The Rise of Identity Politics:

Policy, Political Organization, and

Nationalist Dynamics
Timothy Besley Torsten Persson
LSE IIES, Stockholm University
April 2021

Abstract
The most important global political event since the fall of commu-
nism is the rise of a new kind of politics, which has swept across many
democratic countries. It has two facets. One is a group of angry citi-
zens, who identify with the nation and are suspicious of immigration as
well as the political establishment; the other facet is populist, radical-
right politicians who have shaken up existing party systems and push
an anti-immigration agenda. We develop a stylized dynamic model of
multi-dimensional politics to explore how political cleavages, policies,
and social identities evolve over time. The model allows for endoge-
nous political organization in the form of party leadership changes,
formation of social movements, and/or entry of new political parties.
We study the two-way relationship between politics and nationalism
showing that temporary shocks can have permanent consequences.

We are grateful to Guido Tabellini for perceptive comments on an earlier draft and
to Chris Dann, Azhar Hussain and Linchuan Xu for valuable research assistance. We have
also received helpful feedback from participants in the 2018 Munich Lectures and the 2020
ASSA Congress, meetings in London, Münster, Stockholm, Toronto, and Uppsala as well
as various seminar presentations. We gratefully acknowledge …nancial support from the
ERC and the Swedish Research Coumcil.

1
1 Introduction
Throughout modern history, political con‡ict has mostly re‡ected two key
cleavages. The …rst one – familiar from most political-economics models
(like Meltzler and Richards 1981) – is the tension over redistribution. This
cleavage between rich and poor is largely seen to follow socioeconomic funda-
mentals (class). A second political cleavage, which ebbs and ‡ows in terms of
its importance, is the tension between liberal and conservative social values.
This cleavage is largely seen to follow people’s identities, rather than socioe-
conomic fundamentals, and was an important component of interwar polit-
ical con‡ict. Early post-war con‡icts in most liberal democracies, however,
was mainly running in the class/income dimension. But the most important
global political event since the fall of communism is arguably a new wave of
identity politics found in a wide range of democratic countries. This political
wave has two main components.
One is a group of angry citizens, who identify strongly with their nation
and are suspicious of immigration and globalization as well as the political
establishment. Outside of formal politics, protesters have voiced these con-
cerns through new and growing social groups, such as Pegida in Germany,
Vote Leave in the UK, and various Alt Right groups in the US (Caiani and
Parenti 2016 discuss a number of such groups). Along more conventional po-
litical channels, concerned citizens have increasingly voted for new populist
radical-right parties (Norris and Inglehart 2019, Gidron and Hall 2018, Ryd-
gren 2018). This involves a shift away from the traditional rich-poor cleavage
towards an emphasis on a cleavage between social conservatism and social
liberalism, which is often subsumed in the so-called GAL-TAN dimension
(Kitschelt-McGann 1997). Moreover, among some citizens, this cleavage now
appears more salient.
To illustrate some of these changes, we use information from the V-party
data base on the fortunes and policy stances of parties across European
countries. One clear fact is the rising vote for radical-right parties over the
past 20 years. This fact is illustrated by Figure 1, which uses data for a
sample of 23 countries to show a steep rise in voting for radical right parties
compared to mainstream parties after 2010.1 As we might expect, these
votes are more readily translated into political representation in countries
1
The Appendix gives a list of parties by country that we classify as mainstream and
radical right parties in each country.

2
with proportional representation (PR) rather than majoritarian (plurality)
systems. The striking di¤erence is illustrated in Figure 2, with seat shares
in the 2010-2019 period close to 13 percent for PR countries but less than 1
percent for plurality countries.
The other dimension of the radical-right wave concerns political represen-
tatives. A group of new politicians, who are often members of new political
parties, have shaken up existing party systems (Kitschelt 2018). These new
parties and politicians are pushing a populist (anti-establishment) and na-
tionalist/nativist (anti-immigration) policy agenda. And this has encouraged
some mainstream political parties to shift their positions so as to court so-
cially conservative voters (Wagner and Mayer 2017).
Figure 3 illustrates the last of these facts. Comparing the immigra-
tion policies of the same mainstream parties in the …rst and last elections
since the turn of the century, most of them have moved towards a stricter
anti-immigration stance (as measured by the “v2paimmig”variable from V-
party). But such shifts are only clearly visible in those countries where a
radical-right party exists (the gray markers in the …gure), and especially
when the radical right has formed a governing coalition with one or more of
the mainstream parties (the black markers.)
The reasons for these transformational changes in voter behavior and
party structures have been much debated. Commentators in the public arena,
as well as in academia, have suggested a range of drivers. These include eco-
nomic shifts, like globalization and technological change that have deprived
people in certain groups of jobs, incomes, and status. Another alleged force
is a social shift, fuelled by ‡ows of international migrants who seek a bet-
ter life or refuge from violent con‡icts. Yet another is a range of cultural
shifts, including a drift towards nationalist values, and an associated wave of
identity politics. Finally there are political shifts, such as established parties
having converged and focused on issues of little concern to the disadvantaged,
with new parties …lling the resulting void, or mainstream parties – like the
US Republicans or UK Tories –adapting to voter discontent by a reshaped
candidate and policy mix.
The reality almost certainly involves a combination of such forces, creat-
ing a syndrome with a number of associated features. Untangling the causal
chains between them involves notoriously di¢ cult questions. For example,
what should we make of the fact that some alleged forces behind the radical
right (such as globalization) are long standing, while others (such as refugee
migration) have struck at di¤erent intervals in the last few decades? And,

3
when it comes to organizational changes –like new nationalist social groups,
new radical-right political parties, or an altered composition of existing par-
ties –are these best thought of as drivers, rather than conduits (transmission
channels) or equilibrium outcomes?
What makes the recent developments important is the possibility that
they permanently disrupt politics away from income redistribution towards
a more divisive politics which fractures cohesive elements of society. Such
developments in the interwar period contributed to the political schisms that
led to World War II, but began with certain kinds of politicians gaining
an initial in‡uence via the ballot box. Tides of nationalism, their drivers,
and their in‡uence on policy cannot be studied without engaging with the
dynamics of values and political organization.
In our view, one needs a formal, theoretical approach to study these
di¢ cult questions. However, existing models of political competition are ill-
suited to analyzing the observed changes as they tend to operate with …xed
political preferences. Moreover, they typically do not consider the dynam-
ics of organization such as new party structures, or newly organized social
groups. Existing frameworks may therefore miss mutually reinforcing factors
with implications that will remain hidden in want of a new framework.

Roadmap and scope of the paper To address these shortfalls in our tool
box, this paper develops a theoretical approach to the statics and dynamics
of identity politics. The approach builds on several strands of research both
in political economics and in other branches of social science. However, one
distinguishing feature of the approach compared to prior work is that it allows
endogenous values as well as entry of political movements and parties, which
together allow for path-dependent policy dynamics. Section 2 expands on the
relation of the paper to antecedents inside and outside political economics.
Section 3 formulates our baseline model, a static framework where voters
have con‡icting interests in two dimensions: one rooted in incomes of rich
and poor and the other rooted in identities of nationalists and cosmopolitans.
Each dimension is associated with a policy: more or less redistribution in the
rich/poor dimension and more or less strict immigration control in the nation-
alist/cosmopolitan dimension. These policies are set in electoral competition
between two existing parties, which are formed along income lines and both
run by cosmopolitans. The immigration policies are more likely to accom-
modate anti-immigration preferences of nationalist voters when this policy

4
dimension is more salient to them and/or income polarization is greater.
Section 4 builds three model extensions that all endogenize di¤erent as-
pects of political (and social) organization. In the …rst, parties can no longer
commit to an immigration policy so that stricter policy proposals must be
backed up by higher shares of nationalists on the party’s ballot. This en-
dogenous mix of candidates opens the door to party takeover by a nationalist
politician, making it more costly for a cosmopolitan-run party to accommo-
date nationalists. In the second extension, nationalist citizens may form a
new social group outside of politics. But this group still in‡uences policy
indirectly, as it accentuates its members’ sense of collective identity, mak-
ing their immigration-policy preferences more salient and parties more eager
to compete for their vote. Forming such a group results in a discontinuous
jump to a stricter immigration policy. In the third extension, a new political
party may enter to represent nationalist voters, under both plurality rule and
proportional representation (PR). Such entry also leads to a discontinuous
anti-immigration policy shift that occurs at a lower share of nationalist voters
under PR compared to plurality rule.
Section 5 pursues a dynamic extension where the share of nationalists
evolves over time as new generations are socialized into nationalists or cos-
mopolitans. In the baseline model, long-run immigration policies and na-
tionalist shares are correlated, but this re‡ects a unidirectional link from
(expected) policy to nationalism which is governed by exogenous model pa-
rameters. With endogenous candidate mix and possible party takeover, a
minimal concession to nationalists will eventually generate a nationalistic
long-run outcome with strict immigration policy. With endogenous social
groups or political parties, the model may have alternative steady states
where long-run outcomes depend on initial conditions. More signi…cantly,
there can be path dependence (hysteresis), where group formation and/or
party entry makes the share of nationalists and stricter immigration policy
reinforce each other over time.

2 Related Research
In this section, we relate the modeling in the paper to earlier research. We
begin with antecedents in political economics, then go on to predecessors in
the social sciences at large.

5
Political-economics antecedents The workhorse framework for studying
left-right politics was developed on Downs (1957) and the best known version
uses an economic model based on a single income dimension and a desire for
redistribution (Meltzler and Richards 1981, preceded by Romer 1975 and
Roberts 1977).
These models assume that parties are opportunistic and care exclusively
about winning. Alesina (1988), Calvert (1985) and Wittman (1977), put
forward models where parties have policy preferences. The former imposes
credibility on equilibrium policies, while the latter two introduce uncertainty
about voter behavior. Unlike median-voter models, these predict divergence
in policy platforms across parties. A variety of frameworks along these lines
are discussed in Persson and Tabellini (2000) using a probabilistic voting
framework.
The aforementioned research analyzes the outcomes of two-party compe-
tition under plurality rule with a given set of candidates or parties. Endoge-
nous mixes of candidates o¤ered by parties in the context of two-dimensional
political con‡ict are studied by Roemer (2001) and Levy (2004). Entry of
candidates into electoral competition is studied in Besley and Coate (1997)
and Osborne and Slivinski (1996). Party entry and party formation is also
explored in Feddersen et al (1990). Persson and Tabellini (1999, 2000) em-
phasize how electoral systems help shape the way policies are determined.
In addition to these workhorse models of political competition, our paper
is related to the research on identity in economics introduced by Akerlof and
Kranton (2000, 2010). Providing a link to the political economics literature,
Shayo (2009) proposes a formalized social-identi…cation model, which shows
that people who identify with their nation rather than their class demand
less redistribution. Very recent research has set out to explain the rise of pop-
ulism and nationalist policies via the link between social identity and beliefs
(Gennaioli and Tabellini 2019), or the mapping from social identity to pro-
tectionism (Grossman and Helpman 2020). These papers are complementary
to ours but, as Shayo (2009), they rely on static models.
The focus here is on bringing cultural dynamics and organizational dy-
namics into political economics drawing on earlier work cultural evolution
such as Boyd and Richerson (1985) and Cavalli-Sforza and Feldmann (1981).2
Persson and Tabellini (2021) survey the growing literature on the interaction
2
Such ideas have also been incorporated into economics and game theory (for overviews
see, e.g., Bowles 2004, Sandholm 2010, and Weibull 1995).

6
between culture and institutions from a theoretical perspective (see also Bisin
and Verdier 2017). Applications of these ideas include studying the interplay
between democratic institutions and democratic values (Besley and Persson
2019a), and the coevolution of organization design and organizational culture
(Besley and Persson 2020).

Other social-science antecedents The paper is related to the debates

in political science about the dynamics of issue salience. Kitschelt and Mc-
Gann (1997) discuss two political cleavages: one across left vs. right issues,
the other across liberalism vs. authoritarianism, They argue that policy
preferences have shifted towards the second dimension and that one should
study the strategic responses to this shift by incumbent parties. Similarly,
Wagner and Meyer (2017) stress how incumbent parties have accommodated
the greater salience of new issues by moving “to the right”in the liberalism-
authoritarianism dimension. Carmines and Wagner (2006) review the dy-
namics of changes as “issue evolution”.
Many papers on populism study the drivers of new nationalistic par-
ties (Kitschelt 2018). A key idea is that policies of existing parties con-
verged on programs for economic and political modernization, which opened
the door for new parties with anti-globalization, anti-immigration, and anti-
establishment platforms. (Kitschelt 1995, Kitschelt and McGann 1997, Carter
2005, Hobolt and Tilley 2017). More broadly, a rapidly growing literature
deals with right-wing populism and its origins. Rydgren (2018) includes up-
to-date reviews of di¤erent strands in this literature by political scientists
and sociologists.
Our focus on identity politics relates the paper to a large body of soci-
ological research in identity theory (Stryker and Burke 2000) that focuses
on intragroup structures and behavioral outcomes. Research in social psy-
chology on social-identity theory (Tajfel 1974, Tajfel and Turner 1979) and
self-categorization theory (Turner 1985) focuses on intergroup structures and
cognitive outcomes. A striking …nding is how easily experimental subjects
adopt group-speci…c behaviors that bene…t in-group members at the expense
of out-groups.
Finally, a whole sub…eld of political sociology studies mobilizing social
movements (Walder 2009 reviews this research), with early work by Tilly
(1978), Skocpol (1979), and McAdam (1982). Some later work stresses how
social movements foster collective identities, stronger than individual identi-

7
ties (Malucci 1995). Caiani and della Porta (2018) and Gattinara and Pirro
(2018) suggest that one ought to apply the notion of social movements to
nationalist groups.

