The B.E. Journal of Macroeconomics.

2023; 23(2): 917–958

Contributions

Metodij Hadzi-Vaskov, Samuel Pienknagura* and

Luca Antonio Ricci
The Macroeconomic Impact of Social Unrest
https://doi.org/10.1515/bejm-2022-0039
Received February 18, 2022; accepted January 1, 2023

Abstract: This paper explores the macroeconomic impact of social unrest, using
a novel index based on news reports. It shows that unrest has an adverse effect
on economic activity, with GDP remaining on average 0.2 percent below the pre-
unrest baseline six quarters after a one-standard deviation increase in the unrest
index. Moreover, results are robust to instrumenting via regional unrest to address
potential endogeneity concerns. Unrest “events”, captured by a large change in
the unrest index, result in a more pronounced decline in GDP—a 1 percent reduc-
tion six quarters after the event—but impacts differ by type of event. The adverse
impact of unrest on activity is mainly associated with sharp contractions in man-
ufacturing and services, and consumption. However, it can be mitigated by strong
institutions and by a country’s policy space (such as fiscal space and exchange rate
flexibility).

Keywords: social unrest, economic activity, growth, institutions, policy space

JEL Classification: D74, F50, O11, O47

We thank the editor, Tiago Cavalcanti, and two anonymous referees for excellent suggestions. We are
very grateful to Philip Barrett for kindly sharing the social unrest data and for valuable comments. The
paper also benefitted from helpful comments by Oded Galor, Elena Ianchovichina, Deniz Igan, Aart
Kraay, Inci Otker, Martin Rama, Daniel Riera-Crichton, Bob Rijkers, Guillermo Vuletin, and participants
at the IMF WHD Seminar, the World Bank ABCDE Conference, and the World Bank LAC Chief Economist
Office Seminar. The views expressed in this paper are those of the authors and do not necessarily
represent the views of the IMF, its Executive Board, or IMF management. All errros are our own.

*Corresponding author: Samuel Pienknagura, International Monetary Fund, Washington, USA,

E-mail: Spienknagura@imf.org
Metodij Hadzi-Vaskov and Luca Antonio Ricci, International Monetary Fund, Washington, USA,
E-mail: Mhadzivaskov@imf.org (M. Hadzi-Vaskov), Lricci@imf.org (L.A. Ricci)
918 — M. Hadzi-Vaskov et al.

1 Introduction
The frequency of social unrest has increased recently (Figure 1), with impor-
tant social political, and economic ramifications across the globe. In 2019, social
unrest events affected a large and diverse set of countries, ranging from advanced
economies (France and the People’s Republic of China Hong Kong Special Admin-
istrative Region) to emerging market economies (Chile and Lebanon). The Global
Peace Index (2020) suggests that the number of riots, general strikes and anti-
government demonstrations around the world increased by 244 per cent over the
last decade (2011–2019). Importantly, these unrest events have affected the lives and
livelihoods of millions of persons around the world.
Following a brief decline in the first year of the pandemic, social unrest could be
further accentuated by the COVID-19 pandemic (Barrett and Chen 2021). The Coro-
navirus crisis has been having profound health and socio-economic implications
across various groups in society, with its adverse impact appearing particularly
concentrated among the most vulnerable segments of the population. In this con-
text, recent studies suggest that the pandemic could further exacerbate inequality
and poverty (Furceri et al. 2020; IMF 2020). Thus, to the extent that these socio-
economic variables are important drivers of popular discontent, social unrest could
be further fueled by the pandemic. Preliminary findings for 2021 suggest that the
frequency of unrest events has indeed started increasing (see Figure 1). In addition,
the ongoing cost of living crisis may further raise concerns about social unrest.
In light of these concerns, this paper uses a novel, high-frequency, index of
social unrest constructed by Barrett et al. (2022) to explore the macroeconomic
impact of social unrest episodes. In particular, using this novel index, the paper
applies the local projection method proposed by Jordà (2005) to address the follow-
ing questions: How does social unrest affect economic activity? What sectors and
components of aggregate demand are more likely to be adversely affected? Can
the adverse economic implications of social unrest be dampened by strong institu-
tions and available policy space? Are emerging markets and advanced economies
affected similarly? Do different types of unrest (by intensity and origin) have differ-
ent effects?
Results point to sizeable short-to medium-term economic costs associated with
episodes of social unrest. On average, economic activity declines sharply following
spikes in the unrest index or after an unrest event. For comparison, the average
impact of social unrest events that we estimate in this paper is larger than the
impact on oil exporters’ GDP of a typical oil price plunge and is comparable to the
impact of an oil price plunge combined with a recession (World Bank 2020). It is
also comparable to the impact of a fiscal consolidation of around 1 percent of GDP
(Carriere-Swallow, David, and Leigh 2021).
 0 10 20 30 40 50 60 70 80 90
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
year
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019

Figure 1: Frequency of unrest events. Authors’ calculations based on Barrett, Appendino, Nguyen and de Leon Miranda (2022).
2020
2021
919 — The Macroeconomic Impact of Social Unrest
920 — M. Hadzi-Vaskov et al.

Moreover, results are robust to reverse causality and endogeneity concerns. A

key challenge to identifying the impact of social unrest on GDP growth is that weak
economic activity is arguably a potential trigger of unrest. To tackle this, the paper
follows an instrumental variable approach similar to Acemoglu et al. (2019), where
social unrest is instrumented with waves of unrest at the regional level. As in the
baseline exercise, social unrest leads to a protracted reduction in GDP.
The paper also provides evidence on the channels through which social unrest
affects economic activity and shows that the effects depend on the nature of
the event and on country characteristics. The adverse effect of unrest on GDP
occurs mainly via contractions in manufacturing and services value added (sec-
toral dimension), and consumption (demand dimension). The economic impact of
unrest also differs by type of events: episodes motivated by socioeconomic rea-
sons result in sharper GDP contractions compared to those associated mainly with
politics/elections, while unrest triggered by a combination of both socio-economic
and political issues sees sharpest contractions. Institutions and policy space offer
mitigating factors, as countries with strong institutions and ample policy space
have more modest declines in activity following episodes of unrest. Finally, we find
that the economic impact of unrest is larger in countries that experienced adverse
growth trajectories prior to social unrest, suggesting that unrest compounds the
impact of low growth fundamentals; but the economic impact is still present even
in cases where countries do not experience low growth prior to unrest (further
addressing concerns of reverse causality). By contrast, the impact of unrest does not
appear to differ between countries pursuing fiscal consolidations—GDP systemat-
ically declines both in countries that experience sizeable and swift improvements
in the primary deficits and those that do not.
This paper contributes to the growing literature studying the economic impact
of social unrest and conflict. Studies of the macroeconomic effects of social unrest
are generally limited, with more focus in the literature commonly placed on large,
violent, conflicts. For instance, some studies have found that conflicts are a key hur-
dle to economic growth (Rodrik 1999), produce persistent negative effects on output
(Cerra and Saxena 2008; IMF 2019; Rother et al. 2016), and result in large overall
macroeconomic costs (Novta and Pugacheva 2020). A related strand of the literature
looks at the impact of political instability on economic performance, considering
social unrest as an element that negatively affects growth (Aisen and Veiga 2013;
Alesina and Perotti 1996; Jong-A-Pin 2009). We investigate a broad pool of social
unrest events, which—in addition to civil conflicts—also include considerably less
severe unrest episodes.1

1 See Section 5 in Barrett et al. (2022) for a comparison of the index used in this paper with other
measures of social unrest found in the literature.
 The Macroeconomic Impact of Social Unrest — 921

The pickup in episodes of social unrest around the globe in recent years has
raised the interest in understanding the causes and consequences of these events,
including through recent work at the IMF. Bernal-Verdugo, Furceri, and Guillaume
(2013) use the Cross-National Time Series Data (CNTS) to study the economic impact
of social and political instability and find evidence that these types of events
can have adverse effects on GDP that can last up to three years. More recently,
Barrett et al. (2021) investigate the effects of social unrest on stock markets across
the world. Hlatshwayo and Redl (2021) examine the macro-criticality of social unrest
and use machine-learning techniques to forecast future unrest. Saadi Sedik and Xu
(2020) investigate the dynamics among social unrest, major pandemic in history,
economic growth, and inequality.
Our paper expands this literature along two dimensions. First, by using the
data constructed by Barrett et al. (2022), this paper is able to study the impact of a
broader pool of unrest episodes on activity at a higher frequency, which improves
our ability to identify a causal effect.2 Indeed, compared to existing indicators, the
data by Barrett et al. (2022) provides higher frequency data for a broader set of
unrest events (than the Cross-National Time Series Data – CNTD, annual), broader
coverage (than the Armed Conflict Location and Event Database – ACLED, mainly
focused on SSA) and more objectivity and replicability (than International Country
Risk Guide – ICRG, which relies on subjective assessment). Second, we comple-
ment the work of Bernal-Verdugo, Furceri, and Guillaume (2013), which is closest
to this paper, by disentangling the impact of unrest on activity by looking at differ-
ent demand and supply components and by exploring the potential role of policy
space and institutions in mitigating the adverse impact of unrest.
The rest of this paper is organized as follows. Section 2 describes the data and
empirical methods used in the analysis. Section 3 presents the paper’s main find-
ings. Section 4 explores the channels through which unrest affects activity and the
factors mitigating its impact. Section 5 presents additional robustness exercises.
Finally, Section 6 concludes.

