Last week, I blogged about the idea of having everyone vote on paper ballots with a receipt in response to a recent call for such a system to be implemented. The other component of this proposal that was made was that these paper ballots would be hand counted. I want to focus on this issue of hand counting in part because it gets to a key problem with our voting process – ballot length.

There are several problems with hand counted ballots. As Steve Ansolabehere and Andrew Reeves have found, hand counting is an inaccurate method for counting, especially when compared to electronic counting techniques. Using data from New Hampshire recounts, they found that counts done electronically have less variance than counts conducted by hand.

However, even if hand counting were perfect, there is a quite practical reason for election officials to want to use electronic tabulation: ballot length. In American, we like to vote on a lot of stuff, preferably all at once. For example, the 2004 election in most states included at least three federal offices—the president, the Senate, and the House. It also likely included some state elections for the legislature, possibly for judges, and for state referenda and amendments. At the local level there were also likely some races and issues to be decided.

In non-presidential elections, ballots are even longer. Even though the president is not on the federal ballot, in most states the ballot balloons with elections for governor, lt. governor, attorney general, secretary of state, other state executive branch elected officials, state legislative races, local races, and more referenda.

The math for counting ballots becomes quite complicated once we consider how many races are on the ballot. An extreme example would be Los Angeles County, which may have more than one million ballots cast in any given general election. If there are 10 races on the ballot, this would make for 10 million different counts that would have to be conducted. Even if a jurisdiction is smaller—for example, 10,000 ballots cast—10 races would make for 100,000 different counts that would have to be conducted.

Moreover, ballots in California have candidate lists that are rotated by assembly precinct. For a statewide race, a random draw of letters is done, and candidates with a last name starting with the letter drawn first are listed first in assembly district one, last in district 2, and then move up the list. This also complicates the counting process.

The obvious retort to this problem is that the Europeans and other nations vote on paper ballots and know the results quickly, why can’t we? The answer lies in this ballot length question. In many of these countries, elections are all held on different cycles. A local election is just that—a local election for a single office, or perhaps 2 offices—and each election might be voted on different ballots (you vote a single race per ballot and put each ballot in a different box). A federal election or state election would be similar. Referenda would be voted on separately as well.

In the early 1900s, there was a “Short Ballot” movement, which is well-documented in the American Political Science Review. (For those of you with access to JSTOR, do a search for “short ballot movement.”) It aimed to reduce the number of state and local elected officials. The short ballot was part of a bread progressive agenda that included creating the initiative, referenda, recall, direct primary and secret ballot. Supporters offered several potential benefits to this reform. First, it would increase the responsiveness of the government because it centers power in the governor and ensures that there are not competing power centers in state government. Second, it would make voting easier, since individuals would only be voting on a small set of races where voters would be relatively knowledgeable about the races. Some advocates in the short-ballot movement suggested that there should not be more than five races on a ballot.

In the United States, such a system would mean completely rethinking the election process. Instead of having two major elections in November of even-numbered years, there would be more states looking like New Jersey, where state legislative and gubernatorial elections are in odd-numbered years. Moreover, elections might also look more like they do in the city of Los Angeles, where municipal elections are held in the spring and early summer of odd-numbered years.

What is the problem with implementing this? Cost is one. Local election officials would now run major elections every year, and two major elections—state or federal plus local elections—in some years. This would increase ballot printing costs, since there would be no economy of scale from having multiple races on the ballot. It would also increase poll worker recruitment costs, poll worker payment costs, and polling site access issues. Governments would have to make a commitment to running elections differently and be willing to pony up the money to accomplish this.

Not only would the fiscal costs increase, the costs for voters would increase as well. Voters would have to be willing to perform their civic duty more often, with fewer races on the ballot to bring them out to vote. Given the relatively high number of non-competitive races that exist currently in both the Congress and state legislatures, the benefits of voting might be low relative to the costs of turning out to vote.

One obvious solution to this problem would be to expand alternate voting methods, especially all-vote-by-mail balloting. However, this rounds counter to some of the proponents of the paper balloting, who want voters to cast ballots in person and receive the accompanying receipt.

Given the likely difficulty that would exist in accomplishing these changes in ballot length and election timing, we can expect electronic voting to exist well into the future.