Risk Limiting Audits: Why and How

A recent, paper by the Risk Limiting Audit Working Group, endorsed by The American Statistical Association, articulates and outlines various types of post-election audits, their requirements, and relative advantages. <read>

The paper compares and describes three types of risk limiting audits. One table summarizes and simplifies the amount of work required for each type of audit:

We say the table simplifies the comparison, because there are a lot of details behind accomplishing a ballot level comparison audit, which would be the obvious choice all other things being equal.

CTVotersCount strongly supports advances in electronic auditing which provide the opportunity for states to exploit the advantages of ballot level auditing, perhaps combined with our current selection of 10% of districts statewide after each election, that would likely provide Connecticut with a much more comprehensive, accurate audit than the one we have today…provided it is enacted and accomplished to meet appropriate standards of ballot security and public verifiability.

We will have more to say on this in future posts. The technology is maturing, with several states providing public tests of machine auditing. We can expect bills in the CT Legislature this year authorizing electronic auditing. We will work to see that they are sufficient to provide public confidence.

In addition to discussing the statistical requirements for risk limiting audits, the paper describes additional requirements and issues in areas such as ballot security, results reporting, public confidence, and vote confidentiality. Connecticut has a long way to go in several of these areas. We cannot help but think that our Coalition audit reports contributed to the following statements, in the section entitled Trustworthy audits: the virtue lies in the details:

No matter how attractive the inherent properties of an audit trail [such as paper ballots], it is only as reliable as it is secure. Past elections have been tainted by allegations – and even strong evidence – of ballot box “stuffing” after the election; anecdotes abound of ballots gone lost. Auditing or recounting an untrustworthy audit trail yields untrustworthy results. Moreover, it is highly desirable not only to assert that the audit trail has been secured, but to be able to demonstrate that it has. Some analysts speak of a compliance audit to verify that the preconditions for a risk-limiting audit have been satisfied…

Risk-limiting audits are easy to define, and in broad outline they are fairly easy to implement: Draw a sample, look at the ballots in the sample, do some math to see if more counting is required. However, some implementation details need careful attention.

Public observation and transparency

Risk-limiting audits provide one means for citizens to monitor how well election systems are functioning. Audits provide valuable information to election officials, but crucially, they inform the public and provide evidence as to whether reported election outcomes are correct. The Principles and Best Practices for Post-Election Audits state the case…

These principles impose important responsibilities both on election officials and on public observers. When all parties take these responsibilities seriously – but not grimly – audit observation builds positive relationships between election officials and the citizens they serve.

Good audits are confidence-building exercises; not-so-good audits are more like sullen skirmishes. In the past, some audit observers and would-be observers have reported events like these: never receiving advance notice of audits despite statutory or regulatory requirements; being confined in one corner of a room with no meaningful opportunity to observe; receiving no information about the procedures to be used; having no opportunity to ask basic questions; witnessing unambiguous violations of written procedures but being unable to persuade officials to refer to, or conform with, those procedures. Contrariwise, many other observers have reported interacting cordially with election officials and workers, in some cases politely making suggestions that were immediately adopted, and generally forming a favorable opinion of the audit and other election processes. Clear written procedures, made available in advance of the audit, help observers and other interested citizens understand how the audit evinces the integrity of the results…

Some states provide for partisan observers in certain election audit processes. We recommend that audits be explicitly open to non-partisan observers as well. All interested individuals and groups should be permitted to observe the audit process to the greatest possible extent. Effective audit observation can increase public confidence in the audit and in the integrity of elections, by making the process more transparent and providing an independent verification of the results.

Observability includes not only direct public observation of audits, but clear reporting of the audit findings. Election officials should systematically report audit results, identifying any differences between the audit and voting system counts, and explaining them if possible. These reports need not be long in order to be informative and reassuring.31 The audit results should be forwarded to state election officials, who in turn should compile a summary statewide report – perhaps within 30 days of completion of the audit. Among other things, this report should integrate results from local officials in a consistent, comprehensible, searchable format. Such a report may enable state officials to detect patterns of error they otherwise may have missed, or simply to document how well voting systems performed.

Connecticut law provides for public notification, and today, procedures provide for sufficient public observation. Yet, the public notification provisions are inadequate. The open observation process has led to documentation that the process is profoundly inadequate to provide confidence in our elections and election officials. Observations have also demonstrated the far from adequate ballot security as well. Official reporting of election results and audit results falls short of requirements and rigor

We reiterate that we favor automated auditing for Connecticut. Yet we continue to caution that our support is conditional on the details of any proposed solution and law. We must always “be careful what we ask for”, recognizing the devil is in the details.


Leave a Reply

You must be logged in to post a comment.