Content uploaded by Juan Pablo Ruiz Nicolini
Author content
All content in this area was uploaded by Juan Pablo Ruiz Nicolini on Jul 19, 2019
Content may be subject to copyright.
Ponencia preparada para el XIV Congreso Nacional de Ciencia Política “La política en
incertidumbre. Reordenamientos globales, realineamientos domésticos y la cuestión de la
transparencia”, organizado por la Sociedad Argentina de Análisis Político y la
Universidad Nacional de San Martín, San Martín, 17 al 20 de julio de 2019
Reassessing Ballot Structure
and Electoral Behavior in
Argentina’s Provinces.
(New) Evidence from Natural
Experiments.
SIMPOSIO: Elecciones, reforma política y observación electoral
Juan Pablo Ruiz Nicolini1
1Universidad Torcuato Di Tella (UTDT), Argentina
2juan.ruiznicolini@mail.utdt.edu
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
Abstract. While a previous quasi-experiment research of Salta, Argentina elec-
tions stands that a shift in voting technology and a new ballot structure affected
split-ticket voting, our research goes further. Taking advantage of the opporunity
of natural experiments, this article reasses the effects that new voting systems can
have on seral electoral behaviors. With the two novel provincial cases of Chaco
and Corrientes we are able to (a) integrate different behaviors -not only split-tcket
voting; (b) add geographic variation and electoral regulations and (c) doing so by
relaxing methodological assumptions of previous research.
(91 words)
Keywords
natural experiment - electoral behavior - voting technology - ballot structure -
electoral reform
Note: A previous spanish version was published in Revista SAAP as a research
note (Dodyk and Ruiz Nicolini, 2017). Looking forward to bring some aditional
evidence to support the as-if-random claim of the natural expermiment, this
transalated version is still a work in progress.
2
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
While the incorporation of electronic voting systems (EVS) has been
accelerating in the world, particularly in developing countries, resistance
to their use has increased in many of the most advanced democracies
(Hidalgo, 2010). Some of the eVote pioneers, such as Ireland, the Nether-
lands and Germany reversed their implementation. Lack of guarantees to
ensure the security and integrity of elections appear as the main founda-
tions of those who warn about the risks of incorporating technology into
casting votes procedures (Aguerre, 2017; Feldman et al., 2007). Beyond
that paradox, this technological change is probably the biggest transfor-
mation in the voting instrument since the generalization of the “Australian
Ballot”.
Increasingly, political science literature has focused on less discussed
aspects of these policy reforms: the possible effects that the introduction
of new technologies to cast the vote can have on voting behavior and
the electoral results. This article follows this development and aims to
overcome some existing limitations in the literature.
Taking advantage of the opportunity to exploit natural experiments,
we extend Barnes et al. (2017) quasi-experiment research of EVS in Ar-
gentina s subnational elections. Fundamentally, our research strategy -
based on the as-if random allocation of voters within electoral precints
becasue of their last name3- allows us to sum evidence of multiple voters
behaviors that can be understand because of changes in ballot layouts.
3Previous research on México (Cantú, 2014) and Argentina (Casas et al., 2017) used this same
identification strategy for their analyzes.
3
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
This stochastic voter assignment allows us identify groups of vot-
ers who casted votes with the new voting system (treatment) and others
who did it in the traditional way (control). While in these new cases we
were able to identify variation within electoral precints (where voters are
assigned to polling stations according to their surnames), Barnes et al.
(2017) analysis of the Salta is based on a comparisson between precints,
wich aggregation level has no variation explained by the surnames iden-
tification strategy.
Dunning (2012) argues that the effort to use random processes (or as-
if they were random) to study the effect of causes -instead of trying to con-
trol statistically by covariates- is the main feature of natural experiments.
In fact, that´s what distinguishes natural experiments from other conven-
tional observational studies, such as quasi-experiments or matching de-
signs4. While random assignment makes it possible to control observable
and unobservable factors, quasi-experiments can only attack threats to
the inference of the former, under the assumption that the researcher can
control by relevant known covariables (Dunning, 2012, pp.15-21).
Therefore, the opportunity to exploit a natural experiment to evaluate
the impact that voting instruments can have on the voters behavior is for
us the main virtue of this work, on which we rely on less methodological
assumptions, presenting a more transparent inference mechanism.
