belt and road initiative and china’s image: a big data ... 2018/alicia garcia herr… · • the...

Trade fear behind the image of the Belt and Road Initiative: A Big Data Analysis

Alicia Garcia-Herrero

BRUEGEL and HKUST

1Trade and Investment under ‘One Belt One Road’ and Implications for Hong Kong

Jianwei Xu

Beijing Normal University

With contribution from Hanrui Li

1. Motivation and objective

2. Data and methodology

3. Results

4. Conclusions

Content

Bukit Merah, SingaporeTrade and Investment under ‘One Belt One Road’ and Implications for Hong Kong 2

1. Motivation and objective


Background

• In 2013, China embarked on its grand project of the Belt and RoadInitiative (BRI) to promote infrastructure connectivity in the Eurasianarea. In fact, the Belt and Road Initiatives have been officiallysupported by sixty-six countries (as of May 2018).

• With the massive financing that China is bringing to Belt and Roadcountries to build infrastructure, the recipients should, in principle,view China’s grand plan very positively. However, there is no systemicevidence of how this project fears among recipient countries and,more generally, globally.

• Given the above pros and cons of the BRI, it seems important to offera quantitative assessment of the image of BRI across the globe, bothfor countries which make part of the Initiative but also for others.

• Finally, we investigate empirically what may be the key reasons forsuch perception at the country level

Trade and Investment under ‘One Belt One Road’ and Implications for Hong Kong 4


China 2013 2014 2015 2016 2017 2018

A dynamic view of the Belt and Road

Source: www.gdelt.org & CARTO

http://www.gdelt.org/

Our country classification


BRI countryNon-BRI

country

Narrow BRI countryBroad BRI

country

East Asia &

PacificCentral Asia South Asia

Europe Latin America

and the

Caribbean

Middle East

& North

Africa

Sub-Saharan

AfricaEU Non-EU

Brunei

Cambodia

Indonesia

Korea

Laos

Malaysia

Mongolia

Myanmar

New Zealand

Niue

Papua New

Guinea

Philippines

Singapore

Thailand

Timor-Leste

Vietnam

Kazakhstan

Kyrgyzstan

Tajikistan

Uzbekistan

Afghanistan

Bangladesh

Maldives

Nepal

Pakistan

Sri Lanka

Austria

Bulgaria

Croatia

Czech

Republic

Estonia

Hungary

Latvia

Lithuania

Poland

Romania

Slovak

Republic

Slovenia

Albania

Armenia

Azerbaijan

Belarus

Bosnia and

Herzegovina

Georgia

Macedonia

Moldova

Montenegro

Russia

Serbia

Turkey

Ukraine

Antigua and

Barbuda

Bolivia

Guyana

Panama

Trinidad and

Tobago

Egypt

Iran

Israel

Jordan

Kuwait

Lebanon

Morocco

Oman

Qatar

Saudi Arabia

Syria

United Arab

Emirates

Tunisia

Ethiopia

Kenya

Madagascar

South Africa

Senegal

Rwanda

Bahrain

Bhutan

India

Iraq

Turkmenistan

Yemen

Belgium

Cyprus

Canada

Denmark

Finland

France

Germany

Greece

Japan

Malta

Ireland

Italy

Luxembourg

Netherlands

Norway

Portugal

Spain

Sweden

Switzerland

United

Kingdom

United States

Objective of the project

• Quantitatively assess the global perception of the Belt and Road Initiative.

• Identify the key factor explaining each country’s image of the Initiative and the direction in which the frequency in which these topics are discussed in the news affect the image of BRI in different countries.

• More specifically, the perception can be further captured by answering two questions:

– How well received it is?

– What is the key factor/s behind such image?


2. Data and methodology


Source of big data

Our big data source is Global Database of Events, Language, and Tone (GDELT)

• Content coverage: Print, broadcast, and web news media in over 100 languages

• Time span: Jan 1, 1979 to present; Update every 15 minutes

• Identification of tone: Built-in tonal dictionary


The GDELT

Project

GDELT +

Google

BigQuery

Event

Database

1.0 Event

Database

2.0 Event

Database

Global

Knowledge

Graph (GKG)

2.0 Global

Knowledge

Graph

Visual Global

Knowledge

Graph

Others

GDELT

Summary (API)

