Tuesday, January 15, 2008

NJ Governor Signs Landmark Election Protection Legislation

On Jan. 15, New Jersey Governor Jon Corzine signed S507 into law.

[Note: the text of the law is now available here (PDF)]

I've already written enough about the significance of this bill, what it means to election integrity in the Garden State and the nation, and why it's now the first Law in the nation to seriously address the electronic-vote-counting problem.

What I've not said yet is how appropriate it is for this bill to become law on Martin Luther King, Jr.'s birthday -- not the one that's observed by having a 3-day weekend -- but the actual day, Tuesday, January 15th, 2008, when he would have been 79 years old.

I don't know if this was intentional on the Governor's part. I was not at the bill signing, but it would not surprise me. I do know from having had the privilege of working with Senator Nia Gill, the bill's sponsor who did everything in her power to ensure its passage without any unnecessary compromises, that voting rights and vote-counting rights are inseparably intertwined, as well they should be.

Will this law solve every potential problem with our elections, even if it's copied in all 50 States? Even if Congress begins to take the e-vote-counting problem seriously? No. We know it won't, and some have been all too quick to point this out. It just happens to be a giant step in the right direction, and the people of New Jersey should be proud to have taken it.

Tuesday, January 8, 2008

New Jersey's Post-Election Audit Bill Goes to the Governor!

January 8, 2008

It's alive -- and it's almost a law!

Special-event parking tax surcharges, motor vehicle fines and fees and orthotic appliance health benefit coverage. Such is the business of state government.

But today
at 1:06 AM, at the end of a grueling 12-hour lame-duck session, the New Jersey General Assembly made history by passing A2730/S507, the state's landmark post-election audit bill, by a bi-partisan vote of 53 to 19.

You can watch history in the making here at time index 11:57:35:

The bill -- which is the first in the nation to require hand counts of enough votes to confirm the outcomes reported by electronic vote-counting systems independently of software -- now goes to Governor Corzine for his signature.

Put your hands together for the NJ Legislature!

(Oh, and they also abolished the death penalty. ;-)

Saturday, January 5, 2008

New Jersey's Post-Election Audit Bill and Its Importance to Our Nation

Despite the efforts of election integrity advocates, statisticians and computer scientists, 2007 marked the fifth year since the passage of the Help America Vote Act that the federal government has failed to solve the electronic vote-counting problem. Perhaps more significantly, '07 was the 32nd year since Roy Saltman, working for the National Institute of Standards and Technology's predecessor agency, the National Bureau of Standards, proposed what many believe to be the first workable solution to this problem way back in 1975!

While both Houses of Congress have dithered, on Dec. 13, 2007, the New Jersey Assembly diligently amended their version of the state's post-election audit bill, A2730, to match the Senate version, S507, as amended by Senator Nia Gill. Gill's legislation had already sailed through two Senate committees, and on Dec. 17, the full Senate passed S507 by a vote of 22 to 14. On Jan. 7, 2008 the Assembly will vote on A2730. The bill is expected to pass, and Governor Corzine, who has received many phone calls in support of this bill, is expected to sign it into law.

While there will continue to be calls to replace the state's direct recording electronic (DRE -- pushbutton or touch screen) voting machines with optical scan systems, the audit provisions would apply equally to either system including central-count scanners used to count absentee ballots. Inattentive voters may not verify the paper audit records printed by the DREs, which would be essential to the integrity of the audit, so they will have to be educated and directed to do so. But even those seeking to ban DREs have testified in favor of S507, because they know that optical scanners solve only the electronic-vote-casting problem -- not the electronic-vote-counting one.

Many still have a warm fuzzy feeling when comparing ballot scanners to those used to mark the SATs or other standardized tests, just as some might think that DREs are as reliable as ATMs. But the truth is, both are software-dependent electronic-vote-counting systems, and test after test have shown they were not designed with security in mind.

