tail dependence in reits returns - sec 2 · 2019. 3. 15. · d x in in . in v o d x in in in in . d...
TRANSCRIPT
Tail Dependence in REITs Returns
Kridsda NimmanuntaNIDA Business School, Thailand
Kanak PatelUniversity of Cambridge, UK
SEC Working Paper Forum #922nd October 2015
Overview
Nimmanunta & Patel2
Research QuestionsExisting LiteratureThe Generic Models and Data DescriptionsResults and Discussions
Marginal Distributions of REITsConditional CopulasTail Dependence of REITs with t Copula
Conclusions
Resea
rch Qu
estion
s
Nimma
nunta&
Patel
3Non-n
ormalit
y of as
set ret
urn dis
tributio
n (Mills
(1927)
)No
n-norm
ality o
f asset
return
s Joint
distribu
tionSee
e.g. Lo
nginand
Solnik
(2001)
, Ang
and Ch
en (20
02)A f
all in v
alues o
f asset
s beyon
d some
thresh
olds ca
n trigge
r a fall
in valu
es of ot
her ass
ets tha
t are in
itially w
eakly c
orrelat
ed wit
h the
former
Evidenc
e of ta
il depe
ndence
and a
symme
tric de
penden
ce stru
cture c
an be
found
in almo
st ever
y mark
etInte
rnation
al equi
ty mark
ets (Lo
nginand
Solnik
(2001)
)Equ
ity and
bond
marke
ts in G
-5 coun
tries (H
artmann
et al. (
2004))
Curre
ncy ex
change
marke
ts (Pat
ton (20
06))
Asian
develo
ped fut
ure ma
rkets (
Xuand
Li (20
09))
The US
and E
uropea
n CDS
marke
ts (Co
uderta
nd Ge
x(2008
))
Resea
rch Qu
estion
s
Nimma
nunta&
Patel
4In Real
Estate
marke
t,Kni
ght et
al. (200
5) the
poten
tial be
nefits
of diver
sificatio
n bet
ween re
al estat
e and e
quity m
arkets
during
down
turns
disappe
ars.
Real Es
tate Bo
om-Bu
rstDu
ring the
past t
en yea
rs, real
estate
assets
have te
nded to
mo
ve close
ly toge
ther
Tail De
penden
ce betw
een dif
ferent s
ectors
of rea
l estate
Tail de
penden
ceAsy
mmetr
y
Exist
ing Li
teratu
re Nimma
nunta&
Patel
5Time-v
arying
Condi
tional C
opula (
Patton
(2002)
), exten
ding C
opula (
Sklar(1
959))
The de
penden
ce stru
cture o
f asset
s dynam
ically o
ver tim
e by co
ndition
al on th
e past
informa
tionToo
many
copulas
to cho
ose ->
The co
pula w
ith the
highes
t loglik
elihood
value
Patton
(2006)
finds t
he dep
endenc
e betw
een De
utsche
Mark-
USD an
d Yen-U
SDhigh
ly asym
metric
in one
direct
ion be
fore the
introd
uction
of eur
o curr
ency
margin
ally asy
mmetr
ic in the
other d
irection
therea
fter
Jondea
uand
Rockin
ger(20
06) inv
estigat
e the d
epende
nce am
ong ma
jor sto
ck ma
rkets
Depen
dency b
etween
Europ
ean ma
rkets i
ncreas
es over
the sam
ple pe
riod
Goora
h(2007
) exam
ines th
e depe
ndence
struct
ure be
tween
US and
UK list
ed rea
l estate
com
panies
and RE
ITsSym
meter
isedJoe
-Clayto
n (SJC)
copul
a -bes
t fit to
the da
ily data
from 1
990Q1
to 200
7Q1
Lower t
ail depe
ndence
is gene
rally st
ronger
than th
e uppe
r one
In our p
aper, w
e exam
ine the
tail de
penden
ce withi
n US R
EITs
The G
eneri
c Mod
el
Nimma
nunta&
Patel
6Copul
a-base
d AR, G
JR-t-GA
RCH M
odel
The mo
del co
nsists o
f two c
ompon
ents: M
arginal
sand C
opula
are use
d to cap
ture th
e serial
corre
lation
Gloste
n-Jagan
nathan
-Runkl
e(GJR)
-t-GAR
CH can
captur
e three
feat
ures
Volatil
ity clus
tering
Excess
kurto
sis of t
he retu
rn distr
ibution
Levera
ge effec
tCo
ndition
al Copu
la to d
eal wit
h (asym
metric
) tail d
epende
nceGau
ssian, S
tudent
t, Gum
bel, Fr
ank, Cl
ayton, S
JC
The G
eneri
c Mod
el
Nimma
nunta&
Patel
7Specific
ation o
f the M
arginal
Distri
bution
s
The G
eneri
c Mod
el
Nimma
nunta&
Patel
8Specific
