Turnout Analysis of the August Primary

In a recent press release analyzing turnout in the first August primary, the Secretary of State's office claimed the outcome showed that holding the primary in August instead of September did not depress turnout. But this claim is unfounded. The SoS office doesn't even have historical turnout data for any odd-year primaries before 2005. And this year's primary did have the lowest statewide turnout of any primary for which data is available.

The release, by Elections Director Nick Handy, also made a similarly unfounded implication that mail voting improves turnout:

Among the poll-site counties of King, Kittitas, and Pierce, the projected overall turnout is 25%. [vs. 32% in the rest of the state]. Turnout in the state's two largest counties, King and Pierce, was driven down by poll voters. Combined turnout for poll voters in King and Pierce is expected to reach only 8%, while combined turnout for those voting by mail is likely to reach 33%.

"When voters receive their ballots at their homes, they are more likely to vote," said Handy.

Again, the SoS has no historical data for comparing mail voting with poll voting in Washington. But King County publishes more relevant historical data than the state has. If anything, a review of King County's historical vote-by-mail returns and overall turnout suggests that mail voting does not influence turnout; Turnout varies mainly with voters' level of interest in the races on the ballot (which the Secretary of State acknowledges). The data suggests that it's voters who are more likely to vote in the first place who tend to choose to vote by mail. I haven't seen any data to support the claim that sending someone a ballot makes them more likely to vote.

This scatterplot (click for larger version) plots total turnout by county in every primary for which the Secretary of State's office has data.

The large white circles represent statewide turnout rates, the yellow, red and green circles represent Kittitas, Pierce and King counties respectively and the dark diamonds represent all other counties. This year's primary had the lowest statewide turnout of any primary for which the Secretary of State has data. It's possible that earlier odd-year primaries had even lower turnout, but we don't know. Clearly there's a wide range in turnout from county to county and from year to year. Kittitas County, which still retains poll sites, had above average turnout as it has had in most of the other primaries in the data set. King County had one of the lowest rates of turnout in the state, as it has had in most other primaries (and general elections) in the data set. It's impossible to say how this year's turnout compares to other odd-year primaries, and it's impossible to say how much of the drop-off in turnout from 2005 is attributable to the August date and how much is attributable to reduced interest in the contests on the ballot. But we also don't know enough to claim, as Handy does, that "we're simply seeing the effects of a very typical odd-year primary".

The next two graphs compare turnout in King County with the rest of the state (excluding King) and by mail/poll voters. The top graph shows general elections, the bottom graph primaries. In both graphs the lines mean the following -
Green -- total turnout in King County
Brown -- total turnout statewide (excluding King County)
Dark blue -- turnout among King County mail voters (mail ballots voted / mail ballots issued)
Pink -- turnout among King County poll voters (poll ballots voted / registered voters who weren't issued a mail ballot)
Yellow -- difference between mail turnout and poll turnout
Light blue -- percent of registered voters who were issued a mail ballot (permanent absentee + one-time request)

General Elections 1990-2006 (click for larger version)

Primary Elections 1992-2007 (click for larger version)

Certain patterns come through. Turnout (of both poll and mail voters) is highest in elections where there's high public interest, particularly November 2004, the lowest in off-year primaries. The August 2007 primary had the lowest turnout since the 1999 primary, and the lowest turnout for both poll and mail voters for every election for which such data is available. King County turnout is usually, but not always, a bit below turnout in the rest of the state.

Primary returns suggest some patterns with mail voting. Although the percentage of voters who requested mail ballots increased from 28% in 1998 to 63% in 2007 (light blue line), there's no discernible increase in total turnout during that period. Instead, there only appears to be a noticeable decrease in the percentage of the issued mail ballots that are actually voted. At the same time, the spread between mail-in turnout and poll turnout seems to be falling (yellow line). This is consistent with the hypothesis that sending a voter a ballot does not make that voter more likely to vote, only that motivated voters tend to request mail ballots more than unmotivated voters; and as mail voting has been promoted heavily and more voters have been encouraged to register to vote by mail, casual infrequent voters are becoming an increasingly large percentage of the mail voters. Thus today's mail voters who, as poll voters, only went to the polls at elections they considered most important, are still only participating in key elections and ignoring the others, even if a ballot shows up in their mailbox. I stress that this is only a hypothesis which should be more rigorously tested. But it's a hypothesis that's based on more data than the baseless spin we get from the Secretary of State's office and other vote-by-mail enthusiasts.

Jason Osgood had an earlier response to the Secretary of State's press release.

Data note:
The King County data is from this page, with a few data items filled in from the county election returns archive, public records and newspaper reports. I asked the Secretary of State's office for similar data with historical primary returns and vote-by-mail statistics, but was told that all they had for pre-2005 was this spreadsheet.

Posted by Stefan Sharkansky at September 20, 2007 04:33 PM | Email This