The measurement difficulties arising from relationship‐based business transactions can result in accounting opacity. We test this hypothesis by exploiting a natural experiment. Using a sample of firms that were networked with 45 high‐level Chinese bureaucrats involved in corruption scandals between 1996 and 2007, we examine the patterns in the earnings informativeness of these firms before and after the exogenous break of the networks. We predict that the costs and benefits of business‐politics relationships, which are not measurable by the current accounting systems, diminish the ability of accounting earnings to track a firm's economic performance. In turn, a break in a political relationship due to anti‐corruption enforcement reduces the measurement noise and improves the earnings informativeness. We find that, relative to the matched control firms, there is indeed a significant increase in the earnings informativeness of the networked firms following the public exposure of a scandal. Robustness tests fail to show that the documented improvement in the earnings informativeness is primarily due to systematic changes in the firms’ earnings management behavior or disclosure policies.