3 Baseline Model
In our baseline model, two parties propose policies related to two dimensions
of political con‡ict. One is a traditional left-right dimension, where policy
redistributes income and policy preferences depend on income (class). In
the second dimension, policy regulates the openness to immigration and the
con‡icting policy preferences depend on identity (culture). In particular, we
assume that people identify either as “nationalists” or “cosmopolitans” and
that this rubs o¤ on their policy preferences.3

Two-by-two population sub-groups The baseline model is deliberately

simple in order to home in on the main issues. Thus, it abstracts from many
realistic features and makes strong symmetry assumptions. (Appendix C
shows how these assumptions can be relaxed.)
We begin with a static model of a society with population of size one.
Citizens have traits in two dimensions. One is an economic (class) dimension
where citizens di¤er in their …xed earnings ability. We consider two equally-
sized groups labelled J 2 f1; 2g ; where J = 1 are the “poor”and J = 2 the
“rich”. Incomes are denoted by y J ; with y 1 < y 2 .
The second is a social identity dimension, where some citizens are “nation-
alists”labeled N , while others are “cosmopolitans” labeled C. The fraction
of nationalists is and this fraction is identical among the poor and the
rich (by our symmetry assumptions this will be an equilibrium feature of
our dynamic model with endogenous identities in Section 5). Some of the
population are “irreducibly”nationalist and cosmopolitan, with shares and
1 ; respectively, such that 0 < 13 and 23 < 1: Thus is bounded
between and :
We think about these identities as being individually adopted. (Section
4.2 pursues an extension with endogenous formation of nationalist social
3
Alternative identity-related sources of con‡ict over policy could be captured in similar
ways. These include pollution taxes striking a balance between the views of “environ-
mentalists” and “materialists”, and labor-market regulation striking a balance between
“feminists” and “traditionalists.”

8
groups and collective, rather than individual, identities for members of such
groups).

Two policy dimensions Government determines a redistributive policy

and an immigration policy. The former is a tax on income tax at rate t 2
[0; 1]; which is used to …nance welfare-enhancing government spending or
J
lump-sum transfers. Preferences in this dimension are written as U (t) =
U (t tJ ); a group-speci…c loss (indirect-utility) function, where losses grow
in the distance between t and the group’s interior bliss point tJ . Because
y 1 < y 2 ; we have t1 > t2 : Moreover, the loss function is symmetric in upward
and downward bliss-point deviations. One of our extensions (in Section 4.3)
simpli…es further by making the loss function linear: U (t tJ ) = t tJ :
1 2
In either case, the bliss-point di¤erence t t grows with income inequality
2 1
y y that can shift – in an exogenous and unanticipated way – due to
factors such as technology or globalization.
We think of the immigration policy as a (set of) regulation(s) and denote
the policy by x 2 [0; 1]; with a lower (higher) x representing a stricter (laxer)
regulation that allows less (more) immigration. Nationalists have a bliss
point of x = 0; while cosmopolitans most preferred policy is x = 1. These
preferences are captured by convex loss functions, W (1 x) for C and W (x)
for N: Thus, > 1 is the “salience” of (relative weight on) immigration for
nationalists, which could represent mere sentiments and beliefs. Note that
W is the same function, though it takes di¤erent arguments for groups C and
N . We allow for – exogenous and unanticipated – shifts in ; re‡ecting a
latent shift in the perceived threat of migration, due to, say, higher mobility
among signatories of an international treaty or a worsening global refugee
situation.
This simple speci…cation treats the two policy dimensions as separate and
orthogonal to each other. Someone’s attitude to immigration policy does not
depend on her income. Exogenous links are easy to introduce, at the cost
of more algebra as it breaks the symmetry of the model. In the Appendix
(Appendix C), we consider explicitly such an extension, which generates
broadly the same results as our symmetric baseline (see further below).
It would be more demanding to extend the model with endogenous identity-
based policy preferences, via endogenous economic outcomes (as in Grossman
and Helpman 2020) or endogenous stereotyping (as in Gennaioli and Tabellini
2019). Section 5 o¤ers a dynamic extension where identities and policy pref-

9
erences do evolve over time via a process of cultural evolution, but without
a link between the two policy dimensions.

Overall policy preferences Cosmopolitans from income group J have

policy preferences V C;J (t; x) = U (t tJ ) + W (1 x);while nationalists from
group J have preferences V N;J (t; x) = U (t tJ ) + W (x). To simplify
notation, de…ne class polarization as

= U 0) U (t1 t2 ; (1)

which gives the utility gain from having the group’s preferred redistributive
policy compared to the other income group. Symmetry implies that this gain
is the same for the rich and poor. We assume that

> W (0) W (1) ; (2)

where the right-hand side is the maximal loss to cosmopolitans from immi-
gration policy. This condition implies that cosmopolitans always vote based
on their redistributive policy preferences. Thus, the voting behavior of na-
tionalists is a key aspect of the model.4

Parties There are just two given parties, which are organized along eco-
nomic lines with the poor and rich parties labeled J = 1; 2: We want to
capture an “initial situation”, where (i) the main political con‡ict runs in
the class dimension, and (ii) political elites have a more positive attitude
towards immigration than citizens at large. To that end, we assume that
both parties are led by cosmopolitan members of the income group that the
party represents (Section 4.3. pursues an extension with endogenous entry
of a third nationalist party).
Each party proposes a platform ftJ ; xJ g to maximize the expected utility
of its leadership. The objective function of party J is thus E[U (t tJ ) +
W (1 x)]; where taking expectations re‡ects electoral uncertainty. The only
credible redistributive policies are tJ = tJ . For simplicity, we begin by as-
suming that parties can commit to an immigration policy xJ (but Section 4.1
pursues a citizen-candidate extension where a party must recruit nationalist
candidates to credibly o¤er a lower xJ .)
4
This could be relaxed, but the analysis would become taxanomic depending on which
case applies.

10
Voting Our assumption in (2) implies that all poor (rich) cosmopolitans
vote for party 1 (2). However, nationalists are (prospective) swing voters and
support the party that o¤ers the highest utility allowing for the possibility of
random shocks as in standard probabilistic-voting models, such as Lindbeck
and Weibull (1987) or Persson and Tabellini (2000).
Since the treatment is standard, we relegate details to the Appendix (see
Appendix A). There, we show that party 1’s win probability is given by
a continuous function P (x1 ; x2 ; ) ; which is decreasing in x1 and increas-
ing in x2 : Moreover, immigration policy has larger e¤ects on the probabil-
ity of winning the higher is ; the salience of (weight given to) this pol-
icy among nationalists. By model symmetry, Party 2’s win probability is
1 P (x1 ; x2 ; ) = P (x2 ; x1 ; ) :

Party objectives Party J’s objective is

P (xJ ; xK ; ) Z J (xJ ; xK ); J = 1; 2; (3)

where Z J (xJ ; xK ) = + (W (1 xJ ) W (1 xK )) is the (utility) gain from

winning the election to the cosmopolitan leadership. It has two components:
the gain from your own vs. the other party’s tax policy, and the gain from
your own vs. the other party’s immigration policy.
This objective re‡ects the key trade-o¤ in the baseline model. A more
restrictive immigration policy (lower x) increases the win probability, as the
party becomes more attractive to nationalist voters. However, it imposes a
cost on the party’s loyal (cosmopolitan) voters, which is internalized by party
leaders.

Equilibrium policy The Appendix shows that the party objectives are
(log) supermodular, which is convenient both substantively and analytically.
Substantively, supermodularity makes policies strategic complements: when
one party picks a more restrictive immigration policy, the other party wants
to do the same. Analytically, it implies that a Nash equilibrium exists.
This leads to the following characterization of equilibrium policy.5

Proposition 1 Optimal electoral strategies xb ( ; ) are the same for both

parties and given by a decreasing function h ( ) ; and two bounds m
5
The proof of this and all of the subsequent results are found in Appendix B.

11
 and m such that:
8
< 0 m
b( ; ) =
x h( ) 2 (m; m)
:
1 m:

By symmetry, the two parties o¤er the same policy. The proposition gives
three ranges for ; the product of the immigration salience and class polar-
ization. When is high, both parties cater to nationalist swing voters by
a maximally strict immigration policy: x = 0. In this range, the nationalist
vote is very sensitive to concessions and/or high economic polarization makes
it crucial to win the elections. If is low, the opposite is true, and political
elites ignore nationalist preferences by a maximally lax policy: x = 1. The
intermediate case has an interior solution with x 2 (0; 1) ; which trades o¤
the win probability against the utility cost of policy concessions.
Bounds m and m are precisely de…ned in the Appendix. They re‡ect the
density of swing voters around the symmetric equilibrium – i.e., how much
the vote shifts with a stricter immigration policy. They also re‡ect the extent
to which voters value extreme positions.

Comparative statics The upward-sloping reaction functions make for

simple comparative statics. The equilibrium policy is thus fully determined
by , the product of immigration salience and class polarization:

Corollary 1 Parties set a stricter immigration policy – a lower x – when

immigration salience is higher and class polarization greater, mod-
ulo the interaction between the two.

Figure 4, where an increase in or is equivalent to a move along the

horizontal axis, illustrates this corollary together with Proposition 1.
In the baseline model, it is not the size of the nationalist group, but
the underlying salience and polarization that matter for policy. That said,
neither class polarization nor immigration salience by itself drives the immi-
gration policy of existing parties. Polarization alone is insu¢ cient, as parties
have to attract su¢ ciently many nationalist voters to make a strict immigra-
tion policy worthwhile (i.e., a high enough ). Salience alone is insu¢ cient,
as parties have to care su¢ ciently about the redistributive gains from win-
ning to bear the cost of courting the nationalist electorate (i.e., a high enough
).

12
 Another way to say this is that the two dimensions of policy –while being
economically unrelated –are very much politically related.

A glance at the data Figure 5 considers empirically the evolution of

proxies for the two crucial drivers in the model. As in the model, we thus use
changes in inequality to gauge the traditional left-right class cleavage. The
left panel thus uses data from the Luxemburg Income Studies to examine
changes in inequality over time as measured by the 90-10 ratio –i.e., income
at the 90th percentile relative to income at the 10th percentile. Although
there are exceptions in the sample, our proxy for polarization shifts up on
average during the last twenty years.
The right panel instead looks at changes in immigration salience across
our 23 countries. It uses Eurobarometer data, which asks respondents to
name the top-two policy issues facing the country. Here, we average (over
surveys) the proportions that name immigration in the early 2000s and the
late 2010s. While Spain and the UK are exceptions6 , average immigration
salience went up by a factor of two. Indeed, for many countries, like Germany
and Sweden, the hike –which likely re‡ects the European refugee crisis –is
quite sharp, suggesting that immigration salience has indeed increased.

Implications These comparative statics help us get a perspective on the

debate about the more restrictive stances on immigration. People who look
for single causes might argue that more polarization is a prospective expla-
nation. As we have seen, this may be necessary but not su¢ cient. The
complementarity may shed light on a puzzle regarding the timing of the re-
sponse to polarization – in the model. If higher polarization re‡ected more
inequality because of technology shifts, these occurred much earlier than the
shifts towards stricter immigration control. However, our model says that
has to rise as well to generate an immigration-policy (and political) response
to the economic shocks.
The model may also shed light on observed policy heterogeneities across
countries. At least for some time –even in the run up of the 2015 European
refugee crisis –political elites in some countries, like Germany and Sweden,
appear to have resisted the temptation to court nationalists via immigration
6
It seems likely that for the UK, the 2016 Brexit immigration reduced salience of
immigration since the UK was no longer required to o¤er freedom of movement to EU
citizens post-Brexit.

13
clamp-downs. One reason suggested by the model is weaker polarization in
these countries. However, we can think about the refugee shock as an increase
in that did produce a change in German and Swedish policy. Analogously,
low immigration salience may have held back immigration clamp-downs in
countries like the US and the UK, that had seen substantial increases in
inequality and polarization. This may help us understand why Donald Trump
made immigration a central plank of his two US Presidential campaigns, as
did the Leave supporters in the UK Brexit campaign.
As mentioned earlier, the Appendix generalizes the model to allow for
di¤erent fractions of rich and poor, asymmetric salience between the rich and
poor groups and di¤erent fractions of nationalists among the rich and poor.
This extension shows that the core result on shocks to class polarization and
immigration salience continue to hold under slightly stronger assumptions.
So these insights are not a …gment of a symmetric model. The symmetry is,
however, highly useful for the study of organizational change and nationalist
dynamics.
But the nature and timing of the immigration response may also depend
on other factors that di¤er across these and other countries. In particular,
the next section looks at how political organization might respond to rising
nationalism –e.g., via the candidate mixture of existing parties, or via party
entry in di¤erent electoral systems.

4 Endogenous Political Organization

The advantage of the simple symmetric and separable model of the previous
section is that we can extend the model to study three aspects of endogenous
political organization. In all three cases, we discover richer responses to shifts
in class polarization ( ) and immigration salience ( ), as well the spread of
nationalism ( ). For the moment, we stick with a static model (but Section
4 allows to evolve over time).
First, we relax the commitment assumption in immigration policy and
endogenize the composition – in terms of the candidate mix o¤ered to the
voters –of the two existing parties (Subsection 4.1). In this case, the organi-
zational e¤ect of nationalism occurs inside existing parties. Second, we allow
for the endogenous formation of nationalist social groups. Whether this or-
ganizational change takes place depends on the level of nationalism. While
group formation takes place outside of electoral politics, it indirectly a¤ects

14
immigration policy via the platforms adopted by existing parties (Subsection
4.2). Third, we permit endogenous entry of a third political party, along the
identity-based political cleavage. Moreover, we do so under two electoral sys-
tems. This reorganization of the party system directly reshapes immigration
policy via altered representation (Subsection 4.3).

4.1 Composition of Existing Parties

So far, we have assumed that parties can commit unconditionally to immi-
gration policies, but cannot credibly commit to tax (redistributive) policies.
However, a strict immigration policy o¤ered by a party run by cosmopolitans
may have as little credibility as a low-tax policy o¤ered by a party run by
the poor. This subsection makes the model symmetric in this respect. More
importantly, though, it sheds light on the dynamics of party leadership.
The starting point for our baseline model is a situation where cosmopoli-
tans have captured mainstream political parties. The process which has led
to this is beyond the scope of this, paper but it is likely rooted in a process of
political selection that has increasingly professionalized politics among elites
with experience of higher education. But, as we capture in this section, such
elites face an important trade-o¤ when they respond to political shocks. They
can stick to their true values and retain within-party control. Or they can
accommodate rising nationalist sentiments, encouraging entry by nationalists
in order to make their positions credible and risk losing party control. This
trade-o¤ is what we study here. This opens up one dynamic of change is
when a largely cosmopolitan leadership is displaced. Stark examples of such
events are Donald Trump’s selection as the Republican Party candidate in
the 2016 US presidential election, or Boris Johnson’s 2019 election as the UK
Conservative Party’s leader (and therefore as Prime Minister). The desire to
appeal to nationalist voters after a salience shock and/or greater polarization
on income dimensions can spur such processes of leadership transition and
hence further moves towards more nationalistic politics.

