Down and Out
Author(s): Sharan Grewal and Steve L. Monroe
Source: Comparative Politics , July 2019, Vol. 51, No. 4 (July 2019), pp. 497-518
Published by: Comparative Politics, Ph.D. Programs in Political Science, City University
of New York
Stable URL: https://www.jstor.org/stable/10.2307/26663945
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Comparative Politics, Ph.D. Programs in Political Science, City University of New York is
collaborating with JSTOR to digitize, preserve and extend access to Comparative Politics
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Down and Out:
Founding Elections and Disillusionment with Democracy in
Egypt and Tunisia
Sharan Grewal and Steve L. Monroe
“She was in a state of shock and confusion. [...] It was one thing for the [Muslim]
Brotherhood to win close to 40 percent, but how could 28 percent of her countrymen
vote for ultraconservative Salaﬁ parties? [...] She mourned not only for what she feared
Egypt might become, but for a country that she could no longer recognize, a country
that was no longer really hers. It raised the question: was [democracy] worth it? For
liberals like [her], it apparently wasn’t.”1
How citizens respond to electoral loss is critical to the success of democratic
transitions.2
Supporters of losing parties in founding elections must opt to remain within
the democratic system for a nascent democracy to take root. Where electoral losers
become disillusioned with democracy, they may turn to violence, encourage a coup, or
support the revival of the former regime.
While existing literature on electoral losers’ attitudes toward democracy focuses on
explaining cross-national variation, it does little to help transitional governments
understand where disillusioned losers will be concentrated. Theory and evidence of
subnational variation in losers’ disillusionment could help a democratic government
decide where to target its resources to keep losers invested in the system. For instance,
should the newly-elected government in Egypt in 2012 have prioritized accommodating
losers in Cairo or Luxor? Despite its policy importance, few studies have attempted to
explain subnational variation among losers in new democracies.3
Counter-intuitively, we argue that the most disillusioned losers following a founding
election reside where their parties performed the best subnationally. This result stems
from losers’ expectations of winning at the national level. Where a party is strong locally,
party supporters will be more likely to expect to win nationally. When national electoral
results upset that expectation, they may become disillusioned with democracy as they
realize it will not bring them to power. Where a party is weak locally, on the other hand,
497
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
party supporters already expect to lose nationally, and thus will not become any more
disillusioned when their expectation is simply conﬁrmed. They likely opposed democracy
from the start. We therefore hypothesize that the most disillusioned losers in new
democracies will reside in areas where their party is strongest.
We test this theory in post-Arab Spring Egypt and Tunisia. Islamist political parties
swept these founding elections. Among the losers—non-Islamists—we examine
whether support for democracy decreased the most in governorates where their parties
were strongest. Using pre- and post-election survey data, we ﬁnd that, on average, losers
living in governorates where losing parties performed better became more disillusioned
with democracy after the elections.
Shattered expectations of winning may drive these patterns in democratic
disillusionment. First, losers in governorates where they were strong, and therefore
expected to win, were more supportive of democracy pre-election than losers where
they were weak. Second, the effect of loser strength on democratic disillusionment was
greater for losers who were less aware of national-level politics, and thus whose
expectations primarily stemmed from local-level information. Third, we ﬁnd a similar
expectations bump among electoral winners.
Losers’ disillusionment with democracy did not result in political apathy. Losers in
governorates where their parties were strong also became more supportive of
authoritarian rule. Beyond these attitudinal results, we ﬁnd that governorates where
losers were strong witnessed a greater increase in votes for candidates afﬁliated with the
former autocratic regimes in subsequent elections. These results provide important
predictions on which electoral losers are likely to become disillusioned and spoil a
democratic transition—crucial information for a transitional government seeking to
consolidate democratic governance.
Disillusionment and Democratic Consolidation
A central ﬁnding in the literature on democratic consolidation is that transitions are
more likely to succeed when all players accept the rules of the game.4
For democracy to
take root, those who lose the founding elections must decide to remain within the
system, temporarily assuming the role of the opposition while preparing to contest the
next elections. Political culture models of democratization demonstrate that this
acceptance of democracy must take place not only among the elite but also among the
masses.5
Where the supporters of losing parties become disillusioned with democracy,
they may subvert nascent democratic institutions by turning to violence, or calling for
the return of the former regime. Mass attitudes toward democracy have been correlated
with democratic consolidation and breakdown,6
in particular with military coups7
and
incumbent takeovers.8
Mass attitudes are so integral to the success of a transition that
they are often part of the deﬁnition of democratic consolidation.9
The literature on the “winner/loser gap” has demonstrated that electoral losers
become less satisﬁed with democracy, while winners become more satisﬁed.10
This
498
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
winner-loser gap is especially pronounced where it is most consequential: new
democracies.11
Beyond dissatisfaction with democracy, electoral losers have also been
found to become more risk-seeking12
and willing to protest.13
Why do some electoral losers become disillusioned while others support nascent
democracies? Cross-nationally, the dominant explanation is how badly the losers have
lost the election. The mechanisms are twofold. First, the better the losers do, the greater
their representation and thus buy-in into the new system. Parliamentary systems and
elections based on proportional representation are thought to more effectively keep
losers within the system than winner-take-all, ﬁrst-past-the-post elections, and
presidential systems.14
Second, the better the losers’ performance in the founding
elections, the more likely they are to believe that they can win subsequent elections.15
These two mechanisms suggest that, cross-nationally, the better losers do in founding
elections, the more likely democratic transitions are to succeed.
However, even in lopsided founding elections, not all losers turn against the
system. As we will demonstrate, there remains considerable subnational variation in
losers’ disillusionment with democracy. The prevailing literature thus provides little
guidance for transitional governments about which losers are most likely to play spoiler.
Such information could prove useful in determining which losers the transitional
government should make an extra effort to coopt into the nascent democratic system.
A Theory of Expectations and Disillusionment
We begin with a common, simplifying assumption that voters have relatively stable
preferences over the primary political cleavage. In political economy models of
democratization, for instance, the poor are assumed to prefer redistribution/high tax
rates, while the rich prefer low tax rates.16
In polities where the dominant cleavage is
ideological, such as Islamists versus secularists, we assume that secularists tend to
remain secularists, even if their support for a particular secular party may change. While
party identiﬁcations may be ﬂuid, the preferences and ideologies underlying them are
assumed to be relatively ﬁxed. Accordingly, we deﬁne a winner/loser with respect to the
policy agenda of the party receiving the largest number of votes in an election. If an Islamist
party receives the most votes, then Islamists are the winners and secularists the losers.
With perfect information about who would win and lose a democratic election, few
losers would support democracy.17
However, when there is uncertainty about ensuing
electoral results, losers who believe they may win may support democracy in the run-up
to the ﬁrst elections. Disillusionment, then, could occur when an individual expects to
win but then loses: upon realizing their party or ideology only commands a minority,
they may no longer feel that democracy is in their interest.
Once disillusioned, losers may join forces with those who always opposed
democracy and begin subverting the transition. Indeed, given the frustrated expectations
logic outlined in Ted Gurr’s Why Men Rebel,18
those losers who become disillusioned
with democracy may be even more likely to rebel against the system than losers who
499
Sharan Grewal and Steve L. Monroe
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
opposed democracy from the start. Since Gurr’s writing, there has been extensive work
in psychology and economics demonstrating that unmet expectations breed frustration,
anger, and rebellion.19
Which losers expected to win? Numerous sources can shape expectations of
winning an election. In mature democracies, individuals rely on past electoral results,
opinion polls, and the media.20
In new democracies, however, especially before the
founding elections, there is no credible electoral history or media apparatus to help
predict a party’s strength at the national level.21
Voters’ local context can also inﬂuence expectations of winning. During the 2008
U.S. presidential election, Miller et al. found that individuals in states won by Barack
Obama were more likely to predict Obama’s national victory, while individuals in states
won by John McCain had incorrectly predicted a McCain victory.22
Likewise, in the
most recent U.S. presidential election, urban, coastal Democratic strongholds were
reportedly the most shocked by Donald Trump’s victory, as their local environment
(their “liberal bubble”) had colored their national-level expectations of who would win.
We contend that ahead of founding elections, an individual’s strongest prior for
how she expects her party to perform on a national level likely stems from local context:
how strong her party is in her community. At the local level, she can observe likeminded
partisans in her daily interactions. In the case of religious parties, she can
observe attendance at the local mosque or church. As the campaign season progresses,
she can observe each party’s activity, the size of their local ofﬁces, and their ability to
mobilize supporters. Ahead of the founding elections, individuals’ expectations for how
a party will perform nationally will thus be strongly conditioned by that party’s local
strength.
Where a loser’s party is strong locally, those losers should be more likely to expect
their party to be strong nationally prior to the elections. Accordingly, they should
initially be supportive of democracy as they believe they will win overall. However,
they should then become disillusioned upon learning from the elections that their party
had limited national support. Where a loser’s party is weak locally, on the other hand,
she will tend to believe it is weak nationally as well, and thus feel that democracy is not
in her interests even before the ﬁrst elections. The elections simply conﬁrm her
expectations. Her attitudes towards democracy should remain relatively unchanged.
We therefore hypothesize that:
H1: In new democracies, losers in governorates where their party is strong should
experience more disillusionment with democracy after the founding elections than losers
in governorates where their party is weak.
To test the micro-foundations of the theory, we generate three testable implications:
1. Losers in governorates where they are strong should be more supportive of
democracy pre-election than losers in governorates where they are weak. They
should also be more optimistic about the election and more likely to report they
plan to vote.
500
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
2. The differential effect of local strength on disillusionment with democracy should
be more pronounced among losers who are less politically aware. Losers who are
more politically aware, for instance, by following the news or using the internet/
social media, likely have a better sense of their national electoral weight before the
elections than those who are not.
3. Winners in governorates where they are weak should see a greater increase in
support for democracy than winners in governorates where they are strong.
Mirroring the expectations gap among the losers, winners who are unsure of
winning should see a greater jump in support for democracy upon winning the
elections than winners who had expected to win.
We study the disillusioned because they are the most likely to spoil a democratic
transition. We expect that:
H2: Losers in governorates where they are strong should become more supportive of
autocratic rule.
Finally, to complement these attitudinal results, we predict the following behavioral
outcome:
H3: Losers in governorates where they are strong will be more likely to vote for candidates
representing the former authoritarian regime in subsequent elections.
Uncertainty about how a party will perform in the ﬁrst elections is an important
scope condition to our theory. It is therefore most relevant to new democracies where
the primary political cleavage is economic or ideological, not based on relatively ﬁxed,
ascriptive categories. In new democracies where ethnic cleavages dominate politics, for
instance, each ethnic group might already have a sense of their numbers ahead of the
founding elections.23
We assume, but do not test, that our theory applies to democratic
transitions where the dominant political cleavage is not ascriptive. These include
founding democratic elections in the former Soviet Union, Latin America, and East
Asia.
