Objective To elicit preference weights for a subset of EuroQol five-dimensional (EQ-5D) questionnaire health states from a representative sample for the state of Minas Gerais, Brazil, using a time trade-off (TTO) method and to analyze these data so as to estimate social preference weights for the complete set of 243 states. Methods Data came from a valuation study with 3362 literate individuals aged between 18 and 64 years living in urban areas. The present study was based on quota sampling by age and sex. Face-to-face interviews were conducted in participants’ own homes. A total of 99 EQ-5D questionnaire health states were selected, presorted into 26 blocks of six unique health states. Each participant valued one block together with the full health, worst health, and dead states. Each health state was evaluated by more than 100 individuals. TTO data were modeled at both individual and aggregate levels by using ordinary least squares and random effects methods. Results Values estimated by different models yielded very similar results with satisfactory goodness-of-fit statistics: the mean absolute error was around 0.03 and fewer than 25% of the states had a mean absolute error greater than 0.05. Dummies coefficients for each level within the EQ-5D questionnaire dimensions of health displayed an internally consistent ordering, with the mobility dimension demonstrating the largest value decrement. The values of mean observed transformed TTO values range from 0.869 to−0.235. Conclusions The study demonstrates the feasibility of conducting face-to-face interviews using TTO in a Brazilian population setting. The estimated values for EQ-5D questionnaire health states based on this Minas Gerais survey represent an important first step in establishing national Brazilian social preference weights for the EQ-5D questionnaire.