Popular in Social Sciences

Banks and Policy Organisations

Macroeconomic modelling of CBDC: a critical review

European Central Bank Ulrich Bindseil

Social Identities and Divisions

A new explanation of antisemitism: Jew hatred as a civilisational phenomenon

Boston University Liah Greenfeld

Banks and Policy Organisations

Financing Private Credit

Federal Reserve Bank of New York Nina Boyarchenko

Economics

Failing Banks

Massachusetts Institute of Technology Emil Verner

Culture and Society

The City as a Growth Machine: Toward a Political Economy of Place

New York University Harvey Moltoch

American Politics

Keeping the Republic

Boston College Marc Landy

British Politics

Entrepreneurial thinking: the politics and practice of policy impact

University of Sheffield Matthew Flinders

Public sentiment in times of terror

View more insights in Social Sciences - Economics - Microeconomics

View more figures from University of Sydney

View the page of Ashani Amarasinghe

Read the Study

Next in Economics

Corporate foreign bond issuance and interfirm loans in China

Ugo Panizza | Geneva Graduate Institute

Offshoring and non-monotonic employment effects across industries in general equilibrium

Daniel Baumgarten | Universität Paderborn

Ashani Amarasinghe examines how terror events affect government perceptions. For 11 months after successful attacks, fine-grained analysis indicates 11% more unhappiness. Violence, political institutions, and government competency contribute.

Image courtesy of interviewee. January 9, 2024

Log-in or Sign-up to Faculti
Currently viewing this subject insight as a guest. You have insight(s) remaining for this month. Login to view 8000+ figures on the platform.

Generating Workspace...

Export as PDF

This video video discusses a study on public sentiment towards governments following terrorist attacks, focusing on how citizens react differently based on the context of the event. It contrasts the strong support for the U.S. government after the 9/11 attacks with the backlash against the Spanish government after the 2004 Madrid train bombings.

The researcher examines a global sample of 135 countries from 2002 to 2020, emphasizing short-term public sentiment fluctuations rather than long-term voting data. This video video approach allows for a more immediate understanding of how citizens hold their governments accountable post-attack.

Using a unique dataset called the GAL dataset, which compiles real-time news articles, the researcher quantifies public discontent towards governments by analyzing negative sentiment in events targeting government actions. The study finds that following a terror attack, public discontent tends to rise significantly, with an effect lasting up to five months.

The research also investigates various mechanisms that might drive these sentiments, concluding that the backlash is specifically aimed at government accountability rather than general discontent related to other factors. Ultimately, the findings demonstrate a consistent system of checks and balances where citizens monitor and hold their governments accountable, reflecting significant implications for the citizen-state relationship.

Future research plans include examining public sentiment in other contexts, such as climate change, to further explore how public opinions and government responses interact over time.

Language:

Export as PDF

X-TIMESTAMP-MAP:LOCAL=00:00:00.000,MPEGTS=900000

00:00.185 --> 00:02.245
So the context of the study that,

00:02.265 --> 00:03.725
we are discussing today, the study

00:03.725 --> 00:06.565
that's called public sentiment in times of terror, is

00:06.745 --> 00:08.605
how do pub the people react

00:08.605 --> 00:10.605
to governments following terror attacks?

00:11.095 --> 00:13.485
Right? So, famously,

00:13.485 --> 00:15.085
following the nine 11 terror attack,

00:15.505 --> 00:18.325
the US public really rallied around the government.

00:18.325 --> 00:19.645
There was lots of support

00:19.645 --> 00:22.965
for the government highest level since the 1970s.

00:23.705 --> 00:27.165
But contrast that with the situation in Spain in 2004,

00:27.165 --> 00:30.525
following the Madrid train bombings, where there was a lot

00:30.525 --> 00:32.165
of backlash against the government.

00:32.425 --> 00:35.805
And this terror attack was also cited as a recent

00:35.985 --> 00:39.085
for the government's loss at the subsequent election.

