Content uploaded by Thomas Sattler
Author content
All content in this area was uploaded by Thomas Sattler on Jun 15, 2023
Content may be subject to copyright.
Austerity, Economic Vulnerability, and Populism∗
Leonardo Baccini
McGill University
leonardo.baccini@mcgill.ca
Thomas Sattler
University of Geneva
thomas.sattler@unige.ch
June 15, 2023
Abstract
Governments have repeatedly adjusted fiscal policy in recent decades. We examine the
political effects of these adjustments in Europe since the 1990s using both district-level
election outcomes and individual-level voting data. We expect austerity to increase
populist votes, but only among economically vulnerable voters, who are hit the hard-
est by austerity. We identify economically vulnerable regions as those with a high share
of low-skilled workers, workers in manufacturing and in jobs with a high routine-task
intensity. The analysis of district-level elections demonstrates that austerity increases
support for populist parties in economically vulnerable regions, but has little effect in
less vulnerable regions. The individual-level analysis confirms these findings. Our re-
sults suggest that the success of populist parties hinges on the government’s failure to
protect the losers of structural economic change. The economic origins of populism are
thus not purely external; the populist backlash is triggered by internal factors, notably
public policies.
Keywords: fiscal policy, globalization, automation, political backlash, elections, West-
ern Europe.
∗Previous versions of this paper were presented at the Annual Conference of the Inter-
national Political Economy Society, the Annual Meetings of the American Political Science
Association and the European Political Science Association, the ETH Public Policy lunch
seminar, the Global Research in Political Economy webinar, the Political Science Guest
Speaker Workshop at Stanford University, and seminars at the European University Insti-
tute, Georgetown University, the University of Konstanz, the University of Pittsburgh, the
University of Oxford, Washington University in Saint Louis, and the WZB Berlin Social Sci-
ence Center. We thank the participants at these seminars, Francesco Amodio, Eva Anduiza,
Abel Brodeur, Giorgio Chiovelli, Federico Ferrara, Ken Scheve, Eric Voeten, and Steve Wey-
mouth for comments on this paper and Jane Gingrich for extensive advice on data sources.
We thank Giacomo Lemoli for advice on Stata coding. We thank Akos Mate, Sean Nossek,
and Colin Walder for outstanding research assistance. Leo Baccini acknowledges financial
support from the Canadian Social Sciences and Humanities Research Council (grant no.
430-2018-1145). Thomas Sattler acknowledges financial support from the Swiss National
Science Foundation (grant no. 165480).
1
1 Introduction
Governments have regularly implemented fiscal adjustment measures in recent decades. Ad-
justment policies have strong distributional consequences, especially in contemporary, indus-
trialized democracies. These countries have experienced major economic transformations,
such as globalization or automation, that increase economic insecurity among voters. Pub-
lic safety nets are crucial in this context because they provide insurance against enhanced
economic risk and hence stabilize societies both socially and politically. By contrast, gov-
ernment decisions to cut fiscal spending magnify rather than mitigate the adverse effects of
the ongoing economic transformations. Therefore, government decisions are crucial during
periods of structural economic change.
Economic explanations of populism in recent years have paid surprisingly little attention
to governments and their policy choices. Prior studies have significantly improved our un-
derstanding of political backlash by highlighting how three economic outcomes in particular
– trade shocks, financial crises, and technological innovations – affect voters. This literature,
however, has largely overlooked the role of governments. Yet governments have traditionally
been at the center of analyses of globalization politics (e.g., Mosley, 2003) and should play
an important role in how we think about the economic origins of populism and political
backlash. We therefore examine how government policy, particularly fiscal austerity, affects
voters’ political behavior during periods of enhanced economic risk.
Our analysis concentrates on the impact of fiscal austerity on economically vulnerable
voters. Although fiscal cutbacks are generally national-level decisions that apply to the entire
country, exposure to them varies significantly across regions and societal groups. Cutbacks
primarily affect economically vulnerable voters, who rely on government support to cope
with increased economic risk. By contrast, voters who have sufficient resources to ride out
economic downturns are barely affected by public spending cuts. Austerity policies therefore
1
cause disenchantment primarily among voters who face social decline because they are hit
the hardest by fiscal cutbacks and infer from austerity that their well-being is not a priority
for their government. As a result, vulnerable voters are increasingly swayed by populist
pledges to rectify their economic situation either by reversing spending cuts or by curtailing
globalization as the original source of economic risk.
To identify economically vulnerable voters, we draw on the international political econ-
omy literature, which has investigated the winners and losers from economic transformations
for decades (e.g., Milner, 1988; Frieden, 2000). This research highlights the extent to which
factor endowment (Scheve and Slaughter, 2001), sectoral competitiveness (Jensen, Quinn
and Weymouth, 2017), and occupational characteristics (Gingrich, 2019; Owen, 2020) affect
income and job security. Following these different theoretical logics, low-skilled workers,
workers in manufacturing, and those in routine jobs are particularly vulnerable and suffer
the most from austerity. We therefore expect these workers to be more likely to support
populist parties when the government adjusts fiscal policy.
Empirically, we examine how austerity has affected voting patterns in Western countries
since the early 1990s using both district-level election outcomes and individual-level voting
data. The results of our two-way fixed effects (TWFE) analysis illustrate that austerity
increases support for populist parties in economically vulnerable regions, but has little effect
on voting in less vulnerable regions. Moreover, we find that radical right (but not radical
left) parties gain votes in economically vulnerable regions where austerity measures have
been implemented. Our individual-level analyses confirm these results. Overall, our findings
indicate that fiscal cutbacks and the resulting lack of insurance against economic shocks
contribute significantly to the rise of populist parties and the backlash against globalization.
There are different possible mechanisms that lead to this result, e.g. direct material effects
vs. government responsiveness concerns, and our paper does not adjudicate between these
2
different mechanisms. It is plausible that these different aspects of austerity work together
in ways that expand the appeal of populists.
We implement several additional tests to strengthen our identification strategy. First, we
include lead variables of austerity, which capture anticipatory effects, and NUTS-2 specific
trends, which leave our main results unchanged. Second, we show that our results hold if we
include our measures of economic vulnerability in interaction with election-year fixed effects.
Third, exploiting the fact that European countries implement austerity measures even in
good times, we show that our results are not driven by the occurrence of economic crises.
Put differently, even when macro-economic conditions are normal, economically vulnerable
areas and individuals support populist parties where austerity measures have been imple-
mented. While this additional evidence lends credibility to our findings, we acknowledge
that the assumptions supporting our identification strategy are more demanding than they
would be in a case study with a single episode of austerity, which varies sub-nationally. We
trade-off stronger identification assumptions for a stronger external validity.
Our study makes three main contributions. First, it advances the emerging literature on
the backlash against globalization by moving public policy and governments to the center
of the analysis. There is now a large body of evidence that economically vulnerable vot-
ers increasingly turn towards populist parties (Owen and Johnston, 2017; Jensen, Quinn
and Weymouth, 2017; Gidron and Hall, 2017; Colantone and Stanig, 2018; Gingrich, 2019;
Ballard-Rosa et al., 2021; Milner, 2021; Baccini and Weymouth, 2021). In line with prior
single-country studies (Fetzer, 2019; Wiedemann, 2022), our results show that the success
of populist parties across Europe critically hinges on governments’ failure to protect and
help the losers of structural economic change. The economic origins of populism are thus
not purely external and unavoidable; the populist backlash is triggered by internal factors,
notably public policies.
3
Second, we contribute to the literature on the political effects of fiscal policy by isolating
the impact of fiscal cutbacks on different groups of voters. Although the political risks of
austerity were previously pointed out (Blyth, 2013), empirical tests thus far highlighted the
average response of the electorate to fiscal adjustments (Giger and Nelson, 2011; Grittersov´a
et al., 2016; Arias and Stasavage, 2019; Talving, 2017; Alesina, Favero and Giavazzi, 2019;
Bansak, Bechtel and Margalit, 2021) and how mainstream party economic convergence influ-
ences the choices of voters (H¨ubscher, Sattler and Wagner, 2023). To the extent that voter
heterogeneity is examined, material explanations are dismissed in favor of ideological ones
(Barnes and Hicks, 2018; H¨ubscher, Sattler and Wagner, 2021). To the best of our knowl-
edge, our paper is the first to demonstrate that the economic vulnerability of voters strongly
affects the intensity of their response to fiscal austerity, both regionally and individually.
