrite the problem as D∗ = K(Ψ∗ ).
Where D∗ and Ψ∗ are the adjoint of D and Ψ respectively.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 14 / 50
The Model

The model

The covariance operator :

K: H →H
f → K(f ) = E[(Xi ⊗ Xi )(f )] = E[< Xi , f > Xi ]

Since that K is Hilbert-Schmidt, the spectral decomposition of K is given by



X
K(s, t) = λj φj (s)φj (t) (4)
j=1

where (λj , φj )j≥1 is the eigensystem of K.

The cross-covariance operator :

D: H →H
f → D(f ) = E[(Xi ⊗ Xi+1 )(f )] = E[< Xi , f > Xi+1 ]

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 15 / 50
The Model

The model

Objective : Estimate Ψ and forecast Xi+1 given Xi


Main features of the model : K and D are Hilbert-Schmidt
operators.
Ψ is the parameter of interest.

D∗ = K(Ψ∗ )
We might like to say that Ψ∗ = K −1 ◦ D∗
Issue : ill-posed inverse problem, in general.
If ill-posed : =⇒ unstable estimation of Ψ and bad prediction of
Xi+1 given Xi .

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 16 / 50
The Model

Our estimation approaches

The 4 alternative estimation methods : Functional Principal


Component (FPCA), Functional Partial Least Squared (FPLS),
Functional Tikhonov (FT) and the Functional Landweber-Fridman (FLF)
methods.

Advantages :
Less tuning parameter to control (1 for each method).
Less costly in terms of computation.
Possibility to analyze theoretical properties of estimators (Convergence
rates and asymptotic normality).

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 17 / 50
The Model

Method 1 : FPCA
Project Xn+1 (t) on the first m principal components selected from the
covariance operator of the functional predictor K, φj (s).
Those principal components capture the variations on the predictor variable and
represent the eigenfunctions of K.
The estimation of the autoregressive operator is given by

X
Ψ∗m (s, t) = Qm,j Ψ∗ (vj )(s)vj (t) (5)
j=1

where Qm,j = 1 if 1 ≤ j ≤ m and 0 otherwise.


The empirical version :
m  Pn−1 
1 i=1 Xi+1 (s)ψ̂i,j
X
Ψ̂∗m (s, t) = v̂j (t). (6)
j=1 λ̂j
n−1
Z 1
Where ψ̂i,j = [Xi (t) − X(t)]v̂j (t)dt.
0

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 18 / 50
The Model

Method 2 : FPLS

Z 1
At each step h = 1, ..., m, we identify a new factor fh = Xi (s)φh (s)ds such
0
that the covariance with the functional response is maximized.
Z 1 Z 1 
max cov 2 Xi (s)φh (s)ds, Xi+1 (t)ch (t)dt
φh ,ch ∈L2 ([0,1]) 0 0

subject to ||φh || = 1, ||ch || = 1, and (7)


Z 1Z 1
φ` (s)K(s, t)φh (t)dsdt = 0, ` = 1, ..., h − 1
0 0

φ1 , ..., φh−1 , c1 , ..., ch−1 are already obtained in the h − 1 previous step.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 19 / 50
The Model

Estimation

{φ1 , ..., φh } are the FPLS basis and are also the weights functions
associated with the FPLS components.

The estimation of the autoregressive operator is given by



X
Ψ∗m (s, t) = Qm,j Ψ∗ (vj )(s)vj (t) (8)
j=1
m
Y λj
where Qm,j = 1 − Qm,j = 1 − (1 − ) and θl are the roots of
θl
l=1
the residual polynomials Qm,j .

