That does not completely solve the problem. As Shikano and Mack (2011) show using German federal elections and simulations, even accurate results can appear far from Benford's Law, for reasons other than irregularities. Furthermore, irregularities where not caused by fraud can produce numbers that do not follow Benford’s Law. For example, human vote counters whose mathematical skills are weak may favour certain digits (i.e., “round off”) to simplify computations, without intentionally favouring any of the candidates. The second line of criticism targets the assumption that all irregularities—or even all deliberate fraud—will result in violations of Benford’s Law. Unlike human subjects in laboratory experiments, some electoral fraudsters know what features of the numbers will be inspected, and also have access to tools to generate numbers that will pass statistical tests based on Benford’s law.