American voting relies heavily on technology. Voting machines and ballot counters have sped up the formerly tedious process of counting votes. Yet long-standing research shows that these technologies are susceptible to errors and manipulation that could elect the wrong person. In the 2016 presidential election, those concerns made their way into public consciousness, worrying both sides of the political fence. The uncertainty led to a set of last-minute, expensive state recounts—most of which were incomplete or blocked by courts. But we could ensure that all elections are fair and accurate with one simple low-tech fix: risk-limiting audits.
Risk-limiting audits are specific to elections, but they are very similar to the audits that are routinely required of corporate America. Under them, a random sample of ballots is chosen and then hand-counted. That sample, plus a little applied math, can tell us whether the machines picked the right winner.
In nearly all cases, a risk-limiting audit can be performed by counting only a small fraction of ballots cast. For example, the M.I.T. professor Ron Rivest calculates that Michigan could have checked just 11 percent of its ballots and achieved 95 percent confidence that their machine-counted result correctly named Donald Trump the winner of Michigan's electoral votes. Texas and Missouri, with their wider margins in the presidential race, could have counted a randomly chosen 700 ballots and 10 ballots, respectively, to achieve the same confidence level.