Recall Elections: A Tool of Accountability? Evidence from Peru
| Autor | Michael Haman |
| Cargo | University of Hradec Králové, Hradec Králové, Czech Republic. Email: michael.haman@uhk.cz |
| Páginas | 73-111 |
73
desarro. soc. 71, primer semestre de 2013, pp. x-xx, issn 0120-3584
Revista
Desarrollo y Sociedad Primer cuatrimestre 2021
PP. 73-111, ISSN 0120-3584
E-ISSN 1900-7760
Recall Elections: A Tool of Accountability?
Evidence from Peru
La Revocatoria del Mandato: ¿una herramienta
de rendición de cuentas? La evidencia de Perú
Michael Haman1
DOI: 10.13043/DYS.87.3
Abstract
I examine the role of recall elections in Peruvian municipalities to evaluate
whether they are used as a tool to increase political accountability. I test the
theoretical assumptions of elite mobilization theory and other theories related
to the functions of political parties and invalid votes. I use logistic regression to
analyze a dataset with recall elections until 2013. The results show that recall
elections were held most frequently in municipalities with high local elec-
torate party membership when the percentage of blank votes was low, and
the margin of victory was low in previous municipal elections. The key vari-
ables for the successful removal of a mayor include the political experience
of the organizer of a recall procedure, the number of null votes, and votes for
the winner in previous municipal elections. Future research should build on
these findings and further examine recall elections.
Key words: accountability; democracy; elections; political parties; Peru .
JEL Classification: D72, H70, H72 .
1 University of Hradec Králové, Hradec Králové, Czech Republic. Email: michael.haman@uhk.cz
Este artículo fue recibido el 7 de septiembre de 2019, revisado el 17 de abril de 2020 y finalmente
aceptado el 14 de agosto de 2020.
87
Recall Elections: A Tool of Accountability? Evidence from Peru
74
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Resumen
Examino el papel de las revocatorias en los distritos peruanos para evaluar si se
utilizan como herramienta para aumentar la rendición de cuentas. Utilizo la regre-
sión logística para analizar un conjunto de datos con las revocatorias hasta 2013.
Pongo a prueb a los sup uestos teó ricos de la t eoría de la m oviliz ación de la é lite
y otras teorías relacionadas con las funciones de los partidos políticos y los votos
inválidos. Las revocatorias se celebraron con mayor frecuencia en distritos con un
alto número de miembros de los partidos en el electorado local, cuando el porcen-
taje de votos blancos fue bajo y el margen de victoria fue bajo en las elecciones
municipales anteriores. Las variables clave para la revocación de un alcalde son
la experiencia política de un promotor de la revocatoria, una cantidad de votos
nulos y los votos para el ganador en las elecciones municipales anteriores. Las
investigaciones ulteriores deberían basarse en estas conclusiones y examinar más
a fondo las revocatorias.
Palab ras clave : democracia; elecciones; partido político; rendición de cuentas; Perú .
Clasificación JEL: D72, H70, H72 .
Introduction
In this article, I study subnational-level recall elections in Peru. Recall elec-
tions are sometimes considered a mechanism of direct democracy. For example,
Altman (2011, p. 11) classified recall elections in his typology of the
mechanism of direct democracy as a subtype of popular initiative. However, in
Altman’s (2019, p. 6) later work, he excluded recall from the mechanisms of
direct democracy, considering that citizens do not vote on specific issues but,
rather, that recall targets the elected authorities. Kaufmann, Büchi and Braun
(2008, p. 91) hold a similar position to Altman in considering that recall is not
a mechanism of direct democracy. For the purpose of this study, I too position
recall elections as a special category of elections outside the realm of direct
democracy. The body of literature on recall elections is not particularly large
as these are a scarcely used tool in democracies. Despite this, there have been
some recent publications on the topic (Geissel & Jung, 2018; Holland & Incio,
2019; Welp & Milanese, 2018; Welp & Whitehead, 2020; Whitehead, 2018).
Michael Haman 75
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Recall elections give citizens more control over public administration. It is
they who initiate the process, which, in turn, is meant to serve as an insur-
ance mechanism against inert politicians and institutions (Altman, 2011,
p. 2), rendering such elections a tool of accountability. The greatest obstacle
to examining the impact of recall is the lack of data, as this tool is used in a
few countries and there is often little possibility to conduct a proper analysis.
Peru, however, given its twenty years of practice with local-level recalls, is an
ideal example for this research.
The central research question is: What variables can explain the use of recall
in Peruvian municipalities? The research problem is summarized as: Why is
recall used more in some municipalities and why is it used at all in light of the
fact that it brings high-level instability to the political system? A successful
recall results in shortening the term of office for a mayor or council member,
and there are two possible explanations why such action may be taken: the
first, in the case of Peru, is that recall has generally been used by the opposi-
tion as a mechanism to gain power (Holland & Incio, 2019; Tuesta Soldevilla,
2014; Welp, 2016). I have extended research on recall elections and tested
new variables that are linked to political parties and invalid votes, and cre-
ated a new variable that measures a mayor’s political performance. Hence,
the question of whether recall is used as a mechanism of accountability. In
order to examine these questions, I use a large dataset of different variables
from Peruvian municipalities and municipal elections to test and extend the
current theoretical assumption.
I find that recall elections were held most frequently in municipalities with
high local electorate party membership when the percentage of blank votes
was low, and the margin of victory was low in previous municipal elections.
The key variables for the successful removal of a mayor are the political expe-
rience of the organizer of a recall procedure, the number of null votes, and
votes for the winner in previous municipal elections. Except for the margin
of victory, these findings are a new contribution to the literature related to
political science. Even though these mentioned variables are not part of one
coherent theory, my main argument is that there is one thread that connects
them all. All these variables can potentially predict the success of mobilization,
which, in turn, serves as the thread that links all the variables in this research.
I explain the relationship between mobilization and each of these variables in
greater detail in the theoretical section of this work.
Recall Elections: A Tool of Accountability? Evidence from Peru
76
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
I examine the role of recall elections in Peruvian municipalities and I use
logistic regression to analyze a dataset with recall elections until 2013. I test
the theoretical assumptions of elite mobilization theory and other theories
related to the functions of political parties and invalid votes. Recall elections
were held most frequently in municipalities with high local electorate party
membership, when the percentage of blank votes was low and the margin of
victory was low in previous municipal elections. The performance of mayors and
socio-economic variables such as the Human Development Index, education,
and income, on the other hand, do not affect the likelihood of recall elections
or the removal of mayors. The key variables for the successful removal of a
mayor are the political experience of the organizer of a recall procedure, the
number of null votes, and votes for the winner in previous municipal elections.
In the first section of this work, I present a review of the literature, the theo-
retical framework, and the hypothesis of the paper. In the following section, I
explain the research design and methodology and, in the final section, I inter-
pret the results.
Theories and hypotheses
1) Party competition and membership
When election results are close, competitive elections increase voter turnout
(Blais, 2000; Blais & Dobrzynska, 1998; Caldeira & Patterson, 1982; Franklin
& Hirczy, 1998). Competitiveness mobilizes voters because citizens feel that
their vote is important, leading them to attend elections more frequently.
This argument is based on the theory of rational choice (Downs, 1957),
which assumes that rational voters choose to vote depending on the extent
to which their vote is key to the outcome of the election. In other words, they
believe that, in close elections, their vote may have a more decisive role than
it would if one party dominates the election.
Voters feel election closeness, as do political elites. In the case of close elec-
tions, there is mobilization on the part of the political elites themselves, in
an attempt to win the close elections. (Aldrich, 1993, 1995; Cox & Munger,
1989; Kirchgässner & Schulz, 2005; Rosenstone & Hansen, 1993). Elites invest
far more funds in campaigns and overall mobilization because they have a
Michael Haman 77
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
greater chance of winning. As such, it should be possible to predict the hold-
ing of recall elections by applying elite mobilization theory. It is clear from
previous research that politicians that lose the elections, often attempt to
recall a mayor and councilors of their seats. To do so they need to initiate2
new municipal elections to replace the winning party (Tuesta Soldevilla, 2014,
p. 21; Welp, 2016, p. 12). If the winner of the elections clearly dominates,
the defeated local political leaders should not have the incentive to mobilize
citizens to collect signatures for a recall to try to remove a mayor. If a mayor
is successfully removed, it is also less likely that he or she will win in the new
municipal elections. In a recently published article, Holland and Incio (2019)
confirm that recall is more often held in municipalities where a party wins
by a small margin. However, some key variables were not verified; i.e., party
membership that is closely linked to elite mobilization theory.
Political parties have several functions. One of these is that they mobilize
voters and their members. Often, party members themselves are responsible
for mobilizing voters (Scarrow, 1996; Scarrow & Webb, 2017), and this can
occur in various ways, whether through election campaigns, convincing, donat-
ing money or any other voluntary activity. Mobilizing within local election
campaigns has a key impact on voter turnout and election outcomes (Carty
& Eagles, 1999; Clarke, Sanders, Stewart, & Whiteley, 2004; Denver & Hands,
1997; Gerber & Green, 2000; Hillygus, 2005; Johnston & Pattie, 1997; Karp,
Banducci, & Bowler, 2008).
Peru’s levels of party institutionalization are very low. Indeed, Levitsky (2018)
considers that after the return of democracy in 2001, the political parties in
Peru have tended to be very weak. This is also true in terms of the country’s
party membership, which, as revealed by Došek (2016, p. 179) has the lowest
rates of all the countries in Latin America for which data are available. The
degree of membership in political parties naturally varies across Peru, and dis-
tinctions across municipalities around the country may raise new questions.