2 Data and Methodology

This section describes data sources and the econometric strategy pursued in
the paper. It begins by discussing in detail the social unrest index proposed by

2 A more ambitious exercise would explore high-frequency identification of the causal effect of
unrest on demand proxied by spending (for instance, using credit-cards transaction data); on
methodology, see the pioneering work by Cook and Hahn (1989), Kuttner (2001), and Cochrane and
Piazzesi (2002)
922 — M. Hadzi-Vaskov et al.

Barrett et al. (2022) and other macro data used in the analysis. It then discusses the
various local projection specifications used to study the impact of social unrest.

2.1 Social Unrest Data

Social unrest data comes from Barrett et al. (2022). They propose the reported social
unrest index (RSUI), a monthly news-based index starting in January 1985 that quan-
tifies the extent of social unrest for a large set of countries. The primary source
for news articles is Dow Jones’ Factiva news aggregator. The sample used in the
compilation of the index is restricted to printed articles published in major English-
language newspapers and networks in Canada, the UK and the US.
The RSUI takes the following form:
xi,t 100
RSUIi,t = × (1)
1 ∑
12 x̄ i ∕̄z
12
zt− j
j=1

where xi,t is the article count related to unrest in country i in month t; zt is the
overall article count in period t; and x̄ i and z̄ are the corresponding averages over
all time periods.3
In order to obtain the article counts used in the construction of the RSUI, Bar-
rett et al. (2022) employ a set of search criteria (Table 1). The search criteria for
determining the article count related to unrest in a specific country at a given
month are complex and include both inclusive and exclusive requirements. In par-
ticular, the inclusive requirements aim to select language related to unrest, such
as protests, riots, revolutions, and other forms of civil or domestic unrest. The
exclusive requirements aim to avoid false positives by omitting articles that con-
tain words that may erroneously be associated with unrest or may be related to
anniversaries of previous unrest episodes. In addition, the location criteria ensure
that the article refers to social unrest directed at a specific country, and the word
count ensures a minimum threshold of 100 words for articles to be included. The
RSUI measures the extent of social unrest for each country over time relative to its
country mean. The cross-sectional properties of the RSUI are described in detail in
Barrett et al. (2022).
In addition to constructing the ”index”, Barrett et al. (2022) propose an
algorithm to identify RSUI-implied ”events”, i.e. episodes of unrest that can be

3 Barrett et al. 2022 propose two indices, one which is normalized to have mean 100 (index A in
their paper) and one that is not (index B in their paper). The former is their preferred index, which
is the one we use in this paper.
 The Macroeconomic Impact of Social Unrest — 923

Table 1: Article search criteria.

x it y it zt

Must include Country name AND (“protest∗ ” OR “riot∗ ” OR Country Name “Today”
“revolution” OR ([“civil” or “domestic”] within 10 words AND “today”
of “unrest”)
Must exclude Country-specific terms OR “vote of protest” OR “protest
vote” OR “protestant∗ ” OR “anniversary” OR “war” OR
“memorial” OR “movie”
Location tag Country i Country i
Subject tag Domestic politics or civil unrest
Word count 100+ 100+ 100+

considered sizeable on the basis of large movements in the index. Such an event
has to satisfy the following three criteria:.
– It must be a local peak: RSUI i,t = max{RSUI i,t+1 , RSUI i,t−1 }
– The index must be large enough to satisfy one of the following conditions:
– RSUIi,t > RSUI i,t +(4∗ SD(RSUI i,t ));
– RSUI i,t is in the top 2 percentile;
– RSUI i,t exceeds the 20-year moving average by 4 times the 20-year stan-
dard deviation.
– The unrest article count for the month is at least 10 percent the average monthly
count for the country over the past 12 months.

Importantly, the authors also construct detailed event timelines from independent
narrative sources for more than a dozen country-specific case studies, showing that
these detailed events line up closely with RSUI-implied events. They thus conclude
that large changes in reported unrest are driven by highly newsworthy, real-world
events rather than mismatches, changes in media focus, bias, or other sources of
error.
In addition to these RSUI-implied events, there is a subset of social unrest
episodes located in the very extreme part of the RSUI distribution. We label such
an event that is satisfying all five criteria above as a major event.

2.1.1 Unrest Data Used in the Analysis

In the empirical analysis, we use the RSUI series and apply transformations of the
monthly data, as follows.
924 — M. Hadzi-Vaskov et al.

– First, for each country, we aggregate the RSUI at the quarterly level by taking
the maximum monthly level of the RSUI over the corresponding quarter.4 In this
way, we aim to ensure that a (local) spike in social unrest is reflected in the
quarterly series.5
– Second, we also aggregate the RSUI-implied event dummy constructed by the
authors and the additional major event dummy at the quarterly level (taking
the maximum over the quarter). Notably, we focus on new events—events that
are at least 8 quarters apart from past events. In the analysis, we use both the
RSUI index and the identified events, in separate exercises.

Figure 2 shows the evolution of the RSUI for the selected set of economies affected by
social unrest in 2019 that were mentioned in the introduction. The quarterly value
of RSUI is defined by the maximum of the original monthly RSUI over the 3 months
of that quarter. In the econometric analysis, which is further discussed below, we
exploit the time series properties of the dataset to estimate dynamic panel regres-
sions. See Barrett et al. (2022) for a discussion on the cross-sectional properties of
the distribution of the RSUI, which map a country’s RSUI level to its position in the
cross-sectional distribution.

2.1.2 Categorization of Events

In addition to using the events identified by Barrett et al. (2022), a contribution of

our analysis is to further categorize the events into four groups: political/elections;
socioeconomic; mixed (i.e. both of the above); or none of the above. This is achieved
by consulting contemporaneous news articles and reports to better understand the
underlying causes and triggers of all events that could be identified. In particular,
using contemporaneous news reports, we inspect the context and keywords associ-
ated with the description of each event (such as elections, electoral fraud, food/fuel
price protest, pension, social benefits, economic reforms, austerity, etc.) to deter-
mine the triggers of each of the previously identified events in our dataset.6 While

4 We aggregate at the quarterly level in order to match the GDP components. An alternative exer-
cise could explore high-frequency identification of the causal effect of unrest on demand proxied
by spending (for instance, using credit-cards transaction data); on methodology, see the pioneering
work by Cook and Hahn (1989), Kuttner (2001), and Cochrane and Piazzesi (2002).
5 As in other cases of data aggregation, this approach does not account for timing within the quar-
ter and does not allow to differentiate between effect of a spike in the RSUI at the beginning of
the quarter versus those occurring at the end of the quarter. Nonetheless, this issue may be more
problematic in static regressions rather than the local projections method employed in this study.
6 In particular, each event in the Barrett et al. (2022) dataset has a label (say yellow vests protest),
which we use to further find descriptions of the events.
 Chile France

4000
2000

3000
1500

2000
1000

RSUI
RSUI
500

1000
0
0
1985q1 1990q1 1995q1 2000q1 2005q1 2010q1 2015q1 2020q1 1985q1 1990q1 1995q1 2000q1 2005q1 2010q1 2015q1 2020q1
dateq dateq

Hong Kong SAR Lebanon

2500

4000
2000

3000
RSUI
RSUI

2000
1000 1500

1000
500

0
0
1985q1 1990q1 1995q1 2000q1 2005q1 2010q1 2015q1 2020q1 1985q1 1990q1 1995q1 2000q1 2005q1 2010q1 2015q1 2020q1
The Macroeconomic Impact of Social Unrest

dateq dateq
—

Figure 2: RSUI in selected economies. Authors’ calculations based on Barrett, Appendino, Nguyen and de Leon Miranda (2022).
925
926 — M. Hadzi-Vaskov et al.

certain degree of judgement is necessarily involved and unavoidable in such exer-

cises, we aim to reflect the broadest available evidence that can be found for certain
identified social unrest event in the public sources. All remaining events for which
we could not clearly identify the underlying reasons as political, economic or mixed
are included in a residual category of events that cannot be labeled.
In total, we categorize 490 events as having political/elections as the domi-
nant underlying reasons, 101 socio-economic, 40 mixed political and economic rea-
sons. The underlying reasons for the remaining 205 events could not be clearly
identified.