The political disruptions of austerity can therefore be significant even if the median voter or
the majority of voters support an austerity package.
Finally, our analysis sheds new light on government accountability in open economies.
It suggests that economic policy continues to influence popular evaluations even if voters
hold governments less accountable for economic outcomes in open economies (Hellwig and
Samuels, 2007; Kayser and Peress, 2012). While outcomes convey less information about
policymaker competence in such economies, the policy response to these outcomes still signals
the government’s economic priorities to voters. Vulnerable voters infer from fiscal cutbacks
that the government’s policy position is incompatible with their needs and interests, and
hold it accountable accordingly.
4
2 Austerity and the economic origins of populism
2.1 Fiscal adjustments in times of enhanced economic risk
We define fiscal austerity as a government decision to adjust fiscal policy to reduce the public
fiscal deficit, i.e. the difference between public expenditures and revenues. These decisions
generally center on reducing government spending, e.g. by cutting social security entitle-
ments or public investments, but they can also entail tax increases, such as VAT or income
taxes, to increase public revenues.1A prominent example is the wave of fiscal adjustments
in the wake of the European debt crisis (Copelovitch, Frieden and Walter, 2016). These
recent cutbacks, however, are not unique and represent the peak of a longer-lasting move-
ment towards “permanent austerity” that has been noted for a long time (Pierson, 2001,
ch. 13). As Figure 1 illustrates, most industrialized countries had implemented significant
cutbacks long before the start of the global financial crisis in 2007.2The figure also shows
that adjustments have been quite common throughout Europe, including Germany, Austria,
and the Scandinavian countries.
Fiscal adjustments can affect a wide range of budgetary areas. We focus on the consistent
and strong impact of austerity on policies that are important to vulnerable voters, such as
social transfers and other welfare state policies. As Appendix A at page 2 shows, government
transfers are central to austerity packages: on average, they are cut more than any other bud-
getary category. This is also visible in Figure 2, which illustrates that average social security
transfers in industrialized countries vary considerably over time and declined particularly
1We concentrate on policy choices rather than actual changes in public expenditures and
revenues. The former can be directly attributed to the government, while the latter can also
vary for other reasons, such as macro-economic shocks, which are beyond the government’s
direct control.
2We discuss the measurement of austerity used in this figure in detail in Section 3.
5
Figure 1: Spending Cuts in Industrialized Countries, 1978-2014
Note: Total spending cuts announced by governments, as % of GDP.
Figure shows the cumulative amount, i.e. the sum of all announced cuts,
from 1978 until 2007 (black line) and until 2014 (black and grey lines).
Source: Devries et al. (2011); Alesina, Favero and Giavazzi (2019).
strongly during the 1990s and again from 2013 onwards. These declines in transfers coincide
with the large waves of austerity that governments have implemented in recent decades (see
also Armingeon, Guthmann and Weisstanner, 2016). We find a similar pattern for public
spending on unemployment benefits, education and pensions in Figure C1 in Appendix C at
page 7.3Austerity, therefore, has been associated with cutbacks in public safety nets and
other public schemes that are important for the welfare of economically vulnerable citizens
in open economies.4
3In this figure, we also present the same indicators adjusted for the percentage of entitled
citizens, to rule out that this pattern is simply catching business cycle effects.
4While total social expenditures have gradually increased over time, this is not the case for
the spending items that are central to our argument. The increase in total social expenditures
is mostly due to pensions and health care. It is politically very difficult to make cuts in these
areas related to so-called life-cycle risks (Pierson, 2001). This increases the pressure to
concentrate cuts in other social spending categories that are important for our argument.
6
Figure 2: Austerity and Social Security Transfers Over Time
0
.5
1
1.5
Austerity (% of GDP)
12
13
14
15
16
Social Security Transfers (% of GDP)
1975
1980
1985
1990
1995
2000
2005
2010
2015
2020
Social Security
Austerity
Note: Bars show the average amount of deficit-reducing measures (as % of
GDP) in the countries listed in Figure 1. Black line shows average social
security transfers (social assistance grants and welfare benefits) by general
government (as % of GDP) in the same countries. Sources: Devries et al.
(2011); Alesina, Favero and Giavazzi (2019); Armingeon et al. (2019).
Given how fiscal adjustments affect core fiscal programs, these decisions are generally
intensively debated as part of budgetary debates and in the public discourse. Opposition
parties that expect to benefit politically by exploiting discontent with fiscal austerity tend
to instigate such debates. This creates a public discourse that informs voters of the gov-
ernment’s fiscal plans and how they will affect voters. Political saliency also increases with
the size of the fiscal adjustment because the impact on voters is more serious and therefore
more contentious. This opposition to large austerity packages often manifests itself in the
form of protests, which raise public awareness of the issue (Bremer, Hutter and Kriesi, 2020).
Our qualitative analysis of newspaper reports during major austerity episodes in Europe in
the 1990s and 2000s, described in Appendix B at pages 3-6, confirms this. These findings
demonstrate that austerity measures were often at the center of the political debate, played
an important role in the national media and in electoral campaigns, and that voters were
7
aware of these issues, such as in France in 1997 or 2011/2012, in Italy in 1994/1995, and in
Germany in 2004.
Fiscal adjustments are also politically salient because they take place in a context of
increased social risk. Industrialized economies have experienced major transformations in
recent decades, such as a massive increase in trade, offshoring, and the automation of jobs
(Autor, Dorn and Hanson, 2013). In this context, public safety nets are important to stabilize
countries socially and politically because they help voters cope with economic risk in open
economies (Gingrich and Ansell, 2012; Walter, 2010; Kurer and Gallego, 2019). For instance,
these policies increase support for openness (Hays, 2009; Rickard, 2015) and decrease the risk
of political backlash (Rudra, 2005; Margalit, 2011; Halikiopoulou and Vlandas, 2016; Richtie
and You, 2020; Vlandas and Halikiopoulou, 2022). Yet, austerity has distributional effects
that operate in the same direction as these economic transformations. It magnifies rather
than mitigates the negative economic effects of globalization and technological change and
exacerbates the social decline of vulnerable individuals and communities (Sambanis, Schultz
and Nikolova, 2018).
2.2 Vulnerable voters and their political reactions
We draw on theories of comparative and international political economy to identify the eco-
nomically vulnerable voters that are most affected by austerity. In short, economic vulnera-
bility varies according to 1) skill level, 2) economic sector, and 3) routine job intensity. First,
from a factor logic, low-skilled workers tend to be worse off in open, industrialized economies,
while high-skilled laborers tend to benefit in such countries (Scheve and Slaughter, 2001).
Second, from a sectoral logic, the manufacturing sector faces the greatest competition from
firms in developing and emerging markets, while high-skilled service industries thrive in open
economies (Jensen, Quinn and Weymouth, 2017). Also, small firms in a sector find it harder
to succeed in open economies, while large, productive firms are best positioned to exploit the
8
gains from trade (Baccini, Pinto and Weymouth, 2017). Finally, from an occupational logic,
workers in routine jobs are most likely to lose their jobs due to offshoring or automation
(Gingrich, 2019; Owen, 2020; Gallego, Kurer and Sch¨oll, 2020).
Austerity creates political disenchantment among these vulnerable voters because they
are the most exposed to its material impacts (Wiedemann, 2022). They rely more heavily
on social safety nets and public transfers than more privileged voters, and are directly af-
fected by austerity measures. This particularly applies to large fiscal adjustments that limit
governments’ ability to spare vulnerable voters or compensate them with parallel measures.