Requisites for credible policy Following a citizen-candidate approach,

suppose now that a credible promise of a stricter policy (a lower value of xJ ),
requires the party to …eld a larger share of nationalist representatives –say
on a nationalist list under PR and in a set of single-member constituencies
under plurality rule (the two systems we study in Section 4.3).

15
 To keep things maximally simple, suppose this relation is one-to-one –i.e.,
a more restrictive immigration platform xJ < 1; requires a proportionately
larger share 1 xJ of nationalist candidates. Party leaders now take policy
decisions indirectly, by selecting candidates with di¤erent identities 7 .
In general, a party can be run by a cosmopolitan or nationalist, J 2
fC; N g. However, this subsection begins by focusing on the same baseline as
in Section 2, where both parties are run by cosmopolitans. We thus begin
by asking how the risk of leadership takeover by a nationalist a¤ects the
immigration-policy motives for cosmopolitan-run parties. Leaders have a
personal bene…t from surviving a leadership contest after the policy decision
(speci…c timing below). This modi…es the policy objective in the baseline
model. At the end of the subsection, we consider the behavior of nationalist
party leaders.

The leadership contest We consider a leadership contest in the spirit of

Besley et al (2017). A cosmopolitan-party leader faces a baseline probability
< 1 of losing power to another cosmopolitan, when the party only contains
cosmopolitans (xJ = 1). Parameter can thus be thought about as a generic
measure of leader (un)popularity.
Whenever the legislative representatives contain a set of nationalists, one
of them may take over as leader with a probability proportional to the share of
elected nationalists. A simple model of this is to set that the full probability
of a party leader being ousted at (1 x) + x; where the …rst (second) term
is the probability that a nationalist (another cosmopolitan) takes over the
party. The survival probability of a cosmopolitan party leader is therefore
pC (x) = (1 )x: Let parameter , standing for ego-rent, be the bene…t
of reappointment, which we assume to be the same for the leaders of both
parties.
The earlier party objective (3) is now modi…ed to:

P (xJ ; xK ; ) Z J (xJ ; xK ) + (1 )xJ ; (4)

where the second term re‡ects concerns about retaining the leadership when
pursuing a stricter immigration policy (lower value of xJ ), by …elding a larger
7
Implicitly, this speci…cation corresponds to two assumptions: (i) after the election, the
representatives of the winning party bargain over the implemented policy x, given the pol-
icy preferences held by their identity, (ii) the outcome of this bargaining is monotonically
declining in (1 x):

16
share of nationalist candidates. Since this term is a continuous linear func-
tion, the objective function retains the (log) supermodular property from the
baseline model. A unique Nash equilibrium thus continues to exist.

Timing The timing of the model is now

1. The state of the party leadership characterized by f 1 ; 2g

2. Each party leader chooses a share of cosmopolitan candidates fx1 ; x2 g ;which

translates into electoral platforms for its migration policy.

3. Electoral shocks are realized.

4. The winning party implements its policy platform xJ for J 2 f1; 2g

5. A party leadership contest takes place in each party (In a multiperiod

model, like the one in Section 5, this will determine new values of
f 1 ; 2 g.)

Political equilibrium with cosmopolitan leaders The structure of

equilibrium policy as a function of is essentially as in Proposition 1 and
symmetric across the two parties. Therefore, we focus on the new compar-
ative statics, with respect to and . We denote the equilibrium policy as
b( ; ; ; ); which varies as follows:
x

Proposition 2 With cosmopolitan leaders, both parties pick a common equi-

librium immigration policy, which is (weakly) stricter with lower and
higher .

The threat of losing the leadership reduces the payo¤ to a strict immigra-
tion policy (x < 1): By the upward-sloping reaction functions, this dampens
the both parties’incentives to please the nationalists. The replacement threat
thus widens the range of parameter values for which x b( ; ; ; ) = 1 relative
to the earlier equilibrium xb( ; ): In other words, cosmopolitan leaders are
less likely to make policy concessions.
That said, the model predicts that if an incumbent leader does cater to
nationalists, xb( ; ; ; ) < 1; the party will eventually be taken over with

17
probability 1 x b( ; ; ; ):8 Preparing the ground for the dynamic analysis
to follow in Section 5, we thus have the following comparative static:

Corollary 2 Party takeover by a nationalist is more likely, the higher im-

migration salience and class polarization :

Thus a salience or economic shock can have repercussions beyond policy

responses by changing party control.

Nationalist leaders Consider the policy choices if such a party takeover

has taken place. By an analogous argument to that for a cosmopolitan leader,
a nationalist leader will maximize:

P (xJ ; xK ; ) Z N;J (xJ ; xK ) + (1 )(1 xJ ); (5)

where Z N;J (xJ ; xK ) = + (W (xJ ) W (xK )): This payo¤ is decreasing in

xJ regardless of the other party’s policy –i.e., a nationalist leader perceives
no trade-o¤ between winning and her preferred policy outcome. This means
that she has a dominant strategy: she simply implements her bliss point at
xJ = 0; by …elding a full set of nationalist (and income-level J) candidates.
Such a strategy also fully eliminates the threat of replacement. This (stark)
case will be relevant when we study the dynamic model in Section 5.

4.2 Formation of Nationalist Social Groups

Our baseline model assumes that identifying as a nationalist or cosmopolitan
is an individual act, with no collective element. In this subsection, we allow
nationalists to get together as a group, at a …xed and sunk cost, with a view
to organizing collective action outside formalized party politics. As real-
world examples, we have in mind groups and loose organizations such as:
the Better O¤ Out and Vote Leave groups that were active in the UK Brexit
campaign; the nationalist activist group Pegida that has carried out large
anti-islamic demonstrations in Dresden and other German cities; and the
various nationalist groups, often labeled as Alt Right, that have supported
8
This is because there is always a positive probability of the leader being displaced
even if xJ is very small. We could easily build in a threshold e¤ect, with a positive
probability of displacement only above some value x. This would obviously expand the
range of parameter values where cosmopolitans stay in control.

18
Donald Trump and his policies in the US. These groups have in common
a strong nationalist (nativist) orientation and a rage against the political
establishment not paying attention to the concerns of group members. They
have largely thrived by organizing rallies and protests, with social media
serving as a major coordination mechanism.
As mentioned in the introduction, social movements have been studied
extensively by sociologists. Many have focused on groups on the traditional
left, while fewer studies have been concerned movements on the traditional
right, let alone the radical right. Here, we model the main mechanism sug-
gested in the sociological research on mobilizing social movements, namely
that members collectively identify with the group (Malucci 1995, Flesher
Fuminaya 2014).9 This enhances their sensitivity to immigration policies in-
creasing the salience of this issue for group members and induces political
parties to cater to group members, so that forming a nationalist social group
leads to a discrete jump in immigration policy.

Groups We retain the core elements of the baseline model. However, na-
tionalists may now form a social group at an aggregate cost of F which can be
shared among group members, so the per-capita cost of forming the group is
F= . We focus on the case where F is small but positive, to rule out the triv-
ial case when a group does not form just because collective action is costly.10
This means that a group will not form when it has no political in‡uence.11
In our symmetric framework, either all of the nationalists or none of them
will want to join the group if it does form.
9
We abstract from other channels, such as the traditional role of campaign contribu-
tions to in‡uence election outcomes or policies, as emphasized in the political-economics
literature (Baron 1994, Grossman and Helpman 1996).
10
In practice, the costs of collective action could be substantial, especially in countries
where government puts barriers in the way of civil society. Our results can thus be thought
of as an upper bound on the prospects for group formation. In practice, we also expect the
costs to di¤er substantially across people. If this were the case, we would have a “threshold
model of collective action”(Granovetter 1978) and only a share of nationalists would join
the group.
11
Some of the theories of social movements reviewed in Walder (2009) do not require
a group to be instrumentally powerful. A social movement could be created purely by
a¤ective behavior in response to a sense of grievance whether or not it actually a¤ects
policy.

19
Payo¤s of group members An organized nationalist social group can
create a collective identity due to in-group attachments. We assume that
members of a group internalize the payo¤s of other group members as an
expression of in-group loyalty.12 This may re‡ect formal group activities –
newsletters, advertising, or joint actions – that make members more aware
of being part of a collective. An outright group may also intensify – word-
of-mouth or social-media –communication.
Formally, we write the utility of a nationalist-group member with income
(class) J as
Z
N;J J
V (t; x) = U (t t ) + W (x) + (i)V L (t; x)di; (6)
i

where i runs across all individuals, whether from the same income class
(L = J); or not (L 6= J). To model identi…cation with the nationalistic
in-group, we set the weights (i) to if i 2 N and to 0 if i 2 = N . Parameter
thus re‡ects the in-group loyalty among nationalists.13 It will be the basis
of a social multiplier that can spur the nationalist citizens to get organized.

Spillover to politics Once a group has been formed, the party-J probabil-
ity of winning, once a group has been formed, is given by P (xJ ; xK ; ( ; )) ;
where ( ; ) = (1 + ) captures the enhanced salience of immigration
which is increasing in and for all :14 Intuitively, internalization of in-
group utilities acts as a multiplier on immigration-policy preferences, a mul-
tiplier which grows with the size and loyalty of the nationalist group. The
same kind of multiplier will appear in any model where only a subset of na-
tionalists takes part in the social group, as long as the propensity to join is
increasing in the share of people who (individually) identify as nationalists
:

Timing The timing with possible group formation is as follows:

1. There is a given share of nationalists , salience and economic polar-

ization :
12
This is similar in spirit to the assumption about group-member preferences in Besley
and Persson (2018).
13
Introducing out-group hostility – i.e. (i) < 0 for i 2
= N –.would only serve to
strengthen the mechanism we highlight.
14
See the Appendix for details.

20
 2. Nationalists choose whether to form an interest group, at per-capita
cost F= .

3. Parties o¤er platforms ft1 ; x1 g; ft2 ; x2 g.

4. Individual and aggregate shocks ! and are realized.

5. An election is held where party J wins with probability P (xJ ; xK ; )

or P (xJ ; xK ; ( ; ))

6. Payo¤s are realized.

If the nationalist group does not form, we get the same outcome as in
the baseline model. However, if the group does form, the outcome is similar
to that in Proposition 1, but with e¤ective salience increased to ( ; ) =
(1 + ) > .

Policy and group formation Since ( ; ) > ; it follows from Propo-

sition 1 that formation of an organized nationalist group raises nationalist
in‡uence over immigration policy. Moreover, once the group has formed, a
higher share of nationalists now implies a (weakly) stricter immigration pol-
icy since ( ; ) is increasing in : But there is still a threshold requirement
for and : Speci…cally, if ( ; ) = (1 + ) < m; a nationalist group
will not in‡uence immigration policy. So we still need su¢ ciently high and
in order for nationalism to shape policy. But now the strength of social
ties, , also matters to the strength of nationalist sentiment. If the growth of
social media use have enhanced these ties this is one channel through which
they in‡uence policy.

Equilibrium group formation A nationalist social group will form if the

bene…ts from doing so exceed the cost. Since we focus on the case where costs
are small, i.e. F ! 0, we need only check whether the gain from forming a
group is strictly positive. For this, we must be in the range where an active
nationalist group in‡uences immigration policy.
We now have:

Proposition 3 As F ! 0, then for all < m, a social group will form

when the share of nationalists pass the threshold ^ = m 1 = .

21
 The proposition identi…es a speci…c condition for a group to form. If
> m; then a group will form for all 2 ; and its sole e¤ect is to
magnify the policy concession made to nationalists by incumbent parties. If
(1 + ) > m > ;then the group has to organize for political parties to
o¤er an immigration policy that caters to nationalist preferences (i.e., x < 1).
Thus, the numerical strength of the nationalist group matters for policy. Also
relevant for the entry condition is the social multiplier which determines
the ex post political in‡uence of the group on policy. Formation of a group
is more likely when is high and the power of social media in recent years
may be something that has increased . Finally, it is worth re‡ecting on
why the proposition requires that < m . If immigration salience and
class polarization are very high, then the parties will pick a maximally strict
immigration policy and there is no instrumental reason for nationalists to
form a group.
We now have a comparative static on group formation:
Corollary 3 Given ; a nationalist social group is more likely to form, when
immigration salience is higher and class polarization (inequality) is
greater. Changes in these parameters will further in‡uence immigration
policy once a group has formed.
In this static model, the formation of a nationalist social group will lead to
a discretely stricter immigration policy, as the collective identity makes im-
migration salience jump from to ( ; ) = (1 + ) at = ^ : Moreover,
this multiplier implies that immigration policy is more sensitive to its drivers
with a nationalist group i place. But this result does not nullify the key
insight from the baseline model: nationalists are more likely to obtain policy
concessions whenever class polarization is large ( high) and/or immigration
is more salient for nationalists ( high).

4.3 Entry by Nationalist Political Parties

An important aspect of the rise in nationalism are the populist radical-right
parties that have entered – or grown much stronger – in many countries
during the last 10-15 years. This wave of party entry follows an earlier wave
of green-party entries in many European countries during the 1980s and
1990s.
In spite of its manifest importance, formal models with endogenous party
entry are few and far in between. In this subsection, we extend our model

22
with prospective entry of a third, nationalist party and investigate the way
in which such a party may in‡uence immigration policy. To do so, we have
to analyze how three-party systems shape not just electoral outcomes, but
government formation. Given that nationalist (and green) parties have been
more common in countries where elections are run via proportional repre-
sentation (PR) rather than plurality, we also consider how electoral rules
in‡uence these results.
Here, we show that a necessary condition for a nationalist party to enter
is that it is able to in‡uence policy by being part of the government. In
plurality rule systems, this means holding power and in proportional systems
being part of a coalition. Either way, this is more likely to happen when
there is su¢ cient share of nationalists in the population and, in the status
quo, the existing parties are not accommodating the interests of nationalists.

