Other Resources

A Guide to Elections Forensics - Research and Innovation Grants Working Papers Series by USAID (2017)

Message from the Democracy, Human Rights, and Governance Center Acting Director
This publication, together with the Election Forensics Toolkit DRG Center Working Paper and the online Election Forensics Toolkit, were produced by USAID in partnership with the University of Michigan and the Institute of International Education as part of the Research and Innovation Grants Working Papers Series. [...]

The research presented in this Guide to Election Forensics, in the more detailed DRG Center Working Paper, and in the online tool explains the role election forensics can play in verifying the integrity of election data, demonstrates several statistical tests used in election forensics to verify election data, and illustrates how those tests identify anomalous patterns in the data that can indicate intentional manipulation of results.

Excerpts:
Election forensics is an emerging field in which scholars use a diverse set of statistical tools—including techniques similar to those developed to detect financial fraud—to analyze numerical electoral data and detect where patterns deviate from those that should occur naturally, following demonstrated mathematical principles. Numbers that humans have manipulated present patterns that are unlikely to occur if produced by a natural process—such as free and fair elections or normal commercial transactions. These deviations suggest either that the numbers were intentionally altered or that other factors—such as a range of normal strategic voting practices—influenced the electoral results. The greater the number of statistical tests that identify patterns that deviate from what is expected to naturally occur, the more likely that the deviation results from fraud rather than legal strategic voting. 

Compared to existing methods of electoral transparency such as in-person monitoring and parallel vote tabulations, election forensics has three key advantages: 

1. It relies on objective data, such as reported election results disaggregated to the level of electoral constituencies, precincts, and/or polling stations. 
2. It allows for systematic analysis of reported votes from all contests in all localities. 
3. It products estimates of fraud that include statistical statements about the confidence of conclusions.

However, election forensics also has three disadvantages: 

1. It does not produce definitive proof of fraud, only of statistical anomalies. Finding proof for or explanation of the anomalies could come from in-person electoral monitoring or other social science research on the country or location of interest. 
2. It requires advanced statistical knowledge and substantial computing power. The Election Forensics Toolkit presents one potential solution to this challenge, since academic experts perform the statistical work in-house and provide a report that is accessible and intelligible to practitioners. 
3. It works best with detailed election results—ideally comprehensive polling-place data on turnout, valid ballots, and vote counts for all parties and candidates. The tools can work with less detailed data, but election forensics are of limited use if election results are provided only as a summary at the national level, or not reported at all.

Fraud or Fairytales: Russia and Ukraine's Electoral Experience by Mikhail Myagkov, Peter C. Ordeshook & Dimitry Shakin, Post-Soviet Affairs, 21:2 (2013)

"Of course, 1996 might seem like ancient history, so for a more contemporary anomaly, consider Figure 1, which, for each rayon, plots Putin’s vote in 2004 against his vote in 2000. A normal pattern is a high correlation with data distributed around a 45° line or, if a candidate’s popularity increases (as seems to be the case with Putin), a positively sloping line above the 45° diagonal. Again, this expectation is not met in Russia.  Although the data match expectations for the “dense cloud” of rayons, there is essentially no relationship between his vote in 2004 and 2000 if his vote exceeded 80 percent in 2004. That is, knowing that Putin received 85 percent versus 95 percent of the vote in 2004 in a rayon tells us only that he got more than 40 percent in 2000. At the same time, if we are told that Putin did poorly in a rayon in 2000 (less than 20 percent), we cannot predict anything about his performance in 2004. We appreciate the fact that the absence of meaningful parties diminishes an electorate’s stability, but it is almost as if, among a significant number of rayons, the 2000 and 2004 elections took place in different countries. [...]

However, suppose a subset of rayons are untainted by fraud, but, as is generally assumed to be the case for some republics, that ballot stuffing or allowing people to vote more than once increases turnout in the remaining regions (and we note by way of example that the six regions reporting turnout in excess of 73 percent in 1999 include Dagestan (ranked second), Mordoviya (third), Tatarstan (fourth), and Bashkortostan (sixth)). This second subset might also yield a normal distribution of turnout, but with a mean shifted up in accordance with the magnitude of the fraud committed. Thus, if we fail to separate the two subsets and simply add their distributions, we should observe a multimodal density or one with an overly “fat” or extended tail. And, as Figures 3a and 3b show, that is precisely what we see in Russia’s ethnic republics in 1996 through 2003."

