Go to HHH home page.
Election Academy

Election Geekery 101: Turnout

Bookmark and Share


Yesterday, Twitter user DukeDuluth wrote the following which was picked up and re-Tweeted by many anxious for something to discuss in the run-up to the Iowa caucuses:

State with highest % voter turnout in most recent general election should have first presidential contest. Reward Democracy.

I will admit to having been involved in past efforts to study the primary scheduling process - so I will walk away from any discussion about whether DukeDuluth's idea has merit. But his idea is a golden opportunity to discuss the question of turnout - especially since the concept is going to get a lot of airtime over the next year.

Generally speaking, turnout is the proportion of voters who participate in an election, usually expressed as a percentage.

That's where it stops being simple. Here's why.

First of all, we have to decide what data to put into the fraction used to calculate turnout. The numerator is votes - but which votes? In a Presidential election year, this is usually votes for President, but even this number doesn't likely capture the full extent of voters who cast a ballot and fail to record a vote for president - intentionally or not. As George Mason professor (and turnout guru) Michael McDonald explains:

Some people do not cast a vote, even for president. Some failures to record votes are true errors, such as unrecorded votes originating from the infamous hanging chads of the 2000 Florida election. It is important to realize that some people intentionally abstain. for example, the 2004 presidential election 3,688 Nevadans voted for "None of These Candidates" (Nevada is the only state that allows this option). Under-Votes are such blank or indecernable votes. Over-votes occur when a voter selects multiple candidates when only one is acceptable.

A better measure of participation might be Total Ballots Cast, which includes all under-votes and over-votes, but does not include rejected provisional ballots, which are often cast when a voter's eligibility is in question. (Residual vote is the difference between total ballots cast and vote for highest office.) The problem is that most, but not all, states report the total ballots cast.

The good news is that an increasing number of states report total ballots cast, which makes it possible to estimate the total ballots cast for the few non-reporting states. National total ballots cast is therefore an estimate using the correspondence between the vote for highest office and total turnout for the states that provide both numbers.

Once you've settled on a numerator, it's time to choose a denominator - and here's where it gets really interesting.

I would guess that if you asked the average person how to calculate turnout, they'd say it's the percentage of registered voters who cast a ballot. But as we've discussed several times in past blog posts, states have widely (and wildly) different approaches to voter registration meaning that any attempt to use state-by-state registration as a denominator creates not just apples to oranges but the whole fruit salad.

That's why, at least for a while, turnout mavens used voting age population (VAP) as the denominator - it's a common definition state-to-state (18+, at least following the 26th Amendment in 1971) and makes comparisons more meaningful.

But even VAP has limitations. McDonald again:

Declining turnout rates, post-1971, are entirely explained by the increase in the ineligible population. In 1972, the non-citizen population of the United States was less than 2 percent of VAP and in 2004 it was nearly 8.5 percent of VAP. The percent of non-felons among the VAP have increased from .5 to about 1 percent of the VAP since the mid-1980s ...

State turnout rates are not comparable using VAP since the ineligible population is not uniformly distributed across the United States. For example, nearly 20 percent of California's voting-age population is ineligible to vote because they are not citizens.

Consequently, McDonald and others use a figure that he calls voting eligible population (VEP) which is calculated by removing non-citizens and ineligible felons from the denominator. The resulting percentage -

Highest office or total ballots cast

is typically used to express turnout.

So - back to the idea that started it all ... using this method and McDonald's 2008 turnout data, which state would have the nation's first presidential contest?

  • - Not Iowa - it was 6th (70.3% of VEP);
  • - Not New Hampshire - 4th (72%);
  • - South Carolina? Please ... 42nd (58.8%).

No - the first state would be the one which typically leads the nation in turnout ... Minnesota (78.4%).

Which is home to ... Duluth ... which (presumably) is home to ... DukeDuluth.

Whether or not this is coincidence is left as an exercise for the reader :)


  • There is another denominator used for voter turnout -- the one I use and the one that Walter Dean Burnham pioneered. That is age-eligible population, minus non-citizens , interpolated between Census years. I, and others in and out of the academic community, believe this is a better denominator than McDonald's for several reasons, !. It is the only denominator for which there is available comparable data back to 1870 (and 1860 interpolated backwards). 2. McDonald's VEP does not include other factors that influence the denominator -- Americans living outside the country who are not included in the age eligible but who can vote, naturalized citizens who are naturalized in the year of an election but who are not part of the age-eligible figure and, for some years, under and over counts. 3. Data for many of these factors doesn't exist for many years backwards. The over and under count estimates have only been assayed since 1942. There is a way of estimated those Americans living outside the country who are in military or government services and allocating them to states but no way of doing so for others. Naturalization figures are not available until after the election year as aren't figures on those who have moved or died. There is data about the number of convicted felons back into the 19teens but no compilation of law with respect to voting laws affecting them during incarceration and release going back nearly as far. The book that CQ Press published under my name -- Voter Turnout 1788-2009 was juried and was named one of 23 outstanding reference 2010 books by the American Library Association. I believe that figure that I offer will prove to be much better and more durable than McDonald's.

  • One other point. Total ballots cast, which I also compile, would be a better numerator, but... It has not been available for many states over time, so there are no comparables, Which is why any historical approach uses Presidential vote for presidential elections and highest turnout race by state for mid-terms, including aggregate House vote for those states that don't have statewide contests in a particular year.

  • In the end, it must still come down to "Registered Voters." Since there is only a small minority of states allowing same day registration, these are the only folks who have completed the requirements in their state to lawfully cast a ballot. That is the only way to get a real election day turnout number as all others are not elligible to cast a ballot and therefore, are just as unable to do so as a non-citizen. As for the arguement over ballots cast v. total votes for the highest turnout race, the arguement is naturally skewed toward the ballots cast number rather than the highest turnout race. As an election official, I have seen clear instances where a local ballot issue or contest receives far more votes than any upticket contest. Ballots cast also bears no impact from contests where multiple candidates may be elected, and no impact from the growing use of ranked choice balloting.

  • One odd thing I have noticed is that some college textbooks still ignore McDonald's warnings and still use the outmoded VAP method. One textbook I am looking at right now even cites McDonald's concerns but uses the VAP anyway! (Although they state that turnout rates using VAP and VEP are highly correlated, r=.97 for 2006, but I'd think some patterns would still be altered.) In the end, I don't know why they don't either use McDonald's numbers or just the number of Adult Citizens (which you can easily get from the ACS or CPS).

    There is an additional item about eligibility in the CPS November survey that I don't think anybody has explored: when asked why they are not registered to vote, people can say they are not eligible. They are given several options for responding to the question including having recently moved, but even after removing under-age and non-citizens, you get some checking "not eligible." I believe, but may not recall correctly, that non-white citizens are more likely to check this box. Might be worth exploring by somebody someday.

  • Leave a comment

    Humphrey School Sites
    Humphrey New Media Hub