Expressing Voter Choices in an Authoritarian Election.
Non-conformist Voting in Cuba1
Jorge I. Domínguez2
Ángela Fonseca Galvis3
Chiara Superti4
Abstract
How do voters make choices in single-party authoritarian elections where the number of candidates
equals the number of parliamentary seats? We argue that voter access to information in a limitedinformation environment may permit the expression of some autonomous voter preferences. Thanks to
a unique data set for Cuba’s 2013 National Assembly election, we show how Cuban voters use
information short-cuts to make real choices even in such elections. Locally-rooted politicians, not
national politicians, performed best. Candidates who have been municipal delegates are more likely to
be known in the district because they have competed in municipal elections, which are multi-candidate
albeit single-party. Voters do not seem to vote based on clientelism, nor did we find evidence of fraud
in the reporting of votes. Voters neither rejected nor endorsed the Communist Party; instead, voting
revealed a relatively neutral stance toward the Party, with a mix of support and opposition.
1
2
3
4
We are grateful to Aaron Watanabe for his excellent research assistance. We thank Lisa Blaydes, Shelby Grossman,
Jennifer Pan, Amanda Pinkston, Pablo Querubín, Margaret Roberts, Tesalia Rizzo, Paul Schuler, and Vanessa
Williamson, and participants at the Harvard University Comparative Politics Workshop and the Latin American Politics
Workshop for their helpful comments. All errors are our sole responsibility.
Harvard University
Pontificia Universidad Javeriana
Harvard University
1
1. Introduction
How do voters approach elections for which the number of candidates equals the number of seats? Do
such voters find nooks and crannies in the authoritarian election design to exercise some choice? What
informational clues may they employ? Are there whiffs in their choices that may set up pre-transition
circumstances for subsequent political change? Cuba's 2013 National Assembly election provides us
with a rare opportunity to examine these questions.
The bulk of the scholarly literature on elections in authoritarian regimes focuses on why
authoritarian regime elites would allow and run regular elections (Gandhi & Lust-Okar, 2009), or
examine the limited possibilities for electoral competition in competitive authoritarian regimes
(Levitsky & Way, 2010). As Brownlee (2011, p. 823) puts it, such elections are “political safety
valves.” Such research focuses on outcomes for the political regime, not on voting behavior; our focus
is on voting behavior.
There are two puzzles regarding voting behavior under authoritarian regimes. One is election
turnout. Scholars of democratic societies have long pondered why any voter would turn out to vote
when any voter’s impact on the outcome would be essentially zero (Aldrich, 1993; Downs, 1957, pp.
260-276). This puzzle is only more striking in an authoritarian-regime election where voters are not
compelled to vote and cannot expect to have any impact on who gets elected. The second puzzle is why
voters in an authoritarian regime would cast non-conformist ballots. We call nonconforming voters
those who vote selectively (contrary to the official slogan to vote for the entire official slate) or vote
blank or void their ballot. What lies behind this attempt to express a voter’s preference: a protest, a
response to clientelism, or an actual electoral preference?
In this article we focus on one type of non-conformist voting, selective voting: voters vote for
only one or some but not for all the candidates on the ballot for their district, contrary to the
2
Communist Party’s preference that they should vote for the full slate. Explaining how voters choose to
vote helps to understand why citizens increasingly vote selectively instead of choosing other forms of
protest vote or voting for the entire official slate. We assess four potential explanations for such voting.
One emphasizes clientelistic appeals that candidates may make. This explanation requires that
candidates be able to make such appeals without being arrested and that they have the resources to do
so. Scholars of non-communist authoritarian regimes that schedule elections point often to clientelistic
mechanisms to explain aspects of voting behavior (Blaydes, 2011; Greene, 2007; Magaloni, 2006). A
second explanation highlights partisan signal sending (either out of loyalty or fear). This explanation
requires voters to know who is a communist party leader and vote accordingly. A third explanation
claims that this is a form of protest against the authoritarian regime (Shi, 1999; Southwell & Everest,
1998). Finally, a fourth explanation focuses on access to and use of information in a limitedinformation environment to shape actual electoral preferences; voters express an autonomous choice
based on their own candidate preferences, despite knowing that everyone on the ballot will be elected
and that the National Assembly has little effect on actual policy making. In this interpretation, voters
under some authoritarian regimes exercise choice in their elections in ways that evoke local electoral
behavior in many countries (Blais et al., 2002; Tavits, 2010).
Our empirical findings support this fourth explanation. Thanks to a unique data set for Cuba’s
2013 National Assembly election, we show how Cuban voters appear to be making "choices" even in
elections where the number of candidates equals the number of seats. In such a low information
context, voters use information short-cuts (Shugart et al., 2006). Locally-rooted politicians, not national
politicians, performed best in this Cuban election because voters accessed local information. National
Assembly candidates who have been municipal delegates are more likely to be known in the district
because they have competed in multi-candidate local elections and have earned their reputations.
Voters backed candidates who have grassroots links, experience of municipal multi-candidate electoral
3
contestation (where one candidate wins and the other loses), and do their jobs well -- the voter’s
principal information short-cuts to evaluate retrospective performance.
Cubans re-enacted V. O. Key’s observations made in the 1940s regarding voting behavior in
U.S. southern states. Key wrote about voting in South Carolina, in a section he called "Friends and
Neighbors". "In the absence of well-organized politics," as in the one-party political regime then
prevalent in that state, "localism will play a powerful role in the orientation of the voters' attitudes." In
Alabama, "the friends-and-neighbors pattern reflects the absence of well-organized competing
factions" (Key, 1949, p.132 on South Carolina, p. 41 on Alabama). Cuban National Assembly elections
feature “friends-and-neighbors” effects.
This selective voting does not seem to be either a pure protest against the regime nor a blind
support for communist party members. In 2013 selective voting neither rejected nor endorsed the
Communist Party; instead, voting revealed a relatively neutral stance. Voters sometimes supported and
sometimes failed to support communist party leaders. Signal-sending was blurred. (Until Cuba holds
multi-party elections with full access to competing sources of information, it is impossible to tell how
well communist party candidates truly evoke voter support.) Finally, we also find little evidence of
clientelistic-motivated voting in 2013.
Where do Cuban voters get information? The state owns and operates all television and radio
stations and all the daily newspapers. Individual candidate campaigning is prohibited, as is financing
for individual campaigns. Only the Communist Party of Cuba is legal. In such a low-information
authoritarian context, voters look for the few available clues. National leaders appear in the official
mass media. But voters have two other means to learn about candidates. One is a set of officiallygenerated short biographies of each National Assembly candidate; these are posted in each district and
are visible at or near the polling places. The other is that Cuba’s municipal elections feature two
candidates for every seat, and many of these municipal assembly officials are also nominated to run for
4
the National Assembly. At the local level, therefore, the election and performance of the municipal
assembly official chosen through a local multi-candidate election provides key information — these are
the only politicians voters had chosen in elections featuring inter-candidate choice.
Our findings show that, even in an authoritarian regime with very restrictive rules for elections,
voters find lawful means to express real preferences: they do not vote for the full official slate but,
instead, choose specific candidates, even though such voting does not affect who will be elected. This
is a remarkable expression of citizen agency in an authoritarian regime.
This paper is organized as follows. The next section synthesizes the scholarship on communist
regime elections. The next three sections describe Cuba’s National Assembly and electoral system, the
growth of the nonconformist voting that permits our analysis, and gives evidence that even the current
electoral system already serves to embarrass powerful politicians in Cuba. Sections six through eight
present our hypotheses, data, and variables. Section nine features our analysis regarding the
explanatory importance of prior local electoral experience. Sections ten and eleven explore alternative
explanations, namely, political information, clientelism, and electoral fraud. Section twelve delves
deeper to examine how the communist party affects election. And the last section concludes.
2. Communist regime elections and the Cuban case
We seek to understand voter choices in communist-regime elections. Research on such elections should
ask five questions:
• Are there direct elections for the national parliament?
• Is there more than one party on the ballot and, if so, do the parties belong to the same ruling
coalition or do they compete against each other?
• If parties are not allowed to compete, is there a choice between candidates of the official party?
5
• If there is multicandidate competition under a single-party regime, do the rules protect national
leaders from defeat or embarrassingly low voter support?
• If there is multicandidate competition, does the lowest-voted candidate lose?
Communist regime elections varied considerably. At one end, the People’s Republic of China
does not have direct elections for its national parliament; members of the National People’s Congress
are chosen indirectly. None of the communist systems with direct elections had competitive multiparty
elections. The former Soviet Union had direct elections for the Supreme Soviet, but the number of
candidates equaled the number of seats, with only one candidate per seat per district. Voters could
abstain, vote blank, or void the ballot but had no ability to choose between candidates competing
against each other (Gilison, 1968). In Poland, starting after the 1956 peaceful limited revolution, the
electoral law permitted multiple candidates for each seat and multiple parties to nominate candidates,
but it did not permit opposition parties. There were 750 candidates for 459 seats in the 1957
parliamentary (Sejm) election. Candidates were elected from various parties, with a plurality for the
Communists, but all elected Deputies were members of the government coalition (Staar, 1958).