Global

Online News

Coverage

Television

News

Two approaches to GDELT

GDELT GDELT Summary

Chosen keyword(s)

Key words are restricted within the built-in dictionary for institutions and/or events

For the project, only “One Belt and One Road” is available

No restriction on the choice of key words for search

Content coverage Print, broadcast and online news Online news

Time coverage Jan 1, 1979 to presentThe past 355 days before the date of

search

Extraction method SQL in Google BigQueryApplication Programming Interface

(API)


Table 1 Comparison between two data pools

Our choice sinceconcept not in GDELTdictionary and alsorecent enough to usesummary

We look for several definitions relating to Belt and Road

• One Belt, One Road

– “China’s ‘one belt, one road’ initiative has been much talked about for thepast two years, …”, SCMP Editorial, South China Morning Post, April 02, 2015

• The Belt and Road Initiative

– “In concrete terms, the Belt and Road initiative is an immensely ambitiousdevelopment campaign, …” – Tom Phillips, The Guardian, May 12, 2017

• New Silk Road

– “If the new Silk Road is symbolized by images of stasis – police checkpoints,long lines of trucks and barbed wire – there won’t be much to remember.” –Jonathan Hillman, The Washington Post, February 14, 2018


Which media to focus on

• We use public media as the best channel to reflect the mainstream public opinion.

– Beyond official or elites’ opinions

– Exclude extreme opinions

• Why we exclude social media?

– Avoid noisy and volatile opinions

– Avoid fake news


Construction of the key sentiment measurement


• Tone: as proxy for the image of BRI

• Step 1: calculate the tone of a specific article c in country j

𝑇𝑗,𝑐 =𝑤𝑗𝑝𝑐 −𝑤𝑗𝑛𝑐

𝑤𝑗𝑐

• Step 2: calculate the general tone in a country

wjpc: the number of words with positive sentiment in article c of country jwjnc: the number of words with negative sentiment in article c of country jwjc: the total number of words in article c of country j

𝑇 =1

𝑁𝑐

𝑗

𝑇𝑗,𝑐

3. Results



BRI generally perceived positively

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0

2

4

6

8

10

12

-5.1

7

-4.7

0

-4.2

2

-3.7

5

-3.2

7

-2.8

0

-2.3

3

-1.8

5

-1.3

8

-0.9

1

-0.4

3

0.0

4

0.5

2

0.9

9

1.4

6

1.9

4

2.4

1

2.8

8

3.3

6

3.8

3

4.3

1

4.7

8

5.2

5

5.7

3

6.2

0

Figure 1 The normal distribution of the 130 countries'

sentiment towards the Belt and Road Initiative

Figure 1 reports the summary statistics for our measurement ofthe sentiments across countries, indicating that the Initiative ison average positively received by the world.


BRI perceived similarly among BRI and non BRI countries

In Figure 2, we further compare the tones between BRI countriesand Non-BRI countries. Interestingly, the Belt and Road seemsonly slightly less positive for countries within the Belt and Roadgeographies than the outsiders, but the difference is statisticallyinsignificant.

-4 -3 -2 -1 0 1 2 3 4 5 6

Figure 2 Tone of BRI in BRI country and non-BRI country

BRI country

Non-BRI country

Maldives

Bosnia and Herzegovina

Poland

Laos

Guyana Norway

Liberia

Netherlands

Botswana


Africa has the best image while South Asia the worst

-0.5

0

0.5

1

1.5

2

Central Asia Sub-Saharan

Africa

East Asia &

Pacific

Middle East and

North Africa

Europe(the EU) Europe(the

Non-EU)

North America Latin America

and the

Caribbean

South Asia

Figure 3 Regional comparison of media sentiment on BRI

Media sentiment World average

South Asia fears worst in terms oftheir image of the BRI while Africafears best. Central Asia and Sub-Saharan Africa showed the mostpositive perception of the Belt andRoad, reflecting China’s long-termrelationship with the region evenbefore the establishment of the Beltand Road.


At country level

In fact, nearly all Central Asian countries show very positive attitudestowards the Initiative. East Asia & Pacific show generally strong supportof the BRI. In Europe, the EU countries seem more positive about theBelt and Road than the non-EU European countries. The South Asiancountries take a negative attitude against the Chinese plan.

