Understanding News Geography and Major Determinants of Global News Coverage of Disasters
Understanding
News
Geography
and
Major
Determinants
of
Global
News
Coverage
of
Disasters
Jisun
An
and
Haewoon
Kwak
Qatar
Computing
Research
Institute
(QCRI)
C+J
Symposium
2014
Theory
of
newsworthiness
• Proposed by Galtung and Ruge, 1965
• Frequency, intensity, unambiguity, meaningfulness,
consonance, unexpectedness, and continuity of an
event
• Some characteristics of an actor (e.g. identity)
involved in the event
Effect
of
national
attributes
A guest country
(where a disaster happens)
A host country
(where news media exists)
Effect
of
national
attributes
A guest country
(where a disaster happens)
Population, degree
of press freedom,
geographic size,
economic power, etc
A host country
(where news media exists)
National traits
Effect
of
national
attributes
A guest country
(where a disaster happens)
Population, degree
of press freedom,
geographic size,
economic power, etc
A host country
(where news media exists)
National traits
Relatedness
Geographic distance,
shared language,
volume of trade,
past colonial ties
A
wide
range
of
national
attributes
and
power
relationship
has
been
examined,
but
results
are
inconsistent,
mainly
due
to
cultural
difference.
A
wide
range
of
national
attributes
and
power
relationship
has
been
examined,
but
results
are
inconsistent,
mainly
due
to
cultural
difference.
international
coverage
news
Examined
a
in
country
single
GDELT
dataset
3,574,627
events
(666,150
natural
and
man-‐made
disasters)
happening
in
205
countries
from
April
2013
to
July
2014
News
geography
Which
countries
are
presented
in
one
country’
international
disaster
news
and
to
what
extent?
Global
news
coverage
Which
disasters
are
most
likely
to
be
‘globally’
covered?
GDELT
dataset
3,574,627
events
(666,150
natural
and
man-‐made
disasters)
happening
in
205
countries
from
April
2013
to
July
2014
News
geography
Which
countries
are
presented
in
one
country’
international
disaster
news
and
to
what
extent?
Global
news
coverage
Which
disasters
are
most
likely
to
be
‘globally’
covered
GDELT
dataset
• monitors
the
world's
news
media
in
print,
broadcast,
and
web
formats,
in
over
100
languages
• compiles
a
list
of
every
person,
organization,
location,
themes
(events),
etc.
from
every
news
report
• stretches
back
to
January
1,
1979
and
update
daily
• leverages
Google
Translate
for
Research
to
augment
human
translation
Example
“47
Jihadists
were
killed”
was
found
in
ten
different
articles
that
day
• NumArts=10,
• CountType=KILL,
• Count=47,
• ObjectType=”jihadists”,
Fields
• DATE
• NUMARTS
• COUNTTYPE
-‐ AFFECT,
ARREST,
KIDNAP,
KILL,
PROTEST,
SEIZE,
WOUND..
• COUNTS
• THEMES
-‐ NATURAL
DISASTER,
MANMADE
DISASTER,
…
• LOCATIONS
• OBJECTIVE
-‐ “Christian
missionaries”
from
“20
Christian
missionaries
were
arrested”
• SOURCEURLS
GDELT
3,574,627
events
(666,150
natural
and
man-‐made
disasters)
happening
in
205
countries
from
April
2013
to
July
2014
News
geography
Which
countries
are
presented
in
one
country’
international
disaster
news
and
to
what
extent?
Global
news
coverage
Which
disasters
are
most
likely
to
be
‘globally’
covered
Which
countries
are
presented
in
one
country’
international
disaster
news
and
to
what
extent?
Strong
regionalism
in
reporting
foreign
disasters
All
coefficients
are
positive,
but
less
than
0.5.
Disasters
covered
by
each
region
do
not
overlap
with
one
another.
Strong
regionalism
in
reporting
foreign
disasters
The
unique
interest
of
each
region
makes
it
difficult
for
research
relying
on
a
single
country
or
region
to
obtain
external
validity.
GDELT
3,574,627
events
(666,150
natural
and
man-‐made
disasters)
happening
in
205
countries
from
April
2013
to
July
2014
News
geography
Which
countries
are
presented
in
one
country’
international
disaster
news
and
to
what
extent?
Global
news
coverage
Which
disasters
are
most
likely
to
be
‘globally’
covered?
