Skip Navigation

Voting Machine Security: Where We Stand Six Months Before the New Hampshire Primary

While there has been substantial progress in securing voting machines since 2016, there is still more to do ahead of 2020.

In late July, the Senate Select Commit­tee on Intel­li­gence released its report on the Russian govern­ment’s attacks on Amer­ica’s elec­tion infra­struc­ture.[1] While the report offered dozens of recom­mend­a­tions related to vast and varied elec­tion systems in the United States (from voter regis­tra­tion data­bases to elec­tion night report­ing), it poin­tedly noted that there was an urgent need to secure the nation’s voting systems in partic­u­lar.[2] Among the two most import­ant recom­mend­a­tions made were that states should (1) replace outdated and vulner­able voting systems with “at minim­um… a voter-veri­fied paper trail,” and (2) adopt stat­ist­ic­ally sound audits. These recom­mend­a­tions are not new and have been consist­ently made by experts since long before the 2016 elec­tion.[3]

Last year, Congress provided $380 million to states to help with upgrades, but it wasn’t enough. This analysis, six months ahead of the first primary for 2020, exam­ines the signi­fic­ant progress we’ve made in these two areas since 2016, and it cata­logs the import­ant and neces­sary work that is left to be done.

I. Repla­cing Anti­quated Voting Equip­ment

The lifespan of elec­tronic voting machines can vary, but experts agree that systems over a decade old are more likely to need to be replaced for secur­ity and reli­ab­il­ity reas­ons.[4] We estim­ate that in Novem­ber 2018, 34 percent of all local elec­tion juris­dic­tions were using voting machines that were at least 10 years old as their primary polling place equip­ment (or as their primary tabu­la­tion equip­ment in all vote-by-mail juris­dic­tions).[5] This number includes counties and towns in 41 states.[6]

States, however, have made signi­fic­ant progress in repla­cing such machines since 2016. In 2017, Michigan began repla­cing its aging voting equip­ment statewide while Virginia decer­ti­fied and replaced all paper­less voting machines.[7] In 2018, Arkan­sas completed the replace­ment of its remain­ing anti­quated paper­less voting equip­ment.[8] In several other states, such as Color­ado, Flor­ida, and Nevada, signi­fic­ant replace­ment has happened at the local level since 2016.

These efforts are ongo­ing and expand­ing across the coun­try in 2019, and several states are in route to having new voting machines in time for the 2020 elec­tions. For instance, Ohio approved $114.5 million to replace aging voting machines ahead of the 2020 pres­id­en­tial elec­tion; Cali­for­nia issued a require­ment for counties to replace old voting machines by March of 2020; and North Dakota recently spent $11 million on new voting machines, which the secret­ary of state expects to be in place by the end of the year.[9]

Nearly Half of the States with Paper­less Voting Machines in 2016 Will Have Replaced These Machines by the 2020 Elec­tions

Action is also being taken to replace elec­tronic voting machines that don’t produce a voter veri­fi­able paper record in many of the states that still use them. Experts have long warned that these machines are a secur­ity risk because they do not allow elec­tion offi­cials or the public to confirm elec­tronic vote totals.[10] The Bren­nan Center estim­ates that compared to 2016, when approx­im­ately 27.5 million voters cast their ballots on paper­less machines, a little over half, or as many as 16 million, will do so in 2020—­though that number could go even lower with addi­tional fund­ing from Congress.[11] As the Bren­nan Center has previ­ously found, many elec­tion offi­cials would like to replace their equip­ment before 2020 but do not currently have the funds to do so.[12] A recent survey of elec­tion offi­cials by Politico produced similar find­ings, noting that “many elec­tion offi­cials have been slow to buy paper-based machines,” in part due “to a lack of money”.[13]

In 2016, 14 states used paper­less voting machines as the primary polling place equip­ment in at least some of their counties and towns (Arkan­sas, Delaware, Geor­gia, Indi­ana, Kansas, Kentucky, Louisi­ana, New Jersey, Missis­sippi, South Caro­lina, Pennsylvania, Texas, Tennessee, and Virginia).[14]

Today, just 11 states use paper­less machines as their primary polling place equip­ment in at least some counties and towns, as Virginia, Arkan­sas, and Delaware transitioned to paper-based voting equip­ment in 2017, 2018, and 2019, respect­ively.[15] In addi­tion to these states, three more (Geor­gia, South Caro­lina, and Pennsylvania) have commit­ted to repla­cing equip­ment by 2020. Consequently, we expect the number of states using paper­less equip­ment as primary systems in at least some counties and towns will drop to no more than eight.[16]

Even so, a signi­fic­ant number of voters may not have a paper record of their vote in 2020. Using voter regis­tra­tion and turnout data from the 2016 and 2018 Elec­tion Admin­is­tra­tion and Voting Survey and 2018 voting equip­ment data from Veri­fied Voting, we estim­ate that as many as 12 percent of voters (approx­im­ately 16 million voters) will vote on paper­less equip­ment in Novem­ber 2020. This compares to 20 percent of voters (27.5 million) in 2016.[17]

In its July report, the Senate intel­li­gence commit­tee noted that “paper ballots and optical scan­ners are the least vulner­able to cyber-attack.”[18] The vast major­ity of juris­dic­tions with voter veri­fi­able paper records use such systems. But a minor­ity of juris­dic­tions in 2020 will use Direct-Record­ing Elec­tronic (DRE) machines with voter veri­fi­able paper trails (VVPTs) or Ballot-Mark­ing Devices (BMDs) as their primary polling place voting equip­ment. Approx­im­ately 6 percent of registered voters live in juris­dic­tions where the primary polling place equip­ment will be DREs with VVPTs and around 7.5 percent live in juris­dic­tions where the primary polling place equip­ment will be BMDs.[19]

