[Image courtesy of Eye on The Hill]
Yesterday, my friends and colleagues on Pew’s Election Initiatives team released the first of a series of Election Data Dispatches, which will be dedicated to examining “what data exist, what they say about elections in America and how states and localities use data to increase the efficiency and effectiveness of their election systems.”
You don’t need to go very far back in this blog’s short archives to see how excited I am to see this site make its debut. Take a moment right now and bookmark the page for future reference – I already have.
The first Dispatch looks at cost data collected by the State of North Dakota. For years, the Pew team and I have referred to cost data as the “white whale” of election administration – extremely valuable and eagerly sought, but elusive. Imagine our excitement when we discovered that North Dakota regularly collects such data – and had been doing so for 30 years!
Here’s what Pew has to say about the data:
In North Dakota, research indicates that elections cost significantly more per voter in smaller counties than larger ones.
Less populous counties spent up to $22 per voter to run elections in 2010, while several larger counties spent approximately $3 per voter, according to calculations by the Secretary of State’s office. This parallels a 2001 Voting Technology Project report, which found that one of the least populous counties in North Dakota spent more than $14 per voter, while the largest spent less than $2 per voter in the 2000 election.
North Dakota’s Secretary of State provides a wealth of data on the “cost per vote” for every primary and general election conducted over the past 30 years. The state calculates data from its 53 counties with a simple, one-page election statistics and expense form.
“We’ve been doing it the same way for at least 20 years,” said Lee Ann Oliver, elections director for the state. “The reporting can take a while – some bills don’t come in until several weeks after the election is held – but we ultimately have a 100 percent response rate, with 53 county auditors filling this form out in each of the 53 counties.”
This data raises two issues that are worth mentioning here.
First, election costs are driven not just by financial factors but also by time and geography. Both the findings about primary vs. general elections (posted) and smaller vs. bigger counties (not released) indicate that election costs are not uniform across time or place. Further research will be necessary to dig deeper into the cause, but looking at these initial findings it appears that recurring/fixed costs (which do not vary across time or geography) predominate over marginal costs. This suggests that jurisdictions seeking to control costs must look closely at those line items that are relatively constant over time such as equipment purchase/maintenance. This may explain the the growing buyer’s market in voting technology.
Second, all of us will need to be cognizant of the choice of metric used to evaluate the data. The figures presented in Pew’s North Dakota data are “cost per vote” data derived via this equation:
Election expense/total votes = “cost per vote”
Given what we see in the data, however, such a calculation may be exaggerating the variation between counties and election dates by letting turnout (which election offices cannot control) drive the calculation. A different figure – which offers an opportunity for some “apples to apples” comparisons over time – would be
Election expense/eligible voters = “cost per eligible voter”
Applying this equation to the North Dakota data yields costs per eligible voter that range from a low of 43 cents in the 1996 Presidential primary to $2.09 in the 1998 primary – a much narrower band and one that washes out the unpredictable effects of turnout.
It would be interesting to see how much (if at all) the same calculation would smooth out the variation between counties.
Of course, none of this is the final word on election costs; the best part of Pew’s new Election Data Dispatches (linked again in case you didn’t bookmark it the first time) is that it will give us data we can explore to identify new areas of inquiry.