Our theory is indifferent to the type of ﬁrst election. Whether for a president,
parliament, or constituent assembly, each type of election should play a similar role in
updating voters’ prior expectations of who will win subsequent national elections.
While the type of election certainly affects the overall level of disillusionment among
the losers, we expect similar patterns of subnational variation among losers within each
type of electoral system. Indeed, we ﬁnd similar subnational variation in Tunisia (with a
PR, parliamentary system) and Egypt (with a presidential system and mixed PR/FPTP
parliamentary elections).
None of these hypotheses deny the importance of other factors, such as income or
education, in shaping losers’ democratic attitudes. These factors, however, are generally
ﬁxed in the run-up and follow-up of a founding election. They cannot fully explain why
some electoral losers opt in and then out of the democratic process in the medium and
501
Sharan Grewal and Steve L. Monroe
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
short term. Indeed, it is these disillusioned losers that can tip the political scales and
derail the democratic transition. In highlighting the overlooked informational role of
founding elections, we argue that changes in losers’ expectations of national victory
underlie democratic defectors’ attitudes and behavior precisely when democratic
consensus is needed most: after the ﬁrst election.
Research Design and Analysis
We test this theory in Egypt and Tunisia, the epicenters of the Arab Spring. Popular
protests in early 2011 toppled both countries’ long-time dictators and initiated
democratic transitions. Islamist political parties swept both countries’ founding
elections. In Tunisia, the Ennahda (Renaissance) party won the most votes (37 percent)
in the Constituent Assembly elections. In Egypt, the Muslim Brotherhood won 37.5
percent of the vote in parliamentary elections. The Egyptian Brotherhood’s victory was
augmented by the ultraconservative Salaﬁ Nour party’s second place performance,
adding an additional 25 percent of the vote to the Islamists’ tally. Both countries quickly
polarized along secular-religious lines, making secularists or “non-Islamists”24
the
losers of interest. While the secular-religious cleavage was not the only source of
division, it became the primary cleavage in both countries.25
Egypt and Tunisia are prime locations for testing our theory because the dominant
political cleavage was ideological, for which the informational role of their founding
elections should be particularly strong. Indeed, the extent of the Islamist sweep in both
elections came as a surprise to many secular activists.26
The two countries also capture cross-national variation in loser strength. The
existing literature linking disillusionment with losers’ cross-national strength would
highlight that in Tunisia, the winning Islamists won just 37 percent of the vote, while in
Egypt they won over 60 percent. By selecting two countries in which the losers have
such large differences in strength, we seek to demonstrate that there exists considerable
subnational variation in disillusionment, regardless of how the losers perform on a
(cross-) national level.
Popular accounts provide initial support for our hypotheses. In Egypt, liberal
activists in wealthy Cairo suburbs were shocked that their countrymen would vote in
such numbers for Islamist movements. Liberal politician Mohamed ElBaradei observed
that the revolutionary youth were “decimated” in the elections, disconnected “with the
people on the street,” and felt “let down” by democracy.27
One liberal blogger
noted that “everyone around me is falling apart: physically, psychologically, and
emotionally.” The electoral results made them realize that “you don’t know this country,
that you live in social and cultural ghettos of your own making.”28
Similar shock and disillusionment rocked liberals in the wealthy coastal areas of
Tunisia. Houda Cherif, co-founder of the secular, pro-business party Afek Tounes,
realized after losing the elections that “the secular message was aimed at the elite,” and
made little headway in Tunisia’s more conservative and impoverished interior regions.
502
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
“We were completely wrong.”29
Moez Ali, founder of a secular organization in Tunis,
observed that “there are a lot of people who are disappointed by the result.”30
We use pre- and post-election Arab Barometer survey data to empirically test our
hypotheses. Wave II of the Arab Barometer was conducted in the summer of 2011, prior
to the 2011 Tunisian Constituent Assembly Elections and the 2011–2012 Egyptian
Parliamentary Elections. Wave III of the Arab Barometer, conducted in early 2013,
provides a post-election measure of the same questions with the same sampling
methodology. Figure A.4 in the Appendix displays a timeline of each survey and
election in Egypt and Tunisia.31
Wave II (our pre-election survey) was implemented after Egypt’s March 2011
constitutional referendum. One may worry that results from this referendum may have
already updated losers’ expectations about their national popularity before the Wave II
survey. However, Egypt’s non-Islamist parties were divided on the referendum. This
division clouded assessments of non-Islamist parties’ national weight. Moreover, the
magnitude of the Salaﬁ party’s success and the non-Islamists’ failure in the ensuing
parliamentary elections was largely unexpected.32
Polls showed the Salaﬁ Al-Nour
party with only 9 percent of the vote before the elections.33
It would go on to win 25
percent of the vote. Wave II therefore still reﬂects Egyptian non-Islamists’ democratic
attitudes prior to the shocking returns of the ﬁrst parliamentary elections.
There are two major limitations to the Arab Barometer surveys. First, it is not a
panel dataset. We employ a range of matching techniques (discussed below) to ensure a
proper comparison of electoral losers from before to after the elections. Nevertheless,
our results cannot directly address individual-level changes in attitudes towards
democracy. Second, in an ideal world, our post-election survey would come
immediately after the founding elections to measure the instantaneous feeling of shock
and disillusionment. Wave III of the Arab Barometer, however, was conducted a year
after the founding elections. This timing has both costs and beneﬁts. On the one hand,
there could have been other factors that affected democratic attitudes over the course of
that year. Indeed, Egypt and Tunisia’s deteriorating security and economies from 2011
to 2013 undoubtedly dampened aggregate, national-level support for democracy. We
show that these macro-level changes cannot explain subnational variation in democratic
support. In robustness checks, we ﬁnd that losers did not differ by governorate in their
evaluations of the economy, of security, or in the provision of government services. By
ruling out alternative explanations, we hope to increase conﬁdence that the geographic
variation in disillusionment with democracy stems from different expectations of
winning the founding elections. The late timing of the post-election survey could also be
a beneﬁt. Existing studies that show an instantaneous disillusionment with democracy
invariably provoke questions about this sentiment’s durability. Read in this light, our
study suggests that this effect can endure a year after the founding elections. This
echoes Dahlberg and Linde’s ﬁnding that losers’ disillusionment can last a parliament’s
four-year tenure.34
While imperfect, the timing of these two waves of the Arab Barometer surveys
allows us to cross-sectionally compare electoral losers’ democratic attitudes before and
503
Sharan Grewal and Steve L. Monroe
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
after founding elections. We examine whether support for democracy decreased more in
areas where non-Islamist parties were stronger.
We use speciﬁc questions in the Arab Barometer to identify the losers and their
attitudes (see appendix for the wording of all questions used). Leveraging respondents’
reported vote choice is unfeasible since the Arab Barometer did not ask about previous
vote choice in the post-election survey.35
In addition, only 19 percent of respondents
named a party when asked for their intended vote choice in the pre-election survey. We
use this intended vote choice question to isolate two powerful predictors of vote
choice—trust in the winning Islamist party and support for shari‘a law—which we will
combine to identify non-Islamist respondents.
The ﬁrst predictor is trust in the winning Islamist party (Ennahda in Tunisia and the
Muslim Brotherhood in Egypt), asked on a 4-point scale from “absolutely no trust” to “a
great deal of trust.” Trust in a party is related to, though conceptually distinct from,
party support. While support denotes an evaluation of the party’s performance, trust
refers to “a more generalized feeling” about whether they “can be trusted to take care of
one’s interests.”36
Given that in both Egypt and Tunisia’s transitions the winning parties
had never been in power and had no record of performance, trust is theoretically a more
appropriate indicator.37
The second predictor is the respondent’s support for the implementation of shari‘a
(Islamic law), a hallmark of the Islamists’ agenda.38
As the role of Islam in government
was the dominant axis of contention in both Egypt and Tunisia,39
this question helps
distinguish individuals who support the winning Islamist parties from those who
support the losing non-Islamist parties. Indeed, the secular-religious cleavage was the
primary predictor of vote choice in other post-election surveys.40
Table A.2 in the Appendix demonstrates that both trust in Islamists and support for
shari‘a are strong predictors of intention to vote for the winning Islamist parties. While
the two predictors are highly correlated, we include both to determine the losers for two
reasons. First, there may be hardline Islamists (Salaﬁs) who do not trust the more
moderate Muslim Brotherhood/Ennahda, yet are not non-Islamist losers. Measuring
support for shari‘a allows us to exclude these Salaﬁs and focus on non-Islamists.
Second, the popularity of Ennahda and the Muslim Brotherhood declined considerably
during their time in ofﬁce.41
By the post-election survey, over half of the sample
reported “absolutely no trust” in them. This large, post-election category likely includes
some disaffected Ennahda/Muslim Brotherhood supporters, who are also not nonIslamist
losers. By including support for shari‘a, which remains much more consistent
across the two surveys, we remove these disaffected winners from our sample of
electoral losers. Non-Islamists, then, are identiﬁed as those respondents who exhibited
lower than the median level of trust in Ennahda/the Muslim Brotherhood as well as
lower than the median level of support for the implementation of shari‘a both pre- and
post-election.
Both the trust in Islamists and the support for shari‘a questions suffer from missing
data. Across the two variables, 842 of 4,810 respondents, almost 18 percent, are missing
one or the other. While this rate of missingness is not abnormal, given that we wish to
504
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
test our hypotheses at the subnational level, every observation is important. We use
Multivariate Imputation by Chained Equations (MICE) to account for these missing
data. Imputation is a commonly applied solution to missingness.42
Through predictive
mean matching, MICE calculates predicted values for missing observations based on
their values for other variables in the dataset (excluding our dependent variable of
democratic attitudes). Beyond the standard demographic controls, we include in the
dataset questions regarding individual piety, religion in public life, and the importance
of piety in evaluating political candidates.
With the imputed data, 1,865 (38 percent) of the 4,810 respondents are coded as
non-Islamist. Without imputation, the number of non-Islamists would be 1,398 (35
percent) of the non-missing 3,968. While we also present results with these data
missing, we prefer to use the imputed data to maximize the information available from
the Arab Barometer surveys.
We examine whether losers in areas where they were strong experienced greater
disillusionment with democracy than where they were weak. We therefore subtract
losers’ support for democracy in 2011 from their support for it in 2013. While there is a
number of questions in the Arab Barometer survey that measure support for democracy,
we choose: “To what extent do you think democracy is appropriate for your country?”
for two reasons. First, the answers fall along a 0–10 scale, allowing for greater variance
than the typical Likert scale-style democracy questions. Second, the question explicitly
limits the context “to your country,” instead of considerations of democracy in the
abstract.
The winner/loser gap literature typically measures satisfaction with how democracy
is working, rather than the suitability of democracy. Several studies have found that
individuals may become dissatisﬁed with how democracy works but still support
democracy as a system of rule.43
By choosing this higher bar of the suitability of
democracy, we intend to show that losing transitional elections can induce rejection of
democratic governance in their country as a whole.