These devices collect the voter’s choices and either produce a ballot that is then scanned by the voter in a separ­ate scan­ner (BMDs) or create a “paper trail” that is preserved for poten­tial review later. Experts have warned that some of these paper trails or ballots can be diffi­cult to review.[20] Before purchas­ing such systems, elec­tion offi­cials should consider how easy it will be for voters to review and under­stand the machine marked ballots. In juris­dic­tions where either system is used, elec­tion offi­cials should put in place proced­ures that make it more likely voters will review and catch errors on the paper record, as well as consider addi­tional secur­ity meas­ures recom­men­ded by experts.[21]

II. Progress on Post-Elec­tion Audits

Paper-based systems provide better secur­ity because they create a paper record that voters can review before cast­ing their ballot. Elec­tion offi­cials can review these records during an audit after the elec­tion. However, these paper records will be of little secur­ity value unless they are used to check and confirm elec­tronic tallies.

Tradi­tional post-elec­tion audits, which gener­ally require manual inspec­tion of paper ballots cast in randomly selec­ted precincts or on randomly selec­ted voting machines, can provide assur­ance that indi­vidual voting machines accur­ately tabu­lated votes. Multiple states have employed these audits for over a decade. Currently, 22 states and the District of Columbia have voter veri­fi­able paper records for all votes cast and require post-elec­tion audits of those paper records before certi­fy­ing elec­tion results.[22]

This number should go up to at least 24 states by the Novem­ber 2020 elec­tions, after Pennsylvania and Geor­gia fully trans­ition to paper-based equip­ment.[23] In total, these 24 states and the District of Columbia make up 295 elect­oral votes. The remain­ing 26 states, total­ing 243 elect­oral votes, do not currently require post-elec­tion audits of all votes prior to certi­fic­a­tion. However, there is noth­ing stop­ping most of these remain­ing states from conduct­ing such audits if they have the resources and will to do so.

A New “Gold Stand­ard”: Risk-Limit­ing Audits

Risk-limit­ing audits are a compar­at­ively new proced­ure and offer two import­ant improve­ments to tradi­tional audits. They are gener­ally more effi­cient, requir­ing a review of a smal­ler number of ballots during the audit process. And the stat­ist­ical model­ing used is designed to detect poten­tial inac­curacies in elec­tion outcomes, whether they are the result of acci­dental or inten­tional inter­fer­ence. RLAs can provide assur­ance that the repor­ted winner did, in fact, win the elec­tion, instead of a tradi­tional audit, which only assures offi­cials that machines are work­ing correctly. [24] Because of these features, the Bren­nan Center and many other experts have urged broad adop­tion of RLAs.

States have embraced RLAs at a rapid rate: Color­ado was the first state to imple­ment RLAs in 2017.[25] Not even two years later, more than 12 states are exper­i­ment­ing with the proced­ure in some fash­ion.

Currently, Color­ado and Rhode Island require RLAs imme­di­ately after an elec­tion before results are legally certi­fied; Nevada will do the same start­ing in 2022, thanks to recently passed legis­la­tion.[26] (Local elec­tion offi­cials in Virginia are also required to use the proced­ure, but only once every five years and only after certi­fic­a­tion.)[27] Wash­ing­ton and Ohio allow elec­tion offi­cials to select RLAs from a set of post-elec­tion audit options; Cali­for­nia enacted a similar law last year that will apply for most of 2020.[28]

Elec­tion offi­cials in a slew of these and other states — such as Alabama, Geor­gia, Indi­ana, Michigan, Missouri, New Jersey, Pennsylvania, and Virginia — have either already launched RLA pilots or plan to do so moving forward.[29] Michigan plans to conduct its first state-wide RLA pilot during the March 10 pres­id­en­tial primary, mark­ing the first time that RLAs will be used in conjunc­tion with the pres­id­en­tial elec­tion process in Michigan.[30]

Mean­while, state legis­latures continue to explore the proposal: a bill author­iz­ing RLAs recently passed the Oregon state legis­lature, and lawmakers are consid­er­ing RLA-related legis­la­tion in New Jersey, Ohio, and South Caro­lina.[31] Although some states were able to use the $380 million from 2018 to pilot RLAs, many states had other urgent needs, such as repla­cing aging voting machines and regis­tra­tion data­bases, and have not yet been able to.[32]


While there has been substan­tial progress in secur­ing voting machines since 2016, there is still more to do ahead of 2020. As both the Senate Select Commit­tee on Intel­li­gence and National Academy of Sciences have noted, we should replace anti­quated equip­ment, and paper­less equip­ment in partic­u­lar, as soon as possible.[33] Moreover, of the 42 states that should have paper records of every vote by 2020, 17 are not currently required to conduct post-elec­tion audits pre-certi­fic­a­tion.[34] And as the Bren­nan Center and other organ­iz­a­tions have noted in Defend­ing Elec­tions, elec­tion secur­ity goes far beyond secur­ing voting machines. In addi­tion to the items discussed in this analysis, states and counties need more resources for items like cyber­se­cur­ity support for local elec­tion juris­dic­tions, and upgrades to voter regis­tra­tion data­bases and other crit­ical elec­tion systems.[35]

With research and writ­ing support from William Dobbs-Allsopp.