Archive | Electoral Fraud

The Protesters Are Right: Evidence Suggests More Election Fraud in Last Week’s Elections in Armenia

This is a guest-post from Fredrik M Sjoberg, a Postdoctoral Scholar at Columbia University – The Harriman Institute.


Since last week’s presidential elections (February 18) in Armenia there have been protests around the country alleging widespread election fraud. The presidential campaign was certainly eventful, with one candidate going on a hunger strike and another being shot (non-fatally). The officially reported electoral returns came as a surprise to many. The American-born Raffi Hovannisian from the oppositional Heritage Party, got 37 percent of the votes. Considering that in last year’s parliamentary elections the party got only 3.8 percent of the seats, this can be considered quite a feat. Many questions arise, one of which concerns the extent of election fraud in the 2013 elections compared with last year’s parliamentary elections. I find that there are indications of more fraud in last week’s elections.

First, examining indications of outright fabrication I note a suspicious deviation from the expected uniform distribution in the vote counts for the ruling party in 2013. There is no such pattern in the 2012 data, supporting the hypothesis that there was more fraud in 2013. In the absence of manipulation of vote totals the last digit should follow a uniform distribution of 10 percent in each of the 0-9 digit categories – the vertical bars in the graph (Beber & Scacco, 2012). The deviation from uniform is significant at the 0.1 percent level in the 2013 elections, while non-significant in 2012 (p-value = 0.981)

There is anecdotal evidence of fraud from domestic election observers, like the crowdsourcing platform There are only four incidents of calculation errors or falsification, constituting 1 percent of the 390 incident reports that were categorized. Actually, the most frequently occurring types of incidents reported where either intimidation (11 percent) or vote buying (10 percent). Fraud forensics based on digit-tests would not be able to capture such phenomena.

For the latter type of electoral manipulation we can graph turnout as a kernel density plot to detect anomalies. A Kernel plot is a non-parametric estimation of the probability density function of a random variable. Such a smoothing of the distribution helps us spot irregularities compared with the expected normal distribution. Examining the plots visually reveal an ‘unnatural’ hump at the right-hand tail as we would expect if there was ballot stuffing, or other blatant forms of fraud that would increase turnout. The hump is certainly more articulated in 2013 compared to 2012, again suggesting that there was more fraud in 2013.

In a context where authorities claim that there was no fraud and the opposition alleges widespread fraud, such forensics tools can be helpful as a piece of evidence. These are only diagnostic tools though and should be interpreted with caution. Other organizations have put out similar analyses, for instance Policy Forum Armenia (PFA). It remains to be seen whether the citizens of Armenia overcome the well-known collective-action problem of whether or not to protest. Who knows, the spring might come early to Armenia this year?

Continue Reading

2012 Georgian Parliamentary Election: Fraud Forensics and Map of Opposition Support

This following is a guest-post from Fredrik M Sjoberg, a Postdoctoral Scholar at Columbia University.


The country of Georgia experienced something unique last Monday. The incumbent President conceded defeat in the parliamentary elections. This is the first time ever anywhere in South Caucasus or Central Asia that this happens. Georgia showed that it clearly wants to distance itself from its post-Soviet neighbours where election fraud is widespread. How different were these elections in terms of precinct level dynamics?

Now that the Central Election Commission (CEC) has published the precinct level electoral returns we can use fraud forensics to look for irregularities. Note that it took the CEC relatively long to publish the data, after first claiming that its website was under a Denial-of-service attack (see article). The data used here is based on a download from the official results as of 6pm (EDT) Tuesday October 2, when 92 percent of the precinct level results were in.

First, examining indications of outright fabrication we note a suspicious deviation from the expected uniform distribution. In the absence of manipulation of vote totals the last digit should follow a uniform distribution of 10 percent in each of the 0-9 digit categories (Beber & Scacco, 2012). The deviation from uniform is significant at the 5 percent level.

This is not to say that the results were fabricated, but it does call for more research. Note that none of the major election observation organizations, domestic or international, reported widespread vote-count fraud.

Second, a kernel density plot can also be used as an indication of fraud (Hyde, 2007). A Kernel plot is a non-parametric estimation of the probability density function of a random variable. Such a smoothing of the distribution helps us spot irregularities compared with the expected normal distribution. Examining the plots visually reveal no ‘unnatural’ humps at the right-hand tail as we would expect if there was ballot stuffing, or other blatant forms of fraud, that would benefit one party (or increase turnout).

Finally, let me just illustrate how the support of the opposition is distributed spatially. Georgian Dream, the party led by the billionaire Bidzina Ivanishvili, performed very well and especially so in the proportional component of the mixed electoral system. A map of the support indicates that support was spread out over the whole territory.

No doubt there will be a lot more interesting analyses coming out in the next few days. In the meantime, for a field experiment examining election fraud in the preceding Georgian parliamentary elections in 2008, see the paper Making Voters Count: Evidence from Field Experiments about the Efficacy of Domestic Election Observation.