4Barnes et al. (2017) combine a matching strategy with the Double Difference design for the
Salta case
4
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
Ballots structure matters
Although most research that look for political effects of voting sys-
tems layouts are limited to central countries (Alvarez and Hall, 2008;
Card and Moretti, 2005; Ansolabehere and Stewart, 2005), an incipient
body of work has been developed on cases of less developed countries,
where the expansion in the use of new technologies is more remarkable.
For example, Hidalgo (2010) - with a Regression Discontinuity re-
search design - finds that a shift of voting system in Brazil reduced ar-
tificially the number of votes from the final count5. Similarly, Fujiwara
(2015) extends the analysis and finds effects of this reincorporation of cit-
izens excluded from electoral processes in the provision of public health
services. Ravi et al. (2017) evaluate the India case and find that electronic
voting was effective to reduce fraudulent maneuvers in the electoral pro-
cesses. Zucco and Nicolau (2016) note that although in Brazil the com-
puterization of the vote minimized errors in the tabulation of the results
(reducing the high percentage not counted votes), the pledge of change
was the growth of of party votes percentage that can only be explained
by another type of error.
5He argues that the combination of political machineries linked to conservative parties - which
had control overthe votes counting process-, together with the difficulty of correctly issuing
votes for less-trained electorates, actuallyr educed a substantive proportion of votes with the
old system, biasing political representation of Congress. The change of voting technology
had the effect of incorporating votes into the electoral process, so analogous to what happened
with the elimination of illiteracy requirements or property to vote in the century XIX (Hidalgo,
2010, pp. 2-3)
5
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
In Argentina, voting techonolgy has not undergone drastic changes
for national elections,where party ticket ballot with envelopes is still used.
But a digital revolution grows from below, with several experiencies from
2003 onwards. Provinces and municipalities, for almost 15 years, have
turned Argentina s electoral systems into a virtual laboratory for political
analysts.
A growing body of research exploit this technological transforma-
tions with different cases and research designs. They include techincal
reports of the political implications of the shift from the traditional paper
and envelope system to variants of Australian Ballot during 2011 election
process (Leiras and Calvo, 2011); field experiments that asses the effects
of different e-vote systems on split-ticket behavior (Calvo et al., 2009))
or it s informational cues relevance for voters choice (Katz et al., 2011),
both carried out in parallel to the 2005 legislative election in the City of
Buenos Aires; and, recently, quasi experimental analysis that asses the
impact of a new voting technology on electoral behavior of a real elec-
tion in Salta, where the new techonology was sequentially implemented
during 2011-2015 (Barnes et al., 2017).
This article aims to make a stepforward in this development exploit-
ing the opportunity of two natural experiments in Corrientes city local
elections and Chaco s provincial legislative midterm elections, both dur-
ing 2017 electoral cycle. Our main argument is based on the claim that
there is no relationship between the first letter of the surname of each
voter and our variables of interest, that seek to evaluate voting technol-
ogy impact on electoral behavior.
6
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
Results
In both Chaco and Corrientes 2017 electoral processes, two subsets
of the voting population casted votes with different technologies. There
were precints in which some voted using the new EVS, while others main-
tained the traditional party paper ballots and envelope system. The as-if
random condition, that determine wich voters used each ballot technol-
ogy, permited us identify both treatment and control groups and estimate
differences in outcomes due to treatement effect (voting with an EVS).
This is the reason why we ara able to relax previous methodological as-
sumptions.
The key feature of our research design is that voters werw assigned
to a ballot box (within a polling station) based on the alphabetical order
of their surname. Therefore, it is plausible to argue that there is no rela-
tionship between the assignement and voters behavior.
Argentina´s provinces as if they were laboratories
For the one side, Corrientes elections defined mayor and councilors
positions. The division into two categories made split-ticket voting pos-
sible (voting for different lists in each category). This allowed us esti-
mate the effect of the EVS on split-ticket voting, following Barnes et al.
(2017), making a contribution to the literature offering new evidence de-
rived from a natural experiment. On the other hand, the Chaco provincial
general election was preceded by a primary electoral process (PASO for
its spanish name6). This sequence allowed us to estimate the effect of the
6Primarias, Abiertas Simultáneas y Obligatorias - PASO.
