Todd Donovan of Western Washington University writes to me about alternative voting (or IRV) and how this can be impacted by large numbers of ballots being counted after Election Day:

The long count of votes in Australia is somewhat complicated by the Alternate Vote (AV) system used there.

Under AV (known as Instant Runoff Voting in the US), voters rank their preferences for all House of Representative candidates on their ballot.  In the initial round of vote counts on election night only “ordinary votes” (cast at the polling place) are processed.  The candidate with the least first preference votes is eliminated. A second round of counting then proceeds, with the eliminated candidate’s ballots being examined so that her supporters’ second preference votes may be transferred to the remaining candidates. These transfers continue until someone emerges with a majority.

But as pre-poll, postal, absentee, and provisional ballots are received after election day, the process is ran again. Since the final count can be contingent on the order that lower-ranking candidates are eliminated, everything is ran again from scratch – even in districts where a candidate had reached a majority in previous counts. In the vast majority of districts, the leading candidate on election night will win the seat.  In a seat where the margin is razor thin, additional ballots and new preference transfers can cause things to bounce around a bit.

Cross-posted at Election Updates.

I spent the end of last week in Bellingham, WA (a wonderful city by the way–but keep it a secret!) and experienced the impact of a “slow count” firsthand. At least in Washington, the vote totals are updated daily, and for the press, this seems to provide an ongoing source of breathless coverage, as pundits (my friend Todd Donovan, a professor at Western Washington among them) speculate about the remaining ballots.

But much of our conversation as the week went on centered on the Australian election.

Nearly a week after polling day, as many as two million “special” ballots remain uncounted in the Australian federal election, and the balance of party control remains in question. While the parties continue to maneuver over potential governing coalitions, 14% of ballots cast have yet to be counted.

How did Australia get into this situation?

Observers of elections in the United States have lots of experience with slow counts.

In Alaska’s Senate primary battle between Joe Miller and incumbent Lisa Murkowski, Miller currently holds a 1,900 vote lead with 10,000 absentee votes left to be counted. According to the Alaska Secretary of State’s website, “early” and “in-person absentee” ballots are counted “from election night through up to 15 days after the election.” Unlike many other American jurisdictions, Alaska is a “postmark state”: so long as absentee ballots are mailed on or before Election Day, they will be counted up to the 10th day after the election. The slow count in Alaska is exacerbated by another provision in state law that early in person votes cannot start to be processed until 8 pm on Election Night.

So too in Washington State, voters only need postmark their ballot by Election Day. Election outcomes can play out over a period of several weeks while ballots trickle in to local offices and are processed and counted. The Gregoire/Rossi vote count in Washington’s very close Governor’s race in 2004 was a good example, but ongoing coverage of close races in the state continue to provide examples.

California regularly has slow counts for a different reason. The state allows absentee ballots to be returned to any precinct in the county on Election Day. (121,274 ballots were returned in LA County’s 2008 general election in this way.) Obviously, these ballots cannot be processed — signatures verified, envelopes opened, voter intent determined, and votes counted — until well after the polls close.

So where is Australia? It turns out that, in some ways, Australia has adopted the worst of all rules — at least in terms of determining an election outcome on a timely basis. Like many things in Australia, according to Todd, the country’s election law is a fusion of America and Britain, with a bit of Australasian flair.

The various categories of non-precinct-place voting in Australia are wide-ranging:

  • An absent voter in Australia votes on Election Day, but outside of her home division.
  • An interstate voter casts a ballot outside his home state, but on Election Day. (These citizens can cast a ballot at an ‘interstate voting center’, but we’re not sure how these ballots are counted (or transmitted).)
  • A pre-poll vote is what Americans would understand as an early in-person vote. The ballot is cast in a voter’s home division, before Election Day, at an early voting center or any divisional office.[fn]Both pre-poll and postal voters need to satisfy one of a number of conditions, but these are relatively lax. For example, the conditions include “being 8km away from your polling place on Election Day,” or “traveling or unable to leave work.” A similar condition applies to voters in Virginia who work in DC and it has resulted in a notable spike in “excuse required” absentee voting in that state.[/fn]
  • A postal vote would be understood by Americans to be an absentee ballot. Like Alaska and Washington, a ballot need only be postmarked, not necessarily delivered, by Election Day.

In Australia, these non-precinct-place ballots are known as “declaration votes”. Unlike regular polling-place votes, the Australian Electoral Commission (AEC) is expressly forbidden from even opening the envelopes of declaration votes until after Election Day.

So, like Alaska, California, and other jurisdictions before it, Australia now finds that the outcome of its election hinges on a slow and methodical count of absentee ballots. The diverse voting methods and lenient policies are doubtless a boon for many voters, but as with any aspect of electoral administration, there’s a cost to be borne. In this case, the price is a slow count.

This is not a problem, per se. Indeed, ensuring the widest possible franchise in a country with mandatory voting is a laudable goal. But it is something for election administrators and policymakers to pay heed to. Still, in Australia’s case, the Electoral Commission might look to the example of Oregon (a state with a great deal of practical experience in this area), and allow officials to open, verify, and process — though not actually count — ballots before Election Day.

Crossposted at Election Updates.

A neat use of technology in absentee balloting is going live in Maryland. Beginning with the upcoming gubernatorial primary in September, voters will be able to print the appropriate absentee ballot directly from their own computers.

From the article:

Those who created the program argue that these voters cannot vote in person, may not know what address they will be at on election day, and until now have had to wait for the postal system to deliver their absentee ballot, which can pose problems for overseas voters. By getting their ballot from the Internet, voters no longer have to worry where their ballot should be mailed to or how long it will take to arrive.