Continue Reading

Electoral Integrity Project Requests Beta Testers of Expert Survey

Pippa Norris sends along the following announcement:


We are in the process of developing a new survey monitoring expert Perceptions of Electoral Integrity (PEI) and we need your help!

The survey plans to contact c.40 experts immediately after each national election (constituent assembly, presidential or parliamentary) to gather views about these contests. Experts will be drawn from many different fields, such as scholars, lawyers, election officials and journalists familiar with elections. We will seek a balance of local residents and international experts and we will aim to achieve c.25 replies per country, working with international agencies and regional networks to identify email contacts for suitable experts. The survey is planned to be repeated after each national contest. The aggregate results for each country will be published via the project website, broken down by question category, as well as through a standardized index measuring Perceptions of Electoral Integrity. National coverage is planned to be global. Over time, as results accumulate, we hope that this will generate valuable information for the research and policy-making communities.

It would be greatly appreciated if colleagues could look briefly at the beta testing version of the survey and if you could send us comments.

Colleagues may want to try completing the survey, to see if it works for the last election held in each country, or alternatively just send any criticisms, feedback, and ideas to . Either way, we would greatly appreciate hearing your advice, no matter how critical. We want to develop the right instrument. Opportunities to comment on the beta testing process will close by 1st October 2012. The survey fieldwork will start in January 2013 and the preliminary results will be evaluated at the Electoral Integrity Project’s next annual workshop, to be held at Harvard 3-4th June 2012.

We look forward to hearing from you. In addition, several new draft papers are also now available for comment via the website

Pippa Norris (Harvard University and the University of Sydney), Andrew Reynolds (University of North Carolina) and Jorgen Elklit (Aarhus University)

Continue Reading

Making Voters Count: Field Experiments with Domestic Election Observers

This following is a guest-post from Fredrik M Sjoberg, a Postdoctoral Scholar at Columbia University (who is still disappointed about not being able to present these findings at APSA….). The paper on which this post is based is here.


Why is it that voters in autocratic states report a higher confidence in the honesty of elections, compared to more competitive hybrid regimes? Five-year averages on the Gallup World Poll question about honesty of elections show that the proportion of people that respond affirmatively is 53% in Azerbaijan and Tajikistan, and 39 % in Kazakhstan – all solidly autocratic cases. In the more competitive regimes of Georgia, Ukraine, and Kyrgyzstan, the numbers are paradoxically lower, with 31%, 14%, and 23% respectively. Perhaps the explanation can be found in exposure to voter manipulation? In autocratic countries where elections can be rigged bureaucratically the voters are marginalized. In more competitive regimes where parties have to rely on voter manipulation voters need to be engaged directly. In a recent paper I illustrate the micro-dynamics of fraud with the help of field experiments in five separate elections in three different countries.

For the sake of simplicity, we can separate out bureaucratic fraud, happening inside polling stations, from voter manipulation, happening outside. When faced with a fraud-reducing intervention, like election monitoring, parties as strategic actors can adapt and substitute one form of manipulation for another. When observed by international observers, spending on average less than one hour per polling station, the adaptation can take the form of temporarily suspending bureaucratic fraud. Domestic observers are tougher since they are often assigned to cover one polling station for the entire election day. In addition, domestic observers have more detailed knowledge about malpractices and the local context. Domestic observers, therefore, theoretically reduce bureaucratic fraud, but as a consequence might cause an increase in voter manipulation, a form of perverse empowerment of voters (which in turn might be driving the perceptions about the honesty of elections).

In the first ever cross-national study of domestic observers it is shown that domestic observers do indeed reduce fraud. First, turnout is used as a proxy for ballot box stuffing, especially relevant in countries with a history of inflated turnout. Here randomly assigned observers reduce turnout by 1-3 percentage points:

Second, new fraud forensics techniques, in the form of last digit tests, are used to capture falsification of electoral returns. Also here, observers seem to reduce it as illustrated by the case of Georgia in 2008.

Finally, if parties respond to election observation by engaging voters more directly we should observe a negative turnout effect earlier in the day (when ballots are stuffed in non-observed locations), and a positive effect in the evening (when hitherto reluctant voters are brought out). As it happens, this is exactly what we see in the parliamentary elections in Georgia 2008.

Continue Reading

Post-Election Report: Romanian’s Presidential Impeachment Referendum, and a Request for Help in Identifying Potential Fraud

The following is a guest post from Princeton political scientist Grigore Pop-Eleches on recent political developments in Romania that culminated in a referendum to confirm the impeachment of the president. The material is particularly timely, as the Romanian Constitutional Court is currently hearing challenges to the results of this referendum on the basis of electoral fraud. Although the post is a bit longer than our normal posts, I strongly urge readers with any experience in using analytical techniques to detect electoral fraud to read the post through to the end, and, if you have any thoughts, to either leave them in the comments section or to contact Pop-Eleches directly.


