This article analyzes 156 pre-election surveys conducted in 2010 on candidates for President and Governor in the 27 Brazilian Federal Units. The main objective is to describe and explain differences between survey and ballot results. The study uses Mosteller's Method 3 (MM3) to calculate the overall accuracy of each pre-election survey and suggests the Candidate Error Estimation Method (MEEC) to investigate a possible bias against a party or a group of parties during that election. The results show errors greater than that informed to Brazilian electoral authorities but do not justify calling into question the accuracy or partisan neutrality of the whole set of surveys analyzed. Greater differences were found: i) in surveys conducted many days before the vote, ii) during the first round, iii) in low competitive races, iv) when there were few candidates and v) in elections for Governor.
2010 Brazilian elections; pre-election surveys; methodology; error
