Despite its importance, electoral fraud suffers from a relative lack of attention in the academic literature. Among the main causes is the absence of proper data as well as reliable measures of detecting. The existing methods of fraud detection are more qualitative than quantitative, often based on the subjective assessment of electoral transparency and fairness by observers or other participants of the electoral process, and the results they produce may not always be treated as fully reliable.

The few attempts to analyze rigorously electoral data for the presence of fraud have usually required a large amount of data, which handicaps efforts to measure fraud, proxy it, or even detect it with some confidence. It further precludes implementing reliable empirical research, which in turn discourages efforts towards a theoretical study of the nature and consequences of electoral fraud.

This paper proposes a simple statistical method for testing elections for the presence of ballot stuffing using official detailed electoral data. The method is based on the observation that ballot stuffing increases both turnout and the incumbent‘s vote share in precincts where it occurs.

Hence, precincts with a relatively low reported turnout are more likely to be clean. Using the information on relatively clean precincts, it is possible to simulate counterfactual data for ―infected precincts and compare them with the observed data.

The method is first piloted on artificial and artificially fraudulent real data and subsequently applied to test the fairness of the Russian executive election held in 2004, whose transparency and integrity are dubious. Results strongly reject the hypothesis of no ballot stuffing and suggest that at least 7,000,000 ballots in the 2004 Russian elections were stuffed in favor of the incumbent, which however did not alter the winner of the elections.