2 The new municipal elections (las nuevas elecciones municipales) are very important for the analysis
of the Peruvian recall elections. These by-elections could be invoked until 2013 if a sufficient number
of council members were recalled. The wording of Article 25 of Act No. 23600 is as follows: “Only
when more than one-third of the council is recalled, can new elections be held” (Oficina Nacional de
Procesos Electorales 2013b). In the event that more than one-third of the members of a council were
not removed but some politicians were recalled, substitutes from the party lists would be selected.
Currently, by-elections cannot be held as a law was changed in 2015 and substitutes replace recalled
politicians. I analyze the period until 2013 as I explain later.
Recall Elections: A Tool of Accountability? Evidence from Peru
78
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
This study compares these municipalities, showing significant differences
among them in terms of party membership and its impact on the mobilization
of the electorate. Even though party system institutionalization is low in Peru
(Levitsky 2018) and parties and movements tend to disappear very often, this
does not necessarily mean that party membership cannot offer information
and prediction of electorate behavior. Quite the opposite is true, given that
the mere fact that a higher share of the electorate is even interested in poli-
tics in the form of membership in parties, predicts that there is greater citizen
engagement and possible mobilization. This is true regardless of whether a
party disappears, whether leading voters to have to change their party, or
whether they have to become non-partisan.
In his research of 36 countries, Whiteley (2009, pp. 142–148) reveals that there
are considerable differences between active members of the party, party mem-
bers, former party members and citizens who have never participated in any
political party. Citizens with experience in political parties engage in political
affairs more than those with no experience in this respect. In his multi-level
participation model, he shows that the percentage of party members at the
aggregate level is a strong indicator of the participation of individual citizens.
Thus, party members and citizens in municipalities should be more engaged in
politics. Another of the forms of participation that is examined is the signing of
petitions, which, in Peru, is a crucial argument for why it is more likely that a
sufficient number of signatures could be gathered for the initiation of a recall
in Peruvian municipalities where local electorate party membership is higher.
I argue that there could be important variables other than the closeness of
the previous municipal election for predicting recall. In this respect, I examine
the influence of local electorate party membership. Local political leaders can
use their members as a resource by which they can mobilize the local elec-
torate. This can still happen in the rather low level of party system institu-
tionalization where political parties disappear as they do in Peru. Whiteley’s
(2009) research shows that being a party activist, party member, or ex-member
increases participation and civic engagement in politics. Thus, one may expect
for local political leaders to find it easier to mobilize electorate that has a
higher share of party members as one may assume greater engagement in
politics in the form of signing petitions to hold recall elections. While elite
mobilization theory would explain successful mobilization leading to suc-
cessful initiation of the recall given that the signatures of 25 percent of the
Michael Haman 79
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
electorate are required to initiate such a process, it would not necessar-
ily explain the successful removal of a mayor in a recall election in which a
mayor’s supporters vote alongside the whole local electorate. Therefore, there
must be greater support than 25 percent of the electorate for a mayor to be
removed. The following hypothesis examines the initiation of recall elections.
H1: The greater the party membership in a municipality, the more likely it is
for recall elections to be held.
2) Invalid votes
The next hypothesis concerns a phenomenon that is typical of a number of
Latin American countries; i.e., a large number of invalid votes in elections.
In the case of the Peruvian municipal elections, invalid votes include blank
votes (blancos) and null votes (nulos). A number of studies link citizens’ invalid
votes to their dissatisfaction and therefore as a protest against political elites
(Power & Garand, 2007; Power & Roberts, 1995; Uggla, 2008). Kouba and Lysek
(2019) recently published a comprehensive meta-analysis of invalid voting.
They divide two categories of invalid voting in which voters cast invalid votes
unintentionally or intentionally. They divide intentional invalid voting on the
basis of dissatisfaction into three categories: unsupportive, disempowered,
and disenchanted. All three classifications are based on the level of politi-
cal support and a sense of subjective political empowerment. Disenchanted
voters feel that it makes no difference which party wins an election. Hence, this
type of invalid vote often occurs in political systems with weak competition.
Studies that differentiate blank and null votes are scarce. Most analyze blank
and null votes together and treat them as one (Power & Garand, 2007; Power
& Roberts, 1995). However, I argue that they are not the same and that there
is an important difference between the two. For some, blank votes are a clearer
sign of protest because voters voluntarily and deliberately do not mark any of
the candidates and put a blank ticket in the ballot box. While in the case
of a null vote, the voter can make an involuntary error that invalidates his
vote (Cohen, 2018, p. 403). Driscoll and Nelson (2014) examined the differ-
ence between blank and null votes based on the Bolivian judicial elections.
They investigated Bolivian voters for both types of invalid votes for political
reasons; however, a null vote was more typical for sophisticated voters who
were dissatisfied and thus expressed their protest. In contrast, less-informed
Recall Elections: A Tool of Accountability? Evidence from Peru
80
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
voters gave blank votes for reasons of greater confusion and apathy. Barnes
and Rangel (2018) studied mayoral elections in Chile where they distinguished
between null and blank votes before and after the abolition of compulsory
voting. They found that with and without compulsory voting, both blank
and null votes reduced as electoral competitiveness increased. Peru has a
much higher number of invalid votes, ranking first in Latin America (Kouba &
Lysek, 2016).
Because the number of invalid votes may relate to the municipality’s com-
petitiveness, I examine whether municipalities with a lower number of invalid
votes are more likely to hold a recall. In these cases, citizens should not be as
disenchanted and would not want to waste their votes by making them invalid;
in turn, political leaders should be able to mobilize for a sufficient number
of signatures to hold recall elections, as research suggests that invalid votes
show voter apathy and one may assume that this would make the collection
of signatures less likely. It is important to acknowledge Peru’s compulsory
voting system, which results in it being more costly for a voter to protest by
abstention in elections than casting an invalid vote. At the same time, after
controlling all the variables, especially the High Development Index (highly
correlated3 with levels of education), a significant strength of prediction may
apply for the number of blank votes, but not for null votes. In the case of null
votes, voters can spoil their ballot unintentionally, which is why the hypoth-
esis includes only the blank votes that voters cast deliberately. As in the case
of the first hypothesis, this hypothesis too is only concerned with the initia-
tion of the recall; as the mobilization and initiation of a recall differ from the
vote concerning the fate of a specific mayor.
H2: The more blank votes, the less likely it is for recall elections to be held,
but this does not apply in the case of null votes.
3) Performance of the mayor
One of the basic arguments to support recall is that it gives people the
possibility to remove bad politicians. The reasons for the recall of politi-
cians in the USA are often linked to corruption and the misuse of resources
3 The Pearson correlation coefficient for the years of education variable is 0.894 and for the secondary
education variable, it is 0.742.
Michael Haman 81
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
(Bowler, 2004, p. 209). Recall is considered an insurance mechanism, meaning
that politicians are constantly controlled by the public and any incompetent or
corrupt behavior can lead to a petition for recall (Zimmerman, 2014, p. 112).
The organizers4 of the 2012 recall elections in Peru, gave similar reasons as
the above. For example, the violation of election promises, misappropriation,
incapacity, nepotism, abuse of power or non-transparent municipal manage-
ment (Oficina Nacional de Procesos Electorales 2013a: 81).
Classical economic theory assumes that the voter rewards the incumbent in
good economic times and punishes him in bad times (Lewis-Beck & Stegmaier,
2008). Sakurai and Menezes-Filho (2008) illustrated that spending money
is directly related to the future success of the mayor in Brazilian munici-
palities. This is based on the accountability of politicians. Mayors try to per-
form well because they worry about recalls as mayors that are not effective
are recalled by citizens. However, Holland and Incio (2019) revealed that in
Peruvian municipalities, better mayoral performance does not increase the
likelihood of initiating successful recall elections, but actually decreases
the likelihood of recalling a mayor in an already initiated recall election. I
also examine mayoral performance in a different way as I explain in the next
section. My hypothesis is as follows:
H3: Mayoral performance does not affect the likelihood of holding recall elec-
tions, but it does affect the success in the removal of a mayor in an already
initiated recall election.
4) Organizers of recall elections
In order to confirm the elite mobilization theory in Peru, it is necessary to
investigate recall organizers in particular. Quintanilla (2013) claims that prior
research shows that politicians who lost in previous elections organize recall
elections only in minority cases and the conclusions of previous authors may be
based more on subjective feelings. Holland and Incio (2019, p. 15) support this
and claim that in 18 percent of cases organizers were candidates in the previ-
4 The organizer of the recall (promotor) is a citizen that proposes recalling a mayor to Peruvian autho-
rities. He must obtain the signatures of more than 25 percent of a local electorate and meet all the
requirements established by law (see Tuesta Soldevilla (2014) for a detailed description of the recall
process in Peru).
Recall Elections: A Tool of Accountability? Evidence from Peru
82
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
ous municipal elections between 2002 and 2014. However, Tuesta Soldevilla
(2014, p. 22) argues that if the organizers were not counted only as candi-
dates from the previous elections and their political party membership was
also included, this number would increase greatly. I expand on this issue below.
Past research does not clearly indicate that recall organizers are necessarily
associated with a party. I examine and quantify whether organizers ran for any
election in the past and whether they were members of the party. However,
in my next hypothesis, I examine the relationship between the removal of a
mayor in already initiated recall elections and the electoral experience of the
organizer. I assume that people with political (electoral) experience are more
likely to be successful at mobilizing the electorate than in actually removing
a mayor. This assumption is based on the literature about candidate quality
(Green & Krasno, 1988, 1990; Jacobson, 1990; Lazarus, 2008), which points
to quality candidates being more successful in elections. The quality of can-
didates is usually measured by electoral experience; i.e., candidates running
for office and winning elections. Thus, I expect recall organizers with electoral
experience to be better quality organizers in this respect and more successful
at mobilizing people to remove a mayor as they have electoral experience and
they manage to get enough signatures to initiate recall elections.
