Ask me no questions, I'll tell you no lies.

Since African-Americans began competing in elections in majority white districts, researchers have postulated that some whites were reticent to tell pollsters that they intended to vote for the white candidate, causing polling to be inaccurate. Prior to 2006, it was difficult to determine empirically whether the Bradley Effect existed, and if it did, how powerful of a factor it was, since there were relatively few cases to examine. This study examines the 2006 electoral contests that pitted African-Americans versus whites, and analyzes the Democratic primaries and caucuses in the 2008 presidential race. We find that a Bradley Effect did exist in about half of the 2006 elections and in about one quarter of the 2008 primaries. However, a ‘reverse’ Bradley Effect occurred about as often.