Posted: January 28, 2011
Reflections on Footnote 24 of the 6th Circuit Hunter Decision
The decision issued by the three-judge panel of the 6th Circuit Court of Appeals in the matter of Hunter v. Hamilton County Board of Elections contains a very interesting analysis of problems with Ohio’s law about counting provisional ballots when they are cast in “the right church, wrong pew” (RCWP) One of the things that makes the case so interesting and important is that it gives us a better idea about how Bush v. Gore will be applied in difficult election cases.
Another reason it has garnered such interest is that it grows out of what appears to be a state mandate that some voters be disenfranchised precisely because they have followed the directives of election officials. Close scrutiny of the numbers of ballots that were placed into a mind-numbing array categories reveals that the RCWP problem is a non-trivial one.
Not being an attorney, but an election geek political scientist, a third issue drew my interest — the pattern of use of provisional ballots and the patterns of their rejection. Among those interested in election law, election policy, and election administration, these patterns raise the question of whether provisional ballot policies will have disproportionate effects on particular types of voters.
Precisely this concern is raised in footnote 24 of the Hunter decision. The footnote, in full, reads, as follows:
It is also discomforting that Ohio’s rule that all provisional ballots cast in the wrong precinct must be excluded may fall—at least in this instance—unevenly on voters depending on where the Board directs them to vote. In single-precinct polling places there is less room for error than at the multipleprecinct locations that have caused so much difficulty in this case. As a result, fewer provisional ballots are likely to be counted in multiple-precinct polling places than in those that serve only a single precinct. This disparate impact might not be of constitutional significance everywhere in Ohio, but here Plaintiffs assert that “the polling places where most of the error-infected provisional ballots were cast are in African-American areas of Hamilton Country.” Plaintiffs 2d Br. at 3. It appears, then, that the exclusionary rule in this case may accrue to the detriment of a protected class. (Emphasis added)
The italicized sentence proposes a hypothesis that provisional ballots will more often be rejected in multi-precinct polling places than in single-precinct polling places, suggesting the possibility of the policy producing disproportionate effects in minority and non-minority communities.
Although the decision makes no mention of actual analysis along these lines that was presented at trial (and I would welcome learning if anyone has done it, or has the data that would allow it to be done), it is possible to gain some insight into whether provisional ballots are more likely to be used and to be rejected in multi-precinct polling places by use of the EAC’s 2008 Election Administration and Voting Survey (EAVS).
I have written a brief, nine-page memo on the subject that is available here. Let me summarize the gist of the analysis.
First, to be clear, the EAVS does not provide precinct-level data, which would be necessary to test the footnote 24 proposition directly. (Again, I hope someone out there has the data to share with me.) What it does have is a set of detailed statistics about quantities like turnout, the number of provisional ballots issued to voters, the number of provisional ballots rejected, and the reasons why the provisional ballots were rejected, all at the county level. In addition, it contains data about how many physical polling places were used in the 2008 election in each county, along with the total number of precincts. With these quantities for each county, we can test the proposition offered by footnote 24.
On the whole, the county-level evidence in Ohio for footnote 24 is weak. There is no evidence that counties with more multi-precinct voting locations issue more provisional ballots than those with few consolidated voting locations. The answer to whether counties with more multi-precinct voting locations reject more provisional ballots depends on how you set up the problem — what the denominator is for the rejection rate (all in-person voters or just the number of people issued a provisional ballot) and whether one weights by the number of voters in a county.
I utilized four different ways of examining the correlation between the rejection rate of provisional ballots and the average number of precincts at each polling location. In only one out of four analyses is there a correlation that would pass the traditional tests of statistical significance at the 95% confidence level. That is an analysis that compares the rejection rate, measured as the percentage of provisional ballots cast, with the average number of precincts in each polling place.
The accompanying graph illustrates this analysis.
The data tokens are in proportion to the number of provisional ballots issued in each county. Hamilton County’s location is also given.
Examination of this graph shows a couple of interesting patterns. First, notice that Hamilton County is on the low end of counties in the use of multi-precinct polling places. Second, notice that Hamilton County is close to the mean of Ohio in terms of provisional ballot rejection rates. Thus, if there are outliers in Ohio, in terms of provisional ballots and the RCWP problem, Hamilton County is not the first place to look.
Third, the data provide evidence that the amount of discretion employed by poll workers and election officials in the use of provisional ballots is quite high. Note the tremendous variation in the rejection of provisional ballots. The rejection rate as a percentage of provisional ballots cast ranged from 3.2% (Pike) to 38% (Lawrence). Lawrence County, with the highest rejection rate, is the data point at the top of the graph, almost directly above Hamilton County. So, it has the same degree of potential problems with the RCWP problem, with a rejection rate twice that of Hamilton County.
Looking at other data in the EAVS dataset, it does not appear that Hamilton County in 2008 had a particularly large RCWP problem. The percentage of provisional ballots rejected because they were cast by a registered voter who showed up at the wrong precinct was 21%, compared to the statewide average of 19%. The other two largest jurisdictions in Ohio, Cuyahoga and Franklin Counties, had similar reject rates for this reason, 25% and 18%, respectively. The counties with the highest RCWP problems were Lawrence (38%) and Adams (37%), which had the highest rejection rates overall.
In assessing this empirical analysis, it is important to keep in mind that it was done at the county level using data from the 2008 election. Things may have been different in 2010 (we will know when the EAC releases their 2010 data next fall), and they may have been different if we had done the analysis at the precinct level.
Although there is, at best, limited evidence that multi-precinct polling places are causing provisional ballots problems in Ohio, the best data we have about provisional ballot uses and rejection illustrate a considerable amount of discretion being exercised locally. As an aside,
according to the EAVS dataset, 155 provisional ballots were partially counted in Stark County, which is hard to square with Ohio’s provisional ballot counting laws. Logan County reported 1 provisional ballot that was partially counted. All other counties reported either zero or did not report any number. Careful scrutiny of provisional ballot data may demonstrate other types of discretion that are not always apparent when we look at election administration one precinct at a time.
The Hamilton County case is a nice anecdote that illustrates a larger pattern that emerges when we examine nationwide data about election administration — local election officials do not always feel comfortable implementing election laws with Draconian effects. This leads to local officials — county boards and poll workers — making exceptions that are often well-meaning, but contrary to law, and potentially producing disproportionate effects. The degree to which this happens in practice is an important field of election administration that has been little researched, and therefore rarely addressed when laws are put in place. Research I have done with colleagues has suggested, for instance, that a similar level of discretion is being exercised locally in the implementation of voter identification laws — a small fraction of voters (often white women of a certain age) more easily get off without showing photo identification in states like Indiana and Georgia that have very strict identification requirements. At the same time, voters of all sorts are being demanded a photo identification in states that not only fail to require photo identification, but ban the use.
In an era in which policymakers are paying attention to the integrity of the polling place, something to caution against is passing laws and promulgating regulations that local officials will be unable, or unwilling, to enforce consistently. Hard evidence of voter fraud is difficult to come by. Hard evidence of poll worker discretion is easy to come by. If election administration were as data-driven as other areas of public administration, one would think that we would be tackling the problems related to polling place integrity we know to exist.