7
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
EVS onthe electoral volatility between the PASO candidate selection pro-
cess and the legislative election. At last, following Abdala and Antenucci
(2016), we were also able to estimate for both cases the effect on other
political outcomes, such as party advantages and proportion of positive
votes, determinant for legislative seats distributions.
Corrientes
The election had voting centers (30 out of 798) from two different
electoral precints (3and 5b) that voted with an electronic voting system
(EVS). In each of the precints we identified centers where they voted with
the new system and with the traditional party ballots. As we noted earlier,
who voted in each polling place was defined ex ante by the surname rule.
The prevalence of cross vote in a polling station is not observable
with accuracy based on electoral data, because they report the aggregate
results in each category, but not the lists selection at the individual level,
given that the vote is secret. We can calculate, in any case, a lower bound
for said prevalence, given by the following formula: V Ci=∑k|Iik −
Cik|/2 where the sum goes through the lists kin competition and Iik &Cik
measure the proportion of voters for the klist for mayor and councilors
categories, respectively. We will use this indicator as an estimate.
The case of Corrientes differs from the cases of Salta and Chaco an-
alyzed in previous studies by another of the norms that regulate compe-
tition: the possibility of presenting “mirror lists”7. Given the provincial
legislation, a list was prepared for each of the member parties of the com-
7Throughout the Argentine electoral federalism we find a variety of practices that affect the
clarity of the electoral offer: the ”Ley de Lemas” (Santa Cruz), ”Acoples” (Tucumán) or an-
8
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
peting alliances: the lists presented the same candidates but the design of
each varied according to the name / number assigned to each party (On-
line apendix).
In political terms, this design allows the parties to measure forces
inside their own alliance and push the Mayor candidacy from multiple
labels. Votes for each one of the ”mirrors” allow us distinguish with pre-
cision the contribution of each one to the aggregate of the votes.
Three were the alliances that presented candidates to Intendente. At
the same time these were conformed by a multiplicity of lists: the Frente
para la Victoria by 2; the oficialista Haciendo Corrientes by 12 and its
main contender -winner of the contest-, ECO + Cambiemos by 21. The
Sankey diagram on Figure XX (Online appendix) illustrates the results.
First, the EVS had no significant effect on the proportion of votes to
each candidate for mayor (Figure 1). We also found no evidence that the
result of the election could have been different if the EVS had not been
used in the two centers where the pilot test was carried out. In effect,
the result of simulating the electoral result changing the tables with EVS
by tables without EVS of the same circuit (which, on average, should
have the same result if we discount the effect of the EVS) did not present
significant differences.
other type of ”Colectoras” (Salta) make it more difficult (i) the administration of the electoral
process and (ii) generate confusion in the electorate.
9
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
Fig. 1. EVS effect on the percentage of votes, with confidence intervals of 95%. When a horizontal
black line does not cut to the vertical dotted line it means that the effect on that percentage of votes
is significant.
On the other hand, significant effects are observed on the percentage
of blank votes and null votes, but in the opposite direction (Figure 1).
With EVS there are more blank votes but fewer null votes. This may be
because the same behavior (not voting for any candidate) is expressed
differently with EVS (as a blank vote) than with a paper ballot (less as a
blank vote, and more as a null vote: cut in half, crossed out, etc).
In the case analyzed here, the positive effect on the “blank vote” is
partially offset by the negative effect on the “null vote”. In both circuits
the point estimate of the effect of the EVS on the non-affirmative votes
is positive, although only in the 5B circuit is statistically significant.
The Corrientes analysis allows us to add evidence that the modifica-
tion of the issuance instrument has an effect on the proportion of “ballot
cuted” or split-ticket. For this we considered two types of crossed vote:
10
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
one “between lists” and another “between alliances”, with the intention
of capturing the effect of “mirror lists”.
The first corresponds to the split-ticket itself: opt for a list for mayor
and another for councilor. This is considerably facilitated with the EVS
interface, which does not require scissors (as the traditional party-ballot
system does). The ”ballot cut”, or split-ticket, in this case is done by press-
ing options on a touch screen. The system presents a first screen where
the voter can choose between voting (a) ”Complete List” or (b) ”by Cat-
egories”, although without any clear identification of what will happen
on the next screen. If option (b) is chosen, the voter sees in subsequent
screens all the lists that are presented for each of the categories in dispute
in a given district. The fact that different categories are presented inde-
pendently of each other in each screen raises the chances that voters will
choose alternative political forces.
