The analysis and mitigation of electoral errors: theory, practice, policy.

• Author: Foley, Edward B.

Errors will always plague the counting of votes and, periodically, errors will be big enough to undermine the outcome of a close election. Electoral errors, however, need not be as frequent as they currently are or, when they do occur, as threatening to the legitimacy of an election's result. An analysis of electoral errors, both conceptually and concretely, can lead to proposals that will help reduce their occurrence and their adverse consequences.

This article contributes to this endeavor to minimize the incidence and impact of electoral errors by first defining theoretical methods to measure the extent to which a state's voting process is tainted by error. Second, this article closely examines the way five different states would endeavor to redress four basic types of error that, based on recent experience, might arise. Finally, in light of this analysis, this article proposes specific procedural mechanisms that would tend to protect the legitimacy of an election even when its vote counting is irreversibly infected with error.

1. THEORY

   Since 2000, we have been so used to seeing the voting process malfunction and so focused on fixing the problems that have occurred that we have yet to turn our collective attention to a question of overriding importance: how would we know if the voting process functions properly?

   1. THE INEVITABILITY OF ELECTORAL ERRORS

      Perhaps we know what the voting process would look like if it worked perfectly. No eligible citizen who wants to vote would be prevented from doing so. No ballots except those cast by eligible citizens would be counted as valid votes. No ballot cast by an eligible voter would be tainted by improper influence, whether in the form of financial inducement, coercion, or other inappropriate pressure. The tally of all countable ballots would be entirely accurate. The winner would be determined conclusively before the date for taking office. The loser, while disappointed, would accept the final count as correct, reflecting the prevailing democratic choice of the eligible participants.

      But the voting process will never work perfectly, at least not in any statewide or other large-scale election, like those involving congressional or even state legislative districts. (1) The problem is not simply, or even primarily, one of inaccurate vote-counting machinery. Even if the machine accurately tallies every ballot it receives, and records no extra ("phantom") votes, there is no guarantee that eligible citizens were not inappropriately thwarted from casting a ballot. Registration procedures may prevent otherwise eligible citizens from becoming qualified to cast a countable ballot. Provisional ballots cast by individuals whom officials find to have been unregistered may be set aside as uncountable, although this official finding might be based on an administrative error. (2)

      Alternatively, a polling place may run out of ballots, whether regular or provisional, thereby preventing even admittedly registered voters from casting a ballot. Other polling place problems, like a shortage of poll workers, may cause excessively long lines or, occasionally, prevent a polling place from opening for business at all. Even when a court orders the polling place to stay open for extra hours at the end of Election Day, some registered voters who were turned away when they went to the polling place earlier in the day may be unable to return during the extended hours. Consequently, administrative problems can cause these undeniably qualified citizens to fail in their attempt to cast a ballot.

   2. THE CONCEPTUAL DEVELOPMENT OF AN ELECTORAL ERROR RATE

      As the foregoing shows, there are many ways that the operation of a state's voting system may disenfranchise eligible citizens who undertake the steps expected of them in order to vote. If we can accurately identify the number of individuals disenfranchised by the system in any of these ways for any given election, then we can calculate a Disenfranchisement Rate for that election. This Disenfranchisement Rate would be defined as the ratio of disenfranchised individuals to the number of votes actually counted in the election to determine its winner. (A provisional ballot that should have been counted, but erroneously was not, would increase the Disenfranchisement Rate for that election.) This Disenfranchisement Rate would tell us something significant about how well or poorly the state was operating its voting system, with a lower rate obviously being a sign of a better performing system.

      But this disenfranchisement of eligible citizens who attempt to vote is only one type of error that can infect the voting process. The counting of ballots that are cast by ineligible voters or that otherwise are invalid is another type of error that can undermine the accuracy of the count. It is not necessary that these invalid ballots be cast and counted as a result of fraud. Their inclusion in the total number of ballots counted may be the result of innocent or negligent mistakes. For example, felons, unaware that they are ineligible, may vote, and the system may fail to detect the mistake. A nursing home resident may receive more assistance in casting a ballot than is permitted under state law (or without the procedural safeguards necessary to avert the risk of undue influence).

      The number of invalid ballots cast in most elections in the United States may be relatively small. (3) Nonetheless, it is theoretically possible to calculate precisely an Invalid Ballots Rate for each election. This ratio would identify the percentage of all ballots counted that turned out to have been invalid. Some elections might have higher Invalid Ballots Rates than others, and some states over time might have a higher average Invalid Ballots Rate than other states, suggesting perhaps that their voting systems operate less well on average in this respect than those in other states.

      It is possible to combine an election's Disenfranchisement Rate and Invalid Ballots Rate into an overall Electoral Error Rate. In essence, this number would capture both any "missing votes" that should have been included in the election but were not and all those "improper votes" that should have been excluded yet were counted nonetheless. The Electoral Error Rate would express this sum of all wrongly excluded and included votes as a percentage of all votes actually counted.

      This Electoral Error Rate would be a powerful measure of how well, or poorly, a state's voting system performed its basic function of accurately aggregating the electoral preferences of the eligible citizens endeavoring to participate in democratic decisions. States that averaged lower Electoral Error Rates than others could claim to have better operating systems. (4) In particular, if it turned out that certain procedures significantly reduced a state's average Disenfranchisement Rate while increasing the state's average Invalid Ballots Rate only slightly, that trade-off in the improvement of the state's average electoral Error Rate would show the procedures to be worthwhile. (5) Conversely, if other measures increased a state's Invalid Ballots Rate while yielding essentially no benefits in terms of a lower Disenfranchisement Rate, that measure would be counterproductive in terms of the state's overall Electoral Error Rate. The ability to evaluate different voting procedures in this way might have a salutary effect on public debate concerning the desirability of those procedures.

      In order to turn this Electoral Error Rate into a practical tool for evaluating the actual performance of voting systems, it would be necessary (among other things) to fine-tune what counts as either disenfranchisement or an invalid ballot. Two-hour lines at polling places may constitute disenfranchisement of any qualified voter who can neither wait any longer nor return later during court-ordered extended hours. But what about a one-hour wait? If a polling place never opens because of an unprecedented snowstorm, it is hard to say that the state caused the disenfranchisement of any voters who braved the storm and showed up at the closed precinct attempting to vote. (6) Conversely, if a power outage shuts down the use of electronic voting machines in certain precincts on election day, and the state has no emergency paper ballots as a back-up plan, the disenfranchisement of voters in those precincts might well be attributable to the state and thus included in the Electoral Error Rate.

      Of particular importance in calculating the Electoral Error Rate will be the attribution of errors that cause provisional ballots of unregistered (but otherwise eligible) voters to be rejected. As already suggested, some of these errors may easily be blamed on the state. (For example, state workers misplace registration forms properly submitted by eligible voters.) Responsibility for other errors, however, might be assigned to the voters themselves (if, for example, they make material mistakes in filling out their own registration forms) or third parties (as when groups conducting voter registration drives fail to submit completed forms on time). Similarly, when provisional votes of registered voters are rejected because they are cast in the wrong precinct, should those rejected ballots always be included in the Disenfranchisement Rate (on the ground that the state's voting system prevented the participation of qualified voters who attempted to participate)? Or should disenfranchisement for purposes of the calculation be defined more strictly to encompass only those circumstances in which the state is responsible for the voter's casting the ballot in the wrong precinct? (7)

      Similar judgments would need to be made with respect to the Invalid Ballots component of the Electoral Error Rate. We may be confident that absentee votes "purchased" for twenty dollars apiece should be disqualified as invalid, but what if the inducement is much more subtle (e.g. a pastor's praise...