The field of election forensics employs inferential methods to detect anomalies in voting returns that are indicative of systematic irregularities. So far, scholars have almost exclusively treated individual elections as individual isolated events. I argue that in the presence of concurrent electoral contests on election day, systematic vote alterations can be detected from 'undervoting irregularities' that emerge if protagonists of fraud fail to interfere into multiple races to equal extents. Conceptually, I introduce the distinction between balanced and unbalanced fraud approaches in the presence of several simultaneous electoral contests. Methodically, I develop statistical methodology to detect and quantify systematic interference that stems from unbalanced fraud approaches and evaluate its performance using a series of Monte Carlo simulations. Lastly, I apply the proposed methodology to a number of electoral events from recent Latin American elections and shed new light into current trends of electoral integrity across the Americas. These contributions highlight the relevance of contextual information for the practice of election forensics in general and improve our understanding of undervoting irregularities in particular.