Preliminaries Suppose that nationalist citizens get together and form a

party at a per-capita cost of B= . To simplify the analysis, we specialize re-
distributive preferences to a linear distance function: U t tJ = t tJ :
This will rule out party entry driven by a desire to change the redistribu-
tive policy and thus keeps our focus on nationalism and immigration policy.
When rich and poor nationalists form a party, we assume that they maximize
joint (average) utility among party members. This requires that they bargain
and agree on a tax-rate compromise in the interval [t2 ; t1 ] which we denote
by t~. However, as we show below, the nationalist party could still move away
from this compromise if it negotiates a coalition with another party.
Let e 2 f0; 1g be the decision among nationalists whether to enter (leave)
a new party in the election. When e = 0; we are back to the baseline model
with just two parties, which was fully solved in Section 3.
Because entry, e = 1; is costly, it will only occur when it helps shift
expected policy in a direction preferred by the nationalists. As in the case of
a nationalist social group, we will focus on small aggregate costs of collective
action B ! 0, such that a party will always form when this implies a policy
advantage. Associated with entry will be a probability distribution over
policies (t; x) denoted by q (t; x; ; ; ) :15 Given the model symmetry, the
nationalists include equally many poor and rich individuals. If a nationalist
party enters and competes in the election, we denote the members’average
15
The two speci…c cases we study below generate di¤erent such probability distributions.

23
expected payo¤ by N ( ) ; which we ‡esh out in the applications below.16

Timing The timing of the game is similar to the model with nationalist
groups.

1. There is a given share of nationalists , salience and economic polar-

ization :

2. Nationalists choose whether to form a party at per-capita cost B= .

3. If e = 0; the two existing parties o¤er platforms ft1 ; x1 g; ft2 ; x2 g, in-

dividual and aggregate shocks ! and are realized and the election is
held where party J wins with probability P (xJ ; xK ; ) :

4. If e = 1, the election involves three parties o¤ering policies and there

are no election shocks. If no party alone wins a majority of seats,
policies are determined in post-election coalition formation.

5. Payo¤s are realized.

Party entry A nationalist party enters i¤

B
N( ) > + W (b
x ( ; )) ; (8)
2

where the left-hand side is the expected payo¤ with entry (e = 1) net of
the entry cost, while the right-hand side is the average expected nationalist
payo¤ from equilibrium policy without entry (e = 0) derived in Section 3.17
We now explore the implications of this entry condition under two electoral
rules.
16
In general this is:
!
X U t t1 + U t t2
N( )= + W (x) q (t; x; ; ; ) ; (7)
2
(t;x)2T ( ; ; )

where T ( ; ; ) is a …nite set of policy alternatives (as will be the case below).
17
We assume that no entry occurs in the case of indi¤erence.

24
Plurality-rule elections Consider a simple plurality-rule election, where
all voters cast a ballot for one of the parties in a single electoral district that
coincides with the polity. This election is winner-takes-all, so the party with
the largest vote share wins all seats, forms a government, and acquires an
unchecked right to choose policy.
Whenever a nationalist party has entered, the earlier shocks to nationalist-
voter preferences become irrelevant. Thus the nationalists –like rich and the
poor cosmopolitans before – become loyal voters who sincerely cast their
ballots for the party that represents their own interests.18

Equilibrium policies under plurality rule Sincere voting makes each

party propose the policies preferred by their voters: even though commit-
ments in immigration policy are possible, policy compromise has little value
when all parties faithfully represent their members. Parties 1 and 2 thus o¤er
policy outcomes ft1 ; 1g and ft2 ; 1g respectively. As poor and rich nationalists
have di¤erent views about redistribution, they have to agree on tax policy.
We assume they pick the optimal policy compromise t~; 0 .
In this simple plurality-rule model, the largest party wins the election (or
a lottery decides if two parties have equal numbers of votes). By symmetry,
the poor and rich parties split the cosmopolitan vote (1 ) equally. A new
nationalist party thus wins for sure when > 1=3; while each cosmopolitan
party wins with 50-50 probability when 1=3. Given these electoral
outcomes and equilibrium policies, we can write the post-entry payo¤ as

U t~ + W (0) if > 1=3

N( )= (9)
2
+ W (1) if 1=3:

Equilibrium entry under plurality rule Applying the entry condition

(8), we have:

Proposition 4 Under plurality rule and small entry costs B ! 0; a nation-

b ( ; ) > 0.
alist party enters for all > 1=3; provided that x

Entry requires two conditions to be ful…lled. It only occurs when a na-

tionalist party can collect a plurality of votes. But entry must also trigger
a stricter immigration policy. Recalling the comparative statics in Corollary
18
If we were to maintain some neutral non-loyal voters in the model, this would likely
strengthen the results (on entry) derived below.

25
1, a nationalist party is thus more likely to enter when immigration is not
very salient and class polarization not very strong. If the opposite is true
– meaning that > m – nationalists already get their preferred immigra-
b ( ; ) = 0: There is no point in incurring any entry costs, since
tion policy x
as the two cosmopolitan parties eagerly compete for the nationalist votes.
Nationalist party entry is a substitute for such policy concessions.19

PR elections Suppose the election is still held in a single polity-wide dis-

trict, but parties instead obtain seat shares proportional to vote shares. Once
again, the earlier shocks to nationalist votes are irrelevant when these vot-
ers are represented by their own party as all voters vote sincerely. We also
assume that an equilibrium policy must be backed by at least 50 percent of
the legislators.
These assumptions are consistent with a parliamentary form of govern-
ment under PR elections. With full party discipline, a party can form a
single-party majority government only when it commands more than half
the seats, which requires half the votes. Otherwise, a coalition of two parties
must agree on policy. We can see such policy agreement as the result of
government formation under a rule of positive parliamentarism.20

Equilibrium policies under PR If > 1=2, a nationalist party can form

a single-majority government, and set policy at its preferred point ft~; 0g. In
this case, the logic of entry is the same as under plurality rule, albeit with a
di¤erent threshold value of . The most interesting question is therefore if
entry can occur when < 1=2 and, in particular, when < < 1=3: That
is to say, is entry feasible under PR but not under plurality rule?
When < 1=2, parties 1 and 2 can always form a “cosmopolitan coali-
tion” to defeat the nationalists in the legislature. We that suppose such
a coalition would bargain and maximize its joint surplus by setting policy
at t; 1 . To keep a level playing …eld, assume that t = (t1 + t2 ) =2 – i.e.,
the redistributive compromise is right in between the bliss points of the two
19
This …nding has parallels with Buisseret and van Weelden (2020) who consider entry
by outside candidates in primaries. They show that outsiders favor entering elections
through traditional parties rather than as third-party candidates if and only if polarization
is su¢ ciently high.
20
Positive parliamentarism means that forming a government (recognizing a prime min-
ister) requires an active approval by a majority of the legislative assembly.

26
(equally large) cosmopolitan parties.21
To study the entry of a nationalistic party, we must specify the policy of
a nationalist-cosmopolitan coalition. Intuitively, a new nationalist party can
o¤er something in the redistributive dimension that the other cosmopolitan
party can not. For the latter party represents only one economic class of
voters, while a nationalist party represents both rich and poor nationalists.
The following result describes the policies that emerge when a nationalist
party makes a take-it-or-leave-it o¤er to a cosmopolitan party:

Lemma 1 The most favorable arrangement for a nationalist party when it

bargains with a cosmopolitan party J has t = tJ and
0 if 2 W (0) W (1)
bJ ( ) =
x 1
1 W W (0) 2
otherwise.

Intuitively, the nationalist party may o¤er the preferred redistributive

policy to its partner, in exchange for its own preferred immigration policy.
The cosmopolitan party will accept this o¤er, if the resulting policy makes it
better o¤ than in a redistributive compromise with the other cosmopolitan
party, a condition more likely to hold when economic polarization is large
bJ ( ) in the Lemma). If this participation constraint does not
(the …rst row of x
hold, the nationalist party has to concede some immigration-policy surplus
–i.e., pick xJ > 0 (the second row of x bJ ( ) in the Lemma).
Why is the take-it-or-leave-it assumption and Lemma 1 relevant? A new
nationalist party has a strong bargaining position because it has two options:
it may enter into a coalition and make a deal with either the rich or the
poor. In both cases, the cosmopolitan party gets a better redistributive
policy and the nationalist party gets a tougher immigration policy. It is thus
plausible that the two cosmopolitan parties compete, a la Bertrand, to form
a government coalition with the nationalists.
As the nationalists are equally well o¤ in either coalition, we assume they
choose the coalition partner at random. Putting the pieces together, the
post-entry nationalist payo¤ becomes

2
+ W (0) if > 1=2
N( )= J (10)
2
x ( )) if
+ W (b 2 ; 1=2 :
21
If this did not hold, it would be easier for a nationalist party to form a coalition as it
would choose to coalesce with the party that was forced to make the largest compromise
on redistribution.

27
Equilibrium entry under PR Using the expressions in (10) and the entry
condition (8), we have

Proposition 5 Under PR and B ! 0; a nationalist party enters for all

bJ ( ) ; and for all
2 [ ; 1=2]; provided that x^ ( ; ) > x > 1=2;
provided that x^ ( ; ) > 0.

A nationalist party can always be in a coalition government, but whether

it enters depends on whether the coalition’s policy is better than the status-
quo outcome without entry. The latter is given by x^ ( ; ) de…ned in Propo-
sition 1. Entry under PR is therefore possible in the range 2 ( ; 13 ); whereas
it would not occur under plurality rule. In general with x^ ( ; ) > x bJ ( ), a
nationalist party enters at a lower share of nationalists under PR. Of course,
this theoretical result is entirely in line with the strong empirical pattern
portrayed in Figure 2.

Comparative statics We now ask how party entry changes with a shock
to or . Under plurality rule and low entry costs, the condition for entry
depends only on the size of the nationalist share ; and re‡ect and only
via the side condition that x^ ( ; ) > 0:
Under PR, higher immigration salience by itself may deter entry. This
is because it makes the status-quo policy x^ ( ; ) lower and the incumbent
parties are more likely to accommodate nationalist preferences in immigra-
tion policy. Greater class polarization by itself has an ambiguous e¤ect.
On the one hand, it makes status-quo immigration policy stricter (which de-
ters entry) but, on the other hand, it makes the incumbent parties keener
to compromise with the nationalists and increases the concessions they are
willing to make (which stimulates entry).
However, if we take the complementarity stressed in Section 3 into ac-
count, we obtain a more subtle prediction under both electoral rules. Suppose
that is high but is low enough that < m so we have x^ ( ; ) = 1:
This is the most conducive to entry as the two cosmopolitan parties ignore
nationalists when no nationalist party has entered. At the same time, high
immigration salience makes the nationalists very motivated to enter a new
party.
Comparing to Section 4.2, we thus …nd a similar results for the share of
nationalists: a higher increases the chances for group formation as well as
party entry. But when it comes to polarization and salience, some parameter

28
constellation may favor group formation but disfavor party entry, and vice
versa.

Implications Party dynamics are an important part of the rise of iden-

tity politics. Here, we have stressed that a necessary condition for entry
under any electoral system is that mainstream parties are insu¢ ciently re-
sponsive to concerns of nationalist voters. Now the fraction of nationalist
citizens matters but di¤erently across electoral systems. The fact that
matters since in the next section, will be endogenous so that party forma-
tion becomes part of a dynamically coevolutionary process between values
and political organization.

5 Nationalism Dynamics
We have seen how immigration policy can re‡ect immigration salience among
nationalists, as well as the share of citizens with a nationalist identity. More-
over, the share of nationalist sympathizers appears to have changed over
time in a number of countries. To paint a more complete picture of identity
politics in the model, we therefore identify the share of citizens that identify
as nationalists.
In this section, we develop an evolutionary model of social identity (Sub-
section 5.1). The extension to a dynamic model allows for a richer analysis
of various shocks. In particular, we can ask whether temporary shocks may
have long-run consequences. For example, will a temporary stint in immi-
gration salience permanently alter nationalism and immigration policy? As
we shall see, the answer is no in the dynamic extension of the baseline model
of Section 3 (Subsection 5.2). But all the three extensions in Section 4 turn
out to sow the seeds of path dependence (Subsections 5.3 through 5.5).

5.1 Social Identity Evolution

We now extend our previous static framework to an in…nite sequence of
periods, indexed by s, populated by a stream of sequential generations. In
this subsection, we describe the evolutionary dynamics of social identity in
the context of the baseline model.

Timing In each s, the timing of events is as follows:

29
 1. The polity enters period s with a share s of nationalists in the current
generation.

2. Parties o¤er platforms ft1s ; x1;s g; ft2s ; x2;s g.

3. Individual and aggregate shocks ! and are realized.

4. An election is held where party 1 wins with probability P^ ( ; ) :

5. Payo¤s are realized.

6. Members of the next generation decide to identify as nationalists or

cosmopolitans. This determines s+1 :

Equilibrium …tness We posit a class of dynamics driven by the relative

expected payo¤ from holding one identity rather than another, where the
one that “thrives”grows over time. What drives the socialization process is
thus the expected-utility di¤erence between nationalists and cosmopolitans,
when a fraction s+1 is expected to be nationalists:

s+1 ; ; ^
= W (X s+1 ; ; ) W (1 ^
X s+1 ; ; ): (11)

In writing this expression, we use the fact that in (virtually) all our appli-
cations, parties converge on the same equilibrium immigration policy, such
that we can write x = X ^ ( ; ; ).22

Evolution and socialization Following Besley and Persson (2019b), we

formally consider the transmission of social identities across successive gen-
erations inspired by the research on cultural evolution. At each date, a
generation of “parents”fosters a generation of “children”, with two children
in every family. Reproduction follows a matching process with a fraction
of assortative mating – i.e., parents with the same identity. The remaining
fraction 1 are randomly matched with some mixed-identity couples.
Parents socialize their children. Two parents with the same identity just
pass on this identity to their children. But whether a child with mixed par-
ents becomes a nationalist depends on s+1 ; ; . It also depends on a
22
This is not strictly true for the party-takeover model that can have asymmetric equi-
libria (see Subsection 4.3). But we do not feel that the notational complexity to explicitly
allow for this is justi…ed.

30
family-speci…c shock with in…nite support and c.d.f. G ( ) ;which is sym-
metric around a zero mean with p.d.f. g ( ). The child becomes a nationalist if
s+1 ; ; ; so the probability of this event is G s+1 ; ; : With
a continuum of families, this is the proportion of nationalist children in
mixed-parents couples. Note that G ( ) increases smoothly in with G (0) =
1=2.
The nationalist share at date s + 1; given that s are nationalists at s;
evolves as:
1
s+1 = s +2 s (1 s ) (1 ) G s+1 : (12)
2

To interpret this expression, note that assortative matching preserves the

proportion of nationalists. However, among the randomly matched, a frac-
tion 2s are matched with other nationalists. The fraction of mixed-parent
households is therefore 2 s (1 s ).