Consistent with our expectations, non-Islamists’ Suitability of Democracy ratings
fell considerably pre- to post-election. In Tunisia, the mean suitability of democracy
among non-Islamists dropped from 5.8 to 5.0 points on the 0–10 scale. The mean
suitability of democracy in Egypt fell from 6.6 to 4.6 after the elections. Consistent with
the existing literature’s cross-national predictions, losers’ support for democracy
appears to have fallen by a greater amount in Egypt where the losers performed worse.
We test whether, sub-nationally, losers in areas where they performed relatively
well experienced greater disillusionment than those in areas where their parties
performed relatively poorly. The key independent variable is Loser Strength, which is
the vote share of the non-Islamist parties in the founding elections. While we would
have liked to drill down to the district (markaz/qism) level, we are limited by the survey
data, which provide only seven observations per district. As a result, we collect electoral
data at the governorate (muhafaza/wilayah) level, equivalent to a U.S. state. Tunisia has
twenty-four governorates, while Egypt has twenty-seven. We then identify Arab
Barometer survey respondents by their governorate of residence. With the imputed data,
505
Sharan Grewal and Steve L. Monroe
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
we have an average of thirty-six non-Islamists surveyed per governorate; with the data
missing, thirty non-Islamists.
Figure 1 presents the mean disillusionment with democracy for each governorate in
Egypt and Tunisia. The x-axis represents Loser Strength, while the y-axis represents a
governorate’s mean change in Suitability of Democracy among losers pre- to post-
election.
Much of the data support our theoretical predictions. In Sidi Bouzid, where the
Tunisian revolution began, non-Islamists performed incredibly well in the elections,
receiving a whopping 84 percent of the governorate’s votes, and were hugely
disillusioned thereafter (experiencing an average drop of 4 points on the 0–10 scale).44
More generally, non-Islamists were more disillusioned in the wealthy coastal areas of
Sousse, Monastir, and Mahdia (the “Sahel”) and in the Northwest than they were in the
Islamist strongholds in the neglected interior (e.g., Gabes, Mednine, and Gafsa).
Similarly, in Egypt, non-Islamists were more disillusioned in Cairo and former NDP
strongholds like Minya and Sohag than they were in Islamist strongholds like Beheira,
Kafr el-Sheikh, and Ismailia.45
While these lines of best ﬁt do not include any
individual-level demographic controls,46
they suggest that governorates in which losers
were relatively stronger saw greater disillusionment with democracy.
To include individual-level covariates, we shift the unit of analysis to the individual
respondent and run multivariate linear regressions where the dependent variable is
Suitability of Democracy. Our coefﬁcient of interest is an interaction term between
Time (05pre-election; 15post-election) and Loser Strength. For ease of interpretation,
we recode loser vote share dichotomously, splitting it at its third quartile for each
country.47
For Tunisia, this means loser-strong governorates are coded as those in which
non-Islamists received more than 69 percent of the governorate’s vote share; for Egypt,
this means loser-strong governorates refer to those with more than 45 percent of the
governorate’s vote share. The interaction term thus reveals how much more democracy
scores fall for losers living in loser-strong areas than losers living in loser-weak areas
pre- to post-election. We hypothesize a negative interaction term: losers living in areas
where they are strong should see greater disillusionment with democracy.
Although demographic covariates are relatively balanced in loser-strong vs. loserweak
governorates (see Appendix, Figure A.5), we control for Age, Gender (Female),
Education, Income, Urban, Religion (Muslim), and Governorate.48
Ideally, we would
test our theory with panel data, examining the democratic attitudes of the same
individuals before and after losing the founding elections. However, the Arab
Barometer does not survey the same individuals in both waves. To address this
deﬁciency, we match our samples of pre- and post-election losers along the aforementioned
demographic covariates. We generate covariate balancing propensity scores49
and employ
a “nearest” matching algorithm that removes observations that do not have counterparts in
the other Barometer wave.50
The covariate balance plot (see Appendix, Figure A.6)
suggests that the matching algorithm helped create a similar set of losers pre- and postelection.
Finally, despite our small sample size, we also test the robustness of our theory
by clustering standard errors at the governorate level.
506
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Figure 1 Losers’ Disillusionment with Democracy by Governorate in Tunisia and
Egypt
507
Sharan Grewal and Steve L. Monroe
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table 1 presents our main results. Model 1 is the original data with missing data
excluded. Model 2 imputes the missing data. Model 3 adds the matching algorithm to
the imputed data. Finally, Model 4 employs clustered standard errors on the imputed
and matched data. The interaction term between Loser Strength and Time is negative
and statistically signiﬁcant at the .01 level for Models 1–3, and signiﬁcant at the .05
level for Model 4.
The dotplot below presents the results of Model 3. It plots the average predicted
democracy score for losers in governorates where they are strong and where they are
weak pre- and post-election, holding all other covariates at their means.51
While the
average democracy score for losers in governorates where they are weak drops by about
Table 1 Loser Strength and Disillusionment with Democracy (OLS)
DV: Suitability of Democracy (0-10)
(1) (2) (3) (4)
Time* –0.954***
–0.853***
–1.249***
–1.249**
Loser Strength (0.367) (0.298) (0.345) (0.572)
Time –1.303***
–1.296***
–1.526***
–1.526***
(0.175) (0.148) (0.177) (0.360)
Loser Strength 2.699**
2.811***
3.400***
3.400***
(1.085) (0.990) (1.017) (0.525)
Age 0.097* 0.141***
0.147***
0.147**
(0.057) (0.048) (0.055) (0.067)
Female –0.035 0.027 0.006 0.006
(0.148) (0.124) (0.144) (0.162)
Education –0.060 –0.023 –0.026 –0.026
(0.043) (0.037) (0.044) (0.066)
Income 0.104* 0.103**
0.056 0.056
(0.054) (0.048) (0.097) (0.108)
Urban 0.220 0.025 0.162 0.162
(0.197) (0.159) (0.191) (0.185)
Muslim –0.520 –0.349 –0.346 –0.346
(0.357) (0.287) (0.326) (0.438)
Governorate PE U U U U
Imputation U U U
Matching (Nearest) U U
Clustered SE U
Constant 7.214***
6.584***
7.702***
7.702***
(0.637) (0.554) (0.807) (0.680)
Observations 1,162 1,579 1,218 1,218
R2
0.172 0.167 0.204 0.204
Adjusted R2
0.132 0.138 0.168 0.168
Note: * p,0.1; ** p,0.0 5; *** p,0.01.
508
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
1.5 points (on the 0–10 scale) pre- to post-election, it falls roughly 2.7 points for losers
in governorates where they are strong.
While losers in governorates where they are weak still have objectively lower levels
of support for democracy, what is important is the size of the drop. This is the frustration
caused by unmet expectations. This disillusionment is what is thought—and what we
will show—leads to greater support for a return to authoritarian rule.
Robustness Checks and Alternative Explanations
We analyze Egypt and Tunisia separately to assess if one of the two countries may be
driving these results. While this halves our sample size, Table A.8 in the Appendix
shows that the interaction term remains negative and signiﬁcant in each country,
whether using the original or imputed data.
A respondent’s perception of whether or not their country is a democracy may
affect their expressed support for democracy. The Arab Barometer asks respondents to
rank on a 0–10 scale how democratic their country currently is. While this variable is
potentially post-treatment and thus we do not include it as a control in Table 1, results
are robust to its inclusion. Results also hold when subtracting this 0–10 evaluation from
the 0–10 support for democracy score and using this as the dependent variable (see
Appendix, Table A.9).
Some may wonder if our results hold for voters and non-voters. Consistent with
recent literature,52
results are stronger among losers who reported voting, perhaps
Figure 2 Loser Strength and Disillusionment with Democracy
509
Sharan Grewal and Steve L. Monroe
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
reﬂecting a larger personal stake or investment in the electoral outcome. Results are in
the right direction and in some models marginally signiﬁcant for non-voters as well (see
Appendix, Table A.10).53
There are three primary alternative explanations for our ﬁndings. First, perhaps
losers in loser-strong areas experienced more cases of electoral fraud and became more
disillusioned because of cheating, not losing. Electoral fairness at the local-level has
been shown to color system-level judgments.54
However, Table A.12 (Appendix)
demonstrates that results hold when controlling for actual electoral fraud (Models
1–2).55
Respondents in loser-strong governorates were no more likely to perceive fraud
post-election (Models 3–4).
Redistribution is another alternative explanation. Losers in winner-dominated
governorates may be less disillusioned because the winners were able to reallocate rents
back home. While the authors’ interviews with Brotherhood and Ennahda ofﬁcials
suggest that both countries’ police and bureaucracy largely refused to cooperate with the
winning Islamist parties,56
and thus impeded rent reallocation, we check for these
possibilities using the Arab Barometer data. We compare the attitudes of losers in loserstrong
and loser-weak governorates towards government performance in providing
security and a range of economic and social public goods (see Appendix, Tables
A.13–A.15). Losers in loser-weak governorates were no more likely to report an
improvement than losers in loser-strong governorates for each of these attitudes,
suggesting little reallocation of rents to Islamist strongholds. Differences in the state’s
perceived economic, security, or administrative performance cannot explain our
ﬁndings.
A ﬁnal counter-explanation could be that losers in loser-strong governorates were
ideologically different from losers in loser-weak governorates, perhaps more secular or
more revolutionary. Electoral losers with policy preferences farther from the winning
parties are likely to become more disillusioned.57
However, Table A.16 (Appendix)
shows that loser-strong governorates did not have higher levels of opposition to Egypt’s
and Tunisia’s ousted autocratic leaders, Hosni Mubarak and Zine El Abidine Ben Ali,
prior to the ﬁrst elections. Table A.17 (Appendix) demonstrates that results hold when
controlling for level of trust in the winning Islamist party and support for shari’a. The
underlying political identities or policy preferences of these losers therefore cannot
explain the variation in their disillusionment with democracy.
Mechanisms and Testable Implications
We now validate our proposed mechanism: expectations of winning. If strength at the
local level conditions one’s expectations for winning at the national level, we predict
that:
1. Losers in loser-strong governorates should be more supportive of democracy preelection,
as they expect to win nationally.
510
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
2. The differential effect of local strength on disillusionment with democracy should
be larger for those losers who are politically unaware.
3. Winners in winner-weak governorates should see a greater increase in support for
democracy than winners in winner-strong governorates.
Table 1 conﬁrms the ﬁrst implication. The coefﬁcient for Loser Strength is positive
and signiﬁcant at the .05 or .01 level in all models. Losers in governorates where they
are strong have a higher baseline of support for democracy. Another test of this
implication is that if losers in loser-strong governorates are more likely to think they will
win nationally, they should also be more optimistic about the election and, therefore,
more likely to report that they plan to vote. Table A.3 in the Appendix supports this
proposition: losers in governorates where they are strong are signiﬁcantly more likely to
say they plan to vote pre-election.
