Integrity of elections relies on fair procedures at different stages of the election process, and fraud can occur in many instances and different forms. This paper provides a general approach for the detection of fraud. While most existing contributions focus on a single instance and form of fraud, we propose a more encompassing approach, testing for several empirical implications of different possible forms of fraud. To illustrate this approach we rely on a case of electoral irregularities in one of the oldest democracies: In a Swiss referendum in 2011, one in twelve municipalities irregularly destroyed the ballots, rendering a recount impossible. We do not know whether this happened due to sloppiness, or to cover possible fraudulent actions. However, one of our statistical tests leads to results, which point to irregularities in some of the municipalities, which lost their ballots: they reported significantly fewer empty ballots than the other municipalities. Relying on several tests leads to the well known multiple comparisons problem. We show two strategies and illustrate strengths and weaknesses of each potential way to deal with multiple tests.