Sunday, March 4, 2007

H.R.811: Fact & Friction -- Part I

Now that H.R.811, Congressman Holt's "Voter Confidence and Increased Accessibility Act of 2007", is wending its way through the House with almost 200 co-sponsors, I thought it would be a good idea to blog about some under-reported aspects of the bill and the process that spawned it.

Despite the best intentions of many in the election integrity (EI) community, there has been a relative lack of attention paid to certain crucial problems with the new Holt bill and their potential solutions. I'm speaking primarily about software-independent audits and recounts -- not just "auditability" -- a buzzword bandied about in Congressional testimony and elsewhere.

It's taken years to get members of Congress interested in doing anything about the lack of independent verification of the e-vote tallies generated by touchscreen direct recording electronic (DRE) and optical scan electronic vote-counting systems, but now that we may be on the cusp of doing something (via H.R.811, Rep. Rush Holt's new bill in the House), it doesn't seem to be nearly enough.

Holt's bill, which I had the good fortune to be involved in drafting, leaves much to be desired. And the drafting process itself, while being somewhat open to the EI community, was not as open as advertised. Many important leaders in this movement had not even seen the drafts but more importantly, it was not possible to tell who had and who hadn't because there was no coordinated effort to make this known. This made meaningful collaboration and discussion rather difficult because no one knew who else had a copy of any particular draft and who did not.

Some expert advice was ignored too. How else can one explain the loophole in the bill that allows Election Management Systems such as Diebold's GEMS, ES&S's Unity and Sequoia's WinEDS to be connected directly to the Internet?

Moreover, the final version of the bill was vetted by very few of us and by the time some major mistakes were identified in the bill, it had already been introduced in the House.

One such example is the new language that makes the so-called paper "ballot" the ballot of record only in the event that irregularities are found in an audit or recount. The problem is there is nothing anywhere in the bill to require a recount, no matter how closely a race appears to be decided!

Therefore we are left with at most, a 10% manual audit standing between a correctly tabulated federal election outcome and an incorrect, possibly fraudulent one. Such an audit, which is the only requirement in the bill for a software-independent count, may not be enough to detect an outcome-altering discrepancy, or any discrepancy at all for that matter.

While some states may conduct manual recounts, others will not, or may only conduct recounts using software-dependent methods such as counting cast vote records from DRE memories or rescanning paper ballots.

For example, in Ohio's 15th Congressional District, there was only a 10% hand count of a US House race with a 0.5% margin in 2006. Surprisingly, the Columbus Dispatch reported this as a "recount", but worse than this is the fact that nothing in H.R.811 would have required anything more effective.

So does the audit in the Holt bill guarantee the correctness of outcomes in federal elections? Don't bet on it.

In fact, the 10% "recount" of the Ohio CD-15 race seems to have been a model for H.R.811, which is extremely troubling. If the US Congress, who under the Constitution has the right to judge the elections of its members, is satisfied with only a 10% recount in such close elections, one has to wonder how seriously they are taking this responsibility and for what purpose all those voter-verifiable paper audit records generated as a result of H.R.811 will ever be used.

Although some proposed improvements were included in the bill in one form or another, the most important ones (involving audits and recounts) were never seriously considered, despite some reports to the contrary. The auditing clauses in the bill were based on what seems like an ad hoc approach, requiring more than a minimum percentage of precincts to be sampled randomly, but only in extremely close races, the outcomes of which may not necessarily be confirmed with the mandated audits.

And it gets worse.

The minimum percentage of the audit, which was 2% of precincts in Holt's bill in the previous Congress, H.R.550, has been raised unnecessarily to 3% in this bill. That's an increase of one half. But because this higher minimum applies to every federal contest, it drives up the cost of auditing noncompetitive races or those with thousands of precincts in states such as California. So election officials will have something new to complain about besides having to deal with paper records -- they will be forced to recount many votes unnecessarily. This fruitless exercise will divert resources from the more important task of counting enough votes to confirm the outcomes of closer races where larger audits will be needed.