Table A.4 in the Appendix tests the second implication. If disillusionment is due to
a shattering of locally-generated expectations, then this disillusionment should be less
pronounced among losers who have access to national information: losers who regularly
watch the national news or are active on the internet/social media. Models 1–2 (and
Figure A.7) show that the demoralizing effect of living in loser-strong governorates was
concentrated among losers who do not watch the national news. Losers who self-report
following the national news “to a great extent” were no more disillusioned in loserstrong
than loser-weak governorates. Similarly, Models 3–4 show that losers who are
rarely online became more disillusioned in loser-strong governorates than loser-weak
governorates. However, daily internet users, who are more likely to access national
news, were no more disillusioned in loser-strong or weak governorates. This differential
effect by news and internet consumption suggests that those with purely local, ﬁrsthand
information may have been more shocked by the electoral results.
The third testable implication is a similar expectations gap among winners. Winners
in governorates where they are weak should be uncertain about their likelihood of
national victory ahead of the elections. After being surprised with a victory, they should
become more supportive of democracy. Winners in governorates where they are strong,
by contrast, should already expect to win and thus should experience less change in
their support for democracy before to after the elections. Splitting winner vote share at
its median, the interaction term between Time and Winner Weak should therefore be
positive.
Table A.5 in the Appendix supports this implication. The ﬁrst two models show
the pooled Egypt and Tunisia data (original then imputed), coding winners as the
opposite of losers, either above the median in trust in the winning Islamist parties or
above the median in support for shari‘a. In both models, the coefﬁcient on the
interaction term is positive and signiﬁcant at the .05 level. Winners residing in
governorates where they were weak had a stronger “expectations bump” than winners
where they were strong.
In Egypt, there were two winning Islamist parties: the Muslim Brotherhood (with
37 percent of the vote) and the Salaﬁ Nour party (with 25 percent). In Models 3–6 of
511
Sharan Grewal and Steve L. Monroe
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table 4, we demonstrate that this positive expectations bump holds for each of these
parties separately, using each party’s vote share to code winner-weak governorates. To
capture Muslim Brotherhood respondents, we take those above the median in trust in the
Muslim Brotherhood. To capture Salaﬁ supporters, we take those above the median in
support for shari‘a, but with below the median trust in the Muslim Brotherhood. Across
the original or imputed data, the interaction term is positive and signiﬁcant at the .1 level
for the Muslim Brotherhood in Model 3 and signiﬁcant at the .01 level for Salaﬁs. The
expectations bump is much larger for the Salaﬁs, for whom winning came as more of a
surprise.
These results offer additional evidence against the counter-explanation of rent
reallocation. If winners had reallocated rents back to their strongholds, then winners in
winner-strong governorates would have become more supportive of democracy. By
contrast, it occurred in the winner-weak governorates, implying that it is instead about
expectations of winning.
Disillusionment and Support for Authoritarianism
Some may counter that far from inciting democratic defection, the disillusionment and
frustrations of unmet electoral expectations could breed political apathy. We address
this skepticism in two ways.
First, we use Arab Barometer data to show that disillusioned losers actually became
more supportive of authoritarianism. The Arab Barometer asks respondents how they
would rate “a political system with an authoritarian president (non-democratic) who is
indifferent to parliament and elections.” We rescale this variable from 1 (very bad) to 4
(very good) and then conduct the same analyses as in Table 1 with this dependent
variable.
Table A.6 in the Appendix presents our results. In all but the last model, the
coefﬁcient between time and loser strength is positive and signiﬁcant, suggesting that
losers in governorates where they were strong became more supportive of authoritarian
rule. The coefﬁcient on loser strength is also insigniﬁcant: losers in governorates where
they were strong were no more supportive of authoritarian rule pre-election, just as they
were no more supportive of former autocrats Mubarak or Ben Ali pre-election. The
founding election, however, disillusioned losers in loser-strong governorates and fueled
support for authoritarianism. This rise in support for authoritarianism indicates that
losers’ democratic disillusionment did not induce political apathy but democratic
hostility.
Second, we use voting data to investigate whether these attitudes in fact correlated
with acts of democratic subversion. We test whether disillusioned losers became more
likely to vote for candidates representing the ousted autocratic regime. As the return to
power of the former autocratic regime is one of the primary means by which democratic
transitions fail,58
this is a crucial pathway to study.
512
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Egypt provides a unique case to test this possibility. The country held presidential
elections in May-June 2012 soon after the 2011–2012 parliamentary elections (the
founding elections we have focused on thus far). The presidential elections pitted the
Muslim Brotherhood candidate Mohamed Morsi against a candidate representing the
former regime, retired General Ahmed Shaﬁk. A former Prime Minister, Shaﬁk was
openly labeled as a feloul (one of the remnants of the former regime) and was nearly
disqualiﬁed from the elections by a lustration law.
Shaﬁk campaigned on a platform to “restore the greatness of the old Egypt,”59
pledging to restore order even at the cost of democracy. “Security is one of the top needs
for the Egyptian people,” he claimed. “Democracy has limits. Unrest is not
democracy.”60
In the run-off presidential election, the Muslim Brotherhood’s Mohamed
Morsi narrowly defeated this autocratic challenger with 51 percent of the vote.
In the ﬁrst round of the presidential elections, however, the non-Islamist vote had
been split between Shaﬁk and several more “revolutionary” candidates. The proportion
of the non-Islamist (non-Morsi)61
vote in this ﬁrst round going to Shaﬁk varied widely
by governorate, from 9 percent in Kafr el-Sheikh to 68 percent in Monuﬁa. We
hypothesize that loser-strong governorates, which we found earlier had greater
disillusionment among their losers, will have a higher proportion of losers’ votes
going toward Shaﬁk.
The ﬁrst graph in Figure 3 plots each Egyptian governorate with the proportion
of losers’ votes going to Shaﬁk on the y-axis, and on the x-axis the same loser vote
share in the 2011–2012 parliamentary elections as before. As expected, loser-strong
governorates, which had become more supportive of authoritarianism, also had a
higher proportion of losers’ votes going toward the autocratic candidate. Recoding
loser strength dichotomously at its median, loser-strong governorates have on average
a 15 percent higher proportion of votes going toward Shaﬁk. Model 1 in Table A.7 in
the Appendix ﬁnds that this relationship is signiﬁcant at the .05 level.
One may counter that perhaps these loser-strong governorates had already been
more in favor of autocratic candidates prior to the disillusionment of the founding
elections. To test this possibility, we examine whether loser-strong governorates were
more likely to vote for autocratic parties in the founding 2011–2012 parliamentary
elections. While technically banned from competing, several parties afﬁliated with
remnants of the old regime ran in the elections.62
In total, these remnant parties received
some 8.4 percent of the vote (or 25 percent of the non-Islamist votes). But as Model 2 of
Table A.7 (Appendix) demonstrates, loser-strong governorates were no more likely to
vote for these parties than loser-weak governorates. This suggests that losers in each
type of governorate had equal levels of support for autocratic parties before the
founding elections.
From this baseline of autocratic vote share in the founding elections, we probe
whether support for autocratic candidates increased more in loser-strong governorates
than loser-weak ones. Model 3 of Table 7 reveals that loser-strong governorates saw a
statistically signiﬁcant increase (at the .05 level) in autocratic vote share from the
founding elections to the 2012 presidential elections than loser-weak governorates. The
513
Sharan Grewal and Steve L. Monroe
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Figure 3 Loser Strength and Autocratic Vote Share in Egypt’s 2012 Presidential
Election (First Round)
514
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
second graph in Figure 3 presents these results. While votes going to autocratic
candidates increased by just 7 percentage points on average, some loser-strong
governorates like Monuﬁa and El Sharqiya saw increases of more than 30 percentage
points.
In sum, losers in governorates where they were strong appear to have been more
likely to vote for candidates representing the former autocratic regime. While this
analysis suffers from ecological inference, it is consistent with the individual-level Arab
Barometer survey data suggesting that losers in governorates where they were strong
became more supportive of authoritarianism. Although each analysis may have its
ﬂaws, taken together, the ﬁndings provide consistent evidence from two independent
data sources that disillusioned losers turned not to apathy but to rejection of the
democratic transition.
Conclusion
In new democracies, electoral losers residing in areas where they are strong may
become more disillusioned with democracy than losers in areas where they are weak.
We suggest that expectations drive this effect: ahead of the founding elections, losers in
loser-strong governorates tend to believe they are also strong nationally and are thus
initially supportive of democracy. They then become disillusioned after elections reveal
their true national popularity. Losers in governorates where they are weak, however, are
more likely to already believe they are weak nationally. The elections simply conﬁrm
their expectations.
This ﬁnding has important implications for the consolidation of democratic
transitions. Disillusioned losers have the potential to spoil transitions. We provide subnational
predictions on which losers are likely to initially join the democratic coalition
but then defect (those in loser-strong governorates) and which losers are likely to
oppose democracy from the start (those in loser-weak governorates).
These subnational ﬁndings offer one piece of a broader investigation on the
collapse and consolidation of democratic transitions. A number of other factors—the
national level of disillusionment, losers’ ability to ﬁnd like-minded allies in the military,
etc.—affected why one country in our analysis succumbed to a military coup (Egypt)
while the other remains on the path to consolidation (Tunisia). Yet, despite these crossnational
differences, our analysis uncovers subnational similarities regarding who is
likely to play spoiler in a democratic transition.
These ﬁndings imply that transitional governments ought to make extra effort to
include losers in loser-strong areas within the democratic system. Although losers in
loser-strong governorates may receive representation in parliament, they will remain a
minority in national-level institutions. Decentralizing power, however, especially
through the election of governors and local councils, can empower losing parties
locally. In both Egypt and Tunisia, governors were appointed, and there were no
municipal elections during either transition.63
Since losers in loser-strong areas may win
515
Sharan Grewal and Steve L. Monroe
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
at the local level, decentralization would give these otherwise disillusioned losers some
modicum of power in governorates where they are strong. This buy-in into democracy
at the local level may make electoral losers less likely to oppose the system and subvert
nascent democratic institutions.
NOTES
The authors wish to thank three anonymous reviewers, Elizabeth Baisley, Daniela Barba-Sanchez, Winston
Chou, Daniel Corstange, Kosuke Imai, Amaney Jamal, Hubert Jin, James Lo, Tarek Masoud, Michael
Robbins, Scott Williamson, and audiences at Stanford, Princeton, the Project on Middle East Political Science,
and the 2015 American Political Science Association annual conference.
1. Shadi Hamid, Islamic Exceptionalism: How the Struggle Over Islam is Reshaping the World (New
York: St. Martin’s Press, 2016), 252–53.
