When asking attitudinal questions with dichotomous and mutually exclusive response options, the questions can be presented in one of three ways. A fully balanced question presents both sides of the competing viewpoints. A minimally balanced question presents only one side of the viewpoint but only mentions the competing side very briefly. A third approach is to ask an unbalanced question that only presents one side of the viewpoint while completely ignoring the competing side. Although previous research has compared the fully vs. minimally balanced rating scales, as far as we know, these three types of rating scales have not been tested in a strict experimental setting. Further, previous research has been almost exclusively based on interviewer-assisted telephone surveys, the findings which may not apply to self-administered web surveys. In this study, we report two web survey experiments testing these three types of rating scales among 16 different questions. Overall, the univariate distributions across these three scale balancing types are very similar to one another. Similar patterns are found when breaking down the analysis by respondent’s education level. The results suggest that for self-administered web surveys higher efficiency can be achieved through unbalanced dichotomous rating scale questions.