Posted: October 20, 2010
More Voting by Mail? First, Consider the Hidden Costs
Montana’s Secretary of State Linda McCulloch recently announced she would file a bill with her state legislature to move Montana to all-mail elections. In so doing, she cited the growing number of Montanans who already vote by mail, through the absentee ballot process, and the rising per-voter costs of staffing traditional in-person precincts that fewer voters are using.
The reliance on vote-by-mail has certainly grown. Oregon and Washington now vote entirely by mail; among the remaining states, the percentage of ballots cast absentee (which is almost all done by mail) has grown from 4% in 1992 to 14% in 2008. (About one-quarter of Montanans voted absentee in 2008.)
Election administrators have good reasons to want to move more balloting to the postal service. Vote-by-mail is cheaper. It also removes a major barrier to the consistent application of election laws, by eliminating the need to hire and train precinct poll workers. In these difficult economic times, isn’t a cheaper, administratively simpler method of voting a no-brainer?
Not if the policy costs outweigh the savings. Unfortunately, the policy costs — which come in the form of lost votes, decreased legitimacy, and a further skewing of the composition of the electorate — are real. Also unfortunate is the fact that policy analysts and election geeks (me included) have largely ignored these costs as they have analyzed the growth of voting by mail. One defense against not studying the policy costs of voting by mail is that the data just haven’t been available. However, data that bear on the policy costs of voting by mail have begun trickling in, and the initial look is sobering.
The first cost is lost votes. The problem of lost votes was highlighted in the Florida 2000 recount in Palm Beach County with the appearance of the veritable bestiary of chad — hanging, pregnant, swinging, and the like. Something like pregnant chad is a great example of votes being lost through no fault of the voter, because virtually the only way pregnant chad could be produced was if the holding device had not been properly cleaned. Research over the past decade has established solid evidence that most of what used to show up as “blank,” “over-voted,” and “under-voted” ballots in presidential elections — what is now lumped together under the rubric of “residual votes” — was due to some artifact of the machine the voter was using. A goal of the Help America Vote Act (HAVA) was to retire systems that were prone to inducing residual votes, like Votomatic punch cards, in favor of devices that helped save voters from common errors, such as precinct-count optical scanners or direct-recording electronic devices.
Less obvious, but just as important, HAVA provided incentives and mandates that were aimed at reducing even more important causes of lost votes, by improving voter registration systems and encouraging the improvement of election administration practices.
HAVA was mostly oriented toward making sure that if a voter woke up on Election Day intending to vote for president, the vote would be counted at the end of the day. The Act had a very traditional focus on in-precinct voting. HAVA had virtually nothing to say about the lost votes problem from the mail-in ballot end of the process — probably because the problem was invisible in Florida, as it is nationwide. Very little attention has been paid to this problem in any systematic way. We had no idea how big this problem might be.
Buried deep within the Election Assistance Commission’s (EAC) 2008 Election Administration and Voting Survey (EAVS) is a series of questions that ask local jurisdictions about the workflow associated with absentee voting by mail in 2008. In addition, the 2008 Survey on the Performance of American Elections (SPAE), conducted by my colleagues at the Caltech/MIT Voting Technology Project (VTP), helps provide evidence about voters who asked for absentee ballots, only never to receive them. Putting these studies together, we have the best evidence about how many people requested absentee ballots in 2008 and, in the end, how many were counted.
The top- and bottom-lines are these: it looks like 36 million people requested mail ballots in 2008, whereas only 28 million absentee ballots were counted, leading to a “leakage” in the absentee ballot pipeline of 8 million ballots, or roughly 20% of requests.
Where did the leaks occur? The data from these studies suggest that 4 million ballots were requested but not received, 3 million were transmitted but not returned for counting, and 800,000 ballots were returned for counting, but rejected.
Are these lost votes, in the same sense of pregnant chad? No. Are they the same as lost votes due to registration mix-ups or poor polling place places? Perhaps.
The leaky pipeline statistics quoted above need to be treated as the preliminary estimates they are. Here are some things to consider when assessing these estimates of lost votes via the mails in 2008: (1) the data sources are new and incomplete, (2) respondents to the SPAE may have rationalized why they didn’t vote, (3) some people who requested absentee ballots may have eventually gone in to vote at a precinct, and (4) local officials may have rejected many absentee ballots out of an abundance of caution over concerns about fraud.
These estimates are a beginning, not the end, of a discussion about how many voters who rely on the mails to cast a ballot actually have their vote counted on Election Day.
Still, what if the “real” lost vote rate for absentee ballots wasn’t 20%, but just 2%? It would still be greater than the in-precinct lost vote rate.
These figures only try to estimate votes that are lost through the supply pipeline of mail-in balloting. What about the ballots that are returned and accepted for counting?