At the other end, the 2007 National Assembly election in communist Vietnam featured at least
two candidates for every national parliamentary seat. Most centrally nominated candidates won but just
above half of the locally nominated candidates lost. Only one self-nominated candidate won a National
Assembly seat; all other elected parliamentarians had central or local official sponsorship. Vietnam’s
Communist Party leaders assigned centrally-nominated candidates to safe seats, guaranteeing a good
electoral outcome for them. Members of the Political Bureau and the Central Committee of the
Communist Party performed best, as did incumbents in general. Local politicians, however, ran in lesssafe districts; they performed less well (Malesky & Schuler, 2009; Malesky & Schuler, 2011).
Cuba’s electoral system stands between the ends of the communist-regime spectrum. Unlike
6
China, Cuba has had direct elections for its National Assembly since the early 1990s. Like the Soviet
Union, the number of candidates has equaled the number of National Assembly posts but, unlike the
Soviet Union, Cuba clusters candidates in districts, yielding differentiated outcomes for candidates who
are nevertheless assured of election. As in Poland and Vietnam, Cuban elections permit comparisons
between the shares of votes for each candidate. Unlike in Poland and Vietnam, no National Assembly
Deputy has yet been defeated in Cuba. (Cuba also features multi-candidate single-party local elections.)
We focus on Cuban National Assembly elections because it retains the emblematic communist regime
procedure whereby the number of seats equals the number of candidates yet there is also an important
element of voter choice.
The February 2013 National Assembly election was the first for which the government
published the pertinent data and the first for which official institutions did not intensively campaign in
favor of casting a ballot for the entire official single slate, knowing that every candidate on the official
slate would be elected. In the 2013 election, 1,846,691 voters (23.5 percent) cast nonconforming ballots
(Calculated from “Resultados finales” 2013).
3. Cuba’s National Assembly and Electoral System
In 1976, Cuba created a National Assembly (parliament) and adopted indirect means to choose
Deputies. Voters voted directly only for municipal assembly candidates, who serve for 2.5-year terms,
renewable; municipal assemblies chose the members of the provincial assemblies and the Deputies of
the National Assembly. Deputies serve for five-year terms, renewable. At that first National Assembly,
55.5 percent of the Deputies had also been elected as municipal assembly officials and were thus
performing double duty. The remaining 44.5 percent of the Deputies were government and party
officials, never elected in a direct election, who were chosen to become Deputies by the municipal
7
assemblies (Domínguez, 1978, pp. 243-247). Since 1976, the National Assembly has met twice, at most
three times per year, and each time for just a few days, including the committee hearings in the week
before plenary sessions. Cuba's National Assembly has never defeated a government bill and never
approved a bill introduced by Deputies without prior executive support.
In response to the collapse of European communist regimes in 1989-1991 and the protests at
Tiananmen Square in Beijing, in October 1992 a new electoral law was approved.
1
The following are the key characteristics of the National Assembly’s elections. Items five through
seven indicate the variation that permits our analysis, notwithstanding single-party single-candidateper-post elections.
[1.] Single party. The Communist Party of Cuba is the only lawfully authorized party (Constitution,
Article 5).
[2.] Self-nomination is prohibited. Candidacy commissions screen Deputy candidates for the National
Assembly; commission members are drawn from the officially-sponsored “mass” organizations (Ley
Electoral, Article 68). These commissions recommend to the municipal assembly at least two names for
each candidacy (Ley Electoral, Article 89).
[3.] One post, one candidate. Only the municipal assembly chooses Deputy candidates and it must
choose only one candidate for each post to be filled (Ley Electoral, Article 92). Every candidacy is thus
politically vetted in advance.
[4.] Posting photos and biographies, and limiting campaigns. Only the electoral district commissions -not the candidates-- may draft the biographies following a standard format and post them and the
photos of the candidates (Ley Electoral, Article 30ch) outside each voting location for voters to see
prior to voting. There is no candidate campaigning.
[5.] Multiple candidates per district. Each district votes on at least two candidates for two posts. The
8
most populous electoral districts may elect more Deputies, though the number of candidates always
equals the number of posts to be filled (Ley Electoral, Articles 14 and 15).2
[6.] Voter choice. A voter “may vote for as many candidates as may appear” on the ballot, with an X
next to each name, but if “the voter wishes to vote for all the candidates [the voter] may write an X at
the circle that appears at the top of the ballot” (Ley Electoral, Article 110). The voter may leave the
ballot blank, or void it. To be elected Deputy, a candidate must win more than half of the valid votes
cast; blank and void votes are not counted as valid votes. (In multi-candidate municipal assembly
elections, voters must choose between candidates. Many candidates fail to win a majority of the valid
votes on the first-round election; a second-round election is held between the top two vote-getters.) No
Deputy candidate for the National Assembly has been defeated but the legal possibility exists.
[7.] Candidate variation. Municipal assembly candidacy commissions may choose up to 50 percent of
the Deputy candidates from among the members of the municipal assembly (Ley Electoral, Article 93).
Once elected, they serve simultaneously in the municipal and national assemblies. Deputy candidates
who are not municipal assembly members need not reside in the district where they become candidates.
Therefore, the Cuban electoral law makes it possible for every voter to support el voto unido,
marking an X for the entire official slate, or to vote for some but not all of the candidates even though
all of the candidates will be elected. Voters may thus express support or displeasure at the voting booth
in three ways: voting blank, voiding the ballot, or voting selectively, the sum of which we call
nonconformist voting. Figure 1 shows the ballot for a three-seat district with three candidates. It
prominently features the option for the official united slate front and center, and then it lists the
individual candidates in the event that the voters choose to vote for some but not all of them.
9
Figure 1. Ballot National Assembly Election. The note at the top
says: “to vote for all candidates it suffices to put a cross just in the
circle.” Reproduction from the original ballot in Granma (2013).
Note the relationship between candidacies for National Assembly and municipal assembly
elections. The same Ley Electoral (Article 82) mandates that the number of candidates for the
municipal assemblies must be at least two per local post to be filled. Municipal assembly elections are
multi-candidate albeit single-party, with the same constraints on campaigning noted above. Members of
municipal assemblies are also vetted as candidates for the National Assembly. Therefore, in National
Assembly Deputy elections the amount of voter information regarding the candidates differs. Some
Deputy candidates are national officials in government and party, not elected to the municipal
assemblies; other Deputy candidates are better known at the local level because they had been
previously elected to the municipal assemblies.
Three mechanisms underscore the salience and competitive challenges of municipal assembly
officials. First, municipal assembly delegates are locally elected in multi-candidate elections. They face
competition not only once they are on the ballot but also in getting placed on the ballot to run for re10
election; for the 2015 municipal elections, only 60 percent of the incumbent municipal delegates were
renominated (“Un paso más cerca de elegir,” 2015). Second, the law requires municipal assembly
delegates to be accountable twice a year at a public assembly of constituents and to listen to their
requests. Official statistics aver that over a fifth of the population attends such events (Amaro, 1996).
Between 2012 and 2015, citizens voiced nearly 1.2 million requests to their municipal representatives,
of which three-quarters were reported as having been “solved” (“Elector-planteamiento-delegado…”
2015). National Deputies are not legally required to hold such accountability assemblies or to reside in
the districts from which they are elected. Third, municipal delegates act as problem solvers. As a
Granma columnist put it, voters expect them to become a “one-man band,” noting the risk that
delegates would breach the law (Marrón González, 2015). Because the allocation of resources remains
very highly centralized, no municipal delegate may easily engage in clientelistic electoralist practices
without the risk of arrest (“Las próximas elecciones…” 2015). Nevertheless, jointly the multi-candidate
local election, the accountability assemblies, and the problem-solving work render the municipal
delegate an attractive political figure with significant roots in a community.
4. Voting Context: Non-conformist voting matters
Senior officials keep track of nonconformist voting. For example, following the nationwide April 2000
municipal elections, the president of the national election council (Comisión Electoral Nacional -CEN), Roberto Díaz Sotolongo, noted with satisfaction that voting abstention, and the proportion of
blank and void ballots, had all fallen from the 1997 nationwide municipal election (Table 1). The CEN
President added that voting in Cuba offered the option of voting for no candidate (Mayoral, 2000).
The salience of the elections is also evident because neighborhood Committees for the Defense
of the Revolution, or groups of students, are tasked to contact voters to turn out, and job promotions or
11
the allocation of benefits often include verification of having voted. These contacts increase turnout but
may also create apprehension about the possible adverse consequences of voter abstention (López,
1993, p. 52; Guanche, 2012, p. 78). Moreover, the mass organizations have publicly urged voters to
cast their ballot for the single slate (“Llama la CTC al voto unido,” 2008). The official newspaper's
report on the 2008 National Assembly election highlighted the “triumph of the united vote” (“¡Triunfó
el voto unido!” 2008), thereby calling attention to nonconformist voting.
Cuban scholars also highlight the salience of the official single slate. As Julio César Guanche
puts it, "voters have responded highly and positively to the official requests for the united vote -- for all
the candidates -- alongside with low blank and void voting... The elections have served as ‘virtual
plebiscites...’ on revolutionary rule” (Guanche, 2012, p. 71; see also Valdés Paz, 2009, pp. 149-170).
Nonconformist voting matters.