5.2 Baseline Dynamics

In the baseline model, the expected outcome in period t + 1 is indepen-
dent of t+1 for all values of t+1 : That is, from Proposition 1 we have
X^ s+1 ; ; = x b ( ; ) : Using this in (11), we obtain:

s+1 ; ; = ( ; ) = W (b
x ( ; )) W (1 b ( ; )):
x (13)

By Proposition 1, function ( ; ) more likely takes a positive value the

higher is –i.e., the higher is immigration salience and class polarization
: Function ( ; ) is thus increasing in each argument.23
Given this dynamic process, we obtain the following paths for nationalism:

Proposition 6 From any starting point 2 [ ; 1 ], the polity converges

to maximal nationalism if ( ; ) > 0; but to minimal nationalism
if ( ; ) < 0:

This is a useful benchmark for the results in the next three subsections,
as the only drivers are parameters and . In addition to a¤ecting policy
as in the static model, these parameters now also shape the dynamics of
23
If = 1; then ( ; z) R 0, as x
b ( ; z) R 12 :

31
nationalism. A polity with low immigration salience and low economic po-
larization will not create pro-nationalist dynamics. Nationalism goes to its
minimum ; whenever x^ ( ; ) is su¢ ciently low –i.e., the political equilib-
rium with cosmopolitan parties does not accommodate the policy preferences
of nationalists.

Comparative dynamics Proposition 6 is illustrated in Figure 6. It shows

that permanent shocks to and ; which change equilibrium immigration
policy x^ ( ; ) and thus the relative …tness of nationalism ( ; ) by enough,
can change steady-state nationalism and alter the direction of the dynamics
of nationalism. Speci…cally, suppose that ( ; ) changes to ( 0 ; 0 ) such that

( 0; 0) > 0 > ( ; ) : (14)

The two curves in Figure 6 illustrate the alternative dynamic paths with
these parameters. Their shape re‡ect the fact that, by (12), the growth in
depends not only on G ( ( )) ; but also on (1 ) ; a function which has
a single maximum (at 1=2).
A shift from ( ; ) to ( 0 ; 0 ) will also be associated with a shift in policy.
In particular, a shift from a shrinking to a growing share of nationalists
will coincide with a shift towards a stricter immigration policy. But the
nationalist share is an outcome and not a driver in this process.
To say it di¤erently, the dynamic version of the baseline model entails no
feedback from nationalism to immigration policy. We now show how includ-
ing endogenous political organization – along either of the three extensions
considered in Section 4 –creates such feedback e¤ects.

Implications Although only a benchmark, we view Proposition 6 and the

associated dynamics as representing a particular view of how politics was
evolving in many liberal democracies in the post war era with a move to-
wards cosmopolitan values in an environment where polarization, , and
salience of nationalist sentiment, , were conducive to this. As policy be-
came increasing cosmopolitan not only in terms of immigration, there was
also greater respect for liberal values in the treatment of certain groups. The
fact that mainstream political parties were run by educated cosmopolitan
elites gave their views a policy advantage which translated into a cultural
…tness advantage of cosmopolitanism. The question is whether shocks to
and would be enough to disrupt that path.

32
5.3 Dynamics with Party Takeover
Section 4.1 already showed how cosmopolitan-led parties, which accommo-
date changing preferences by running more nationalist candidates, can have
transitions to nationalist leaders. We also showed that a nationalist leader
will choose a maximally strict immigration policy x = 0; implemented by a
set of all-nationalist candidates.
In a dynamic version of the model, this e¤ect may cause a permanent
shift of identities and policies.

Proposition 7 As long as the initial parameters are such that the two cosmopolitan-
led parties set x^ ( ; ; ; ) < 1; the polity will eventually converge to
maximal nationalism from any starting point 2 [ ; 1 ]:

Here, the each of the two parties with cosmopolitan leaders has a positive
initial probability 1 x^ ( ; ; ; ) of a shift to a nationalist leader. Suppose
only one party makes the ‡ip and shifts its immigration policy to x = 0 by
appointing only nationalist candidates. Then, log supermodularity (comple-
mentarity) implies that the other party responds to this nationalist takeover
by adopting a more anti-immigration stance (a lower x), implemented by a
higher share of nationalist candidates. Sooner or later, both parties will thus
become fully nationalist, with maximal nationalism becoming an absorbing
state with x1 = x2 = 0 and

( ; ; ) = W (0) W (1) > 0; (15)

for all ( ; ; ) :
The polity thus ends up in an equilibrium like the upper line in Figure
6 where the arrows point to the right and the dynamics lead to maximal
nationalism . Moreover, these dynamics will continue even if immigra-
tion salience and/or class polarization fall back at a level that would imply
x^ ( ; ; ; ) = 1 in a setting with two parties with cosmopolitan leaders.

Implications Party composition and takeover can respond more e¤ectively

than pure electoral competition to nationalist sentiment, as they recompose
and displace the elites who make policy. Our model predicts that the lead-
ership/elite displacement will be followed by a nationalist backlash, which
reverses previous trends towards cosmopolitanism – we can thus have path
dependence in nationalism and immigration policy. We leave it to the reader

33
to judge whether this mechanism resonates with the developments of the US
Republican Party after the 2016 election of Donald Trump, and of the UK
Conservative Party after the Brexit vote.
Of course, our simple model does not include any prospective responses
by cosmopolitan elites. If they were far sighted (and could commit policy),
they could resist accommodation of nationalists even at an electoral cost.
Moreover, they could marshal the support of cosmopolitans to form a new
party in a mirror image of the party entry that we studied in Section 4.3.
However, if the share of nationalists is falling, the window of opportunity
for such party re-entry may be quite narrow. Although our analysis is simple,
it suggests that we ought to put more focus on internal party dynamics in
understanding the dynamics of policies and identities.

5.4 Dynamics with Social Group Formation

Section 4.2 already showed how a nationalist social group can form endoge-
nously and how this is more likely with a high share of nationalists . The
policy change – from x b ( ; ) to x
b ( ( ; ) ; ) – that accompanies group
formation can happen in real time if is growing, even without any shift in
and . In this subsection, we show that this coevolution of policy and na-
tionalism opens the door to divergent dynamics with multiple steady states
and path dependence.

Nationalist cultural …tness Without an existing group, the relative …t-

ness of being a nationalist is still given by (13). When a nationalist group
has formed (and is expected to be maintained at s + 1), …tness is modi…ed
to:
( ; ; )= s+1 ; ; = W (b
x s+1 ; ; b
) W (1 x s+1 ; ; ):
(16)
Since ( ( ; ) ; ) ( ; ) the sense of collective identity created by an
organized social group creates a …tness advantage to nationalists. However,
the signs of the expressions in (13) and (16) remain a priori ambiguous.

Equilibrium dynamics We focus on the case where the threshold above

which a nationalist group forms is interior, namely
hm i1
> 1 > . (17)

34
In this case, the possible long-run outcomes are given by:

Proposition 8 If (17) holds, the dynamic model with endogenous groups

has three possible steady states:

1. If ( ( ; ) ; ) < 0; the polity converges to the unique steady-

state = from all 0 2 [ ; ]: No nationalist group will form
and any existing group will eventually be disbanded.
2. If ( ; ) > 0, the polity converges to the unique steady-state
= for all 0 2 [ ; ]: A nationalist group will form along
the equilibrium path.
3. If ( ( ; ) ; ) > 0 > ( ; )), there is a critical value ^ 2 [ ; ]
such that if 0 ^ then the polity converges to the maximal-
nationalism steady state , and a nationalist group forms along the
equilibrium path. If 0 < ^ ; the polity converges to the minimal-
nationalism steady state without a nationalist group.

In the …rst case of Proposition 8, policy does not move far enough if a
nationalist group forms to create a …tness advantage to a nationalist identity.
In the second case, the opposite is true and nationalism grows monotonically
over time, and eventually creates a incentive for a nationalist social group to
form. Once the group has formed, the complementarity between salience and
group size sets and the share of nationalists and immigration policy start to
reinforce each other.
The third case is arguably the most interesting, as the steady-state out-
come depends on the starting point. Here, too, the prospective complemen-
tarity between the nationalist share and immigration policy plays a key role.
A polity that begins with a high enough share of nationalists will see that
share rise over time. This, in turn, guarantees that eventually crosses the
threshold for nationalist group formation (if it has not reached that point
already). The group reinforces nationalism by causing a shift to a stricter
immigration policy, which starts a two-way feedback process between and
x: However, if the polity begins with su¢ ciently weak pro-nationalist mo-
mentum, any nationalist group is eventually disbanded, the polity loses its
nationalist orientation, and immigration policy becomes more open.
Once a polity has crossed threshold ^ , no forces restrain nationalism. A
necessary condition for this outcome is that a group forms. This is because
( ( ; ); ) < 0 implies ( ; ) < 0 –i.e., nationalism cannot take o¤, even

35
if a nationalist group (counterfactually) exists. The third case in Proposition
8 is illustrated in Figure 7, where the blue curves show the dynamics around
a threshold at ^ with arrows indicating the possible paths of nationalism. In
a polity where > ^ nationalism will grow, while it will shrink for initial
values < ^ .

Path dependence One implication of Proposition 8 is the possibility of

path dependence (hysteresis), which leads to a permanent e¤ect from a tem-
porary shock. This is illustrated in Figure 7, where the red curves illus-
trate the response to a temporary shock from ( ; ) to ( 0 ; 0 ) such that
( 0 ; 0 ) > 0. The shifts makes s grow over time. If it grows beyond ^
and ( [1 + ^ ] ; ) > 0; as in the …gure, there is a permanent e¤ect on the
trajectory for – nationalism continues to grow even after a reverse shock
when and/or fall back from ( 0 ; 0 ) to ( ; ).
This theoretical logic can be related to the practical attempts of combat-
ting the extreme nationalism that had formed in some countries in the 1930s.
In post-war Japan and Germany, the allied forces imposed policies equivalent
to x = 1: This helped quell nationalist sentiments in young generations that
were forming their identities afresh. On top of that, nationalist groups were
outlawed. The model could rationalize such legal constraints as a method
to reduce nationalist …tness. It could also illustrate the role that a supra-
national body –like the EU –might play during events like the recent refugee
crisis. By trying to maintain a policy with x = 1; such a body might prevent
nationalist sentiments from translating into policy in individual countries.
More generally, the logic of Proposition 8 shows that social organization –
a part of civil society –creates path dependence if it a¤ects the way citizens
with a certain outlook (in our case nationalism) can interact to mobilize.
Such forces have doubtless been at work in recent political developments
and may have been enhanced by social media – if we think of those as a
positive permanent shock to the sense of collective identity, ; which lowers
the threshold in (17).

5.5 Dynamics with Party Entry

Section 4.3 relied on an extension of our baseline model where a new (third)
political party can enter (at a cost) to pursue a nationalist policy agenda.
We showed that such entry requires the nationalists in the population to
reach a certain threshold share, which is dependent on the electoral rule. In

36
this subsection, we study how the nationalist share and immigration policy
coevolve over time and interact with party entry, in a dynamic version of the
party-entry model.

Dynamics with latent party entry Using the results on party entry in
Section 4.3, we can rewrite (13), our expression for the (relative) …tness of a
nationalist identity as:
P
(x;t)2T ( ; ; ) [ W (x) W (1 x)] q (t; x; ; ; ) if e
( ; ; )=
( ; ) otherwise.
(18)
This is a piece-wise continuous function with an upward jump at e – i.e.,
the electoral-rule and parameter-dependent threshold share for the entry a
nationalist party that is contained in Propositions 4 and 5.
The jump in the …tness of nationalism at e re‡ects that – under both
electoral rules –party entry is instrumental and only occurs when it triggers
a stricter immigration policy. Note that Proposition 5 implies that we may
have ~ = , such that nationalist party is always active under PR.
Using (18) in (12), we obtain:

Proposition 9 The dynamic model with endogenous parties has three pos-
sible steady states:

1. If (e; ; ) < 0 the unique steady-state has = for all 0 2

[ ; ] and no nationalist party enters. If 0 > ~ ; the nationalist
party exits along the equilibrium path, as passes ~ from above.
2. If ; ; > 0, the unique steady-state has = for all
0 2 [ ; ] and if 0 < ~ a nationalist party enters along the
equilibrium path, as passes ~ from below.
3. If (e; ; ) > 0 > ; ; , the polity converges to with a
nationalist party if 0 e and to without a nationalist party if
0 < e.

As with social-group formation, a growing share of nationalists can change

political organization. Here, this amounts to the creation of a three-party
system due to the entry of a nationalist party. Entry feeds back to ex-
pected policy and the share of nationalists. Drawing on Propositions 4 and

37
5, party entry occurs only when the pre-entry equilibrium does not recog-
nize the forces of nationalism and exhibits a lax immigration policy (a high
value of x). If a nationalist party enters, this is associated with a stricter
immigration policy, which provides a boost to nationalist identity and anti-
immigration preferences, a feedback that can create path dependence. Such
path dependence will de…nitely arise in the third case of Proposition 9, which
is analogous to the third case of Proposition 8.

Party entry and comparative dynamics How shocks to and shape

nationalism and policy dynamics is quite subtle and dependent on the elec-
toral rule. With plurality, a shock to and can have an e¤ect on if
( ; ) > 0 which may eventually trigger formation of a nationalist party.
Once > 1=3, that party remains in place even if the shock to and re-
verses. Nationalist policy is thus boosted permanently. This is very similar
to the mechanism of social-group formation.
Under PR, entry is less sensitive to : A boost to and resulting in
( ; ) > 0 now makes the share of nationalists cross the higher threshold
of = 1=2 to have a permanent e¤ect which outlasts the shock. If a party
is active for all , a positive shock to can actually cause the nationalist
party to exit, but only if it pushes down x^ ( ; ) all the way to 0: If x^ ( ; )
is initially close to 1, a positive shock to can induce entry. In both cases,
we will see a process towards nationalist dynamics, which may or may not
reverse if the shock goes away.

Insights In spite of the subtlety, a key insight from Proposition 9 is that

endogenous party entry can generate path dependence for policy, nationalism,
and the party system. We may thus have to study party entry in a dynamic
model to fully grasp the sustainability of nationalism and immigration policy.
The results in this subsection also show why it is important to interpret
events through the lens of a conceptual framework. It is tempting to argue
that nationalist parties cause harsher immigration policies and promote na-
tionalist sentiments. But in our framework the party system is more of an
outcome, or a mediating variable. The true causes are the fundamentals,
and : Similarly, the strength of nationalism, , may appear as a short-
term cause. But as our model brings out, it should also be thought of as a
malleable long-run outcome. In particular, one may want to claim that
determines immigration policy x: In our setting, this claim would su¤er from

38
a di¢ cult chicken-and-egg problem, as these outcomes coevolve in a two-way
dynamic process.