Here is where we encounter the promise of HAVA in polling-place voting falling short in the case of mail-in balloting. HAVA required that ballots voted in a precinct have some safeguard against over-voting; no such safeguards exist for voting by mail. Not surprisingly, the residual vote rate for absentee balloting is greater than for in-precinct voting. For instance, in Florida in 2008, the residual vote rate for Election Day voting in Florida was 0.54%, compared to 0.78% for absentee voting. While not huge, this difference does represent a total of almost 4,600 ballots, a number that well exceeds the vote margin of some recent close elections in that state.
The second cost is the legitimacy of election outcomes. This is where one claim of Secretary McCullough needs to be examined, which is that after visiting groups involved in voting in Oregon, they “loved voting by mail.”
Well, yes and no.
In the Survey of the Performance of America Elections (SPAE), which I mentioned above, we asked voters nationwide whether they favored a series of voting reforms, including whether elections should all be run by mail. Nationwide, the proposal was a dud, garnering support from only 15% of all respondents. A majority of voters in only two states favored vote-by-mail — not surprisingly, in Oregon (64%) and Washington (51%). The next-closest state was Arizona, at 27% support. (Montana was number six, at 22% support.)
While Secretary McCullough was referring to groups in Oregon who administer elections as “loving” vote-by-mail, not the citizens at large, it is hard to characterize all of Oregon, and certainly Washington, as embracing it. Most support it, but even in Oregon — a state that adopted vote-by-mail by referendum — it seems that a solid minority wish the Beaver State continued to use traditional polling places.
This division over whether voting by mail is a good idea in Oregon and Washington has consequences for how voters judge the election outcomes. The SPAE asked respondents if they were confident their vote was counted as cast. Nationwide, the percentage answering “very confident” was 64%. In Washington, the percentage was 48%; in Oregon, it was 59%. What is interesting is that in these two states confidence in the vote count was related to whether one favored voting by mail, which is not true of voters nationwide. Among the residents of Oregon and Washington, supporters of voting by mail were close to the national average in trusting the quality of the vote count. Among the opponents, less than half were confident their vote was counted as cast.
This suggests that one effect of sanctioning universal voting by mail in Oregon and Washington, has been the creation of a solid minority of voters who doubt the wisdom of the move. Many of them are convinced that the vote count is flawed. As a result, Oregon and Washington voters express among the least confidence that their vote was counted as cast among voters nationwide, as least as measured in this survey.
The third cost is in the composition of the electorate. The best research into the effects of vote-by-mail suggests three things. First, the only elections in which turnout is clearly higher is in local elections, where turnout is already homeopathic. Second, in the biennial general elections, turnout generally does not go up; people who used to drop out of the electorate are more likely to be retained, but new people are not systematically added to the electorate. This leads us to the third point: over time, the net effect of voting by mail is to shift the composition of the electorate toward people who are wealthy and well educated. Voting by mail doesn’t expand the electorate overall, while making it less representative.
What should Montana, or any state considering going over entirely to vote by mail, do? Should they abandon the thought? Not necessarily. The purpose of pointing out the (until now) hidden potential policy costs of universal vote-by-mail elections is to make citizens and policymakers more aware of the potential costs, in terms of democratic values, that this form of voting imposes. Perhaps the trade-offs are worth it. Perhaps the perils could be minimized with some creative thinking.
There is also a middle alternative, between continuing to support the dual system of highly decentralized precincts plus mail-in absentees, on the one hand, and complete vote-by-mail, on the other. That is to allow absentee balloting to continue as-is, while moving to a small number of vote centers, dispersed around a county, that are opened for early voting a couple of weeks ahead of the traditional Election Day. (Some states already have these, but often call them “satellite absentee ballot centers.”) On Election Day, use only the vote centers.
A vote center is an office that is set up to allow a voter who lives anywhere in a county to receive the appropriate ballot and cast it on the spot. Implementing vote centers requires an investment of money, intensive communication with voters, and the deployment of new technologies, such as electronic poll books and ballot-on-demand systems. Even with these investments, local governments could still reap most of the savings and administrative improvements envisioned by vote-by-mail if they just closed hundreds of small precincts and opened a few vote centers. That way, voters who wished to avail themselves of the communal aspects of voting, or avail themselves of the safeguards against over-voting, could do that. And, there is even evidence that Election Day voting centers increase voter turnout.
State and local election officials are coping with budgetary shortfalls, just like the rest of state and local governments, and this work is rarely given adequate funding, even in the best of times. Proposals to save money through precinct consolidations and more voting by mail will only increase in number. As this happens, it will be important for everyone who cares about economy in government and the quality of elections to take a serious look at the hidden costs of these changes. Otherwise, we risk a new lost-votes epidemic of our own making.