Statistical Detection of Systematic Election Irregularities: Three Essays on Supervised and Unsupervised Machine Learning Approaches and the Attitudinal Consequences of Exposing Cheating | Lion Behrens | PhD Dissertation, University of Mannheim (2023)

Excerpt:

"The method that I develop builds on the fact that in many countries, elections are not held as singular events. Rather, concurrent electoral contests which are administered side-by-side often take place simultaneously. I show that undervoting irregularities, which emerge if the same polling station documents different turnout levels across different electoral events, can be exploited for the detection of systematic irregularities if the extent of undervoting is related to the winner’s vote share. I present a semi-parametric simulation model to estimate the share of polling stations with vi undervoting irregularities at which vote shares were tainted. I apply this approach to a novel data set of recently contested Ecuadorian elections which report large extents of undervoting and simulated data for which the degree of fraud is known. I find that the proposed method reliably reverse-engineers true shares of fraud in synthetic data and that the empirical patterns which are inherent to Ecuadorian voting returns are well explained by systematic manipulation. [...]

Finally, I assess the attitudinal consequences of confronting individuals with credible information on systematic electoral manipulation (Chapter 4). In the past, scholars have shown that consciousness of election fraud lets individuals withdraw support from candidates, institutions and governments that are supposedly involved in manipulation. Together with my co-author Viktoriia Semenova, I argue that election fraud information will let individuals extrapolate legitimacy loss even to political institutions that are unrelated to electoral events and lead to decays of trust in the political system as a whole. Second, we argue that these spillovers are crucially shaped by the reactions of other political actors, as within-system corrections like court punishments of alleged fraud perpetrators can mitigate decays in diffuse support."

 Elections Without Democracy: The Menu of Manipulation | by Andreas Schedler  Journal of Democracy | Johns Hopkins University Press (2002)

"The idea of democracy has become so closely identified with elections that we are in danger of forgetting that the modern history of representative elections is a tale of authoritarian manipulations as much as it is a saga of democratic triumphs. Historically, in other words, elections have been an instrument of authoritarian control as well as a means of democratic governance.

Since the early days of the "third wave" of global democratization, it has been clear that transitions from authoritarian rule can lead anywhere. Over the past quarter-century, many have led to the establishment of some form of democracy. But many others have not. They have given birth to new forms of authoritarianism that do not fit into our classic categories of one-party, military, or personal dictatorship. They have produced regimes that hold elections and tolerate some pluralism and interparty competition, but at the same time violate minimal democratic norms so severely and systematically that it makes no sense to classify them as democracies, however qualified. These electoral regimes do not represent limited, deficient, or distorted forms of democracy. They are instances of authoritarian rule. The time has come to abandon misleading labels and to take their nondemocratic nature seriously.  [...]

Their dream is to reap the fruits of electoral legitimacy without running the risks of democratic uncertainty. Balancing between electoral control and electoral credibility, they situate themselves in a nebulous zone of structural ambivalence. Delimiting the blurry frontiers of electoral authoritarianism cannot help but be a complex and controversial task. " 

Get Out the Vote (or Don’t): Evaluating Manipulation Methods in Russian National Elections from 2000 to 2021 | Chloe Duval  | Bachelor's Thesis, University of Chicago (2022)

"Informational Autocrats: Elections in Authoritarian Regimes 

To examine the reasoning behind and the extent of the Kremlin’s shift from manipulating Putin’s vote share to manipulating turnout, an understanding of the existing literature on authoritarian states like Russia contextualizes why elections are manipulated in the first place. A new form of autocracy emerged in the 21st century led by informational autocrats. Guriev and Treisman (2019) offer a strong case for why autocrats like Putin choose to have elections and legislatures even if the results are falsified. In contrast to early authoritarian leaders like Stalin and Mao, informational autocrats choose to concentrate their power by using democratic institutions instead of terror. The authors argue the key to regimes like Putin’s is manipulating information to “survive by leading citizens to believe – rationally but incorrectly – that they are competent and public spirited” (Guriev and Treisman 2019, 101). Of course, applying the characteristics of informational autocracy outlined by Guriev and Treisman to a regime such as Putin’s, which historically has relied on television for its information control, has implications as Russia is an informational autocracy with an arguably shrinking information gap between the public and elites.

Informational autocrats often use institutions like elections to maintain legitimacy, consolidate their power, and signal their political strength to the public and bureaucrats alike. Alberto Simpser, in his book Why Governments and Parties Manipulate Elections: Theory, Practice, and Implications (2013), focuses specifically on elections as tools for autocrats to maintain legitimacy and pushes back against the previous academic consensus: that autocrats excessively manipulate them just to win. Putin is a popular politician in Russia, and his approval ratings remain consistently high.8 Since his first term as president in 2000, Putin would be more likely than not to win a free and fair election. Simpser explains why autocrats continue to manipulate elections they could win fairly: “electoral manipulation can potentially yield substantially more than simply winning the election at hand ... [including] to discourage opposition supporters from turning out to vote or to protest; to convince bureaucrats to remain loyal to the government...” (Simpser 2013, 3). 

