Zhi Hao, Author at Goody Feed

You probably already know that sample counts in Singapore's elections are incredibly accurate. In fact, they’re so spot-on that some people go straight to bed once all the sample counts are out. But are they really that accurate? Before that, let's look at what they are. What Are Sample Votes? First, what exactly are sample counts, and why do we have them? While the public has only seen sample counts since GE2015, they were actually conducted even before that—they just weren’t published. Initially, the purpose wasn’t to give an early result to the public. Instead, it was used as a quality check: if the sample count and the actual result differed greatly, something might be wrong. Let’s say after counting 100 votes, a fictional party called the Cats’ Action Party has 68% of the vote. If the final result ends up showing they only got 19%, that’s a red flag. Sample counts help prevent such mismatches. But once the internet made it easy for unofficial sources to leak or speculate on early results, the authorities decided in 2015 to release official sample counts to reduce misinformation and guesswork during counting. So how exactly are these sample counts done? What Are Sample Counts Done It’s actually quite simple. Each SMC or Group Representation Constituency (GRC) has several polling stations. The bigger the constituency, the more polling stations it has. From each station, 100 ballot papers are randomly picked and counted. The results are then weighted based on the number of voters at each station. All this is done in front of the candidates and counting agents to ensure transparency. In essence, it’s a statistically sound snapshot—like testing a small portion of a big soup to know the flavour. Why is it so accurate, though? Well, that’s where statistics come in. Statisticians use formulas to determine the margin of error, and in Singapore’s sample counts, the margin is typically around 4%, with a 95% confidence level. That means 95 out of every 100 sample counts will fall within 4% of the final result. If the sample count shows a party has 54%, the actual result is very likely to be between 50% and 58%. It’s a bit like TV viewership ratings before the digital age. Companies didn’t track every single viewer—they sampled a small group and used that data to estimate the total audience. Similarly, Singapore’s elections don’t need to sample every vote to give us a reliable early picture. However, it’s still not perfect. That 95% confidence level also means 5 out of 100 sample counts could fall outside that 4% range. And that’s exactly what happened in three SMCs during GE2020. GE2020 Sample Counts That Are Not Within 4% Margin of Error In Kebun Baru SMC, the sample count predicted 68% for the PAP, but the final result was 62.97%—a difference of 5.03%. In Pioneer SMC, the sample count was 66%, while the final result was 61.98%, a difference of 4.02%. In Punggol West SMC, the sample count showed 65%, but the actual result was 60.97%, differing by 4.03%. So yes, it can be wrong—but rarely. Interestingly, SMCs tend to show larger deviations compared to GRCs. This is likely because GRCs have more polling stations, which means more sample data and thus more accuracy. So if you’re watching the elections closely, here’s a tip: stay up late for the final results of closely contested SMCs or GRCs, especially if the sample count margin is small. In GE2020, for example, West Coast GRC had a sample count of 52% for the PAP, and Sengkang GRC had 53% for the Workers’ Party—both within the margin of error and both nail-biters. Finally, if you're curious about which GRCs or SMCs might see a close fight this time, you’ll want to check out this video: https://www.youtube.com/watch?v9CY-yXJvRKQ