Posted: November 10, 2013
Of Bouncing Balls and a Big Blue Shift
It is a fortuitous coincidence that the University of Virginia’s Journal of Law & Politics has just published a piece of mine that shows the relevance of the current vote-counting process in Virginia’s Attorney General election to what might happen if the 2016 presidential election turns on a similar vote-counting process in Virginia. (The piece was presented at a symposium on “The Voting Wars” that took place last spring at UVa.)
The piece is an analysis of shifts that have occurred in presidential elections between initial returns and final certified margins of victory. The piece analyzes these shifts for all presidential elections going back to 1960, for potential swing states—including Virginia—as well as all states. The major takeaway from the piece is that something has changed in the canvassing process since 2000 that makes it significantly more likely that a Democratic candidate for president will be able to gain large number of votes during the canvassing process.
Consider Colorado, for example. As shown in Table A, on page 511 of the piece, prior to the 2004 election the Republican candidate for president gained during the canvass every time except 1964, the year of LBJ’s landslide. Then, in 2004, Kerry gained over 30,000 votes during the canvass—not enough to overcome Bush’s initial lead of over 130,000 votes (see Appendix B, page 543), but a significant cut in that lead nonetheless. Then in 2008, Obama padded his lead during the canvass in Colorado by over 70,000 votes, huge amount. He did not do quite as well in the state in 2012, extending his lead there by “only” 26,884 votes.
The fact that in the three most recent elections Kerry and Obama have been able to gain significantly during the canvass in Colorado, whereas Democrats generally could not do so before, suggests that something new is occurring in the canvassing process that did not happen prior to 2004. Is it a consequence of HAVA and its nationwide requirements concerning provisional ballots? Is it increased reliance on absentee voting? A combination of both, or some other factors?
My piece explores these issues for all swing states as well as the nation as whole. But more data, and more sophisticated statistical analysis of the numbers, will be necessary to draw any definitive conclusions. Charles Stewart and I plan to collaborate on future efforts to see what more illumination we can tease out of available evidence.
Meanwhile, with current attention focused on the Virginia Attorney General’s election, here are some observations.
The AG race is undoubtedly important in its own right. But it is also a chance to see how Virginia’s procedures would handle a super-close presidential election. Whenever there has been a vote-counting dispute in Ohio recently (as there have been in some congressional and local races), the participants involved—including the lawyers—know that they are potentially setting precedents that might affect a presidential election. Indeed, the Sixth Circuit’s 2011 decision concerning provisional ballots in a judicial election, Hunter v. Hamilton County Board of Elections, was a major Equal Protection precedent for litigation that occurred in Ohio during the 2012 presidential campaign.
Virginia, albeit to a lesser extent than Colorado, seems consistent with the general trend showing that Democratic presidential candidates now are able to do better during the canvass than they usually did in the past. Obama was able to gain almost 80,000 votes during the canvass there in 2008 and over 40,000 in 2012. But Kerry did not gain during the canvass in 2004; Bush did by almost 10,000 votes. And both Carter and Clinton were able to gain during the canvass in the three years that those Democratic candidates won the White House: 1976, 1992, and 1996. Thus, Virginia’s numbers suggest that what may matter more there is the prevailing political atmosphere at the time: the candidate that wins the White House tends to gain during Virginia’s canvass. But, interestingly, it is not uniformly a situation in which the candidate who won the state added to his lead during the canvass: although Obama won Virginia in both 2008 and 2018, Carter and Clinton lost Virginia in 1976, 1992, and 1996. And it is worth noting that Carter and Clinton’s gains in Virginia during the canvass were miniscule in comparison to Obama’s: about 500 votes in 1992, about 1200 in 1976, and less than 2000 in 1996.
What may be most intriguing about Virginia’s numbers is the possibility they suggest of a Democrat being able to come from behind and overtake a Republican opponent during the canvass. Suppose, for example, that in 2016 Hillary Clinton trails Chris Christie in Virginia by just 1000 votes on the morning after Election Day—and that whoever wins Virginia wins the White House. The numbers from previous presidential elections in the state suggest that Clinton might well overtake Christie during the canvass. (I make this observation, not from a partisan perspective, but as an analyst concerned at how rules governing the vote-counting process might affect conduct by actors, especially partisans, within that process. Any rules that potentially have an effect of engendering distrust in the vote-counting process are a cause for concern, and if one party has an asymmetrical advantage as a result of some rules, there is a risk that this asymmetry might breed distrust.)
It is with what might happen in 2016 in mind that I’ve been watching the numbers move during the canvass in this year’s Virginia Attorney General race.
I don’t profess to have followed all the numerical shifts during the canvass as closely as others have, especially Dave Wasserman of Cook Political Report (whom Chuck Todd of NBC News has nominated for the yet-to-be-created position of national director of counting votes). Wasserman has been keeping a Google Doc of all these vote changes so far and tweeting as they come in. (This VA AG canvass has put Twitter in a whole new light for me.) My impression, from keeping an eye on his great work as well as related tweets from others, is that the lead has gone back and forth repeatedly since Election Day, sometimes because of an accumulation of small shifts and sometimes because of single larger ones (like the so-called “Seven Corners Surprise”). So far, I haven’t detected a distinct trend in either direction, and I haven’t seen shifts of the magnitude that have occurred in presidential elections.
Of course, the canvass is not yet finished. As I understand it, counties must certify their results this coming Tuesday, but the State Board of Election has until Nov. 25 to certify a statewide result, and a recount most probably will follow that. Thus, final certification of the election is likely weeks away. And, if recent reports are any indication, there may be litigation over provisional ballots. As Rick Hasen observes, if it is true that the State Board of Elections has ordered a change in the procedures by which provisional ballots may be validated, that rule change potentially raises federal constitutional questions under both Bush v. Gore and the earlier important appellate precedent of Roe v. Alabama.
But for now I’m curious about the “bouncing ball” quality of the Virginia AG canvass so far in comparison to the “big blue shift” that I generally observed in my analysis of the numbers from presidential elections. If there is a difference, what accounts for it? Are presidential elections simply distinct from other statewide races, especially in “off year” elections—or “off off year” ones, as some people call elections in odd-numbered years when there are no federal races on the ballot?
Since working on this “Big Blue Shift” piece, I’ve wanted to expand the project by collecting equivalent data for gubernatorial, U.S. Senate, and other major statewide elections. It will be interesting to see if in each state there are similar shifts in the canvass each year regardless of which statewide race is at issue (in other words, for example, a presidential year whether the state is also electing a governor or U.S. Senator). Or, conversely, perhaps there are race-distinctive patterns that emerge even within particular states. Maybe, perhaps, this “Big Blue Shift” is a uniquely presidential phenomenon. But if so, why?
So stay tuned. Both to the Virginia AG race specifically and to the implications it might have for the understanding of the vote-counting process more generally.
There is so much more to learn about how the vote-counting process actually works in our fifty states, as well as the extent to which this process may have changed in each state since 2000. As Doug Chapin, Heather Gerken, and others like to say, we really need more data on how our whole election administration works. This point certainly applies to the canvassing of returns after Election Day and the effect the canvass can have in super-close races.
Edward B. Foley is Director of the Election Law @ Moritz program. His primary area of current research concerns the resolution of disputed elections. Having published several law journal articles on this topic, he is currently writing a book on the history of disputed elections in the United States. He is also serving as Reporter for the American Law Institute's new Election Law project. Professor Foley's "Free & Fair" is a collection of his writings that he has penned for Election Law @ Moritz. View Complete Profile