The public, like the bureaucracy, poses a potential problem for an autocrat like Putin. Andrew Little unveils a second motivation autocrats may have to manipulate elections: to keep the public from staging a revolution. It is beneficial for an autocrat to change election results to distort their signal to the public. When deciding whether to protest, the public evaluates the costs and benefits of doing so after seeing election results. Strong election results may dissuade them, since they suggest the autocrat is more popular than anticipated. Little establishes that “the information provided by elections does not only affect incumbent behavior as emphasized in past work, but also all political actors from elites to citizens” (Little 2011, 273).

Electoral Manipulation and Regime Support: Survey Evidence from Russia by Ora John Reuter and David Szakonyi | Higher School of Economics (2020)

Abstract 

Does electoral fraud stabilize authoritarian rule or undermine it? The answer to this question rests, in part, on how voters evaluate regime candidates who engage in fraud. Using a survey experiment conducted after the 2016 elections in Russia, we find that voters withdraw their support from ruling party candidates who commit electoral fraud. This effect is especially large among strong supporters of the regime. Core regime supporters are more likely to have ex ante beliefs that elections are free and fair. Revealing that fraud has occurred significantly reduces their propensity to support the regime. These findings illustrate that fraud is costly for autocrats not just because it may ignite protest, but also because it can undermine the regime’s core base of electoral support. Because many of its strongest supporters expect free and fair elections, the regime has strong incentives to conceal or otherwise limit its use of electoral fraud.

Excerpt:
The effects of increasing awareness of electoral fraud are largest among core regime supporters. In electoral authoritarian regimes, regime partisans are more likely to believe ex ante that elections are conducted fairly. This can happen for a number of reasons. Regime supporters are more exposed (and possibly susceptible) to regime propaganda, and partisanship biases may inhibit the internalization of rumors about fraud. Alternatively, they may support the regime precisely because they believe it is holding free and fair elections. Given these pre-conceived notions, the regime’s core supporters will be most likely to punish regime incumbents when fraud is revealed to them. By contrast, swing or weakly aligned voters are already skeptical about electoral integrity. Hence, the revelation of fraud will do less to affect their vote choice. Expectations of electoral fraud are already factored in for these voters.

Using data from the same survey, we report several other findings in support of our main arguments. First, the vast majority of Russians express moral disapproval of electoral fraud, regardless of their affinity for the regime in power. Second, a surprisingly large share of Russians believe that elections are held honestly, and more importantly for this study, regime supporters are much more likely to believe that elections are free and fair. Finally, we find that learning about fraud by UR candidates produces a much larger reduction in support among strong regime backers than it does among weakly aligned voters. We conclude that if information on fraud were to become widespread in Russia, the size of Putin’s electoral coalition would diminish significantly. We replicate these findings with a second survey experiment conducted in Russia in May 2018, which also examines how an individual’s likelihood of voting depends on perceptions of fraud. Our findings demonstrate that excessive use of fraud can destabilize autocracy not just because it leads to mass protest, but also because it erodes the regime’s electoral base.

How trustworthy are electronic voting systems in the US? by Elizabeth Clarkson | American Statistical Association | Royal Statistical Society (2015) 

Excerpt:

"When you do your civic duty, and cast your vote for the various candidates and public propositions at an electronic voting machine, how confident are you that the results will be tabulated honestly? 

If you feel less than sanguine about it and do a bit of the research to assuage your doubts, be prepared to feel even less confident afterwards. After years of casual research, the results I found have led me to file a lawsuit requesting access to the records needed to perform an audit myself. 

My statistical analysis shows patterns indicative of vote manipulation in machines. The manipulation is relatively small, compared with the inherent variability of election results, but it is consistent. These results form a pattern that goes across the nation and back a number of election cycles. I’ve downloaded data and verified the results from several states for myself. Furthermore, the manipulation is not limited to a single powerful operator. My assessment is that the data reveals multiple (at least two) agents working independently to successfully alter voting results. [...]

I want to emphasize, as I always try to do, that statistics don’t prove vote fraud. These statistics show that patterns exist in the data that correlate the type of electronic voting system in use with the %R vote changing with the total votes cast.

Such patterns are examples of what we might expect to see if some voting systems were being sabotaged, but that doesn’t mean that no other explanations are possible for these patterns. Voting machine manipulation is, in my opinion, the most likely explanation for these patterns. The most common pattern supports Republican candidates, but Democratic candidates are sometimes the beneficiary. 

The only way to prove vote fraud is through a post-election audit demonstrating significant deviations from the reported totals."