2.2 Macroeconomic Data and Sample

We include all 89 countries for which RSUI, quarterly GDP, and commodity terms-
of-trade data are available over the period 1990–2019 (Table 2). We exclude fragile
states—those labeled as fragile states by the World Bank in at least one year since
2006—in order to avoid our results being driven by countries in enduringly fragile
social conditions or facing prolonged episodes of violence, as we are interested in
a broad category of unrest not in extreme conflict situations (for which there is
extensive literature as discussed in the introduction).
We collect quarterly GDP and other national accounts data from national
sources (seasonally adjusted by authorities if available, otherwise we use Haver for
seasonal adjusted data). Data series on debt-to-GDP come from the Fiscal Monitor
database. Rule of law measures come from the World Governance Indicators (Kauf-
mann, Kraay, and Mastruzzi 2010), and data on exchange rate regimes comes from
Ilzetzki, Reinhart, and Rogoff (2019). We retrieved data series for confidence indi-
cators from Haver. Data on uncertainty comes from the latest vintage of the World
Uncertainty Index proposed by Ahir, Bloom, and Furceri (2022). Commodity terms-
of-trade data comes from Gruss and Kebhaj (2019). Data on bilateral distance and
contiguity comes from CEPII’s Geodist dataset. Finally, data on primary balances
comes from the IMF’s World Economic Outlook dataset.

2.3 Econometric Approach

We assess the macroeconomic impact of episodes of social unrest using the local
projection method proposed by Jordà (2005). The method’s flexibility has the advan-
tage that it does not constrain the shape of the impulse response functions. The
benchmark specification at a quarterly frequency is as follows:

∑
h
yi,t+h − yi,t−1 = 𝛼ih + 𝛾th + 𝛽 h unresti,t + 𝜃 Xi,t + 𝜇 j,h unresti,t+ j + 𝜀i,t+h (2)
j=1
 The Macroeconomic Impact of Social Unrest — 927

Table 2: Sample.

32 advanced economies 57 emerging markets and low-income countries

Australia (1990Q1-2019Q4), Austria Albania (2008Q1-2019Q4), Argentina

(1996Q1-2019Q4), Belgium (1995Q1-2019Q4), (1990Q1-2019Q4), Azerbaijan
Canada (1990Q1-2019Q4), Cyprus (2001Q1-2019Q4), Bahrain (2008Q1-2019Q4),
(1995Q1-2019Q4), Czech Republic Belarus (2009Q1-2019Q4), Bolivia
(1996Q1-2019Q4), Denmark (1990Q1-2019Q4), (1990Q1-2019Q4), Brazil (1990Q1-2019Q4),
Estonia (1995Q1-2019Q4), Finland Bulgaria (1995Q1-2019Q4), Chile
(1990Q1-2019Q4), France (1990Q1-2019Q4), (1990Q1-2019Q4), China (1992Q1-2019Q4),
Germany (1990Q1-2019Q4), Greece Colombia (2000Q1-2019Q4), Croatia
(1995Q1-2019Q4), Hong Kong SAR (2000Q1-2019Q4), Ecuador (1990Q1-2019Q4),
(1990Q1-2019Q4), Ireland (1995Q1-2019Q4), Egypt (2006Q3-2019Q4), Ghana
Israel (1995Q1-2019Q4), Italy (2006Q1-2019Q4), Guatemala
(1990Q1-2019Q4), Japan (1990Q1-2019Q4), (2001Q1-2019Q4), Honduras
Korea (1990Q1-2019Q4), Latvia (2000Q1-2019Q4), Hungary (1995Q1-2019Q4),
(1995Q1-2019Q4), Lithuania (1995Q1-2019Q4), India (1996Q2-2019Q4), Indonesia
Netherlands (1990Q1-2019Q4), New Zealand (1990Q1-2019Q4), Iran (2004Q2-2019Q4),
(1990Q1-2019Q4), Norway (1990Q1-2019Q4), Jordan (1992Q1-2019Q4), Kazakhstan
Portugal (1995Q1-2019Q4), Singapore (1999Q1-2019Q4), Kenya (2009Q1-2019Q4),
(1990Q1-2019Q4), Slovakia (1995Q1-2019Q4), Kuwait (2010Q1-2019Q4), Kyrgyz Republic
Slovenia (1995Q1-2019Q4), Spain (2012Q1-2019Q4), Macedonia
(1995Q1-2019Q4), Sweden (1990Q1-2019Q4), (2000Q2-2018Q3), Malaysia (1991Q1-2019Q4),
Switzerland (1990Q1-2019Q4), UK Mexico (1990Q1-2019Q4), Moldova
(1990Q1-2019Q4), United States (2010Q1-2019Q4), Montenegro
(1990Q1-2019Q4) (2011Q1-2019Q4), Morocco (1998Q1-2019Q4),
Nicaragua (2006Q1-2019Q4), Panama
(2007Q1-2019Q4), Paraguay (1995Q1-2019Q4),
Peru (1990Q1-2019Q4), Philippines
(1990Q1-2019Q4), Poland (1995Q2-2019Q4),
Qatar (2010Q1-2019Q4), Romania
(1995Q1-2019Q4), Russia (1995Q1-2019Q4),
Rwanda (2006Q1-2019Q4), Saudi Arabia
(2010Q1-2019Q4), Senegal (2017Q1-2019Q4),
Serbia (1995Q1-2019Q4), South Africa
(1990Q1-2019Q4), Sri Lanka (2010Q1-2019Q4),
Tanzania (2010Q1-2019Q3), Thailand
(1993Q1-2019Q4), Tunisia (2000Q1-2019Q4),
Turkey (1990Q1-2019Q4), Uganda
(2008Q1-2019Q4), Ukraine (2001Q1-2019Q4),
United Arab Emirates (2012Q1-2019Q4),
Venezuela (1997Q1-2018Q4), Vietnam
(2008Q4-2019Q4), Zambia (2013Q1-2019Q3)
928 — M. Hadzi-Vaskov et al.

where t captures time, i denotes countries, h is the time horizon in quarters after
unrest, y is the variable of interest (GDP, sectoral value added, demand compo-
nents, confidence); unrest is either the index proposed by Barrett et al. (2022)
(RSUI) or the event dummy; 𝛼ih are the horizon-specific country fixed effects; 𝛾th are
horizon-specific quarter/year fixed effects, and X it are a set of controls that includes
past values of the dependent variable, past values of terms-of-trade growth, and
past values of the RSUI. As proposed by Teulings and Zubanov (2014), we include
∑h
the term j=1 𝜇 h, j unresti,t+ j to control for the fact that unrest events may be per-
sistent. When we study the impact of the index, this term includes forward values of
the index; when we study the impact of the event dummy, this term includes future
events (regardless of whether they are new or not). Excluding this term could bias
our estimates of 𝛽 h , which is our parameter of interest.
One potential caveat of a causal interpretation of the econometric approach
described above is the potential reverse causality running from growth perfor-
mance to social unrest and the potential omitted variable bias. We tackle these
concerns by implementing an IV approach that exploits regional waves of social
unrest. In addition, we show evidence that the impact of unrest on GDP does not
seem to be driven by preceding periods of weak activity or by fiscal consolidations,
two factors that may result in both low growth and in social unrest. More details
are presented later in the paper.

