Disadvantaged students perform differentially worse when randomly given a financially salient mathematics exam. For students with socioeconomic indicators below the national median, a 10 percentage point increase in the share of monetary themed questions depresses exam performance by 0.026 standard deviations, about 6 percent of their performance gap. Using question-level data, I confirm the role of financial salience by comparing performance on monetary and highly similar non-monetary questions. Leveraging the randomized ordering of questions, I identify an effect on subsequent questions, providing evidence that the attention capture effects of poverty affect policy relevant outcomes outside of experimental settings. (JEL G53, I21, I24, I32, J13, O15)