From an exhaustive Jun 2009 study of election auditing options, HR 2894 "does not limit risk".

[clear eye]

Implementing RiskLimiting Audits in California

by: Joseph Lorenzo Hall^1,2, Luke W. Miratrix³, Philip B. Stark³, Melvin Briones⁴, Elaine
Ginnold⁴, Freddie Oakley⁵, Martin Peaden⁶, Gail Pellerin⁶, Tom Stanionis⁵, and
Tricia Webber<sup>6

1</sup>University of California, Berkeley; School of Information
²Princeton University; Center for Information Technology Policy
³University of California, Berkeley; Department of Statistics
⁴Marin County, California; Registrar of Voters
⁵Yolo County, California; County Clerk/Recorder
⁶Santa Cruz County, California; County Clerk

UC Berkeley School of Information Report 2009032
18 June 2009

2.2.3 Federal Legislation
...{top p.9}The bill allows followup if errors are discovered during the audit, but the auditors are not required to expand the audit. Because the audit need not progress to a full hand count even when large errors are found, the Holt bill does not limit risk.

If the political will were there, this would easy to remedy. Simply add the requirement for the hand count, when, according to a valid statistical formula, the errors may have compromised the outcome. However, the bill as it stands, does not do the job.

[Bill Bored]

1. There's only one practical way to limit risk:

[Bill Bored]

2. Not so fast, or "exhaustive"....

Here's what Holt's alternative audit method actually says:

(b) Use of Alternative Mechanism-

‘(1) PERMITTING USE OF ALTERNATIVE MECHANISM- Notwithstanding subsection (a), a State may adopt and apply an alternative mechanism to determine the number of voter-verified paper ballots which will be subject to the hand counts required under this subtitle with respect to an election, so long as the alternative mechanism uses the voter-verified paper ballots to conduct the audit and the National Institute of Standards and Technology determines that the alternative mechanism is in accordance with the principles set forth in paragraph (2).

‘(2) PRINCIPLES FOR APPROVAL- In approving an alternative mechanism under paragraph (1), the National Institute of Standards and Technology shall ensure that the audit procedure will have the property that for each election--

-snip-

‘(B) the reported election outcome will have at least a 95 percent chance of being consistent with the election outcome that would be obtained by a full recount.

http://www.govtrack.us/congress/billtext.xpd?bill=h111-2894

Wouldn't that limit the risk of an incorrect outcome to no more than 5%?

[clear eye]

4. Not sure.

I expect they are talking about the audit mechanism prescribed in the bill, not some possible alternative method. Couldn't the study be saying that even when it is clear by the results of the audit that something has gone very wrong, and the original result can't be certified, Holt's bill doesn't spell out what has to happen next? Could it be that the results of an audit which would be valid if occurences were random (unplanned machine glitches, etc.) is no longer reliable in a case of intentional tampering with software patches, or mass flipping by a machine in the middle attack as is suspected to have happened in Ohio in 2004? Perhaps my quote is the statisticians' polite way of avoiding insulting the creators of HR 2894's recommended methodology. They may have meant it doesn't hold up for systematic tampering; that when something like that appears to have ocurred, only a full hand count will produce reliable results? Or maybe it means that the bill assumes that if one machine in a precinct w/ three is audited and found to have a significant error, the entire precinct will be recounted and on that basis the results will be 95% guaranteed, but unfortunately doesn't spell it out.

If school were in session, I would try to contact the two statisticians from Berkeley. I may email them anyway w/ your excerpt from the bill. If they answer, I'll let you know what they said.

[emlev]

3. K&R