If with the traditional system a voter has to make a conscious effort
to vote two different lists in two categories in dispute (use the scissors
to separate categories in a ballot that brings them together), with the new
system the situation is reversed: the difficulty it goes in the opposite di-
rection, since if they want to vote for the same party the voters have to
“gather” the votes of their selection in each of the categories (Leiras and
Calvo, 2011). The effect of the EVS is statistically significant in the two
precints analyzed when we do it with every list.
Finally, the second type of split-ticket vote (“between alliances”) cor-
responds to voting a candidate for mayor and a list of councilors that goes
with another mayor candidate. For example, a vote to Tassano (ECO-
11
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
Cambiemos) for mayor, and, for councilors, to a list that does not got him
as a candidate (that is, none of the 23 lists that do have him as a can-
didate). Given the system of mirror lists, this split-tiket voting is more
“substantive”: it expresses a clear split of the electoral preference in the
two categories, which is not merely a consequence of the multiplicity of
lists in which each candidate appears. The effect of the EVS on this mea-
sure (more restrictive) is smaller, but points estimates are positive for both
precints and is statistically significant in one of them.
Chaco
In Chaco s provintial mid term elections a EVS was partially adopted
in 641 over 3104 ballot centers of 12 electoral precints. These were dis-
tributed in 10 ”first category municipalities” (more than twenty thousand
inhabitants). The partial implementation allow us estimate the effect of
the EVS on electoral behavior.
Eleven parties presented candidates on the PASO election, but only
2 had real internal competition and others presented only one list: the
provincial government, Frente Chaco Merece Más (FCHMM), which pre-
sented 11 lists, and the oficialismo at the national level, Cambiemos, that
presented 3. The FCHMM obtained the highest number of votes (47%),
followed by Cambiemos (32%), Partido del Obrero (PO, 4%) and the
Kirchnerist Frente Grande party (3%). The effective number of internal
lists8submitted by FCHMM was 2.24, with a list that took 66% of the
votes to the front, and another 10 that were between 1% and 5%. The
8Calculated as 1/ ∑kv2
k, where vkis the proportion of votes received by the list k.
12
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
competition within Cambiemos was lower: the effective number of lists
was 1.34, with a list that took 86% of the front’s votes; the rest took 8%
and 6%
As in Corrientes case, we estimate the magnitude of the effect of the
EVS on parties or fronts votes shares (FCHMM, Cambiemops, PO, and
others, that is, the sum of the others), blank vote, null votes and non-
affirmative votes (white and null addition). We also estimate the effect on
the difference |Pk−Gk|between PASO results (Pk) and general election
ones (Gk) of each party k, which indicates the magnitude of vote volatilty
between both elections.
Since we have 12 precints, we use a mixed-effects model, in which
the constant term (which measures the average value of the variable under
the traditional voting system) varies between precints (fixed effect per
precints), and the effect of the EVS is random between precints. We call
βto the mean effect of the EVS, which is the parameter of interest. Figure
2 shows estimates with a 95% OLS confidence interval.
No afirmativos
Nulos
Blanco
Otros
PO
Cambiemos
FCHMM
−5.0 −2.5 0.0 2.5 5.0
PASO
No afirmativos
Nulos
Blanco
Otros
PO
Cambiemos
FCHMM
−5 0 5
Elecciones legislativas
No afirmativos
Nulos
Blanco
Otros
PO
Cambiemos
FCHMM
−4 −2 0
Diferencia
Fig. 2. EVS effect on the percentage of votes, with confidence intervals of 95%. When a horizontal
black line does not cut to the vertical dotted line it means that the effect on that percentage of votes
is significant.
13
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
In contrast with Corrientes estimations, the EVS has a significant ef-
fect on the vote for most competitive parties: in both the PASO and the
general elections the EVS caused an average loss of 5% (over total votes)
aCambiemos, and in the general election a 6% advantage for the FChMM.
On the other hand, in the PASO the Partido del Obrero -PO obtained a
1% advantage associated with the EVS, while other parties obtained an
advantage of 2%. At last, the EVS had a zero average effect on PO votes
in the general election, and a negative effect (-1%) on smaller parties.