2.3.1 State-dependent Local Projections

In addition to the benchmark regression presented previously, we explore specifica-

tions that condition the response of activity to social unrest to particular state vari-
ables (see below). Indeed, one of the main advantages of the local projection method
in estimating the effects of shocks is its flexibility in dealing with non-linearities and
state dependency (Ramey and Zubairy 2018). Hence, the typical specification of the
regression equation with state-dependent variables takes the following form:
[
yi,t+h − yi,t−1 = F(zi,t−1 ) 𝛼high
h
,i + 𝛾high,t + 𝛽high unresti,t + 𝜃high Xi,t
h h

]
∑
h
j ,h
+ 𝜇high unresti,t+ j
j=1
[
+ (1 − F(zi,t−1 )) 𝛼loh 𝑤,i + 𝛾loh 𝑤,t + 𝛽loh 𝑤 unresti,t + 𝜃lo𝑤 Xi,t
]
∑
h
+ 𝜇loj,h𝑤 unresti,t+ j + 𝜀i,t+h (3)
j=1
 The Macroeconomic Impact of Social Unrest — 929

where F (zi,t−1 ) is a smooth function of the state variable taking a value between
0 and one; and z is a normalized version of the state variable such that F(0) = 21 .
When presenting results, we plot estimates for 𝛽high
h
and 𝛽loh 𝑤 , the two parameters
capturing differences in response to unrest across states. While many state vari-
ables could affect the impact of social unrest on growth, in this analysis we are
primarily interested in state variables that reflect the economies’ institutional and
policy settings. In what follows we present results for three state variables:.
– Rule of law;
– Debt level;
– Exchange rate flexibility

These are broad indicators that have the advantage of being available for a large
set of countries and for a relatively long period of time (most of the time horizon
in our analysis).7 Each of them represents a different perspective of the two dimen-
sions that we explore as potential intermediaries in the relationship between social
unrest and growth (namely institutions and policy space). In the exercise, each of
the state variables is normalized such that the standardized variables (z) have mean
zero and a standard deviation equal to 1. In other words, countries with low (high)
rule of law/debt levels are those that have values below (above) the average in our
sample. In the case of exchange rate flexibility, F (z) is an indicator function that
takes the value of one if the “fine” exchange rate classification constructed by Ilzet-
zki, Reinhart, and Rogoff (2019) takes a value higher than 9,8 and zero otherwise. In
turn, as is common in the literature, we assume:
( ) ( ( ))
F(zi,t−1 ) = exp −𝜆0 zi,t−1 ∕ 1 + exp −𝜆0 zi,t−1 (4)

where 𝜆0 takes value 1.5 (as in Auerbach and Gorodnichenko 2012).9

3 Results
This section presents the estimated impact of both changes in the RSUI index and
the social unrest events on GDP. It also presents results from an IV exercise aimed
at addressing endogeneity concerns.

7 In Hadzi-Vaskov, Pienknagura, and Ricci (2021) we also study the differential impact of social
unrest on economic activity for alternative structural factors (labor market flexbility and product
market competition).
8 A country with classification of nine is one that has a “preannounced crawling band that is wider
than or equal to +/−2 percent”.
9 Results are robust to alternative choices for 𝜆0 .
930 — M. Hadzi-Vaskov et al.

3.1 Impact Based on the Unrest Index (RSUI)

Figure 3 presents our baseline results of the impact of the RSUI on economic activity
for 16 quarters. We show the impact of RSUI in terms of deviations of GDP from the
pre-unrest period (baseline). It shows that GDP experiences a steady decline follow-
ing a change in the RSUI of one standard deviation in the RSUI for about six quarters,
and recovers thereafter. For the overall sample, GDP declines by over −0.15 percent
qoq on impact. After 6 quarters, quarterly GDP remains about 0.2 percent below its
pre-unrest level. The impact of social unrest becomes statistically non-significant
after about 10–12 quarters, a finding that is consistent with Bernal-Verdugo, Furceri,
and Guillaume (2013). Going forward, the rest of the paper will focus on the 12-
quarter horizon. However, estimates point to potentially large accumulated GDP
losses due to social unrest.
How large of a change corresponds to one standard deviation in the RSUI? Our
calculations suggest that a change of one standard deviation is equivalent to the
protests following the Peña Nieto election in 2012 or Chile’s presidential election
protests in 2013. For comparison, the protests of July 2019 in Hong Kong SAR and
the yellow vest protests of 2018 in France resulted in an increase of 4 standard devi-
ations in the RSUI, while the events of October/November 2019 in Chile resulted in
an increase of RSUI of 10 standard deviations.

3.2 Impact Based on Unrest Events

While the analysis so far has concentrated on changes in the unrest index, as mea-
sured by the RSUI, this section explores the impact of RSUI-implied events, identified
in line with the criteria explained in Section 2.1. Accordingly, Figure 4 documents
the impact of social unrest events on economic activity. We restrict the analysis to
new (both overall and only major) events, according to the criteria presented in
Section 2.3, to distinguish these events from those that are part of a persistent wave
of unrest.
We confirm that social unrest events exert a negative impact on GDP, and these
effects are statistically significant. An RSUI-implied event lowers GDP by about 0.6
percent on impact relative to the baseline, and this effect grows to approximately 1
percent after 1 year, while it tends to dissipate after two and half years. Not surpris-
ingly, major events lead to even larger GDP contractions. Moreover, similar as the
results for changes in the RSUI, these (major) unrest events have a persistent effect
on economic activity, even beyond three years.
 .1
-.1 0
Change relative to GDP in t-1 (in percent)
-.2 -.3
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Quarters since unrest

90 percent confidence band Average effect

baseline
The Macroeconomic Impact of Social Unrest
—

Figure 3: Impact of social unrest on GDP (response to a one st. dev. change in the RSUI). Shaded area indicates 90 percent confidence
interval with standard errors clustered at the country and year level.
931
 New Event New Major Events
932

.5
.5
—

0
-.5

-.5
M. Hadzi-Vaskov et al.

-1

-1
-1.5

-1.5
Change relative to GDP in t-1 (in percent)
Change relative to GDP in t-1 (in percent)
-2

-2
-2.5

-2.5
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Figure 4: Impact of social unrest conditional on the extent of the event (response after a new event identified by algorithm). Shaded
area indicates 90 percent confidence interval with standard errors clustered at the country and year level.
 The Macroeconomic Impact of Social Unrest — 933

3.3 Addressing Endogeneity Concerns: IV Estimations

The results presented thus far have been interpreted as a causal relationship
running from unrest to growth, but this interpretation is subject to caveats. One
concern is that unrest may be preceded (maybe caused) by stints of low growth.
If this were the case, our results for the effect of social unrest on growth may be
picking up the persistence of adverse growth events (reverse causality). Another
possibility may be that both unrest and growth are driven by variables that are
not properly captured in the econometric specification (omitted variables). For
example, fiscal austerity plans can lead to both bouts of social unrest (Ponticelli
and Voth 2020) and economic contractions, in which case the relationship between
unrest and growth would be spurious. However, it is important to note that recent
work shows that episodes of unrest are hard to predict using observable data,
including GDP growth (see Saadi Sedik and Xu 2020; and Hlatshwayo and Redl 2021),
decreasing the likelihood that our results are driven by reverse causality or omitted
variables,
Nonetheless it is important to address the endogeneity concern. For this pur-
pose this section presents results for a panel regression instrumental variables (IV)
strategy where we exploit the timing of unrest episodes across countries.10 More
specifically, we instrument social unrest in country i with current and past episodes
of social unrest in nearby countries. As documented in Barrett et al. (2022), episodes
of social unrest generate spillovers at the regional level: social unrest in one country
is associated with a 1 percent increase in the probability that neighbors experience
social unrest in the next six months. This identification strategy (namely, identi-
fication through regional waves) has been also used to study the causal effects
of democratization on growth (Acemoglu et al. 2019) and to study the impact of
fiscal austerity on social unrest (Ponticelli and Voth 2020). For robustness, when
operationalizing the instrument, we construct three versions of the variable cap-
turing “regional waves”. The first is a distance-weighted measure of social unrest
for country i’s partners, as follows:

∑ 1∕ log(disti, j )
unresti−,ti,W = ∑ unrest j,t (5)
j∈W−i
(1∕ log(disti,k )
k ∈W−i

where W−i is the set of all countries with available data on social unrest exclud-
ing country i, unrest j,t is either RSUI or the RSUI-implied events, and disti, j is
the bilateral population-weighted distance between country i and country j, as

10 Section 5 presents additional robustness exercises aimed at studying the role of factors that may
affect growth and unrest simultaneously.
934 — M. Hadzi-Vaskov et al.

presented in the CEPII Geodist dataset. The second measure is a similar index using
only regional partners (where regions are defined using the IMF regional classi-
fication), unrest− i,t
i,R
. And the third one is an index that only takes into account
contiguous countries (countries sharing a land border), regardless of their region,
unrest−
i,t
i,C
. With those indices, we follow an IV strategy where RSUI and the RSUI-
implied events are instrumented using, in turn, unrest−
i,t
i,W
and its lagged values,
unrest−i,t
i,R
and its lagged values, and unrest−
i,t
i,C
and its lagged values.11 In the first
stage we also include region-time fixed effects to capture any regional time-varying
variable that may affect waves of unrest.
The results confirm the finding that both spikes in the RSUI and episodes of
social unrest lead to lower GDP compared to countries that do not suffer social
unrest.12, 13 A one standard deviation increase in RSUI leads to a 0.2 to 0.3 percent
decline in GDP on impact, depending on the instrument used (Figure 5).14 As in the
baseline specification, the impact of social unrest in the IV specification are also
persistent over our 6-quarter window.
Similar results are obtained when studying the impact of RSUI-implied unrest
events (Figure 6). New episodes of social unrest lead to a 1 to 2 percent decline in
GDP on impact relative to baseline, and the impact grows to 2 to 3 percent after 6
quarters. The estimated effects are roughly two to three times as large as those in
the baseline specification. Figure 7 shows that the same holds when analyzing the
impact of major events.

