Election-day lines in the Global South are often cast as a gauge of citizens’ commitment to democracy. The European Parliament observer report on the 2015 Nigerian election noted that “[l]arge numbers of citizens had to queue for much of the day when seeking to cast their votes […] Their patience, enthusiasm, and civic commitment have been an example to the world.” By contrast, election-day lines in the recent U.S. elections were put forth as an affront to democratic practice — voting should be easy! These disparate views of election-day lines mask a common cause: election administration choices. My recent research in Comparative Political Studies reveals how election-day lines — and the policy decisions behind them — shape voter behavior and electoral outcomes.
In the study of African politics, the role of elite behaviour in shaping voter behavior (particularly with respect to violence, fraud, and vote-buying) has been extensively studied. Election administration — the practical organization and implementation of elections — receives much less attention, particularly in low- and middle-income countries outside of the global North. This focus is problematic because election administration in any context shapes voter behavior and, thus, election outcomes. In this post, I discuss the impacts of one specific implication of election administration — election-day lines — on voter behavior in Kenya’s August 2017 elections.
Managing Election-Day Queues
In Kenya, politicians have long decried the waits faced by voters at the polls. In 1983, future President Mwai Kibaki explained, “if you are forming only one queue… [and] go to only one ballot box to cast your vote, this will lead to a slowing down of the whole system”.
“We know many people who have refused to cast their votes… because they waste about an hour queueing on the line”. –– Hon. Samuel Chepkonga (Ainabkoi), Chair of the Committee on Justice and Legal Affairs, Parliamentary Hansard, 2016.
Indeed, election commissions worldwide are often concerned with reducing voter wait times. Many commissions use a simple rule to manage polling centre queues: if a polling place has more than some predetermined number of voters, then additional staff, ballot boxes and voting queues are added at that polling place.
In the run-up to the 2017 election, the Kenyan legislature implemented a similar rule, capping the number of registered voters per polling place at 700. At polling centres with more than 700 registered voters, an additional queue was added, along with election officials, ballot boxes, and voting materials. This cut line lengths in half for polling centers with just over 700 registered voters, relative to those with just under 700 registered voters.
How Does Queueing Affect Voting?
Election administration decisions — like this 700-voter threshold for shortening queues — are important because they can affect voter behavior. Studying the effect of queuing on election outcomes is challenging, since lines may form for any number of reasons, and those reasons may also relate to turnout or vote choice. For example, local politicians may be particularly good (or bad) at mobilising voters in some areas. In areas with good mobilizers, we might see an increase in turnout, as well as an increase in lines. Naively, this might suggest that lines actually increase turnout. Moreover, the polling centre itself may be well (or poorly) organised or located, leading to idiosyncratic differences in queues as well as turnout. In order to account for such concerns, I use the Kenyan case to make a targeted comparison of polling centres just below and just above the 700-voter threshold.
This approach, called a regression discontinuity design, provides a more credible estimate of the causal impact of election-day queues on voter behavior. Polling centres with a total number of registered voters just below (e.g. 698) and just above (e.g. 702) the threshold likely share attributes (such as gender or ethnic composition, median age, poverty, or population density) that could also influence voter turnout. I compare election outcomes at polling centres with one long line (and a lengthier wait time) just below the threshold to polling centres just above the threshold with two shorter lines (and shorter wait times). This comparison shows that, among polling centres around the 700-voter threshold, moving from one queue to two increases turnout by 2.4%.
Another implication of queues is that election results themselves will be delayed if the presiding officer must finish processing voters prior to counting the votes. In fact, Kenyan law anticipates this reality, ensuring that citizens in line to vote when a polling station closes at 5 pm will be allowed to cast a ballot. Using the same discontinuity approach, I show that the time between polling centre closing and reporting of results is over 1.5 hours later, on average, in single-queue centres relative to dual-queue centers.
Threshold Rule, Partisan Impacts
Surprisingly, polling centres just above the cut-off had 6.7% higher vote share for incumbent president Uhuru Kenyatta. Waiting time is unlikely to influence a voter’s vote choice given that voters are unlikely to think “What a short wait! I’ll vote for Kenyatta.” Rather, the density and number of polling centres in Kenyatta-supporting areas meant those areas had more dual-queue centres relative to other areas. As a result, Kenyatta’s supporters were the primary beneficiaries of shorter queues.
Election administration is important precisely because it can affect voter behavior; thus, non-partisan election administration is a very important tool in maintaining the strength of democratic elections. Qualitative data does not support the contention that the choice of a 700-voter threshold was the result of partisan manipulation by the election commission or Parliament. Neither interviews with election officials nor committee and parliamentary debates suggest that the threshold was intended to differentially benefit the incumbent. In fact, the threshold was supported by incumbent and opposition MPs as well as civil society groups.
Threshold rules are common. I found similar results in Zimbabwe where, using a threshold of 1,000 registered voters, polling centres just above the threshold recorded a 1.1% higher voter turnout. This effect, in contrast to the Kenyan results, worked against the incumbent. The implications also reach beyond these specific elections as many other countries use similar rules, including Nigeria, Malawi, Colombia, and Mexico.
Election Administration Matters
What steps can be taken to minimise the unintended consequences of the threshold rule? First, if financially feasible, election commissions could expand the number and locations of polling centres. This would have the dual benefit of shortening queues on average, and decreasing the distance citizens must travel to cast a ballot. Evidence from South Africa shows that this approach increases turnout, but may also profoundly shape the composition of the electorate. Second, election commissions could evaluate whether polling centres in different areas should have different thresholds to minimise the time needed for voters to cast their ballot, though this approach could be politically volatile. Third, further lowering the threshold might ameliorate some of the differential benefits, though at significant administrative cost.
A renewed focus on the study of election administration could empower scholars, NGOs, and governments to improve election integrity. To make country-specific recommendations about election administration, however, we need publicly-accessible data. Open election data is valuable not just to ensure accountability of leaders and quality of results, but also to evaluate and improve the electoral process. In addition, regular and systematic data on election-day voter experiences (like South Africa’s Election Satisfaction Survey) is essential for monitoring and improving the electoral process.
J. Andrew Harris is an Assistant Professor of Political Science, New York University – Abu Dhabi.