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 20 / 50
The Model

Estimation

The empirical version :


n ˆ  Pn−1
Qm,j

i=1 Xi+1 (s)ψ̂i,j
X
Ψ̂∗m (s, t) = v̂j (t). (9)
j=1 λ̂j
n

m
ˆ Y λ̂j
where Q m,j = 1 − Q̂m,j = 1 − (1 − ) and
l=1
θ̂l
Z 1
ψ̂i,j = [Xi (t) − X(t)]v̂j (t)dt.
0

θ̂l is the smallest eigenvalue of the Hankel matrix derived from the
estimation procedure.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 21 / 50
The Model

Method 3 : Landweber iteration

Take the inital value as Ψ0 (s, t) = d ∗ D(s, t)


For each h = 1, ..., m, we have

Ψh (s, t) = (1 − dK)(Ψh−1 )(s, t) + dD(s, t) (10)

d is a positive parameter such that 0 < ||K||op < 1/d


From the estimation, we obtain :

X
Ψ∗m (s, t) = Qm,j Ψ∗ (vj )(s)vj (t) (11)
j=1
 
m
where Qm,j = 1 − (1 − dλj )

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 22 / 50
The Model

Estimation

The empirical version is given by :


n ˆ  Pn
Qm,j

i=1 Yi (s)ψ̂i,j
X
Ψ̂∗m (s, t) = v̂j (t) (12)
λ̂j n
j=1
 
ˆ m
where Qm,j = 1 − (1 − dλ̂j ) .

Alternative version
m
X
Ψ̂∗m (s, t) = d (I − dK̂)l−1 ◦ D̂(s, t) (13)
l=1

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 23 / 50
The Model

Method 4 : Functional Tikhonov

Another alternative approach proposed by Benatia, Carrasco and


Florens (2017)
We have
Ψ∗α (s, t) = ((αI + K)−1 ◦ D)(s, t); (14)
α is a positive parameter which converge to zero as n goes to
infinity.
From the estimation, we obtain :

X
Ψ∗α (s, t) = Qm,j Ψ∗ (vj )(s)vj (t). (15)
j=1

λj
where Qm,j = .
λj + α

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 24 / 50
The Model

Estimation

The empirical version is given by :


n ˆ  Pn
Qα,j

i=1 Yi (s)ψ̂i,j
X
Ψ̂∗m (s, t) = v̂j (t) (16)
λ̂j n
j=1

ˆ λ̂j
where Q α,j = .
λ̂j + α
Alternative version

Ψ̂∗α (s, t) = ((αI + K̂)−1 ◦ D̂)(s, t). (17)

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 25 / 50
Asymptotic results

Assumptions

Xi follows a FAR(1)
Z process.
A1. ||Ψ||HS < 1, X 2 (t)dt < +∞, E[||X||4 ] < ∞, E[εi |X] = 0.
S
A2. εi are strong white noises. E[ε2i |X] = σ 2 , E[ε4i |X] < +∞

X < Ψ∗ , φj >2
A3. We assume that Ψ satisfies < ∞ with β ≥ 0.
j=1 λβj
A4. The eigenvalues of the covariance operators K and K̂ are distinct.
λ1 > λ2 > ... > 0 and λ̂1 > λ̂2 > ... > λ̂m .
nλm
A5. → +∞.
m

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 26 / 50
Asymptotic results

Asymptotic results : FPCA and FPLS

Under assumptions A1 to A5, we have


   
m β
M SEF P CA = Op + Op λm+1 . (18)
λm n
Under assumptions A1 to A5, if ||K||op < 1, then
   
m β
M SEF P LS = Op + Op λm+1 . (19)
θm n
where θm is the smallest root of the residual polynomial Q̄m,j .

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 27 / 50
Asymptotic results

Asymptotic results : FLF and FT

Under assumptions A1 to A4, if ||K||op < 1, then


 2  
m −2γ
M SEF LF = Op + Op m . (20)
n

Under assumptions A1 to A4,


if β > 1    
1 min{β,2}
M SEF T = Op + Op α . (21)

if β < 1
αβ
   
M SEF T = Op + Op α β . (22)
nα2

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 28 / 50
Asymptotic results

Asymptotic results

−β
If m ∼ n1/(2+2β) then M SEF LF ∼ n 1+β .
−min{β,2}
For β > 1, if α ∼ n−1/(2+2β) then M SEF T ∼ n 1+min{β,2} , (see
Benatia et al. (2017)).
For β < 1, if nα2 → ∞, M SEF T ∼ αβ .
To get more information about the optimal number of functional
components m for the FPCA and FPLS, it is necessary to set
some additional assumption on the eigenvalues.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 29 / 50
Asymptotic results