Political scientist Chuck Bullock notes that a longer, early voting period in Georgia did not result in an increase in turnout (full story is in the Moultrie Observer).

Chuck’s quote:

It seems logical if you extend the voting time you increase voting,” he said. “It turns out it doesn’t work that way. The early research suggested it did bump up participation. All the later research discounts that. More convenience and increasing the time didn’t increase the vote.”

A story in The Plain Dealer notes that voting errors increase as the use of vote by mail increases. 

What’s interesting about this story is that it focuses on a basic mechanical “mistake,” meaning that voters failed to include a proper form of identification or did not use the secrecy envelope.  The much more common error results from voters mis-marking the ballot (over and undervotes).  Residual vote studies have long shown higher rates of voter errors whenever a central count system is used, where voters are not given real-time feedback on the accuracy of their use of the ballot.

Apologies for being a little late on this one, but it is worth linking. As New York begins to move from older voting technology (lever machines) to modern optical-scan machines, NYU’s Brennan Center is drawing attention to a potential programming problem:

Under the new system voters will fill out a paper ballot and then “scan” them into an electronic machine. The State and City Boards have set up the new machines so that they do not give voters adequate warning of “overvotes”– ballots that cannot be read in full because the machine reads the ballot as having too many votes for a particular contest. Instead of returning the ballot, as is done in many other jurisdictions, in New York the ballot will be retained, and a computer screen with present the voter with a confusing message that includes a green “cast” button. Voters are not told that if they press the green button, their vote will not count.

What’s striking about this problem is that it is easily avoidable. Election jurisdictions are able to control whether the machine “warns” voters, or not. It’s an question of programming, not a technological limitation. Indeed, the Center notes:

The only other time these voting machines have been used in the same way in a major election (13 counties in Florida in 2008), they produced overvote rates almost 14 times higher than expected, with thousands of votes for the presidential contest rejected – in comparison to almost no votes rejected in the 36 counties that automatically returned the ballots.

A delay in counting the “deluge” of mail in ballots in last Tuesday’s California primary has sparked calls for County Clerk Barbara Dunmore’s resignation.  Delays in counting ballots in California has been recognized for a long time.  One of the main causes is that California allows citizens to drop off their absentee ballots “in person” on election day at any local precinct.  Administratively, this means that, at the end of the day, all of these ballots need to be transported to the central counting location, validated, opened, and processed. 

CA is also a voter-intent state, which only further slows the processing of vote by mail ballots, where citizens are more prone to make stray marks and errors that are not flagged by optical scanning readers.

I can’t imagine how this could be done in any large county by the next morning, as some state legislators apparently want.  There is a clear tension between a speedy final count and a very generous, vote anywhere and in any way system like currently exists in many California counties.

Full story here

This story out of Passaic County is not another story about absentee ballots and vote fraud.  But it is a cautionary tale about how important ballot handling procedures can be when new voting systems are implemented.

The basic summary is this: a county clerk found 49 uncounted mail in ballots while “handling” the ballots after the election.  Even though the envelopes had been time-stamped indicating that they had arrived on time, the clerk chose not to count them because the election results had already been announced.  A judge overturned this decision, and once counted, a different winner was declared.

What I found most curious, and disturbing, about the story is this quote:

Ken Hirrman, an office administrator with the Passaic County Board of Elections, said he discovered the 49 ballots Tuesday while handling the mail-in ballots.

Hirmann said he noticed the uncounted ballots because they were enclosed in thicker envelopes, indicating that they had not been opened and counted.

I have witnessed a lot of vote by mail and absentee balloting systems and have interviewed dozens of election officials about their administrative procedures.  I can’t imagine putting in place a system whereby the situation above could possibly occur.  This means that the ballot, still inside the secrecy sleeve, and then still inside the stamped envelope, somehow made it through the slicing process, the signature verification process, the separation of the outside envelope from the inner ballot process, and finally the tallying process, and no one noticed that there were still intact envelopes in the batch?

New Jersey has only recently gone to no-excuse absentee and permanent absentee.  I hope they have also paid attention to some of the long established ballot handling procedures put in place in CA, OR, WA, IA, and many other states.

Paul Gronke, Director of EVIC, was quoted in the New York Times today on the potential problems with a top-two primary system in California.

There have been accusations of absentee vote fraud in Lincoln County, WV, when 75% of absentee ballots that were requested were returned, and 90% of those favored one group of candidates.

The response of county officials is not encouraging:

“Our office encouraged every voter to participate in the democratic process, whether it be early voting at the courthouse, at local voting precincts or absentee voting,” Scraggs read from the statement. “An increase in any of the options is an encouraging sign that the democratic process is alive and well.

“Therefore, who would think that having more people vote is a bad thing?” Scraggs read.

This is a wonderful example of how election forensics can detect likely election fraud.  While the technology is complicated (see Walter Mebane’s papers and Alvarez, Hall, and Hyde’s book), the intuition behind the statistics is simple.

In short, 75% absentee ballot return is not “high” or “low” unless you compare it to some standard.  The problem is that it’s impossible to discover absentee ballot return rates from the state’s website.  Since overall turnout was juist 42.99% in 2008, however, that 75% return rate does seem high (keep in mind that absentee voters have actively requested a ballot, already indicating their preference for voting).

As to the totals for one faction vs. the other, the county elections website shows no vote totals at all for 2008!  This makes judging the 2010 results a bit sketchy. 

Data = transparency!

Featuring WPMU Bloglist Widget by YD WordPress Developer