We ran a simulation based on the mixed effects model to predict what
the election results would have been if the EVS had not been applied (see
online appendix). The model assumes that precints in which the EVS was
fully implemented the effect on the vote follows the same distribution
(modeled as a normal) than precints in which the implementation was
partial. Doing this we extrapolate the inference of the effect of the EVS on
the precints in which the implementation was partial to those in which it
was fully implemented. Based on 20,000 iterations, the conclusion is that
having voted with the traditional voting system, the FCHMM would have
obtained 8 seats instead of 9, and Cambiemos, 7 instead of 6. We affirm,
then, with 77% certainty, that the EVS benefited the ruling party, granting
it a seat lost by the opposition. On the other hand, in all the simulations
PO party obtained one seat, due to the almost negligible effect of the EVS
on its vote.
In line with Corrientes estimations, the EVS had a significant im-
pact on blank vote (4% in the PASO election and 2% in the general elec-
tions), which was partially offset by a reduction in null votes. Summed
14
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
up, non-affirmative votes percentage (white and null) grew on average by
2.5points in the PASO and in 0.5in the general election.
On the other hand, we estimate that EVS had a negative impact on
cross-election vote, our volatility meassure:FCHMM fall by an average
of 3% between PASO and general elections, 2% the difference for Cam-
biemos,0.5% dor PO and 1.5% for minor parties. However, because vot-
ers for both elections were not the same (turnout grew by 12% between
elections), these results should not be interpreted as a fall in the propor-
tion of voters who changed their vote due to the EVS.
Discussion
It is usual that when those in charge of electoral governance debate
the implementation of electronic voting systems, the main arguments re-
volve around issues of tech security and integrity of electoral processes.
However, increasingly political science research focuses more on the po-
tential effects that new voting technologies can have on voters behavior
and electoral outcomes. Following the last, this artilce provides evidence
of how the change in voting instrumetns may influence the way voters
interact with technology and results derived from that process.
While some of our results are consistent with previous findings, in
this article we were also able to provide new evidence and resolve some
limitations in the literature. On the one hand, our research design takes
the opportunity to make an analysis based on natural experiments. This
opportuinity allowed us to present results that derive from a more trans-
15
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
parent research design in which the inference process is based on fewer
assumptions.
At the same time, we were able to integrate in the analysis a variety of
electoral behaviors (up to here analyzed separately) for two different sce-
narios (with different regulations and contexts). First, the analyzed cases
confirm findings suggested in Leiras and Calvo (2011) and analyzed in
Abdala and Antenucci (2016) with respect to the decrease in null votes
(caused by the device interface) and the increase in blank votes. Con-
sidering these variables together, they have a positive impact on non-
affirmative votes proportion, which reduces the base on which seats are
distributed in legislative elections.
Secondly, contrary to our Corrientes report and Abdala and Antenucci
(2016) findings regarding Chaco´s elections in 2015, we find a significant
effect on the proportion of votes received by frontrunners in 2017 Chaco´s
election. Depending on whether votes were casted with the traditional
party ticket system or an EVS, we find differences for main contenders
support. With these data, we ran a simulation to evaluate the effect on the
distribution of seats and we found a high probability that the first force
obtained an extra seat at the expense of its main opponent that could be
explained by the use of an EVS.
We also estimate for Chaco elections the split between electoral turns
(which functions as an indicator of volatility) and find significant effects
for several political parties.
Finally, in the case of Corrientes we follow Barnes et al. (2017) strat-
egy. Like them, we find a positive effect of using an EVS on the split-
16
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
ticket in the same electoral turn. Grow that could open representative pos-
sibilities to relatively weak parties in the legislative arena. This is Barnes
et al. (2017) argument when analyzing Salta´s case. For these authors,
the central explanation behind the increase in the split-ticket is based on
strategic voters, who find an oportunity to maximize the performance of
their votes in the costs reduction that a new voting technology permits.
Some of the results presented here (such as the significant effect of EVS
on competitive forces between electoral shifts) suggest that other expla-
nations are worthwhile. While future work should investigate and provide
evidence on some of the causal mechanisms behind these findings, these
evaluations allow us to argue that designs matter: the devil is in the de-
tails.
17
Bibliography
Abadie, A. (2005). Semiparametric Difference-in-Differences Estima-
tors. The Review of Economic Studies, 72(1):1–19.