As I've said on several occasions in correspondence about H.R.811, the minimum audit is too big; the maximum audit is too small; sometimes the audits will be just right, but this will be purely coincidental.

This author, along with others who have written election auditing protocols, has suggested a proper way to conduct audits by basing the sample sizes on the probability of detecting outcome-altering miscounts. Variations in precinct size also need to be taken into account. This was repeatedly brought to the attention of Mr. Holt's office but we were told only that other experts would be addressing the auditing issue.

Finally, shortly after the auditing language in the bill was released, a letter to Mr. Holt penned by those with whom he apparently had consulted, showed that the audits in the bill will not provide a high level of confidence that races were correctly decided in many cases. The letter also stated that attempting to prevent an unacceptable level of error on electronic voting machines through audits may be, "too administratively burdensome."

Rather than throwing in the towel, we will offer a different assessment of the auditing problem and Mr. Holt's solution in Part II of this series.

Fortunately, it's not too late to rectify the errors in H.R.811, but this depends on using the bill as a vehicle for amendments, and also trying to improve the Senate version now under construction by Senator Dianne Feinstein's office.

Here are some comments produced by Teresa Hommel, to which I contributed, with links to other useful suggestions for amending H.R.811. The bill with all its faults still has the potential to help solve the e-vote-counting problem by requiring voter-verifiable paper records, but it also has the potential to do away with software-independent recounts of those records in federal elections by sending a clear message that a 10% hand count is good enough for the US Congress.

The bill's supporters are quick to point out that it's "better than nothing", but personally, I expected more from a member of Congress such as Mr. Holt who, to his credit, has taken so much of an interest in this issue -- especially now that his party is in power.

Click here to read Part II of this series.

7 comments:

Anonymous said...

Big Stanislevic fan here. How long have you been hiding this blog?!?!

NC Voter

Severina said...

I'm worried that all the changes that you propose cannot be incorporated into the legislation soon enough--you seem confident that they can. Are there precedents? 811 has only 200 or so supporters as is; will the revised bill win or lose supporters? Rudy Giuliani is way ahead in the polls of presidential campaigners; McCain is behind but still ahead of any Democrat. Not that I'm partisan per se, but . . .

Anonymous said...

Howard,

What do you think of this:

"Arizona law requires manual audits, but only if the political parties provide representatives to conduct them. As a result, only a third of counties audited ballots after the 2006 election."

from electionline. org

jerry d

Howard Stanislevic said...

So much for the one-precinct-per county minimum audit! I guess that law needs to be re-examined but I'm sure it was fought for by many dedicated activists who deserve a lot of credit for trying.

Howard Stanislevic said...

Severina,

There are numerous provisions in the bill as written that might not be implemented in time for 2008, or that aren't well understood or defined. I don't think trying to improve it will necessarily make anything take any longer, and clarifying the language might avoid some lawsuits, which would actually speed implementation.

Paul said...

What about the whole (ill advised) idea of advertising, via a publication of a statute in the United States Code, the primary way in which we are going to attempt to catch election cheaters? Doesn't this make it a lot easier if an embezzler, for example, knows the methods that will be used to investigate potential embezzlement?

Why not instead, (assuming auditing makes sense at all, a separate debate) that we just specify by statute a level of confidence that must be achieved and let the auditors apply the laws of statistics to achieve that statutory directive, thus preserving at least a little bit of discretion within statistical allowances?

Howard Stanislevic said...

You mean like, "99% certainty that the outcome of every federal election is correctly decided?" What a great idea Paul! Wish I'd thought of that! ;)

A 50-state analysis of this approach will be coming up in Part III of this series. I don't want to give too much away, but suffice it to say that it wouldn't even COST much more than what Mr. Holt has planned for us.

Meanwhile, I hope you will read Part II. And please don't let the audits in this bill give auditing a bad name.