2. Adam Przeworski, Democracy and the Market: Political and Economic Reforms In Eastern Europe
and Latin America (Cambridge: Cambridge University Press, 1991); Juan Linz and Alfred Stepan, Problem of
Democratic Transition and Consolidation: Southern Europe, South America, and Post-Communist Europe
(Baltimore: Johns Hopkins University Press, 1996); Christopher Anderson, Andre Blais, Shaun Bowler, Todd
Donovan, and Ola Listhaug Losers’ Consent: Elections and Democratic Legitimacy (Oxford: Oxford University
Press, 2005).
3. Important exceptions include Jeffrey Conroy-Krutz and Nicholas Kerr, “Dynamics of Democratic
Satisfaction in Transitional Settings: Evidence from a Panel Study in Uganda,” Political Research Quarterly,
68 (September 2015), 593–606; Luigi Curini and Willy Jou “The Conditional Impact Of Winner/Loser Status
And Ideological Proximity On Citizen Participation,” European Journal of Political Research, 55 (November
2016), 767–88; and Andrew Daniller, “Can Citizens Care Too Much? Investment in Election Outcomes and
Perceptions of Electoral Legitimacy,” Electoral Studies, 44 (December 2016), 151–61.
4. Przeworski, 1991; Linz and Stepan, 1996; Anderson et al., 2005.
5. Gabriel Almond and Sidney Verba, The Civic Culture: Political Attitudes and Democracy in Five
Nations (Princeton: Princeton University Press, 1963); Larry Diamond, Developing Democracy: Toward
Consolidation (Baltimore: John Hopkins University Press, 1999).
6. Ronald Inglehart and Christian Welzel, Modernization, Cultural Change, and Democracy (Cambridge:
Cambridge University Press, 2005); Pippa Norris, Democratic Deﬁcit: Critical Citizens Revisited (Cambridge:
Cambridge University Press, 2011).
7. James Gibson, “A Mile Wide But An Inch Deep (?): The Structure of Democratic Commitments in the
Former USSR,” American Journal of Political Science, 40 (May 1996), 396–420.
8. Damarys Canache, “From Bullets to Ballots: The Emergence of Popular Support for Hugo Chavez,”
Latin American Politics and Society, 44 (Spring 2002), 69–90.
9. Wolfgang Merkel, “The Consolidation of Post-Autocratic Democracies: A Multi-Level Model,” Democratization,
5 (Autumn 1998), 33–67; Diamond, 1999.
10. Christopher J. Anderson and Yuliya V. Tverdova, “Winners, Losers, and Attitudes about Government
in Contemporary Democracies,” International Political Science Review, 22 (October 2001), 321–38; Christopher
J. Anderson and Silvia Mendes, “Learning to Lose: Election Outcomes, Democratic Experience and
Political Protest Potential,” British Journal of Political Science, 36 (January 2006), 91–111; Shane Singh,
Ignacio Lago, and Andre Blais, “Winning and Competitiveness as Determinants of Political Support,” Social
Science Quarterly, 92 (September 2011), 695–709; Shane P. Singh, Ekrem Karakoc, and Andre Blais,
“Differentiating Winners: How Elections Affect Satisfaction with Democracy,” Electoral Studies, 31 (March
2012), 201–11; Andre Blais, Alexandre Morin-Chasse, and Shane P. Singh, “Election Outcomes, Legislative
Representation, and Satisfaction with Democracy,” Party Politics, 23 (March 2017), 85–95.
11. Anderson et al., 2005; Anderson and Mendes, 2006; Luigi Curini, Willi Jou, and Vincenzo Memoli,
“Satisfaction with Democracy and the Winner/Loser Debate,” British Journal of Political Science, 42 (April
2012), 241–61. Stefan Dahlberg and Jonas Linde, “The Dynamics of the Winner-Loser Gap in Satisfaction
with Democracy: Evidence from a Swedish Citizen Panel,” International Political Science Review, 38 (November
2017), 625–41.
516
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
12. Shaun Bowler and Todd Donovan, “Reasoning about Institutional Change: Winners, Losers and
Support for Electoral Reforms,” British Journal of Political Science, 37 (July 2007), 455–76.
13. Anderson and Mendes, 2006; Curini and Jou, 2016.
14. Alfred Stepan and Cindy Skach, “Constitutional Frameworks and Democratic Consolidation: Parliamentarianism
versus Presidentialism,” World Politics, 46 (October 1993), 1–22; Juan Linz and Arturo
Valenzuela, The Failure of Presidential Democracy (Baltimore: Johns Hopkins University Press, 1994).
15. Przeworski, 1991.
16. See, e.g., Carles Boix, Democracy and Redistribution (Cambridge: Cambridge University Press,
2003); Daron Acemoglu and James Robinson, Economic Origins of Democracy and Dictatorship (Cambridge:
Cambridge University Press, 2006).
17. Some losers may still support democracy for its guarantee of rights and freedoms. These individuals
should not become disillusioned with democracy pre- to post-election, and thus, while important, are tangential
to our question.
18. Ted Gurr, Why Men Rebel (Princeton: Princeton University Press, 1970).
19. Morty Bernstein and Faye Crosby, “An Empirical Examination of Relative Deprivation Theory,”
Journal of Experimental Social Psychology, 16 (September 1980), 442–56; Heather Smith, Thomas Pettigrew,
Gina Pippin, and Silvana Bialosiewicz, “Relative Deprivation: A Theoretical and Meta-Analytic Review,”
Personality and Social Psychology Review, 16 (August 2012), 203–32.
20. Andre Blais and Marc Bodet, “How Do Voters Form Expectations about the Parties’ Chances of
Winning the Election?” Social Science Quarterly, 87 (September 2006), 477–93; Michael Miller, Guanchun
Wang, Sanjeev Kulkarni, H. Vincent Poor, and Daniel Osheron, “Citizen Forecasts of the 2008 U.S. Presidential
Election,” Politics and Policy, 40 (December 2012), 1019–52; Thomas Zerback, Carsten Reinemann,
and Angela Nienierza, “Who’s Hot and Who’s Not? Factors Inﬂuencing Public Perceptions of Current Party
Popularity and Electoral Expectations,” International Journal of Press/Politics, 20 (October 2015), 458–77.
21. Guillermo O’Donnell, Philippe Schmitter, and Laurence Whitehead, Transitions from Authoritarian
Rule: Prospects for Democracy (Baltimore: John Hopkins University Press, 1986), 61.
22. Miller et al., 2012.
23. Kanchan Chandra, Why Ethnic Parties Succeed (Cambridge: Cambridge University Press, 2007).
24. Despite advocating for a reduction in the role of religion in politics, many of these losing parties would
not use the term “secular” themselves. We therefore prefer to use the term “non-Islamists.”
25. Shadi Hamid, Temptations of Power (Oxford: Oxford University Press, 2014); Chantal Berman and
Elizabeth Nugent, “Deﬁning Political Choices: Tunisia’s Second Democratic Elections from the Ground Up,”
Brookings Institution Analysis Paper, 38 (May 2015), 1–28; Jason Brownlee, Tarek Masoud, and Andrew
Reynolds, The Arab Spring: Pathways of Repression and Reform (Oxford: Oxford University Press, 2015).
26. Kenneth Perkins, A History of Modern Tunisia (Cambridge: Cambridge University Press, 2014);
Thanassis Cambanis, Once Upon a Revolution: An Egyptian Story (New York: Simon and Schuster, 2015).
27. “Egyptian youth ‘decimated’ in parliamentary elections says Mohamed El Baradei,” The Telegraph,
Dec. 5, 2011.
28. Mahmoud Salem, “Underneath,” December 20, 2011, http://www.sandmonkey.org/2011/12/20/
underneath/.
29. Marc Fisher, “In Tunisia after the Arab Spring, Islamists’ New Freedoms Create New Muslim Divide,”
Washington Post, Apr. 28, 2012.
30. See Elizabeth Dickinson, “Secularists Voice Dismay at Election Gains for Tunisia’s Islamist Party,”
Christian Science Monitor, Oct. 25, 2011.
31. Due to space constraints, the Appendix is not in the print version of this article. It can be viewed in the
online version, at www.ingentaconnect.com/cuny/cp.
32. Cambanis, 2015.
33. Danish-Egyptian Dialogue Institute (DEDI), “3rd National Voter Survey in Egypt,” November 12,
2011.
34. Dahlberg and Linde, 2017.
35. Using membership in the winning and losing parties is also not feasible, as party membership was
lower than 4 percent in both the pre and post-election surveys.
36. David Easton, “A Re-Assessment of the Concept of Political Support,” British Journal of Political
Science, 5 (October 1975), 449.
37. Given that there were tens of secular parties, the Arab Barometer chose to ask for trust in the winning
Islamist party rather than each secular party.
517
Sharan Grewal and Steve L. Monroe
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
38. We average the following four Arab Barometer questions to create an index for support for shari‘a:
whether shari‘a should be the basis of law, the basis of penal law, the basis of personal status law, and the basis
of inheritance law.
39. Hamid, 2014; Brownlee et al., 2015.
40. Ekrem Karakoc, Talha K¨ose, and Mesut ¨Ozcan, “Emigration and the Diffusion of Political Salaﬁsm,”
Party Politics, 23 (November 2017), 731–45; Ege Ozen, “Egypt’s 2011–2012 parliamentary elections,”
Mediterranean Politics (Firstview 2017), 1–26.
41. Tarek Masoud, Counting Islam (Cambridge: Cambridge University Press, 2014).
42. Gary King, James Honaker, Anne Joseph, and Kenneth Scheve, “Analyzing Incomplete Political
Science Data: An Alternative Algorithm for Multiple Imputation,” American Political Science Review, 95
(March 2001), 49–69.
43. Renske Doorenspleet, “Critical Citizens, Democratic Support and Satisfaction in African Democracies,”
International Political Science Review, 33 (June 2012), 279–300; Norris, 2011; Conroy-Krutz and
Kerr, 2015.
44. Results hold when removing Sidi Bouzid as an outlier.
45. See, e.g., Atef al-Saadawy, “What the Brotherhood Lost,” Carnegie Endowment for International
Peace, http://carnegieendowment.org/sada/48832. The Brotherhood was founded in Ismailia by Hassan alBanna,
who hailed from Beheira.
46. Lines of best ﬁt contain subregion ﬁxed effects, but no individual demographic covariates.
47. Results hold when coding it at the median or when leaving the variable continuous.
48. Results hold without governorate ﬁxed effects (see appendix, Table A.11), or with random effects for
governorate (available from authors).
49. Kosuke Imai and Marc Ratkovic, “Covariate Balancing Propensity Score,” Journal of the Royal
Statistical Society, 76 (January 2014), 243–63.
50. Results are robust to alternate matching techniques including full and genetic matching.
51. As we cannot average across the governorate ﬁxed effects, this particular ﬁgure presents the predicted
probabilities for secularists in Cairo (loser strong) v. Minya (loser weak).
52. Conroy-Krutz and Kerr, 2015; Timothy Rich, “Losers Consent or Non-Voter Consent? Satisfaction
with Democracy in East Asia,” Asian Journal of Political Science, 23 (August 2015), 243–59.