This post is a bit of an odd mix. First, I will try to report on the ongoing political crisis in Romania, which raises a lot of interesting questions for political scientists who study democratization, European integration and political institutions. Secondly, however, I hope to get some advice from readers of the blog about ways one might be able to establish the existence and magnitude of electoral fraud. This is not simply an academic challenge but might have political implications given that (as I explain below) on August 31st the Romanian Constitutional Court will decide on the validation of the referendum and, therefore, statistical evidence of turnout fraud uncovered before that date could influence the outcome. (Needless to say, this also raises the stakes of being wrong…)

Over the past five months, Romania has undergone what has arguably been the most serious political crisis since the miners’ strikes of the early 1990s. Whether or not one wants to go as far as calling it a constitutional coup d’etat, the government of PM Victor Ponta has used its newly established parliamentary majority to launch a massive attack on the political institutions that were not yet under its control, particularly the Presidency and the Constitutional Court. While the details and democratic implications of these moves have been discussed widely in the international press – for a good overview see Kim Scheppele’s posts on Paul Krugman’s NYT blog (here and here) – in this blog post I want to focus on one of the crucial elements of this political strategy: the referendum to confirm the impeachment of President Traian Basescu, which was voted by a wide parliamentary majority on July 6, 2012, and took place on July 29.

In order to validate Basescu’s impeachment, the referendum needed to fulfill two conditions: a turnout of 50%+1 of registered voters and a majority of yes-votes among voters to the referendum question: “Do you agree with the dismissal of the President of Romania Mr Traian Băsescu?” After initially trying to get rid of the 50% turnout requirement (despite a ruling by the Constitutional Court that such a change would be unconstitutional), the Ponta government caved in to EU pressure and stuck with the 50% turnout rule. Given that public opinion polls suggested that roughly two thirds of likely voters would support the dismissal, it quickly became clear that the crucial question would be whether the government would succeed in getting 50% of voters to show up on Election Day. Consequently, Basescu reversed his initial plea for voters to turn out and vote no, and instead endorsed the rational strategy of encouraging his supporters to stay home in order to deprive the referendum of its 50% quorum. In the end, on August 1 the Romanian Electoral Commission announced a turnout of 46.24% with 87.6% of those voting answering “yes”.

Given that this turnout was clearly short of the 50% threshold, most observers expected the Constitutional Court decision on whether the referendum should be validated to be a mere formality. However, the Ponta government made a surprise move in that it decided to contest the results of the referendum before the Constitutional Court, which decided to postpone the final decision about the referendum validation until September 12 (though it was later moved up to August 31).  The Romanian government motivated its injunction by claiming that the electoral lists overstate the number of eligible voters, because they do not accurately reflect the number of deaths and the numbers of Romanians living abroad whose citizenship has lapsed (Romanians have the right to vote abroad as long they have not given up citizenship and their papers have not expired). While it is probably true that Romanian electoral rolls are in need of updating, this decision is bizarre for at least two reasons: First, the government is trying to change the denominator to the turnout calculation post facto, which to the best of my knowledge is unprecedented (at least in democratic countries) and is particularly odd because the government had organized the referendum and issued self-congratulatory statements about how well it had been organized. Second, the government is trying to use the data from the 2011 census to contest the electoral lists, even though one of the two main parties in the governing coalition, the ex-communist Social Democratic Party (PSD), had previously tried to boycott the census and has since questioned its accuracy. The government even went as far as to try to conduct a new “mini-census” by instructing local mayors to get a sense of the number of people living in their towns, but seems to have backed away from this effort after a the Constitutional Court made it clear that it had not asked for such a measure and after the Interior Minister resigned citing undue pressures on his office. However, it is unclear what the government will try next, and the Constitutional Court has complained for the second time to the Venice Commission  about the political pressures bent on undermining its independence.

But while much of the focus in recent days has been on the denominator of the turnout statistic, I think that there are good reasons to focus at least some attention on the numerator: the votes that were purportedly cast in the referendum. While elections in Romania have generally improved over the course of the last two decades, electoral fraud at the local level still occurs fairly regularly. (After the 2009 presidential elections, which President Basescu narrowly won, Ponta famously said that “their” electoral fraud system had worked better.)

There are three main ways that electoral fraud can happen in the Romanian context. First, there could be outright ballot-stuffing at the polling station, and while a number of incidents of this kind were reported in the Romanian press, the presence of opposition observers at most polling stations should have limited the magnitude of this particular method.  Second, since Romania has no centralized system to detect double voting, it is possible for citizens to engage in multiple voting by voting at different polling stations on Election Day. Such “electoral tourism” has traditionally been organized by political parties of all stripes (who will bus supporters around), but this time around its utility was limited to opponents of Mr. Basescu, i.e. largely the governing USL coalition and its smaller allies. Third, local authorities intent on falsifying the ballot could use the mobile urn method to boost turnout. The mobile urn, which is intended for voters who cannot physically get to the polling station on voting day, offers an opportunity to avoid the watchful eye of polling station observers.

