The science of opinion measurement often advances because of big mistakes. Three of the biggest mistakes are easy to identify. Did we just make a fourth?