53. To capture voters, we subsetted the data to those who intend to vote in the pre-election survey and
those who report having voted in the post-election survey.
54. Jonathan Hiskey and Shaun Bowler, “Local Context and Democratization in Mexico,” American
Journal of Political Science 49 (January 2005), 57–71.
55. In Tunisia, this refers to seats won by Aridha Chaabia (the Popular Petition) that the electoral commission
nulliﬁed on account of fraud (5 out of 24 governorates). In Egypt, this refers to districts with sufﬁcient
electoral irregularities to warrant a re-run of their election in later phases of the parliamentary elections (8 of 21
governorates).
56. Interviews with Muslim Brotherhood minister Amr Darrag, Istanbul, July 11, 2016; Ennahda Prime
Minister Hamadi Jebali, Sousse, December 17, 2015.
57. Singh et al., 2012; Curini and Jou, 2016; Daniller, 2016.
58. Milan Svolik, “Which Democracies Will Last? Coups, Incumbent Takeovers, and the Dynamic of
Democratic Consolidation,” British Journal of Political Science, 45 (October 2015), 715–38.
59. Liam Stack and David Kirkpatrick, “Dark-Horse Candidates Add to Egypt’s Suspense,” The New York
Times, May 19, 2012.
60. “Where they stand—Egyptian candidates Shaﬁq and Mursi,” BBC News, Jun. 6, 2012.
61. Results hold when calculating the non-Islamist vote as non-Morsi and non-Aboul-Futouh. AboulFutouh
was endorsed by the leadership of the Salaﬁ Nour party, but very few of its base actually turned out for
him, seeing him as too liberal.
62. These parties include the Freedom, National of Egypt, Democratic Peace, Egyptian Citizen, Union,
New Independents, Modern Egypt, Conservative, and Reform and Development parties.
63. Tunisia plans to hold municipal elections in May 2018.
518
Comparative Politics July 2019
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
APPENDIX
1. Timeline of Arab Barometer Surveys
2. Arab Barometer Questions
Sample: Electoral Losers
c [Trust] “I will name a number of institutions, and I would like you to tell me to
what extent you trust each of them” (to a great extent, to a medium extent, to a
limited extent, absolutely do not trust)
s The Muslim Brotherhood (Egypt only)
s Al-Nahda Movement/Party (Tunisia only)
c [Shari’a Index] “To what extent do you agree or disagree with each of the
following principles in the enactment of your country’s laws and regulations?”
(strongly agree to strongly disagree)
s “The government and parliament should enact laws in accordance with Islamic law.”
s “The government and parliament should enact penal laws in accordance with
Islamic law.”
s “The government and parliament should enact personal status laws (marriage,
divorce) in accordance with Islamic law.”
s “The government and parliament should enact inheritance laws in accordance
with Islamic law.
Dependent Variables
c [Suitability of Democracy] “Suppose there was a scale from 1-10 measuring the
extent to which democracy is suitable for your country, with 1 meaning that
democracy is absolutely inappropriate for your country and 10 meaning that democracy
is completely appropriate for your country. To what extent do you think
democracy is appropriate for your country?”
Figure A.4 Timeline
1
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
c [Support for Authoritarianism] “I will describe different political systems to you,
and I want to ask you about your opinion of each one of them with regard to the
country’s governance – for each one would you say it is very good, good, bad, or
very bad? [...] A political system with an authoritarian president (non-democratic)
who is indifferent to parliament and elections.”
Robustness checks
c [Perception of Fraud] “In general, how would you evaluate the last parliamentary
elections that were held on (date of the last elections)? [Completely free and fair,
minor breaches, major breaches, not free and fair].
c [Curr Econ] “How would you evaluate the current economic situation in your
country?”
c [Curr Safety] “Do you currently feel that your own personal as well as your
family’s safety and security are ensured or not?”
c [Retro Safety] “Compared to this time last year, do you feel that your own
personal and your family’s safety and security are now. . . (better, same, worse)”
c “I am going to ask a number of questions related to the current government’s
performance. How would you evaluate the performance of the current
government’s [Insert]?”
s [Jobs] “Creating employment opportunities”
s [Narrow] “Narrowing the gap between rich and poor”
s [Manage] “Managing the process of democratic transition process”
c “I will ask some questions about your ability to obtain certain services. Based on
your actual experience, how difﬁcult or easy is it to obtain...
s [Health] “Appropriate medical treatment in a nearby clinic or public
(government) hospital.”
s [Police] “Assistance from public security (the police) when needed.”
s [Redress] “Access to the concerned ofﬁcial to ﬁle a complaint when you feel that
your rights have been violated.”
c [Needs] I will read you some statements related to your household income. Which
of these statements comes closest to describing your household income?
s Our household income covers our expenses well and we are able to save.
s Our household income covers our expenses without notable difﬁculties.
s Our household income does not cover our expenses and we face some difﬁculties
in meeting our needs.
s Our household income does not cover our expenses and we face signiﬁcant
difﬁculties in meeting our needs.”
Voters v. Non-voters
c [Pre-Election, Egypt] “Will you vote in the next parliamentary elections?”
c [Pre-Election, Tunisia] “Will you participate in the upcoming elections of the
Constituent Assembly (October 23rd, 2011)?”
2
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
c [Post-Election] “Did you vote in the last parliamentary elections that were held on
(date of the last elections)”
News and Internet
c “To what extent do you follow political news in your country?" (to a great extent,
to a medium extent, to a limited extent, I don’t follow political news at all).
c “Do you use the internet?” (Daily or almost daily; at least once a week; at least once
a month; a few times a year; I do not use the Internet).
Satisfaction with Former Regime
c [Egypt] “Suppose that there was a scale from 1-10 to measure the extent of your
satisfaction with the Mubarak regime, in which 1 means that you were absolutely
unsatisﬁed with the regime and 10 means that you were very satisﬁed, to what
extent were you satisﬁed with the Mubarak regime? (Egypt only)”
c [Tunisia] “Suppose that there was a scale from 1-10 to measure the extent of your
satisfaction with the Ben Ali regime, in which 1 means that you were absolutely
unsatisﬁed with the regime and 10 means that you were very satisﬁed, to what
extent were you satisﬁed with the Ben Ali regime? (Tunisia only)”
Questions used for Multiple Imputation
c [Individual Piety] Do you...? (always, most of the time, sometimes, rarely, never)
s Pray daily.
s Attend Friday prayer/Sunday services.
s Listen to or read the Quran/the Bible.
c [Religion in Public Life] To what extent do you agree or disagree with the
following statements? (strongly agree to strongly disagree):
s Religious leaders (imams, preachers, priests) should not interfere in voters’
decisions in elections.
s Your country is better off if religious people hold public positions in the state.
s Religious leaders (imams, preachers, priests) should have inﬂuence over
government decisions.
s Religious practices are private and should be separated from social and political
life.
s Religious associations and institutions (excluding political parties) should not
inﬂuence voters’ decisions in elections.
s Mosques and churches should not be used for election campaigning.
• Arrange the following characteristics in order of their importance for a person to be
qualiﬁed for political leadership in the country. [...] Piety.
3
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Demographic Controls
c Age
c Gender:
c Monthly household income in 2011 US Dollars
c Level of education:
s Illiterate/Literate
s Elementary.
s Preparatory/Basic.
s Secondary.
s Mid-level diploma/professional or technical (not included in Tunisia)
s BA
s MA and above
c Urban/Rural
c Religion
s Muslim
s Christian
s Other
4
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
3. Predicting Vote Choice
Trust in Islamists and support for shari’a are strong predictors of intended vote choice:
Table A.2 Effect of Trust and Shari`a on Vote Choice (OLS)
DV: Vote Choice (15Islamist, 05Non-Islamist)
Tunisia Egypt
(1) (2) (3) (4)
Trust in Ennahda/ 0.192***
0.193***
0.133***
0.122***
Muslim Brotherhood (0.019) (0.021) (0.034) (0.037)
Support for Shari‘a 0.163***
0.164***
0.178***
0.156*
(0.042) (0.048) (0.059) (0.083)
Age 20.022 0.006
(0.022) (0.046)
Female 20.010 20.103
(0.048) (0.110)
Education 20.120 20.152
(0.109) (0.206)
Income 0.132 0.833
(0.564) (1.046)
Urban 20.054 0.052
(0.056) (0.106)
Muslim 20.004 0.039
(0.423) (0.256)
Constant 20.423***
20.317 20.582***
20.802*
(0.114) (0.516) (0.216) (0.475)
Observations 375 321 103 95
R2
0.309 0.317 0.201 0.218
Adjusted R2
0.306 0.300 0.185 0.145
Note: * p,0.1; ** p,0.05; *** p,0.01.
5
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
4. Covariate Balance
Figure A.5 Covariate Balance in Loser-Strong v. Loser-Weak Governorates (Pre-
Election)
Figure A.6 Covariate Balancing among Pre- and Post-Election Losers (Nearest
Matching)
6
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
5. Mechanism Tests
Table A.3 Loser Strength and Likelihood of Voting Pre-Election (logit)
DV: Will you vote?
Pre-Election
(1) (2)
Loser Strength 0.955* 0.682*
(0.492) (0.359)
Age 0.226**
0.237***
(0.110) (0.085)
Female –1.066***
–0.901 **
(0.281) (0.215)
Education 0.156**
0.166**
(0.077) (0.065)
Income 0.126* 0.138**
(0.065) (0.056)
Muslim 0.479 0.214
(0.460) (0.420)
Urban 0.550* 0.021
(0.311) (0.246)
Subregion FE U U
Imputation U
Constant –0.952 –0.713
(1.381) (0.891)
Observations 528 779
Log Likelihood –198.461 –319.804
Akaike Inf. Crit. 434.922 677.607
Note: * p,0.1; ** p,0.05; *** p,0.01.
7
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table A.4 The Moderating Effect of News and Internet
DV: Suitability of Democracy (0-10)
(News) (No News) (Internet) (No Internet)
Time* 0.347 20.980***
20.834 20.783**
Loser Strength (0.720) (0.342) (0.804) (0.326)
Time 21.001***
21.408***
20.281 21.561***
(0.353) (0.166) (0.379) (0.164)
Loser Strength 0.178 3.704***
4.825* 2.507**
(1.664) (1.257) (2.634) (1.063)
Age 0.111 0.114**
0.089 0.148***
(0.109) (0.055) (0.154) (0.054)
Female 0.139 0.068 20.010 0.100
(0.292) (0.138) (0.321) (0.137)
Education 20.209**
0.018 20.002 20.027
(0.082) (0.043) (0.120) (0.042)
Income 0.232**
0.077 20.097 0.151**
(0.109) (0.054) (0.128) (0.052)
Urban 20.147 0.005 20.471 0.064
(0.382) (0.176) (0.458) (0.175)
Muslim 21.078 20.405 20.623 20.320
(0.732) (0.313) (0.790) (0.310)
Governorate FE U U U U
Imputation U U U U
Constant 8.053***
6.336***
7.994***
6.306***
(1.248) (0.633) (1.442) (0.609)
Observations 387 1,192 289 1,290
R2
0.225 0.206 0.252 0.195
Adjusted R2
0.107 0.169 0.095 0.161
Note: * p,0.1; ** p,0.05; *** p,0.01.