Putting the pieces together Summing up the insights from this section,
large enough shocks to ultimate drivers – here called class polarization ( )
and immigration salience ( ) –can have long-run consequences on national-
ism and immigration policy. If these shocks are permanent, then this result
is unsurprising and would appear already in a static model (if we interpret
equilibria under alternative parameter values as alternative steady states).
The value added of our dynamic analysis is that we can also study tem-
porary shocks to these drivers. We have seen that endogenous political orga-
nization can serve as a mediator, whereby temporary shocks to class polar-
ization or immigration salience can induce long-term changes in nationalism
and immigration policy. By the results in Propositions 4-9, we expect such
path dependence to more likely occur via nationalist takeovers of existing
parties under plurality rule, but via entries of new nationalist parties under
PR.

6 Final Remarks
Part of the success of some democracies around the world has stemmed from
building common-interest solutions to policy-making of the kind stressed in
Besley and Persson (2011). When economic polarization and salience over
other divisive dimensions of policy are low, this kind of politics is easier to
sustain. But in the past twenty years, we have begun to see a shift towards
a certain kind of identity politics. The forces that have shaped this dynamic
are still debated and whether this is a short-term shock that will reverse,
or a form of more disruptive long-run change is a key question. Its answer
depends on the resulting dynamics of values and political organization.
To gain a better understanding, we have developed a novel framework
that highlights organizational change in tandem with the dynamics of iden-
tity politics. This framework captures some elements of the clash between
incumbent cosmopolitan elites – who control major parties – and an in-
creasingly nationalist electorate. It also helps us uncover the possibilities
of path dependence where temporary shocks can have permanent political
consequences.

39
 We have focused on just two underlying drivers. Changes in parameter
represent shifts in nationalist sentiments, which could re‡ect the European
refugee crisis and/or increased salience fostered by social-media interactions.
Changes in parameter embody a range of factors promoting economic po-
larization, such as the fallouts from globalization or the recession and policies
following the …nancial crisis. We study the comparative statics, as well as
the comparative dynamics, of shocks to these parameters.
Distinctive of our dynamic framework is that it allows non-linear re-
sponses at “critical junctures.”These can induce organizational change, even
for small changes in caused by underlying shocks to and . The model
helps us think about the timing of events. Observers frequently argue that
pressures on manufacturing and other industries due to immigration and
trade liberalization predate recent political developments. However, this is
exactly what we should expect with threshold e¤ects of the kind suggested
by our model. Tensions could build gradually, but suddenly erupt into orga-
nized political activity and policy change. Another core aspect of the model
–still rare in political economics –is that shocks may trigger the formation
of new nationalist social groups and/or the entry of new nationalist political
parties. The model suggests that immigration-policy shifts are more likely to
be part of incumbent party platforms in plurality-rule electoral systems, but
by new nationalist parties in PR electoral systems.
Much has been made of the so-called “cultural backlash”in the wake of a
shift towards more liberal values among political elites (Inglehart and Norris
2019). In our framework, the capture of incumbent parties by cosmopolitans
cause liberal policies towards immigration, which promote a cultural drift
towards cosmopolitanism. A backlash can arise in the model, e.g., from a
shock like the European refugee crisis (a higher value of ). When polariza-
tion (parameter ) is high enough, this can produce a policy reversal towards
stricter immigration policy (lower x). Such a shock may also produce polit-
ical reversals: the formation of new nationalist groups and/or the entry of
nationalist parties can set in motion a self-reinforcing growth of nationalism.
Although we do not have data to support this idea, the descent into fascism
in the middle of the 20th century –which followed a period of tolerance and
openness – shows that we should not assume that all cultural trends are
unidirectional.
A number of extensions would add complexity and realism to our frame-
work. We have already relaxed our baseline symmetric structure in the Ap-
pendix. A more extensive treatment can be found in Besley and Persson

40
(2021), where we allow nationalists to belong disproportionately to di¤er-
ent income groups, introduce a less stylized model of income inequality, and
permit the salience of nationalism to di¤er across economic groups. Such ex-
tensions can help explain why the mainstream left has su¤ered more than
the mainstream right from the rise of nationalistic forces.
Perhaps the biggest intellectual challenge posed by the events that moti-
vate the paper is to build richer models of political dynamics. The progression
of salient political cleavages and of political organization have been explored
in political-science writings, but they are not really part of the political-
economics tool kit. History starkly reminds us how important it is to have a
grasp of such dynamics. In earlier generations, it took a world war and a long
period of conscious interventions to deconstruct the 1930s institutions that
had evolved to support nationalism. Unless today’s political elites under-
stand and act upon the dynamics that underpin the rise of identity politics,
they may see history repeat itself.

41
References
[1] Akerlof, George and Rachel Kranton (2000). “Economics and Identity,”
Quarterly Journal of Economics 115(3), 715–753.
[2] Akerlof, George and Rachel Kranton (2010). Identity Economics: How
Our Identities Shape Our Work, Wages, and Well-Being, Princeton, NJ:
Princeton University Press.
[3] Algan, Yann, Sergei Guriev, Elias Papaioannou, and Evgenia Passari
(2017). “The European Trust Crisis and the Rise of Populism,” Brook-
ings Papers on Economic Activity 48(2), 309-400.
[4] Autor, David. H., David Dorn, Gordon Hanson, and Kaveh Majlesi
(2020). “Importing Political Polarization? The Electoral Consequences
of Rising Trade Exposure,”forthcoming in American Economic Review.
[5] Baron, David (1994). “Electoral Competition with Informed and Unin-
formed Voters”, American Political Science Review 88(1), 33-47.
[6] Barth, Fredrik (1998). Ethnic Groups and Boundaries: The Social Or-
ganization of Culture Di¤erence, Prospect Heights, IL: Waveland Press.
[7] Becker, Sascha. O., Thiemo Fetzer, and Dennis Novy (2017). “Who
Voted for Brexit? A Comprehensive District-Level Analysis,”Economic
Policy 32(92), 601–650.
[8] Besley, Timothy and Stephen Coate (1997). “An Economic Model of
Representative Democracy,” Quarterly Journal of Economics 112(1),
85-114.
[9] Besley, Timothy, Olle Folke, Torsten Persson, and Johanna Rickne
(2017). “Gender Quotas and the Crisis of the Mediocre Man: The-
ory and Evidence from Sweden,” American Economic Review 107(8),
2204-2242.
[10] Besley Timothy and Torsten Persson (2011). Pillars of Prosperity: The
Political Economics of Development Clusters, Princeton, NJ: Princeton
University Press.
[11] Besley, Timothy and Torsten Persson (2019a). “Democratic Values and
Institutions,”American Economic Review: Insights 1(1), 59-76.

42
[12] Besley, Timothy and Torsten Persson (2019b). “The Dynamics of En-
vironmental Politics and Values,” Journal of the European Economic
Association 17(4), 993–1024.

[13] Besley, Timothy and Torsten Persson (2020). “Organizational Dynam-

ics: Culture, Design, and Performance,”unpublished manuscript.

[14] Besley, Timothy and Torsten Persson (2021). “Identity Politics and the
Decline of the Left,”unpublished notes.

[15] Bisin, Alberto and Thierry Verdier (2011). “The Economics of Cultural
Transmission and Socialization,” Chapter 9 in Benhabib, Jess (ed.),
Handbook of Social Economics, Volume 1A, Amsterdam: North-Holland,
Elsevier.

[16] Bisin, Alberto, and Thierry Verdier (2017). “On the joint Evolution of
Culture and Institutions,”Working Paper, No. 23375, National Bureau
of Economic Research.

[17] Bowles, Samuel (2004). Microeconomics: Behavior, Institutions and

Evolution, Princeton, NJ: Princeton University Press.

[18] Boyd, Robert and Peter Richerson (1985). Culture and the Evolutionary
Process, Chicago, IL: University of Chicago Press.

[19] Buisseret, Peter and Richard Van Weelden (2020). “Crashing the Party:
Elites, Outsiders and Elections,”American Journal of Political Science
64(2), 356-370.

[20] Caiani, Manuela and Linda Parenti (2016). European and American Ex-
treme Right Groups and the Internet, London: Routledge. e-book.

[21] Caiani, Manuela and Donatella della Porta (2018). “The Radical Right
as Social Movement Organizations,”in Rydgren, Jens (ed.), The Oxford
Handbook of the Radical Right, Oxford, UK: Oxford University Press.

[22] Caplin, Andrew and Barry Nalebu¤ (1991). “Aggregation and Imperfect
Competition: On the Existence of Equilibrium,” Econometrica 59(1),
25-59.

43
[23] Carmines, Edward G. and Michael W. Wagner (2006), “Political Issues
and Party Alignments: Assessing the Issue Evolution Perspective,”An-
nual Review of Political Science 9, 67-81.

[24] Cavalli-Sforza, Luigi L. and Marcus Feldman (1981). Cultural Transmis-

sion and Evolution, Princeton, NJ: Princeton University Press.

[25] Feddersen, Timothy J., Itai Sened, and Stephen G. Wright (1990). “Ra-
tional Voting and Candidate Entry under Plurality Rule,” American
Journal of Political Science 34(4), 1005-1016.

[26] Gattinara, Pietro C. and Andrea L.P. Pirro (2018). “The

Far Right as a Social Movement,” European Societies
https://doi.org/10.1080/14616696.2018.1494301.

[27] Gennaioli, Nicola and Guido Tabellini (2019). “Identity, Beliefs, and
Political Con‡ict,”unpublished manuscript.

[28] Gidron, Noam and Peter A. Hall (2018). “Populism

as a Problem of Social Integration,” available at
https://scholar.harvard.edu/…les/hall/…les/gidronhallmay2018.pdf.

[29] Granovetter, Mark (1978). “Threshold Models of Collective Behavior,”

American Journal of Sociology 83(6), 1420-1443.

[30] Grossman, Gene and Elhanan Helpman (1996). “Electoral Competition

and Special-Interest Politics,” Review of Economic Studies 63(2), 265-
286.

[31] Grossman, Gene and Elhanan Helpman (2020). “Identity Politics and
Trade Policy,”forthcoming in Review of Economic Studies.

[32] Hobolt, Sara and James Tilley (2016). “Fleeing the Centre: The Rise of
Challenger Parties in the Aftermath of the Euro Crisis,”West European
Politics 39(5), 971-991.

[33] Inglehart, Ronald and Pippa Norris (2017). “Trump and the Populist
Authoritarian Parties: The Silent Revolution in Reverse,” Perspectives
on Politics 15(2), 443-454.

44
[34] Inglehart, Ronald and Pippa Norris (2019). Cultural Backlash: Trump,
Brexit and Authoritarian Populism, Cambridge, UK: Cambridge Uni-
versity Press.
[35] Kitschelt, Herbert (1993). “Class Structure and Social Democratic Party
Strategy,”British Journal of Political Science 23(3), 299-337.
[36] Kitschelt, Herbert and Anthony McGann (1997). The Radical Right in
Western Europe: A Comparative Analysis, Ann Arbor, MI: University
of Michigan Press.
[37] Kitschelt, Herbert (2018). “Party Systems and Radical Right-wing Par-
ties,”in Rydgren, Jens (ed.), The Oxford Handbook of the Radical Right,
Oxford, UK: Oxford University Press.
[38] Levy, Gilat (2004). “A Model of Political Parties,”Journal of Economic
Theory 115(2), 250-277.
[39] Levy, Gilat and Ronny Razin (2019). “Echo Chambers and Their E¤ects
on Economic and Political Outcomes,” forthcoming in Annual Review
of Economics.
[40] Lindbeck, Assar and Jörgen Weibull (1987). “Balanced Budget Redistri-
bution as the Outcome of Political Competition,” Public Choice 52(3),
273-297.
[41] McAdam, Doug (1982). Political Process and the Development of Black
Insurgency, 1930–1970, Chicago, IL: University of Chicago Press.
[42] Meltzler, Allan M. and Scott F. Richard (1981). “A Rational Theory of
the Size of Government”, Journal of Political Economy 89(5), 914-927.
[43] Mutz, Diana C. (2018). “Status Threat, Not Economic Hardship, Ex-
plains the 2016 Presidential Vote,”PNAS 115(19), E4330-E4339.
[44] Osborne, Martin J. and Al Slivinski (1996). “A Model of Political
Competition with Citizen Candidates,”Quarterly Journal of Economics
111(1), 65-96.
[45] Persson, Torsten and Guido Tabellini (1999). “The Size and Scope
of Government: Comparative Politics with Rational Politicians,” 1998
Marshall Lecture, European Economic Review 43(4-6), 699-735.

45
[46] Persson, Torsten and Guido Tabellini (2000). Political Economics: Ex-
plaining Economic Policy, Cambridge, MA: MIT Press.
[47] Persson, Torsten and Guido Tabellini (2021). “Culture, Institutions, and
Policy”, Chapter 15 in Bisin, Albert and Giovanni Federico (eds.), Hand-
book of Historical Economics, Amsterdam: North-Holland, Elsevier.
[48] Pettigrew, Thomas F. (2017). “Social Psychological Perspectives on
Trump Supporters,” Journal of Social and Political Psychology 5(1),
107-116.
[49] Reny, Tyler T., Loren Collingwood, and Ali A. Valenzuela (2019). “Vote
Switching in the 2016 Election: How Racial and Immigration Attitudes,
Not Economics, Explain Shifts in White Voting,”Public Opinion Quar-
terly 83(1), 91-113.
[50] Roberts, Kevin (1977). “Voting over Income Tax Schedules,”Journal of
Public Economics 8(3), 329-340.
[51] Roemer, John E. (2001). Political Competition: Theory and Applica-
tions, Cambridge, MA: Harvard University Press.
[52] Romer, Thomas (1975). “Individual Welfare, Majority Voting, and the
Properties of a Linear Income Tax,”Journal of Public Economics 4(2),
163-185.
[53] Rueda, David (2005). “Insider-outsider Politics in Industrialized Democ-
racies: The Challenge to Social Democratic Parties,”American Political
Science Review 99(1), 61-74.
[54] Rydgren, Jens (2007). “The Sociology of the Radical Right,” Annual
Review of Sociology 33(1), 241-262.
[55] Rydgren, Jens (ed.) (2018). The Oxford Handbook of the Radical Right,
Oxford, UK: Oxford University Press.
[56] Sandholm, William H. (2010). Population Games and Evolutionary Dy-
namics, Cambridge, MA: MIT Press.
[57] Shayo, Moses (2009). “A Model of Social Identity with an Application
to Political Economy: Nation, Class, and Redistribution,” American
Political Science Review 103(2), 147-174.