These measures then raise doubts among vulnerable voters that governments are committed
to make globalization a success for everybody, including the economically vulnerable. They
infer that government parties are more responsive to economically more privileged voters
who are less affected or potentially benefit from austerity than to vulnerable voters when
fiscal trade-offs sharpen the divide between them (Bartels, 2008).5
Although we focus on voter characteristics (i.e. the demand side), political parties (i.e.
the supply side) play an important role in our argument in two ways. First, as we discuss
in greater detail in Appendix B at pages 3-6, populist parties have increasingly positioned
themselves against fiscal austerity. In sum, leftist populist parties have opposed austerity on
average, although there was a tendency towards acceptance in the early 2000s. Right-leaning
populist parties were more accepting of austerity in the past, but have become increasingly
critical over the past two decades. This characterization is consistent with our qualitative
analysis of major austerity periods in Europe, and is in line with recent findings that radical
right parties oppose austerity measures proposed by government parties (Enggist and Ping-
5This implies that vulnerable voters respond in similar ways in different types of welfare
states because they react to how austerity measures affect their well-being. Our empirical
analysis examines whether this is indeed the case.
9
gera, 2022). This, in turn, is consistent with new findings that populist parties quickly adapt
to shifting voter opinions and emphasize new issues that allow them to challenge established
parties (De Vries and Hobolt, 2020).6Populist parties, therefore, allow vulnerable voters to
express anti-government sentiment.
Second, our argument requires that many vulnerable voters see a lack of alternatives
among traditional, non-populist parties (H¨ubscher, Sattler and Wagner, 2023). This is
plausible because a broad consensus among traditional (non-populist) political parties has
supported the recent austerity waves (Blyth, 2013; Hopkin, 2020).7Again, our qualitative
analysis of traditional parties’ positions on major austerity episodes in European countries
confirms this (see Appendix B at pages 3-6). These results demonstrate that traditional
political parties regularly supported austerity measures. If they opposed them, they did so
while in opposition, but supported them as government parties in earlier or later austerity
episodes. This gives voters fewer opportunities to sanction governments, for example by vot-
ing for the non-populist opposition, especially when the pro-austerity consensus cuts across
political camps. But despite the importance of supply-side politics, our point is that vul-
nerable voters are more likely than their better-off counterparts to support populist parties
when they lack non-populist, anti-austerity alternatives.
More generally, voters’ responses can vary across contexts, countries and time periods,
6Dissatisfied voters can also abstain to express their discontent, and may vote for populist
parties at a later point in time. We empirically examine this possibility to the extent that
our data allows.
7There are multiple possible reasons for this tendency, including financial pressure (Haller-
berg and Wolff, 2008; Sattler, 2013; Barta and Johnston, 2018), international integration
(Mosley, 2003; Konstantinidis, Matakos and Hutlu-Eren, 2019), the diffusion of pro-austerity
ideas (Blyth, 2013) and institutional constraints (Bodea and Higashijima, 2017).
10
depending on the existence of populist parties, non-populist parties’ past involvement in
implementing austerity measures, the type of austerity, the nature of welfare states and eco-
nomic conditions. We explore these various moderating factors in the empirical analysis, but
our key goals are to identify which types of voters are affected most strongly and to examine
how their reactions differ from those of non-vulnerable voters. We expect that economically
vulnerable individuals, on average, are more likely than economically safe individuals to vote
for populist parties after an austerity package is implemented. Similarly, populist vote share,
on average, should increase more in economically vulnerable than in economically prosperous
electoral districts after an austerity package is introduced.
3 District-level elections
The first part of our analysis examines district-level election results in 12 Western European
countries and (up to) 195 NUTS-2 regions. Our time span covers (up to) 99 elections between
1986 and 2018. We focus on elections for the lower house of the legislature. Each country
appears only in years in which elections are held. The data on party vote shares on the
district level is from the Constituency-Level Elections Archive (CLEA) database (Kollman
et al., 2019).
3.1 Data
Measuring populism. Our main outcome variable measures support for populist parties in
an electoral district in a given election. To compute this variable, we first match the CLEA
data with the Global Party Survey’s (Norris, 2019) classifications of political parties on an
11-point populism scale.8This allows us to calculate a populism score for each district-
8In this dataset, parties are classified according to a range of dimensions based on expert
surveys. The conceptualization and operationalization of populism relies on Norris and In-
glehart (2019), which treats populist rhetoric as antithetical to pluralist rhetoric. Populist
11
election. This score is the weighted average of the populism scores of all parties in the
district-election where parties are weighted by their vote shares. This variable theoretically
ranges from 0 (pluralist parties receive all the votes) to 10 (populist parties receive all the
votes). This measure varies across electoral districts and over time.9We label this variable
Populism Score.
We also examine support for radical parties, many of which take a populist position.
Using data from PopuList (Rooduijn et al., 2019), we evaluate the share of votes for radical
left and right parties, which allows us to explore how austerity influences ideology.
Figures C2 and C3 in Appendix C at page 8 display the distribution of our outcome vari-
able across NUTS-2 regions and over time. The figures illustrate that half of the countries
in the sample had already experienced a surge of votes for populist parties in the 1990s and
not only during the past decade.
language “typically challenges the legitimacy of established political institutions and em-
phasizes that the will of the people should prevail,” while pluralist language “rejects these
ideas, believing that elected leaders should govern constrained by minority rights, bargaining
and compromise, as well as checks and balances on executive power” (Global Party Survey
codebook, p. 10). Populist rhetoric is measured from 0 (less populist) to 10 (more populist).
9The Global Party Survey’s classification of parties is fixed since it is difficult to judge the
degree of populist rhetoric in the more distant past using current expert surveys. Nonetheless,
populism scores vary over time and across districts when party vote shares in a district
change. Our measure thus captures the demand-side effects that arise when voters switch to
a different political party, and rules out supply-side effects that are created when mainstream
parties become more populist. This generates more conservative estimates. We also examine
the vote share of strongly populist parties, which arguably includes those that have been
populist for the entire period. We also rely on different (but related) time-varying measures,
such as Colantone and Stanig (2018)’s nationalism score, and find similar effects.
12
Measuring austerity. We measure austerity as the amount of deficit-reducing policies
implemented by the government in a given time period (Devries et al., 2011; Alesina, Favero
and Giavazzi, 2019). This indicator is based on government policy documents (e.g. bud-
getary reports) and reports from international organizations, such as International Monetary
Fund country reports, to identify the timing and magnitude of a fiscal adjustment package.
It captures policy decisions to reduce public spending or increase taxes as announced by the
government and recorded in these documents, and indicates by how many percentage points
these policies are expected to reduce the deficit (as % of GDP). The indicator captures the
year when the policy change takes effect and has the advantage of directly capturing the
government’s policy decision.10
We use the cumulative amount of austerity implemented between the two elections, i.e.
the previous election in our dataset and the election for which we examine votes.11 Note
10The data distinguishes between the year of announcement and the year of implementa-
tion. In case of multi-year adjustment plans, a large part of the plan is usually implemented
in the year in which it is announced. Some of the announced policy changes, however, only
take effect in later years. Where this is the case, we use the year in which the policy is effec-
tively implemented. Since we measure austerity across electoral periods rather than years,
the announcements and implementation mostly coincide in our dataset.
11Originally, the austerity variable was an annual time series for each country: it captures
the amount of deficit-reducing measures that a government implements in a particular year.
We sum these annual values for each electoral period, which gives us the total amount of
austerity (as % of GDP) implemented during an electoral period. This is a straightforward
way to attribute the annual consolidations to an election period in years without elections.
It is trickier for election years where we had to make some judgment calls. We manually at-
tributed fiscal consolidations in election years to one of the two election periods as accurately
as possible.
13
that this measure is continuous; 0s indicate governments that do not implement austerity
measures, i.e. our control group. The fact that austerity measures are continuous implies
that treated units receive the treatment with different degrees of intensity: some austerity
policies are mild, whereas others are quite severe. We label this variable Austerity.
In additional analyses, we also use a disaggregated version of this indicator that distin-
guishes between the number of deficit-reducing measures that are due to spending cuts vs.
tax increases. Moreover, we build a variable that captures the share of austerity measures
that entail spending cuts over total austerity measures. Similarly, we build a dummy that
takes a value of 1 if austerity measures involving spending cuts are larger than those con-
cerning tax increases. The data comes from Alesina, Favero and Giavazzi (2019). These
last two variables allow us to explore whether cutting welfare expenditure triggers a stronger
demand for populism than increasing taxes.
Figure C4 in Appendix C at page 9 displays the temporal evolution of our austerity
variable by country. There is evidence that the intensity of this measure varies quite dra-
matically among countries and over time. Overall, it illustrates that European governments
have implemented austerity measures very frequently over the past three decades.
