Vot­ing au­dit falls short

Two vot­ing watch­dogs say the state is fail­ing to take true ad­van­tage of the cer­tainty pro­vided by new pa­per bal­lots

Baltimore Sun - - FROM PAGE ONE - Philip B. Stark (stark@stat.berke­ley.edu) is a pro­fes­sor of sta­tis­tics and as­so­ciate dean of math­e­mat­i­cal and phys­i­cal sciences at the Univer­sity of California, Berke­ley. Poorvi L. Vora (poorvi@gwu.edu) is a pro­fes­sor of com­puter sci­ence at The Ge­orge Was

At the Board of Pub­lic Works Oct. 19 meet­ing, mem­bers passed without dis­cus­sion a pro­posal by the State Board of Elec­tions to pay Clear Bal­lot Group Inc. $275,000 for an “in­de­pen­dent and au­to­mated so­lu­tion to ver­ify [the] ac­cu­racy” of the state’s elec­tion re­sults.

Seems rea­son­able, right? Es­pe­cially now that the term “rigged” fre­quently pre­cedes “elec­tion” in this year’s cam­paign rhetoric. The only prob­lem is it won’t work.

We have some ex­pe­ri­ence to back this judg­ment: Be­tween us, we have helped au­dit about 20 con­tests in sev­eral states and de­signed au­ditable vot­ing sys­tems. Meth­ods de­vel­oped by one of us are in laws in two states.

It’s great that Mary­land vot­ers get to vote on pa­per bal­lots this year; pa­per bal­lots that vot­ers can check are the best ev­i­dence of “the will of the peo­ple.” Mary­land’s bal­lots will be scanned and then counted elec­tron­i­cally. As re­quired by hard-won state leg­is­la­tion passed in 2007, the pa­per bal­lots will be stored se­curely as durable ev­i­dence of what vot­ers wanted.

The next step in en­sur­ing that the elec­tronic count shows who re­ally won is to man­u­ally re­view some of the pa­per bal­lots through an au­dit. But the re­cently pro­posed post-elec­tion “au­dit” falls short; it will not look at the marked pa­per bal­lots. In­stead, Clear Bal­lots’ “ClearAu­dit” soft­ware as­sumes the state’s vot­ing sys­tem scanned ev­ery bal­lot per­fectly and uses that in­for­ma­tion in its re­view. But no sys­tem is per­fect; mis­takes hap­pen, equip­ment mal­func­tions. And some peo­ple want to make it look like the right­ful win­ner lost.

There’s no good rea­son not to use the ac­tual bal­lots in the au­dit. Other states re­view the pa­per bal­lots to en­sure that any tab­u­la­tion er­rors didn’t change the out­come of an elec­tion. And mod­ern au­dits can be highly ef­fi­cient; they re­view only a small ran­dom sam­ple of bal­lots.

It is good that the board plans to re­view all votes, races and coun­ties. The pro­posed au­dit­ing tech­nol­ogy can de­tect many types of er­rors. But re­ly­ing on the scans — which are as vul­ner­a­ble as any other com­puter data — lim­its the kinds of prob­lems the re­views can de­tect. Th­es­cans aren’t like pho­to­graphs; they can dif­fer due to ma­chine er­ror, tam­per­ing or human er­ror (for in­stance leav­ing out a batch of bal­lots or scan­ning the same batch twice).

A ro­bust sta­tis­ti­cal au­dit of the elec­tronic re­sults against the pa­per bal­lots can pro­duce strong ev­i­dence that elec­tion out­comes are cor­rect; it can also cor­rect in­cor­rect out­comes. In this con­tentious elec­tion, it is ex­tremely im­por­tant to Mary­land and the na­tion to au­dit elec­tion re­sults against the ac­tual pa­per bal­lots. It is not too late to plan and con­duct a real au­dit. We­would be happy to help.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.