Traditionally – up to the local elections that took place in June 2012 – it would have been possible to detect evidence for both of the latter two options through the data reported by the Central Electoral Commission, which reported for each polling station the number of voters on supplementary voting lists (i.e. who were not on permanent voting lists at that polling station) and the number of votes cast through mobile urn. However, just prior to the referendum, the Romanian government passed ordinance #41 that modified the law governing the reporting requirements of polling stations and eliminated the reporting requirement for the supplementary lists and mobile urns! Not surprisingly, the opposition immediately decried the change as an invitation to electoral fraud.

Given this situation, the question is how one would go about establishing whether electoral fraud took place and what its likely magnitude was. In addition to the intentional scarcity of data about the composition of the local vote, the task is complicated by the fact that many of the traditional diagnostics used in traditional elections don’t easily translate to the referendum situation. Thus, the correlation between turnout and the vote share of the winner, which might catch ballot-stuffing efforts of dominant parties such as United Russia, is unlikely to work in the context of a referendum, where the rational strategy of would-be cheaters could well be to add a mix of yes and no ballots in order to drive up the turnout (since the share of the yes-vote was not the main concern).

Another option would be to compare the polling-station reported turnout levels to the normal distribution, which is generally assumed to hold in free and fair elections, may not hold in the context of this referendum, where the dominant strategy of part of the electorate would be to avoid voting altogether. This issue is complicated by the fact that in some places – including in Sibiu, from where I am writing this post – voters received fake mailings and text messages, purportedly written on behalf of the pro-Basescu camp, in which they were encouraged to go vote against the suspension. Thus, the national-level turnout figures are not normally distributed but they do not display the sorts of abnormalities that were seen as evidence of fraud in the Russian case (see Scott Gehlbach’s post on Monkeycage).

However, things get more interesting when we look at the subnational level. As can be seen in this map of turnout by county, turnout was well above average in the southern part of the country and low in Transylvania, and particularly in the counties with high Hungarian population shares.

But while these overall patterns are not surprising, because the South has traditionally been more supportive of Mr. Ponta’s ex-communist PSD party and voted in lower numbers for Mr. Basescu in the 2009 presidential elections, what is striking is the very high turnout in three of the southern counties: Olt (73.9%) Teleorman (70.2%) and Mehedinti (69.2%), where turnout was between 15-20% higher than even in the neighboring Southern counties.  Things get even stranger when we compare the distribution of turnout at the polling station level for the different counties. Thus, whereas, in other Southern counties, such as Arges and Dolj, turnout frequencies resembled the national pattern, in the three Southern outliers the distributions peaked at much higher levels and then dropped off quite sharply. Furthermore, whereas in Arges and Dolj only about 5% of polling stations reporting turnout above 85%, in Olt over 40% of polling stations were above 85% and over 15% of polling stations with at least fifty registered voters reported turnout above 100%!

The reason why such turnout figures are even mathematically possible, is that Romania’s electoral law allows voters in national elections who are out of town on voting day to vote at any polling station (in Romania or abroad) by just presenting a valid national ID or passport. As a result, in some of the polling stations that were set up in tourist areas – and not surprisingly the Romanian government did its best to set up as many as possible, including in restaurants and hotels on the Black Sea Coast – it was not uncommon to observe turnout rates of 200-300% (including 10 polling stations above 1000% turnout). Such tourist-targeting voting stations may of course be vulnerable to multiple votes and while the total number of such stations was limited, polling stations with turnout above 200% accounted for about 135,000 votes or about .75% of registered voters. The even more important concern are the roughly 366,000 voters (2% of registered voters) who voted in the 657 polling stations reporting between 90-200% turnout, especially given that many of these are in small villages that do not represent obvious tourist destinations (except perhaps of the electoral variety). Remarkably, more than half (56%) of the votes in this latter category came from the three Southern counties – Olt, Teleorman, and Mehedinti – even though the three counties accounted for only 5.3% of registered voters. However, this obviously does not mean that fraud occurred in all of these high turnout polling stations or that it was limited to them (see e.g. the 50% “bump” in Mehedinti, which resembles patterns in Russia).

As I mentioned at the outset, I would appreciate any suggestions about strategies to go about this – especially if it could lead to some results before the Constitutional Court’s decision on August 31st.