ation o
f the C
onditio
nal Co
pula
The G
eneri
c Mod
el
Nimma
nunta&
Patel
9Specific
ation o
f the C
onditio
nal Co
pula
The G
eneri
c Mod
el
Nimma
nunta&
Patel
10The Est
imation
of Par
amete
rs : Two
-step M
LUsi
ng the
exten
sion of
Sklarth
eorem
(Patto
n (2006
))
Hence
, the lo
g-likeli
hood fu
nction
can be
expre
ssed a
s
Data
Descr
iption
s
Nimma
nunta&
Patel
11Daily r
eturns
of the
U.S. eq
uity RE
ITs fro
m 2000
to 201
1:Ind
ustrial
& Office
, Reside
ntial, R
etail an
d Hote
l & Lod
gingTot
al Retu
rn Indi
ces of
FTSE/N
AREIT
North
Americ
a (Base d
ate 1st
Jan 200
0 at
1,000)
Data
Descr
iption
s
Nimma
nunta&
Patel
12Daily r
eturns
of the
U.S. eq
uity RE
ITs fro
m 2000
to 201
1:Ind
ustrial
& Office
, Reside
ntial, R
etail an
d Hote
l & Lod
ging
Data
Descr
iption
s
Nimma
nunta&
Patel
13 Data
Descr
iption
s
Nimma
nunta&
Patel
14
Margi
nal D
istrib
ution
s of R
EITs
Nimma
nunta&
Patel
15AR ter
ms bec
ome in
significa
nt after
includ
ing GA
RCH t
erms
GARC
H(1,2)
for Re
sidentia
l REIT
GARC
H(2,1)
for Ind
ustrial
& Office
REIT a
nd Ret
ail REIT
GARC
H(1,1)
for Ho
tel & L
odging
REIT
In cont
rast to
Cotter
& Stev
enson
(2007)
Levera
ge effec
t is fou
nd for
all four
REITs
Indust
rial &
Office
REIT h
as the
highest
levera
ge effec
tSign
ificant S
tudent
tdegr
ee of f
reedom
Implyin
g the e
xcess k
urtosis
of all f
our RE
ITs
Margi
nal D
istrib
ution
s of R
EITs
Nimma
nunta&
Patel
16
Margi
nal D
istrib
ution
s of R
EITs
Nimma
nunta&
Patel
17Condi
tional V
olatilit
y for ea
ch err
or ter
m of RE
ITs ret
urns
Cond
itiona
l Cop
ulas
Nimma
nunta&
Patel
18
Cond
itiona
l Cop
ulas
Nimma
nunta&
Patel
19 Cond
itiona
l Cop
ulas
Nimma
nunta&
Patel
20
Cond
itiona
l Cop
ulas
Nimma
nunta&
Patel
21 Cond
itiona
l Cop
ulas
Nimma
nunta&
Patel
22
Cond
itiona
l Cop
ulas
Nimma
nunta&
Patel
23 Cond
itiona
l Cop
ulas -
Summ
ary
Nimma
nunta&
Patel
24TV-Stu
dent tc
opula, w
hich ha
s symm
etric t
ail depe
ndence
, provi
des
the hig
hest lo
g-likeli
hood v
alue, A
IC, BIC
in all c
ases.
Hence
, the p
air-wis
e cont
empor
aneous
depen
dence
structu
res of
REITs No
t signif
icantly
asymm
etric
Unlike
other s
tudies
in vario
us asse
t mark
ets tha
t find th
e asym
metric
dep
endenc
e struc
ture
The sam
ple inc
ludes b
oth the
boom
and b
urst in
which
REITs
moved
togeth
er close
ly. As
REITs
are hig
hly cor
related
The gen
eral sh
ape of
the de
penden
ce stru
cture b
ecome
s more
imp
ortant
Cond
itiona
l Cop
ulas
Nimma
nunta&
Patel
25The sha
pe of t
he dep
endenc
e struc
ture o
f REITs
is close
r to tha
t of Stu
dent ta
nd Gu
assian
copulas
.Sca
tter Plo
ts of th
e Actu
al Prob
ability
Integra
l Value
and th
e Simu
lated Va
lue fro
m Six
Copul
as: Re
sidentia
l REIT v
s Indus
trial &
Office
REIT
Tail D
epend
ence
of RE
ITs
Nimma
nunta&
Patel
27
Linear
Corre
lation
of RE
ITs
Nimma
nunta&
Patel
28 Conc
lusion
s
Nimma
nunta&
Patel
29Using
daily d
ata of f
our U.S
. equity
REITs
(Jan 2
000 -D
ec 201
1)Lev
erage e
ffects f
ound in
every s
eries
Tail de
penden
ce wit
hin U.S
. REITs
is not s
ignifica
ntly
asymm
etric
The fea
ture fo
und in
REITs
over o
ur sam
ple pe
riod is t
he tail
depen
dence
during
both th
e uptu
rn and
downtu
rn.Ho
wever, t
he co-
movem
ents in
down
turn a
re clear
ly stro
nger
due to
leverag
e effec
ts
Q & A
Nimma
nunta &
Patel
30Conta
ct:Kr
idsda
Nimm
anun
taEm
ail: k
ridsda
.nim@
nida.a
c.th
Or Dr
Kana
k Pate
lEm
ail: k
p100
05@c
am.ac
.uk