H4: In successfully initiated recall elections, the organizers with political expe-
rience are more successful in recalling a mayor.
Data and research design
I examine the period of recall elections from 2001 to 2013 in Peruvian munici-
palities that coincide with municipal elections from 1998 to 2010. However,
as I will explain later, I was able to use regression models only for some peri-
ods given the lack of the necessary data and I use correlations for different
periods as I do not have complete data to draw up comprehensive models.
Unfortunately, I was also unable to analyze the first recall elections under the
new rules in 2017, as they were held only in 27 municipalities (1.6%). This
is too small a number with which to be able to use the statistical methods I
employ, and given this small number, they can be considered rare events in
terms of statistics. In the previous period, recall elections were held in 24% of
the municipalities, and this sharp contrast makes it impossible to include both
Michael Haman 83
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
periods in the same model. For a sample of this size, qualitative methods may
be more appropriate. Moreover, due to the changes in the regulations, for the
opposition to initiate a recall seems to make less sense5. I therefore focus on
a period when the institutional rules were the same.
Dependent variables
I use two basic analyses and two dependent variables. The first analysis deals
with the likelihood of holding recall elections and I examine the difference
between the municipalities where recall elections were held after the municipal
elections. If recall elections were held, I code this as 1; if not, I code it as 0. The
second analysis examines the likelihood of the mayor’s removal if recall elec-
tions are already initiated. Again, the successful removal of a mayor is coded
as 1, and failure as 0. These processes differ greatly from each other and there-
fore need to be examined separately as was mentioned in the theory section.
In the case of recall elections, citizens vote for each politician separately.
However, I only examine a mayor in the second analysis because it is the most
important office in the municipality. Recall elections held in 1146 municipal-
ities until the year 2013 involved 1113 mayors (Infogob-JNE, 2017). During
recalls, voters can make decisions regarding multiple politicians in the same
recall; however, I do not examine the 4083 councilors. Current research shows
that councilors face recalls mainly because the opposition attempts to initiate
new municipal elections6 (Tuesta Soldevilla, 2014, pp. 25–27). Thus, it is not
really their performance that is involved and the model would be much more
complicated to interpret because it would also be necessary to distinguish
possible councilors who do not come from the mayor’s party.
Independent variables
My first independent variable is the percentage of party members in the local
electorate. Information about this variable is not available for earlier than
2010. The other two independent variables are the percentage of blank and
5 Law 30315 (Ley 30315) of 2015 determined that recall elections can be held only in the third year of
mayor’s term. Also, this law no longer allowed the initiation of new municipal elections as I explain in
the first footnote.
6 For by-elections, at least one third of the council must be recalled.
Recall Elections: A Tool of Accountability? Evidence from Peru
84
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
null votes. As I discuss in the theory section, blank and null votes are different
and I therefore examine them separately. The third independent variable is the
mayor’s performance. As the first variable for measuring performance, I use
the investments made by the mayor in the first year of being in office. If no
investment is made, the inhabitants of the municipality would see this because
roads, schools, or other public buildings and areas would not be repaired or
built. I measure this variable as the share of investments in the first year of
a mayor’s term in office over the previous two years. This variable, which
compares a period within a municipality is ideal, as candidates for mayor gen-
erally promise investment and if they then invest less than the previous mayor
(in the case of reelected incumbents, then the same mayors in the previous
period), citizen dissatisfaction can be expected. The advantage of using this
variable is that it compares periods within each municipality, thus limiting the
need to control for financial transfers from the central government (inputs)
as it is reasonable to expect that municipalities receive more or less the same
financial resources every year. The potential changes in the distribution of
financial resources from the central government would reflect all municipali-
ties. Thus, during the first year of their term in office, mayors should have to
perform to similar standards. Local government performance and efficiency
literature very often includes an investment in infrastructure or other public
services as an output variable to measure performance (Narbón‐Perpiñá &
De Witte, 2018). This is the direction I follow in this study.
In the Peruvian context, one of the mayor’s main tasks is to plan and imple-
ment the budget. Therefore, the second variable measuring performance relates
to the budget and it is a measure of how much the original budget has been
spent successfully in the first year in office. In Peru, for the sake of trans-
parency, individual budget items are documented and, to combat corruption,
mayors face obstacles in actually spending these resources (Sexton, 2017,
pp. 11–23). In Peru, this variable is often used7 as an indicator of performance
and also included as such in research (Holland & Incio, 2019; Leon & Kleine-
Rueschkamp, 2018; Pique, 2019; Sexton, 2017). If the mayor and his team are
not proficient in the administration, it is quite likely that the planned budget
will not be exhausted and that the planned projects will not be executed. I
use the first year in office (2011) to measure the performance because it is a
7 For example, for budget execution as an indicator of performance also see two press articles: Fernández
(2016) and “Regiones y municipios del sur” (2017).
Michael Haman 85
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
sufficient time period for citizens to evaluate the mayor’s performance, and
it is when a recall procedure could be initiated. In the following year (2012),
a recall election could be held in a municipality. I use the first year in office
(2011) also as a measurement for the recall elections held in 2013 as I examine
whether recall elections were held in the municipality during one mayor’s term
in office (2011-2014). It is also logical that in the 2013 recall elections, the
mayors were judged for the first year and the mobilization process for the recall
could have already begun in 2012, but no formal obligations were complied
by for the 2012 recall; the recall was therefore delayed in the given munici-
pality. The budget figures available for all municipalities only date as far back
as 2008. The first year in the mayor’s office after the municipal elections in
2006 was in 2007, making it impossible to examine the period between 2006
and 2010 based on performance data. Holland and Incio (2019) also used the
execution of a budget in their study as a variable and they deal with this lack
of data using by the proxy from 2008.
The last independent variable involves the recall organizer. I code a per-
son with electoral experience as 1, if they ran in an election. This includes
elections prior to 20128 at local, regional, and national levels. Information
about recall organizers is provided by the National Jury of Elections (Jurado
Nacional de Elecciones, JNE).
Control variables
I use several control variables that draw from substantial research that can
generally influence electoral behavior, and are thus also validalso for elections
in Peru (Geys, 2006; Holland & Incio, 2019; Kouba & Lysek, 2019). The first
control variable concerns socioeconomic factors. Traditionally, citizens with
lower levels of education, lower income, and lower social status are consid-
ered to be underrepresented among politically active citizens (Verba & Nie,
1972). Higher education itself is considered to be an important variable in
terms of understanding protest (Barnes et al., 1979). I use the United Nations
Development Programme (UNDP) Human Development Index (HDI) as a proper
measure of a municipality’s social and economic development. Also, based
on the analysis of the data from the Latin American Public Opinion Project
8 In models, I analyzed recall elections held in 2012 and 2013. Therefore, I examined pre recal l elec tion data.
Recall Elections: A Tool of Accountability? Evidence from Peru
86
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
(LAPOP), it is clear that in municipalities where GDP is rising, more citizens
trust the municipal government (Montalvo, 2010).
Other control variables include election closeness and the effective number
of parties (ENP). Closeness is measured as the percentage difference between
the second and the first candidates in the election. The percentage difference
is a good measure to be used in the Peruvian context, as the winner always
receives more than half the seats in the municipality and the electoral system
is heavily disproportionate. I include the ENP mainly because of party coor-
dination. Successful electoral strategic coordination decreases the number of
political parties competing in elections (Cox, 1997, p. 4). ENP can measure
this and it should be easier to organize and mobilize to initiate recall elections
when there are fewer political parties. I use the winner’s percentage of the
vote for models that deal only with successfully initiated recall elections. In
this case, there was already enough coordination for a successful recall and,
because all citizens can vote, the winner’s last percentage in the election is
more important. A mayor must receive more than 50 percent of valid votes
to be recalled and 25 percent of the signatures of the electorate are enough
to initiate a recall.
I use turnout as an additional variable to control the possible effects of politi-
cal participation. Spivak (2004) finds that turnout is an important factor for
the likelihood of recall elections in some states in the United States, as the
required percentage of signatures to initiate recall elections in some states
depends on the turnout in previous elections. Therefore, when abstention in
the previous elections is high then it is easier for organizers to collect the
required number of signatures as the number is low. However, voting in Peru
is compulsory, and the required number of signatures is fixed at 25% of the
electorate. Thus, turnout should be a less important variable for the likelihood
of recall in Peru than in some states in the United States. I include the size of
the municipality, density, and urbanization to control for the potentially easier
organization of the recall in smaller areas and in high-density ones.
Next, I control for the relationship between politics at national and local
levels. The most important political party in Peru of the 20th and 21st centuries
is APRA (Partido Aprista Peruano); President Alan García’s party and where his
term ended in 2011. The second-order election theory, explored especially in
the European Parliament elections, posits that voters often try to punish the
Michael Haman 87
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
ruling parties in non-national elections (Oppenhuis, Van der Eijk, & Franklin,
1996). The greatest number of mayors came from APRA. I control this poten-
tial effect on the likelihood of a recall, and I use control variables for gender
and age. In this respect, men are still perceived by a part of the population
as better political leaders than women. However, according to the World
Values Survey, in Peru, 30.3 percent of people strongly disagreed with this
claim in 2012; 44.8 percent disagreed, and 18.9 percent either strongly agreed
or agreed (Inglehart et al, 2014).
Age is very often used as a variable to explain the likelihood of protest. Young
people participate in protests more than the elderly (Barnes et al., 1979;
Dalton, 1996). JNE offers information about the age and gender of candi-
dates, and considers political candidates under the age of 30 as young (joven).