Asymptotic normality : FPCA, FLF and FT

Assume that A1 to A4 (A1 to A5) for FLF and FT (for FPCA) hold, if
E(||Xi ||4 < ∞ and E(||Xi ||2 ||εi ||2 ) < ∞, then

n(Ψ∗δ − Ψ∗δ ) → N (0, Ωδ ). (23)

where δ = m for the FPCA and FLF and δ = α for the FT.
 
Ωδ = E (ε + Kδ−1 ◦ Ψδ (Xi )) ⊗ Kδ−1 (Xi )) ⊗ ¯
 
(ε + Kδ−1 ◦ Ψδ (Xi )) ⊗ Kδ−1 (Xi ))
   
− E Kδ (Ψ∗ ) ⊗ E(Xi ) ⊗E ¯ Kδ (Ψ∗ ) ⊗ E(Xi ) .

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 30 / 50
Simulation results

Simulation results

The following data generating process is considered


Z 1
Xn+1 (t) = Ψ(s, t)Xn (s)ds + εn+1 (t), n = 1, ..., N. (24)
0

(t2 + s2 )
Gaussian Kernel : Ψ(s, t) = C exp{− }, with C chosen
2
such that ||Ψ||HS = 0.5.
Identity operator : Ψ(s, t) = C,
Sloping plane (t) : Ψ(s, t) = Ct,
Sloping plane (s) : Ψ(s, t) = Cs
with (s, t) ∈ ([0, 1])2 and C a constant useful to normalize the
autoregressive operator.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 31 / 50
Simulation results

Simulation results

3 error process configurations :


The Brownian motion (nondifferentiable) innovation :
ε(1) (t) = W (t) − tW (1), where W is the standard Wiener process
Smoothed√ and differentiable
√ √innovation :
ε(2) (t) = ξ1 2sin(2πt) + ξ2 2 λcos(2πt), where ξ1 and ξ2 are
independent standard normal and λ = 0.5.
Combo innovation : ε(3) (t) = aε(1) (t) + (1 − a)ε(2) (t), where
a ∈ [0, 1] is a real constant.
N = 50, 100, 200 and 500.
A grid of 1000 points on the interval [0,1].

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 32 / 50
Simulation results

Simulation results
Estimation error criteria :
s
Z 1Z 1 2
M SE = Ψ̂(s, t) − Ψ(s, t) dsdt.
0 0
Z 1Z 1
AD = Ψ̂(s, t) − Ψ(s, t) dt

0 0

Ψ̂(s, t) − Ψ(s, t)
Z 1 Z 1

RAD(Ψ) = dsdt
0 0 |Ψ(s, t)|
Prediction error criteria :
s
Z 1 2
En = X̂n (t) − Xn (t) dt
0
Z 1
Rn = X̂n (t) − Xn (t) dt

0
By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach
June 12, 2019 33 / 50
Simulation results

Simulation results
The Gaussian Kernel with brownian motion innovations

Figure: Comparison of the different estimation techniques. Gaussian Kernel


with n = 50, and ε(1)

MSE RAD AD
3
4 6
2.5
3 2
MSE

RAD
4

AD
2 1.5

2 1
1
0.5

FPCAFPLS FT FLF FPCA


FPLS FT FLF FPCA
FPLS FT FLF

Rn En
2
2

1.5
Rn

En

1.5

1
1

FPCA
FPLS FT FLF FPCA
FPLS FT FLF

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 34 / 50
Simulation results

Simulation results

Figure: Comparison of the different estimation techniques. Gaussian Kernel


with n = 500, and ε(1)