00:39.425 --> 00:41.965
So there's two distinct terror attacks.

00:42.315 --> 00:44.565
They may be different along many dimensions,

00:44.625 --> 00:46.645
but one thing that stands out is

00:46.645 --> 00:50.325
that the public's reaction deferred enormously, right?

00:50.585 --> 00:52.685
So this then led me to, you know,

00:53.045 --> 00:56.285
investigate this question further on whether there is any

00:56.285 --> 00:59.965
particular trends that we can observe where in with respect

00:59.965 --> 01:01.525
to how the public,

01:01.595 --> 01:03.885
respond following terror attacks, right?

01:03.905 --> 01:07.485
So in order to examine this question, I look at,

01:07.645 --> 01:11.165
a global sample of countries, 135 countries over the years,

01:11.185 --> 01:12.725
2002 to 2020.

01:13.425 --> 01:17.165
And the real novelty in my study actually comes from the

01:17.165 --> 01:19.445
fact that I'm looking at short-term

01:20.165 --> 01:23.405
fluctuations in public sentiments towards governments.

01:23.625 --> 01:25.125
I'm not looking at voting data

01:25.305 --> 01:27.525
or any kind of annual long-term data,

01:27.745 --> 01:31.125
but I'm really trying to exploit the truly short-term

01:31.325 --> 01:33.685
dynamics of the citizen's state relationship

01:33.835 --> 01:35.045
following terror attacks.

01:35.345 --> 01:39.085
So terrorism has been an area of interest for economics

01:39.105 --> 01:42.405
for a long, long time, and you find many,

01:42.405 --> 01:45.485
studies looking at the effects of terror attacks on say,

01:45.485 --> 01:49.125
employment wages, GDP migration, many areas.

01:49.545 --> 01:51.645
But then within this literature,

01:51.945 --> 01:56.905
my work actually contributes to the kind of niche,

01:56.905 --> 01:59.785
component which looks at the outcomes of terrorism

01:59.925 --> 02:01.785
and how terrorism affects political

02:01.985 --> 02:03.425
outcomes in particular, right?

02:03.805 --> 02:05.065
So, but as I mentioned

02:05.065 --> 02:06.585
before, most of what has been done

02:06.645 --> 02:08.705
so far relies on voting data.

02:09.005 --> 02:11.785
For example, Alvo 2011 or

02:11.925 --> 02:13.065
or Bali 2007,

02:13.095 --> 02:15.585
they look at this particular Madrid train attacks

02:15.885 --> 02:17.305
and look at how the,

02:17.305 --> 02:19.305
incumbent government was actually voted out in

02:19.305 --> 02:20.385
the subsequent election.

02:20.765 --> 02:23.505
But the problem with using voting data is that, you know,

02:23.785 --> 02:25.505
elections happen only five or six years.

02:26.005 --> 02:28.945
Um, and then there's lots going on in between,

02:29.325 --> 02:32.225
and it's very difficult for us to say it's this one event

02:32.575 --> 02:34.625
that affected the voting outcome, right?

02:34.625 --> 02:36.185
That seems a little farfetched.

02:36.405 --> 02:38.945
And there's actually quite a bit of literature that sh

02:39.235 --> 02:41.825
shows that events that happened closer

02:41.885 --> 02:43.825
to the election actually has more salience

02:43.825 --> 02:44.905
in the voters' minds.

02:45.325 --> 02:48.145
Um, and then, you know, people actually react to

02:48.145 --> 02:49.465
what happened closer to the election,

02:49.465 --> 02:51.665
and they kind of forget what happened further away.

02:52.005 --> 02:55.145
So that's why looking at these short-term dynamics becomes

02:55.145 --> 02:56.385
really, really important.

02:56.885 --> 03:00.805
Um, if you look at short-term dynamics, then you can observe

03:00.945 --> 03:02.445
how the gov people respond

03:02.445 --> 03:05.725
to the government's actions consistently throughout time

03:05.785 --> 03:08.685
and not being limited to, election cycles.