I measure the percentage of young people that run for office on all lists in
a municipality as well as their gender. Subsequently, I consider whether the
mayor is a female, which I code as 1, or male, which I code as 0, and whether
he or she is under the age of 30, coded as 1, or over it, coded as 0. I also con-
trol for whether the mayor won the previous municipal election (incumbent).
I use financial transfers per voter9 from the state to municipalities as another
control variable. The Peruvian Ministry of Finance reallocates funding to indi-
vidual municipalities based on multiple criteria. This is revenue to which every
municipality is entitled; however, in areas where resources are mined, they
also receive mining revenue (canon). Thus, municipalities do not receive the
same amount of funds in proportion to their populations.
I control for the previous use (before 2010) of recall elections in a municipality
by counting all past recall elections. According to the psychological concept of
adaptive learning, voters can adopt a habit (Bendor, Diermeier, & Ting, 2003;
Kanazawa, 2000; Sieg & Schulz, 1995), so that citizens who voted in the past
are more likely to do so in the future. Thus, a municipality’s past experience
can influence the likelihood of new recall elections. I examine this also in the
provinces by first adding up the total number of recall elections in the prov-
ince, then deducting the number of recalls in the municipality from this sum
9 I chose to use the electorate and not the population. In Peru, the closest census to this period was
in 2007 and I would have to use an estimate. Therefore, I used the electorate. However, the Pearson
correlation coefficient between the estimated population and the electorate is of 0.980. I verified this
by checking that using the other would not change the results of the research.
Recall Elections: A Tool of Accountability? Evidence from Peru
88
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
in the province, and finally dividing it by the number of municipalities ana-
lyzed in the province. Citizens from neighboring municipalities in the province
can learn about the recall and use this tool themselves.
The final control variable involves irregularities in elections. I measure this
variable as the presence of incidents registered by the National Directorate
of Inspection and Electoral Processes (Dirección Nacional de Fiscalización y
Procesos Electorales, DNFPE) during municipal elections. I code the presence
of incidents as 1 and their absence as 0. In this case, the DNFPE identified
no incidents serious enough to invalidate whole elections, but only incidents
which were considered a problem during the electoral process.
Models
I use logistic regression, which is the ideal tool in this case as holding recall
elections and the mayor’s removal are dichotomous dependent variables.
I create several regression models to verify the hypotheses and search for
associations between variables in the correlations (in the Appendix, Table 5).
Because of the higher number of variables, I place emphasis on avoiding mul-
ticollinearity. Accordingly, I treat variables based on theoretical considerations
very cautiously in individual regression models, and conduct a number of tests
to avoid possible multicollinearity10.
Four logistic regression models include as many variables as possible and
examine the period after the municipal elections in 2010. I include all vari-
ables in correlations that were at my disposal from municipal elections since
1998 in the Appendix as a way of controlling logistic regressions that exam-
ine the recall only after 2010. In the rare11 case that municipal elections were
annulled, I do not examine that municipality in the study period. I explain all
variables and provide information on their source in Table 4 of the Appendix.
10 The maximum value of VIF (Variance Inflation Factor) in all models was 2.72 and in most cases below
1.5. I also tested condition index and no models reached critical values (Belsley, Kuh & Welsch, 2005).
11 For example, I examine 1605 municipalities after the 2010 municipal elections among all 1639 elections
that were held.
Michael Haman 89
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Results
I first analyze all 1113 recall elections in which the mayor faced a recall from
1997 to 2013 and the organizers of these recalls. I found that 56% of the
organizers ran for some election in the past and 50.3 percent of them were
party members. The politicians running in elections and party members are
not exclusive categories as a politician does not have to be a member of a
party to run in elections and a party member does not necessarily run in elec-
tions. However, organizers who already ran for elections or were members of
the parties constituted 74.4%. Only a quarter of the organizers had no links
to political parties and never ran in elections. I confirm the assumption that
Peru’s recall elections are a highly politicized instrument. It seems that recall
is used by political actors who take advantage of this tool to regain power
themselves or to help their party do so.
Table 1 offers descriptive statistics of variabl es that I use in the logistic regres-
sions. The models (Table 2) unambiguously show that the political-electoral
variables are very important in terms of the likelihood of holding recall elec-
tions after the 2010 municipal elections. The first two models confirm that
the greater the party membership of the electorate in the political parties,
the greater the likelihood of holding recall elections. Even though Peru is a
country with a low level of political party institutionalization, it seems that
party membership of a local electorate is an important variable in explaining
recalls. Citizens in more politically active municipalities use the recall mecha-
nism more often, which is quite logical as they are more interested in politics
and are more easily mobilized. Local political leaders may use their members
as a resource to obtain a sufficient number of signatures to initiate the recall
procedure. These findings are in line with Whiteley’s (2009) research on party
membership and mobilization. Whiteley (2009) finds that party membership
at the aggregate level is a significant variable for the prediction of participa-
tion including mobilization for signing petitions. At the same time, in line with
the elite mobilization theory, political leaders try harder to initiate a recall
if the election results of the municipal elections are close. This supports the
first hypothesis.
Blank votes are statistically significant in the first two models and in all elec-
tions since 1998 that are included in the correlations. The more blank votes
there are, the lower the likelihood of holding recall elections. However, this is
Recall Elections: A Tool of Accountability? Evidence from Peru
90
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Table 1. Descriptive statistics
N Minimum Maximum Mean Std. Deviation
Party membership 1605 0.54 46.84 9.87 6.70
Blank votes 1605 1.35 25.60 9.08 3.74
Null votes 1605 0.00 61.33 6.87 5.83
Execution of budget 1605 11.40 99.40 73.70 15.66
Closeness 1605 0.00 56.22 9.45 8.66
ENP 1605 1.52 11.07 4.30 1.41
Turnout 1605 64.86 96.12 86.25 4.25
Size of electorate – logged 1605 4.91 13.27 7.84 1.25
HDI 1605 0.09 0.80 0.33 0.13
Experience 1605 0.00 4.00 0.46 0.73
Experience – province 1605 0.00 2.00 0.40 0.31
Female 1605 0.00 1.00 0.04 0.19
Young mayor 1605 0.00 1.00 0.02 0.14
Female candidates 1605 24.00 55.56 36.94 3.79
Young candidates 1605 9.09 54.76 26.35 6.28
Irregularities in elections 1605 0.00 1.00 0.05 0.21
Financial transfers per voter 1605 3.77 10.16 7.22 0.76
APRA 1605 0.00 1.00 0.08 0.27
Incumbent 1605 0.00 1.00 0.20 0.40
Rural area 1605 0.00 98.69 53.71 30.46
Density – logged 1605 -2.12 10.18 3.10 1.84
Investment difference 1605 0.00 2.00 0.66 0.46
Winner 1605 14.56 78.11 35.64 10.07
Electoral experience 376 0.00 1.00 0.40 0.49
not true in the case of null votes. Null votes do not show significance in the
first two models. These models, therefore, confirm that these may be distinct
tools in the political system. Peruvian voters that have the potential to mobilize
do not necessarily want to cast blank votes in municipal elections. However,
voters may make a mistake during their voting, invalidating their votes and
Michael Haman 91
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
making them null or they may possibly feel more dissatisfied with the over-
all political system. Null votes are not statistically significant to predict the
likelihood of holding recall elections. Voter errors could be explained by socio-
economic variables (mainly education), which is controlled in the models by
the HDI variable. On the other hand, Driscoll and Nelson (2014) came up with
the finding that voters cast blank votes because of political apathy in Bolivia.
In contrast, Peruvian blank votes are associated with higher mobilization of
citizens for holding a recall election in municipalities. It would seem that
under compulsory voting, fewer blank votes within the electorate can pre-
dict greater engagement in politics in the form of recall elections. The higher
number of invalid votes is often linked with higher dissatisfaction in politics
and is also connected with party competition (Kouba & Lysek 2019). These
models therefore confirm that a higher number of intentional blank votes
increases the likelihood of recall elections because a high number of blank
votes is likely connected with greater dissatisfaction and interest in politics.
Therefore, a highly dissatisfied electorate with more blank votes is not actu-
ally interested in the recall procedure. This supports the second hypothesis.