In the early 1990s, the novelty was the direct election for national deputies, which gave voters
the option to cast nonconforming national votes for the first time. Although Cuba’s government and
economy were shaken badly by the collapse of the Soviet Union and its allies, Cuban leaders had faith
that their communist party was different. Cuban scholars had carried out detailed research for the 1989
municipal elections. They found, first, that the communist party as an institution did not add much
value to a candidacy. Only two percent of 150 voters surveyed mentioned membership in the
communist party as a "desirable" quality in a municipal assembly candidate. Second, voters supported
their friends and neighbors who were good in the community, worked hard, and were interpersonally
effective. Third, seven out of ten Cuban municipal assembly members elected in 1989 were communist
party members, who were good workers, friends, and neighbors. Communist party candidates, as
human beings, added value to a party that lacked popular prestige (Dilla, González, & Vicentelli, 1994,
pp. 69-73). Cuban leaders hoped that party members could rescue the party out of the crisis blamed on
the European communist collapse.
12
Table 1. Nonconforming voting for Cuba’s National Assembly elections (percentages)
1993
2008
2013
Selective vote
4.6
8.7
17.6
Blank and void vote
7.0
4.8
5.9
Nonconforming vote
11.6
13.5
23.5
Source: Granma, March 11, 1993; January 30, 2008; and February 8, 2013.
In the first direct election for the National Assembly (1993), blank and void voting – pure protest
voting – exceeded the selective voting (Table 1). It took voters until the 2008 election, the first under
Raúl Castro, to prefer the selective vote, which became the dominant means for nonconforming voting
by the 2013 National Assembly election. In 2013, the official press refrained from calling for el voto
unido for the first time ever (August 2014, 89), though the ballot had not changed. In 2013, nearly a
quarter of Cuban voters were nonconformists -- over 1.8 million voters. Cuban officials and the press
reported the results with no comment. They could have celebrated a democratic opening; they could
have evinced embarrassment that the united slate had fallen significantly. They did neither.
By 2013, Cuban voters had reasons for concern and hope. Cuba's gross domestic product per
capita hardly grew since the 2008-09 financial crisis (Pérez Villanueva, 2015). This slowdown
coincided with Raúl Castro's replacement of his brother Fidel as Cuba's president (acting president
August 2006-February 2008, president since then) and preceded the spike in nonconformist voting in
the 2013 election.
By 2013, the voters may have felt freer to choose. Cuba experienced a gradual opening of the
public sphere under Raúl Castro's presidency. There was a more open and vigorous debate at the
University of Havana regarding economic and social policies. In the late 1990s, Temas a social
13
sciences journal written for a general university-educated reader began publication and, during Raúl
Castro's years, widened its scope. In 2005, the Roman Catholic Archdiocese of Havana launched a
magazine, Espacio Laical, to address contemporary topics in Cuban society, economy, politics, and
diaspora, not just religious topics. In 2008, the communist party's newspaper, Granma, resumed
publication once a week of letters-to-the-editor, which it had suspended a quarter century earlier (“De
la Dirección,” 2008). Subsequently, Granma’s digital edition began to publish comments on some of its
articles. In 2013, Cuba had only 90 personal computers per thousand people and 261 people per
thousand had some Internet access (Organización Nacional de Estadística e Información, 2013, Table
17.4); thus mainly middle-class users may access these digital editions. Yet, those who do, argue
vigorously. For example, in the four days following publication of a new foreign investment law in
2014, comments posted on Granma's site sharply criticized aspects of the new law (“Texto de la ley...”
2014). Similarly in February 2015, in anticipation of upcoming municipal elections, the newspaper
Juventud Rebelde (“Concluyó entrevista…” 2015) hosted an online forum about the Cuban electoral
system that generated significant criticisms. (Internet traffic has grown as well outside official sites.)
Within the Castro family, President Raúl Castro's daughter, Mariela, became a public advocate for gay
rights, helping to shift the Cuban government from its hitherto homophobic policies (1960s to 1980s).
In a seemingly off the cuff remark at the December 2010 National Assembly plenary meeting, Raúl
Castro remarked, “Why do we have to butt in on people's lives?” Granma reported that the National
Assembly burst into applause.
This greater openness of the public sphere and the serious economic constraints are consistent
with Brownlee’s (2011) argument that elections in non-communist authoritarian regimes open a
political safety valve that allows officials to collect information from this behavior, which in different
ways Chinese officials do as well (King, Pan & Roberts 2013).
14
5. Embarrassing the Powerful?
In Vietnam, the Communist party protected many central national leaders from electoral
embarrassment. In Cuba, the Communist party protected fewer. Consider candidates from the fourteenmember Communist Party Political Bureau (the country’s top political leaders) who ran in 2013 in
Havana province. Marino Murillo, the economic policy tsar, came in third and last in District 1, Plaza
de la Revolución, overtaken by two municipal assembly delegates. His first vice minister, Adel
Yzquierdo, also came in third and last in District 1, Playa. Foreign Minister Bruno Rodríguez also
came in third and last in District 2, Diez de Octubre, behind a municipal delegate. Lázara López Acea,
First Secretary of the Communist Party in Havana province, came in fourth and last in District 2,
Boyeros; the visible award-winning chief of the national weather bureau won the district. Armed
Forces Minister General Leopoldo Cintras and National Assembly President Esteban Lazo won their
districts in Arroyo Naranjo but each just by two percentage points.
Political Bureau members who ran for the National Assembly in eastern Cuba did better but
won no landslides. In Santiago province, Interior Minister General Abelardo Colomé won District 1 in
Contramaestre, as did Army Corps General Álvaro López Miera in District 5, City of Santiago, each by
a percentage point over the next candidate. National Second Secretary of the Communist party, José
Ramón Machado, led by three points in District 1, Guantánamo.
The worst outcome was Salvador García Mesa’s, long-time leader of the Central Confederation
of Labor (CTC), who trailed by nearly nine percentage points, in a two-member district, Santa Cruz del
Sur, Camagüey. A strong endorsement, in Santa Clara’s District 3, went to Miguel Díaz-Canel, Raúl
Castro’s designated successor, who led the next candidate by thirteen points.
President Raúl Castro, with 98.046 percent of the votes in San Luis, led the other candidate by
twelve percentage points. His brother Fidel got 94.727 percent in District 7, City of Santiago, both
15
brothers running in Santiago province. Nationally, Raúl was only the fourth highest vote winner.
In short, unlike in Vietnam, Communist party national leaders were not shielded from mediocre
election results. If the candidate with the fewest votes in a district would have been barred from
becoming a Deputy, then five of the fourteen Political Bureau members would have been defeated.
These results also suggest that the votes were counted as they were cast. Running in Havana province
was tough. Had Cuban leaders behaved like Vietnamese leaders, more Cuban leaders would have been
candidates in eastern Cuba, where the Castro brothers and top Generals won. These observations
preview our statistical analysis: top national leaders did acceptably but did not outperform.
6. Hypotheses
How are Cubans using the selective voting? Whom do they reward when they choose not to cast a
“united vote”? We assess two intuitions. The first intuition is the standard model for communist regime
elections, including Vietnam. Incumbents, Communist party members, and national politicians should
do best systematically and win the most selective votes — the first three hypotheses. Voters support
such candidates because they are loyal to the regime or because they fear retribution. They feel that
selecting to support such candidates from the group of candidates on the ballot will protect them and
recognize their loyalty. On the other hand, voters may systematically shun party and regime insiders,
which would falsify these three hypotheses; they discard fears and in protest select out those candidates
who are party members or regime incumbents.
The second intuition is that voters use the selective vote, not for systematic support or protest
but to express a more nuanced choice of candidates. The 1992 Cuban electoral law sought to build
regime support from the bottom up and permit Cubans to express candidate preferences. Thus Cubans
16
should be more likely to support municipally-rooted National Assembly Deputy candidates as well as
those who earned job-related merits — the fourth through sixth hypotheses. The challenge is how to
distinguish the variation observed in the distribution of the selective votes from what might be
observed as a result of clientelism. Do local connections between candidates and voters spur patronage
networks?
Here is how we test these different intuitions:
Fear/loyalty versus protest:
Hypothesis 1. Incumbents running for re-election should win more votes.
Hypothesis 2. Communist party Deputy candidates should obtain more votes than noncommunist-party candidates.
Hypothesis 3. Members of the Political Bureau, the Council of State, the Cabinet, and the party's
Central Committee should win more votes than those who are not.
Preferences expression versus Clientelism:
Hypothesis 4. Deputy candidates previously elected at multi-candidate municipal assembly
elections should win more votes in Deputy elections than Deputy candidates not previously
elected to local office. Voters value candidates for their local roots and for their winning an
inter-candidate election.
Hypothesis 5. Voters should support people who get things done. Those who have earned awards
for their hard work or served in international missions should win more votes.
Hypothesis 6a. If clientelism were at work, it should be evident everywhere, and distance to
Havana should not matter for the varying vote share between municipal delegates and nonmunicipal delegates. Local candidates should have the same advantage everywhere. The
17
relationship between distance and premium for local candidates should not be statistically
significant.
Hypothesis 6b. If information rather than clientelism is at work, voters close to Havana will have
access to more information for both local and national candidates. Therefore, the premium of
local candidates will not be apparent near Havana. Farther away from the capital, the visibility
of local candidates will gain them a significant premium in vote share over less known national
candidates. The interaction between distance and municipal delegates should be positive and
statistically significant.
