Avoiding Election Bias

By Habib Olapade
Habib Olapade is a second-year law student at Yale University.

In January 2008, Hillary Clinton surprised most political commenters by winning the New Hampshire primary. Clinton’s chances were thought to be severely diminished after Barack Obama defeated her in the Iowa Caucus. Indeed, some pre-election polls in New Hampshire showed Clinton trailing Obama by as much as 13 points. Although political pundits attempted to explain this surprise by pointing to polling errors, groupthink among reporters, identity politics, and a critical, decisive mass of undecided voters, one explanation has been neglected: ballot order.

Before 2008, New Hampshire created a two-step process for primary ballots. First, state officials randomized the name order, and then they rotated name order on ballots precinct by precinct. This method ensured that each candidate appeared first on the ballot for approximately the same number of voters. The New Hampshire State Secretary refused, though, to rotate names during the 2008 primary. The result? Joe Biden was listed first, followed by Hillary Clinton in the fourth slot, and Barack Obama in the eighteenth position. This ordering triggered the ‘primary effect’ phenomenon whereby individuals are more likely to select one of the first choices on the ballot.  Judging from historical data, this may have given Clinton as much as a three-point bump – her margin of victory – on Election Day.      

There are several possible reasons for the primary effect. Democracy demands a lot of citizens who have little incentive to search for political information. Obtaining political information requires the expenditure of scarce resources, and rational citizens will look for ways to minimize those costs. As a result, citizens often delegate the procurement and analysis of political information to political parties, interest groups, the government, and other specialists. In single-district elections, heuristics such as party affiliation, incumbency, and likeability help voters narrow their choices and make decisions. In this low information environment, the primary effect may be due to a confirmation bias; instead of looking for reasons to not select a candidate, voters may be searching for positive reasons to select the candidate instead. As a decision-maker moves down a list, he or she becomes less likely to generate reasons for selecting an item due to fatigue and short-term memory constraints. Another possible cause of the primary effect, is satisficing behavior – selecting the first option that meets a minimum threshold. Voters tend to be more susceptible to the primary effect when they have no information about a race, are ambivalent between the candidates, or are participating in down-ballot, nonpartisan, primary races. Of course, none of these categories are mutually exclusive.    
     
States and their political subdivisions have nearly exclusive control over voting technology, registration criteria, voter identification, ballot rules, and other facets of election administration. [1] Because states are, “laboratories of democracy,” this plenary control has resulted in fifty different and highly unique election administration regimes. [2] Commonalities abound, though. There are four different types of ballot order rules: some states permit local election officials to order the candidates, others place the incumbent candidate or party first, a handful rank the candidates in alphabetical order by last name, and a few create a random order by lottery. [3] The first two methods provide the largest opportunity for intentional bias and have been held unconstitutional by lower federal courts and state courts of last resort. [4] The other two have been upheld against constitutional attack across the board for the most part. [5]

The ballot order effect has a statistically significant and unconstitutional effect on elections. [6] The Supreme Court has held that the right to vote is a fundamental right that is protected under the Fourteenth Amendment’s Equal Protection Clause and that severe infringements on that right are subject to strict scrutiny. [7] When they are not accompanied with a precinct-by-precinct rotation scheme, all four ballot order rules fail this test. Indeed, an equal protection jurisprudence that accepts primary effect bias caused by alphabetical or lottery systems, but rejects that same bias when it is caused by clerical preference or partisan ordering laws, rests on a distinction without a constitutional difference. If primary bias inflicts a constitutionally cognizable injury that cannot be justified with an acceptable state interest, regardless of the circumstances, it makes no difference how the state throws its punches – an injury is an injury. Under these circumstances, rotation and randomization are appropriate remedies because if each candidate appears first on the ballot for an equal number of voters, each candidate benefits and suffers from the primary effect equally. [8]

A cynic might retort that alphabetical ordering and lottery laws that cause  primary bias cannot be unconstitutional because these laws do not limit who can be on the ballot or keep any voter from choosing a preferred candidate. However,  what these arguments miss is how that alphabetical ordering and lottery schemes do not treat similarly situated candidates and voters equally. This results from the way the ballot scheme systematically steers some proportion of voters to certain candidates over others. This harm is intolerable for three reasons. First, state and local jurisdictions can correct this error with precinct by precinct name rotation which is usually just a small fraction of their total election administration budget. [9] Second, randomization procedures without rotation further no cognizable state interest other than resource conservation. [10] The state’s interest in being cheap is not weighty when it is severely restricting its citizens’ right to vote.

Finally, to the extent that the primary effect causes some voters to not select a candidate that, ceteris paribus, they would have selected; this scheme dilutes the voting power of the disadvantaged candidate’s voters. This is so for two reasons. First, the advantaged candidate receives some votes, not because of her policy positions, campaign strategy, or experience but rather solely, by virtue of a random, state generated polling feature: her ballot placement.[ 11] This tactic makes it harder for opposing supporters to aggregate their votes and select their candidate of choice. [12] Second, a ballot ordering scheme unaccompanied by name randomization only permits one candidate to benefit from primary bias, which accentuates the bias’ effect on the election. This one-sided advantage is significant because name randomization equalizes and eliminates primary bias by ensuring that each candidate benefits and suffers from primary bias equally by appearing on the ballot first for approximately the same number of voters.    

Our elections are a product of our election laws. Because this country contains close to 13,000 election districts nationwide, ballot order procedures can vary between and within states. This variation can introduce constitutionally unacceptable election administration errors that threaten the legitimacy of the American people’s political choices. Non-randomized ballot order presents such a risk. While “there is no constitutional right to a wholly rational election based on reasonable consideration of the issues,” there is a constitutional right to rational election laws.       

[1] U.S. Const. art. I § 4.
[2] New State Ice Co. v. Liebmann, 285 U.S. 262, 311 (1932) (Brandeis J. dissenting).
[3] Laura Miller, Election by Lottery: Ballot Order, Equal Protection, and the Irrational Voter, 13 N.Y.U.J. Legis. & Pub. Pol’y 373, 376-378 (2010).
[4] Sangmeister v. Woodard, 565 F.2d 460 (7th Cir. 1977); Gould v. Grubb, 536 P.2d 1337 (Cal. 1975).
[5] Schaefer v. Lamone, No. L-06-896, 2006 U.S. Dist. LEXIS 96855 (D. Md. Nov. 30, 2006); Koppell v. New York State Bd. of Elections, 382 F. Supp. 2d 382 (S.D.N.Y. 1998).  
[6] Joanne M. Miller & Jon A. Krosnick, The Impact of Candidate Name Order on Election Outcomes, 62 PUB. OPINION Q. 291, 293-94 (1998).
[7] Harper v. Virginia Board of Elections, 383 U.S. 663 (1966).
[8] Supra note 3, at 396-402.
[9] Supra note 3, at 403.
[10] Clough v. Guzzi, 416 F. Supp. 1057, 1068 (D. Mass. 1976).
[11] McLain v. Meier, 637 F.2d 1159, 1167 (8th Cir. 1980).
[12] See, e.g., Timmons v. Twin Cities Area New Party, 520 U.S. 351 (1997).
Photo Credit: Flickr User The White House

The opinions and views expressed through this publication are the opinions of the designated authors and do not reflect the opinions or views of the Penn Undergraduate Law Journal, our staff, or our clients.