The same security problems that exist with DREs are present with optical scanners. Both use the same centralized election management systems (typically a single PC for an entire county -- sometimes poorly secured), and with either type of system, the combination of the secret ballot and the trade-secret vote-counting software makes it almost impossible for anyone to know that their vote is being counted as cast. Add to this the fact that error-free software is beyond the state of the art, so even with the best of intentions, it's not possible to know who really won an election counted only electronically with software.

Fortunately the solution to this problem is relatively simple. Not surprisingly, it has little to do with computer science, information technology or software -- and it's not very expensive either. Researchers at Northeastern University and MIT (Aslam, Popa and Rivest) have demystified the statistical procedure involved so that it can be implemented using high-school-level math and a hand calculator. Statisticians, auditors, and other advocates across the nation will now be able to explain this procedure to election officials and average voters.

The solution is to hand count enough votes, at a cost of about 10¢ apiece, to find out who won each audited election contest. And New Jersey's post-election audit bill, S507/A2730 will become the first law in the nation to require confirmation of electoral outcomes by using such a procedure, independently of software.

Rather than live in a world of demon-haunted elections, New Jersey is poised to take the lead in solving the electronic vote-counting problem. If their bill becomes law, it could serve as a model for the nation, with some state-specific tweaks of course. Even now the key provisions of this bill are being copied and tailored to meet other states' requirements.

With the help of public-spirited statisticians, advocates, and election officials, similar provisions can be implemented in any state where there are voter-verified paper records or ballots. Congress might be persuaded to be part of this process, perhaps to help fund it, but with the cooperation of states and counties, federal intervention may not even be necessary to institute this long overdue reform.

The NJ post-election audit bill is a victory for the people of the Garden State; their friends from around the country who helped draft it and showed up to testify in its favor; the election officials who provided valuable insights and consultation; the legislature who passed the bill; and most of all, the bill's sponsor, Senator Nia Gill.

Friday, January 4, 2008

New Jersey's Post-Election Audit Bill: Fact & Friction

A number of people have asked for a copy of S507. Now that it's passed the New Jersey Senate, it can be found on the NJ Legislature's website with the latest amendments here:


The Assembly has an identical version, A2730, amended to match S507 on Dec. 13, 2007:


The bill has passed the full Senate and the Assembly will vote on January 7.

Some folks who follow election integrity legislation have expressed some concerns about this bill. I would suggest that before venturing an opinion as to the adequacy (or lack thereof) of this legislation, everyone should first read the bill, with at least some understanding of its provisions.

Just to correct some misinformation that has been circulated thus far:

1. The bill does require a statistical audit to confirm outcomes of elections reported by electronic vote counts by using hand-to-eye counts of VVPRs and VVPBs. There are NO caps on the size of these audits -- even the initial ones. The size of the audit of each contest would have to be determined by the margin of victory, the number of precincts, the number of votes cast in each precinct and some scientifically reasonable assumptions that set a floor, but not a ceiling, on what the audit board may do.

2. The reason there are 2 different statistical power levels (99% vs. 90%) is that each of these applies to elections for different offices. This is clearly spelled out in the bill.

3. The definition of statistical power, as it relates to S507/A2730 audits can be found here. The language in the bill uses the 100% hand count as the "gold standard" to which the effectiveness of the audit is to be compared. All hand counts must be hand-to-eye counts (i.e., not relying on optical scanners or other software).

4. In NJ, an Election District is just another name for a precinct. It is not a Legislative or Congressional district. The Attorney General is the state's chief election official, but this role may revert back to the Secretary of State.

5. There are over 6,000 Election Districts in NJ, so, e.g., the minimum audit of 2% of these precincts per S507/A2730 would be at least 120 precincts statewide; more for close races and due to stratification of the samples by Legislative or Congressional district.

6. S507/A2730 requires all types of ballots counted electronically to be audited, including emergency, provisional, absentee, military and overseas federal ballots.

7. S507/A2730 requires all ballots cast and counted at the precincts to be subject to audit (random or targeted selection) and compared to precinct totals announced by the counties.

