The National High School Exam is the second largest large-scale exam in the world. In Brazil, the students' performance in the test is the main selection criteria to access higher education. The association between student performance and socioeconomic status in large-scale exams has been investigated by researchers in the education field since the late 1960s. To overcome this influence, researchers need find new alternatives. Therefore, this work presents a research that aims to analyze physics questions in ENEM in order to find examples less associated with context variables. The use of questions with low association with socioeconomic status in the ENEM can lead this assessment to be more socially fair, after all, it would select students mostly by proficiency in a certain area of knowledge and not by socioeconomic status. For this, in order to better define the socioeconomic level of the students participating in the exam, we carried out a Full-Information Factor Analysis (FIFA) on the socioeconomic questionnaire. The results of the analysis lead to seven groups of bivariate socioeconomic levels, obtained by the mix of economic and institutionalized cultural capitals (one of the dimensions of bourdieusian cultural capital). Then, we obtained the proportion of correct responses in the physics questions for each of the capital levels. The questions with lower variability in these proportions between groups are the least associated with context variables. In this article, we identified and analyzed items with these characteristics in the 2009, 2012 and 2015 ENEM tests.