Quick counts and the integrity of the 2024 election

Amid allegations of electoral fraud, data analyses point to the general reliability and accuracy of quick counts, and that the real specter the public needs to watch for in future elections is information disorder.

The informal quick count results released by a number of pollsters indicate a victory for Prabowo Subianto and Gibran Rakabuming Raka in the 2024 general election with around 57-58 percent of the vote.

The Centre for Strategic and International Studies-Cyrus Network (CSIS-CN) quick count on Feb. 14, which sampled 2,000 polling stations proportionally distributed across 38 provinces, found that the Prabowo-Gibran presidential pair had won 58.25 percent of the vote.

As this article was being written, however, the General Elections Commission (KPU) had reached a real count of 78.05 percent and Prabowo-Gibran had garnered 58.83 percent of valid votes.

The questions, then, are whether the official results of the simultaneous presidential and legislative elections will differ significantly from the quick count and how the future of Indonesia’s elections should be evaluated. On Feb. 29, the Safer Internet Lab (SAIL), a joint initiative involving the CSIS and Google, held a seminar titled “Multi-Party Collaboration to Maintain the Integrity of the 2024 Election Data”. During the seminar, the initiatives for monitoring the 2024 election were explained and civil society groups shared their experience in maintaining the integrity of election data and detecting problems in vote counting and election-related information disorder.

Overall, the 2024 election demonstrated three best practices that can be replicated in the future: the role of quick counts as a tool to control election results, the rise of civil society initiatives in monitoring the elections and the decline in the spread of hoaxes and information disorder.

Using a margin of error (MoE) for quick counts ranging between 0.6 percent and 1 percent depending on the sample size used, the likelihood of different results in the presidential and legislative elections is minimal, no more than 1 percent.

To assess the accuracy and precision of the quick count results, we can compare the quick count data and the KPU’s official results for the 2019 election. For instance, the difference between CSIS-CN’s quick count and the KPU’s election results stood at a mere 0.12 percent. Considering the average quick count results conducted by 12 pollsters at that time, the discrepancy was around 0.8 percent.

In the 2024 presidential election, the average MoE between the quick count results from 10 pollsters and the KPU’s tally was less than 1 percent. Therefore, I can predict that the quick count results pointing to a Prabowo-Gibran victory will be unlikely to deviate significantly from the KPU’s real count.

Quick counts, also known as parallel vote tabulation (PVT), were first used in the 1986 Philippine elections to detect fraud.

The Operation Quick Count launched by NAMFREL successfully indicated the electoral fraud in the vote counting by COMELEC, which declared Ferdinand Marcos Sr as the winner. In contrast, NAMFREL’s findings indicated victory for Corazon Aquino. This discrepancy led to the EDSA People Power Revolution that year.

In Indonesia, quick counts have been in use for the past 20 years, starting with the first direct presidential election in 2004. That means that the quick count process has been tested in five national elections. They have also been used to predict the results of hundreds of regional elections (Pilkada) since 2005.

With such extensive use, the accuracy of quick count results is hard to challenge.

In the 2024 election, CSIS and several other pollsters conducted quick counts to project the results. Any discrepancies in the vote count can be verified easily by comparing the C1 vote tabulation forms signed off by polling station workers and official KPU data.

Research on election integrity has progressed rapidly, along with the methodologies to detect electoral fraud.

In 2024, several initiatives were undertaken to monitor the elections, such as Jaga Suara (protect voices) and Kawal Pemilu (control elections). This year’s Jaga Suara data analysis, conducted from Feb. 15 to 20 and covered 355,000 polling stations (43.28 percent of all polling stations), indicated that despite some errors in tabulation, the irregularities were insignificant.

The errors had occurred due to the limitation of the optical character recognition (OCR) technology in scanning images into numbers, which the KPU has addressed.

Furthermore, these tabulation errors did not favor or disadvantage any presidential pair. For instance, the discrepancy in the tallies for Anies Baswedan-Muhaimin Iskandar was around 1 percent, for Prabowo-Gibran 0.4 percent and for Ganjar Pranowo-Mahfud MD 1.41 percent.

In general, I found that hoaxes and information disorder had declined this year compared with previous elections. During this year’s election, information disorder was unorganized and did not last long. This conclusion was arrived at after analyzing data from reports of hoax that had been verified as misinformation by the Communications and Information Ministry.

The ministry monitored election-related hoaxes from Oct. 17, 2023 to Feb. 28, 2024, a timeframe deemed crucial to observing the pattern of hoaxes spread during the election. The analysis covered the period from two days before the registration of presidential and vice presidential candidates until two weeks after election day.

The analysis categorized the data according to issues, time and the scope of national and local dissemination. It was generally found that during this period, most of the information disorder was election-related (33.98 percent).

This was followed by social issues such as education, health and environment (33.10 percent), political and government issues (21.54 percent), financial issues (7.01 percent) and then the Rohingya refugee issue (4.38 percent).

In the future, attempts in information disorder are expected to increase locally, although most information disorder circulates nationally. When the data is broken down, information disorder related to political/government issues and social issues are shown to spread more extensively at the local level. Meanwhile, election disinformation tends to occur nationally.

Looking at its pattern in the past, information disorder has tended to increase during a crucial political event, such as presidential and vice presidential nominations and the election debates in January. Information disorder is therefore expected to intensify before the KPU announces the official election results on March 20.

*** The writer is head of the Politics and Social Change Department at the Centre for Strategic and International Studies, Jakarta.

This article was published in thejakartapost.com with the title “Quick counts and the integrity of the 2024 election”. Click to read: https://www.thejakartapost.com/opinion/2024/03/05/quick-counts-and-the-integrity-of-the-2024-election.html.


Pakarti Centre Building
Jl. Tanah Abang 3 No. 23-27
Jakarta, 10160

© 2024 · Safer Internet Lab