03:08.705 --> 03:10.645
And that's really the new contribution

03:10.645 --> 03:12.005
that I'm trying to make with this paper.

03:12.425 --> 03:13.485
And, I've,

03:13.665 --> 03:17.925
my broader research area agenda is also linked to these kind

03:17.925 --> 03:20.845
of, you know, short term,

03:21.325 --> 03:23.845
fluctuations in the citizen state relationship.

03:24.105 --> 03:26.245
You can imagine that the key challenge was

03:26.245 --> 03:29.045
that there exists almost,

03:29.105 --> 03:32.965
no data on consistently measuring public sentiments towards

03:32.965 --> 03:35.845
government as very fine temporal intervals, right?

03:36.225 --> 03:39.605
So I had to come up with a new, kind of quantification

03:39.785 --> 03:41.005
to answer this question.

03:41.185 --> 03:43.885
And actually, in my previous paper,

03:43.885 --> 03:47.005
diverting Domestic Turmoil, which came out in 2022 Journal

03:47.005 --> 03:48.325
of Public Economics, that's

03:48.325 --> 03:49.845
where I first introduced this idea

03:49.945 --> 03:53.885
of a quantified indicator, a quantified consistent indicator

03:53.945 --> 03:56.885
of public discontent, towards the government.

03:57.185 --> 04:00.365
And I use this massive dataset called the GAL Dataset,

04:00.375 --> 04:05.005
which is a real time, data set of news media articles

04:05.005 --> 04:06.605
that are being reported across the world.

04:06.715 --> 04:08.205
It's updated every 15 minutes.

04:08.545 --> 04:12.045
And then what G DAL does is it takes all the key information

04:12.045 --> 04:15.285
from these news articles, say, who are the actors involved?

04:15.305 --> 04:17.525
Who was it targeted, against,

04:17.525 --> 04:18.885
which country, which location?

04:19.305 --> 04:22.645
And, you know, what, what kind of an event was it?

04:22.645 --> 04:25.125
Was it a positive event or an aggressive event?

04:25.465 --> 04:27.885
And it also assigns a sentiment score to each

04:27.885 --> 04:30.045
of these events called the, Goldstein score.

04:30.505 --> 04:34.565
So in my, dataset to quantify public discontent,

04:34.845 --> 04:37.765
I filtered through this massive g data set,

04:37.865 --> 04:39.245
and I retained the subset

04:39.345 --> 04:42.045
of events which were particularly targeted towards the

04:42.045 --> 04:43.045
government of a country.

04:43.585 --> 04:46.085
And then I take the share, I take the number

04:46.085 --> 04:48.645
of events which were targeted against the government,

04:48.705 --> 04:51.725
and which had a very high negative sentiment score,

04:51.985 --> 04:54.885
and I expressed that as a, as a share

04:55.105 --> 04:56.525
of the total number of events

04:56.525 --> 04:57.805
that were targeted towards the

04:57.805 --> 04:58.965
government in the same period.

04:59.345 --> 05:02.285
And that's how I arrive at this quantified indicator

05:02.385 --> 05:03.605
of public dis content.

05:04.185 --> 05:06.565
And I have this dataset as, let's say, if I want

05:06.565 --> 05:08.925
to quantify public discontent at the day level

05:09.025 --> 05:11.805
or the week level or the month level, I can still do that

05:12.025 --> 05:13.485
for the purpose of this study.

05:13.805 --> 05:16.405
I use monthly, deed data,

05:16.825 --> 05:18.245
for public this content.

05:18.385 --> 05:20.925
So that's the key outcome variable of my study.

05:21.385 --> 05:23.685
And then obviously, because I'm looking at terrorist

05:23.685 --> 05:27.085
attacks, I did data on, terror events.

05:27.345 --> 05:30.245
And, for that I use data from the global terrorism

05:30.355 --> 05:32.205
dataset, which is like the gold standard

05:32.305 --> 05:33.605
for terrorism datasets.