Abdala, B. and Antenucci, P. (2016). Evaluación de efectos de la boleta
electrónica. evidencia experimental de las elecciones de chaco en 2015.
Revista SAAP, 10(2):339.354.
Aguerre, T. (2017). Voto electrónico: un debate entre lo seguro y lo mod-
erno. In Voto electrónico. Una solución en busca de problemas, pages
37–53. Fundación Vía Libre, Buenos Aires.
Alvarez, R. M. and Hall, T. (2008). Electronic Elections: The Perils and
Promises of Digital Democracy. Princeton University Press.
Ansolabehere, S. and Stewart, C. (2005). Residual Votes Attributable to
Technology. The Journal of Politics, 67(2):365–389.
Barnes, T. D., Tchintian, C., and Alles, S. (2017). Assessing Ballot Struc-
ture and Split Ticket Voting: Evidence from a Quasi-Experiment. The
Journal of Politics, 79(2):439–456.
Brady, H. E. and Collier, D. (2010). Rethinking social inquiry: Diverse
tools, shared standards. Rowman & Littlefield Publishers.
Busaniche, B., editor (2017). Voto electrónico. Una solución en busca de
problemas. Tren en Movimiento, fundación vía libre edition.
Calvo, E., Escolar, M., and Pomares, J. (2009). Ballot design and split
ticket voting in multiparty systems: Experimental evidence on infor-
mation effects and vote choice. Electoral Studies, 28(2):218–231.
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
Cantú, F. (2014). Identifying Irregularities in Mexican Local Elections.
American Journal of Political Science, 58(4):936–951.
Card, D. and Moretti, E. (2005). Does Voting Technology Affect Election
Outcomes? Touch-screen Voting and the 2004 Presidential Election.
Working Paper 11309, National Bureau of Economic Research. DOI:
10.3386/w11309.
Casas, A., Díaz, G., and Trindade, A. (2017). Who monitors the moni-
tor? Effect of party observers on electoral outcomes. Journal of Public
Economics, 145(Supplement C):136–149.
Chaparro, E. (2016). El sistema de voto electrónico en la ciudad de
buenos aires. una “solución” en busca de problemas. Fundación Vía
Libre.
Dodyk, J. and Ruiz Nicolini, J. P. (2017). Enchufes, espejos y tijeras:
efectos del diseño de las boletas sobre el comportamiento electoral.
Revista SAAP (ISSN 1666-7883), 11(2):365–386.
Dunning, T. (2010). Design-based inference: Beyond the pitfalls of re-
gression analysis? In Rethinking social inquiry: Diverse tools, shared
standards, pages 273–311.
Dunning, T. (2012). Natural Experiments in the Social Sciences.A
Design-Based Approach. Cambridge University Press.
Feldman, A. J., Halderman, J. A., and Felten, E. W. (2007). Security
Analysis of the Diebold AccuVote-TS Voting Machine. In Proceedings
of the USENIX Workshop on Accurate Electronic Voting Technology,
EVT’07, pages 2–2, Berkeley, CA, USA. USENIX Association.
Fujiwara, T. (2015). Voting Technology, Political Responsiveness, and
Infant Health: Evidence From Brazil. Econometrica, 83(2):423–464.
19
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
Gertler, P. J., Martinez, S., Premand, P., Rawlings, L. B., and Vermeersch,
C. M. (2016). Impact evaluation in practice. World Bank Publications.
Hidalgo, F. D. (2010). Digital democracy: the consequences of electronic
voting technology in Brazil. New Faces in Political Methodology III,
Penn State Quantitative Social Science Initiative, Pennsylvania State
University.
Katz, G., Alvarez, R. M., Calvo, E., Escolar, M., and Pomares, J.
(2011). Assessing the Impact of Alternative Voting Technologies on
Multi-Party Elections: Design Features, Heuristic Processing and Voter
Choice. Political Behavior, 33(2):247–270.
King, G., Keohane, R. O., and Verba, S. (1994). Designing social inquiry:
Scientific inference in qualitative research. Princeton university press.
Leiras, M. and Calvo, E. (2011). La forma de votar importa. El impacto
de los nuevos instrumentos de votación sobre la conducta electoral en
las provincias argentinas. Technical report, CIPPEC - COPEC.