Figure 4 Distribution of sentiment within 130 countries and regions

4-2 620-4

Negative Positive

Central Asia

Sub-Saharan Africa

East Asia & Pacific

Middle East &

North Africa

Europe (EU)

Europe (Non-EU)

North America

Latin America

& Caribbean

South Asia

UzbekistanTajikistan

BotswanaTanzaniaZimbabweSouth Africa

Australia Thailand Vietnam Laos

Iraq Syria Yemen Jordan

Sweden Finland Portugal Netherland

Bosnia and Herzegovina Ukraine Turkey Belarus

Canada US

Guyana Venezuela Jamaica

Maldives Bhutan Pakistan Afghanistan

Poland

Vanuatu


Regional ComparisonSoutheast Asia

Top Positive Tone

Laos 3.01

Top Negative Tone

Philippines 0.22

0

1

2

3

4

Laos Indonesia Viet Nam Brunei Cambodia Myanmar Singapore Thailand Malaysia Philippines

Tone

Average Line

Note:

Tone for BRI country


Regional ComparisonEurope

Top Positive Tone

Netherlands 3.12

Top Negative Tone

Bosnia and Herzegovina -2.43

-2.8

-1.8

-0.8

0.2

1.2

2.2

3.2

Ne

the

rlan

ds

Po

rtu

gal

Uzb

eki

stan

Slo

ven

ia

Kyr

gyzs

tan

Be

laru

s

Cyp

rus

Kaz

akh

stan

Ital

y

Ge

org

ia

Cro

atia

Alb

ania

Luxe

mb

ou

rg

Fran

ce

Gre

ece

Serb

ia

Tajik

ista

n

Lith

uan

ia

Aze

rbai

jan

Fin

lan

d

Mac

edo

nia

Arm

en

ia

Slo

vaki

a

Spai

n

Un

ite

d K

ingd

om

Turk

ey

Au

stri

a

Latv

ia

Ge

rman

y

Esto

nia

Bu

lgar

ia

Swed

en

Hu

nga

ry

Swit

zerl

and

Mo

nte

neg

ro

Ro

man

ia

De

nm

ark

Ru

ssia

Mo

ldo

va

Irel

and

Cze

ch R

ep

ub

lic

Be

lgiu

m

Ukr

ain

e

No

rway

Po

lan

d

B&

H

Tone

EU&BRI Non-EU&BRI EU&Non-BRI Non-EU&Non-BRI Average Line (Europe & Central Asia)


Regional ComparisonMiddle East

Top Positive Tone

Jordan 2.47

Top Negative Tone

Iraq -1.19

-2

-1

0

1

2

3

Jord

an

Yem

en

UA

E

Egyp

t

Bah

rain

Leb

ano

n

Syri

a

Isra

el

Sau

di A

rab

ia

Iran

Qat

ar

Om

an

Ku

wai

t

Iraq

Tone

Average Line

Note:



Regional ComparisonAfrica

Top Positive Tone

Botswana 4.97

Top Negative Tone

South Africa -0.16

-1

0

1

2

3

4

5

6

Bo

tsw

ana

Lib

eria

Tan

zan

ia

Ch

ad

Zam

bia

Eth

iop

ia

Rw

and

a

Mal

awi

Nig

eria

Zim

bab

we

Gh

ana

Som

alia

Cam

ero

on

Mau

riti

us

Mad

agas

car

Ken

ya

Sen

ega

l

Alg

eria

Uga

nd

a

Sou

th A

fric

a

Tone

Average Line

Note:


Tone for non-BRI country


Regional ComparisonCentral and South America

Top Positive Tone

Brazil 1.26

Top Negative Tone

Guyana -2.26

-2.50

-2.00

-1.50

-1.00

-0.50

0.00

0.50

1.00

1.50Tone

Average Line

Note:


Tone for non-BRI country


Image of the BRI at country level

Difference in the BRI image are very large and sometimes unexpected.Both extremes are broadly based in Europe and Asia, which meansChina’s Belt and Road Initiative has particularly penetrated the tworegions but received much-divided opinions within them. Theimplementation of the Belt and Road Initiative seems to be heavilyinfluenced by the local characteristics.