Measuring economic vulnerability. To measure economic vulnerability, we follow
the international political literature on the distributional effects of globalization and au-
tomation. We use the share of unskilled workers and the share of workers in manufacturing.
Low-skilled workers have been negatively affected by both competition with cheap labor from
emerging markets and technological shocks, whereas the manufacturing sector has been par-
ticularly hard hit by trade liberalization over the past 30 years. The data comes from
Colantone and Stanig (2018) and varies by NUTS-2 regions.12 We map each district to its
12The share of low-skilled workers is the share of employees with a lower secondary ed-
14
NUTS-2 region to merge the outcome variable with variables capturing economic vulnera-
bility.
We use the share of workers exposed to automation as a further proxy for economic vul-
nerability. To build this variable, we rely on the EU Labour Force Survey, which is a large
household sample survey providing quarterly results on 1) the labor participation of people
aged 15 and over and 2) people outside the labor force. For each labor participant, we have
information on her occupation, which we match to the routine task intensity (RTI) score
developed by Goos et al. (2014). Following Autor (2013), we take the top 33% most routine-
task-intense occupations and count their employment as routine jobs. We then calculate the
share of workers in the most RTI occupations in each NUTS-2 region.
Note that we use economic vulnerability variables at their baseline value at the beginning
of our period of analysis; this value does not change over time.13 We label these variables
Share of Low-Skilled Workers,Share of Manufacturing Workers, and Share of Workers Ex-
posed to Automation. Figure C5 in Appendix C at page 9 displays the geographic distribution
of the share of low-skilled workers across NUTS-2 regions.14
ucation and below in a region, according to Eurostat Regional Statistics. The share of
manufacturing workers is the share of employees in a region who work in the manufacturing
sector, according to Eurostat and national statistical offices. The manufacturing sector is
identified using NACE two-character alphabetical codes (DA to DN).
13We use the baseline value to avoid that a time-varying vulnerability measure picks up the
effect of austerity. Due to fast-moving economic transformations, such as deindustrialization
and automation, vulnerability at the baseline value may deviate from vulnerability in the
later periods that we examine. This is more relevant for the share of manufacturing and RTI
workers than for share of low skilled workers. To overcome this challenge, we use several
measures of vulnerability, which capture different parts of the society.
14The figures for the other two variables are available upon request.
15
3.2 Empirical strategy
Our analysis at the district level is a standard TWFE with a continuous treatment. We
estimate the following baseline model:
ycd,t =α+Xcr (d),baseline ×Austerity′
c,tβ+γct +δr+ϵcd,t ,(1)
where ycd,t is our outcome variable capturing the vote share of populist parties in each district
in each election year. Xcd(r),baseline is a matrix including our measures of economic vulner-
ability at baseline. The function r(d) maps district dto its NUTS-2 region r.Austerityc,t
is a continuous variable scoring strictly positive values in election years in which austerity
measures are implemented. The key coefficient of interest is β, which estimates the inter-
action term between the two main independent variables. It reflects how the impact of
national-level austerity measures varies across districts with different degrees of economic
vulnerability.
We are unable to estimate the coefficient of Xcr(d),baseline alone, because it is absorbed by
(NUTS-2) region fixed effects, i.e. δr. Similarly, we are unable to estimate the coefficient of
Austerityc,t alone, because it is absorbed by country-year fixed effects, i.e. δc,t. These fixed
effects net out time-invariant differences across districts as well as time-variant differences
across countries. The term ϵcd,t captures any unaccounted-for variation.
In augmented model specifications, we enrich our baseline model with potential con-
founders. We include a China shock variable as in Colantone and Stanig (2018). Moreover,
we include foreign direct investment (FDI) inflow, FDI outflow, and export growth to ac-
count for regional economic conditions. We also include the share of foreign-born people
as a proxy for migration. This set of controls is at the baseline, i.e. they vary only at the
NUTS-2 level. Thus, we interact each of these controls with our austerity variable to esti-
16
mate their effects. We run ordinary least squares (OLS) regressions with robust standard
errors clustered at the country-election year level.
3.3 Identification
Identifying the effect of austerity presents at least three challenges. First, in an effort to rule
out anticipatory effects, we include lead variables that fake austerity measures before they
are implemented. If we find that these are significant, this would be a clear indication of
the presence of anticipatory effects because it would indicate that areas with large shares of
manufacturing and low-skilled workers support populism regardless of the presence of aus-
terity measures, which we do not observe. We also include NUTS-2 specific trends and find
no evidence of pre-trend effects, except for Share of Workers Exposed to Automation. Fur-
thermore, we include our economic vulnerability variables in interaction with election-year
to fixed effects to test whether economic vulnerability alone drives our results. We find no
evidence that this is the case. In addition, we show that the results are similar if we include
constituency fixed effects. Appendix D at pages 14-17 reports the results of all of these tests.
In a second challenge, austerity is a potential outcome of negative economic conditions.
Thus, economic crises may trigger support for populism among vulnerable voters. To ad-
dress this point, we leverage the fact that austerity measures do not perfectly correlate with
negative economic conditions in Western European countries. In other words, while auster-
ity correlates negatively with economic growth and fiscal balance, our data indicates that
such measures have also been implemented during periods of economic stability and growth.
Thus, to determine whether periods of economic crisis are driving our estimates, we run our
main models on two sub-samples: 1) observations experiencing sluggish economic growth and
negative fiscal balance and 2) observations experiencing average or fast economic growth and
average or positive fiscal balance.15
15We use the value of the lower quartile to split the sample.
17
Third, it seems likely that governments strategically implement austerity measures. In
particular, they probably anticipate the negative electoral consequences of austerity mea-
sures and time their implementation to mitigate voters’ responses. For instance, there is a
clear tendency for governments to implement austerity policies early and avoid them later
in the electoral cycle, especially if their legislative majority is at risk (H¨ubscher and Sattler,
2017). We note that governments’ strategic behavior leads us to underestimate the effect of
austerity on vulnerable voters.
Our goal is to test the general relationship between austerity and populism in a broad
range of countries and periods. Cleaner identification strategies are possible for specific,
well-selected austerity episodes in specific, well-selected countries (see, for instance, Fetzer
2019). For one, our continuous treatment (i.e. austerity) switches on and off and it does not
vary sub-nationally, though its interaction with economic vulnerability does. The question
arises to what extent these results are unique to a particular case, or whether they apply
to a broader range of countries and time periods. This leads to a well-documented trade-
off between internal and external validity. For a broad, comparative analysis over time,
identification is more difficult and requires stronger assumptions. Yet in exchange, we are
able to explore the extent to which austerity contributes to populism in general, or only in
particular and unique circumstances.
3.4 Results
Populism. Table 1 reports the results of our main analysis. The coefficient of the interaction
between variables capturing economic vulnerability and Austerity is positive and significant,
as expected, in both the baseline models (Models 1–3) and in models that include controls
(Models 4–6). The share of low-skilled workers, share of manufacturing workers, and share
of workers exposed to automation give similar results; their coefficients remain positive and
18
significant, even when we include both at the same time on the right-hand side of the models
(Model 7).16
Table 1: Austerity and Populism: Main Results
(1) (2) (3) (4) (5) (6) (7)
Share of Low Skilled Workers*Austerity 1.041** 1.037** 0.604†
(0.356) (0.302) (0.345)
Share of Manufacturing Workers*Austerity 1.248** 1.459** 1.107**
(0.334) (0.456) (0.341)
Share of Workers Exposed to Automation*Austerity 1.581** 1.525*
(0.513) (0.620)
Constant 4.352** 4.439** 4.198** 4.360** 4.502** 4.205** 4.290**
(0.105) (0.058) (0.154) (0.121) (0.059) (0.199) (0.110)
Observations 14,110 14,158 14,435 11,607 11,607 11,583 14,110
R-squared 0.867 0.868 0.870 0.847 0.849 0.848 0.868
Controls No No No Yes Yes Yes Yes
NUTS2 fixed effects Yes Yes Yes Yes Yes Yes Yes
Country-election year fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
OLS
Populism Score
Robust standard errors in parentheses † p<0.1, * p<0.05, ** p<0.01
Note: OLS regressions with robust standard errors clustered by county-election year in
parentheses. The unit of observation is NUTS-2-election year. The outcome variable is
populism score. The key independent variable is the share of low-skilled workers, the share
of manufacturing workers or the share of workers exposed to automation interacted with
austerity measures. District-level controls are interacted with austerity.
