This paper proposes coarse wage-setting as an explanation for why wages tend to cluster at round numbers. Using data on the contracted salaries of 280 million new hires from Brazil, I first establish that the distribution of salaries presents stark bunching at round numbers. In the data, more than a third of all workers are hired at a round-numbered salary. This finding is inconsistent with the wage distribution predicted by canonical wage-formation models. Reduced-form findings reveal that firms that tend to hire workers at round-numbered salaries are less sophisticated in observable characteristics (e.g., they are smaller and have less hiring experience) and, conditional on firm sophistication, experience worse outcomes (e.g., they have lower growth and survival rates). These firms also engage in nonstandard behavior in other settings (e.g., they are more likely to pay round-numbered wage increases). These facts are consistent with the bunching being driven by firm coarse wage-setting and inconsistent with firms optimally responding to a worker behavioral bias. Motivated by the reduced-form findings, I develop a wage-posting model in which optimization costs lead to the adoption of coarse wage-setting. The model delivers three predictions that I test and for which I find support using two research designs. This provides additional evidence for the coarse wage-setting hypothesis. Finally, I quantify some consequences of coarse wage-setting for relevant economic outcomes. Coarse wagesetting generates within-firm wage compression, increases nominal wage stickiness, and interacts with policies that affect the wage distribution, such as changes in the minimum wage.