Hypothesis 6c. If information rather than clientelism is at work, the relationship between distance
and premium for local candidates presented in 6.b should be attenuated by education. Areas
with high education should have access to more information about all candidates, which
compensates for the gap in visibility between local and national candidates. Therefore, the
premium that municipal delegates receive should be increasing along with distance to Havana
and should be negative with regard to education (e.g., the premium is attenuated in areas with
high education).
Hypothesis 6.d: If clientelism explains the observed difference in vote shares for municipal
delegates, then we should expect to detect a systematic difference in electoral fraud.
For control purposes, we also look at the provincial assemblies. Cubans vote directly for
members of the provincial assemblies. In provincial elections, the number of candidates also equals the
number of seats; other constraints on nominations and campaigns apply as well. We expect that a
candidacy for the National Assembly is neither stronger nor weaker if the National Assembly candidate
had been previously elected to a provincial assembly. Unlike the multi-candidate municipal elections,
the provincial elections are neither contested between candidates at risk of defeat nor connect to
18
grassroots bases. Voters obtain little additional information from provincial assembly elections.
In 2013 President Raúl Castro indicated his election priorities. He was pleased that women were
48.86 percent of the Deputies, black and mestizo candidates summed to 37.9 percent, and 67.26 percent
of the Deputies were on their first term (Castro, 2013).
7. Data3
We collected and coded the officially-published biographies of all National Assembly members elected
in 2013. The biographies include a photograph, information about the life of the candidates,
educational background, employment, awards won, participation in various organizations, and previous
political experience. In 2013, 612 candidates entered the assembly, of whom 301 were women and 500
had a university degree (Panel A in table 2 has descriptive statistics of the candidates). One third of the
candidates mention CCP membership and 27% were incumbents. We also have data on each
candidate’s vote share in 2013; it ranged from 99.993 percent to 66.629 percent. Both the biographies
and the vote shares were published by Granma, the official newspaper of the Cuban Communist Party.
The data about the municipalities come from the Anuarios Estadísticos Municipales 2013,4 while their
distance from the capital city of each municipality is taken from Google maps (see descriptive statistics
in Panel B of table 2).
19
Table 2. Descriptive Statistics
Panel A: Candidates
Standard
Deviation
Mean
Vote
Color (1)
Education (2)
0.89
4.74
2.01
Number of
candidates
Gender (3)
International Experience
Awards
FAR
Members of Cabinet
Communist Party
Incumbent
Delegado Municipal
Delegado Provincial
311
156
92
82
31
191
168
222
67
Total
612
0.04
2.67
0.63
Percentage of
total
50.82
25.49
15.03
13.40
5.07
31.21
27.45
36.27
10.95
Panel B: Municipalities in Table 4 (4)
Standard
Deviation
Mean
% SB Graduates (5)
Distance to Havana km/100
Distance to Havana min/100
Total
0.75
3.91
2.74
0.08
3.05
2.16
88
Notes: (1) The color skin scale goes from 1 to 11, with 11
representing the darkest skin color
(2) Education: 0: less than high school, 1: high school, 2: college, 3:
graduate school
(3) Gender: 1: Male, 0: Female.
(4) Data for the municipalities used in the estimations. There are >=2
candidates/municipality, so the total number of observations is 347
(5) SB Graduates represents the fraction of the population 15-19 (in
2013) that graduated from "secundaria básica" between 2010 and
2013.
Sources: Official candidate biographies (Granma) and Organización
Nacional de Estadística e Información
8. The Variables
We examine whether a Deputy candidate’s participation in a prior multi-candidate local municipal
election affects the likelihood of winning more votes in a Deputy election where no candidate loses.
Does prior experience as delegado municipal, as compared to other types of experiences in the National
20
Assembly or in the provincial assembly, which lack multi-candidate competition, affect the vote shares
in the National Assembly Deputy elections?
We control for gender and race, collected from the candidates' biographies. The race variable
was assigned by the coders who looked at the pictures and by comparing the candidates to a skin color
chart (LAPOP Barómetro de las Américas 2012);5 in some specifications, we dichotomized this
variable using only black and white.
We control for level of education. The majority of representatives (500) have at least a
university degree and 120 have a post-graduate degree. We assigned a score of 0 to those with lower
than a high-school level of education, 1 for high-school only, 2 for university degrees, and 3 for higherthan-university degrees.
We distinguish with dummy variables those candidates who have performed some
internationalist service reported in the biography. These experiences include high-ranking roles such as
ambassador; international academic training and degrees; or development tasks ("misión
internacionalista") in various countries.
We identify whether an individual was awarded some special recognition, the most common of
which, the "Vanguardia Nacional" (National Vanguard), is awarded based on outstanding performance
in different realms.
We insert variables that describe a parliamentarian’s previous political or institutional
experience, such as having been (or being) part of the FAR, the “Revolutionary Armed Forces,” or
being (or having been) a member of the Cuban Communist Party (CCP). Within Communist Party
membership, we distinguish between the posts held by the members.
We code the previous experiences of the elected members of an assembly. We include
provincial delegates elected previously through a non-competitive election for the provincial assembly,
21
incumbents from the same National Assembly running for reelection, and municipal delegates who
have been through a multi-candidate municipal-assembly election.
To test our hypotheses, we run models with different specifications. The first model, with
district fixed effects, compares the vote share of the few individuals who run within the same district.
This model keeps constant the political, economic, demographic, and cultural features of each district
that might otherwise impact voting behavior and takes the following form:
VoteShareid = β0 + β1DelegadoM i + β2 DelegadoPi + γ ' X i + ρd + ε id
(1)
where VoteShareid is the candidates’ vote share in their own district, DelegadoMi refers to delegados
€
municipales and DelegadoPi refers to delegados provinciales, and are defined as described above. Xi is
a vector of control variables (including membership to the Communist Party).
A second model analyzes the vote share's determinants across all candidates (within provinces).
This model allows us to control for the district magnitude (i.e. the number of seats assigned to each
district), which ranges from 2 to 5. The median district is size 3, but the most common district
magnitude is 2. Given this distribution of district magnitudes, we operationalized this concept into a
dichotomous variable (Md) that identifies large districts as those where size is greater than 2.
VoteShareid = β0 + β1DelegadoM i + β2 DelegadoPi + β3 M d + β 4 M d × DelegadoM i + γ ' Xi + ρd + εip (2)
In the section testing the explanatory mechanism, we include two variables at the municipal
level. We calculate education as the number of junior high-school graduates (grade 7th to 9th) who
graduated between 2010 and 20136 as a proportion of the population between the age of 15 and 19 for
2013 in a municipality (SBm).7
We measure distance of any municipality to the centrally located Plaza de la Revolución in the
City of Havana, calculating from Google maps distance in kilometers and in minutes of driving; the
latter measures the challenges of transportation (Havanam, measured in hundreds of kilometers and
22
hundreds of minutes).
The dependent variable is the individual candidate vote share across all districts. Figure 2 shows
its distribution, which goes from a minimum of .67 to a theoretical maximum of 1. The majority (95%)
of the candidates receive a vote share between 80 percent and 97 percent. We estimate the following
models:
VoteShareid = β0 + β1DelegadoM i + β2 DelegadoPi + β3 DelegadoM i × Havanam +
γ ' X i + ρd + ε ip
(3a)
VoteShareid = β0 + β1DelegadoM i + β2 DelegadoPi + β3 DelegadoM i × SBm +
+β4 DelegadoM i × Havanam + γ ' X i + ρd + ε ip
(3b)
VoteShareid = β0 + β1DelegadoM i + β2 DelegadoPi + β3 DelegadoM i × SBm +
β4 DelegadoM i × Havanam + β5 DelegadoM i × SBm × Havanam + γ ' X i + ρd + ε ip
(3c)
€
€
€
Figure 2. Distribution of vote share. The vertical lines represent one and two
standard deviations from the mean of the distribution
23
9. Does Experience Matter?
In Figure 3, we show the distribution of vote share for candidates who had been elected delegados
municipales in multi-candidate election only, those who were delegados provinciales only, and those
who were never elected in any election. In general, experience in electoral politics pays off. The
difference between those who had previous experience and first-time candidates is substantively
important and statistically significant at 1%, which represents a 1/4 of a standard deviation of the
distribution of votes.
In Table 3, we test the claim that the prior experience of having won a multi-candidate local
election, defeating another candidate for the same post, and being elected delegado municipal,
increases the vote share of a Deputy candidate. The coefficient on the dummy variable delegado
municipal is systematically larger and more significant statistically than other explanatory variables
across all specifications.
6
4
2
0
Density
8
10
12
Delegado Municipal only
Delegado Provincial only
New to Electoral Politics
0.75
0.80
0.85
0.90
0.95
Vote Share
Figure 3. Distribution voter share of delegados provinciales, municipales and new candidates.
24
In the first two columns, we control for district fixed effects. In the first column, we compare
only the different previous political experiences known to the voter. Previous experience as an elected
municipal assembly delegate plays the most important role, while being an incumbent is helpful. In the
second column, we test for all the information known to the voter through the biographies posted at the
voting precinct. Previous contested election as a municipal assembly delegate remains significant and
most important (Figure 4 shows the coefficients for the different political experience variables). Voters
also reward candidate merit, such as winning a workplace award and some international experience.
