Elections vs. Banking - Not Such A Bad Analogy After All

There has been a considerably amount of brouhaha over assertions made yesterday by King County Executive Ron Sims that "we had an accuracy rate that any bank would envy.". His comment was based on an error rate of 0.2%, and was received by detractors as a laughable statement.

At first glance one would tend to agree, right? After all, in the run up to the elections everyone was talking about improving machine voting and wondering why it couldn't be as accurate and trustworthy as a bank ATM machine. You'd also think that a 0.2% error rate in banking would be a noticeable problem that people would lose their jobs over.

Yet, if you dig a little deeper one discovers that in actual fact, the banking industry analogy isn't really a bad one.

During a "discussion" at the Sound Politics blog, a comment was left by Norm pointing to an article from August 2004 at Bank Technology News, entitled On-line Billing: Customer Literacy, Software Cut Errors, which includes the following

According to TowerGroup, about 36 million U.S. households-about one third of the total number-will pay bills on-line in 2004. Errors will account for about 0.2 percent of those 1.7 billion on-line transactions. "There is always going to be an error rate that they have to live with," notes Beth Robertson, senior analyst at TowerGroup. But she says rates are remarkably low and roughly half of what they were in 2001.

[emphasis added]

Oh my! That's actually quite shocking on several fronts: 1) the 0% banking error rate myth is debunked, 2) it is a seemingly fully automated process, and 3) perhaps Ron Sims isn't the idiot people claim him to be.

The same article goes on to state,

Nonetheless, a 0.2 percent rate is still about 3.4 million payment mistakes that occur either because of customer error or some other glitch in the system, including problems stemming from billers who may switch bank accounts and fail to notify banks so they can change the routing of customer payments. Banks also may fail to cancel a recurring payment either because the request wasn't processed in time or the customer didn't make the change correctly on the front end.

So the banking systems seem to be plagued by the same type of errors we've seen in the elections:

- customer error, much like voters who couldn't fill out their ballots correctly, or didn't sign them.

- glitches, not unlike voting machines that counted wrong, or lost votes, or couldn't read valid votes marked in pencil.

- switched bank accounts, kind of like change of addresses that produced confusion.

- failure to cancel a recurring payment, similar to deaths that went unrecorded, and felons who were kept on the voter lists.

That's pretty uncanny!

The article sums it up by saying, "the equation will likely never be perfect."

Another article I found from 2003 by Elizabeth Judd at the BAI Web site appears to confirm that bank systems are not infallible (What is the world coming to? Next I'm probably going to discover there is no Santa Claus!):

To those who balk at the thought of electronically-generated errors, he says "no human agent is 100% accurate" either. Customer reps give wrong answers on the phone as well, and from a cost/benefit perspective, some level of error may be acceptable. The question is: What is that level of error?

NetBank's Johnston remains conservative on that point. "I don't think you can plan to make errors with your customers 5% of the time. Yes, all humans make errors, but our error rate with deposits is 0.2%." Unless automated systems can make the error rate nearly infinitesimal, she says she will still want a rep to review all responses.

[emphasis added again]

If you recall detractors of the second, manual recount in November stated that it could never be more accurate than a machine count, yet even the banking industry realizes that humans are needed to improve the accuracy rate of automated systems.

I'd like to also point out this Proposal for Universal Voting in which the author laments a 1% error rate in our current systems. Two comments, 1) the author here also claims a 1% error rate in the banking industry would be unacceptable, but I don't believe he would have expected that 0.2% is apparently common, and 2) compared to his 1% error rate, King County's 0.2% error rate is not bad, a point election officials nationwide have tried to make.

If Sims is to be accused of anything at yesterday’s press conference it is probably having a poor feel for how the manner in which he made his arguments would come across, and how receptive people would be to his comments. Perception is 99.8% of reality, and other analogies could have been better, even if it appears now that his was really quite reasonable.

If you'd like an example of a really bad analogy I'd have to suggest the one County Councilwoman Jane Hague made, when she said,

"There are two things that should be managed to zero defects. One is nuclear technology; the other is elections"

Comparing nuclear technology and elections is laughable. Elections are not a matter of life and death should there be an error, and one does not expect 100% accuracy. I am not suggesting we shouldn’t strive for 100% accuracy, of course we should try, but the reality is we will likely never get there. We can be assured that a future election will inevitably produce close results, so we need to set an acceptable error rate standard, and recognize that the margin of error may again be greater than the vote difference, because no standard can avoid that possibility.

Look for additional commentary on this at Also Also and HorsesAss.org.