Virtually everywhere I've ever worked has had an unwritten but widely understood informal policy of placing a multiple on predicted effort for both new code/features and bug fixing to account for Hofstadter's law.