Incumbent advantage vanishes, and none of the other political or demographic variables matter. In the
third and fourth columns, we do not control for district dummies but only for provincial dummies to
verify the relationship between the vote share and the number of seats assigned in the different districts
(M); we cluster the standard errors at the district level. Larger district magnitude is negatively
correlated with the average vote share received by the candidates: votes are dispersed between more
candidates, and other variables matter less, not surprisingly.
Membership in the Communist Party, being a government minister or a military officer, or
having been elected in noncompetitive provincial assembly elections, was never statistically
significant. That is, voters sometimes supported candidates with such traits but just as often they did
not. No demographic or socio-economic variable has any predictive power. The clearest finding is that
candidates who have previously won a local contested multi-candidate election are also likely to obtain
more votes in a National Assembly election.
25
Table 3. Experience in Electoral Politics
Dependent Variable: Candidate's Vote Share
(1)
(2)
(3)
(4)
Delegado Municipal
0.00914***
(0.003)
0.0115***
(0.004)
0.00519
(0.004)
0.0121*
(0.007)
Delegado Provincial
0.00331
(0.006)
0.00440
(0.006)
0.00500
(0.005)
0.00518
(0.005)
Incumbent
0.00855**
(0.004)
0.00490
(0.004)
0.000960
(0.004)
0.000896
(0.004)
Communist Party
0.00533
(0.004)
0.00585
(0.004)
0.00347
(0.003)
0.00346
(0.003)
Ministro
0.00464
(0.008)
-0.000470
(0.008)
0.00878
(0.006)
0.00800
(0.006)
Color
0.0000724
(0.001)
0.000132
(0.001)
0.000162
(0.001)
Gender
-0.00125
(0.004)
-0.00149
(0.003)
-0.00111
(0.003)
Education
-0.00154
(0.003)
-0.00272
(0.002)
-0.00276
(0.002)
International Experience
0.00791*
(0.004)
0.00461
(0.003)
0.00437
(0.003)
FAR
0.00804
(0.006)
0.00544
(0.004)
0.00552
(0.004)
Awards
0.0108**
(0.005)
0.00514
(0.004)
0.00476
(0.004)
-0.0190***
(0.004)
-0.0148***
(0.005)
M>2
M>2 * Delegado Mun.
District FE
Province FE
Observations
-0.0113
(0.007)
Yes
No
612
Yes
No
612
No
Yes
612
No
Yes
612
Standard errors in parentheses
Errors in columns 3 and 4 are clustered at the district level.
* p<0.10, ** p<0.05, *** p<0.01
26
Figure 4. Premium in vote share from having been a municipal delegate, a provincial delegate, or a
National Assembly incumbent. These coefficients are taken from the full model with district fixed
effects in column 2 from table 3.
10. The Grassroots Premium: Information Short-Cut or Clientelism?
Grassroots links thus impart a premium to National Assembly candidacies. Is that because municipal
officials who subsequently run for National Assembly Deputy succeed at vote buying and voter
coercion? Or is it because voters recognize candidates with local links more readily and value them for
the local engagement — grassroots links as an information short-cut for quality candidates? We show
that the information short-cut explanation works.
Cuba’s authoritarian election context makes available little political information. In
municipalities peopled with less well-educated individuals who are less capable of being exposed to
wider political information, voters are more likely to be aware mainly of the local career of National
27
Assembly Deputy candidates. In contrast, in municipalities with highly educated people or in areas
closer to the capital city of Havana, it is easier to obtain alternative sources of information about
candidates and to have been exposed more to the national candidates who work in Havana. This
implies that the grassroots National Assembly election premium drops in these instances of high
education and proximity to Havana. The observable implication is a positive relationship between
longer distance and higher vote share for municipal delegates, and a negative relationship between vote
share and higher education of voters. In this respect, education and distance may be substitutes in the
process to acquire information.
Alternatively, a clientelistic explanation of the premium imparted by grassroots links to
National Assembly candidates focuses on the ability of local politicians to access patronage, pressure
voters, and buy votes everywhere. Note, however, that budget revenues for all Cuban local
governments in 2013 were just $670 million dollars; most of these funds were transfers from the
national government for budgeted activities (computed from Organización Nacional de Estadística e
Información, 2013, Table 6.5). If clientelism were the correct explanation, we should not observe any
relationship between the vote for candidates with previous municipal assembly experience and distance
to Havana, once we show that such distance has no relationship (for up to 99th quantile of the data) with
education, which is indeed the case (see Figure 5). Alas, the closer to Havana, the greater the relative
access to clientelistic resources; thus, if clientelism were at work, closeness to Havana should impart an
even greater grassroots National Assembly election premium, which we will show it does not.
A first step is to show (Figure 6) that there is a systematic increase in vote share received with
an increase in distance from the municipality to Plaza de la Revolución, the headquarters of the
Communist Party in Havana. In this non-parametric representation that does not assume any
relationship between the data, the increased difference between municipal delegates and non-municipal
delegates in their National Assembly candidacies is clear up until approximately 350 kilometers from
28
Havana: municipal delegates get more votes in such National Assembly elections if they are far from
Havana. This figure confirms hypothesis 6.b while rejecting 6.a.
Figure 5. Relationship between distance and education.
There is no relationship except for last three observations
outlier.
●
●
●
●
0.93
●
Municipal
delegates ●
●
●
Non−municipal
delegates●
●
●●
●
Distribution
of distance
●
●
●
0.92
●
0.90
0.91
●
●●●
●
●
●
●
●
●●
0.89
0.88
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
0.87
●
●
● ●
●
● ●
●
●
●
●
●
●
●
●
●
●
Vote share
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
0●
●
200
●
●
●
400
●
●
600
800
1000
●
●
Distance in Km from Havana
Figure 6. Distance from Havana and voter share
29
A more rigorous analysis, with district fixed effects, confirms that municipal delegates have a
significant National Assembly election advantage in areas far from Havana. The evidence displayed in
Figure 7 shows how having been a delegado municipal is positively correlated with vote share. This
correlation increases with distance from Havana and is statistically significant only in districts that are
at least 250 kilometers from the capital; in the districts close to Havana delegados do not appear to
have any advantage (see Table A1 in the appendix, columns 1 and 2). This is consistent with Figure 6.
In hypothesis 6c we argued that, if it is information and not clientelism what gives delegados
municipales an advantage, then in areas with high levels of education we should find a lower premium
for delegados, even if far from Havana. Education would compensate for the distance from the center
of government. Figure 8 shows the marginal effect of being a delegado in districts with low levels of
education (the municipalities in the bottom quartile), conditional on distance. Here the marginal effect
of being a delegado increases in distance from Havana, but it is less steep than in Figure 7 and only
significant at the 90% level for distances greater than 600 kilometers (the effect is no longer significant
after about 800 kilometers). For districts with higher levels of education, the premium increases as
distance to Havana increases, but the effect is no longer significant (see Table A1 in the appendix,
columns 3 and 4).
Finally, Figure 9 shows the marginal effect of being a delegado, for different levels of education
and conditional on distance to Havana. As education increases the effect of being a delegado decreases
(see the graph on the right), and the effect is only significant at the 90% level for areas with low
education (left graph) (see Table A1 in the appendix, columns 5 and 6).
30
Marginal Effect of Delegado Municipal on Vote Share
Marginal Effect of Delegado Municipal
-.01
0
.01
.02
.03
.04
VoteShare=f(Delegado Mun., Distance Km)
0
2
4
6
Distance to Havana (km)
Marginal Effect of Delegado Mun.
Lower Bound 95% CI
8
10
Upper Bound 95% CI
Figure 7. Marginal Effect of Being a Delegado Municipal, Conditional on Distance
from Havana (in hundreds of kilometers). Corresponds to the results from Column 1
in Table 4.
Marginal Effect of Delegado Municipal on Vote Share
Marginal Effect of Delegado Municipal
-.01
0
.01
.02
.03
VoteShare=f(Delegado Mun., Distance Km, Education)
0
2
4
6
Distance to Havana (km)
Effect of Del. for Grads at 25th Percentile
Lower Bound %90 CI
8
10
Upper Bound %90 CI
Figure 8. Marginal Effect of Being a Delegado Municipal for Districts at the 25th
Percentile (and below) of Education, Conditional on Distance to Havana (in
hundreds of kilometers). Corresponds to Column 3 in Table 4.
31
Marginal Effect of Delegado Mun. on Vote Share
Distance and Levels of Education
Distance and Levels of Education
0
Marginal Effect of Delegado Mun.
-.02
0
.02
.04
Marginal Effect of Delegado Mun.
.005
.01
.015
.02
Marginal Effect of Delegado Mun. on Vote Share
1
2
3
4
5
6
7
Distance to Havana (km)
SB Grads at 25th Percentile
Lower Bound %90 CI
8
9
10
Upper Bound %90 CI
1
2
3
4
5
6
7
Distance to Havana (km)
SB Grads at 25th Percentile
SB Grads at 75th Percentile
8
9
10
SB Grads at 50th Percentile
Figure 9. Marginal Effect of Being a Delegado Municipal for Districts at the 25th Percentile (left)
and for Districts at the 25th, 50th and 75th Percentiles (right) of Education, Conditional on Distance to
Havana (in hundreds of kilometers). Corresponds to Column 5 in Table 4.