MSE RAD AD
0.6
0.7
1.2
0.6 0.5
1
0.5
MSE

RAD
0.4

AD
0.8
0.4
0.6 0.3
0.3
0.4 0.2
0.2
FPCA
FPLS FT FLF FPCA
FPLS FT FLF FPCA
FPLS FT FLF

Rn En

1.2
1
Rn

En

1
0.8

0.6 0.8

FPCA
FPLS FT FLF FPCA
FPLS FT FLF

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 35 / 50
Simulation results

Simulation results
Gaussian Kernel with smooth innovations :

Figure: Comparison of the different estimation techniques. Gaussian Kernel


with n = 50, and ε(2)

MSE RAD AD
2500
4000 5000

4000 2000
3000
MSE

1500

RAD
3000

AD
2000
2000 1000
1000 500
1000

0 0 0
FPCA
FPLS FT FLF FPCA
FPLS FT FLF FPCA
FPLS FT FLF

10 58 Rn 10 58 En
10
6
8

4 6
Rn

En

4
2
2

0 0
FPCAFPLS FT FLF FPCA
FPLS FT FLF

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 36 / 50
Simulation results

Simulation Results

Figure: Comparison of the different estimation techniques. Gaussian Kernel


with n = 200, and ε(2)

MSE RAD AD

2000 2500

2000 1000
1500
MSE

RAD
1500

AD
1000
1000 500
500 500

0 0 0
FPCA
FPLS FT FLF FPCA
FPLS FT FLF FPCA
FPLS FT FLF

10 45 Rn 10 46 En
2

10 1.5
Rn

En

1
5
0.5

0 0
FPCAFPLS FT FLF FPCA
FPLS FT FLF

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 37 / 50
Simulation results

Simulation Results
Identity Kernel

Figure: Comparison of the different estimation techniques. Identity Kernel


with n = 50, and ε(1)

MSE RAD AD
2.5
1.2
1.5
2 1

0.8
MSE

RAD
1 1.5

AD
0.6
1
0.5 0.4
0.5
0.2

FPCA
FPLS FT FLF FPCA
FPLS FT FLF FPCA
FPLS FT FLF

Rn En
4
4
3
3
Rn

En

2
2

1
1

FPCAFPLS FT FLF FPCA


FPLS FT FLF

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 38 / 50
Simulation results

Simulation results

Figure: Comparison of the different estimation techniques. Identity Kernel


with n = 500, and ε(1)

MSE RAD AD

0.6 1 0.5

0.5
0.8 0.4
MSE

RAD

AD
0.4
0.6 0.3
0.3
0.4 0.2
0.2
0.2 0.1
FPCA
FPLS FT FLF FPCA
FPLS FT FLF FPCA
FPLS FT FLF

Rn En
1.1
0.9

1
0.8
Rn

En

0.7 0.9

0.6 0.8
FPCA
FPLS FT FLF FPCA
FPLS FT FLF

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 39 / 50
Empirical data analysis

The data

The data of S&P 500 Index future is used to analyze the intraday
returns predictability.
The available sample is from 01/01/2008 to 12/31/2017 collected
from the website www.backtestmarket.com.
Useful to control for other effects : the volatility, the volume, the
aftermath financial crisis, the macroeconomic news release effect
(Federal Open Market Committee (FOMC), Consumer Price Index
(CPI), Growth Domestic Product (GDP)), the infrequent
rebalancing.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 40 / 50
Empirical data analysis

The simple regression

The considered period is 2015 - 2017.


The sample size for this regression is N = 750. This sample is
split into two parts : the regression part (N 1 = 650 days) and the
Validation and testing part (N 2 = 100 days).

Z 1
Xn+1 (t) = Ψ0 (t)+ Ψ(s, t)Xn (s)ds+εn+1 (t), n = 1, ..., 650. (25)
0

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 41 / 50
Empirical data analysis

The simple regression

Figure: The estimated Autoregressive operator. Year 2015 - 2017


Estimated Kernel (FPCA) Estimated Kernel(FPLS)
1 1
next day (9:30 AM - 4:00 PM)

next day (9:30 AM - 4:00 PM)


0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Current day (9:30 AM - 4:00 PM) Current day (9:30 AM - 4:00 PM)

Estimated Kernel (FT) Estimated Kernel (FLF)


1 1
next day (9:30 AM - 4:00 PM)

next day (9:30 AM - 4:00 PM)


0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Current day (9:30 AM - 4:00 PM) Current day (9:30 AM - 4:00 PM)

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 42 / 50
Empirical data analysis

Results

The FT and the FLF methods display the same estimation results that are
significantly different to the one of FPCA and FPLS.