0

1

2

3

4

5

6

Bo

tsw

an

a

Ne

the

rla

nd

s

Lib

eri

a

Ta

nza

nia

La

os

Mo

rocc

o

Jord

an

Po

rtu

ga

l

Ch

ad

Za

mb

ia

Figure 5 Media sentiment for the

most positive countries

Tone(BRI countries) Tone(Non-BRI countries)

-3

-2.5

-2

-1.5

-1

-0.5

0

Ma

ldiv

es

Bo

snia

an

d

He

rze

go

vin

a

Gu

ya

na

Po

lan

d

No

rwa

y

Ira

q

Ku

wa

it

Bh

uta

n

Ind

ia

Ec

ua

do

r

Figure 6 Media sentiment for the

most negative countries

Tone(BRI countries) Tone(Non-BRI countries)

Source: https://www.gdeltproject.org/

https://www.gdeltproject.org/


Empirical analysis: Descriptive statistic to chose key factors

To investigate the relationship between the image of the BRI andthe factors that may affect such image, we select the countrieswith the highest and lowest sentiment towards the BRI in eachregion. Within the selected sample, we found most of the BRI-related articles contain two keywords: trade and investment.

Source: https://www.gdeltproject.org/

Keywords Tone Proportion (%)‘trade’ only in BRI-related news 0.58 22.9

‘investment’ only in BRI-related news 0.57 11.3Both ‘trade’ and ‘investment’ in BRI-related news 0.88 40.2

Others and BRI 0.27 25.6BRI (Total) 0.58 100

Table 1 Descriptive statistic of the tone and proportion of trade, investment, and other topics

https://www.gdeltproject.org/


Empirical analysis: Modeling

To assess how the readings of ‘trade’ and ‘investment’ affect thecountries’ perception of the BRI, we use the following econometricmodel to analyze their influences.

0 1 2 3i i i i iTone BRI Trade Investment = + + + +

• Tonei is the sentiment of the BRI in country i, in other words, theimage of the BRI in country i.

• BRIi is a dummy variable with its value is set to 1 if country i is inthe BRI geographies and 0 if not. To maintain consistency with thecoverage of the sentiment variable, we define the BRI countriesusing its official status (classified by China’s official Belt and Roadwebsite) until April 25th, 2018.

• Tradei represents the proportion of the BRI-related newsmentioning the word ‘trade’ to all the BRI-related news, depictingthe influence of trade in the local media reporting the BRI.

• Investmenti represents the proportion of the BRI-related newsmentioning ‘investment’ to the total BRI-related news.


Empirical analysis: Results

• First, here is no statistical difference between the BRI and non-BRIcountries as regards their perception of the Belt and Road.

• Second, the key factor explaining the differences in the perceptionof BRI in our sample is trade and, more especially, how frequentlytrade is mentioned in the media relating to BRI. The morefrequently trade is mentioned in the media (trade intensity), theworse perception a country tends to have about BRI.

• Third, while the coefficient before the investment is also negative, itis not statistically significant for all the specifications.

Independent variable BRI only Trade only Investment only All controls

BRI-0.117(0.20)

-0.249(0.20)

Trade (%)-0.019(0.01)

-0.030*(0.01)

Investment (%)-0.004(0.02)

-0.016(0.01)

R2 0.0025 0.0305 0.0009 0.08Obs 130 118 113 107

p<0.001: ***; p<0.01: **; p<0.05: *; p< 0.1: ^

Table 2 Robust OLS regression result

4. Conclusions


Based our bigdata analysis of news across the globe, we concludethat:

• The key finding is that most regions in the world are holding a positiveview on China’s Belt and Road Initiative, although wide differencesappear across regions and countries and some do have a negativeimage of BRI.

• Second, there is no significant difference in the perception of the BRIbetween the countries who are officially part of the initiative and others.

• Third, ‘Trade’ is the key factor explaining the BRI image across countries.Investment seems to be less relevant as it is not found statisticallysignificant.

• Fourth, zooming in into one country, Sri Lanka, it seems that bothawareness as well as the tone can vary quite substantially over timefollowing events/relations with China.


Key take-aways

• Time extension, using big query but problem with the definitionof Belt and Road

• More focus on particular countries

• Extend the realm from BRI to China economic relations with eachcountry or region

• Check to what extent fears relate to trade are grounded onactual data on trade, such as a growing trade deficit with Chinaor the like.


Possible extensions

Comments welcome!

Prof Alicia Garcia-Herrero


Jianwei Xu

With contribution from Hanrui Li

belt and road initiative and china’s image: a big data ... 2018/alicia garcia herr… · • the...

Documents