EVENTS

Date:	Monday, June 21, 2004
Time:	10:15 AM -- 2:45 PM
Location:	Wohlstetter Conference Center, Twelfth Floor, AEI 1150 Seventeenth Street, N.W., Washington, D.C. 20036

June 2004

Preventing Another Florida: Will the Changes Make Things Better?

For better or worse, the battle over the 2000 Florida election still remains with us.  Billions of dollars are being spent by states on changing their voting machine systems in order to correct the perceived problems experienced in Florida.  At a June 21 AEI conference, experts addressed questions of who won the Florida election, whose votes went uncounted, and what has been done and what should be done to prevent these problems in the future.

Panel I: Who "Won" the Election and Whose Votes Weren't Counted?

Dan Keating
Washington Post

The details of the 2000 election in Florida paint a very sloppy picture. Simply looking at the counting of ballots is cause for alarm: Tampa double-counted votes, Palm Beach County erased the results of precincts, and Polk County lost ballots--only to find them during the recount. Absentee ballots were given to citizens who did not even request them. Even the final certified vote had a typo of three hundred votes.
 
There was no state audit to check for errors. Florida was such a mess that the error margin far outweighs anything one could find to determine what really happened. The recount was done by a media consortium. The consortium came back with an analysis of the 175,000 over-votes and under-votes, and this data was released publicly. The data was not simply aggregated but reported to the chad level.

Punch-card ballots have a higher error rate, but electronic voting machines are not a panacea because different machines have different ways of changing your vote. Electronic voting machines also have bad security. The recounts were very subjective since people disagreed on intent on approximately one-third of the reclaimable votes. Allowing write-in votes cost thousands of votes.

Mark Seibel
Knight-Ridder

President Bush is the president by accident, by unplanned and unintended events. Machines do not work sometimes: the stylus could not be long enough, the thickness of the ballot varies, or there could be a maintenance issue with machines. The voters may have made mistakes: some voted for both Bush and Gore, and others voted for anyone but Bush or Gore. Election officials failed to properly count ballots that the machines did not. Clearly, the results could have been different.

The U.S. Supreme Court did not give Bush the presidency; the result would have been the same since the Florida high court's ruling was flawed. Sixty-four thousand under votes showed that had all dimples been counted as votes, Bush would have won by even more. However, most counties did not even follow Florida election law. Florida law states that before a ballot is counted, it must be checked to see if there is a discernible vote, but only thirty of Florida's sixty-seven counties actually did that.

We will never know with certainty; therefore, we must pay closer attention to how people vote. There should be registration requirements on equipment. We need a procedure for close elections. We should not fight things out in court because we make mistakes in court.

John R. Lott Jr.
AEI

Polls in Florida stayed open until 8 p.m. local time. However, Florida is situated in two time zones: eastern and central. All major networks mistakenly reported that the polls were closed in Florida although they were still open in Florida's Panhandle. Gore was declared the winner before the polls closed in an area that is approximately two-thirds Bush.

Surveys show that the last hour of polling is the busiest. This would disproportionately hurt Bush since most voters in the Panhandle are Bush voters. For example, people are less likely to vote in California if the winner is already declared. The western Panhandle cited heavy turnout during the day, but the turnout rate was well below the rest of the state at night.

We find that that if one thousand more African-Americans are added we get 2.6 more non-voted ballots. However, if you break it down by party affiliation, these non-votes tend to be African-American Republicans.

There is no systematic relationship between education and non-voted ballot rates. Rich people have a higher non-voted ballot rate than poorer people. These issues tend to be very random. Also, voting machines and ballot design only explain about eleven percentage points of the variation in non-voted ballots.  Race only adds about 2 percentage points.  In fact, the non-voted ballot rate is all over the place.

Henry Brady
University of California, Berkeley

The butterfly ballot caused many problems. Buchanan's vote was essentially a "statistical anomaly." Many of the people who voted for Buchanan ended up voting for other Democrats in statewide elections. At least two thousand Democratic voters voted for Buchanan on the butterfly ballot.

If we look at adjacent counties in Illinois, Madison and St. Claire--Madison switched to optical scan, but St. Claire stayed with votomatic cards. Madison switched to optical scan and did much better; the residual vote rate decreased. St. Claire ended up doing worse with punch cards. It has to be the voting machines that are getting the residual vote rates down.

Punch cards fail to record voter intent. Cards do not have candidate names on them and there is no check of voter intent. We need a system that is transparent. Personal characteristics such as disability or age make it harder for people to use these systems.

Lott gets a different result because he argues that if we change the ethnicity or skin color of minorities without changing residence, income, or education, residual votes would not be reduced. If machines other than punch cards are adopted, minorities would have fewer lost votes. Machines must improve accuracy and must have feedback for voters.

Ilya Somin
George Mason University

Where one stands on the Florida controversy generally correlates to whom you voted for during the 2000 election. It is quite likely that had there been a recount as requested by Gore, Bush still would have prevailed. Therefore, while Bush v. Gore is not above criticism, it is reasonable to say that the U.S. Supreme Court decision did not determine the outcome of the election.

The fact that we only have aggregate data at the county level by race limits the use of data. With aggregate data we are unable to the match individual ballots to individual voters by race; this makes it difficult to draw conclusions about race and error.

We may not be able to design a voting system that can prevent future Floridas. People looked at ballots very subjectively; we need better technology to limit the potential for human error.

