Thursday, July 13, 2006

Turnout: high and low these days?

Conventional wisdom says voter turnout is lower than ever, always getting lower, and democracy's doomed. This CW says the 1992 and 2004 presidential turnouts were higher than usual, but still dismal compared to the good old days from 1950 to the early 1960s. Is that really true? Maybe not so much.

The doom approach comes from comparing voter turnout to total resident Voting Age Population (VAP), which is often done in the media. But the proportion of people who can't vote among the VAP has been increasing rapidly due to the ever larger proportion of non-citizens living in the US, and of people in prison, on parole or probation (the 'ineligible'). Prof. Michael McDonald, an election expert at George Mason University and Brookings, has found that accounting for the reduction in the proportion of citizen eligible voters in the VAP makes a big difference. Citizens overseas are eligible to vote, so they should be included too (but as a proportion of total population, they have been roughly constant, so this makes little difference in looking at trends).

Votinglinks downloaded his data and played around (data from ) . Graph 1 below, shows some of MdDonald's results in a different way. You can see that ineligibles (dark circles) has increased steadily as a proportion of the VAP. The total ineligible and non-citizen proportion of VAP (open circles) went from about 4% to nearly 10% over the last twenty-five years.

Graph 1.-- Proportion of ineligible persons and non-citizens of VAP in US.

Lets call the VAP adjusted for ineligibles, resident non-citizens and citizens overseas the Voting Eligible Population (VEP). Graph 2 below shows a recalculation of McDonald's results on voter turnout using the VEP (the line with the black circles). Turnout looks better. There is a zigzag pattern due to the off-year elections, but turnout in 1992 and 2004 was about 60%. That's almost as high as the glory days of the 1950s. McDonald has a series of turnout ratios going back to 1948 (Take a look: In McDonald's graph there is a long low turnout period starting after 1968 from which we may be recovering, finally. We were almost back in 1992, and 2004. Turnout is actually quite volatile, and can jump from low to high within one election cycle. The 1992 election had the highest turnout in 25 years and is sandwiched between two elections with the lowest turnouts in about 40 years of presidential contests. Using McDonald's, data, the average change over four years between similar elections is about 4% of the vote, about 9 million for presidential, and 4 million for off-year elections.

Graph 2.--Voter turnout and registration ratios using the VEP.

But are things getting totally peachy again? Maybe not, in terms of what proportion of registered voters show up at the polls. Look at the line with open circles in Graph 2. That is the ratio of registered voters to the VEP, using FEC data on registration downloaded from Infoplease ( Registration has been going up quite a bit faster than the turnout as proportion of VEP. And the registered voter to VEP ratio seems quite a bit more stable than the turnout ratio. It might be fun to look at turnout as a proportion of registered voters. It might be informative too, since it shows voting as a proportion of people interested enough (or badgered enough during GOTV campaigns) to get ready to actually vote. That's in Graph 3 below, using same data as above.

Graph 3.--Registration to VEP, and turnout to registration ratios.

Note: some years not shown due to incomplete data.

Graph 3 shows a totally different picture. When you look at voter turnout as a proportion of registered voters, there has been a steep decline since at least the early 1960s. Turnout as a function of registration is also quite volatile: in 1992 it was higher than it had been in twenty years, and right after the worst turnout to registration ratio that we can see. Also note that while both the 1992 and 2004 elections were close to the high turnout ratios of forty and fifty years ago, only 1992 was high in terms of turning out registered voters. All elections after 1992, both presidential and congressional, had low turnouts relative to the pool of registered voters available on election day. It doesn't seem to be related to any change in aggregate number of registered voters, since that has been steadily increasing at about the same rate since 1986. The turnout to registation ratio is slowly increasing, but the larger gap between turnout and registration seems roughly constant.

For the overall health of democracy, the total turnout as a proportion of the VEP is most important. But for campaign work close to an election, turnout as a proportion of registered voters is all that matters, since that is the pool of potential voters. Seems like there is a message here regarding the importance of GOTV efforts after the registration deadline has past.

No speculations offered here as to why for any of this. I am not an expert on the politics of it, and there must be several plausible stories, given all the serious history and campaign tricks that happened over this period. Whether the increasing prison population, etc., is a good thing or not, or is related is a separate issue. So, will wrap up with what I think are the main lessons, just looking at the numbers:

  • Turnout as proportion of VEP is lower than in 1950s and 1960s, but seems to be gradually recovering to those previous high levels.
  • Turnout as proportion of VEP and registration is quite volatile.
  • Registration rates highest ever, and increasing at a constant rate.
  • Turnout as a function or registered voters is low compared to 1960s, and seems to have fallen off a cliff after 1992.

Votinglinks has not figured out comments in the few days since this blog started, but the e-mail address is at the bottom of the page. Comments, thoughts or pointing out errors would be greatly appreciated.