How Big Should an Election Audit Be

How Big Shoul d an E lection A udit Be?Fixed R ate A udits D o N ot Work F or E lectionshttp:/electionarchive.org/ ucvA nalysis/US/paper-a udits/ElectionA uditEstimator.pdf Kathy D oppkathy@e lectionarchive.orgJanuary 17, 2007AbstractThis paper pre sents a simple form ula for estimating vot e count audit sample sizes to achieve any de sired certainty for e nsuring t he integrity of election out comes. T he form ula described in this paper for 1estimating e lection a udit sample sizes, was derived first by Rona ld Ri vest.In pa rticular, this paper bri efly shows how to de rive an estimate for vot e count audit sample sizes:1.to achieve any de sired proba bility of de tecting vot e miscount that could alter an election outcome, and22.to incorporate the principle of maximum vot e shift per machine vot e count that is required for logical consistency w hen auditing e lections Vote count audit sample sizes must be based on m argins between the leading t wo candidates because the smaller the margin, the smaller the amount of miscount that could w rongly alter the election out come, and t he larger the audit sample must be to de tect the smaller num ber of corrupt counts. In conjunction w ith this paper, a spreadsheet is available to calculate exact minimum election a udit sample sizes necessary to ensure the integrity of election out comes. M arian Beddill helped to craft this 3new version of a n earlier audit calculator s preadsheet created by Dopp. ...