Continue Reading

How many votes would the PRI have had to purchase in the 2012 Mexican Election? A Response to Simpser

Tolga Sinmazdemir, a Post-Doctoral Research Associate at Washington University in St. Louis in the Department of Political Science, sends along the following response to Alberto Simpser’s now very well read Monkey Cage guest post on the 2012 Mexican elections. In the post, Simpser attempted to estimated the cost of purchasing enough votes to swing the election to the PRI’s presidential candidate, and then speculated about the feasibility of actually purchasing that many votes. The post attracted many comments, a good number of which raised the question of whether the PRI could have “purchased” the elections in ways that move beyond direct vote-buying. In this response, however, Sinmazdemir addresses Simpser’s argument largely on its own terms by simply questioning whether he got the actual number of votes that would need to be purchased correct, a point that does not seem to have been raised in the comments in reaction to the original post:

Having read Alberto Simpser’s post titled “Could the PRI have bought its electoral result in the 2012 Mexican election? Probably Not.”, I was a little puzzled by the number (3.2 million) he uses for the minimum number of votes that PRI needed to buy to get the election results they did. I am an expert neither on Mexican politics nor on elections so I wanted to join the conversation on this subject. It seems to me that this number lies somewhere between 1.6 and 6.4 million. Let me explain why.

As Alberto Simpser also points out, the minimum number of votes PRI needed to buy depends on two things: 1. How voters would have voted if there were no bribes. 2. Whom PRI targeted with their bribes (PRD voters or those who would have stayed home and those who would have voted for a third party)

Now we know that PRI received 3.2 million more votes than PRD.

Assuming for now that PRI targeted exclusively either PRD voters, or those who would have stayed home and those who would have voted for a third party, and that bribes are 100 % effective, the 3.2 million vote margin could have happened in four different ways depending on what we assume about these two things:


1. Without any bribes, PRD would have received 3.2 million more votes than PRI.
2. PRI targets 3.2 million PRD voters who change their votes to PRI.


1. Without any bribes, PRD would have received 3.2 million more votes than PRI.
2. PRI targets 6.4 million voters who would have stayed home or would have voted for a third party, so they vote for PRI because of the bribe.


1. Without any bribes, PRD would have received the same number of votes as the PRI.
2. PRI targets 3.2 million voters who would have stayed home or would have voted for a third party, and they vote for PRI because of the bribe.


1. Without any bribes, PRD would have received the same number of votes as the PRI.
2. PRI targets 1.6 million PRD voters who change their votes to PRI.

So, depending on these two factors, the minimum number of voters needed to be bribed fluctuates between 1.6 and 6.4 million. And as scenarios three and four illustrate, if we assume that clean elections were close, and if PRI targeted both groups (those who would have voted for PRD and those who would have stayed home or voted for a third party) and succeeded, then the number would be somewhere between 1.6 and 3.2 million. If they targeted only those who would have stayed home or voted for a third party, the number would be 3.2 million. As I understand from Alberto Simpser’s response to one of the comments on the figure of 16 % he uses for vote-buying effectiveness, PRI has likely targeted both groups although some scholars object to the idea of parties bribing partisans of their rivals. But the bottom line is, as far as I can see, the right number for voters bribed that should be used for calculating the lowest possible cost of buying the elections should be 1.6 million, not 3.2 million.

This obviously changes our estimate of the total cost of buying the election. With the lower bound of 1.6 million rather than 3.2 million, and using the figures that Simpser uses in the post for the cost of buying votes (700 pesos) and its effectiveness (16 %), the total estimated cost of stealing the election becomes half a billion dollars. At 100 % effectiveness, the estimated cost becomes 1,120 million pesos which is less than the total campaign funding of PRI-PVEM coalition as reported by Simpser (1,390 million pesos).

Therefore, this analysis makes me reach the conclusion that if the result of a clean election was close between PRI and PRD, and if PRI focused heavily on potential PRD voters effectively (to me as a total outsider, this seems consistent with the image of the PRI’s fraud “machinery” used by some commentators), then the total cost of buying the elections was less than the total campaign funding of PRI-PVEM, and PRI buying the election results in 2012 becomes less improbable.

Continue Reading

Post-Election Report II: Revisiting Fraud and the 2012 Mexican Presidential Election

In view of the large amount of attention yesterday’s post on the potential for vote buying in the 2012 Mexican presidential election has received, we are very pleased to have a second follow up post-election report on this election from  Marco A. Morales, a Ph.D candidate in the Department of Politics at New York University, that addresses more directly the question of the potential for and consequences of electoral fraud in this election. In addition to his graduate work in political science, Morales has also served as a public official under both PRI and PAN administrations, most recently – under the current PAN administration – as Press Counsellor and Spokesman for the Mission of Mexico to the United Nations, and Director General for Political Analysis at the Office of the Mexican Presidency.


The official tally for the Mexican presidential election has now been released. After all votes have been counted (and recounted whenever there were discrepancies in the tallies) the results advanced by the PREP and the expedite count remain virtually unchanged: PRI’s Peña Nieto has 38.21% of the vote, PRD’s López Obrador 31.59%, and Vázquez Mota 25.41% of the vote. This means that a difference of over 3.3 million votes between the front-runner and the second place remains virtually unchanged. (Or in other words, that there was human error on the vote counts, but no systematic error).