05:33.605 --> 05:36.485
And then I combine that with lots of other data sets,

05:36.485 --> 05:39.285
which help me examine mechanisms

05:39.585 --> 05:43.005
and, you know, eliminate certain, things

05:43.005 --> 05:47.085
that one could suggest are going on, like

05:47.185 --> 05:49.365
for these results to, appear.

05:49.665 --> 05:51.925
And then in terms of the methodology I use,

05:52.045 --> 05:53.805
I use a combination of event study

05:54.025 --> 05:57.685
and difference indifferent, methodologies where I'm,

05:57.965 --> 06:01.045
I take a two-step approach to my identification.

06:01.185 --> 06:05.395
So firstly, I look at, I just compare country months

06:05.485 --> 06:08.275
where terror attacks occurred with country months

06:08.275 --> 06:10.515
where no terror attacks occurred, right?

06:10.695 --> 06:13.195
So, okay, in one country month there was a terror attack in

06:13.195 --> 06:14.475
another, there was no terror attack.

06:14.475 --> 06:15.835
And then I look at, okay,

06:16.105 --> 06:17.955
what happens when there's a terror attack?

06:17.955 --> 06:20.195
What's the difference? And then I can understand whether

06:20.195 --> 06:21.675
people rally around the government

06:21.695 --> 06:23.915
or whether people start criticizing the government more.

06:24.105 --> 06:25.875
However, this kind of comparison

06:26.415 --> 06:28.675
is not exactly an appropriate comparison

06:28.775 --> 06:29.875
for causal purposes,

06:29.875 --> 06:33.395
because there could be so many reasons why these two units

06:33.415 --> 06:36.195
of observation are different from each other, right?

06:36.375 --> 06:39.115
So we are not exactly comparing apples to apples, right?

06:39.415 --> 06:41.555
So therefore, I go the next step

06:41.975 --> 06:44.515
and use the identification strategy, which was,

06:44.565 --> 06:46.715
which has been little bit used in the literature,

06:46.715 --> 06:48.875
for example, most recently in broader 2018,

06:49.285 --> 06:53.595
where the comparison is between country months,

06:53.595 --> 06:56.675
where there was a successful terror attack against country

06:56.675 --> 06:58.835
months where there was a failed terror attacks.

06:59.095 --> 07:01.635
So we are essentially restricting our analysis

07:01.655 --> 07:03.995
to country months where terror attacks occurred,

07:04.455 --> 07:06.515
but for some random reason,

07:06.895 --> 07:10.035
the attack was successful in one set of country months,

07:10.295 --> 07:13.155
and the attack failed in the other set of country months.

07:13.415 --> 07:16.875
So the key assumption here is that,

07:16.875 --> 07:18.155
conditional on the timing

07:18.295 --> 07:19.635
and location of the event,

07:19.745 --> 07:21.995
whether a terrorist attack is successful

07:22.055 --> 07:23.595
or not is entirely random.

07:24.015 --> 07:26.835
And this random nature of my identification allows me

07:26.835 --> 07:30.635
to make causal, interpretation of my estimates.

07:30.945 --> 07:34.155
Primarily, what I find is that following a terror attack,

07:34.365 --> 07:36.515
there is a ri huge rise in public

07:36.565 --> 07:38.075
discontent towards the government.

07:38.095 --> 07:39.595
So people are actually very unhappy

07:39.595 --> 07:41.395
with the government following terror attacks.

07:41.655 --> 07:44.075
So remember at the beginning I mentioned that it could be

07:44.075 --> 07:45.555
that people really support the government,

07:45.555 --> 07:47.635
or it could be that people really criticize the government.

07:47.635 --> 07:51.315
So on average, the effect that materializes is actually

07:51.665 --> 07:54.475
that, people are actually holding the government

07:54.475 --> 07:56.435
accountable and criticizing the government

07:56.495 --> 07:57.755
for not keeping them safe.