Table 2. Logistic regressions and Peruvian recall elections
DV: Recall elections after 2010
municipal elections
DV: Mayor was recalled in a recall
election
Model 1 Model 2 Model 3 Model 4
Party membership 0.026*
(0.010)
0.025*
(0.010)
-0.020
(0.020)
-0.019
(0.020)
Blank votes -0.080***
(0.020)
-0.082***
(0.020)
0.062
(0.041)
0.063
(0.040)
Null votes -0.007
(0.011)
-0.007
(0.011)
0.040*
(0.020)
0.040*
(0.020)
Execution of budget -0.001
(0.004) --0.002
(0.009) -
Closeness -0.058***
(0.010)
-0.058***
(0.010) - -
ENP -0.096
(0.055)
-0.094
(0.055) - -
Turnout 0.023
(0.016)
0.021
(0.016)
-0.023
(0.033)
-0.020
(0.033)
Size of electorate -0.280**
(0.088)
-0.283**
(0.087)
-0.199
(0.193)
-0.193
(0.191
HDI -0.609
(0.778)
-0.638
(0.775)
1.629
(1.722)
1.785
(1.722)
(Continued)
Recall Elections: A Tool of Accountability? Evidence from Peru
92
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Table 2. Logistic regressions and Peruvian recall elections
DV: Recall elections after 2010
municipal elections
DV: Mayor was recalled in a recall
election
Model 1 Model 2 Model 3 Model 4
Experience 0.411***
(0.083)
0.415***
(0.083)
-0.035
(0.162)
-0.036
(0.162)
Experience – province 0.776***
(0.215)
0.770***
(0.215)
0.315
(0.413)
0.307
(0.411)
Female 0.547
(0.321)
0.552
(0.323)
-0.647
(0.689)
-0.636
(0.688)
Young mayor 0.194
(0.468)
0.204
(0.470)
-0.67
(1.133)
-0.751
(1.146)
Female candidates -0.018
(0.019)
-0.017
(0.019)
0.324
(4.115)
0.071
(4.131)
Young candidates 0.014
(0.010)
0.014
(0.010)
1.448
(1.961)
1.316
(1.966)
Irregularities in
elections
0.987***
(0.273)
1.002***
(0.274)
-0.919
(0.598)
-0.964
(0.598)
Financial transfers
per voter
-0.352**
(0.119)
-0.322**
(0.116)
-0.109
(0.282)
-0.121
(0.271)
APRA -0.267
(0.245)
-0.247
(0.245)
0.068
(0.481)
0.029
(0.483)
Incumbent 0.257
(0.166)
0.274
(0.166)
0.464
(0.319)
0.412
(0.319)
Rural area -0.002
(0.003)
-0.003
(0.003)
-0.002
(0.006)
-0.001
(0.006)
Density -0.094
(0.052)
-0.093
(0.052)
0.108
(0.112)
0.124
(0.112)
Investment difference - -0.198
(0.145) -0.252
(0.297)
Winner - - -0.043*
(0.017)
-0.043*
(0.017)
Organizer – electoral
experience - - 0.578*
(0.262)
0.575*
(0.260)
Constant 3.289
(1.993)
3.297
(1.900)
2.529
(4.428)
2.042
(4.099)
Cases 1605 1605 376 376
Nagelkerke’s R2 0.212 0.213 0.113 0.116
Note: Standard errors in parentheses, *p <0.05, **p <0.01, ***p <0.001
Source: Author
The first and second models show that the smaller the difference between the
first and the second party, the more likely it for recall elections to be called.
Michael Haman 93
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
So far, researchers (Holland & Incio, 2019; Tuesta Soldevilla, 2014; Welp,
2016) have considered political competition in the municipality to be one of
the most important factors in explaining recall in Peru. I too confirm these
conclusions. The difference between the first and the second party after the
municipal elections is key to explaining recall, as is the ENP. However, in con-
trast, the lower the ENP, the greater the chance for holding recall elections.
If support is concentrated in a few parties, there is no doubt that it is easier
to agree and coordinate together and obtain the necessary number of signa-
tures to hold recall elections. In municipalities with a large number of parties,
recall organization is much more complicated. Also, as shown in the Appendix,
I find a statistically significant correlation between competitiveness and the
holding of all recall elections for all elections since 1998.
So far, the hypotheses have concerned the probabilities of holding recall elec-
tions. I now turn to variables that were assumed to influence the removal of
a mayor. This is not the same as holding recall elections. For the successful
organization of recall elections, it is necessary to mobilize and obtain 25 per-
cent of the signatures of the electorate. For the successful removal of a mayor,
he must be judged by the citizens and not only by the electorate who initi-
ated the recall procedures. This is examined in models three and four. Unfor-
tunately, as in previous models, all variables are only available for 2010 and
beyond. However, all variables from 1998 that were not included in the mod-
els are listed in Table 5 of the Appendix.
Concerning socioeconomic variables, municipal development has no effect on
the likelihood of holding recall elections and the recalling of a mayor. The HDI
alone was used in the regressions, which was highly correlated with the other
variables based on which this index is composed: life expectancy, percentage
of people with secondary education, average years of education and family
income per capita. None of these variables reached significant correlations in
any elections from 2002 onward; some of them are only statistically signifi-
cant in certain elections, but since this minimal correlation changes across
the election, they cannot in any way be considered significant. By including
the HDI variable in all models, where all the variables are controlled, it does
not reach significance. Therefore, the municipality and the education levels of
local citizens do not affect the likelihood of holding recall elections or whether
the mayor is removed.
Recall Elections: A Tool of Accountability? Evidence from Peru
94
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
The most important question for supporters of a recall is whether we can
consider the recall to be an accountability mechanism. It is very difficult to
support this based on the recalls after municipal elections in 2010. The budget
execution variable does not reach significance in any of the models, which is
also the case for the variable examining the difference between the number
of investments in the mayor’s first year in office with the previous two years.
Table 2 also shows performance comparisons in percentiles. Five percent of
the worst mayors in terms of budget execution faced recall elections in 21.3
percent of cases but they were recalled only in 3.8 percent of cases. In con-
trast, 5% of the best performing mayors were recalled in 7.5 percent of cases.
Something similar is true for a variable called difference in investment, which
should be crucial for citizens as it can be used to compare the periods and see
the investments with their own eyes. Yet, the result is the same: 5% of the
worst mayors were recalled only in 3.8 percent of cases, while 5% of the best
mayors were recalled in 7.5 percent of cases.
Table 3. Performance of elected mayors after the municipal elections in 2010
% mayors Execution of budget Difference in investment
Top Recall elections % Recalled % Recall elections % Recalled %
5 26.3 7.5 23.8 7.5
10 29.4 8 21.9 6.3
20 28.3 8.1 22.7 5.9
30 24.9 6.7 24.1 6.4
Bottom
30 20.5 5.8 23.9 5.4
20 19 5.6 24.6 5.9
10 19 5.5 22.5 5.6
5 21.3 3.8 28.7 3.8
Source: Author
The third hypothesis is therefore refuted because after municipal elections
in 2010, the likelihood of removal in initiated recall elections for poorly
performing mayors was not greater. Unfortunately, it is not possible to exam-
ine this finding for previous years due to insufficient data. These conclusions
confirm what other researchers have suggested in previous literature (Tuesta
Michael Haman 95
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Soldevilla, 2014; Welp, 2016). Overall, findings in relation to recall elections in
Peru are not what would be expected according to classical economic theory,
in which voters reward the incumbents in good economic times and punish
them in bad times (Lewis-Beck & Stegmaier, 2008). In Brazilian municipalities,
Sakurai and Menezes-Filho (2008) find that spending money is directly related
to a mayor’s future success and is therefore seen as a form of accountability.
However, the worst performing Peruvian mayors are not punished any more
than the better performing ones. Thus, in the case of Peru, these findings are
an important contribution to the literature.
But how can we explain the fact that the mayor is successfully recalled? If
the organizer is a politically experienced person, this increases the chances
of successfully recalling the mayor in the third and fourth model by about
80 percent12 compared to when the organizer never ran for election prior to
recall elections in 2012 and 2013. Table 6 of the Appendix offers descriptive
statistics based on organizers’ varying electoral experiences. For example, it
shows whether organizers ran in the 2010 municipal elections or any other.
It also examines whether the organizers previously ran for the office of mayor. In
all cases, the organizers with electoral experience successfully removed
mayors more often than organizers without electoral experience. These findings
support the fourth hypothesis. Tuesta Soldevilla (2014, p. 22) suggested in his
research that the linkage between an organizer’s party and electoral history
is an important variable based on which to understand the recall procedure.
I too confirm that it is a significant predictor of a successful recall. It seems
that a person that is more engaged in politics is able to both convince enough
citizens to recall a mayor and to organize the process. These findings follow
Whiteley’s research (2009) of 36 countries to some extent. By using the multi-
level model, he finds that at individual level, being a party activist, member or
ex-member of a party increases civic participation and engagement in politics.
I add to his conclusion, the new insight that organizers with electoral experi-
ence are more successful in recalling mayors than unexperienced organizers.
A greater number of null votes increases the likelihood of the successful
removal of a mayor. In this case, even though blank votes go in the same
positive direction as null votes, this variable does not reach statistical signif-
icance. After controlling socioeconomic variables (such as education within
the HDI) that could have an influence on an involuntary voter error. It seems
12 Odds ratio from logistic regressions.
Recall Elections: A Tool of Accountability? Evidence from Peru
96
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
that null votes better explain voters’ dissatisfaction in municipalities, which
they manifest during recall elections in removing a mayor. As I mentioned in
the theory section, most studies do not differentiate between blank and null
votes. The results of this study show that there may be good reason to distin-
guish between these two forms of invalid votes in future research.
Only one other variable is a statistically significant predictor for successful
recall and that is the percentage of votes in the preceding municipal elections.
Mayors who have won more votes in the preceding municipal election had a
more secure position during successfully initiated recall elections.
The result of the regression shows that the municipalities that experienced
some irregularities13 during the 2010 elections were over 120 percent more
likely to hold a recall than municipalities with no problems. This is undoubt-
edly connected with the great political competition in municipalities, in which
the electoral contest can sometimes exceed what is allowed. The turnout did
not significantly explain the likelihood of holding recall elections. In this case,
voter turnout is not a variable that would increase the likelihood of holding
recall elections.
Financial transfers per voter appear to be relevant in the first two models to
explain holding recall elections. The amount of financial transfers is mainly
related to the size of the municipalities if the other variables are controlled;
thus, the holding of recall elections is not more frequent in municipalities with
higher financial transfers by the state per voter. The size of a municipality’s
electorate is a statistically significant variable in the first two models and this
is not surprising. Indeed, this is consistent with previous research (Tuesta Sol-
devilla, 2014; Welp, 2016). In smaller municipalities, it is easier to obtain the
necessary number of signatures to organize a recall.
The young candidates variable is not statistically significant in any of the mod-
els and the young mayor variable is not statistically significant in the regres-
sion models. The gender of mayors and the gender of candidates in municipal
elections are not statistically significant variables for explaining why a recall
may be initiated. Although there is a positive relationship between a female
mayor and the probability of recall elections, this does not reach statistical
significance in any of the models and nor does political membership in APRA.