Yet, for outside observers, this could seem like a convoluted process taking an exceptional amount of time. And that would be a correct observation. Nevertheless, that is what the electoral law dictates. So what exactly is going on, and why?

A bit of Mexican electoral history could provide some necessary context. Decades of one-party dominance supported by tampering with the vote created a plethora of allegories for common electoral practices that seem drawn from Magical Realism literature: the “crazy mouse” (ratón loco), where voters run around “like crazy” trying to find the booth where they should vote until they quit tired from never finding the right place; the “carrousel”, whereby groups of voters cast a ballot in different voting booths along a pre-set route; the “tamale” (tamal), whereby a voter stacks the booth with votes that he had previously brought with him; the “pregnant ballot boxes” (urnas embarazadas); whereby the ballot boxes are brought to the polling station already stuffed with votes for one candidates; the “shoe ballot boxes” (casillas zapato), ballot boxes where all votes are cast for one candidate; the “shaved-off voters” (votantes rasurados), whereby partisan of the opposing parties are eliminated from the list of voters; and all these allegories are perpetrated by “racoons” (mapaches), electoral “alchemists” who tamper with the votes in order to get the desired outcomes.

With these precedents, it is no surprise that recent Mexican electoral history is characterized by attempts to foster trust in electoral outcomes, primarily by coping with three main forms of electoral fraud: controlling who votes, tampering with the vote count, and vote buying.

The first type of fraud – on the voter list – has been addressed through a voter registration list that is administered by an autonomous electoral authority, and audited by all political parties. Similarly, voters on the registration list receive a voter ID card that is a requisite to vote, and get an indelible (acid) mark on their right thumb after voting.

►    On this election, about 1.2 million randomly selected citizens counted the votes of more than 50 million voters.

The second type of fraud – in the counting of votes – has been addressed with multiple layers of scrutiny. Polling stations are manned by randomly selected citizens, whose performance can be audited in situ by party representatives and citizen observers, especially when they produce vote tallies. Days later, ballots can be recounted on every district whenever an objection to the tally is presented, when the amount of ballots does not correspond to the tally, or when the difference between the first and second place on the district is less than 1% of the votes.

►    On this election, nearly 2 million party representatives and 32 thousand citizens observed vote counts, and over 132 million ballots were recounted to correct vote tallies.

Photo: José Merino

To further enhance the credibility of the results, upon closing the polls on Sunday, an electronic system – the PREP – begins adding up the tallies from polling booths in the country, providing an unofficial count of votes when it closes 24 hours later with over 90 percent of polling places accounted for. Similarly, a statistical exercise – an “expedite count” (conteo rápido) – that uses a large “representative” sample of polling places (7,000 in 2012) provides a fairly accurate estimate with a rather small margin of error, of the outcome of the election before midnight on Election Day.

►    On this election, nearly 141 thousand polling stations were accounted for by the PREP. The expedite count sampled over 7,000 polling stations rendering results with a +/-0.5 margin of error. Both results coincided with one another, and with the official results of the election.

The third one – vote buying – has been addressed by making it a criminal offense that is to be investigated by a Special Prosecutor. But as many have pointed out, the challenge on this specific area is considerable as evidence should show beyond doubt that the “incentive” did change the original vote intent and that these “sold” votes changed the outcome of the election.

In other words, there are two different matters regarding vote buying: whether someone “bought” votes, and whether vote-buying was extensive enough to alter the outcome of the election. The first one – buying or coercing votes – is punishable under Mexican criminal law. The second effect – whether vote buying altered the outcome of the election – is determined by the Electoral Court. What must be clear is that the existence of the first does not automatically translate into the second. The devil, as it stands, is in the detail.

To summarize, there were no complaints of the first type of fraud on this election. And while there were plenty of complaints of the second type of fraud, they were for the most part addressed and dispelled in the process that just ended. What lies ahead are complaints of the third type of fraud: massive vote buying.

The leftist candidate, López Obrador, announced this morning that he will seek to annul the election, arguing that there was massive vote buying – over 5 million votes – by the PRI candidate Peña Nieto.

It is yet unclear what will be the decision of the Electoral Court, even if the evidence of vote buying were compelling. The 2007 Electoral Reform rendered most previous rulings on this matter invalid, so we have little to rely on for analysis. The Court could rule that there was indeed vote buying, but lacking evidence of its magnitude or effects, rule that it cannot determine whether it changed the outcome of the election. Or it could, in a historic ruling, determine the opposite.

The important thing is that the electoral process in Mexico is not formally over until the Electoral Court (a special court in the Judiciary branch that deals exclusively with electoral matters) declares a winner. And that will not happen until the last complaint is resolved … which could take months – as was the case in 2006 – but must happen before September 6th.

Continue Reading

Could the PRI have bought its electoral result in the 2012 Mexican election? Probably Not.

The following is a guest post from University of Chicago political scientist Alberto Simpser. The first chapter of his forthcoming book, Why Governments and Parties Manipulate Elections: Theory, Practice, and Implications, can be found here.