4 Exploring Heterogeneity
So far we have shown that, on average, social unrest has an adverse and protracted
impact on GDP. Moreover, we showed that, while endogeneity may be a concern,
results are robust to an IV strategy that exploits regional waves of unrest. The results
in Section 3, however mask potential heterogeneity on the economy-wide impact
of social unrest and on the role of country characteristics. This section explores
several layers of heterogeneity, which help understand the channels through which
unrest affects activity and the role that country fundamentals have in mitigating the
economic impact of unrest.

11 In each case we instrument the unrest variable with both RSUIi−,ti and two lags and eventi−,ti and
two lags.
12 In all cases, the hypothesis of weak instruments is rejected.
13 IV results for other outcome variables are available upon request.
14 Hadzi-Vaskov, Pienknagura, and Ricci (2021) show results using a GDP-weighted index of unrest,
with similar quantitative impacts.
 Baseline Result Instrument: RSUI and events in countries in

.2
the world, weighted by distance

.2

0
0

-.2
-.2

-.4
-.4

-.6
-.6
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)
Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Instrument: RSUI and events in countries in Instrument: RSUI and events in contiguous countries
the region, weighted by distance

.2

.2
0

0
-.2

-.2
-.4

-.4
-.6

-.6
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)

Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
The Macroeconomic Impact of Social Unrest

baseline baseline
—

Figure 5: Impact of social unrest on GDP, instrumental variables approach (response to a one st. dev. change in the RSUI). Shaded
area indicates 90 percent confidence interval with standard errors clustered at the country and year level.
935
 Baseline Result Instrument: RSUI and events in countries in
936

4
the world, weighted by distance

2
—

0
0

-6 -4 -2
-6 -4 -2
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)
M. Hadzi-Vaskov et al.

Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Instrument: RSUI and events in countries in Instrument: RSUI and events in contiguous countries

4
the region, weighted by distance

4
2

2
0

0
-6 -4 -2

-6 -4 -2
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)

Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Figure 6: Impact of social unrest on GDP, instrumental variables approach (response to a new RSUI-Implied event). Shaded area
indicates 90 percent confidence interval with standard errors clustered at the country and year level.
 Baseline result Instrument: RSUI and events in countries in
the world, weighted by distance

-8 -7 -6 -5 -4 -3 -2 -1 0 1
-8 -7 -6 -5 -4 -3 -2 -1 0 1
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)
Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Instrument: RSUI and events in countries in Instrument: RSUI and events in contiguous countries
the region, weighted by distance

-8 -7 -6 -5 -4 -3 -2 -1 0 1
-8 -7 -6 -5 -4 -3 -2 -1 0 1
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)

Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
The Macroeconomic Impact of Social Unrest

baseline baseline
—

Figure 7: Impact of social unrest on GDP, instrumental variables approach (response to a new RSUI-Implied major event). Shaded area
indicates 90 percent confidence interval with standard errors clustered at the country and year level.
937
938 — M. Hadzi-Vaskov et al.

4.1 Differences Across Sectors and Demand Components, and

Potential Channels of Transmission
The assessment of the effect of changes to the RSUI on aggregate economic activity
may hide heterogeneity across sectors of the economy and across aggregate demand
components. These differences may stem from the fact that episodes of social unrest
may affect unevenly some areas (urban vs rural areas) or economic agents with
distinct patterns of sectoral specialization and consumption. Understanding these
differences may also provide a better understanding of the mechanics through
which unrest affects aggregate growth and its potential impact on medium-term
performance. Hence, in Figures 8 and 9 we present results for GDP’s components
from the sectoral/supply side and demand side, respectively.
On the one hand, Figure 8 shows that the adverse effects on GDP appear to
be driven by sharp contractions in services and manufacturing. The effects after
6 quarters appear to be similar in magnitude across the two sectors and slightly
higher than for GDP (about 0.4 percent relative to baseline, respectively). How-
ever, the decline in services activity appears to be more immediate while that of
manufacturing appears to be more gradual. On the other hand, the impact on agri-
culture is not significant at any point over the horizon. This result is not surprising
given that social unrest is typically associated with larger urban centers and loca-
tions with higher concentration of services activity and manufacturing production,
and thereby more likely to directly impede the normal functioning of these sec-
tors; unless social unrests escalate and jeopardize the stability and security of the
whole country, they are unlikely to pose commensurate obstacles to agriculture
production.
Turning to the demand-side components, Figure 9 shows that the effect of social
unrest is more evident on consumption. The impact on consumption is negative
and significant over most of the six quarters following unrest. While unrest seems
to lower investment as well, this effect is not statistically significant. The differ-
ence in the response of consumption and investment suggests that the impact of
social unrest on perturbating medium-to long-term expectations (critically impor-
tant for investment) is more limited. The lack of statistically significant impact of the
unrest index on investment certainly requires further investigation.15 Inevitably,
however, average effects likely mask country-specific experiences, some of which
may be characterized by substantial deteriorations of long-term expectations and
sharper contractions of investment relative to consumption. Finally, both exports
and imports fall following spikes in the social unrest index.

15 This is particularly in light of its impact on uncertainty elaborated below and the relationship
between uncertainty and investment as in Bloom (2009).
 Agriculture Manufacturing

.5
0 .5

0
-.5

-.5
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to level in t-1 (in percent)

Change relative to level in t-1 (in percent)
Quarter since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Services

-.5 0 .5
-1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to level in t-1 (in percent)

Quarters since unrest

90 percent confidence band Average effect

The Macroeconomic Impact of Social Unrest

baseline
—

Figure 8: Impact of social unrest on economic sectors (response to a one st. dev. change in the RSUI). Shaded area indicates 90
percent confidence interval with standard errors clustered at the country and year level.
939
 Consumption Investment
940

1
1
—

.5
.5

0
0

-.5
-.5

-1
-1
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to level in t-1 (in percent)

Change relative to level in t-1 (in percent)
Quarters since unrest Quarters since unrest
M. Hadzi-Vaskov et al.

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Exports and Imports

-1 -.5 0 .5 1
-1 0 1 2 3 4 5 6 7 8 9 10 11 12
Quarters since unrest

Change relative to level in t-1 (in percent)

CI exports Exports
Imports CI Imports
CI Imports baseline

Figure 9: Impact of social unrest on demand components (response to a one st. dev. change in the RSUI). Shaded area indicates 90
percent confidence interval with standard errors clustered at the country and year level.
 The Macroeconomic Impact of Social Unrest — 941

Figure 10 documents the impact of social unrest on uncertainty as well as, con-
sumer and business confidence. The left panel shows that changes to the RSUI are
associated with a significant increase in the World Uncertainty Index for the cor-
responding economy for a few quarters, before dying out during the first year. In
addition, social unrest is associated with a drop in consumer confidence, albeit
generally not statistically significant.
Consistent with the findings above, we find that new unrest events result in
a decline in manufacturing and services value added (Figure 11), a contraction
in consumption, exports and imports (Figure 12), and a reduction in consumer
confidence (Figure 13). We do not find a statistically significant effect of unrest
events on agriculture value added, investment, and business confidence.