13 However, they were not serious enough for the elections to be annulled. An example fo this kind of
irregularity would be the campaigning of a political party in locations where it was not allowed.
Michael Haman 97
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
The experience with the recall variable is statistically significant in the munic-
ipality itself and in the province. For every experience with recall elections
(between 1997 and 2010) in a municipality, in the first two models, the chances
are increased by about 50 percent that there will be recall elections after the
municipal elections in 2010. This implies that the recall cannot be consid-
ered a tool that would be used occasionally in municipalities and serve as a
warning to politicians. On the other hand, if a municipality has already expe-
rienced a recall, it is more likely to use it again than a municipality without
this experience. At the same time, the experience with this mechanism in the
province shows that it reinforces awareness of this possibility and increases
the likelihood of holding recall elections. Also, if the mayor was incumbent, it
is positively associated with holding recall elections, but not significantly so.
Conclusion
What is the evaluation of Peruvian recall elections from 2001 to 2013?
Recall cannot be described as a tool that would increase trust in the political
system and parties. Peru is one of the countries with the least trust in its
political parties, its democratic system and administration of municipalities
in all of Latin America (Carrión, Zárate, & Zchmeister, 2015). The possibility of
recall elections, which should give citizens more power to control politicians,
has not changed this reality. By holding recall elections, individuals primar-
ily seek to gain power and so the recall cannot be described as a tool of civil
control. These conclusions are based on the fact that the variables describing
the mayor’s performance in office were not relevant for explaining the likeli-
hood of holding a recall election or, especially, his removal. In contrast, party
membership, the number of blank votes and political competition do predict
the likelihood of recall elections in Peru.
The findings regarding party membership are an important contribution.
Political parties have a number of functions and are an important part of the
political system, and therefore the weakness of political parties can have a
considerable impact on the functioning of the political system. Political scien-
tists have long analyzed European political parties; however, within the field
of political science, studies outside this region are being encouraged (Heidar,
2006, p. 312). Today, the phenomenon of political parties in Latin America has
yet to be sufficiently explored (Ribeiro & Locatelli, 2018, p. 4). As such, this
Recall Elections: A Tool of Accountability? Evidence from Peru
98
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
paper is a positive contribution to the current literature on the “black box” of
party organization (Levitsky, 2001).
In the case of successfully initiated recalls, only two other independent
variables were important in explaining the removal of a mayor when the recall
election was held. The first variable is logical and represents the percentage
of votes in the previous municipal elections. The mayor who won with fewer
votes would have fewer supporters, thereby reducing the likelihood that he will
succeed during the recall election. The second variable involves the organizer
of the recall procedure. If the organizer ran in the past elections, the success
rate of the mayor’s removal increases. Similarly to the previous variable, this
variable too shows the importance of political experience.
Future research should continue on in the direction of this article and exam-
ine the influence of party membership at local level and researchers should
examine the blank and null votes as separate variables. In the future, when
there is more data on recalls, it would be interesting to see whether this tool
can truly improve accountability.
Acknowledgments
This research was supported by a grant from the Philosophical Faculty, Uni-
versity of Hradec Králové.
The author would like to thank the anonymous reviewers for their suggestions
and comments to the original version of this paper.
References
1. Aldrich, J. H. (1993). Rational Choice and Turnout. American Journal of
Polit ical Sci ence, 37(1), 246–278. https://doi.org/10.2307/2111531
2. Aldrich, J. H. (1995). Why parties? : the origin and transformation of
political parties in America. Chicago: University of Chicago Press.
3. Altman, D. (2011). Direct democracy worldwide. Cambridge: Cambridge
University Press.
Michael Haman 99
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
4. Altman, D. (2019). Citizenship and contemporary direct democracy.
Cambridge: Cambridge University Press.
5. Barnes, S. H., Kaase, M., Allerbeck, K. R., Farah, B. G., Heunks, F. J., Ingle-
hart, Ronald F Jennings, M Kent Klingemann, Hans Dieter Marsh, A., &
Rosenmayr, L. (1979). Political Action: Mass Participation in Five Western
Democracies. Beverly Hills Calif: Sage Publications.
6. Barnes, T. D., & Rangel, G. (2018). Subnational Patterns of Participation:
Compulsory Voting and the Conditional Impact of Institutional Design.
Polit ical R esear ch Qua rterl y, 71(4), 826–841. https://doi.org/10.1016/j.
lungcan.2014.12.004
7. Belsley, D. A., Kuh, E., & Welsch, R. E. (2005). Regression diagnostics:
Identifying influential data and sources of collinearity. Hoboken, NJ:
John Wiley & Sons.
8. Bendor, J., Diermeier, D., & Ting, M. (2003). A behavioral model of turnout.
American Political Science Review, 97(2), 261–280.
9. Blais, A. (2000). To vote o r not to vo te? : the merits and limits of rational
choice theory. Pittsburgh (Pa.): University of Pittsburgh Press.
10. Blais, A., & Dobrzynska, A. (1998). Turnout in electoral democracies.
European Journal of Political Research, 33(2), 239–261. https://doi.
org/10.1111/1475-6765.00382
11. Bowler, S. (2004). Recall and representation Arnold Schwarzenegger
meets Edmund Burke. Representation, 40(3), 200–212. https://doi.
org/10.1080/00344890408523266
12. Caldeira, G. A., & Patterson, S. C. (1982). Contextual influences on partici-
pation in US state legislative elections. Legislative Studies Quarterly, 7(3),
359–381. https://doi.org/10.2307/439363
13. Carrión, J., Zárate, P., & Zchmeister, E. (2015). Cultura política de la
democracia en Perú y en las Américas, 2014: gobernabilidad democrática
a través de 10 años del Barómetro de las Américas. Lima: Instituto de
Estudios Peruanos.
Recall Elections: A Tool of Accountability? Evidence from Peru
100
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
14. Carty, R. K., & Eagles, M. (1999). Do local campaigns matter? Campaign
spending, the local canvass and party support in Canada. Electoral Studies,
18(1), 69–87. https://doi.org/10.1016/S0261-3794(98)00044-4
15. Clarke, H. D., Sanders, D., Stewart, M. C., & Whiteley, P. (2004). Polit ical
choice in Britain. Oxford: Oxford University Press.
16. Cohen, M. J. (2018). Protesting via the Null Ballot: An Assessment of
the Decision to Cast an Invalid Vote in Latin America. Polit ical Behav ior,
40(2), 395–414. https://doi.org/10.1007/s11109-017-9405-9
17. Cox, G. W. (1997). Making votes count : strategic coordination in the
world’s electoral systems. Cambridge, U.K.; New York: Cambridge Univer-
sity Press.
18. Cox, G. W., & Munger, M. C. (1989). Closeness, Expenditures, and Turnout
in the 1982 U.S. House Elections. The American Political Science Review,
83(1), 217–231. https://doi.org/10.2307/1956441
19. Dalton, R. J. (1996). Citizen politics : public opinion and political parties in
advanced industrial democracies (2. e d.) . Ch ath am N.J .: C hat ham Hou se.
20. Denver, D., & Hands, G. (1997). Modern constituency electioneering : local
campaigning in the 1992 general election. London: F. Cass.
21. Došek, T. (2016). Party Membership in Latin America Party Strategies and
the Role of Party Members. Taiwan Journa l of Democracy , 12(1), 169–198.
22. Downs, A. (1957). An economic theory of democracy. New York: Harper.
23. Driscoll, A., & Nelson, M. J. (2014). Ignorance or Opposition? Blank
and Spoiled Votes in Low-Information, Highly Politicized Environ-
ments. Political Research Quarterly, 67(3), 547–561. https://doi.
org/10.1177/1065912914524634
24. Fernández, C. (2016, December 24). ¿Cuánto invirtió en obras tu distrito?
Aquí te lo contamos. El Comercio. Retrieved from https://elcomercio.pe/
lima/invirtio-obras-distrito-contamos-155429-noticia/
Michael Haman 101
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
25. Franklin, M. N., & Hirczy, W. P. (1998). Seperated Powers, Divided Govern-
ment, and Turnout in U. S. Presidential Elections. American Journal of
Polit ical Sci ence, 42(1), 316–326. https://doi.org/10.2307/2991758
26. Geissel, B., & Jung, S. (2018). Recall in Germany: explaining the use of a
local democratic innovation. Democratization, 25(8), 1358–1378. https://
doi.org/10.1080/13510347.2017.1398735
27. Gerber, A. S., & Green, D. P. (2000). The effects of canvassing, telephone
calls, and direct mail on voter turnout: A field experiment. American Polit-
ical Science Review, 94(3), 653–663. https://doi.org/10.2307/2585837
28. Geys, B. (2006). Explaining voter turnout: A review of aggregate-level
research. Electoral Studies , 25(4), 637–663. https://doi.org/10.1016/j.
electstud.2005.09.002
29. Green, D. P., & Krasno, J. S. (1988). Salvation for the Spendthrift
Incumbent: Reestimating the Effects of Campaign Spending in House
Elections. American Journal of Political Science, 32(4), 884. https://doi.
org/10.2307/2111193
30. Green, D. P., & Krasno, J. S. (1990). Rebuttal to Jacobson’s “New Evidence
for Old Arguments.” American Journal of Political Science, 34(2), 363.
https://doi.org/10.2307/2111451
31. Heidar, K. (2006). Party membership and participation. In R. S. Katz &
W. J. Crott y ( Eds.), Handbook of party politics (pp. 301–315). London:
Sage Publications.
