November 15, 2003
More troubling voter technology evidence from California recount statistics
As we know, exit polling showed a 2.6% intentional undervote rate on question 1 of the California recall. The average state final figures show an average undervote rate of 4.6%. I had been using the preliminary figures from the state to argue that the ACLU was right in challenging the punch card voting in six California counties, because of their much higher than average vote count. The final statistics vindicate the ACLU. Los Angeles, the largest county, had an 8.9% undervote figure. (Disclosure: I filed a brief supporting the ACLU in this case.)
Anthony Argyriou and Eugene Volokh point out an equally disturbing trend: an underrate of 0 in three counties (Alameda, Kern, and Plumas), all of which used Diebold-made voting machines. See the Secretary of State's undervote figures here.
Alameda and Plumas were the only counties to use Diebold DRE machines. Kern used Diebold-made optical scan machines. Other counties using those same types of machines as Kern had more typical undervote rates. (Here is the list of voting machines by county.)
What does this tell us? First, an investigation is absolutely essential. I recall that preliminary figures gave Alameda a 0.4% undervote rate on the first question. How did this change? Are there any factors other than malfunctioning voting technology that might explain this result? One thing that anecdotal evidence suggests is that because the touch screen prompts voters to avoid unintentional undervoting, it might prompt voters who intended to undervote not to do so---perhaps in an effort to finish voting and leave the polling place.
Is this an argument for the need for a voter verified paper trail? I think we need to wait to hear what the experts from NIST tell us after their conference. Most unusual is the result from Kern using the optical scan ballots. Because the result is inconsistent with the results reported from other counties using the same technology, the explanation likely is not the vote casting machinery. Perhaps there was a problem with the vote counting machinery in Kern.
As I said, the most important thing is an investigation, and a quick one. If there are problems that can be identified and resolved before the 2004 elections, we would all be better off.