To ease the interpretation of the interaction term, we plot the estimates of the share of
low-skilled workers. Figure 3 displays the linear predictions of Populism Score for different
low-skilled workers in the case of austerity measures at their minimum and in the case of an
average value of austerity measures. These figures reveal three key points. First, support for
populism is always higher with austerity than without, but the difference is small in areas
that are not economically vulnerable. Second, support for populism does not increase in
regions with high shares of low-skilled workers with minimum levels of austerity, i.e. the
linear prediction is a flat line. Third, Populism Score increases dramatically in regions with
16We are unable to include the share of workers exposed to automation with the other two
measures of economic vulnerability due to their very high collinearity, i.e. ρ > 0.9.
19
high shares of low-skilled workers with (average) austerity measures.17
The magnitude of these effects is substantial. In countries implementing (average) aus-
terity measures, Populism Score increases by 4.6%, moving Share of Low-Skilled Workers
from one standard deviation below the mean to one standard deviation above it. Conversely,
in countries implementing minimum levels of austerity measures, Populism Score increases
by a mere 0.2%, moving Share of Low-Skilled Workers from one standard deviation below
the mean to one standard deviation above it.18
The role of crises. To assess whether crises are a potential confounder, we re-run our
main models, splitting the sample according to macro-economic conditions, i.e. low and high
fiscal balance (Table 2). As expected, the effects are larger when fiscal balance is negative
than when it is positive. However, our main findings remain unchanged even if there are
no economic crises, i.e. when macro-economic conditions are sound. Our results are similar
even when we interact fiscal balance with the share of low-skilled workers (Model 3). In
short, our results are not a by-product of the correlation between austerity and economic
crisis: austerity independently sways voters toward populism.
17In Appendix D at pages 12 and 13, we show the linear predictions of regions with
high economic vulnerability (one standard deviation above the mean) and low economic
vulnerability (one standard deviation below the mean) with minumum or average value of
austerity. Results indicate that support for populism increases in both sets of regions, but
the increase is significantly larger in regions with high economic vulnerability.
18The result is similar for Share of Manufacturing Workers and Share of Workers Exposed
to Automation (see Appendix D at page 10). These effects are in line with the effects
estimated by Colantone and Stanig (2018), who leverage a single event, i.e. the China trade
shock, over a relatively short period of time, whereas the occurrence of austerity is more
frequent in our sample and our time span is longer.
20
Figure 3: Austerity and Populism: Share of Low-Skilled Workers
4 4.5 5 5.5
Linear Prediction of Populism Score
0 .8.1 .2 .3 .4 .5 .6 .7
Share of Low Skilled Workers
Austerity (min) Austerity (mean)
0 5 10 15
Frequency of Share of Low Skilled Workers (%)
Note: Linear predictions refer to Model 2 in Table 1. The dashed line
reports the linear predictions of minimum level of austerity for different
shares of low-skilled workers in NUTS-2 regions. The continuous line
reports the linear predictions of average level of austerity for different
shares of low-skilled workers in NUTS-2 region. 90% confidence interval.
21
In Appendix E at page 18-20, we show that our results are similar if we use (i) other
proxies of economic vulnerability and (ii) economic growth rather than fiscal balance.
Table 2: Austerity and Populism: The Role of Fiscal Balance
(1) (2) (3)
Low Fiscal Balance
High Fiscal Balance
Full
Share of Low Skilled Workers*Austerity
1.115†
1.040†
0.950*
(0.564)
(0.525)
(0.371)
Share of Low Skilled Workers*Fiscal Balance
-0.057
(0.068)
Constant
4.551**
4.277**
4.367**
(0.240)
(0.133)
(0.106)
Observations
3,497
10,602
14,110
R-squared
0.789
0.884
0.867
Controls
No
No
No
NUTS2 fixed effects
Yes
Yes
Yes
Country-election year fixed effects
Yes
Yes
Yes
OLS
Populism Score
Robust standard errors in parentheses † p<0.1, * p<0.05, ** p<0.01
Note: OLS regressions with robust standard errors clustered by county-election year in
parentheses. The unit of observation is NUTS-2-election year. The outcome variable is
the share of votes for populist parties. The key independent variable is the share of low-
skilled workers interacted with austerity measures. District-level controls are interacted with
austerity.
Types of austerity. In Table 3, we distinguish between different types of austerity. We
show that our results are similar if we rely on spending cuts rather consolidation (Models
1 and 2). Moreover, Models 3 and 4 indicate that measures of austerity involving spending
cuts trigger more support for populism than measures of austerity concerning tax increases
(Model 2). Similarly, when spending cuts are larger than tax increases, support for populism
increases more than vice versa (Model 3).19 Moreover, Models 5 and 6 show that the effect
19Government transfers, which are relevant for job security, and government consumption,
which disproportionately hits lower-income voters, drive these results.
22
is similar for countries with low and high levels of welfare spending.20 This indicates that
changes in welfare support matter more than levels: current welfare state support is the
reference point of a vulnerable voter against which she evaluates governments, and when the
government cuts this baseline level of welfare, the voter reacts.21
In addition, we run our main model specification, replacing Austerity with two dummies
to indicate high and low austerity measures. We use the average value of (strictly positive)
Austerity to create these two dummies. Models 7 and 8 report the results, which show that
severe austerity measures are driving our results. This finding has three implications. First,
it seems to indicate that voters are more likely to be aware of and observe the economic
effects of large consolidation measures, and therefore to electorally sanction the parties that
implement them. Second, to the extent that large consolidation measures are typically im-
plemented by cross-party agreement, there is evidence that the lack of alternatives to the
dominant narrative in favor of austerity fuels populism because dissatisfied voters have no
other means of expressing their discontent. Yet, we note that this is only the case for vul-
nerable voters, as our theory suggests. Third, severe austerity measures make it particularly
difficult for governments to spare, with targeted policies, economically vulnerable workers,
which may explain why they turn their vote to populist parties. In Appendix E at page 21,
we show that our results are similar when we use other proxies for economic vulnerability.
Austerity and ideology. Table 4 reports the results of support for radical left and
20We use data on baseline social expenditure to distinguish between countries with low
(i.e. below average) and high (above average) levels of welfare spending.
21Theoretically, the role of welfare state size is ambiguous. On the one hand, voters in
large welfare states may react more strongly to austerity because they have higher expecta-
tions on state support. On the other hand, voters in small welfare states may react more
strongly because their situation becomes even more precarious after an equivalent reduction
in support.
23
Table 3: Austerity and Populism: Types of Austerity
(1) (2) (3) (4) (5) (6) (7) (8)
Full Full Full Full
Low Welfare
High Welfare
Full Full
Share of Low Skilled Workers*Austerity (cuts) 1.089* 1.470**
(0.545) (0.412)
Share of Low Skilled Workers*Share of Spending cuts 1.779*
(0.690)
Share of Low Skilled Workers*Predominantly Spending cuts 1.011*
(0.493)
Share of Low Skilled Workers*Austerity (consolidation) 0.923† 1.118*
(0.509) (0.492)
Share of Low Skilled Workers*Austerity (low) -0.373 -0.482
(0.546) (0.345)
Share of Low Skilled Workers*Austerity (high) 1.655** 1.492**
(0.559) (0.468)
Constant 4.431** 4.362** 4.368** 4.451** 5.206** 3.824** 4.459** 4.523**
(0.114) (0.119) (0.065) (0.041) (0.120) (0.162) (0.134) (0.102)
Observations 14,110 11,607 12,439 12,439 5,334 8,776 14,110 11,607
R-squared 0.867 0.847 0.871 0.871 0.856 0.809 0.868 0.848
Controls No Yes Yes Yes No No No Yes
NUTS-2 fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Country-election year fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
OLS
Populism Score
Robust standard errors in parentheses † p<0.1, * p<0.05, ** p<0.01
Note: OLS regressions with robust standard errors clustered by county-election year in
parentheses. The unit of observation is NUTS-2-election year. The outcome variable is the
share of votes for populist parties. The key independent variable is the share of low-skilled
workers interacted with types of austerity measures. District-level controls are interacted
with austerity.
