One of the challenges faced by advocates of election audits and transparency is that current voting systems each record and store election file data in unique ways. This is no surprise given that vendors have long claimed that their systems are proprietary. But the current model of storing election data in ways that prevent easy sharing and analysis is proving difficult for election officials, statisticians, election integrity advocates, and even voting systems vendors. Because of these problems, serious discussion is taking place about what can be done about standardizing election data.
Often, within a single state there are many different voting systems from multiple vendors. At the same time, many elections, including most federal and statewide races, cross election jurisdictions so that votes for the same race are reported in different ways, depending on the system type used in each district. Even a single polling place may have different types of equipment – an optical scanner and a touch screen device for accessible voting for example – which report results in incompatible ways but which must be combined after the polls close.
On election night, results from different systems are sent to a central location where the results need to accurately collected, totaled, and reported to the media and the public. But when election officials have to deal with data in several different formats these tasks become difficult if not impossible. The problem is further exacerbated by several types of voting systems which export data only in PDF format. This format, well know to computer users everywhere, is designed to be read by humans but is not easily imported into spreadsheets to total detailed election results.
Effective, flexible audits also require a standard data format. Having election data reported in a common format would make it possible to perform auditing using results from small units, even at the granularity of a single machine, and allow quick consolidation of results for election night audits. This means audits can be done at a significantly lower cost. Not only would audits benefit, but a data standard would make it far easier and for members of the public and the media to access and analyze post-election data, increasing transparency and providing independent oversight of reported results.
Verified Voting’s work on developing a standard for election data began in 2008 with comments on the Voluntary Voting System Guidelines. In October, Verified Voting gave a presentation at a National Institute of Standards and Technology meeting on standard data formats. Conference participants were enthusiastic about developing a common format and working to extend and improve the Election Markup Language (EML). EML has been under development since 2001 by the OASIS Election and Voter Services Technical Committee, which includes representatives from voting systems vendors, government officials, computer scientists and data experts, and holds a great deal of promise for the standardization of election data.
In November, Verified Voting representatives presented and participated in a meeting at the American Statistical Association’s national office in Alexandria, Virginia which was organized to discuss the latest research on election auditing. Participants included election officials, statisticians, political scientists, and election integrity advocates who have been working for several years developing principles and techniques for verifying election results. In a statement released after the meeting, participants noted that effective auditing requires common data formats, calling for “…standard election result reporting formats, with support for standardized identification of contests and candidates…”
Much remains to be done towards standards development, and several hurdles still stand in the way before agreement among election stakeholders is reached. It now seems clear that a standard format for election data would improve the accuracy, transparency, and auditing of modern elections.