Problem gambling is becoming a growing concern in the United States because of the proliferation of, and state support for, gambling opportunities. The economic cost along with the physical and mental health problems associated with problem gambling make it necessary to study how problem gambling can be reduced. Our study examines whether financial literacy could be a means to reducing gambling frequency in the United States. We use data from the Preference Parameter Study of Osaka University, Japan, and apply instrumental variable probit regression models. The results show that, generally, financial literacy does not have a relationship with gambling frequency, but the relationship is significant in the states where electronic gambling machines (EGMs) are available. The results imply that gamblers are irrational and fail to assess the risks of gambling as well as the probabilities that maximize expected payoffs. It appears that gamblers’ psychological gain from gambling outweighs the negative expected utility when there is easy access to gambling. Thus, rationality with regard to gambling decisions does not work unless the easy access to EGMs is controlled. Our results further show that males, older people, people with higher household income, and people who have easy access to gambling are likely to be frequent gamblers.