The PRI won the presidency in the July 1st Mexican election, but the dust has not settled yet (see Marco A. Morales’s excellent post on this). Given the safeguards in place and the electoral commission’s (IFE) stewardship, election fraud almost certainly did not occur on a large scale. Vote buying, however, apparently did, as numerous media reports and YouTube videos suggest. A serious allegation is circulating: that the PRI in fact would have lost a clean election to the runner up, the PRD’s Andrés Manuel López Obrador (AMLO), and that its electoral result stemmed from massive vote buying. I have no evidence about whether this allegation is in fact true. Rather, in this note I explore the plausibility of the accusation by estimating how much money it would have cost the PRI to obtain the result it did. The credibility of the ensuing estimates, of course, depends on the solidity of the underlying assumptions. By clarifying my assumptions, I hope to spark informed debate on the question.

I use three pieces of information: the number of votes that the PRI would have needed to buy in order to obtain the result it did, the effectiveness of vote buying (i.e. the rate at which vote buying attempts translate into actual votes), and the cost per vote. I leave out other potentially important factors, such as payments to vote brokers and embezzlement of funds by these brokers.

The PRI obtained about 3.2 million votes more than the PRD (equivalent to about 6.5% of cast votes). Because the PRD could also have bought votes, as some have alleged it did, this figure is a lower bound on the amount of votes that the PRI would have needed to buy in order to upend the election outcome (assuming, for simplicity, that no other forms of vote manipulation were used).

The effectiveness of vote buying depends on a variety of factors. For one thing, ballot secrecy means that many citizens can take the money and vote their conscience. The PRI is alleged to have violated ballot secrecy through the use of cell phone pictures, as well as children who allegedly accompanied voters into the booth; but it is unlikely to have been able to verify every single vote it bought. Effectiveness also depends on the counterfactual scenario: whether and how the bribe recipient would have voted in the absence of a bribe. Even with perfect observability and enforcement, it is possible to waste a bribe on a PRI loyalist who planned to vote anyway. I do not have data on effectiveness for Mexico, so I use an estimate from the literature, based on a survey of Argentinean citizens in 2000-2001 (Brusco, Nazareno and Stokes 2004). Only 16% of surveyed citizens who received a bribe said that it influenced their vote. There exist additional factors that could impact effectiveness, such as embezzlement by brokers. I ignore those in my calculations here (Wang and Kurzman 2007 estimate money “leakage” in a 1993 Taiwan election at upwards of 45%).

Estimates of the cost per vote bought range from 100 to 1,800 pesos. A YouTube video suggests that 100 was low enough to insult the recipient, while 1,800 (a figure implied in recent allegations by AMLO about the amount of money spent and the number of votes bought by the PRI in the state of Mexico) was substantially more than at least some brokers appear to have received for their services. I use 700 pesos, reported on some internet videos, as my best guess (as a point of comparison, a study of the 2000 Mexican election suggests a price per vote of 250 to 500 pesos).

Based on these assumptions, the estimated cost to the PRI of stealing the election through vote buying is 3.2 million actual votes * 700 pesos per vote bought / 16% actual votes per vote bought = 14,000 million pesos approximately (roughly equivalent to USD $1 billion). This is an underestimate insofar as it leaves out brokerage costs and assumes that the PRD bought zero votes. It is an overestimate insofar as vote-buying effectiveness was greater than I have assumed, or if the price per vote figure is too high. At 100% effectiveness, the estimate is 3.2 * 700 = 2,240 million pesos.

What do costs of this magnitude imply? As a point of comparison, the total campaign funding of the PRI-PVEM coalition was about 1,390 million pesos (according to Alianza Cívica’s July 3 press bulletin), far less than even the smaller figure estimated here. This suggests that, if the PRI bought its electoral result, it must have done so with unaccounted-for funds. It is not inconceivable that the PRI should have access to such extra-official sums (the former PRI governor of Coahuila, for example, is involved in a debt and fraud scandal involving thousands of millions of pesos). Nevertheless, the magnitude of the requisite sums suggests that these should be difficult to conceal from authorities, should these wish to investigate the matter.

This analysis suggests a dilemma for the parties accusing the PRI of vote buying: the graver the accusation, the more scandalous, but also the less feasible it would appear to be for the PRI to have pulled it off. The runner-up AMLO today said that 5 million votes were stolen (presumably by the PRI). As a matter of strategy, he might have been better off leveling the more modest accusation that the PRI bought just enough votes to overturn a PRD victory. Doing so, however, would have amounted to an admission that the result of a clean election – which the PRD claims to have won – was close. I emphasize that I have no information about whether the PRI, the PRD, or other parties did in fact buy votes, with what money, and on what scale.