32. Hillygus, D. S. (2005). Campaign effects and the dynamics of turnout
intention in election 2000. Journal of Politics, 67(1), 50–68. https://doi.
org/10.1111/j.1468-2508.2005.00307.x
33. Holland, A. C., & Incio, J. (2019). Imperfect Recall: The Politics of Subna-
tional Office Removals. Comparative Political Studies, 52(5), 777–805.
https://doi.org/10.1177/0010414018797939
34. Infogob-JNE. (2017a). Retrieved from https://infogob.jne.gob.pe/
Recall Elections: A Tool of Accountability? Evidence from Peru
102
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
35. Infogob-JNE. (2017b). Consulta popular de revocatoria 2017. Retrieved
from https://infogob.jne.gob.pe/Analisis/Reportes
36. Inglehart, R., Haerpfer, C., Moreno, A., Welzel, C., Kizilova, K., Díez-
Medrano, J., Lagos, M., Norris, P., Ponarin, Ed., Puranen, B. & et al. (eds.)
(2014). World Values Survey : Round Six - Countr y-Pooled Dataf ile Version.
Madrid: JD Systems Institute.
37. Jacobson, G. C. (1990). The Effects of Campaign Spending in House Elec-
tions: New Evidence for Old Arguments. American Journal of Political
Science, 34(2), 334. https://doi.org/10.2307/2111450
38. Johnston, R. J., & Pattie, C. J. (1997). Where’s the difference? Decomposing
the impact of local election campaigns in Great Britain. Electoral Studies,
16(2), 165–174. https://doi.org/10.1002/jcb.10473
39. Kanazawa, S. (2000). A new solu tion to the collective action proble m: The
paradox of voter turnout. American Sociological Review, 65(3), 433–442.
https://doi.org/10.2307/2657465
40. Karp, J. A., Banducc i, S. A., & Bowler, S. (2008). Getting out the vote:
Party mobili zatio n in a comp arativ e persp ective . British Journal of Political
Science, 38(1), 91–112. https://doi.org/10.1017/S0007123408000057
41. Kauf mann, B., Büchi, R., & Brauneisen, N. (2008). Guidebook to direct
democracy in Switzerland and beyond. 2008 Edition. Bern: Initiative and
Referendum Inst. Europe.
42. Kirchgässner, G., & Schulz, T. (2005). Expected closeness or mobilisation:
why do voters go to the polls? Empirical results for Switzerland, 1981-
1999 (No. 1387). Munich.
43. Kouba, K., & Lysek, J. (2016). Inst itutional deter minants of invalid voting
in post-communist Europe and Latin America. Electoral Studies, 41,
92–104. https://doi.org/10.1016/j.electstud.2015.11.021
44. Kouba, K., & Lysek, J. (2019). What Affects Invalid Voting? A Review and
Meta-Analysis. Government and Opposition, 54(4), 745–775. https://doi.
org/10.1017/gov.2018.33
Michael Haman 103
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
45. Lazarus, J. (2008). Why Do Experienced Challengers Do Better than
Amateurs? Pol itic al Beha vior , 30(2), 185–198. https://doi.org/10.1007/
s11109-007-9046-5
46. Leon, G., & Kleine-Rueschkamp, L. (2018). Accountability, Political
Capture and Selection into Politics: Evidence from Peruvian Municipali-
ties Accountability, Political Capture and Selection into Politics: Evidence
from Peruvian Municipalities. Retrieved from https://www.barcelonagse.
eu/sites/default/files/working_paper_pdfs/1041.pdf
47. Levitsky, S. (2001). Inside the black box: Recent studies of Latin American
party organizations. Studies in Comparative International Development,
36(2), 92–110. https://doi.org/10.1007/bf02686211
48. Levitsky, S. (2018). Peru: The Institutionalization of Politics without
Parti es. In Party Systems in Latin America: Institutionalization, Decay,
and Collapse (pp. 326–358). Cambridge: Cambridge University Press.
49. Lewis‐Beck, M. S., & Stegmaier, M. (2008). The Economic Vote in Tran-
sitional Democracies. Journal of Elections, Public Opinion and Parties.
https://doi.org/10.1080/17457280802227710
50. Montalvo, D. (2010). Understanding Trust in Municipal Governments.
Latin American Public Opinion Project, Insights(35), 1–5. Retrieved from
http://www.vanderbilt.edu/lapop/insights/I0835en.pdf
51. Narbón‐Perpiñá, I., & De Witte, K. (2018). Local governments’ efficiency:
a systematic literature review—part I. International Transactions in
Operational Research, 25(2), 431–468. https://doi.org/10.1111/itor.12364
52. Oficina Nacional de Procesos Electorales. (2013a). Nuevas Elecciones
Municipales y Segunda Consulta Popular de Revocatoria Del Mandato
de Autoridades Municipales de Julio 2013, Lima: Oficina Nacional de
Procesos Electorales, Retrieved from https://www.web.onpe.gob.pe/
modEducacion/Publicaciones/L-0092.pdf
53. Oficina Nacional de Procesos Electorales. (2013b). Ley N. 26300. Retrieved
from https://www.web.onpe.gob.pe/modCompendio/html/procesos_elec-
torales/ley_derechos_particicontrol_ciudadano.html
Recall Elections: A Tool of Accountability? Evidence from Peru
104
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
54. Oppenhuis, E., Van der Eijk, C., & Franklin, M. N. (1996). The party context:
Outcomes. In C. Van der Eijk & M. N. Franklin (Eds.), Choosing Europe?
The European Electorate and National Politics in the Face of Union (pp.
287–306). Ann Arbor: The University of Michigan Press.
55. Pique, R. (2019). Higher pay, worse outcomes? The impact of mayoral
wages on local government quality in Peru. Journal of Public Economics,
173, 1–20. https://doi.org/10.1016/j.jpubeco.2019.01.005
56. Power, T. J., & Garand, J. C. (2007). Determinants of invalid voting in Latin
America. Electoral Studies, 26(2), 432–444. https://doi.org/10.1016/j.
electstud.2006.11.001
57. Power, T. J ., & Robe rts, J. T. (199 5). C ompul sory Voting, Inv alid Ballot s,
and Abstention in Brazil. Politi cal Res earch Qu arte rly, 48(4), 795–826.
https://doi.org/10.1177/106591299504800407
58. Quintanilla, A. (2013). El debate sobre la revocatoria y las reformas de
la ley 26300. In Eduardo Toche Medrano (Ed.), Perú H oy. Su surr os de sde
Babel (pp. 103–125). Lima: Desco.
59. Regiones y municipios del sur no pudieron gastar todo su presupuesto.
(2017, December 7). La República. Retrieved from https://larepublica.
pe/sociedad/1155111-regiones-y-municipios-no-pudieron-gastar-
todo-su-presupuesto-este-ano/
60. Ribeiro, P. F., & Locatelli, L. G. B. (2018). The Proof of the Pudding: Members,
Money and Party Organizational Strength in Latin America. https://doi.
org/10.13140/RG.2.2.15490.99522
61. Rosenstone, S. J., & Hansen, J. M. (1993). Mobilization, participation, and
democracy in America. New York: Macmillan.
62. Sakurai, S. N., & Menezes-Filho, N. A. (2008). Fiscal policy and reelection
in Brazilian municipalities. Public Choice, 137(1–2), 301–314. https://doi.
org/10.1007/s11127-008-9329-3
Michael Haman 105
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
63. Scarrow, S. E. (1996). Pa rtie s an d th eir memb ers: Or ganiz ing for vic tory
in Britain and Germany. Oxford: Oxford University Press.
64. Scarrow, S. E., & Webb, P. D. (2017). Investigating Party Organization:
Structures, Resources, and Representative Strategies. In S. E. Scarrow, P.
D. Webb, & T. Poguntke (Eds.), Organizing Political Parties: Representa-
tion, Participation, and Power (pp . 1– 27) . Ox for d: O xfo rd U ni ver sit y Pr ess .
65. Sexton, R. (2017). The Unintended Effects of Bottom-Up Accountability:
Evidence from a Field Experiment in Peru. Retrieved from https://www.
povertyactionlab.org/sites/default/files/research-paper/The-Unintended-
Effects-of-Bottom-Up-Accountability_Conflict_Peru_October2017.pdf
66. Sieg, G., & Schulz, C. (1995). Evolutionary dynamics in the voting game.
Public Choice, 85(1–2), 157–172.
67. Spivak, J. (2004). California’s Recall: Adoption of the “Grand Bounce”
for Elected Officials. California History, 82(2), 20–63. https://doi.
org/10.2307/25161725
68. Tuesta Soldevilla, F. (2014). Perú: entre la participación y la gobernabilidad
local (1997-2013). In Y. Welp & S. Uwe (Eds.), La dosis hace el veneno:
análisis de la revocatoria del mandato en América Latina, Estados Unidos y
Suiza (p p. 7– 30) . Qu ito : Der ech os r ese rvad os I nst itu to d e la D emo cra cia .
69. Uggla, F. (2008). Incompetence, Alienation, or Calculation? Explaining
Levels of Invalid Ballots and Extra-Parliamentary Votes. Comparative
Polit ical Stu dies, 41 (8), 1141–1164.