11. Clientelism through electoral fraud?
Clientelistic networks may also work through electoral fraud. In some countries, stronger networks of
clientelism and patronage translate into vote buying on Election Day; in Cuba, we lack the means to
study vote buying or ballot stuffing. However, electoral fraud may happen through other methods
(Lehoucq, 2003), one of which is the direct manipulation and inflation of the reported electoral results.
Municipal delegates might have better connection with local bureaucrats and polling station officials
and, hence, have more chances to manipulate the electoral results to report higher percentage votes.
This assumes that district officials report the proportions and the count; candidates would have the
opportunity to manipulate the reported vote shares.
We looked for evidence of electoral manipulation in our data. We follow the example of work
done on Iran by Bernd Beber and Alexandra Scacco (Beber and Scacco 2009; 2012). They investigate
the distribution of the last digits of vote counts reported. These should follow a uniform distribution.
Their research relies on psychological studies to find and test specific biases in human-made numbers
32
that exist even when individuals try to reproduce randomness. These studies show that humans are not
good at reproducing randomness because they favor certain numbers, produce pairs of adjacent
numbers more often, and repeat numbers less often than we would observe by chance (Boland &
Hutchinson, 2000).
Our paper differs from these other studies because the only available public data is the vote
share for each candidate, on which we focus to study possible reported vote manipulation. The vote
share is provided for five decimal digits. If the goal of electoral fraud is to show better results, then the
reported vote share will be manipulated accordingly. National leaders and specific candidates will
produce vote shares different from those actually reported; they may in order to look better.
We check whether the distribution of the last digits is significantly different from a uniform
distribution and whether the number of equal or adjacent numbers is different from what would be
expected to be observed by chance. In addition, we study the relationship between average digits and
candidate features with OLS regression. Given that vote share differs from a vote count, we focus on
the numbers that we expect could be changed to show better results, namely, the second, third and
fourth decimal, while avoiding the last reported decimal which could be affected by different levels of
precision and choice of rounding. So in a proportion of 98.352% (that is, 0.98352) we focus on the 8, 3,
and 5.
In figure 10 we show that the distribution of the decimals (2nd, 3rd, and 4th) is not distinguishable
from a uniform distribution.8 The p values reported in the picture represent the probability of observing
that this distribution of decimals might come from a uniform distribution. None of the decimals
considered have distributions that appear “suspicious” of human manipulation. Similarly, in Figure 11
we show that we cannot detect any significant deviation from the expected mean (red horizontal line)
of any province by examining the mean of the second and third decimal across provinces. The plotted
33
confidence intervals for each province show that there is no significant difference across provinces and
or the expected mean.
Figure 10. Distribution of the second, third and fourth decimal of
candidates’ vote aggregate at the national level.
34
Figure 11. Mean decimal across the different provinces. The horizontal line is the
expected mean of 4.5.
Given that there is no systematic manipulation of the publicly reported vote, we check whether
municipal delegates make for a systematic difference compared to the other candidates. Are the
municipal delegates presenting better results for themselves thanks to their local connections? Are they
fixing their vote share to make their results more favorable? We find no evidence of this. As shown in
Figure 12, the distribution of decimals for municipal delegates follows a uniform distribution as much
as that of non-municipal delegates. A more systematic test of this difference is provided by Table 4. All
test parametrically whether there are systematic differences between municipal delegates and other
candidates. The OLS regressions check whether, within districts, the municipal delegates tend to have
higher second or third decimals, accounting for their level of success (vote share). The third column
also tests whether there is a significant relationship between the third and second decimal digits to
inflate the vote share for municipal delegates and finds no evidence of that. The last two columns (4
and 5) simply compare with the actual vote share.
35
Table 4. OLS regressions to show no clear tendency to have higher decimals in
the vote share reported for municipal delegates compared to those reported for
other candidates.
36
Figure 12. Distribution of decimal digits of the vote shares of municipal delegates.
We also seek to verify whether the rate of adjacent numbers between decimal digits and the rate of
repetition of decimal digits are compatible with rates produced by a random distribution of digits. We
compare the second and third decimal digits and the third and fourth. In Figure 13 we report the rates of
adjacencies and repetitions in our dataset and compare them with the proportions of adjacent and
repeated digits in 1000 datasets with 612 pairs of numbers, each simulated from uniform distributions.9
The Figure 13 shows that the values found in our dataset are compatible with a random distribution.
We remain unable to detect any manual manipulation of the reported vote shares.
37
Figure 13. Proportion of pairs with adjacent numbers and repetition. The
distribution of proportions plotted here comes from 1000 datasets with 612 pairs
of numbers, each simulated from a uniform distribution (with replacement). The
proportion of adjacent and repeated number in our dataset is reported relative to
this distribution.
We run the same test on municipal delegates. We compare the proportion of adjacent and equal
pairs between municipal and non-municipal delegates. In Figure 14 we find no systematic difference
from the expected proportion from the simulation and also no systematic difference between the two
types of candidates (all differences tested with a t.test delivered a p value larger than .10).
38
Figure 14. Proportion of adjacent and equal number in the pairs of digits
considered in our dataset. Comparison between municipal delegates and nonmunicipal delegates. On the X axis we plotted the dummy variable equal to 1
for municipal delegates.
In sum, this section has found no evidence of manipulation of the vote shares reported to the public and
no such systematic manipulation done by municipal candidates who run for the National Assembly.
12. Unpacking the CCP influence (or lack thereof)
Simple membership to the communist party is not a good predictor of vote share. Candidates do not get
more or less votes based on that. The Communist party matters greatly; 191 of the 612 candidate
39
biographies mention party membership. Of these, 89 have been delegates to the party’s national
congress, 23 serve or have served in the party’s political bureau, and 56 serve on the party’s Central
Committee; the rest held various roles at the provincial or municipal level.
In this section, we explore further in order to distinguish roles within the party and find two that
send a stronger signal of quality or attachment to the territory than mere membership. Table 5 shows a
variable for high-ranking national level positions (NationalPartyi, party central committee and political
bureau) and another for local-level party posts (LocalPartyi, party municipal committee or bureau). We
control for district fixed effects in all specifications. We unpack the influence of the party, that is,
between highly visible (local and national) and less-visible party leaders, and ascertain whether this
distinction has an impact on the vote share apart from being or having been a municipal elected official.
By controlling also for party roles, we also verify that this party-role variable does not cancel
the significance of the delegado municipal variable. We estimate the following model:
VoteShareid = β0 + β1DelegadoM i + β2 DelegadoPi + β3 LocalParty i + β4 NationalParty i +
γ ' X i + ρd + ε id
(4)
Here Xi is a vector of control variables and does not include membership to the Communist
€
Party.
Several key findings survive this test (see Table 5 and Figure 10) across all specifications. Prior
election in a multi-candidate contest at the local level remains statistically and substantively significant,
contributing to a higher vote share. Earning awards for good work also has a positive impact on vote
share, as does having had international experience. Demographic factors are not significant, nor having
been elected in noncompetitive provincial assembly elections.
We interpret these finding as evidence that the support for the communist party is not a blind
sign of loyalty or fear, but it is conditional on the visibility and connection of these candidates to
40
specific territories. The local or the national party variables are always positive; voters are not rejecting
communist party candidates but they are making distinctions between them. In such a low-information
environment, the party members at the middle ranks are less well known to the voters, which is why the
table 3 coeff column 4
table 5 coeff column 6
0.02
●
0.01
●
0.00
●
−0.01
Vote Share Premium
0.03
0.04
local-level or the national-level candidates earn a higher vote share.
CPP member
CPP Local Role
CPP National Role
Figure 15. Coefficients of generic CPP member and from local or national CPP
member from different models. The bars represent a .1 confidence interval of
significance
41
Table 5. Unpacking the CCP
Dependent Variable: Candidate's Vote Share
(2)
(3)
(4)
(5)
(1)
(6)
Delegado Municipal
0.00808**
(0.003)
0.0110***
(0.004)
0.0105***
(0.003)
0.0121***
(0.004)
0.0104***
(0.003)
0.0123***
(0.004)
Delegado Provincial
0.00430
(0.006)
0.00597
(0.006)
0.00488
(0.006)
0.00550
(0.006)
0.00552
(0.006)
0.00652
(0.006)
Local Level Party Official
0.0145**
(0.006)
0.0182***
(0.006)
0.0115*
(0.006)
0.0157**
(0.006)
National Level Party Official
Incumbent
0.0106***
(0.004)
0.00603
(0.004)
0.0194***
(0.006)
0.0157**
(0.007)
0.0173***
(0.007)
0.0122*
(0.007)
0.00460
(0.004)
0.00244
(0.005)
0.00515
(0.004)
0.00294
(0.005)
Color
-0.0000177
(0.001)
-0.0000434
(0.001)
-0.0000206
(0.001)
Gender
-0.00170
(0.004)
-0.00142
(0.004)
-0.00189
(0.004)
Education
-0.00171
(0.003)
-0.00154
(0.003)
-0.00184
(0.003)
International Experience
0.00859**
(0.004)
0.00709*
(0.004)
0.00802*
(0.004)
FAR
0.0101*
(0.006)
0.00603
(0.006)
0.00735
(0.006)
Awards
0.0114**
(0.005)
0.00983*
(0.005)
0.0112**
(0.005)
District FE
Observations
Yes
612
Yes
612
Yes
612
Yes
612
Yes
612
Yes
612
Standard errors in parentheses
* p<0.10, ** p<0.05, *** p<0.01
42
Conclusion
On election day, Cuban voters remember that they live in an authoritarian political regime yet they
understand that they can exercise some choice. We focus on selective voting in a single-party election
where the number of candidates equals the number of seats but voters may vote for some but not all
candidates on the ballot.
