Skip Navigation

Common Ballot Design Flaws and How to Fix Them

Poor design can confuse voters and cause their votes to be discounted, but common problems are often easily resolved.

Published: February 3, 2020

Prob­lem No. 1: Contests placed in the same column as instruc­tions

When instruc­tions for voters are placed in a single ballot column and contests directly below them, voters are likely to over­look those contests. Such a design viol­ates guidelines issued in 2007 by the U.S. Elec­tion Assist­ance Commis­sion.

Problem 1 Illustration: Contests placed in the same column as instructions Oxide Design Co.
Click image to enlarge.


The prob­lem: Broward County, Flor­ida, placed voter instruc­tions in the left-hand column and contests directly below them. Since the instruc­tions were in three languages, voters had to get through a consid­er­able amount of text before arriv­ing at the Senate contest beneath them. Voters may have never even looked at the bottom of the long ballot page — which might have hung off the edge of the booth — and instead begun with the governor contest in the second column. Other Flor­ida ballots had shorter instruc­tions and did a better job of seper­at­ing them from the contests.

The result: The place­ment of the contests in the same column as the instruc­tions, combined with long, multiple-language instruc­tions, proved consequan­tial. The differ­ence between the under­vote rate in the Senate contest and the governor’s contest was much higher in Broward Coun­try than in other Flor­ida counties. In Broward County, 3.5 percent fewer votes were cast for senator than for governor, while in other counties, the differ­ence was negli­gible. This repres­ents approx­im­ately 25,000 lost votes in the Senate contest, in which there was a margin of victory of approx­im­ately 10,000 votes.

Prob­lem No. 2: Candid­ates for the same office split onto differ­ent pages or columns

Split­ting candid­ates for the same office onto differ­ent pages or columns tends to invite over­votes, in which voters make too many selec­tions for a contest, usually acci­dent­ally (we illus­trated the issue in our 2012 report Better Design, Better Elec­tions). When all candid­ates for one office are listed in a single column, voters asso­ci­ate them as belong­ing to the same voting task. Conversely, when candid­ates for the same office are listed in multiple columns, voters have a harder time seeing the contest’s bound­ar­ies and might inter­pret the differ­ent columns as belong­ing to differ­ent voting tasks.

Problem 2 Illustration: Candidates for the same office split onto different pages or columns Oxide Design Co.
Click image to enlarge.

The prob­lem: In 2016, counties in Cali­for­nia faced the chal­lenge of fitting 34 candid­ates running in the U.S. Senate primary onto the ballot. Counties took differ­ent approaches to this design chal­lenge: 33 of the state’s 58 counties listed the Senate candid­ates in more than one column and the rest fit the Senate candid­ates in a single column on one ballot page.

The result: Accord­ing to an analysis by Profess­ors David C. Kimball and Martha Kropf, the over­vote rate was 3.6 percent in counties that listed the Senate candid­ates in multiple columns and 0.8 percent in counties that listed the Senate candid­ates in one column on a single page. Ballots with a multi­column format accoun­ted for close to 215,000 lost votes in the contest, accord­ing to a report by research­ers Davit Avagyan and Philip Muller.

When counties anti­cip­ate this type of prob­lem, they should conduct usab­il­ity test­ing to determ­ine the best route. Los Angeles County’s mech­an­ical voting machines (retro­fit­ted from punch cards) required two entire pages to fit the Senate candid­ates, so the county mitig­ated the issue by includ­ing a large red warn­ing icon with instruc­tions on the first page to vote for only one candid­ate, as well as design elements to help voters under­stand that the contest contin­ued onto a second page. The solu­tions were designed with input from voters through a process of test­ing ideas, eval­u­at­ing the results, and iter­at­ing the design. Although split­ting the candid­ates still led to high resid­ual vote rates, the situ­ation could have been much worse if Los Angeles County and other counties in the state had not designed their ballots to minim­ize the prob­lem or educated voters.

Prob­lem No. 3: Multiple contests placed on the same elec­tronic ballot screen

Just as with paper ballots it is best to display candid­ates for the same office on the same page or column, it is best when using elec­tronic ballots to place candid­ates for just one contest on a single computer screen. This is a basic prin­ciple of inter­face usab­il­ity; auto­mated teller machines (ATMs), for example, gener­ally ask one ques­tion at a time and proceed to a new screen only after the user has answered the previ­ous ques­tion. People are far more likely to miss ques­tions if they are asked to answer more than one at a time.

Problem 3 Illustration: Multiple contests placed on the same electronic ballot screen Oxide Design Co.
Click image to enlarge.

The prob­lem: In the 2018 midterm elec­tion, at least some of Geor­gi­a’s elec­tronic ballots contained multiple contests on a single screen. The elec­tronic inter­face in some counties displayed candid­ates for the lieu­ten­ant governor contest on the same screen as candid­ates for other contests.