James Garand
Louisiana State University

There is a big problem when we make statements about individuals based on aggregate data. This is known as the ecological fallacy problem. We do not have data on African-American Republicans, so we cannot know if African-American Republicans were more likely to cast a non-voted ballot.

The voter capacity theory, which argues that spoiled ballots are the result of voter characteristics, has some sway. In fact, racial effects continue to be observed even if we control for these capacity variables. The inference is drawn that voter capacity does not explain racial effect.

There are big holes in the voter-discrimination model. For example, we would expect to see higher spoiled ballot rates in counties that have a Republican supervisor, but we do not. In fact, black election supervisors presided over higher rates of spoiled ballots. We do not find evidence that black voters live in areas where machines are more likely to fail. So, the level of evidence for the discrimination hypothesis is not sufficient.

Panel II: What Should Be Done to Stop These Problems in the Future?

Michael Shamos
Carnegie Mellon University

Physical ballots are troublesome for voting, as they require a two-phase operation: First, the voter must mark the ballot.  Second, a machine must read the ballot, which is highly problematic.  Furthermore, the paper ballots must be stored in the event of a recount.  Electronic voting machines offer a much better system, but there needs to be universal standards and testing.  Direct Recording Equipment (DRE), commonly known as "touch-screen" voting machines, have been successfully used since 1980.  Only in the past year have people been concerned with their security.  Adding a paper trail to DRE is inadvisable as it brings back the same problems as paper ballots.  DREs present less risk than paper ballots, but testing and qualification standards must be stringently set.  Independent Test Authorities (ITAs), who certify DREs, should make their testing standards public.

Testing is not rigorous enough right now; in most states, the secretary of state tests the machines for barely five minutes.  In others, a vendor can simply show up with a letter from an ITA certifying that the DRE has been tested. In Alabama, the vendor just needs to guarantee that the DRE has been tested and works properly.  DREs should be tested before, during, and after the election, The Federal Voting Systems Standards needs to set protocols for testing, as well as machine storing and handling.  There should only be one ITA since in a competitive system one will inherently become known for more lax testing procedures.  Furthermore, all DRE software should be open source, as there is no need for trade secret protections since the software is all basically the same.

Adam Cohen
New York Times

Computer errors with DREs, such as those in Florida, have lowered voter confidence.  The computer codes for electronic voting machines should be open source and available for public inspection.  Slot machines in Las Vegas have exceedingly more stringent testing and qualification standards than that for voting machines.  The testing is done publicly in state labs with all employee identities open.  Adding to DRE trepidation, uncertified software patches have been put into certified machines, and sometimes these patches have originated outside the United States.  The voting machine companies hire the ITAs, which presents a conflict of interest.  Software errors in DREs are extremely difficult to find, and computer scientists are extremely skeptical of electronic voting.  If receipts are provided with an ATM transaction, how is it justifiable to not provide one with something as important as voting too?  The main groups against paper trails are voting companies who do not want flaws in their systems uncovered and election officials who fear the added workload.  Electronic voting and paper are stronger together.

John R. Lott Jr.
AEI

There are 32 million Americans who are still going to use punch-card ballots in the upcoming 2004 election, while 50 million Americans are poised to use electronic voting methods.  With DRE machines, it is impossible to have over-votes, and they also greatly reduce the chance of non-voted ballots.  There is a great amount of roll-off moving down the ballot; this means that the further down a ballot you go, the likelihood of non-voting rises.  DREs have a high occurrence of roll-off since voters need to navigate through multiple screens to vote an entire ballot.  Furthermore, it takes time for DREs to refresh their screens and review their choices before the ballot is cast.  Roll-off is much lower in punch-card style voting systems compared to other methods.  African-Americans have a higher roll-off rate, but there is no systematic relationship between race, income, or education in the non-voting ballot rate.  

Dana Walch
Ohio Director of Election Reform

DREs have been used for more than twenty years without compromise, and in fact all voting systems have the same problems.  Paper trails for DREs are advisable, but more importantly the quality and testing of machines need to be increased.  There is no perfect voting system, and the importance of well-trained poll workers should not be overlooked, since election officials--not vendors--conduct elections.  Therefore, the volunteer election workforce must be trained exceedingly well to handle problems arising from electronic voting machines.  Most of the election problems are people-related, not equipment-related.  Furthermore, all equipment needs to be tested prior to the vendor getting paid.

Stephen Knack
World Bank

There are inherent trade-offs between the various different types of voting systems that election officials must consider before choosing which system to employ.  Non-voted ballots would be greatly reduced by punch-card style voting; however, there are other issues to consider.  The timeliness of results is extremely important as the media and the public wants to know the outcome of elections right away, and electronic machine voting is much quicker.  Hand-counted paper ballots would reduce the non-voted ballot rate the most, but the extended amount of time it takes to tally these votes renders it implausible.  Electronic machines, which count much quicker, have the problem of being expensive, and if a polling center has fewer machines, it runs the risk of longer lines and therefore decreased voter turnout.  Furthermore, too many candidates on a ballot render the ballot confusing.  The threshold for mandatory recounts should be lower, even though the likelihood of it changing the election result is slim.  It would be advisable to wait to replace all voting systems nationwide until the costs and benefits of various systems can be analyzed.  Also, it is more likely that there will be a conflict between the popular vote and the Electoral College vote than that voting technology issues will occur.

AEI interns Neil Dutta and Aarif Morbi prepared this summary.