4.2 Impact of Different Type of Unrest Events

Social unrest can be motivated by various socio-economic and/or political reasons
and sparked by various triggers. Hence, in Figure 14 we provide a comparison of
the impact on economic activity of different types of unrest events (defined by
their underlying reasons/triggers, as described in Section 2.1). There are several
interesting findings. First, all three types of events analyzed—socio-economic, polit-
ical, and mixed—lead to persistent reductions in economic activity. Second, unrest
episodes motivated by socio-economic issues lead to sharper GDP contractions than
episodes related to politics/election. Third, episodes triggered by a combination
of socio-economic and political factors are associated with the largest GDP con-
tractions. Finally, it seems that other events that cannot be labeled as triggered
by socio-economic or political factors do not bear a negative effect on economic
activity.
As discussed, differences in GDP’s response to different social unrest triggers
are potentially subject to endogeneity concerns. In particular, social unrest events
triggered by socio-economic causes could be associated with past weak activity,
which itself may affect future activity. To dispel these concerns, Section 3.3 docu-
ments robustness to instrumental variable regressions, while Section 5 shows that
the impact of unrest is robust to controlling for past sluggish growth.

4.3 Differences Across Income Groups

While the results so far have included all economies in our dataset, Figure 15 makes
a distinction between advanced economies (AEs) and emerging markets and devel-
oping economies (EMDEs). The findings in Figure 15 suggest that the adverse effects
of social unrest are evident in all countries regardless of income levels. However, the
effect on impact is twice as large in EMDEs relative to AEs, and remains somewhat
larger throughout the window of analysis.
 World Uncertainty Index Consumer Confidence
942

.5 1
—

-.5 0

Change relative to level in t-1

Change relative to level in t-1

-.01 -.005 0 .005 .01 .015

-1.5 -1
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
Quarters since unrest Quarters since unrest
M. Hadzi-Vaskov et al.

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Business Confidence

0 .5 1
Change relative to level in t-1
-.5
-1 0 1 2 3 4 5 6 7 8 9 10 11 12
Quarters since unrest

90 percent confidence band Average effect

baseline

Figure 10: Impact of social unrest on uncertainty and confidence (response to a one st. dev. change in the RSUI). Shaded area
indicates 90 percent confidence interval with standard errors clustered at the country and year level.
 Agriculture Manufacturing

2
1 2

1
0

0
-1

-1
-2

-2
-3

-3
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to level in t-1 (in percent)

Change relative to level in t-1 (in percent)
Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Services

-3 -2 -1 0 1 2
-1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to level in t-1 (in percent)

Quarter since unrest

90 percent confidence band Average effect

The Macroeconomic Impact of Social Unrest

baseline
—

Figure 11: Impact of unrest events across sectors (response after a new event identified by algorithm). Shaded area indicates 90
percent confidence interval with standard errors clustered at the country and year level.
943
 Consumption Investment
944

2
2
—

0
0

-2
-2

-4
-4

-6
-6
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to level in t-1 (in percent)

Change relative to level in t-1 (in percent)
Quarters since unrest Quarters since unrest
M. Hadzi-Vaskov et al.

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Exports Imports

2
2

0
0

-2
-2

-4
-4

-6
-6
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to level in t-1 (in percent)

Change relative to level in t-1 (in percent)
Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Figure 12: Impact of unrest events across demand components (response after a new event identified by algorithm). Shaded area
indicates 90 percent confidence interval with standard errors clustered at the country and year level.
 Consumer Confidence Business Confidence

2
2

0
0

-2
-2

Change relative to level in t-1

Change relative to level in t-1

-4
-4

-6
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
Quarters since unrest Quarter since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline
The Macroeconomic Impact of Social Unrest
—

Figure 13: Impact of unrest events on confidence (response after a new event identified by algorithm). Shaded area indicates 90
percent confidence interval with standard errors clustered at the country and year level.
945
 Response to Events Triggered by Politics Response to Events Triggered by Social Issues
946

5
5
—

0
0

-5
-5

-10
-10

-15
-15
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)
Quarters since unrest Quarters since unrest
M. Hadzi-Vaskov et al.

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Response to Events Associated with Both Response to Events that Cannot be Labeled

5
5

0
0

-5
-5

-10
-10

-15
-15
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)
Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline

Figure 14: Impact of different types of events on GDP. Shaded area indicates 90 percent confidence interval with standard errors
clustered at the country and year level.
 Advanced Economies Emerging Markets

.1

.1
0

0
-.1

-.1
-.2

-.2
Change relative to GDP in t-1 (in percent)
Change relative to GDP in t-1 (in percent)
-.3

-.3
-.4
-.4
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
Quarters since unrest Quarters since unrest

90 percent confidence band Average effect 90 percent confidence band Average effect
baseline baseline
The Macroeconomic Impact of Social Unrest
—

Figure 15: Impact of social unrest on GDP by income group (response to a one st. dev. change in the RSUI). Shaded area indicates 90
percent confidence interval with standard errors clustered at the country and year level.
947
948 — M. Hadzi-Vaskov et al.

4.4 The Role of Institutions and Policy Space

Figure 16 presents results focusing on the role of institutional settings and policy
space in mitigating the impact of social unrest on economic activity, based on the
state-dependent local projections discussed above. Strong institutions—measured
here by the level of rule of law—could play an important role in allowing a robust
engagement of civil society, which is crucial in mature democracies, without leading
to disproportionate economic costs (see Acemoglu and Robinson 2019). Similarly,
having ample policy space may allow countries to buffer the direct economic impact
of social unrest or to meet the demands posed by society.
Indeed, the result in Figure 16, left panel, shows that countries with strong insti-
tutions see no decline in activity in the aftermath of episodes of unrest relative
to baseline. By contrast, countries with weak institutions experience a 0.4 percent
decline in activity six quarters after episodes of unrest. Similarly, economies with
ample policy space—measured by the public debt level and the degree of exchange
rate flexibility—are found to better cope with the adverse impact of social unrest
episodes.16

5 Robustness to Confounding Factors

This section presents two additional exercises aimed at reinforcing the main find-
ings of the paper—the adverse impact of social unrest on economic activity—by
ruling out the possibility that they are mainly driven by other factors, thus further
contributing to dispel the concerns of endogeneity. First, it studies the impact of
social unrest in countries that were experiencing low growth prior to the social
unrest event and in those that were not. Second, it studies the impact of social unrest
on growth in both countries that experience large year-on-year improvements in
primary fiscal balances (likely linked to fiscal consolidations) and those thatdid not.
We present analysis controlling for prior adverse growth events, both using
the event dummy and the RSUI. We begin by defining prior low growth events
on the basis of country-specific negative deviations of country growth from the
country average growth for at least two out of the last three quarters before the
social unrest.17 Using this definition, 28 percent of the new RSUI-implied events are

16 The threshold for the debt to GDP ratio to define whether a country has high or low debt is 54
percent. As mentioned, this is the average in our sample.
17 The results are robust to alternative definitions of a low growth evento, based on: growth devi-
ations from country-specific moving averages, as well as just negative growth in the period before
the social unrest.
 By Rule Law By Debt Level

.5
.5

0
-.5 0

-.5
-1

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)

-1
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
Quarters since unrest Quarters since unrest

By XR-flexibility

-.5 0 .5
Change relative to GDP in t-1 (in percent)
-1
-1 0 1 2 3 4 5 6 7 8 9 10 11 12
The Macroeconomic Impact of Social Unrest

Quarters since unrest

—

Figure 16: Impact of social unrest conditional on institutions and policy space (response to a one st. dev. change in the RSUI). Black
(red) lines correspond to weak (strong) institutions/policy space. Shaded (dashed lines) area indicates 90 percent confidence interval
949

with standard errors clustered at the country and year level.