24
right parties. There are two take-away findings. First, only radical right parties gain from
austerity; we find no effect for radical left parties (Models 1 and 2).22 Second, whether left
or right governments implement austerity measures does not generally affect explanations
of support for radical right/left parties (Models 3–6).23 This finding is confirmed when we
use the share of manufacturing workers to proxy for economic vulnerability. For the share of
workers exposed to automation, we find some evidence that radical right (left) parties gain
electoral consensus when mainstream left (right) parties implement austerity (see Appendix
F at page 25).
Furthermore, we illustrate how austerity affects a set of socio-economic policies (see Table
F1 in Appendix F at page 23). We replace our main outcome, Populism Score, with outcomes
that capture support for parties that are against international trade, the EU, migration, and
minorities, and that support conservative values. Our results indicate that where austerity
measures are implemented, economically vulnerable areas experience a surge of support
for parties advocating autarky or conservative values, as well as Eurosceptic parties and
anti-migration parties. Radical right parties typically hold such positions, which implies
that austerity causes economically vulnerable voters to move rightwards on socio-economic
issues.
Robustness checks. We perform four additional robustness checks and report the re-
sults in Appendix G at pages 27-34. First, our results are similar if we use other measures
of support for populism. Second, our results are similar if we use the raw value of auster-
ity rather than its logged value. Third, we find no evidence that austerity affects turnout.
22These results are robust to the use of other proxies for economic vulnerability and dif-
ferent measures of support for radical left and right parties (see Appendix F at pages 24 and
26).
23Left/right incumbency measures the ideology of the cabinet before the election using the
average left–right position of all parties in government. The data comes from the Compara-
tive Manifestos Project.
25
Table 4: Austerity, Economic Vulnerability, and Radical Parties
(1) (2) (3) (4) (5) (6)
Full Left Incumbent Right Incumbent Full Left Incumbent Right Incumbent
Share of Low Skilled Workers*Austerity -0.006 -0.084** 0.013 0.152** 0.183* 0.136*
(0.071) (0.025) (0.054) (0.049) (0.071) (0.055)
Constant 0.042* 0.070** 0.029 0.017 0.025 0.012
(0.021) (0.006) (0.018) (0.014) (0.016) (0.019)
Observations 14,111 5,624 8,478 14,111 5,624 8,478
R-squared 0.640 0.739 0.612 0.833 0.918 0.767
Controls No No No No No No
NUTS-2 fixed effects Yes Yes Yes Yes Yes Yes
Country-election year fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
OLS
Share of Votes for Radical Left Parties Share of Votes for Radical Right Parties
Robust standard errors in parentheses † p<0.1, * p<0.05, ** p<0.01
Note: OLS regressions with robust standard errors clustered by county-election year in
parentheses. The unit of observation is NUTS-2-election year. The outcome variables are
the share of votes for radical left and right parties. The key independent variable is the
share of low-skilled workers interacted with austerity measures. District-level controls are
interacted with austerity.
Fourth, we show that our results are not driven by the post-2010 period. Finally, we show
that our results hold if we exclude one country at a time; thus they do not depend on the
inclusion of any specific country in our sample.
4 Individual-level voting
Our individual-level analysis includes 12 Western European countries and (up to) 46,918
respondents for whom we have data. Our time span covers (up to) eight waves of the
European Social Survey (ESS) administered between 2002 and 2016. Below, we describe the
data and the empirical strategy and report our main results. In line with the district-level
analysis, we exploit variation in individual exposure to national austerity policies depending
on our measures of economic vulnerability. We expect economic vulnerability to moderate
the extent to which national austerity measures affect individuals’ tendency to vote populist.
26
4.1 Data
Our main outcome variable measures ESS respondents’ support for populism. We use each
party’s populism score as described in the previous section and match it to the party for
which the respondent voted in the most recent election before the ESS survey.24
To measure austerity, we rely on the variable described in the previous section. We cap-
ture the austerity packages that were implemented during the electoral period that is leading
up to the election recorded in a particular ESS wave.25 In our main model specification, we
use a dummy that takes a value of 1 if any austerity measure was introduced during the
electoral period before the election that we analyze.26 To capture economic vulnerability, we
use the number of years of education of each respondent, which identifies low-skilled workers
in line with the district-level analysis. Years of education is homogenous across countries
that have different education systems. We split this variable into three dummies: Lower
Secondary (less than 10 years of education), 2) Upper Secondary (10–15 years of education),
and 3) Tertiary (more than 15 years of education). Tertiary is the baseline category in the
analysis; i.e. it is the excluded variable. Furthermore, we use a dummy coded as 1 if respon-
dents work in manufacturing. This variable is built on the NACE trade category reported
in the ESS.
24For instance, ESS wave 6 from 2012 captures the vote of Irish respondents in the 2011
national election.
25In the example of the 2011 Irish election recorded in ESS wave 6, the austerity variable
reflects the fiscal consolidation that the Irish government implemented between the preceding
election in 2007 and the 2011 election.
26In additional analyses (available upon request), we show that our results are virtually
the same for education if we use a continuous measure of austerity. They are weaker for
manufacturing and RTI, though the sign of the main coefficient remains the same.
27
We also include a variable measuring exposure to automation at the individual level. Fol-
lowing Goos, Manning and Salomons (2014), we convert varying occupational measures into
a 2-digit ISCO-88 code and link it to an aggregated RTI index.27 Then, following Gingrich
(2019), we aggregate the RTI measure into five quintiles, rescaled to 0 (least affected) to 1
(most affected), which allows us to identify broad categories of exposure.28
4.2 Empirical strategy
In line with the district-level analysis, our analysis at the individual level is a standard
TWFE. We estimate the following baseline model:
yic,w =α+Xic,wζ′+Xic,w ×Austerity′
c(i),wη+γcw +ϵic,w ,(2)
where yic,w is our outcome variable, which captures respondent’s isupport for populism in
wave w.Xic,w is a matrix that includes our measures of economic vulnerability: education,
manufacturing, and exposure to automation.29 Austerityc,w is a dummy scored as 1 if coun-
try cimplements austerity measures in ESS wave w. The function c(i) maps respondent i
to its country c. In this analysis, the key coefficient of interest is η, which estimates the
interaction term between the two main independent variables. We are unable to estimate
the coefficient of Austerityc,w alone, because it is absorbed by country-year fixed effects, i.e.
27The RTI index categorizes occupations based on the skills most affected by automation
in the 1980s and 1990s. This measure is missing for three major occupational groups (ISCO
23, 33, and 61), which are excluded from the analysis.
28The results are similar if we use the continuous version of the RTI developed by Goos,
Manning and Salomons (2014).
29We are unable to use the baseline values of our measures of economic vulnerability, since
the ESS is a repeated cross-section rather than a panel: different respondents take part in
each wave.
28
γc,t. The term ϵic,w captures the residuals.
In the augmented model specifications, we enrich our baseline model with a host of
individual-level characteristics including gender and age, which absorb an important vari-
ation of our outcome. We also add dummies for retired respondent, student, unemployed
respondent, self-employed respondent, and respondent working in services. We interact each
of these controls with Austerity to estimate their effects. We run OLS regressions with robust
standard errors clustered at the country-wave level.30
4.3 Results
We report the results in Table 5. In all models, the interaction terms between economically
vulnerable individuals and austerity measures are positive and statistically significant, con-
firming the results of the district-level analysis. Importantly, the coefficient of the interaction
term remains the same when we include individual-level controls interacted with austerity.
To ease the interpretation of the results, we plot the estimates of the interaction term
between education and austerity in Figure 4. The figure displays the marginal effect of Lower
Secondary Education on Populism Score with and without austerity. Less-educated individ-
uals are significantly more likely than highly educated individuals to support populist parties
if austerity measures have been implemented. Without austerity measures, less-educated in-
dividuals are not more likely to support populist parties than highly educated individuals.31
The effect is sizable: Populism Score is six times higher with austerity than without. We also
find that individuals with an upper secondary education are more likely to support populist
parties where austerity policies have been introduced, though the magnitude of the effect is
significantly smaller than that found for individuals with a lower secondary education.32
30All estimates include post-stratification weights, including design weights.