Figure A.7 The Moderating Effect of News
8
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table A.5 Winner Strength and Winners’ Support for Democracy (OLS)
DV: Support for Democracy (0-10)
Egypt and Tunisia
Muslim Brotherhood
(Egypt) Salaﬁs (Egypt)
(1) (2) (3) (4) (5) (6)
Time* 0.488**
0.392**
0.661* 0.498 1.433***
1.400***
Winner Weak (0.199) (0.193) (0.379) (0.349) (0.501) (0.501)
Time 21.825***
21.771***
21.776***
21.839***
23.035***
23.127***
(0.141) (0.140) (0.283) (0.260) (0.352) (0.351)
Winner Weak 20.884 20.665 1.404**
1.106**
20.299 21.518
(0.612) (0.643) (0.565) (0.526) (1.958) (1.685)
Age 0.139***
0.125***
0.064 0.071 20.031 20.037
(0.040) (0.039) (0.069) (0.065) (0.099) (0.099)
Female 0.056 0.063 20.405**
20.382**
0.011 0.015
(0.099) (0.096) (0.171) (0.159) (0.238) (0.237)
Education 0.048 0.015 20.037 20.032 20.081 20.070
(0.031) (0.030) (0.047) (0.044) (0.073) (0.072)
Income 0.066* 0.056 0.167***
0.182***
0.113 0.084
(0.037) (0.035) (0.050) (0.046) (0.081) (0.080)
Urban 20.123 20.043 0.263 0.122 0.073 0.008
(0.124) (0.120) (0.212) (0.200) (0.300) (0.298)
Muslim 0.385 20.172 20.045 20.076 24.567* 0.359
(0.342) (0.389) (0.422) (0.391) (2.649) (1.579)
Governorate FE U U U U U U
Imputation U U U
Constant 6.390***
7.140***
5.497***
5.587***
10.970***
6.294***
(0.556) (0.575) (0.742) (0.685) (2.737) (1.814)
Observations 2,614 2,768 749 851 487 497
R2
0.163 0.159 0.222 0.213 0.330 0.320
Adjusted R2
0.145 0.142 0.189 0.185 0.287 0.277
Note: * p,0.1; ** p,0.05; *** p,0.01.
9
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
6. Authoritarianism Results
Table A.6 Loser Strength and Support for Authoritarianism (OLS)
DV: Support for Authoritarianism (1-4)
(1) (2) (3) (4)
Time* 0.227**
0.226***
0.138* 0.138
Loser Strength (0.105) (0.083) (0.081) (0.146)
Time 20.206***
20.179***
20.159***
20.159
(0.050) (0.042) (0.039) (0.097)
Loser Strength 20.027 20.118 0.109 0.109
(0.313) (0.280) (0.275) (0.124)
Age 0.026 0.020 0.018 0.018*
(0.016) (0.013) (0.013) (0.010)
Education 20.015 20.018* 20.017 20.017
(0.012) (0.010) (0.011) (0.013)
Income 0.022 0.024* 20.028 20.028
(0.016) (0.014) (0.023) (0.032)
Female 0.044 0.030 0.057 0.057
(0.042) (0.035) (0.035) (0.042)
Urban 20.041 20.060 20.006 20.006
(0.052) (0.044) (0.046) (0.052)
Muslim 20.051 20.030 0.010 0.010
(0.102) (0.082) (0.085) (0.150)
Governorate FE U U U U
Imputation U U U
Matching (Full) U U
Clustered SE U
Constant 1.549***
1.527***
1.669***
1.669***
(0.182) (0.156) (0.189) (0.248)
Observations 1,193 1,611 1,611 1,611
R2
0.136 0.112 0.104 0.101
Adjusted R2
0.096 0.082 0.074 0.074
Note: * p,0.1; ** p,0.05; *** p,0.01.
10
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
7. Robustness Checks
These robustness checks include the following:
Table A.8 demonstrates that the main results hold when analyzing Egypt and Tunisia
separately.
Table A.9 demonstrates that the main results remain when controlling for one’s
evaluation of how democratic their country is, and when subtracting this evaluation
from their suitability of democracy and using that as the dependent variable. Table A.10
shows that effects hold when analyzing voting and non-voting losers separately. Table
A.11 reveals that results hold without governorate ﬁxed effects.
To counter alternative explanations, Table A.12 shows that results hold when
controlling for electoral fraud, and that loser-strong governorates were no more likely
to perceive fraud. Similarly, Tables A.13-15 ﬁnd that losers in loser-strong governorates
did not perceive a greater deterioration in their country’s economic performance,
security environment, or provision of government services than losers in loser-weak
governorates. Table A.16 shows that losers in loser strong governorates were no more
revolutionary (opposed to former autocrats) than losers in loser weak governorates.
Finally, Table A.17 shows that results hold when controlling for trust in Islamists and
support for shari’a.
Table A.7 Disillusionment and Votes for Autocratic Candidates in Egypt (OLS)
DV: Proportion of losers voting for autocratic candidates:
Pres Elcx 2012 Parli Elex 2011-12 Change in Proportion
(1) (2) (3)
Loser Strength 0.152**
0.001 0.151**
(0.060) (0.043) (0.068)
Constant 0.249***
0.253***
20.004
(0.042) (0.030) (0.046)
Observations 19 19 19
R2
0.272 0.00002 0.228
Adjusted R2
0.230 20.059 0.183
Note: * p,0.1; ** p,0.05; *** p,0.01.
11
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table A.8 Democratic Disillusionment in Egypt and Tunisia, Separately (OLS)
DV: Support for Democracy (0-10)
Egypt Tunisia
(1) (2) (3) (4)
Time* –1.297***
–1.082***
–0.974* –0.771*
Loser Strength (0.482) (0.414) (0.573) (0.427)
Time –1.912***
–2.095***
–0.351 –0.511**
(0.215) (0.201) (0.302) (0.220)
Loser Strength 3.202***
3.208***
1.176 –0.818
(1.077) (0.985) (1.181) (0.772)
Age 0.162**
0.138**
–0.071 0.053
(0.071) (0.065) (0.093) (0.072)
Female –0.026 0.009 0.055 0.132
(0.180) (0.162) (0.251) (0.189)
Education –0.061 –0.024 –0.063 –0.051
(0.051) (0.047) (0.084) (0.063)
Income 0.182***
0.196***
0.021 0.056
(0.056) (0.052) (0.177) (0.117)
Urban 0.222 0.095 0.361 0.065
(0.218) (0.201) (0.327) (0.246)
Muslim –0.596* –0.407 1.221 –0.301
(0.346) (0.201) (1.819) (1.273)
Govcrnorate FE U U U U
Imputation U U
Constant 7.003***
6.598***
4.133* 5.418***
(0.674) (0.607) (2.151) (1.478)
Observations 718 848 444 731
R2
0.233 0.260 0.132 0.104
Adjusted R2
0.201 0.234 0.066 0.064
Note: * p,0.1; ** p,0.05; *** p,0.01.
12
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table A.9 Extent to Which Country is Currently a Democracy
Dependent variable:
Suitability of Democracy Suitability Extent
(1) (2) (3) (4)
Time* –1.053***
–0.97***
–1.169***
–1.060***
Loser Strength (0.349) (0.284) (0.412) (0.339)
Extent 0.359***
0.348***
(0.031) (0.026)
Time –0.609***
–0.756***
0.629***
0.241
(0.177) (0.146) (0.196) (0.168)
Loser Strength 2.893***
3.16***
3.170***
3.692***
(1.026) (0.939) (1.213) (1.118)
Age 0.044 0.088* –0.051 0.006
(0.054) (0.046) (0.064) (0.055)
Female –0.027 –0.018 –0.014 –0.062
(0.140) (0.119) (0.166) (0.141)
Education –0.055 –0.025 –0.052 –0.025
(0.041) (0.036) (0.048) (0.043)
Income 0.114**
0.12***
0.151**
0.157***
(0.052) (0.046) (0.061) (0.055)
Urban 0.270 0.118 0.356* 0.303*
(0.174) (0.151) (0.206) (0.180)
Muslim –0.799**
–0.628**
–1.312* –1.183***
(0.339) (0.276) (0.399) (0.328)
Governorate FE U U U U
Imputation U U
Constant 5.712***
5.279***
2.960***
2.744***
(0.618) (0.537) (0.713) (0.629)
Observations 1,154 1,557 1,154 1,557
R2
0.264 0.256 0.132 0.102
Adjusted R2
0.228 0.229 0.090 0.070
Note: * p,0.1; ** p,0.05; *** p,0.0l.
13
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table A.10 Disillusionment with Democracy Among Voting and Non-Voting Losers
(OLS)
DV: Suitability of Democracy
Voters Non-Voters
(1) (2) (3) (4)
Time* –0.728* –0.820**
–1.264 –1.460*
Loser Strength (0.406) (0.336) (1.094) (0.820)
Time –1.404***
–1.355***
–0.724 –0.788**
(0.197) (0.171) (0.467) (0.352)
Loser Strength 3.177***
3.257***
0.068 0.825
(1.175) (1.060) (2.782) (2.658)
Age 0.076 0.122**
0.049 0.118
(0.066) (0.056) (0.131) (0.104)
Female –0.125 –0.044 0.402 0.421
(0.169) (0.143) (0.356) (0.282)
Education –0.083* –0.026 –0.097 -0.082
(0.049) (0.042) (0.102) (0.085)
Income 0.140**
0.134**
–0.067 -0.014
(0.060) (0.054) (0.129) (0.110)
Urban 0.204 –0.052 0.117 0.211
(0.207) (0.180) (0.467) (0.362)
Muslim –0.528 –0.506 –0.280 -0.222
(0.382) (0.320) (0.969) (0.700)
Governorate FE U U U U
Imputation U U
Constant 7.155***
6.713***
7.963**
6.670***
(0.714) (0.626) (1.473) (1.208)
Observations 882 1,178 276 395
R2
0.179 0.179 0.262 0.2-11
Adjusted R2
0.127 0.140 0.109 0.129
Note: * p,0.1; ** p,0.05; *** p,0.01.