70. United Nations Development Programme (UNDP). (2013). Informe Sobre
Desarrollo Humano Perú 2013 Cambio Climático y Territorio: Desafíos
y Respuestas Para Un Futuro Sostenible, Retrieved from http://www.
pe.undp.org/content/peru/es/home/library/poverty/Informesobredesar-
rollohumano2013/IDHPeru2013.html
71. Universidad de San Martín de Porres. (2016a). Retrieved from http://usmp.
edu.pe/idp/wp-content/uploads/2016/09/densidadvb.xlsx
Recall Elections: A Tool of Accountability? Evidence from Peru
106
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
72. Universidad de San Martín de Porres (2016b). Retrieved from: http://
usmp.edu.pe/idp/wp-content/uploads/2016/09/perruravb.xlsx
73. Universidad de San Martín de Porres. (2016a). Retrieved from http://usmp.
edu.pe/idp/wp-content/uploads/2016/09/densidadvb.xlsx
74. Verba , S., & Nie, N . (1972 ). Part icipa tion in Am eric a-Pol itica l dem ocra cy
and social equality. University of Chicago Press.
75. Welp, Y. (2016). Recall refere ndums in Peruvian municip alities: a politica l
weapon for bad losers or an instrument of accountability? Democratiza-
tion, 23(7), 1162–1179. https://doi.org/10.1080/13510347.2015.1060222
76. Welp, Y., & Milanese, J. P. (2018). Playing by th e rule s of the game:
partisan use of recall referendums in Colombia. Democratization, 25(8),
1379–1396. https://doi.org/10.1080/13510347.2017.1421176
77. Welp, Y., & Whitehead, L. ( Eds.). (2020) . The Politics of R ecall Electi ons.
Cham: Palgrave Macmillan. https://doi.org/10.1007/978-3-030-37610-9
78. Whitehead, L. (2018). The recall of elected officeholders the growing
incidence of a venerable, but overlooked, democratic institution. Democ-
ratization, 25(8), 1341–1357. https://doi.org/10.1080/13510347.2018.1
455665
79. Whiteley, P. (2009). Party membership and activism in comparative
perspective. In J. DeBardeleben & J. H. Pammett (Eds.), Activating the
Citizen (pp. 131–150). New York: Palgrave Macmillan.
80. Zimmerman, J. F. (2014). Recall, The, Second Edition Tribunal of the People.
Albany: SUNY Press.
Michael Haman 107
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Appendix
Table 4. All variables
Variable Description Source
Political variables based on the previous municipal elections
Effective number of parties Laakso and Taagepera (1979) Author based on Infogob-
JNE (2017a)
Winner Share of votes in percentage points Infogob-JNE (2017a)
Closeness Victory margin between first and second
party in percentage points
Author based on Infogob-
JNE (2017a)
Incumbent Dichotomous variable - mayor is
incumbent (1) Infogob-JNE (2017a)
Turnout In percentage points Infogob-JNE (2017a)
Blank votes In percentage points Infogob-JNE (2017a)
Null votes In percentage points Infogob-JNE (2017a)
Female Dichotomous variable - mayor is female (1) Infogob-JNE (2017a)
Young mayor Dichotomous variable - mayor is younger
than 29 years (1) Infogob-JNE (2017a)
Female candidates Share of female candidates in all party-lists
in municipal election, in percentage points
Author based on Infogob-
JNE (2017a)
Young candidates Share of young candidates in party-lists in
municipal election, in percentage points
Author based on Infogob-
JNE (2017a)
Incidents in elections
Dichotomous variable - existence (1)
of incidents registered by the Dirección
Nacional de Fiscalización y Procesos
Electorales during elections
Infogob-JNE (2017a)
APRA Dichotomous variable - mayor was elected
for Partido Aprista Peruano (1)
Author based on Infogob-
JNE (2017a)
National party
Dichotomous variable - mayor was elected
(1) on the list of national party - party was
in Congress after legislative elections in
2006 or 2011
Author based on Infogob-
JNE (2017a)
Economic, social and political variables of municipality
HDI
Human Development Index; the following
years were used for organizing recall
elections: after municipal election of 2010 -
2012; after 2006 - 2007; after 2002 - 2003;
the same years were used for other variables
from the UNDP
UNDP (2013)
(Continued)
Recall Elections: A Tool of Accountability? Evidence from Peru
108
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Table 4. All variables
Variable Description Source
Economic, social and political variables of municipality
Expectancy of life Expectancy of life in years UNDP (2013)
Secondary education Population with secondary education in
percentage points UNDP (2013)
Years of education Average years of education (citizens over
25 years) UNDP (2013)
Income per capita growth Growth of family income from 2003 to
2012 per capita
Author based on UNDP
(2013)
Income Family income per capita UNDP (2013)
Party membership
Share of electorate in a municipality
listed as members of a political party
in percentage points
Infogob-JNE (2017a)
Size of electorate Number of votes in a municipality - logged Infogob-JNE (2017a)
Density Logged - population / km2 Universidad de San Martín
de Porres (2016a)
Rural area Population living in rural area in a
municipality in percentage points
Universidad de San Martín
de Porres (2016b)
Financial transfers per
voter
Logged - financial transfers to a
municipality from the state per
registered voter (2011)
Author based on Ministry
of Economy and Finance
(2017); Infogob-JNE (2017a)
Performance by mayor
Execution of budget
Measurement of performance - successfully
spend initial plan ned bu dget (2011), in
percentage points
Ministry of Economy and
Finance (2017)
Difference in investment
Share of investment in public projects made
by newly elected mayor (2011) to former
mayor in two previous years (2009, 2010) -
max. value - 2, I set value 2 to 65 cases with
value over 2 to avoid high skewness of data
Author based on Ministry of
Economy and Finance (2017)
Investment per voter Logged - investment in public projects per
registered voter 2011
Author based on Ministry of
Economy and Finance (2017)
Variables about recall
Experience Number of recall elections in a municipality
to 2010
Author based on Infogob-
JNE (2017a)
(Continued)
Michael Haman 109
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Table 4. All variables
Variable Description Source
Experience - province
Number of recall elections in province to
2010 (excluding municipality) divided by
the number analyzed of municipalities in
the province
Author based on Infogob-
JNE (2017a)
Electoral experience (in
correlations)
Dichotomous variable - organizer ran in any
elections in Peru (from 1990s to 2014)
Author based on Infogob-
JNE (2017a)
Electoral experience
(regression models)
Dichotomous variable - organizer ran in any
elections in Peru prior to 2012
Author based on Infogob-
JNE (2017a)
Table 5. Correlations – recall elections
DV: Recall elections after municipal
elections
DV: Mayor was recalled in recall
elections
2010 2006 2002 1998b2010 2006 2002 1998b
Effective number of
parties -0.035 -0.020 -0.006 0.048 0.131* 0.171** 0.171* 0.173*
Winner -0.045 -0.048 -0.046 -0.087** -0.148** -0.137* -0.174* -0.176*
Closeness -0.142** -0.088** -0.096** -0.097** -0.085 -0.014 -0.108 -0.142
Turnout 0.070** 0.040 0.089** 0.001 -0.034 0.078 -0.056 0.096
Blank votes -0.160** -0.051* -0.052* -0.058* 0.085 0.093 0.041 -0.046
Null votes -0.049* -0.032 -0.018 0.000 0.081 0.091 0.095 0.075
Size of electorate -0.185** -0.183** -0.141** -0.147** 0.020 0.132* 0.041 0.150
Incumbent 0.028 -0.010 0.053* -0.001 0.044 0.125* -0.098 0.042
HDI -0.033 -0.070** 0.054* 0.050 -0.084 -0.079
Expectancy of life 0.011 -0.020 0.016 -0.043 0.029 -0.125
Secondary education -0.025 -0.036 0.032 0.019 -0.087 -0.122
Years of education 0.000 -0.029 0.046 0.037 -0.130* -0.109
Income -0.047 -0.097** 0.056* 0.067 -0.037 0.000
Income per capita
growtha-0.045 0.088
Party membership 0.192** -0.057
Financial transfers
per voter 0.056* -0.025
(Continued)
Recall Elections: A Tool of Accountability? Evidence from Peru
110
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Table 5. Correlations – recall elections
DV: Recall elections after municipal
elections
DV: Mayor was recalled in recall
elections
2010 2006 2002 1998b2010 2006 2002 1998b
Execution of budget 0.057* -0.018
Experience 0.219** -0.038
Experience – province 0.203** 0.001
Density -0.090** 0.068
Rural area -0.032 -0.024
Investment per voter 0.036 -0.003
Difference in
investment -0.005 0.033
Female 0.051* -0.051
Young mayor -0.005 -0.022
Female candidates -0.056* -0.003
Young candidates 0.059* 0.024
Irregularities in
elections 0.104** -0.052
APRA -0.007 0.005
National party -0.010 0.102*
Organizer – electoral
experience 0.149** 0.151** 0.066 0.077
N 1605 1615 1622 1523 376 302 205 155
Note: a N for this variable is 1600 a 373; some municipalities were created after 2000.
b Only municipalities that held only one round of elections and had all data available were analyzed.
** Correlation is significant at the 0.01 level.,*Correlation is significant at the 0.05 level
Source: Author
Michael Haman 111
desarro. soc. 87, bogotá, primer cuatrimestre de 2021, pp. 73-111, issn 0120-3584, e-issn 1900-7760, doi: 10.13043/dys.87.3
Table 6. The organizers of recall elections – their electoral experience and success
in the removal of a mayor
Electoral experience No electoral
experience
Cases with el.
experience
Type of electoral experience Failure Success Failure Success N %
Candidate in the 2010 municipal
elections 69.1% 30.9% 77.6% 22.4% 81 21.5%
Candidate for mayor in the 2010
municipal elections 67.9% 32.1% 77.1% 22.9% 53 14.1%
Candidate in any election prior
to 2012 69.7% 30.3% 79.9% 20.1% 152 40.4%
Candidate for mayor in any election
prior to 2012 68.1% 31.9% 77.6% 22.4% 72 19.1%
Note: N – 376 cases of recall election 2012 and 2013
Source: Author
