2. Inbound
Tourism
Demand
Modelling;
Gothenburg
case
• I.
Introduc+on
&
Literature
Reviewed
• II.
Theore+cal
Framework
Determinants
of
Tourism
demand
func1on
and
other
aspects
of
Tourism
demand
modelling
• III.
Methodology
and
Diagnos+c
test
procedure
• IV.
Results
and
Analysis
Data
and
summary
sta1s1cs
Results
from
the
Interna1onal
Markets
Results
from
the
Na1onal
Regional
Markets
• V.
Conclusions
3. Introduc+on
• The
research
project
focus
on
the
descripGon
of
the
determinants
and
the
tourism
demand
funcGon
in
Gothenburg
region
as
desGnaGon,
– Measured
by
guest
night
producGon
and
– The
general
to
specific
modelling
approach
from
the
top
5
markets:
Sweden,
Norway,
UK,
USA
and
Germany.
•
I
have
worked
with
a
Gme
series
annually
data
from
1982
to
2013,
– Ordinary
least
squares
(OLS)
– Autoregressive
distributed
lag
model
(ADLM)
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
4. 0
50
100
150
200
250
300
Thousands
guest
night
produc+on
Norway
Germany
UK
USA
Figure
1
Guest
Nights
produc+on
in
Gothenburg
region,
from
1982
to
2013
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
5. Literature
Reviewed
• The
gap
in
the
literature
reviewed;
– Lack
of
diagnosGc
checking
in
the
empirical
studies.
– The
supply
side
of
the
market
is
not
the
favorite
research
topic
for
modelling
and
forecasGng
tourism.
– New
soXware
improvements
make
possible
to
work
with
the
computable
modelling
approach.
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
6. Theore+cal
Framework
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
Determinants
of
tourism
demand
func1on
• The
inbound
demand
funcGon
for
the
tourism
product
in
desGnaGon
• i
=
Gothenburg,
by
residents
of
origin
j
=
USA,
UK,
Germany,
Norway
and
Sweden,
is
given
by
Qij
=
f
(Yj,
Pij,
Psj)
Yj
is
the
level
of
income
in
origin
country
j
Pij
is
the
price
of
tourism
in
desGnaGon
i,
for
residents
in
origin
country
j
Psj
is
the
price
of
tourism
in
the
subsGtute
desGnaGon
s,
for
resident
in
origin
country
j
7. Tourism
Demand
Modelling
• The
most
basic
relaGon
between
the
variables
is
the
linear
relaGonship,
Song
and
Wi^
(2000)
and
Song,
Wong
and
Chong
(2003).
Qjt
=
α0
+
α1Yit
+
α2Pjt
+
α3Psjt
+
ęit
• RelaGve
consumer
price
of
tourism
in
Sweden
for
InternaGonal
market
Pijt
=
(CPISW/EXSW)/(CPIj/EXj),
j
=
1,
2,
3,
4
• RelaGve
consumer
price
of
tourism
in
Sweden
for
DomesGc
Market
Pit
=
(CPISW/EXSW),
t
=
1982,
…,
2013
• RelaGve
consumer
price
of
tourism
in
Norway
for
InternaGonal
Market
Psjt
=
(CPInw/EXnw)/(CPIj/EXj),
j
=
1,
2,
3,
4
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
8. Methodology
The
“tradiGonal”
tourism
demand
modelling
proceeds
with
the
following
steps:
a)
Formulate
hypotheses
based
on
classic
microeconomic
theory
b)
Through
the
decided
model´s
funcGonal
form,
present
data
and
esGmate
the
coefficients
c)
Once
esGmate
the
model´s
coefficients,
generate
forecast
for
the
desGnaGon
using
the
ElasGcity
concept:
η
=
(α)*(Prj/Qjt)
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
9. • Hypothesis
I:
The
Engel
curve
suggest
that
if
the
price
of
tourism
is
held
constant,
an
increase
in
tourists´
income
will
result
in
an
increase
in
the
demand
for
tourism
to
the
desGnaGon
provided:
tourism
is
a
normal
or
necessary
good
• Hypothesis
II:
If
the
price
of
tourism
in
desGnaGon
1
increases
while
price
of
tourism
in
desGnaGon
2
and
consumers´
income
in
the
origin
country
remain
unchanged,
tourist
will
“switch”
from
going
to
desGnaGon
1
to
desGnaGon
2,
and
therefore
the
demand
for
tourism
to
desGnaGon
1
will
decrease
• Hypothesis
III:
With
respect
to
the
demand
for
tourism
to
desGnaGon
1,
the
effect
of
a
price
change
in
desGnaGon
2
can
have
either,
a
posiGve
or
negaGve
effect
– If
desGnaGon
2
is
subsGtute
for
desGnaGon
1
the
demand
for
tourism
to
desGnaGon
1
will
move
in
the
same
direcGon
as
to
the
price
change
in
desGnaGon
2.