Based on the estimations using the FLF and FLF we have:

The cumulative return for the first half of the current day is positively correlated to
the next day.

The cumulative return for the second half of the current day is positively
correlated to the first half of the next day and predict negatively the second half
of the next day.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 43 / 50
Empirical data analysis

The simple regression

Figure: The estimated Functional R-Squared. Year 2015 - 2017


Functional R-squared(FPCA) Functional R-squared (FPLS)
0.05 0.08

0.04
0.06
R-squared(R2(t))

R-squared(R2(t))
0.03
0.04
0.02

0.02
0.01

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Time of the day (9:30 AM - 4:00 PM) Time of the day (9:30 AM - 4:00 PM)

Functional R-squared (FT) Functional R-squared (FLF)


0.06 0.03

0.05 0.025
R-squared(R2(t))

R-squared(R2(t))
0.04 0.02

0.03 0.015

0.02 0.01

0.01 0.005

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Time of the day (9:30 AM - 4:00 PM) Time of the day (9:30 AM - 4:00 PM)

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 44 / 50
Empirical data analysis

Results

In terms of predictive R2 we obtain

The FPLS tend to catch a remarkable R2 of 8% in the first and last hours of the
trading session, that is almost 4 times the one obtained by Gao et al.(2018) and
twice the one obtained by Zhang et al. (2019).

FPCA and FT method can reach an R2 of 5% and 6% respectively in the


morning and 2.5% in the ending hour of a trading session.

The FLF display an R2 value of 2.5% (nearly similar to the one obtained by Gao
et al.(2018)).

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 45 / 50
Empirical data analysis

The main causes of the intraday momentum

Bogousslavski (2016) : infrequent rebalancing, late-informed effect


(slow diffusion and acceptance of news release).
Chan (2013) : market manipulation of high-frequency traders, forced
sales or purchases of assets

Renault (2017) : high-frequency investor sentiment.


Other tests : the volatility, the volume, the aftermath financial crisis, day
effect, , the macroeconomic news release effect .

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 46 / 50
Empirical data analysis

The main causes of the intraday momentum

The regression equation is given by


Z 1
Xn+1 (t) = Ψ0 (t)+ Ψ(s, t)Xn (s)ds+ρ1 Ie,n +ρ2 Ie,n ∗Rn +εn+1 (t). (26)
0

n = 1,...,750. where Ie,n is a dummy variable taking the value 1 if the


concerned event happened at day n and 0 otherwise. Rn the cumulated
return at the closing of the day n.
ρ2 capture the interaction effect between Ie,n and Rn .

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 47 / 50
Conclusion

Conclusion

Forecasting the shape of the stock market could be very useful to


analyze the return predictability.
We use a novel approach (Functional autoregressive model).
This approach can help investors to optimize on their investment
strategy and portfolio rebalancing.
The FLF method tend to outperform the others in terms of
estimation of the parameters.
It is possible to reach remarkable value of the predictive R2 (up to
8%).
simulations and empirical analysis are promising.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 48 / 50
Conclusion

Still ongoing

Run the robustness check of our empirical analysis (day effect,


High volatility day effect, Macroeconomic news effect, High
Volume day effect, financial crisis effect ...)
Economic significance analysis : Examine the performance of the
identified investing strategy.

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 49 / 50
Conclusion

THANKS FOR YOUR KIND ATTENTION

By Idriss TSAFACK UdeM Intraday Market Momentum : A functional Econometric approach


June 12, 2019 50 / 50

Das könnte Ihnen auch gefallen