Sample-based observation projections of disputed poll


Independent electoral watchdog, the Zimbabwe Election Support Network (Zesn), released its sample-based observation (SBO) projections of the presidential election, which were consistent with the official results announced by the Zimbabwe Election Commission (Zec).


Zesn though could not definitively confirm whether or not a presidential runoff election should be held. While the statement was embraced by some, it also torched a storm regarding the methodology used to arrive at the projected results.

The July 30 general elections were meant to signal a break from Zimbabwe’s past of disputed electoral outcomes. However, as the election results started trickling in, the stage was set for yet another disputed electoral outcome.

The opposition coalition, the MDC Alliance, has alleged vote rigging and vote manipulation in favour of the ruling Zanu PF party candidate.
All political contestants are entitled to their day in court and the MDC Alliance’s concerns should be heard and adjudicated by a court of law.
As allegations of vote manipulation and concerns over the transmission and management of the results persisted, opposition supporters took to the streets only to be brutally squashed by the military, resulting in dozens of people being injured and the death of six.
The unfortunate events of August 1 have certainly taken Zimbabwe backwards especially given the relative peace and tolerance that prevailed during the campaign period.
A lot of questions have been raised about the efficacy of the SBO methodology used to arrive at the projected results. SBO is an advanced and proven election monitoring methodology employed around the world that allows non-partisan citizen observers to systematically assess the quality of opening, voting, closing and counting processes.
Using this methodology, an SBO can accurately project voting results, serving as an independent check on the official election results.

SBOs can confirm the accuracy of official results, or expose manipulation during tabulation process if it occurs. SBOs use direct observations by trained observers and do not ask voters for opinions.
SBOs deploy observers to a representative sample of polling stations carefully selected by statisticians and use rapid reporting technology to transmit observation data quickly and completely to a central data center.

This enables observers to provide the public with more accurate and comprehensive information about the conduct of an election for the entire country. The findings are based on sound and systematic data (and not anecdotal information or opinions). SBOs, which are sometimes referred to as Parallel Voter Tabulation (PVT) or the quick count in some countries, are designed to collect two types of information which are qualitative and quantitative. Qualitative information is on the conduct of the election day process.

In this case, observers use a standardised form that allows them to examine the integrity of the voting and counting processes for the entire country. This informs the observer organisation about whether to have confidence in the vote results by evaluating the conduct of the election day process.
On the other hand, quantitative information is on the outcome of the elections (results) and turnout figures. This data is examined only after the information on the conduct has demonstrated that the voting and counting processes went well.
Zimbabwe is not the first country to have had an SBO conducted, in Africa this methodology has been used in countries such as Ivory Caost (2016), Malawi (2009 and 2014), Zambia (2008, 2011, 2015 and 2016), Ghana (2008, 2012 and 2016), Uganda (2011), and Nigeria (2011, 2012 and 2015).
In all these cases the SBO accurately projected the vote tabulation outcomes including presidential results.
SBOs can project voting results and voter turnout with a high level of precision (i.e., low margins of error). Sample-based election observation is grounded in statistics such as the law of large numbers. Deploying observers to all approximately 10 000 polling stations boosts confidence of the voters and electoral stakeholders that the process is subject to scrutiny.
However, limited funding affected the ability to deploy comprehensively at all polling stations. Thus the SBO, while complementing the larger observation efforts, focusses on a more manageable group of well-trained, representatively deployed observers that can transmit complete and accurate data on the electoral process quickly. A lot of queries were raised with regards to the sample size of 750 polling stations. First, the sample size for the SBO was similar to that of SBOs from other countries and for past elections in Zimbabwe. SBOs generally have sample sizes between 500 and 1 000.
Second, the true sample size is actually the number of registered voters at sampled polling stations not the number of polling stations — in this case the true sample size is 383 272 voters.
More importantly, the sample size is a determinant of the precision (how large is the margin of error), but it does not affect the quality of the SBO. The quality of the SBO estimates is based on how representative the 750 sampled polling stations and 382 272 sampled voters are of the entire country.
The tables provided at the end of Zesn’s statement (and repeated below) clearly show that the sample is truly representative of Zimbabwe. For every province the percent of sampled polling stations and registered voters at sampled polling stations closely matches the percentage for all polling stations and all registered voters.
For example, 14% of all polling stations are in Midlands (1 540 divided by 10 985) and 14% of sampled polling stations are in Midlands (105 divided by 750).
Similarly, 11,5% of registered voters are in Mashonaland West (655 133 divided by 5 695,706) and 11,5% of registered voters at sampled polling stations are in Mashonaland West (44 113 divided by 383 272). It is this distribution that gives confidence that the SBO findings are truly representative of all polling stations and all registered voters.
Much focus was given to the results table in the SBO statement and all the other elements of the observation included were largely disregarded.
It is important to note that Zesn did not endorse the results announced but simply stated that they were consistent with the SBO range. This provides confidence that the results were tabulated correctly. However, as stated, the SBO projections did not rule out the possibility of a runoff and indicated that the results could have fallen either way.
Further, SBO findings on the conduct of voting and counting as well as reports by Zesn’s long-term observers (LTOs) show both improvements, but continued shortcomings in Zimbabwe’s elections. Analysis of the results should not be conducted in isolation from the broader election day and pre-election context.
In this post-election period, it is imperative to maintain peace, and exercise restraint, and allow for resolution of electoral challenges through the channels provided for in the law.

The Zimbabwe Election Support Network (Zesn), formed in 2000, is a coalition of 31 non-governmental organisations formed to co-ordinate activities pertaining to elections.