31Recall that we obtain these marginal effects controlling for student status, unemployed
status, and age. Thus, years in school does not proxy for (youth) unemployment.
32The result is similar for manufacturing workers and for workers exposed to automation
29
Table 5: Austerity and Populism: Individual-level Analysis
(1) (2) (3) (4) (5) (6)
Lower Secondary Education
0.054
0.047
(0.055)
(0.059)
Upper Secondary Education
0.096*
0.090*
(0.037)
(0.037)
Manufacturing
-0.011
-0.016
(0.021)
(0.021)
RTI
-0.026
0.011
(0.035)
(0.039)
Lower Secondary Education*Austerity (dummy)
0.221**
0.254**
(0.083)
(0.089)
Upper Secondary Education*Austerity (dummy)
0.164**
0.173**
(0.062)
(0.062)
Manufacturing*Austerity (dummy)
0.093*
0.085*
(0.038)
(0.039)
RTI*Austerity (dummy)
0.092†
0.111*
(0.050)
(0.055)
Constant
4.621**
4.471**
4.843**
4.670**
4.779**
4.521**
(0.040)
(0.162)
(0.005)
(0.147)
(0.017)
(0.168)
Observations
86,939
86,717
82,317
82,128
72,613
72,449
R-squared
0.301
0.303
0.299
0.301
0.299
0.301
Controls
No
Yes
No
Yes
No
Yes
Country-wave fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
OLS
Populism Score
Robust standard errors in parentheses † p<0.1, * p<0.05, ** p<0.01
Note: OLS regressions with robust standard errors clustered at the country-wave level in
parentheses. The unit of observation is individual-survey wave. The outcome variable is
populism score. The key independent variables are economic vulnerability variables inter-
acted with austerity measures. Individual-level controls are interacted with austerity.
30
Figure 4: Austerity and Populism: Lower Secondary Education
-.1 0 .1 .2 .3 .4
Marginal Effect of Lower Secondary Education on Populism Score
No Austerity Austerity
0 20 40 60
Frequency of Austerity (%)
Note: Marginal effects refer to Model 2 in Table 5. The marginal effects of
Lower Secondary Education are reported for individuals living in countries
with no austerity and with austerity. The histograms show the distribution
of observations without and with austerity. 90% confidence interval.
31
Finally, we perform a large number of robustness checks in line with the district-level
analysis (see Appendix H at pages 35-41). All these tests leave our results unchanged. All
in all, the individual-level analysis confirms our district-level findings: austerity increases
support for populism more among the losers than among the winners from globalization and
automation.
5 Conclusion
This paper examines the political effects of fiscal austerity in open economies. It shows that
economically vulnerable voters – i.e. low-skilled workers, workers in the manufacturing in-
dustry, and workers in routine jobs – increasingly turn to populist parties when governments
implement fiscal cutbacks. We find this effect for both district-level election and individual-
level voting data in Western European countries since the 1990s. Austerity has distributional
effects that magnify, rather than mitigate, the negative economic effects of globalization and
technological change for many workers. These voters therefore begin to question government
promises to make globalization a success for everyone.
These results imply that economic policy and government decisions play a crucial role in
the mechanism that led to the backlash against globalization. Governments have a variety
of ways to moderate the adverse effects of globalization and technological change. But if
they fail to use these means to compensate voters for the increased social risk they face in
open economies, populist parties will be able to exploit the growing anti-globalization senti-
ment among dissatisfied voters. The economic origins of populism, therefore, are not purely
external or unavoidable. Public policies – especially austerity policies – are crucial because
they undermine the ‘embedded liberalism’ compromise of the postwar period that protected
(see Appendix H at pages 35 and 36.)
32
vulnerable workers from the enhanced social risks inherent in open economies (Bisbee et al.,
2020).
These findings also have important implications for government policy after the Covid-19
crisis. Governments have spent large amounts to reduce the pandemic’s economic impact.
A crucial, long-term question is how to deal with the public debts this has generated. Our
results demonstrate that a return to austerity policies after the crisis would be very politi-
cally contentious. While government spending has helped prevent large-scale economic and
political destabilization, the pandemic has had very unequal effects across societal groups
(Bambra, Lynch and Smith, 2021). If vulnerable groups are left to repay the bulk of the cost
of these government interventions, this will likely fuel further support for populist rhetoric
and populist parties.
Our study examines the overall effect of austerity on votes in different political and eco-
nomic contexts. We evaluate how reactions to austerity vary across these contexts, but
our research design cannot provide definitive answers to this question. We leverage a large
amount of regional and individual variation for our key moderating factor, economic vulner-
ability, but there is less variation in our data on national contextual factors. Future research
should further explore how the political–economic context matters, in addition to the voter
characteristics that we study.
33
References
Alesina, Alberto, Carlo Favero and Francesco Giavazzi. 2019. Austerity: When It Works
and When It Does Not. Princeton: Princeton University Press.
Arias, Eric and David Stasavage. 2019. “How Large Are the Political Costs of Fiscal Aus-
terity.” Journal of Politics 81(4):1517–1522.
Armingeon, Klaus, Kai Guthmann and David Weisstanner. 2016. “Choosing the Path of
Austerity: A Neo-Functionalist Explanation of Welfare-policy Choices in Periods of Fiscal
Consolidation.” West European Politics 39(4):628–647.
Armingeon, Klaus, Virginia Wenger, Fiona Wiedemeier, Christian Isler, Laura
Kn¨opfeland David Weisstanner and Sarah Engler. 2019. “Comparative Political Data
Set 1960-2017.” Electronic Database, available at http://www.cpds-data.org.
Autor, David H., David Dorn and Gordon H. Hanson. 2013. “The China Syndrome: Local
Labor Market Effects of Import Competition in the United States.” American Economic
Review 103(6):2121–2168.
Baccini, Leonardo, Pablo M. Pinto and Stephen Weymouth. 2017. “The Distributional
Consequences of Preferential Trade Liberalization: Firm-Level Evidence.” International
Organization 71(2):373–395.
Baccini, Leonardo and Stephen Weymouth. 2021. “Gone for Good: Deindustrialization,
White Voter Backlash, and U.S. Presidential Voting.” American Political Science Review
(Online First).
Ballard-Rosa, Cameron, Mashail Malik, Stephanie Rickard and Ken Scheve. 2021. “The
Economic Origins of Authoritarian Values: Evidence from Local Trade Shocks in the
United Kingdom.” Comparative Political Studies (Online First).
34
Bambra, Clare, Julia Lynch and Katherine E. Smith. 2021. The Unequal Pandemic: Covid-10
and Health Inequalities. Bristol University Press.
Bansak, Kirk, Michael Bechtel and Yotam Margalit. 2021. “Why Austerity? The Mass
Politics of a Contested Policy.” American Political Science Review (Online First).
Barnes, Lucy and Timothy Hicks. 2018. “Making Austerity Popular: The Media and Mass
Attitudes Towards Fiscal Policy.” American Journal of Political Science 62(2):340–354.
Barta, Zs´ofia and Alison Johnston. 2018. “Rating Politics? Partisan Discrimination in Credit
Ratings in Developed Economies.” Comparative Political Studies. 51(5):587–620.
Bartels, Larry M. 2008. Unequal Democracy: The Political Economy of the New Gilded Age.
Princeton University Press.
Bisbee, James, Layna Mosley, Thomas B. Pepinsky and B. Peter Rosendorff. 2020. “Decom-
pensating Domestically: The Political Economy of Anti-Globalism.” Journal of European
Public Policy 27(7):1090–1102.
Blyth, Mark. 2013. Austerity – The History of A Dangerous Idea. Oxford University Press.
Bodea, Cristina and Masaaki Higashijima. 2017. “Central Bank Independence and Fiscal
Policy: Can the Central Bank Restrain Deficit Spending?” British Journal of Political
Science 47(1):47–70.
Bremer, Bj¨orn, Swen Hutter and Hanspeter Kriesi. 2020. “Dynamics of Protest and Electoral
Politics in the Great Recession.” European Journal of Political Research 59(4):842–866.
Colantone, Italo and Piero Stanig. 2018. “The Trade Origins of Economic Nationalism: Im-
port Competition and Voting Behavior in Western Europe.” American Journal of Political
Science 62(4):936–953.