Consider instead the standard skepticism regarding elections in communist political regimes,
which the following quotation reveals: “A few questions are still hotly debated among students of the
[Cuban] political system, but certainly the nature of [Cuban] elections is not one of them. Everyone
agrees that they are more interesting as a psychological curiosity than as a political reality.” This
opened a 1968 article on Soviet elections published in The American Political Science Review (Gilison,
1968, p. 814). In the brackets above, we substituted “Cuban” for “Soviet” because the same dismissive
lack of study has applied to Cuba's national elections. In this article, we demonstrate that Cuban
elections already matter.
We considered two puzzles: why do citizens vote in a context where their choice has no impact
on the outcome and why would they select some candidates over others? Our evidence addresses the
second puzzle. We show that voters express real preferences by rewarding candidates for either their
grassroots links or some merits. They act based on the information available to voters about the
candidates; in 2013, the voters were not mobilized to vote for the single official slate nor were they
responding to clientelistic relationships. We find that Cuban voters value political experience and
accomplishment in their politicians and, in this very low political information context, they use
information short-cuts to find candidates to support. In particular:
• Voters are likely to reward those National Assembly candidates who had first proven
43
themselves by winning a prior multi-candidate local election where some candidates win and
other candidates lose (hypothesis 4)
• Voters are likely to reward those who get things done, that is, they have earned awards or, less
markedly, performed international service (hypothesis 5).
• In areas with fewer well educated voters, or lower likelihood of exposure to national politicians,
the grassroots links matter the most, and these are developed through multi-candidate local
electoral competition and constituency responsiveness (hypothesis 6 a, b, and c).
In addition, demographic variables are never statistically significant in the analysis of vote shares
for individual candidates. Serving as government minister is also statistically insignificant, as is serving
as a military officer in most specifications. Voters ignore prior entirely-noncompetitive election to
provincial assemblies in their assessments of candidates. There is no incumbent advantage in most
specifications in standing for re-election, and mere affiliation with the communist party has no
statistically significant impact on vote shares — the first three hypotheses are rejected.
Other negative findings are important as well. Clientelistic favors and fear or blind loyalty for
the regime are not the main incentive when voting. Nor did we find evidence of fraudulent reported
vote manipulation. Our findings are consistent, instead, with an information mechanism: voters prefer
delegados municipales over other candidates especially in areas far from Havana and with low average
education where local candidate visibility is their most rewarding feature.
These findings, in turn, shed some light on the first puzzle, namely, why would voters even
bother to vote selectively in this context. Many voters seek to send a signal to national leaders,
especially about candidates more than about the communist party. This selective vote is more common
in more educated areas and closer to the Capital City where voters inform political elites without
punishing or rejecting the communist party. Furthermore, as our forthcoming work will show, selective
44
voting prevails in provinces other than from those where blank voting is most frequent; it can be argued
that blank voting, not selective voting, is a clearer sign of political protest and rejection of the ruling
party. Yet, blank and void voting actually declined between the 1993 and 2013 elections.
In turn, this permits two speculations, and a call for further research, regarding Communist
Party and government motivations to forego the mobilization strategy employed in past National
Assembly elections to permit a wider use of the selective vote. First, the selective vote implies greater
acquiescence to the political regime than the protest-marked blank vote. And, second, the selective vote
permits national leaders to ascertain the relative popularity of cadres, especially that of the municipal
delegates running for National Assembly Deputy (for a similar motivation in China, allowing some but
not all forms of expression on the Internet, see King, Pan & Roberts, 2013).
If the data were to become available, further research should explore why being a municipal
delegate is so helpful to win a higher share of the vote. Is it name recognition that comes from having
previously won a local multi-candidate election? Is it a “friends and neighbors” familiarity in small and
especially remote districts, unrelated to elections? Is it that these local delegates reside in the districts
where they run for Deputy in contrast to the higher officials who do not live in the district? The data
currently available do not permit us to sort these interpretations.
From the perspective of the continuation of an authoritarian regime, some news is good. Voters
do not rush to support newcomers whom they do not know. Voters do not repudiate Communist
national leaders. Voters are likely to support local communist party officials or national communist
party officials at the expense of the more opaque middle party ranks. This is why we had to unpack the
implications of our third hypothesis. Voters accept the validity of officially-bestowed awards and
officially-sponsored international service, which have been important policies across the decades. And
the 1992 electoral reform was productive to support the political regime because voters value the
capacity to send their favored local officials as Deputies to the National Assembly.
45
For the proponents of change, some news is also good. Voters reward those who submit their
candidacy to voters where there is such competition, namely, at the local level. The reward for the
award-winners also implies support for merit, not just for power-holding. The lack of substantial
support for Government Ministers also augurs better for change. Further exposure to information (more
education, proximity to Havana) throws an election more open. The local-competition and the awards
variables do better than, or as well as, any of the party leadership variables when all factors are
considered.
Yet, there are unanswered questions. Does the 1992 electoral reform strengthen the political
regime because it worked as intended, or is it a political change that rulers will be unable to control if
nonconformist voting continues to grow? Is the lack of support for electoral newcomers, evident in a
low-information authoritarian regime, likely to be reversed and newcomers thus be welcomed if there
is a further political opening? Should the veterans on the Political Bureau give way to politicians with
local electoral experience? Will the reward for achievement extend to those who succeed in Cuba’s
emerging private sector?
To paraphrase the opening line of The Communist Manifesto, there is a specter haunting Cuba.
It is the specter of political change. Neither regime supporters nor regime detractors have a good handle
on the trajectory of change.
46
APPENDIX
Table A1. Investigating the mechanisms: education and distance10
(1)
Distance in
Km/100
Dependent Variable: Candidate's Vote Share
(2)
(3)
(4)
(5)
Distance in Distance in Distance in Distance in
Min/100
Km/100
Min/100
Km/100
(6)
Distance in
Min/100
Delegado Municipal
0.00617
(0.006)
0.00642
(0.006)
0.0360
(0.034)
0.0359
(0.035)
0.00365
(0.067)
0.00504
(0.069)
Distance X Delegado Mun.
0.00130
(0.001)
0.00176
(0.002)
0.00146
(0.001)
0.00199
(0.002)
0.00669
(0.008)
0.00900
(0.011)
-0.0477
(0.048)
-0.0473
(0.048)
-0.00473
(0.090)
-0.00640
(0.092)
-0.00680
(0.009)
-0.00908
(0.013)
Delegado Mun. X % of SB Graduates
Distance X % of SB Graduates X Delegado Mun.
Communist Party
0.00705
(0.006)
0.00702
(0.006)
0.00533
(0.007)
0.00529
(0.007)
0.00518
(0.007)
0.00514
(0.007)
Delegado Provincial
0.00456
(0.007)
0.00457
(0.007)
0.00315
(0.008)
0.00313
(0.008)
0.00330
(0.008)
0.00330
(0.008)
Incumbent
0.00347
(0.005)
0.00350
(0.005)
-0.00345
(0.006)
-0.00337
(0.006)
-0.00379
(0.006)
-0.00371
(0.006)
District Fixed Effects
Observations
Yes
606
Yes
606
Yes
347
Yes
347
Yes
347
Yes
347
Standard errors in parentheses
Errors clustered at the district level
District Fixed Effects
Candidate controls in all columns: color, gender, education, international experience, FAR, Minister, awards
* p<0.10, ** p<0.05, *** p<0.01
47
NOTES
1
For the law in effect for the 2013 election, "Ley Electoral" 2005; for a sympathetic account, August
2013, chapter 7; for a critique published in Cuba, Rafuls Pineda 2014.
2
Municipalities and districts are not necessarily the same. More populous municipalities are divided in
multiple districts. Smaller municipalities usually represent a single district.
3
The data are available upon request to the authors; please email corresponding author. The data will
be made public upon publication.
4
Available at http://www.one.cu/aedm2013.htm.
5
We used the same chart used by the 2012 LAPOP (Latin American Public Opinion Project) survey. It
has a scale of 11 skin colors, which is used to classify the respondents of the survey.
6
This measure represents that fraction of first time voters or young voters (ages 15 to 19) that finished
their basic education by the time of the election; it was the only one available across the highest
number of municipalities.
7
A few municipalities are assigned a value that is actually larger than one. This could be due to a
mistake in the data or to the fact that a large group of students graduating in those years were actually
older than their cohort. It is impossible for us to distinguish between the two so we keep the value as
they are.