950 — M. Hadzi-Vaskov et al.

preceded by low growth events. This strategy is implemented both for the analysis
based on the RSUI and the RSUI-derived new events.
More specifically, when using “new event” dummies we run the following
regression:

yi,t+h − yi,t−1 = 𝛼ih + 𝛾th + 𝛽sg

h
sgeventi,t + 𝛽sh seventi,t + 𝛽 gh geventi,t + 𝜃 Xi,t

∑
h
+ 𝜇 j,h unresti,t+ j + 𝜀i,t+h (6)
j=1

where sg eventi,t is a dummy taking value one if country i experienced both a social
unrest event in period t and two or three quarters of below-average growth in the
three quarters preceding the unrest event. The dummy variable s eventi,t takes
value of one if country i experienced a social unrest event and did not experience
two or three quarters of below average growth in the three quarters preceding the
identified unrest event. Finally, the dummy variable g eventi,t takes value of one
if country i experienced two or three quarters of below-average growth prior to
period t and did not experience a social unrest event. The excluded group includes
countries that neither experienced unrest nor low growth. Such a regression allows
us to show the results for four mutually exclusive cases: i) only social unrest event;
ii) only growth event; iii) both events; and iv) no event. When using the RSUI, the
econometric specification takes the following form:
[
yi,t+h − yi,t−1 = 𝛼ih + 𝛾th + lowi,t 𝛿loh 𝑤 + 𝛽loh 𝑤 RSUIi,t + 𝜃lo𝑤 Xi,t
]
∑
h
+ 𝜇loj,h𝑤 unresti,t+ j
j=1
[
+ (1 − lowi,t ) 𝛿nolo
h
𝑤 + 𝛽nolo𝑤 RSUIi,t + 𝜃nolo𝑤 Xi,t
h

]
∑
h
j ,h
+ 𝜇nolo 𝑤
unresti,t+ j + 𝜀i,t+h (7)
j=1

where lowi,t is a dummy variable that takes value of one if country i experienced
two or three quarters of below average growth prior to the event. Thus, 𝛽loh 𝑤
captures the effect of an increase in the social unrest index (RSUI) in countries with
prior low growth, and 𝛽nolo
h
𝑤
captures the effect of an increase in the social unrest
index (RSUI) in countries with no prior low growth.
Figure 17 shows the impact of social unrest on economic activity after control-
ling for prior adverse growth events, as defined above (note that all four effects
 Impulse response to events of both Impulse response to events of
social unrest and low growth low growth and no social unrest

0 1
0
-1
-2
-3

-4 -3 -2 -1
-4
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)
Quarters since unrest Quarter since unrest

90 percent confidence band social unrest and low growth 90 percent confidence band no social unrest and low growth

baseline baseline

Impulse response to events of Difference in impulse response between events of low

social unrest and no low growth growth and social unrest and low growth and no social unrest

1
1

0
0

-4 -3 -2 -1
-4 -3 -2 -1
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)
Quarters since unrest Quarters since unrest

90 percent confidence band social unrest and no low growth 90 percent confidence band difference

baseline baseline
The Macroeconomic Impact of Social Unrest
—

Figure 17: Impact of social unrest on GDP after controlling for prior adverse growth events (response to growth and social unrest
events). Shaded area indicates 90 percent confidence interval with standard errors clustered at the country and year level.
951
952 — M. Hadzi-Vaskov et al.

are estimated simultaneously). The top panels show that adverse growth events
are persistent, as expected, but the decline in GDP is larger when growth events
are combined with episodes of social unrest. Crucially, the lower left panel shows
that social unrest significantly affects growth even if not preceded by adverse
growth events, with a GDP contraction of about 1 percent after 6 quarters. Similarly,
the lower right panel shows that social unrest leads to a larger decline in eco-
nomic activity in countries experiencing below average growth prior to the unrest
event compared to countries that experience below average growth but no social
unrest event. The last two results point to the robustness of our results to reverse
causality, by confirming both the standard prior that an unrest event preceeded by
low growth is associated with stronger effect on growth, and the case where unrest
not preceeded by low growth still affects growth subsequently.
Figure 18 shows similar analysis using the RSUI index, focusing on the
”differential impact” of spikes in the social unrest index in countries with low
growth prior to the spike versus those that did not have low growth prior to the
spike. In countries with low prior growth (black line), GDP gradually falls after a
one standard deviation increase in the RSUI, and after six quarters it stands approx-
imately −0.2 percent below the baseline (countries with low growth and no spike
in the index). Countries that did not experience an episode of low growth prior
to the spike in RSUI (red line) also suffer a decline in activity. When experienc-
ing unrest, low growth countries face a larger contraction on impact than no-low
growth countries, but the medium-term effects are similar for both groups.
Both the analysis of events and the RSUI confirm that our results are robust
to controlling for past growth dynamics. Indeed, we find that social unrest has an
adverse impact on growth regardless of whether the country was experiencing low
growth prior to unrest or not.
Next, we focus on the concern that our baseline results could be driven by a
potential driver of both unrest and activity: fiscal consolidations. While there is a
longstanding debate regarding the impact of consolidations on activity, there are a
number of studies suggesting that fiscal consolidations have an adverse effect on
economic activity (although the effect would depend on the composition of the con-
solidation as discussed in Alesina, Favaro, and Giavazzi 2019). At the same time,
episodes of fiscal consolidation are often linked with protests and socio-political
instability (Ponticelli and Voth 2020).
To study whether our results are picking-up the link between fiscal austerity
and economic activity and unrest, respectively, we estimate an econometric specifi-
cation that distinguishes between the effect of social unrest in countries that expe-
rience episodes of fiscal austerity and those that do not. In particular, we estimate
the following specification:
 -.2 .1 -.1 0
Change relative to GDP in t-1 (in percent)
-.3
-.4
-1 0 1 2 3 4 5 6 7 8 9 10 11 12
Quarters since unrest

CI Low growth Low growth

No low growth CI No low growth
CI No low growth baseline
The Macroeconomic Impact of Social Unrest
—

Figure 18: Impact of social unrest on GDP after controlling for prior adverse growth events (response to a one st. dev. change in the
RSUI). Shaded area and dashed lines indicate 90 percent confidence interval with standard errors clustered at the country and year
953

level.
954 — M. Hadzi-Vaskov et al.

[
yi,t+h − yi,t−1 = 𝛼ih + 𝛾th + consi,t 𝛿cons
h
+ 𝛽cons
h
RSUIi,t + 𝜃cons Xi,t

]
∑
h
j ,h
+ 𝜇cons unresti,t+ j
j=1
[
+ (1 − consi,t ) 𝛿nocons
h
+ 𝛽nocons
h
RSUIi,t + 𝜃nocons Xi,t

]
∑
h
j ,h
+ 𝜇nocons unresti,t+ j + 𝜀i,t+h (8)
j=1

where consi,t is a dummy variable taking value one if the country’s primary fiscal
deficit over GDP increased by more than 2 percentage points from one year to the
other, and unresti,t captures either RSUI or the RSUI-implied dummy.
The results in Figure 19 show that social unrest has an adverse effect on eco-
nomic activity, regardless of whether a country is immersed in a fiscal consolidation
effort, though the effect appears somewhat stronger under consolidation (the black
and red lines represent the effects of unrest, in the presence or absence of fiscal
consolidation respectively). Hence, the adverse impact of social unrest on economic
activity estimated in countries that are not experiencing episodes of fiscal auster-
ity suggests that social unrest affects GDP beyond the potential role played by fiscal
consolidations.
 Response to a one st. dev. change in the RSUI Response to a New RSUI-Implied Event

.2 0
0

-.2
-2

Change relative to GDP in t-1 (in percent)

Change relative to GDP in t-1 (in percent)

-.4
-4

-.6
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
Quarters since unrest Quarters since unrest

CI fiscal consolid. fiscal consolid. CI fiscal consolid. Fiscal Consolid.

no fiscal consolid. CI no fiscal consolid. No fiscal Consolid. CI No fiscal Consolid.
CI no fiscal consolid. baseline CI No fiscal Consolid. baseline
The Macroeconomic Impact of Social Unrest
—

Figure 19: Impact of social unrest on GDP after controlling for fiscal consolidations. Shaded area and dashed lines indicate 90 percent
confidence interval with standard errors clustered at the country and year level.
955
956 — M. Hadzi-Vaskov et al.