14
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table A.11 Disillusionment with Democracy without Governorate FE (OLS)
DV: Suitability of Democracy
(1) (2) (3) (4)
Time* –0.621* –0.643**
–1.096***
–1.096*
Loser Strength (0.349) (0.291) (0.352) (0.632)
Time –1.376***
–1.376***
–1.469***
–1.469***
(0.169) (0.144) (0.181) (0.361)
Loser Strength –0.091 –0.076 0.363 0.363
(0.264) (0.213) (0.291) (0.380)
Age 0.095* 0.153***
0.149***
0.149**
(0.057) (0.049) (0.056) (0.074)
Female 0.005 0.048 0.017 0.017
(0.147) (0.125) (0.147) (0.175)
Education –0.044 0.006 –0.0003 –0.0003
(0.043) (0.037) (0.045) (0.065)
Income 0.117**
0.107**
0.049 0.049
(0.052) (0.047) (0.087) (0.104)
Urban 0.205 0.113 0.277* 0.277
(0.148) (0.127) (0.149) (0.205)
Muslim –0.392 –0.177 –0.094 –0.094
(0.321) (0.260) (0.303) (0.492)
Imputation U U U
Matching (Nearest) U U
Clustered SE U
Constant 5.789***
5.230***
5.581***
5.581***
(0.506) (0.436) (0.693) (0.866)
Observations 1,162 1,579 1,217 1,217
R2
0.096 0.095 0.115 0.115
Adjusted R2
0.089 0.090 0.109 0.109
Note: * p,0.1; ** p,0.05; *** p,0.01.
15
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table A.12 Disillusionment with Democracy Accounting for Fraud (OLS)
Dependent variable:
Suitability of Democracy Free and Fair? (Post-Election only)
(1) (2) (3) (4)
Time* 20.753**
20.846***
Loser Strength (0.354) (0.295)
Time 21.333***
21.319***
(0.169) (0.143)
Loser Strength 0.220 0.238 0.079 0.179
(0.311) (0.260) (0.154) (0.127)
Fraud 0.261 0.307* 20.341***
20.299***
(0.184) (0.163) (0.122) (0.105)
Age 0.098* 0.133***
0.045 0.062**
(0.057) (0.048) (0.035) (0.029)
Female 20.031 20.015 0.079 0.057
(0.148) (0.125) (0.089) (0.076)
Education 20.046 20.006 20.049* 20.017
(0.043) (0.037) (0.028) (0.023)
Income 0.111**
0.067 20.049 0.002
(0.053) (0.047) (0.066) (0.049)
Urban 0.350**
0.239 20.098 20.126
(0.168) (0.145) (0.103) (0.089)
Muslim 20.502 20.377 0.540**
0.429**
(0.335) (0.273) (0.251) (0.182)
Subregion FE U U U U
Imputation U U
Constant 5.211***
5.242***
2.727***
2.343***
(0.722) (0.601) (0.589) (0.446)
Observations 1,162 1,573 614 797
R2
0.123 0.123 0.247 0.272
Adjusted R2
0.107 0.111 0.223 0.254
Note: * p,0.1; ** p,0.05; *** p,0.01.
16
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
TableA.13EvaluationoftheEconomyinLoserStrongv.WeakGovernorates(OLS)
Dependentvariable:
Curr.EconNeedsJobsNarrow
(1)(2)(3)(4)(5)(6)(7)(8)
Time*0.158*0.162**
20.19520.1460.0990.1080.0660.121
LoserStrength(0.089)(0.072)(0.122)(0.097)(0.100)(0.083)(0.099)(0.080)
Time20.703***
20.659***
20.05020.05820.689***
20.631***
20.704***
20.676***
(0.043)(0.036)(0.058)(0.049)(0.048)(0.042)(0.047)(0.040)
LoserStrength0.0190.1861.642***
1.485***
0.0740.0270.2180.238
(0.269)(0.250)(0.368)(0.339)(0.301)(0.283)(0.295)(0.271)
Age0.026*0.036***
0.060***
0.038**
–0.00050.0130.0230.030**
(0.014)(0.012)(0.019)(0.016)(0.016)(0.013)(0.015)(0.013)
Female20.066*20.0380.0590.078*0.0080.0030.0220.033
(0.036)(0.030)(0.049)(0.041)(0.040)(0.035)(0.040)(0.034)
Education0.0020.0010.107***
0.118***
–0.002–0.004–0.0020.003
(0.010)(0.009)(0.014)(0.012)(0.012)(0.010)(0.012)(0.010*
Income0.036***
0.037***
0.149***
0.152***
0.039***
0.035**
0.0240.027**
(0.013)(0.012)(0.018)(0.016)(0.015)(0.014)(0.015)(0.013)
Urban20.03920.0340.122**
0.162***
–0.106**
–0.053–0.071–0.057
(0.044)(0.061)(0.050)(0.050)
Muslim0.150*0.127*–0.098–0.0980.1050.152*0.0600.028
(0.086)(0.070)(0.118)(0.095)(0.097)(0.080)(0.095)(0.077)
GovernorateFEUUUUUUUU
ImputationUUUU
Constant1.423***
1.405***
0.2930.2841.691***
1.524***
1.659***
1.569***
(0.155)(0.137)(0.213)(0.186)(0.175)(0.156)(0.172)(0.150)
Observations1,2181,7061,2111,7131,2021,6651,1851,628
R2
0.3000.2560.2260.2140.2590.2090.2610.230
AdjustedR2
0.2680.2320.1900.1890.2250.1830.2270.204
Note:*p,0.1;**p,0.05;***p,0.01.
17
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
TableA.14EvaluationofSecurityinLoserStrongandWeakGovemorates(OLS)
Dependentvariable:
CurrSafetyRetroSafetyPolice
(1)(2)(3)(4)(5)(6)
Time*20.0270.07220.0960.0710.1200.150
LoserStrength(0.114)(0.091)(0.107)(0.085)(0.164)(0.132)
Time20.549***
20.480***
0.0100.01020.348***
20.283***
(0.054)(0.046)(0.051)(0.043)(0.078)(0.066)
LoserStrength0.865**
0.910***
20.21420.21220.73820.521
(0.343)(0.317)(0.323)(0.296)(0.492)(0.457)
Age0.0040.0120.0210.029**
0.0360.015
(0.018)(0.015)(0.017)(0.014)(0.025)(0.022)
Female20.06720.0630.082*0.0540.0580.099*
(0.046)(0.038)(0.043)(0.036)(0.066)(0.056)
Education20.01320.0010.0140.0130.0010.017
(0.013)(0.012)(0.013)(0.011)(0.019)(0.017)
Income0.044***
0.041***
20.01020.0120.097***
0.095***
(0.017)(0.015)(0.016)(0.014)(0.024)(0.022)
Urban20.00720.01320.08320.06320.183**
20.158**
(0.057)(0.049)(0.054)(0.046)(0.082)(0.070)
Muslim0.0480.0930.1420.0900.00420.099
(0.109)(0.090)(0.103)(0.084)(0.157)(0.129)
GovernorateFEUUUUUU
ImputationUUU
Constant1.763***
1.683***
1.112***
1.127***
1.732***
1.837***
(0.198)(0.175)(0.187)(0.163)(0.284)(0.250)
Observations1,2131,6981,2091,6941,2131,697
R2
0.2390.2090.1420.1250.1530.111
AdjustedR2
0.2040.1830.1020.0970.1140.082
Note:*p,0.1;**p,0.05;***p,0.01.
18
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UT6 12:34:56 UTC
All use subject to https://about.jstor.org/terms
TableA.15EvaluationofHealth,Redress,andManagementofTransition(OLS)
Dependentvariable:
HealthRedressManage
(1)(2)(3)(4)(5)(6)
Time*0.206*0.268***
20.22220.1840.0150.062
LoserStrength(0.108)(0.089)(0.163)(0.129)(0.108)(0.090)
Time20.736***
20.710***
20.147*20.153**
21.059***
20.982***
(0.052)(0.045)(0.077)(0.065)(0.052)(0.046)
LoserStrength0.1600.10620.53620.2730.3960.380
(0.324)(0.305)(0.488)(0.444)(0.314)(0.292)
Age0.0050.0090.070***
0.046**
0.0030.031**
(0.017)(0.015)(0.025)(0.021)(0.017)(0.015)
Female20.0010.0180.235***
0.220***
0.01920.011
(0.044)(0.038)(0.065)(0.054)(0.043)(0.038)
Education20.01120.0160.0310.044***
0.0020.004
(0.013)(0.011)(0.019)(0.016)(0.013)(0.011)
Income0.049***
0.041***
20.01520.0010.050***
0.042***
(0.016)(0.015)(0.024)(0.021)(0.016)(0.015)
Urban20.100*20.06320.11820.132*20.101*20.084*
(0.054)(0.048)(0.081)(0.068)(0.055)(0.048)
Muslim20.0400.0630.2250.249**
20.03520.028
(0.103)(0.086)(0.156)(0.126)(0.103)(0.085)
GovernorateFEUUUUUU
ImputationUUU
Constant1.811***
1.758***
1.909***
1.856***
2.023***
1.905***
(0.188)(0.169)(0.283)(0.244)(0.186)(0.166)
Observations1,1921,6441,2071,6871,1241,498
R2
0.2590.2260.1150.0860.3910.358
AdjustedR2
0.2240.2000.0750.0570.3610.335
Note:*p,0.1;**p,0.05;***p,0.01.
19
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table A.16 Losers’ Support for Former Autocratic Leaders; Pre-Election Survey
(OLS)
DV: Support for Former Autocratic Leaders (1-10)
(1) (2)
Loser Strength 0.398 0.377
(0.645) (0.595)
Age 0.008 0.014
(0.069) (0.054)
Female 0.231 0.221
(0.181) (0.137)
Education 20.062 20.041
(0.050) (0.040)
Income 0.010 20.008
(0.048) (0.041)
Urban 20.460**
20.255
(0.208) (0.166)
Muslim 20.313 20.526*
(0.345) (0.306)
Governorate FE U U
Imputation U
Constant 2.156***
2.165***
(0.745) (0.669)
Observations 527 770
R2
0.114 0.096
Adjusted R2
0.027 0.036
Note: * p,0.1; ** p,0.05; *** p,0.01.
20
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms
Table A.17 Disillusionment with Democracy controlling for Ideology (OLS)
DV: Suitability of Democracy
(1) (2)
Time* 20.977***
20.956***
Loser Strength (0.368) (0.301)
Time 21.240***
21.172***
(0.194) (0.162)
Loser Strength 2.639**
2.887***
(1.086) (0.991)
Trust in Islamists 0.077 0.148
(0.232) (0.194)
Support for Shari’
a 0.190 0.310**
(0.172) (0.145)
Age 0.096* 0.119**
(0.057) (0.048)
Female 20.035 20.004
(0.148) (0.124)
Education 20.055 20.013
(0.043) (0.037)
Income 0.108**
0.059
(0.054) (0.047)
Urban 0.216 0.104
(0.184) (0.157)
Muslim 20.698* 20.668**
(0.393) (0.317)
Governorate FE U U
Imputation U
Constant 6.713***
6.049***
(0.793) (0.675)
Observations 1,162 1,573
R2
0.173 0.170
Adjusted R2
0.132 0.140
Note: * p,0.1; ** p,0.05; *** p,0.01.
21
This content downloaded from
78.128.147.204 on Tue, 02 Mar 2021 14:46:15 UTC
All use subject to https://about.jstor.org/terms