– if
tourist
tend
to
travel
both
desGnaGons
together,
i.e.,
the
desGnaGon
are
complementary
to
each
other,
tourism
demand
to
one
desGnaGon
will
move
in
the
opposite
direcGon
to
the
change
in
price
of
tourism
in
the
other
Hypotheses
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
10. • The
inbound
tourism
demand
equaGon
The
staGc
model
by
origin
country,
OLS
regression
qjt
=
βo
+
β1
Yjt
+
β2
Pjt
+
β3
Psjt
+
ε
it
j=1,
2,
3,
4,
5
The
ADLM
model
by
origin
country,
Newey-‐West
regression
qjt
=
α0
+
α1Yjt
+
α2Yjt-‐1
+
α3Pjt
+
α4Pjt-‐1
+
α5
Pst
+
α6
Pst-‐1
+
υjt
j=1,
2,
3,
4,
5
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
Diagnos+c
test
procedure
11. • OLS;
Best
Linear
Un-‐bias
EsGmator
BLUE
– Gauss-‐Markov
assumpGons
– t-‐test
and
p-‐value,
significance
level
• Lagrange
MulGplier;
autocorrelaGon
– Breusch
(1978)
and
Godfrey
(1978)
– Variance
–
Covariance
matrix
• Newey-‐West
– AutocorrelaGon
– HeteroscedasGcity
– Spurious
regression
• Unit
root
test
-‐
CointegraGon
Dilemma
on
instruments,
How
to
measure?
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
12. Ordinary
least
squares,
OLS
• BLUE-‐Gauss
Markov
– A1:
E(ui|Xi)
=0
– A2:
(Xi,Yi)
are
independent
thus
(Xi,ui)
are
independent
(endogeneity)
– A3:
Var
(ui)
=
σ2
(homoskedasGcity)
– A4:
Cov
(Ui,Uj)
=
0,
(no-‐autocorrelaGon)
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
14. Measures
of
fit;
R-‐square
• R2
=
ESS/
TSS
– Explained
sum
of
squares/Total
sum
of
squares
• R2
=
1
–
(SSR/
TSS)
– Sum
of
squares
residuals/Total
sum
of
squares
20. • An
important
observaGon
is
that
the
ADLM
performs
stable
the
regional-‐domesGc
forecast.
• I
can
describe
that
1%
change
increase
in
the
relaGve
consumer
price
of
tourism
in
Sweden,
will
decrease
the
domesGc
demand
0.7%
or
20,931
guest
night
producGon
in
2014.
• Furthermore,
1%
change
increase
in
the
relaGve
consumer
price
of
tourism
in
Norway,
will
decrease
the
domesGc
demand
in
0.08%
the
guest
night
producGon.
• Regarding
to
the
regional
market
with
Norway,
I
can
describe
that
1%
change
increase
in
the
relaGve
consumer
price
of
tourism
in
Sweden
for
Norwegians,
decrease
the
travelers´
demand
to
Gothenburg
0.35%
or
921
guest
night
producGon
in
2014.
Results
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
21. Step
forward
• An
important
conclusion
is
that
the
ADLM
is
not
stable
in
the
internaGonal
forecast.
The
posiGve
aspect
inside
this
negaGve
outcome
is
related
to
the
distance
factor.
– Likely
to
deal
with
omi^ed
variable
bias
related
to
travel
costs
and
distance
from
the
desGnaGon
point
• PotenGal
soluGons,
Song,
Wong
and
Chong
2003:
– Aggregate
the
seasonality
effects
through
the
monthly
or
quarterly
database,
use
TVP
as
instrument
of
esGmator
– Re-‐design
the
relaGve
consumpGon
price
of
tourism
in
the
desGnaGons
– Use
an
alternaGve
raGo
relaGon
in
the
subsGtute
desGnaGon
as
follows;
Psjt
=
(CPInw/EXnw)
wj
,
where,
wj
=
(TGNsj
/
Σsj=1,2
TGNsj)
• W
is
the
proporGon
of
internaGonal
guest
nights
producGon
from
origin
country
j
into
the
subsGtuGon
desGnaGon,
related
with
the
summary
of
guest
nights
producGon
of
the
two
desGnaGons
from
the
same
origin
country
j.
Inbound
Tourism
Demand
Modelling;
Gothenburg
case
22. Inbound
Tourism
Demand
Modelling;
Gothenburg
case
Antal
gästnäder
per
månad,
2008
–
2014,
SCB.
Millions
gästnä^er
0
0.5
1
1.5
2
2.5
2008M01
2008M04
2008M07
2008M10
2009M01
2009M04
2009M07
2009M10
2010M01
2010M04
2010M07
2010M10
2011M01
2011M04
2011M07
2011M10
2012M01
2012M04
2012M07
2012M10
2013M01
2013M04
2013M07
2013M10
2014M01
2014M04
2014M07
2014M10
2015M01
2015M04
Millions
Stor-‐Göteborg
Västra
Götalands
län
VGR-‐SG