As an aside, from a theoretical perspective, it is interesting to note that, based on the internet videos, some voters appear to have been offended when discovering that they were offered too low a bribe in exchange for their vote. This is consistent with the idea that voter compliance in vote buying transactions might stem, at least in part, from normative or expressive behavior. Also, it suggests that a marginal change in the size of the individual bribe could have a non-linear effect on the effectiveness of vote buying efforts: a zero bribe could obtain more votes than a low, offensive one, which in turn could obtain fewer votes than a higher, non-offensive bribe.

Continue Reading

Implications of Finding that Webcams Shift Locus of Electoral Fraud

This previous Monkey cage guest post reported on new research from Fredrik Sjoberg suggesting that when Azerbaijani authorities installed webcams in some precincts in a recent election, it likely reduced overt forms of electoral fraud such as ballot stuffing while simultaneously increasing more subtle fraud such as falsification of precinct level results.

This result turns out to overlap very well with a speculation I made here at The Monkey Cage regarding Putin’s incentives for installing webcams in all of Russia’s polling places for the 2012 Russian presidential election. At the time I wrote:

An alternative explanation, however, might be that the Kremlin was seeking to avoid the mechanism by which fraud was revealed following the parliamentary elections, that is the use of individual of cell phones to capture visible fraud in polling places by polling workers who believed they were not being observed. If we assume that the motivation for local level officials to manipulate vote totals (e.g., to win the favor of the Kremlin) had not changed, then the webcams would provide a very powerful incentive for local officials to find other ways of manipulating results than the blatant forms of ballot stuffing that appeared online following the December parliamentary election.

This appears to be exactly what happened in Azerbaijan, which in turn raises the very interesting question of whether the Russians knew what had happened in Azerbaijan when deciding to install the webcams in Russia.

It also raises a very tricky question for anyone advocating for free and fair elections in countries with less than stellar records in this regard. Should webcams in polling stations be embraced as a technology that at the very least decreases one form of electoral fraud? Or perhaps should they be a cause for concern as a technology that is likely to replace a more easily observable (and easier to publicize) form of fraud—ballot stuffing—with one that is more subtle and less observable: the manipulation of precinct level results. Random digit tests are great for academics, but they don’t really compete with YouTube videos of ballot stuffing for convincing citizens of a country that their leaders have cheated. If we want to take this one step further, then we could argue for a variety of reasons (see this post by NYU political scientist Andrew Little here at The Monkey Cage for example) that by making local agents engage in a type of fraud that is less likely to be publicly discovered, webcams could perhaps make leaders more likely to engage in fraud than otherwise.

Continue Reading

Webcams and Election Monitoring: Shifting, not Stopping, Fraud

This is a guest-post from Fredrik M Sjoberg, currently a Postdoctoral Scholar at NYU. The paper on which this post is based is here.

In recent years, non-democratic rulers have surprisingly begun to embrace fraud-reducing technologies, like web cameras or transparent ballot boxes. Most recently Russia had webcams installed across the vast territory (see this earlier post at The Monkey Cage). Ukraine and Albania are also supposedly considering it. Little is known about the effects it has on reducing fraud. Yes, earlier studies have shown that turnout is reduced in the presence of webcams, but that the ruling party vote share is not necessarily affected (e.g., see here). However, relying solely on proxy measures like turnout and vote shares is problematic since both clearly depend on many different things, some of which might not be related to fraud at all. In any case we would need to account for historical turnout and vote share.

With the help of new fraud identification techniques, I argue that the installation of web cameras in polling stations changes how fraud is conducted. First, we need to recognize that election fraud comes in different forms. Ballot box stuffing, multiple voting, and fabrication of vote are not necessarily affected equally by a fraud-reducing technology like webcams.

Micro-level evidence from Azerbaijan, the first country to install web cameras at the precinct level, suggests that ballot box stuffing is reduced, as measures by turnout, but that authorities instead focus on fabricating the vote totals. Miscounting of the ballots can be detected by digit-based tests where the distribution of the last digit is compared to a uniform distribution (the green line in the graph). As Beber & Scacco show in a recent article, absent outright fabrication of vote totals we expect a uniform distribution on the last digit (Political Analysis, 2012, What the Numbers Say: A Digit-Based Test for Election Fraud). In Azerbaijan the distribution on the last digit in webcam and non-webcam precincts is given by the graph:

  • Note: the data is here matched (CEM) based on size, district, historical turnout and competitiveness to account for biases in how the cameras were allocated.

In precincts with no webcam there is no significant deviation, as indicated by a p value of .6088 (chi square goodness of fit test). In the monitored group, on the other hand, there is a deviation from the uniform, significant at the 5-percent level. This suggests that the officially reported vote totals were fabricated in precincts with a webcam.

By replacing one form of fraud with another incumbents are able to prevent vote share losses, while contributing a veneer of legitimacy by self-initiating anti-fraud measures. It therefore seems like a win-win for the autocrat (and for the company that managed to sell 100,000 webcams to Putin)

Continue Reading