07:57.975 --> 07:59.835
That's the, the first kind of result,

07:59.835 --> 08:01.235
the baseline estimates that I find.

08:01.575 --> 08:03.395
So this basically suggests that, you know,

08:03.395 --> 08:06.075
there's a very strong direction of the effect,

08:06.295 --> 08:09.875
and that this effect is actually very important even in the

08:10.025 --> 08:11.395
very short run, right?

08:11.415 --> 08:14.395
For example, in the event study estimates, I find that,

08:14.395 --> 08:16.355
public discontent rise for up

08:16.355 --> 08:18.755
to five months following a terror attack.

08:18.815 --> 08:20.315
Now, if you were using election data,

08:20.495 --> 08:21.875
we wouldn't observe this, right?

08:21.875 --> 08:24.155
So that's where the real novelty comes in.

08:24.775 --> 08:29.675
And, following this, I go on to explore what kind

08:29.675 --> 08:32.275
of mechanisms could be driving these estimates, right?

08:32.495 --> 08:34.715
So is it government accountability itself

08:34.935 --> 08:36.875
or could it be that, for example,

08:36.875 --> 08:38.835
people feel sympathy with terrorists?

08:39.375 --> 08:42.075
Or could it be that there is a general increase in this

08:42.075 --> 08:44.475
content not related to the terror attack, right?

08:44.575 --> 08:48.195
So it could be like discontented disco content targeted at

08:48.195 --> 08:50.315
opposition or some other parties.

08:50.695 --> 08:53.715
But then I'm able to rule out these kind of hypothesis

08:53.715 --> 08:56.765
because I can quantify public sentiments towards terror

08:56.765 --> 08:59.925
groups, public sentiments towards businesses, elites,

08:59.925 --> 09:01.405
opposition, and I'm unable

09:01.405 --> 09:04.205
to find any effects on these outcome variables

09:04.205 --> 09:05.285
following a terror attack.

09:05.305 --> 09:08.045
So suggests this, you know, overall suggests

09:08.045 --> 09:10.205
that this is actually, discontent,

09:11.005 --> 09:13.565
specifically targeted towards the government, right?

09:14.105 --> 09:18.205
And then I also compile a number of data sets on,

09:18.205 --> 09:20.085
let's say country level characteristics,

09:20.305 --> 09:21.725
attack level characteristics,

09:21.905 --> 09:23.685
and leader level characteristics.

09:23.985 --> 09:25.365
So I find that,

09:25.365 --> 09:28.725
public discontent rises specifically in countries which have

09:28.785 --> 09:32.765
non-democratic political institutions, in countries

09:32.765 --> 09:36.285
with low levels of military capacity, for example.

09:36.465 --> 09:37.565
So all of this suggests

09:37.565 --> 09:40.245
that countries a country level characteristics matter.

09:40.705 --> 09:43.645
And then I also find that people are more likely

09:44.465 --> 09:48.485
to be forgiving towards the government if there is a female

09:48.505 --> 09:50.765
leader in power, or if there's a new

09:51.145 --> 09:52.525
or young leader in power,

09:52.865 --> 09:55.445
or if the leader has some military experience.

09:55.865 --> 09:58.725
So people actually take into consideration all these

09:58.725 --> 10:01.005
available information when,

10:01.405 --> 10:04.045
deciding on their reaction towards a government following a

10:04.045 --> 10:05.645
terror attack and taking together,

10:06.315 --> 10:07.925
eliminating certain channels

10:07.925 --> 10:11.845
and providing support in favor of certain channels.

10:12.105 --> 10:14.405
Um, together the overall picture is

10:14.405 --> 10:17.645
that people actually do hold the government accountable,

10:17.665 --> 10:21.605
and that seems to be the key, underlying mechanism,

10:22.185 --> 10:23.485
in my estimates.