46
[58] Skocpol, Theda R. (1979). States and Social Revolutions: A Compara-
tive Analysis of France, Russia, and China, New York, NY: Cambridge
University Press.

[59] Stryker, Sheldon and Peter J. Burke (2000). “The Past, Present, and
Future of an Identity Theory,”Special Millennium Issue on the State of
Sociological Social Psychology, Social Psychology Quarterly 63(4), 284-
297.

[60] Tilly, Charles (1978). From Mobilization to Revolution, Reading, MA:

Addison-Wesley.

[61] Tajfel, Henri (1974). “Social Identity and Intergroup Behaviour,”Social

Science Information 13(2), 65-93.

[62] Tajfel, Henri and John C. Turner (1979). “An Integrative Theory of
Intergroup Con‡ict,” in W.G. Austin and S. Worcel (eds.), The Social
Psychology of Intergroup Relations, Monterey, CA: Brooks/Cole.

[63] Turner, John C. (1985). “Social Categorization and the Self-concept: A

Social Cognitive Theory of Group Behaviour,”Lawler, Edward J. (ed.),
Advances in Group Processes, volume 2, Greenwich, CT: JAI Press.

[64] Sandholm, William H. (2010). Population Games and Evolutionary Dy-

namics, Cambridge, MA: MIT Press.

[65] Vives, Xavier (2005). “Games with Strategic Complementarities: New

Applications in Industrial Organization,” International Journal of In-
dustrial Organization 23(7-8), 625-637.

[66] Vliegenthart, Rens and Hajo G. Boomgaarden (2007). “Explaining the

Rise of Anti-Immigrant Parties: The Role of News Media Content in the
Netherlands,”Tijdschrift voor Communicatiewetenschap 26(2), 404-417.

[67] Wagner, Markus and Thomas M. Meyer (2017). “The Radical Right as
Niche Parties? The Ideological Landscape of Party Systems in Western
Europe, 1980-2014,”Political Studies 65(1), 84-107.

[68] Walder, Andrew (2009). “Political Sociology and Social Movements,”

Annual Review of Sociology 35, 393-412.

47
[69] Weibull, Jörgen W. (1995). Evolutionary Game Theory, Cambridge,
MA: MIT Press.

48
 Appendix
A Model Details
A.1 Baseline Nash Equilibrium
Electoral strategies As party objectives coincide with voter utilities, all
the poor (rich) cosmopolitans always vote for party 1 (2), provided that
(2) holds. While the cosmopolitans are loyal voters, the nationalists be-
come swing voters: they vote for the party that o¤ers them the highest
utility. But these utilities are subject to random shocks, as in the standard
probabilistic-voting models (e.g., Lindbeck and Weibull 1987, Persson and
Tabellini 2000).
Formally, let vJK be swing-voter utility o¤ered by party J to group K
nationalists:
vJK (tJ ; xJ ) = U (tJ tK ) + W (xJ ). (19)
For example, a nationalist from group 1 (the poor) votes for party 1 if v11 +
!+ v21 ; where the idiosyncratic shock ! in favor of party 1 has symmetric
c.d.f. H (!) ; with unimodal p.d.f. h (!) ; and the aggregate shock in favor
of party 1 has log-concave symmetric c.d.f. G( ) with p.d.f. g( ).
As the nationalist shares among poor and rich are equal, the parties o¤er
total swing-voter utilities
1 1 1
v(x1 ) = (v + v12 ) = [U (0) + U (t1 t2 )] + W (x1 ) (20)
2 1 2
1 1 1
v(x2 ) = (v + v22 ) = [U (0) + U (t2 t1 )] + W (x2 ) (21)
2 2 2
we write these utilities as v(xJ ); because the …rst term in square brackets
is exogenous and equal across parties by U (t1 t2 ) = U (t2 t1 ): Therefore,
the only way for parties to o¤er more or less swing-voter utility is to vary
their migration policy. The maximum swing-voter utility on o¤er to the
nationalists is thus 12 [U (0) + U (t1 t2 )] + W (0); while the minimum is
1
2
[U (0) + U (t1 t2 )] + (W (1).

Winning probabilities The poor-party vote share among nationalists is

n1 = H ( + v(x1 ) v(x2 )). It wins the election if its overall vote share is

49
larger than that of the rich party –i.e., if
1 1
(1 ) + n1 (1 ) + (1 n1 ) :
2 2
1
The condition is thus n1 2
; or

1 1
+ v(x1 ) v(x2 ) H = 0; (22)
2
where the equality takes advantage of our assumption that c.d.f. H (!) is
symmetric.
As the aggregate shock is drawn from c.d.f. G ( ) ;the win probability
for poor party 1 is
P (x1 ; x2 ) = G (v(x1 ) v(x2 )) = G ( (W (x1 ) W (x2 )) . (23)
This probability is independent of , the population fraction of nationalists.
This is because both parties symmetrically lose loyal cosmopolitan voters
when the share of nationalist swing voters rises.24

Gains from winning To study the political equilibrium, we write the

surplus from winning the election for party 1 as
Z 1 (x1 ; x2 ) = + W (1 x1 ) W (1 x2 ):
Similarly, the surplus for party 2 is
Z 2 (x1 ; x2 ) = + W (1 x2 ) W (1 x1 ):
We can thus write the poor party’s objective function as P (x1 ; x2 ) Z 1 (x1 ; x2 )
plus a constant.
Condition (2) implies that both parties want to win the election for all
xJ 2 [0; 1]. Note that W (1 x) is increasing in x – i.e., a policy closer
to the cosmopolitan loyal voters’ bliss point x = 1. But the probability
of winning is decreasing in x – i.e., a policy closer to the nationalist swing
voters’bliss point x = 0. This means that parties face a standard trade-o¤:
pushing policy closer to the party-members’preference diminishes the chance
of winning the election.
24
Symmetry of G implies that the win probability of rich party 2 is:

1 P (x1 ; x2 ) = G ( (W (x2 ) W (x1 )) .

50
Political equilibrium We now look for a Nash equilibrium in the choices
of fx1 ; x2 g ; with

x1 2 arg max [Z 1 (x1 ; x2 )]G[ (W (x1 ) W (x2 )] (24)

x2[0;1]

x2 2 arg max [Z 2 (x1 ; x2 )]G[ (W (x2 ) W (x1 )] : (25)

x2[0;1]

The symmetry of economic payo¤s U ( ), and c.d.f. G; imply that parties

face symmetric problems.
A political equilibrium is de…ned by a winning probability for the poor
party (party 1) and two immigration policies
n o
b b1 ( ; ) ; x
P ( ; );x b2 ( ; ) :

We have:

Lemma 1 A Nash equilibrium exists and is unique.

Proof. The electoral game is log supermodular

L (x1 ; x2 ) = log[ + W (1 x1 ) W (1 x2 )] + log G ( (W (x1 ) W (x2 ))

@ 2 L(x1 ;x2 )
with @x1 @x2
=
h i
g( (W (x1 ) W (x2 ))
@ @W (x) @W (x)
G( (W (x1 ) W (x2 ))
@W (x) @x1 @x2
+ [ +W (1 x1 )1 W (1 x2 )]2 @W@x
(1 x) @W (1 x)
1 @x2
@ 2 L(x1 ;x2 )
> 0: The …rst term in @x1 @x2 is positive, because distribution G
is log
concave and W (x) is decreasing in x; while the second term is positive because
W (1 x) is increasing in x: We can then appeal to the …xed-point theorem
of Topkis (Vives 2005, Caplin and Nalebu¤ 1991).
This relies on the fact that the electoral-strategy game is (log) supermod-
ular. Thus the immigration policies are strategic complements: one party
pursuing a more nationalistic policy raises the motive of the other party to
do the same. The strategic complementarity re‡ects two e¤ects. A tougher
migration policy by one party: (i) increases polarization, which induces the
other party to compete more intensely for o¢ ce; (ii) appeals more to swing
voters and hence reduces the other party’s probability of winning, lowering
the cost of setting a policy against the interests of its loyal voters. Supermod-
ularity also makes it straightforward to handle corner solutions where x = 0
or x = 1.

51
Characterization The political equilibrium has a very simple form. Given
the model symmetry, it is perhaps not surprising that the two parties choose
the same migration policies xb1 ( ; ) = xb2 ( ; ) = x
b( ; ) and have the same
b 1
chance of winning the election P1 ( ; ) = 2 : To develop some intuition for
the next result, consider the …rst-order condition for party 1 (that de…nes its
reaction function)

W1 x (1 x1 ) G ( (W (x1 ) W (x2 )) (26)

+ [ + W (1 x1 ) W (1 x2 )] Wx (x1 ) g ( (W (x1 ) W (x2 )) Q 0.

The …rst term is the cost of conceding to the nationalists while the second
term is the bene…t of a larger win probability.
As a preliminary step, de…ne function h (m) for m 2 [m; m] from

W1 x(1 h (m)) g (0)

=m ;
Wx (h (m)) G (0)

where m = WW1 x (0)

x (1) g(0)
= G(0) and m = W1 x (0) g(0)
Wx (1)
= G(0) : It is straightforward to see
that h ( ) is a decreasing function.

A.2 Probability of Winning in the Group Model

Evaluating (6) for each income class K = 1; 2 and summing across across
these classes gives
1
v(xJ ) = (1 + )f [U (0) + U (t1 t2 )] + W (xJ )g:
2
The previous utility expressions in (20) and (21) are thus multiplied by 1+ ;
an expression which is increasing in and . Repeating the steps outlined in
Subsection 2.1, and exploiting the model symmetry, the poor party’s proba-
bility of winning, in the presence of a nationalist group, is now

P (x1 ; x2 ) = G[v(x1 ) v(x2 )] = G[ ( ) (W (x1 ) W (x2 ))]

where ( ) = (1 + ) is the enhanced salience of nationalism once nation-

alists operate as an organized group. Clearly, the function ( ) is increasing
in and for all .

52
References
[1] Caplin, Andrew and Barry Nalebu¤ (1991). “Aggregation and Imperfect
Competition: On the Existence of Equilibrium”, Econometrica 59(1), 25-
59.

[2] Lindbeck, Assar and Jörgen Weibull (1987). “Balanced Budget Redis-
tribution as the Outcome of Political Competition,” Public Choice 52(3),
273-297.

[3] Persson, Torsten and Guido Tabellini (2000). Political Economics: Ex-
plaining Economic Policy, Cambridge, MA, MIT Press.

[4] Vives, Xavier (2005). “Games with Strategic Complementarities: New

Applications in Industrial Organization,” International Journal of Indus-
trial Organization 23, 625-637.

B Proofs
Proof of Proposition 1 To prove the proposition, it su¢ ces to note that
at an interior solution (26) can be arranged so that

W1 x (1 h ( )) g (0)
= :
Wx (h ( )) G (0)
W1 x (1) g(0)
Moreover, since h ( ) is increasing, there exists m = Wx (0)
= G(0) and m =
W1 x (0) g(0)
Wx (1)
The result follows by noting that we obtain a corner solution if
= G(0) :
> m and < m.

Proof of Proposition 2 Recall that the party J’s objective here is

P (xJ ; xK ; )Z J (xJ ; xK ) + (1 )xJ ;

where P (xJ ; xK ; ) = G[ (W (xJ ) W (xK ))], and Z J (xJ ; xK ) = + W (1

xJ ) W (1 xK ). Hence an interior solution xJ will satisfy the …rst order
condition:

g[ (W (xJ ) W (xK ))] Wx (xJ )[ + W (1 xJ ) W (1 xK )]

G[ (W (xJ ) W (xK ))]W1 x (1 xJ ) + (1 ) = 0:

53
In equilibrium, both parties pick a common immigration policy, and hence
the optimal policy x^( ; ; ; ) should satisfy
(1 ) = G(0)W1 x (1 x^( ; ; ; )) g(0) Wx (^
x( ; ; ; )):
@x^ @x^
To obtain @
and @
, we di¤erentiate the both sides and get
@ x^
1 = [ G(0)W(1 x)(1 x) (1 x^) g(0) Wxx (^
x)] ;
@
and
@ x^
= [ G(0)W(1 x)(1 x) (1 x^) g(0) :
Wxx (^
x)]
@
Since W (:) is convex, we have @@x^ < 0 and @@ x^ > 0. This says that both parties
will pick a stricter immigration policy with higher bene…t of reappointment
and lower (higher popularity). Intuitively, the party leader now face
a tradeo¤ between attracting more nationalist voter on the one hand, and
pleasing cosmopolitans and avoiding losing power in the party on the other
hand. Now with higher popularity in the party (i.e. lower ) and better
reappointment bene…ts, the leader is ensured to be able to marginally pick
stricter immigration policy for a better chance of winning in the general
election.

Proof of Proposition 3 To prove this proposition, observe that for a

nationalist, the payo¤ without a group is
U (0) + U (t1 t2 )
+ W (^
x ( ; )) ;
2
while the net payo¤ with such a group is
U (0) + U (t1 t2 ) F
+ W (^
x( ( ) ; )) :
2
Recall that as F goes to zero, a nationalist group will form if and only if
[W (^
x( ( ); )) W (^
x( ; ))] > 0:
Note that x^( ; ) = 0 when m, and x^ is (at least weakly) increasing in
both its arguments. Hence when m, [W (^ x( ( ); )) W (^ x( ; ))] =
0, and there will be no group formation. Hence nationalist group will form
if and only if both of these conditions are satis…ed: (1) (1 + ) > m and
(2) < m.

