Even though the election has already been held, problems keep emerging. The most recent is the issue of trust among part of the public toward the quick count. Prabowo Subianto asked his supporters not to trust survey institutions and to keep focusing on the tally conducted by the KPU. Almost all quick counts from public survey institutions stated that Joko Widodo-Ma’ruf Amin were leading according to the quick count. However, according to Prabowo's camp, the results differed from the quick count carried out by his internal team. "I just want to give an update that based on the real count, we are already at the 62 percent position. This is the result of the real count," said Prabowo at a press conference on Jalan Kertanegara, Jakarta, on Wednesday night (17/4/2019). The Chairman of the Indonesian Association for Public Opinion Surveys (Persepi), Philips J. Vermonte, denied Prabowo Subianto's statement suggesting that survey institutions had produced a bogus quick count. Persepi held a press conference titled 'Expose Data Hasil Quick Count Pemilu 2019' at Morrissey, Jakarta, on Saturday (20/4/2019).
In this press conference, several survey institutions that are members of Persepi published the sampling methodology they used. Fundamentally, the quick count is the process of recording the vote acquisition results at Polling Stations (TPS) that are used as samples. To obtain the number and locations of the TPS that must be used as samples, several stages are required. There are several techniques commonly used in quick count sampling. Generally, they are a combination of several techniques such as stratified, cluster, and multistage. The main concept of stratified sampling is to group the population into strata or identical sub-populations. Therefore, the samples taken from each stratum can represent its population. The quick count population is the valid votes of voters spread across all TPS in Indonesia.
Referring to the quick count methods from the survey institutions we collected, SMRC and Indikator use the stratified systematic cluster random sampling method. The population is the valid votes from all TPS across Indonesia; they then group the TPS by the RI electoral district (Dapil) region and rural-urban status. After that, from the TPS sub-groups, samples are drawn randomly in proportional numbers so that the TPS to be used as samples are selected. Meanwhile, Indo Barometer also uses the stratified systematic random sampling technique. The population is first grouped into sub-populations with the same criteria or strata. After that, from the sub-population, samples will be drawn randomly and proportionally, following the sample size. This random drawing is only carried out on the first element of the selected sample, while subsequent elements are chosen systematically according to a certain pattern. The quick count results between one institution and another are usually different, but if the method is correct, the results will fall within a similar range of figures.
The quick count results of 6 survey institutions show the winning figure of the Jokowi-Ma’ruf pair at 54 to 55 percent, while Prabowo-Sandiaga at 44-45 percent. Looking at the track record, the quick counts released by credible survey institutions in the elections that have taken place in Indonesia did not deviate far from the KPU's real count. Does the KPU's Real Count Differ from the Quick Count? PKS politician Hidayat Nur Wahid, through his Twitter account @hnurwahid on 26 April 2019, questioned the accuracy of the survey institutions' quick count results in Bengkulu, which differed from the KPU's real count. “Prabowo-Sandi's victory according to the KPU Real Count in Bengkulu. Presidential candidate 02: 50.12%, and presidential candidate 01: 49.88%). That is Different in Principle & a Far-Off Figure from the Quick Count Results by the Survey Institution[s]: Presidential candidate 01 wins with 58.78%, presidential candidate 02 is only given: 41.22%. So?” The KPU's tally as of 29 April 2019 for the Bengkulu Province region was already 100 percent. The tally of the 6,165 TPS that came in shows the Prabowo-Sandiaga pair leading with a vote acquisition of 585,689 or 50.13 percent, while the Joko Widodo-Ma’ruf pair obtained 582,572 votes or 49.87 percent. Of the four survey institutions that provided estimates of the presidential election vote acquisition, the quick counts of Indikator and Poltracking in Bengkulu showed estimates that missed the mark. Indikator predicted Jokowi-Ma’ruf's vote acquisition at 52.61 percent, while the estimated Jokowi-Ma’ruf vote acquisition from the survey institution Poltracking was 58.87 percent. At a glance, the values are very different from the KPU's real count results. However, we need to look at the margin of error (MoE) and its sample. Compared to Central Java Province, which has 115,407 TPS, the number of samples in Bengkulu Province is very different. Sampling is done proportionally or following the population figure, so the smaller the population, the fewer samples taken. The margin of error figure also differs between one province and another. For quick counts at the national level, survey institutions on average set the margin of error value at around ±1%, while at the provincial level, the margin of error value varies because the distribution of sample TPS becomes smaller. In Central Java, the margin of error of the Indikator survey was ±1.2%, while in Bengkulu Province it was ±7.32%. The difference between the figure 52.61 percent (Indikator's quick count for Bengkulu) and 49.87 percent (the KPU's real count for Bengkulu) is still within the MoE range.
Finally, there is the matter of significance. SMRC, Indikator, and Poltracking conducted a significance test of the difference in vote acquisition between the two pairs in each province with a significance level of 0.05. If the significance value is less than 0.05 (p≤0.05), there is a possibility that the position obtained by a candidate pair at the time of the quick count result will not change. However, if the significance value is greater than 0.05, there is a high probability of a change in position. In short: if, according to the quick count in a province, Jokowi's or Prabowo's victory is significant, that is also the position that will occur in the real count. The explanation above is only part of the many things that need to be considered in a research result. Every credible survey institution will of course transparently display the characteristics of the respondents and the data processing and analysis methods used. Examine and consider the details above if you wish to assess the credibility of a quick count result.