10:23.635 --> 10:25.765
Basically, the conclusion here is that yes,

10:25.765 --> 10:28.685
people do hold their government accountable,

10:28.785 --> 10:31.245
and that this is actually an important part

10:31.345 --> 10:35.365
of the short term relationships between, people

10:35.425 --> 10:36.645
and their government, right?

10:36.705 --> 10:38.485
So this, we observed that people

10:39.125 --> 10:41.165
actually keep monitoring the behavior

10:41.185 --> 10:43.165
of the government even in the short term

10:43.345 --> 10:46.085
and not necessarily when it's closer to any election.

10:46.545 --> 10:48.805
So this is especially important in the context

10:49.065 --> 10:51.365
of the citizen state relationship

10:51.425 --> 10:53.045
and how governments use their power.

10:53.395 --> 10:55.325
It's important that the government know

10:55.635 --> 10:58.845
that they are consistently being monitored by the people.

10:58.845 --> 11:00.325
There's a consistent system of checks

11:00.345 --> 11:02.365
and balances that operates by the public.

11:03.105 --> 11:07.125
Um, and, and, that probably can be a deterrent

11:07.185 --> 11:10.605
for the government if, if they're to make suboptimal,

11:10.605 --> 11:14.045
levels of, you know, suboptimal decisions, so to say.

11:14.345 --> 11:18.725
So basically the whole idea is that I want to convey is

11:18.725 --> 11:20.765
that this study provides, some

11:20.765 --> 11:24.285
of the first evidence about the short term,

11:24.925 --> 11:27.725
feedback loop between people and the government, right?

11:27.905 --> 11:31.405
And in my broader research agenda, I'm,

11:31.545 --> 11:34.285
now using this public disco content index,

11:34.305 --> 11:36.405
to look at extended areas.

11:36.745 --> 11:38.405
For example, how does the public

11:38.545 --> 11:41.605
and government interact when it comes to the climate change,

11:42.145 --> 11:43.205
dialogue, right?

11:43.625 --> 11:47.645
And how do, the government respond when they know

11:47.645 --> 11:49.125
that people are unhappy with them.

11:49.505 --> 11:52.845
Um, so basically that's, that's the broad research

11:52.845 --> 11:56.445
Agenda that I'm pursuing with the, temporarily

11:57.005 --> 11:59.525
granular data on public

12:00.065 --> 12:03.925
and how public, provide their feedback to governments

12:03.925 --> 12:07.765
and how governments, then respond to how, public,

12:08.065 --> 12:09.685
to public sentiments, right?

12:09.985 --> 12:12.165
So this feedback loop is really important

12:12.225 --> 12:15.805
to be observed in the fine temporal, dimensions.

12:15.945 --> 12:19.045
And for example, I'm working now on extending this

12:19.045 --> 12:22.165
to other areas such as the climate change discussion,

12:22.165 --> 12:24.445
where it's very important for governments

12:24.505 --> 12:27.365
and, public to interact,

12:27.385 --> 12:29.365
and many other aspects of this,

12:29.365 --> 12:31.405
citizen state relationship, which can be

12:31.925 --> 12:34.085
observed at these really fine temporal dimensions.

Faculti AI can make mistakes

Faculti

Popular in Social Sciences

Macroeconomic modelling of CBDC: a critical review

Financing Private Credit

Failing Banks

The City as a Growth Machine: Toward a Political Economy of Place

Keeping the Republic

Entrepreneurial thinking: the politics and practice of policy impact

Public sentiment in times of terror

Next in Economics

Guide

Platform and Category Pages

Contact Us

FAQs

What is Faculti?

How do you select interviews?

Interview Process

Page Structure

Popular in Social Sciences

Guide

Platform and Category Pages

Contact Us

FAQs

What is Faculti?

How do you select interviews?

Interview Process

Accessibility Options

Contrast Adjustment

Text Size Adjustment

Text Spacing Adjustment

Line Height Adjustment

Cursor Size Adjustment

Page Structure Options

Page Structure