While the negative consequences of corruption are widely acknowledged, citizens often dismiss or overlook fraudulent or suspicious cases of public fund abuse. In a meta-review, Incerti (2020) finds that although respondents express strong disapproval of corrupt candidates in survey experiments, corrupt candidates are, on average, penalized by zero percentage points in field experiments. This raises a critical question: how can survey experiments be made more representative of real-world voting behavior? This paper takes a first step toward addressing this gap by examining individuals' sensitivity to the costs of voting for a corrupt versus a clean in-party candidate. I hypothesize that while respondents generally dislike corrupt candidates, they may tolerate some level of corruption if switching to an honest candidate incurs significant costs. I categorize these costs into four types: (1) cognitive costs of updating beliefs about a known candidate; (2) personal and ideological costsassociated with increasing the out-party’s chances of winning; (3) social desirability costs tied to revealing support for corrupt politicians; and (4) conformity costs related to deviating from widely accepted norms.To quantify individuals' sensitivity to these costs, I employ an incentive experiment that modifies the multi-arm design used by Janas et al. (2024), adapting it into a single-vignette conjoint design. The experiment randomly varies four factors—each corresponding to one of the four types of costs. By quantifying the real-world costs associated with punishing corruption, this project addresses a critical gap in the literature and provides new insights into the mechanisms driving voter tolerance for corrupt politicians.