Barcodes on ballots do not introduce new risks. With or without a barcode, there is a legitimate concern that ballots may be misprinted such that they are read differently by tabulators and voters. The solution to this problem is not to reject barcodes, but instead to run risk-limiting audits.
----
In 2020, 95% of US voters will have a paper ballot. In some cases, they will fill out this ballot by hand. In other cases, they will use a ballot-marking device to make their selections using a touchscreen, then print the ballot and cast it.
Most ballot-marking devices print, in addition to plain text of the voter’s selections, a barcode that contains a copy of those choices, for the purpose of rapid tabulation. These barcodes have become a source of concern for some. We believe this controversy around barcodes is a distraction, because:
The point of a paper ballot is to allow the voter to directly verify that their vote was correctly recorded. Why, then, do ballot-marking devices use a barcode at all? Why can’t they simply use the text?
Barcodes are used because scanning them is fast and reliable, even when they’re smudged or partially damaged. Accuracy and speed are particularly important in election tabulation. The technology to achieve this same accuracy and speed with optical character recognition (OCR) is not yet sufficiently perfected.
In the future, the situation is likely to evolve. We will likely have ballot-marking devices that print only text and scanners that read the same text that was verified by voters with the same speed and accuracy as barcodes. We just don’t have this technology yet.
First, it’s important to immediately address one unfounded fear: no deployed voting machine has ever printed only a barcode. Every ballot-marking device prints readable text that the voter can verify.
That said, in the case where there is both human-readable text and a barcode, the concern of a mismatch between the human-readable text and the barcode is understandable. It could happen, and most voters would not detect this mismatch.
This concern, while valid, is not the emergency some have made it out to be. Such a mismatch leaves an indelible trace. During any post-election audit, mismatches between human-readable text and barcode on a significant enough number of ballots will be evident. The problem is thus detectable and easily remedied: the voter-verified text is clearly the voter’s intent. (Not to mention that a mismatch of this kind would lead to an in-depth investigation of the vendor.)
Tabulation of hand-marked paper ballots is not done by reading the text printed on ballots, either. Instead, timing marks along the edges of the ballots are used to locate bubbles, and the tabulator determines which candidate was selected based on whether the expected position of the corresponding bubble is filled. As a result, a number of things can go wrong with bubble ballots.
A ballot could be misprinted with the wrong candidate order. Current tabulators would not detect this flaw. While a voter would visually recognize their vote as being for Joseph, the tabulator would record the vote for Alice.
Much more subtly, a bubble ballot could be misprinted, with one bubble slightly out of alignment with its corresponding timing mark so that, even if the candidate order is correct, the tabulator would look for a filled bubble a few millimeters off from where the bubble is actually printed, causing some voter marks to go unread. This is much harder to detect than a barcode mismatch, because the misprint is so subtle.
Ultimately, any high-speed tabulation in use today is vulnerable to mistakes in ballot printing, whether that mistake is made by the ballot-marking device or by the ballot printer shop. Until we have better OCR technology, the solution to this issue lies in post-election auditing.
Whether we use barcode or bubble ballots, we should verify that the tabulator declared the correct winner as dictated by what voters verified on paper, even if ballots contain printing errors. The way to achieve this is with a risk-limiting audit.
With a risk-limiting audit, barcodes and timing marks don’t matter, as they are completely ignored. Instead, a bipartisan group of individuals reviews a subset of ballots with their own eyes, evaluating ballots the same way voters did. The risk-limiting audit process ensures that the sample of ballots selected for review is statistically rigorous.
We shouldn’t spend much time worrying about barcodes on ballots. They do not introduce any meaningful new risk compared to bubble ballots. Election integrity efforts should instead focus on risk-limiting audits.