48
8
Our paper focuses on the vote share. To test whether it is plausible to expect the decimal of proportion
to be uniformally distributed, we simulate 100 normal distributions with mean and standard deviation
equal to that of vote share in our dataset; we check the distribution of their decimal digits. We found
that 92% of the distributions of the second-decimal digits are uniform (p value of the chi-square test
larger than 0.05), as are also 96% of the third-decimal digit distributions, and 94% of the fourthdecimal digit distributions.
9
We simulated 612 pairs of numbers from uniform distributions and we calculated the proportions of
pairs that have adjacent numbers (n, n+1 or n-1) and repeated numbers (n, n). We did this same
exercise 1000 times and plot each proportion in Figure 3 to create a distribution of reference for our
proportions in our dataset.
10
The estimations in this table use the aggregate data of graduates for the years 2010-2013. The
denominator in all the calculations for the education variable is the population age 15-19 in 2013. The
estimations using data from each of the four years (2010, 2011, 2012, 2013) independently are
available upon request.
49
REFERENCES
Aldrich, J. (1993). Rational Choice and Turnout. American Journal of Political Science 37: 246-78.
Amaro, N. (1996). Decentralization, Local Government, and Citizen Participation in Cuba. Cuba in
Transition (Association for the Study of the Cuban Economy): 262-282.
August, A. (2013). Cuba and Its Neighbours: Democracy in Motion. New York: Zed Books.
________ (2014). Democracy Still in Motion: The 2013 Election Results in Cuba. International Journal
of Cuban Studies 61 (Spring): 87-94.
Beber, B., & Scacco, A. (2012). What the numbers say: A digit-based test for election fraud. Political
Analysis, mps003.
__________________ (2009, June 20). The Devil Is in the Digits: Evidence That Iran's Election Was
Rigged. The Washington Post.
Blais, A., Gidengil, E., Dobrzynska, A., Nevitte, N., & Adeau, R. (2003). Does the local candidate
matter? Candidate effects in the Canadian election of 2000. Canadian Journal of Political
Science/Revue canadienne de science politique, 36(03), 657-664.
Blaydes, L. (2011). Elections and Distributive Politics in Mubarak’s Egypt. New York: Cambridge
University Press.
50
Boland, P. J., & Hutchinson, K. (2000). Student selection of random digits. Journal of the Royal
Statistical Society: Series D (The Statistician), 49(4), 519-529.
Brownlee, J. (2011). Executive Elections in the Arab World: When and How do they Matter?
Comparative Political Studies 44(7): 807-828.
Castro, R. (2013, February 25). La mayor satisfacción es la tranquilidad y serena confianza que
sentimos al ir entregando a las nuevas generaciones la responsabilidad de continuar construyendo el
socialismo. Granma. http://www.granma.cubaweb.cu/2013/02/25/nacional/artic04.html accessed
2/25/2013.
¿Cómo
votar
este
domingo
3
de
febrero?
(2013,
February
1).
Granma.
http://www.granma.cubaweb.cu/2013/02/01/nacional/artic08.html
Concluyó entrevista on-line sobre el sistema electoral cubano. (2015, February 28). Juventud Rebelde.
http://juventudrebelde.cu/cuba/2015-02-28/que-desea-saber-sobre-el-sistema-electoral-cubano accessed
03/09/15.
De la Dirección. (2008, June 27). Granma. http://www.granma.cubaweb.cu/secciones/cartasdireccion/art-009.html accessed 6/27/08
Dilla, H., Núñez, G. G., & Vicentelli, A. T. (1990). Participación y desarrollo en los municipios
cubanos. Caracas: Fondo Editorial Tropykos.
51
Domínguez, J. I. (1978). Cuba: Order and Revolution. Cambridge: Harvard University Press.
Downs, A. (1957). An economic theory of democracy. New York: Harper.
Elector-planteamiento-delegado: una relación que no termina ahí. (2015, March 29). Granma.
http://www.granma.cu/elecciones-en-Cuba-2015-03-29/elector-planteamiento-delegado
Gandhi, J., & Lust-Okar, E. (2009). Elections under authoritarianism. Annual Review of Political
Science 12: 403-422.
Gilison, J. M. (1968). Soviet Elections as a Measure of Dissent: The Missing One Percent. The
American Political Science Review 62: 814-826.
Greene, K. (2007). Why Dominant Parties Lose. Cambridge University Press.
Guanche, J. C. (2012). La participación ciudadana en el Estado cubano. Temas 70: 69-79.
Las próximas elecciones reafirmarán el apoyo del pueblo a la Revolución. (2015, March 27). Granma.
http://www.granma.cu/cuba/2015-03-27/las-proximas-elecciones-reafirmaran-el-apoyo-del-pueblo
Lehoucq, Fabrice. "Electoral fraud: Causes, types, and consequences." Annual review of Political
Science 6.1 (2003): 233-256.
Karklins, R. (1986). Soviet Elections Revisited: Voter Abstention in Noncompetitive Voting. The
52
American Political Science Review 80: 449-470.
Key, V. O., Jr. (1949). Southern Politics in State and Nation. New York: Alfred A. Knopf.
King, G., Pan J., & Roberts M. (2013). How Censorship in China Allows Government Criticism but
Silences Collective Expression, The American Political Science Review 107(2): 326-343.
Levitsky, S., & Way L. A. (2010). Competitive Authoritarianism: Hybrid Regimes after the Cold War.
New York: Cambridge University Press.
Ley electoral. (2005, November 26). Granma.
http://www.granma.cubaweb.cu/secciones/elecciones2005/art121.html accessed 11/26/2005.
Llama la CTC al voto unido. (2008, January 14). Granma.
http://www.granma.cubaweb.cu/2008/01/14/nacional/artic02.html accessed 1/14/2008.
López, R. (1993). Un domingo distinto. In Elecciones en Cuba: ¿farsa o democracia? Melbourne:
Ocean Press.
Magaloni, B. (2006). Voting for Autocracy. Cambridge University Press.
Malesky, E., & Schuler, P. (2009). Paint-by-Numbers Democracy: The Stakes, Structure, and Results
of the 2007 Vietnamese National Assembly Election, Journal of Vietnamese Studies 4(1): 1-48.
53
________ (2011). The Single-Party Dictator’s Dilemma: Information in Elections Without Opposition,
Legislative Studies Quarterly 35(4): 491-530.
Marrón González, K. (2015, February 19). ¿Quién quiere ser delegado? Granma.
http://www.granma.cu/opinion/2015-02-19/quien-quiere-ser-delegado
Mayoral, M. J. (2000, April 25). 98.06: Masiva respuesta electoral. Granma.
http://www.granma.cubaweb.cu/25abr00/nacional/articulo14.html accessed 06/25/2000.
Organización Nacional de Estadística e Información. (2014). Anuario estadístico de Cuba, 2013.
Havana. http://www.one.cu/aec2013/esp/17_tabla_cuadro.htm accessed 07/22/2015.
________ (2012). Censo de Población y Vivienda, Cuba, 2012. Havana.
http://www.one.cu/aec2013/esp/07_tabla_cuadro.htm ,
http://www.one.cu/aec2013/esp/20080618_tabla_cuadro.htm accessed 11/26/2014
http://www.one.cu/censo2012.htm accessed 11/26/2014.
________ (2014). Anuarios estadísticos municipales, 2013. Havana. http://www.one.cu/aedm2013.htm
Pérez V., Everleny, O. (2016). The Cuban Economy: Trajectory, Change and Perspectives. Cuban
Studies 44 (forthcoming).
Rafuls Pineda, D. (2014). El sistema electoral cubano: de la representación formal a la participación
real. Temas 78: 64-71.
54
Resultados finales oficiales de las elecciones. (2013, February 8). Granma.
http://www.granma.cubaweb.cu/2013/02/08/nacional/artic09.html
Staar, R. (1958). Elections in Communist Poland. Midwest Journal of Political Science 2(2): 200-218.
Shi, T. (1999). Voting and Nonvoting in China: Voting Behavior in Plebiscitary and Limited-Choice
Elections. Journal of Politics 61(4): 1115-1139.
Shugart, M. S., Valdini M. E., & Suominen, K. (2005). Looking for Locals: Voter Information
Demands and Personal Vote-Earning Attributes of Legislators under Proportional
Representation. American Journal of Political Science 49(2): 437-449.
Southwell, P. L., & Everest M. J. (1998). The Electoral Consequences of Alienation: Nonvoting and
Protest Voting in the 1992 Presidential Race. Social Science Journal 35(1): 43-51.
Tavits, M. (2010). Effect of Local Ties On Electoral Success and Parliamentary Behaviour The Case of
Estonia. Party Politics 16(2): 215-235.
Texto de la Ley No. 118 de la Inversión Extranjera. (2014, April 21). Granma.
http://www.granma.cubaweb.cu/2014-04-16/asamblea-nacional-del-poder-popular
accessed 4/21/2014.
Triunfó el voto unido. (2008, January 22). Granma.
55
http://www.granma.cubaweb.cu/secciones/elecciones/84.html accessed 1/22/2008.
Un paso más cerca de elegir. (2015, March 31). Granma.
http://www.granma.cu/elecciones-en-cuba-2015-03-31/un-paso-mas-cerca-de-elegir
Valdés Paz, Juan. 2009. El espacio y el límite: Estudios sobre el sistema político cubano. Panamá: Ruth
Casa Editorial.
56