54
Proof of Proposition 4 With > 1=3, then for small enough B, the
condition
B
[W (0) W (b
x ( ; ))] >

b ( ; ) > 0. Moreover, if
holds for small enough B; as long as x 1=3,
entry is always deterred because
B
[W (1) W (b
x ( ; ))] <

b ( ; ) 2 [0; 1].
for all x

Proof of Lemma 1 The best possible outcome for the nationalists in a

coalition with party J is given by

[U (t t1 ) + U (t t2 )]
max + W (x) (27)
(t;x) 2
subject to U t tJ + W (1 x) U t + W (0) :

[U (t t1 )+U (t t2 )]
Now note that given our assumption on preferences, 2
is
independent of t for t 2 [t2 ; t1 ]. To …x ideas, consider J = 1 (the argument
for J = 2 is analogous). This means that t = t1 and x b1 ( ) = 0 solves (27) if

U (0) + W (1) U t~ + W (0) : (28)

Because U (0) = 0 and

t1 + t2 t2 t1
U t t1 = t1 = = :
2 2 2
we need that
W (0) W (1) .
2
b1 ( ) is implicitly de…ned by
If (28) does not hold, the solution for x

W (1 x) = W (0) :
2
bJ ( ).
This expression yields the formula for x

55
Proof of Proposition 5 We look at entry at N b J policies: tJ ; x
bJ ( ) .
Hence, we postulate equilibrium policy probabilities

1 if > 1=2
q t~; 0; ; ; =
0 if 1=2.

and
0 if > 1=2
q tJ ; x
bJ ( ) ; ; ; = J = 1; 2:
1=2 if 1=2.
Putting these pieces together, we have equation (10).

2
+ W (0) if > 1=2
N( )= J
2
x ( )) if
+ W (b 2 ; 1=2 .

If > 1=2, then entry will occur if:

B U (t1 t2 ) + U (0)
+ W (0) > + W (^
x ( ; ))
2 2
= + W (^
x ( ; )) ;
2
which always holds as B ! 0. For < 1=2, entry requires

B U (t1 t2 ) + U (0)
xJ ( ))
+ W (b > + W (^
x ( ; ))
2 2
= + W (^
x ( ; ))
2
or
B
bJ ( )
W x W (^
x ( ; )) > ;

bJ ( ) < x^ ( ; ).
which holds as B goes to zero whenever x

Proof of Proposition 6 To see why this proposition is true, note that

if ( ; ) < 0, then (12) implies s+1 < s for all 2 ; 1 . But if
( ; ) > 0, then for all 2 ; s+1 > s :

56
Proof of Proposition 7: Denote the state by f 1;s ; 2;s g where J;s 2
fC; N g. Under mild conditions, party takeover will always result in nation-
alism convergence to maximum nationalism in the long-run. This gives four
possible states and hence four possible values of . To study these, we need
to specify what will happen in di¤erent kinds of political equilibria since,
above, we only considered the case where the state was fC; Cg. In general,
let ~ ( ; : 1;s ; 2;s ) denote the cultural …tness of the nationalists in di¤erent
states. Then observe that
~( ; : C; C) = W (b
x ( ; ; ; )) W (1 b ( ; ; ; ))
x

in the political equilibrium described by Proposition 2. The sign of this is

uncertain a priori.
With fN; N g then x b ( ; ; ; ) = 0. Hence
~( ; : N; N ) = W (0) W (1) > 0.

To complete the analysis, we only need to consider what happens when one
party has a cosmopolitan leader and faces a party with a nationalist leader.
Without loss of generality, let 1 = C and 2 = N . Then x2 = 0 and the
optimal choice of x1 is

x~ ( ; ; ; ) 2 arg max f[ + W (x) W (0)] G[ (W (x) W (0)] + (1 )xg :

x2[0;1]
(29)
The supermodular structure implies that x~ ( ; ; ; ) < x b( ; ; ; ).
b( ; ; ; ) < 1, then any party run
To complete the proof, note that for x
by a cosmopolitan has a positive probability of becoming nationalist in each
period and the lower bound on this probability is p = 1 (1 b( ; ; ; ) >
)x
0. Then after date S the lower bound on the probability that a party has a
cosmopolitan leader is
S
p ! 0 as S ! 1:
Hence in the limit all leaders are nationalists and

s+1 ; ; = W (0) W (1) > 0

and ! .

57
Proof of Proposition 8 Incorporating institutional dynamics in response
to , modi…es the timing as follows:

1. The polity arrives to period s with a given share of nationalists s

equally split among the rich and the poor, and with given values of
salience and economic polarization :

2. Nationalists choose whether to organize a group at per-capita cost

F= s . (They can also abandon a pre-existing group.)

3. Parties o¤er platforms ft1s ; x1;s g; ft2s ; x2;s g.

4. Individual and aggregate shocks ! and are realized.

5. An election is held where party 1 wins with probability P ( ; ) or

P^ ( ( ) ; )

6. Payo¤s are realized.

7. The next generation of citizens decide to identify as nationalists or

cosmopolitans. This determines s+1 :

The …rst and second parts are straightforward. If ( ( ) ; ) < 0 then

( ; ) < 0. Hence, whether or not an interest group forms, (12) implies
s+1 < s for all . If ( ; ) > 0, then for all 2 ; s+1 > s using
(12). To prove the third part, de…ne
m 1
( ( ) ; ) for 1
( ; ; )=
( ; ) otherwise.

This function is increasing and hence can switch sign at most once. From
(12) s+1 > s ; if and only if ^h ( ; ; ) > 0.i As ( ( ) ; ) > 0 > ( ; ))
by (17), there must exist H 2 m 1 1 ; such that ( (~ ) ; ) > 0 and
by (12) we have s+1 > s . When ( ; ) < 0, (17) implies that there exists
m
L such that s+1 < s for all L 2 ( ; 1 1 ]. Hence, there must exist
a unique value ^ 2 ; at which ( ; ; ) switches from strictly negative
to strictly positive so that s+1 > s for > ^ and s+1 < s for < ^ .

58
Proof of Proposition 9 The proof is similar to Proposition 8. The …rst
and second parts are straightforward. If (^ ; ; ) < 0; then no nationalist
party forms and (12) implies s+1 < s for all . If ; ; > 0, then for
all 2 ; s+1 > s using (12). To prove the third part, recall that

P
(x;t)2T ( ; ; ) [ W (x) W (1 x)] q (t; x; : ; ) if ^
( ; ; )=
( ; ) otherwise.

It follows that ( ; ; ) = ; ; < 0 for all < ^ . Hence from (12)

s+1 > s if and only if ^ . When ( ; ; ) > 0; s+1 < s for < ^ .

C Relaxing Symmetry
We now show that the core static model can be generalized in three directions
to allow for (still exogenous) prospective correlations between economic sta-
tus and identity. First, we let immigration salience di¤er across nationalists
of di¤erent economic groups, with K denoting the immigration salience in
group K = 1; 2: Second, we let the share of nationalists di¤er across groups,
with the share in group K denoted by K : Third, we let 1=2 be the
fraction of poor citizens in the population.
As in the core model, we suppose that cosmopolitans are loyal voters.
The nationalists are still swing voters: they vote for the party that o¤ers
them the highest utility. With the earlier assumptions, = 1 + (1 ) 2
is the fraction of swing voters, although this group now includes two types
with di¤erent immigration salience.

Swing voters To study swing-voter behavior, let vJK be swing-voter utility

o¤ered by party J to group K nationalists:

vJK (xJ ) = U K (t^J t^K ) K

W (xJ ) . (30)

As in the baseline model, a nationalist from group 1 (the poor) votes for
party 1 if v11 + ! + v21 ; where the idiosyncratic shock ! in favor of party
1 is uniformly distributed to [ 1= ; 1= ]. and the aggregate shock in favor
of party 1 has a log-concave symmetric c.d.f. G( ) with p.d.f. g( ).

59
 The parties o¤er swing-voter utilities
1 1
v1 + (1 ) 2 2
v1 = 1
U (0) + (1 ) 2 U (t^1 t^2 )]
1 1
+ (1 ) 2 2 W (x1 )
and
1 1
v2 + (1 ) 2 2
v2 = 1
U (0) + (1 ) 2 U (t^1 t^2 )]
1 1
+ (1 ) 2 2 W (x2 )
The left party’s share of swing voters is
1
1
n1 = 1+ + [W (x1 ) W (x2 )]
2
2
(1 ) 2
+ 1+ [W (x1 ) W (x2 )]
2
= (1 + ( + # [W (x1 ) W (x2 )])) :
2
In this expression, # = [ 1 (1 ) 2 ] = measures whether nationalists
are more prevalent in the poor or rich population (equal to zero in the sym-
metric case) while = ( 1 1 + (1 ) 2 2 )= measures average immigra-
tion salience among swing voters.

Win probabilities The poor party wins the election if its overall vote
share is larger than that of the rich party –i.e., if
1 2
1 + n1 (1 ) 1 +( n1 ) :
This boils down to
2 1
(1 ) (1 ) (1 )
# + [W (x1 ) W (x2 )]
= + [W (x1 ) W (x2 )] :
In this expression,
2 1
(1 ) (1 ) (1 ) #
= + (31)

measures which party has an innate electoral advantage. The case > 0
refers to one where this advantage lies with the poor party. Note also that
is higher (lower) when there are more poor (rich) nationalists.

60
 As aggregate shock is drawn from c.d.f. G ( ) ;the win probability for
poor party 1 is
P (x1 ; x2 ; ) = G ( [W (x1 ) W (x2 )]) . (32)
This is increasing and in x2 x1 : Thus if the rich party has a more (less) pro-
immigration stance than the poor party this raises (cuts) the win probability
of the rich party.

Equilibrium immigration policies As in the symmetric model, we look

for a Nash equilibrium in the choices of fx1 ; x2 g ; with
x1 2 arg max Z 1 (x; x2 ) G ( [W (x) W (x2 )]) (33)
x2[0;1]

x2 2 arg max Z 2 (x1 ; x) G ( + [W (x1 ) W (x)]) : (34)

x2[0;1]

A political equilibrium results in a win probability for the poor party (party
1) and two immigration policies
n o
Pb ( ; ; ) ; x
b1 ( ; ; ) ; x
b2 ( ; ; ) :

The log super-modularity properties of the symmetric model carry over this
non-symmetric model. Note that if 1 = (1 ) 2 then = 0, and we are
back to the symmetric model except that replaces . We can thus conclude
that if = 1=2 and 1 = 2 , the (a)symmetry of immigration preferences
does not matter for the political equilibrium.

Consequences of asymmetry? What if we depart from that assumption?

The most natural case to focus on is the case where nationalists are more
common among the poor: 1 > 2 : In this case, > 0 if > 1=2, and the
poor party has a an innate political advantage among swing voters.
The …rst-order conditions, at an interior solution for both parties are:
W1 x (1 x1 ) G ( (W (x1 ) W (x2 ))
+ [ + W (1 x1 ) W (1 x2 )] Wx (x1 ) g ( (W (x1 ) W (x2 )) = 0.
and
W1 x (1 x2 ) G ( + (W (x1 ) W (x2 ))
+ [ + W (1 x2 ) W (1 x1 )] Wx (x2 ) g ( + (W (x1 ) W (x2 )) = 0.

61
 Now, we show that > 0 implies that x b2 ( ; ; ) < xb1 ( ; ; ) – i.e.,
the rich party will o¤er the strictest immigration policy to compensate for
its innate political disadvantage. To see this, observe that we can write the
…rst-order conditions as
' (x1 ) = ( (W (x1 ) W (x2 )) Z 1 (x1 ; x2 ) (35)
and
' (x2 ) = ( + (W (x1 ) W (x2 )) Z 2 (x1 ; x2 ) (36)
g( )
where ( ) = G( )
and ' (x) = W1 x (1 x) =Wx (x) are both decreasing func-
tions. Suppose that, contrary to the claim, xb2 ( ; ; ) b1 ( ; ; ): Then
x
( (W (x1 ) W (x2 )) < ( + (W (x1 ) W (x2 )), Z 1 (x1 ; x2 ) < Z 2 (x1 ; x2 )
and ' (x1 ) > ' (x2 ). This implies that
( (W (x1 ) W (x2 )) Z 1 (x1 ; x2 ) = ' (x1 )
' (x2 )
= ( + (W (x1 ) W (x2 )) Z 2 (x1 ; x2 )
which yields a contradiction. Hence the left party will have a more pro-
immigration stance than the right party if > 0.

Comparative statics Now consider the comparative statics for changes in

or . Given the supermodular structure, we will get unambiguous compar-
ative statics if the right hand side of (35) and (36) are increasing in and .
Speci…cally, fx1 ; x2 g will then be decreasing in f ; g as in the symmetric
case.
For the asymmetric case, there are two e¤ects to consider.
The …rst e¤ect is a direct and coincides with that in the symmetric model,
where the right-hand side of (35) and (36) increase leading to a lower x1 and
x2 . This operates via the fact that multiplies everything on the right hand
side and Z 1 (x1 ; x2 ) and Z 2 (x1 ; x2 ) are both increasing in .
The second, indirect e¤ect renders the overall e¤ect of a change in immi-
gration salience or class polarization ambiguous. Consider …rst an increase
in : Since x1 > x2 , and ( (W (x1 ) W (x2 )) is increasing in , so
the indirect e¤ect reinforces the fact that (35) is increasing in . However,
the same e¤ect decreases ( + (W (x1 ) W (x2 )) and hence decreases
the left-hand side of (36), all else equal, going against the direct e¤ect. In-
tuitively, as goes up and as x1 > x2 , the advantage of the poor party
decreases and this encourages the rich party to increase x2 .

62
 Consider instead an increase in : In addition to the direct e¤ect present
in the symmetric model, this shift has a similar indirect e¤ect. Speci…cally,
( + (W (x1 ) W (x2 )) on the right-hand side of (35) is decreasing in
when # > 0;while ( (W (x1 ) W (x2 )) on the right-hand side of (36)
is increasing in . Intuitively, as goes up and as x1 > x2 , the advantage
of the poor party goes down all else equal. This encourages that party to
lower x1 . If increases, this increases the advantage of the poor party and
encourages it to raise x1 , all else equal.
Putting this together, a su¢ cient condition for an unambiguous compara-
tive static in is that the right-hand side of (36) is increasing in ; meaning:

( + (W (x1 ) W (x2 )) Z 1 (x1 ; x2 )

+ 0( + (W (x1 ) W (x2 )) Z 1 (x1 ; x2 ) [W (x1 ) W (x2 )] > 0:

A su¢ cient condition for an unambiguous comparative static with respect to

is that the right-hand side of (35) is increasing in ; meaning:
0
( (W (x1 ) W (x2 )) + ( (W (x1 ) W (x2 )) Z 1 (x1 ; x2 ) # > 0:

Note that in the special case where follows a Laplace distribution with
mean zero, i.e.
1
2
exp b if > 0
G( ) =
1 12 exp b
if 0;
then the equilibrium is symmetric even with asymmetries. Moreover, the
comparative statics are unambiguous since ( ) = 1b .

63