Morgan, S. L. and Winship, C. (2014). Counterfactuals and causal infer-
ence. Cambridge University Press.
Ravi, S., Debnath, S., and Kapoor, M. (2017). The impact of Electronic
Voting Machines on electoral frauds, democracy, and development.
Ruiz Nicolini, J. P. (2017a). El impacto de enchufar los votos. Evalu-
ación del voto cruzado entre categorías en las elecciones de la provin-
cia de Salta. Master’s thesis, Universidad Torcuato Di Tella. DOI:
10.13140/RG.2.2.24737.20326.
Ruiz Nicolini, J. P. (2017b). Fuego amigo en torno a la BUE. El Estadista.
Zucco, C. and Nicolau, J. M. (2016). Trading old errors for new er-
rors? The impact of electronic voting technology on party label votes
20
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
in Brazil. Electoral Studies, 43(Supplement C):10–20.
21
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
APPENDIX
A. Mirror ballots (Corrientes)
Fig. 3. Mirror Ballots designs. Top: party ballot. Down: EVS. Source:El Litoral & Junta
Electoral Provincial de Corrientes - http://www.juscorrientes.gov.ar/elecciones/
elecciones-municipales-corrientes-4-junio-2017/
22
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
B. Mirror Ballots and party coalitions in Corrientes - Mayor
(Intendente) section. Sankey Diagram
Fig. 4. Left (origin): individual ticket that carried the same candidate for mayor within each al-
liance (Mirror ballot). Right (destiny): electoral alliances posited by a candidate for mayor (ECO-
Cambiemos,Frente Para la Victoria yHaciendo Corrientes). Values represent votes count of each
ticket and alliance. Source: http://elecciones2017.corrientes.gov.ar/totalianza.
php
23
Reassessing Ballot Structure and Electoral Behavior J.P. Ruiz Nicolini
C. Electronic Voting System (EVS)
In both provinces the new voting instruments was partially carried
out. In Corrientes only a couple of schools used the EVS for the capital
city municipal elections. In Chaco, with record of previous pilot trials,
the EVS was used in several municipalities; in some cases whole precints
voted with the new technology, in others only with the traditional paper
ballots and some mixed technology (which we use here).
Corrientes Chaco
Municipialities 1 10
Precints 2 12
Ballot boxes 30 / 798 641/3104
(% ballot boxes)(3.7%) (20%)
The two precints in Corrientes case are 3 and 5b.
The selected circuits in the Chaco case are: 119, from Charata; 22A,
from Fontana; 62, from General José de San Martín; 136, from Juan José
Castelli; 117, from Las Breñas; 104, of Machagai; 88, of Presidencia
Roque Sáenz Peña; 98, of Quitilipi; 13A, 16A and 8A, of Resistencia;
and 78, of Villa Ángela.
D. Electoral Files
Electoral legislation that rules over Chaco and Corrientes electoral
files are consistent with the National Electoral Code that regulates na-
tional elections. In it´s 29º article it states that ”the voter register will be
24
J.P. Ruiz Nicolini Reassessing Ballot Structure and Electoral Behavior
ordered according to territorial demarcations, the corresponding polling
stations and alphabetical by last name”.
Corrientes Electoral Code (Decreto-Ley N° 135/01)9and it´s Elec-
toral Law (Ley N.4169)10, wich regulate electoral processes within the
provinces, establish in their second chapters the ordering that electoral
files must follow, specifying the subdivisions within which all the voters
are located: first demarcating electoral sections within the territory; inte-
grated these by electoral circuits and, finally, ordering alphabetically to
each one of the registries.
This identification allows us to argue that the assignment of certain
voters to a school in which they vote with one or another system - that
is, the determination of treatment and control groups - can be considered
as-if-random; and therefore, that groups are equivalent in expectations -
both for potential outcomes and for unobservable covariates. Under these
conditions, on average, any observable difference of a variable of interest
will respond to the causal effect of the treatment, which is what we seek
to estimate.
9Source: Observatorio Electoral Argentino (CIPPEC) http://oear.cippec.org/
wp-content/uploads/2015/09/CODIGO-ELECTORAL-PROVINCIAL.pdf
10 Source: Tribunal Electoral de Chaco www.electoralchaco.gov.ar/images/pdfs/2014/
Ley_4169.pdf
25