Copelovitch, Mark, Jeffry Frieden and Stefanie Walter. 2016. “The Political Economy of the
Euro Crisis.” Comparative Political Studies 49(7):811–840.
35
De Vries, Catherine and Sara Hobolt. 2020. Political Entrepreneurs: The Rise of Challenger
Parties in Europe. Princeton University Press.
Devries, Pete, Jaime Guajardo, Daniel Leigh and Andrea Pescatori. 2011. “A New Action-
Based Dataset of Fiscal Consolidation.” IMF Working Paper 11/128.
Enggist, Matthias and Michael Pinggera. 2022. “Radical Right Parties and Their Welfare
State Stances: Not so Blurry After All?” West European Politics 45(1):102–128.
Fetzer, Thiemo. 2019. “Did Austerity Cause Brexit?” American Economic Review
109(11):3849–3886.
Frieden, Jeffry A. 2000. Global Capitalism: Its Fall and Rise in the Twentieth Century.
Norton & Company.
Gallego, Aina, Thomas Kurer and Nikolas Sch¨oll. 2020. “Neither Left-Behind nor Superstar:
Ordinary Winners of Digitalization at the Ballot Box.” Journal of Politics (online).
Gidron, Noam and Peter A Hall. 2017. “The Politics of Social Status: Economic and Cultural
Roots of the Populist Right.” British Journal of Sociology 68:S57–S84.
Giger, Nathalie and Moira Nelson. 2011. “The Electoral Consequences of Welfare State
Retrenchment: Blame Avoidance or Credit Claiming in the Era of Permanent Austerity.”
European Journal of Political Research 50(1):1–23.
Gingrich, Jane. 2019. “Did State Responses to Automation Matter for Voters?” Research
and Politics 6(1):1–9.
Gingrich, Jane and Ben Ansell. 2012. “Preferences in Context: Micro Preferences, Macro
Contexts, and the Demand for Social Policy.” Comparative Political Studies 45(12):1624–
1654.
36
Goos, Maarten, Alan Manning and Anna Salomons. 2014. “Explaining Job Polariza-
tion: Routine-Based Technological Change and Offshoring.” American Economic Review
104(8):2509–2526.
Grittersov´a, Jana, Indridi H. Indridason, Christina C. Gregory and Ricardo Crespo. 2016.
“Austerity and Niche Parties: The Electoral Consequences of Fiscal Reforms.” Electoral
Studies 42:276–289.
Halikiopoulou, Daphne and Tim Vlandas. 2016. “Risks, Costs and Labour Markets: Explain-
ing Cross-National Patterns of Far Right Party Success in European Parliament Elections.”
Journal of Common Market Studies 5(3):636–655.
Hallerberg, Mark and Guntram Wolff. 2008. “Fiscal institutions, fiscal policy and sovereign
risk premia in EMU.” Public Choice 136(3):379–396.
Hays, Jude C. 2009. Globalization and the New Politics of Embedded Liberalism. Oxford
University Press.
Hellwig, Timothy and David Samuels. 2007. “Voting in Open Economies – the Electoral
Consequences of Globalization.” Comparative Political Studies 40(3):283–306.
Hopkin, Jonathan. 2020. Anti-System Politics: The Crisis of Market Capitalism in Rich
Democracies. Oxford University Press.
H¨ubscher, Evelyne and Thomas Sattler. 2017. “Fiscal Consolidation under Electoral Risk.”
European Journal of Political Research 56(1):151–168.
H¨ubscher, Evelyne, Thomas Sattler and Markus Wagner. 2021. “Voter Responses to Fiscal
Austerity.” British Journal of Political Science 51(4):1751–1760.
H¨ubscher, Evelyne, Thomas Sattler and Markus Wagner. 2023. “Does Austerity Cause
Polarization?” British Journal of Political Science (Online First).
37
Jensen, J. Bradford, Dennis P. Quinn and Stephen Weymouth. 2017. “Winners and Losers in
International Trade: The Effects on US Presidential Voting.” International Organization
71(3):423–457.
Kayser, Mark A. and Michael Peress. 2012. “Benchmarking across Borders: Electoral
Accountability and the Necessity of Comparison.” American Political Science Review
106(3):661–684.
Kollman, Ken, Allen Hicken, Daniele Caramani, David Backer and David Lublin.
2019. “Constituency-Level Elections Archive.” Ann Arbor, MI: Center for Po-
litical Studies, University of Michigan [producer and distributor]. Retrieved from
http://www.electiondataarchive.org.
Konstantinidis, Nikitas, Konstantinos Matakos and Hande Hutlu-Eren. 2019. ““Take Back
Control”? The effects of Supranational Integration on Party-System Polarization.” Review
of International Organizations 14(2):297–333.
Kurer, Thomas and Aina Gallego. 2019. “The Distributional Consequences of Technological
Change: Worker-Level Evidence.” Research and Politics (online).
Margalit, Yotam. 2011. “Costly Jobs: Trade-Related Layoffs, Government Compensation,
and Voting in U.S. Elections.” American Political Science Review 105(1):166–188.
Milner, Helen V. 1988. Resisting Protectionism: Global Industries and the Politics of Inter-
national Trade. Princeton University Press.
Milner, Helen V. 2021. “Voting for Populism in Europe: Globalization, Technological
Change, and the Extreme Right.” Comparative Political Studies 54(13):2286–2320.
Mosley, Layna. 2003. Global Capital and National Governments. Cambridge University
Press.
38
Norris, Pippa. 2019. “The Global Party Survey (V1.0).” Available at:
www.globalpartysurvey.org.
Norris, Pippa and Ronald Inglehart. 2019. Cultural Backlash: Trump, Brexit and the Rise
of Authoritarian Populism. Cambridge University Press.
Owen, Erica. 2020. “Firms vs. Workers? The Politics of Openness in an Era of Global
Production and Automation.” Paper presented in the Global Research in International
Political Economy seminar, June 10.
Owen, Erica and Noel Johnston. 2017. “Occupation and the Political Economy of Trade:
Job routineness, offshoreability and protectionist sentiment.” International Organization
71(4):665–699.
Pierson, Paul, ed. 2001. The New Politics of the Welfare State. Oxford University Press.
Richtie, Melinda and Hye Young You. 2020. “Trump and Trade: Protectionist Politics and
Redistributive Policy.” Working Paper, UC Riverside and New York University.
Rickard, Stephanie J. 2015. “Compensating the Losers: An Examination of Congressional
Votes on Trade Adjustment Assistance.” International Interactions 41(1):46–60.
Rooduijn, Matthijs, Stijn Van Kessel, Caterina Froio, Andrea Pirro, Sarah De Lange, Daphne
Halikiopoulou, Paul Lewis, Cas Mudde and Paul Taggart. 2019. “The PopuList: An
Overview of Populist, Far Right, Far Left and Eurosceptic Parties in Europe.” Available
at: www.popu-list.org.
Rudra, Nita. 2005. “Globalization and the Strengthening of Democracy in the Developing
World.” American Journal of Political Science 49(4):704–730.
Sambanis, Nicholas, Anna Schultz and Elena Nikolova. 2018. “Austerity as violence: mea-
suring the effects of economic austerity on pro-sociality.” EBRD Working Paper No. 220.
39
Sattler, Thomas. 2013. “Do Markets Punish Left Governments?” Journal of Politics
75(2):343–356.
Scheve, Kenneth and Matthew J. Slaughter. 2001. “What Determines Individual Trade-
Policy Preferences?” Journal of International Economics 54(2):267–292.
Talving, Liisa. 2017. “The Electoral Consequences of Austerity: Economic Policy Voting in
Europe in Times of Crisis.” West European Politics 40(3):560–583.
Vlandas, Tim and Daphne Halikiopoulou. 2022. “Welfare State Policies and Far Right Party
Support: Moderating ‘Insecurity Effects’ Among Different Social Groups.” West European
Politics 45(1):24–49.
Walter, Stefanie. 2010. “Globalization and the Welfare State: Testing the Microfoundations
of the Compensation Hypothesis.” International Studies Quarterly 54(2):403–426.
Wiedemann, Andreas. 2022. “The Electoral Consequences of Household Indebtedness under
Austerity.” American Journal of Political Science. Online First.
40