The result: Four percent fewer votes were cast for lieu­ten­ant governor than for governor, and 2.7 percent fewer votes for lieu­ten­ant governor than for secret­ary of state. That repres­ents 103,290 to 159,024 lost votes for lieu­ten­ant governor in a contest where the margin of victory was 123,172 votes. Since Geor­gi­a’s voting systems in 2018 did not have a paper backup, it is diffi­cult to know with certainty the cause of the high under­vote rate, but history suggests that the place­ment of two contests on one screen could have played a major role.

Prob­lem No. 4: Published sample ballots differ from actual ballots

Sample ballots allow voters to famil­i­ar­ize them­selves with the layout of contests and candid­ates, the differ­ent voting tasks, and the ballot language. As such, they should look like the ballots that voters will use on Elec­tion Day.

Problem 4 Illustration: Published sample ballots differ from actual ballots Oxide Design Co.
Click image to enlarge.

The prob­lem: In the 2018 general elec­tion, ballots in the city of Memphis, Tennessee, were differ­ent than the sample ballots Shelby County published for the city prior to early voting. Whereas the ballot ques­tions had “yes” or “no” response options on the sample, they had “for” and “against” response options on the actual ballot. The sample also failed to mimic the layout of the elec­tronic ballot and there­fore did not provide voters with an approx­im­a­tion of what they would encounter in the polling place. Instead, the sample ballot only listed the contests and candid­ates.

The result: During early voting, voters encountered ballots that were differ­ent than the sample they might have reviewed ahead of time. The discrep­ancy in language, a local voting reform group argued, compoun­ded the issue of already mislead­ing and confus­ing ballot ques­tions. Accord­ing to Elena Delavega, a professor at the Univer­sity of Memphis, “mater­ial meant for the general public should be writ­ten at 900 Lexile, or the equi­val­ent of a fourth-grade read­ing level,” yet the ballot ques­tions were writ­ten at 2,000 Lexile.

Prob­lem No. 5: Diffi­cult absentee ballot envel­opes

The use of vote-at-home ballots (ballots submit­ted by mail or at a drop-off loca­tion, often before an elec­tion) has been stead­ily increas­ing for the past couple of decades. The propor­tion of such ballots has nearly doubled, from 13 percent of votes cast nation­wide in 2004 to 23 percent in 2018. But they are likely to go uncoun­ted if voters fail to prop­erly fill out and sign the envel­ope contain­ing them. Of the returned-by-mail ballots in the 2018 elec­tion, 91.8 percent were coun­ted, 1.4 percent were rejec­ted, and 6.8 percent were not categor­ized as either coun­ted or rejec­ted, accord­ing to the 2018 Elec­tion Admin­is­tra­tion and Voting Survey

Problem 5 Illustration: Difficult absentee ballot envelopes Oxide Design Co.
Click image to enlarge.

The prob­lem: In the 2018 midterm elec­tion, design flaws made the absentee ballot envel­ope in Gwin­nett County, Geor­gia, confus­ing to voters. First, there was no X or box to draw voters’ atten­tion to the signa­ture line. The instruc­tions below the signa­ture line read “signa­ture or mark of elector,” but voters rarely think of them­selves as “elect­ors.” Finally, the envel­ope’s language format was incon­sist­ent: the English and Span­ish voter oaths were placed on top of each, but the English and Span­ish oaths for the “person assist­ing elector” were placed side by side.

The result: In the weeks lead­ing up to the 2018 general elec­tion, reports surfaced about the alarm­ing rate at which Gwin­nett County was reject­ing mail-in ballots. Our analysis revealed that the county rejec­ted a total of 1,690 mail-in ballots ― 21 percent of those rejec­ted in Geor­gia during that elec­tion ― even though Gwin­nett County accoun­ted for only 10 percent of the mail-in ballots submit­ted in the state. Among the issues that led to the rejec­tions were miss­ing birth dates, address discrep­an­cies, and signa­tures that were miss­ing or did not match those on regis­tra­tion records.

Absentee ballot envel­ope design flaws have been addressed else­where. An increas­ing number of states notify voters of prob­lems and allow voters to fix (or “cure”) them. In 2014, a change in Flor­ida elec­tion law allowed voters to cure envel­opes by complet­ing an affi­davit. 

This new law promp­ted Escam­bia County, Flor­ida, to review and redesign its absentee ballot envel­ope. An envel­ope intro­duced for the 2016 primary includes a check­list on the outside, instruc­tions with simple illus­tra­tions, good line spacing, larger text, bullets to make the voter’s oath easier to read, and color to high­light the signa­ture area. Compared to the 2012 pres­id­en­tial elec­tion, the 2016 pres­id­en­tial elec­tion saw an 18 percent drop in absentee ballots miss­ing a signa­ture and a 70 percent drop in rejec­tions over­all after miss­ing signa­tures were cured.