6 Conclusions
Using a novel, timely and transparent indicator for social unrest based on media
word count metrics, this study investigates the macroeconomic effects of social
unrest. The analysis provides several contributions to the existing literature—it
includes a broader pool of social unrest events and investigates the impact on a
broad set of higher-frequency economic indicators.
The analysis draws several key conclusions. First, economic activity declines
following a one standard deviation increase in the social unrest index, with GDP
remaining on average 0.2 percent below the pre-unrest level after 6 quarters. Sec-
ond, the adverse effect on GDP is driven by sharp contractions of services and
manufacturing (from sectoral perspective) and consumption (from demand side).
Third, social unrest is found to dampen (consumer) confidence and raise uncer-
tainty. While unrest is found to adversely affect country groups regardless of their
income level, the effect upon emerging markets seems to be (about two times) larger
than in advanced economies. Fourth, strong institutions and available policy space
(fiscal space and exchange rate flexibility) can dampen the adverse effects of unrest.
Fifth, the effects are larger when considering major events of social unrest—which
are increases of at least 4 standard deviations in the RSUI—as, on average, GDP
remains 1 percent below baseline after 6 quarters of such a social unrest event.
Hence, the effect of such major effects is more than proportional than implied by
the results for RSUI above, and consistent with the notion that major events have
relatively larger impact. For comparison, the impact of social unrest events is larger
than the impact on oil exporters’ GDP of an oil price plunge and is comparable to the
impact of an oil price plunge combined with a recession (World Bank 2020). It is also
comparable to the impact of a fiscal consolidation of around 1 percent of GDP (Carri-
ere-Swallow, David, and Leigh 2021). The impact of unrest on GDP is consistent and
quantitatively similar to the findings Barrett et al. (2021), who study the behavior
of stock market returns after episodes of social unrest. Sixth, unrest episodes moti-
vated by socio-economic considerations result in sharper GDP contractions com-
pared to those associated mainly with politics/elections; and episodes triggered by
a combination of both socio-economic and political issues see the sharpest contrac-
tions. Finally, the results are robust to controlling for endogeneity, not only via the
implementation of an instrumental variables strategy that exploits regional waves
of social unrest, but also by controlling for previous adverse growth events (thereby
addressing some of the possible reverse causality concerns) and for episodes of
fiscal consolidation.
The findings in this paper also point at possible areas for future research.
For instance, future work may want to focus on further disentangling the het-
erogenous effect of different types of unrest, such as highly disruptive and violent
 The Macroeconomic Impact of Social Unrest — 957

events versus peaceful demonstrations.18 The findings also suggest that further
research would be valuable to better understand the impact of social unrest on
investment.

References
Acemoglu, D., and J. Robinson. 2019. The Narrow Corridor: States, Societies, and the Fate of Liberty. New
York: Penguin Press.
Acemoglu, D., S. Naidu, P. Restrepo, and J. A. Robinson. 2019. “Democracy Does Cause Growth.”
Journal of Political Economy 127 (1): 47 − 100,.
Ahir, H., N. Bloom, and D. Furceri. 2022. The World Uncertainty Index. Cambridge: NBER WP 29763.
Aisen, A., and F. J. Veiga. 2013. “How Does Political Instability Affect Economic Growth?” European
Journal of Political Economy 29: 151 − 67,.
Alesina, A., C. Favero, and F. Giavazzi. 2019. “Effects of Austerity: Expenditure- and Tax-Based
Approaches.” The Journal of Economic Perspectives 33 (2): 141 − 62,.
Alesina, A., and R. Perotti. 1996. “Income Distribution, Political Instability, and Investment.” European
Economic Review 40: 1203 − 28,.
Auerbach, A. J., and Y. Gorodnichenko. 2012. “Measuring the Output Responses to Fiscal Policy.”
American Economic Journal: Economic Policy 4 (2): 1 − 27,.
Barrett, P., M. Appendino, K. Nguyen, and J. de Leon Miranda. 2022. “Measuring Social Unrest Using
Media Reports.” Journal of Development Economics 158,.
Barrett, P., M. Bondar, S. Chen, M. Chivakul, and D. Igan. 2021. “Pricing Protest: The Response of
Financial Markets to Social Unrest.” IMF Working Paper 2021/079. Washington.
Barrett, P., and S. Chen. 2021. “Social Repercussions of Pandemics.” IMF Working Paper 2021/021.
Washington.
Bernal-Verdugo, L., D. Furceri, and D. Guillaume. 2013. “The Dynamic Effect of Social and Political
Instability on Output: The Role of Reforms.” IMF Working Papers 2013/091. Washington.
Bloom, N. 2009. “The Impact of Uncertainty Shocks.” Econometrica 77 (3): 623 − 85.
Carriere-Swallow, Y., A. C. David, and D. Leigh. 2021. “Macroeconomic Effects of Fiscal Consolidation in
Emerging Economies: New Narrative Evidence from Latin America and the Caribbean.” Journal of
Money, Credit, and Banking 53 (6): 1313 − 35,.
Cerra, V., and S. C. Saxena. 2008. “Growth Dynamics: The Myth of Economic Recovery.” The American
Economic Review 98 (1): 439 − 57,.
Cochrane, J. H., and M. Piazzesi. 2002. “The Fed and Interest Rates—A High-Frequency Identification.”
The American Economic Review 92 (2): 90 − 5,.
Cook, T., and T. Hahn. 1989. “The Effect of Changes in the Federal Funds Rate Target on Market
Interest Rates in the 1970s.” Journal of Monetary Economics 24: 331 − 51,.
Furceri, D., P. Loungani, J. D. Ostry, and P. Pizzuto. 2020. “Will Covid-19 Affect Inequality? Evidence
from Past Pandemics.” Covid Economics 12: 138 − 57.
Gruss, B., and S. Kebhaj. 2019. “Commodity Terms of Trade: A New Database.” IMF Working Paper
19/21.

18 See the study on country risk by Hassan, Schwedeler, Schreger and Tahoun (2021).
958 — M. Hadzi-Vaskov et al.

Hadzi-Vaskov, M., S. Pienknagura, and L. A. Ricci. 2021. “The Macroeconomic Impact of Social Unrest.”
IMF Working Paper 21/135. Washington.
Hassan, T. A., J. Schreger, M. Schwedeler, and A. Tahoun. 2021. “Sources and Transmission of Country
Risk”. NBER Working Paper No. 29525.
Hlatshwayo, S., and C. Redl. 2021. “Forecasting Social Unrest: A Machine Learning Approach.” IMF
Working Paper 21/263. Washington.
Ilzetzki, E., C. M. Reinhart, and K. S. Rogoff. 2019. “Exchange Arrangements Entering the 21st Century:
Which Anchor Will Hold?” Quarterly Journal of Economics 134 (2): 599 − 646,.
IMF. 2019. “The Economic Consequences of Conflicts.” In Chapter 2 in Sub-Saharan Africa Regional
Economic Outlook: Recovery amid Elevated Uncertainty. Washington: International Monetary Fund.
IMF. 2020. World Economic Outlook: A Long and Difficult Ascent. Washington: International Monetary
Fund.
Institute for Economics and Peace. 2020. Global Peace Index 2020: Measuring Peace in a Complex World.
Sydney: Institute for Economics and Peace.
Jong-A-Pin, R. 2009. “On the Measurement of Political Instability and its Impact on Economic Growth.”
European Journal of Political Economy 25: 15 − 29,.
Jordà, O. 2005. “Estimation and Inference of Impulse Responses by Local Projections.” The American
Economic Review 95: 161 − 82,.
Kaufmann, D., A. Kraay, and M. Mastruzzi. 2010. “The Worldwide Governance Indicators: Methodology
and Analytical Issues.” In Policy Research Working Paper WPS 5430. Washington: World Bank.
Kuttner, K. N. 2001. “Monetary Policy Surprises and Interest Rates: Evidence from the Fed Funds
Futures Market.” Journal of Monetary Economics 47: 523 − 44,.
Novta, N., and E. Pugacheva. 2020. “The Macroeconomic Costs of Conflict.” IMF Working Paper
20/110. Washington: International Monetary Fund.
Ponticelli, J., and H.-J. Voth. 2020. “Austerity and Anarchy: Budget Cuts and Social Unrest in Europe,
1919 2008.” Journal of Comparative Economics 48 (1): 1 − 19,.
Ramey, V. A., and S. Zubairy. 2018. “Government Spending Multipliers in Good Times and in Bad:
Evidence from US Historical Data.” Journal of Political Economy 126 (2), 850 − 901.
Rodrik, D. 1999. “Where Did All the Growth Go? External Shocks, Social Conflict, and Growth
Collapses.” Journal of Economic Growth 4: 385 − 412,.
Rother, B., G. Pierre, D. Lombardo, R. Herrala, P. Toffano, E. Roos, G. Auclair, and K. Manasseh. 2016.
“The Economic Impact of Conflicts and the Refugee Crisis in the Middle East and North Africa.”
IMF Staff Discussion Note 16/08. Washington: International Monetary Fund.
Saadi Sedik, T., and R. Xu. 2020. “A Vicious Cycle: How Pandemics Lead to Economic Despair and Social
Unrest.” IMF Working Paper 20/216. Washington.
Teulings, C. N., and N. Zubanov. 2014. “Is Economic Recovery a Myth? Robust Estimation of Impulse
Responses.” Journal of Applied Econometrics 29: 497 − 514,.
World Bank. 2020. Global Economics Prospects. Washington: World Bank.

Supplementary Material: This article contains supplementary material (https://doi.org/10.1515/

bejm-2022-0039).