8. S507/A2730 requires "batching" of emergency, provisional, absentee, military and overseas federal ballots counted centrally at the county level so they can be randomly audited (or targeted for an audit) by comparing the vote totals of each audited batch, as reported by the optical scanner at the time the ballots were scanned, to hand-to-eye counts of the same ballots. Every ballot in each of these audit units selected for auditing must be hand counted. The use of batch sampling was necessary because NJ does not sort absentee ballots by precinct. The size of each batch is based on the average precinct size but the batches do not include votes already cast and counted at the precincts, as they are covered by the above audit of precincts.

9. S507/A2730 requires the establishment of an independent audit team and empowers them to oversee audits of any precincts or ballot batches in which they believe votes may have been miscounted in addition to the random audits. This team may promulgate additional regulations, subject to public comment, which may go beyond the strict requirements of the bill, and to implement those requirements. They must have verifiable expertise in statistics and auditing.

10. At minimum, S507/A2730 requires the initial sample size of the audit (which is at least 2%, but could be much larger in close races) to be doubled if a 0.1% change in the vote share of any candidate is detected in the sample. The audit board is empowered to order any additional audits they deem necessary to confirm the outcome of the elections prior to certification up to and including a full hand-to-eye recount. The above doubling in the case of a relatively small error rate in the sample is a floor -- not a ceiling.

11. Full manual recounts are still allowed under existing NJ law, but S507/A2730 allows hand-to-eye counts from the audits of precincts to be used in lieu of recounts of the same precincts, unless a court rules otherwise.

12. At minimum, S507/A2730 allows the same observers permitted to attend recounts to observe audits. Election officials may allow additional observers.

There are other good provisions in this bill, many of which are being adopted in other states to solve their e-vote-counting problems. This bill deserves our support, and once again, I'd suggest that before venturing an opinion about it, people should read it, and do so in the context of New Jersey's other election laws, procedures and voting systems.

Far be it from this writer to suggest that one size fits all. But this will be the first post-election audit law to require electoral outcomes to be confirmed independently of software. In other words, it's well worth emulating.

Iowa's Open-Source Precinct Aggregation Audit Tool

Thanks to Jerry Depew from IowaVoters.org for giving me the opportunity to use this cool tool.

It's all right here: http://www.iowacaucusresults.com

The above URL is a HTML frame. There's nothing to it. But there are a bunch of JavaScripts accessed from this URL contained within the above frame: http://s3.amazonaws.com/caucuspublic/index_live.html
That one's a little harder to remember.

In English, it's open-source precinct aggregation audit software and it can be used for the whole state of Iowa! (Democratic precincts only at the moment.)

In better English, they post the precinct tallies on this website in real time, so anyone who's at the precinct can verify them. If the tallies don't match, people will know. All that's needed is a web browser.

I was able to check Jerry's precinct in Iowa, where I had people (well Jerry anyway) on the ground, to see that the results on the website were the same as at the precinct. And they were! Imagine that.

And I just signed up for access to the "post-caucus analysis" part; I'm supposed to get a password tomorrow.

Note: this is NOT an Election Management System on the Internet; it's just a reporting tool (website).

There Ought To Be A Law

Why should you care?

Because this is free precinct tally aggregation software, and there ought to be a law about having to use it for every election -- not just the Iowa Democratic Caucuses.

It may cost $1-Billion to get paper ballots in every precinct, and even with a statistical audit (which is really the only kind), we still won't know for sure who won our elections without checking precinct aggregation too. At this point, the latter can be done for free, using this Iowan software. As a taxpayer, I'm really interested in the FREE solution, since neither that nor the billion-dollar one solves the e-vote counting problem on its own. But together, they might!

Let's not let our obsession with paper ballots cloud the issue. Any kind of voting system can be rigged by messing with the precinct aggregation -- even Hand Counted Paper Ballots. There are 1,781 precincts in the State of Iowa (and about 180,000 nationwide). I know at least one of them could have been aggregated correctly. Now we just have to check the rest and see if they all add up. I'm hoping the password to the post-caucus analysis will allow that.