Which
one
would
be
more
likely
to
be
covered
by
more
counties?
“A
baby
killed
by
her
mum”
or
“200
people
killed
by
flooding
in
Malaysia”
Which
one
would
be
more
likely
to
be
covered
by
more
counties?
“A
baby
killed
by
her
mum”
or
“200
people
killed
by
flooding
in
Malaysia”
Why
would
one
be
covered
more
than
the
other?
“200
people
killed
by
flooding
in
Malaysia”
school
girls
Why
would
one
be
covered
more
than
the
other?
“200
people
killed
by
flooding
in
Malaysia”
2000 school
girls
Why
would
one
be
covered
more
than
the
other?
“200
people
killed
by
flooding
in
Malaysia”
2000 school
girls USA
Why
would
one
be
covered
more
than
the
other?
“200
people
killed
by
flooding
in
Malaysia”
2000 school
girls USA
Would
this
then
get
more
coverage?
Determinants
of
global
news
coverage
We
build
a
hierarchical
(mixed-‐effect)
multiple
regression
model
where
Dependent
variable:
Number
of
countries
covering
a
disaster
Determinants
of
global
news
coverage
We
build
a
hierarchical
(mixed-‐effect)
multiple
regression
model
where
Dependent
variable:
Number
of
countries
covering
a
disaster
Independent
variable:
1) the
attributes
of
a
nation
e.g.,
GDP,
population,
etc
2) the
attributes
of
a
disaster
e.g.,
disaster
type,
the
number
of
affected
people,
etc
3) logistics
of
news
gathering
e.g.,
whether
INAs
report
a
disaster
or
not
Candidate
independent
variables
National
variables:
GDP
(gross
domestic
product)
per
capita,
GNI
(gross
national
in-‐
come)
per
capita,
military
expenditure,
population,
land
size,
population
density,
merchandise
exports
(US$),
merchandise
imports
(US$),
number
of
scientific
journal
publications,
unemployment
rate,
foreign
aid
received
(US$),
Internet
use
(per
100
people),
mobile
cellular
subscriptions
(per
100
people),
homicide
rate
(per
100,000
people),
index
of
press
freedom,
the
world
giving
index,
and
the
political
stability
index.
Disaster
variables:
a
disaster
type,
a
fine-‐grained
subtype
of
a
disaster,
the
unexpectedness
of
a
disaster,
the
number
of
people
involved
in
the
disaster,
the
type
of
people’s
involvement
in
the
disaster
(denoted
as
count
type),
and
the
type
of
people
affected
by
the
disaster
(denoted
as
object
type).
Logistic
of
news
gathering
variable:
one
binary
variable
to
show
whether
a
disaster
is
reported
by
any
of
international
news
agencies
(Agence
France-‐
Presse
(AFP),
Associated
Press
(AP)
and
Reuters)
Determinants
of
global
news
coverage
We
build
a
hierarchical
(mixed-‐effect)
multiple
regression
model
where
Dependent
variable:
Number
of
countries
covering
a
disaster
14
independent
variable:
1) the
attributes
of
a
nation
e.g.,
GDP,
population,
etc
2) the
attributes
of
a
disaster
e.g.,
disaster
type,
the
number
of
affected
people,
etc
3) logistics
of
news
gathering
e.g.,
whether
INAs
report
a
disaster
or
not 1
control
variable:
A
country
a
disaster
happened
• Disasters
in
countries
with
high
population
tend
to
be
more
globally
covered.
• The
more
people
killed
or
affected
by
a
disaster,
the
more
countries
the
disaster
will
be
covered
in.
• Kidnap
is
one
of
the
disaster
types
that
gets
the
most
attention
from
other
countries.
• Some
types
of
disasters
are
more
likely
to
be
chosen
than
others.
For
example,
Aftershocks
and
Flooding
types
tend
to
be
reported
in
more
countries
than
the
other
types.
Three
national
variables,
which
are
log(population),
mobile
subscription,
political
stability,
explain
3.1%
of
the
variance.
Disaster
variables
are
added
to
the
first
model
The
characteristics
of
a
disaster
themselves
explain
an
additional
4.3%
of
the
variance.
Summary
• GDELT,
a
valuable
resource
to
study
news
flow
around
the
world
• A
strong
regionalism
in
reporting
disasters
• A
prominent
role
of
INAs
in
global
news
coverage
of
disasters