Here’s Why Sample Counts Are So Accurate When it Comes to Polling

For the people who sleep through the elections night, you might not even know that the sample count exists.

For people who didn’t, they’re wondering if the sample votes for Sengkang GRC would be similar to the actual votes, and had to keep themselves awake because the final results came later than your food delivery.

So…

What Is Sample Count?

Reader Bao: Simi sample? Like supermarket sample can eat meh?

No, my dear Bao. This is just a count they do at the start using a random bundle of 100 papers from each polling station, then weighted according to the number of votes. Basically, it’s kind of like an “early” election result.

Why A Sample Count?

Reader Bao: Har? Got final count already, sample count for what? Later see the result and cannot sleep how?

You see, unlike our dear Bao who sleeps well no matter what, there are some who die die cannot sleep unless they know the election result. Some might even be so excited they start believing any fake election results or anyhowly coming up with election results and spreading them.

So to prevent people from doing that, there’s the sample count.

Actually, this is quite new. GE 2015 is the first time sample counts were released to all GRCs and SMCs. Before that, only the ELD can use the sample counts. Then in 2011, candidates, not Mr Bao, can ask them for the sample.

But since people start to spread more and more things on the internet that even POFMA is implemented, the gahmen thought might as well just release the sample count.

Not my words hor. Institute of Policy Studies senior research fellow Gillian Koh say one. That Singaporeans have become more “impatient for news” and “more networked”.

How Accurate Are The Sample Counts?

Reader Bao: Pui! Take 100 votes only? You think accurate meh?

Well, actually… Since we already have the final results, let’s compare and see see.

Here’s the sample count versus final count for some polling stations I anyhowly picked.

Jurong
Sample
Number of Electors: 131,058
PAP: 75%
RDU: 25%
Final
PAP: 74.62%
RDU: 25.38%

Marymount
Number of Electors: 23,431
Sample
PSP: 46%
PAP: 54%
Final
PSP: 44.96%
PAP: 55.04%

Pioneer
Number of Electors: 24,653
Sample
CHEANG PENG WAH: 2%
PSP: 32%
PAP: 66%
Final
CHEANG PENG WAH: 2.78%
PSP: 35.24%
PAP: 61.98%

West Coast
Number of Electors: 146,089
Sample
PAP: 52%
PSP: 48%
Final
PAP: 51.69%
PSP: 48.31%

As you can see, pretty accurate hor?

Which leads to the question of 2020:

Why Are Sample Counts So Accurate?

Reader Bao: Holy- What magic did you use?!

The magic is used… is called math. More accurately, statistics.

You see, assuming that you take the 100 votes randomly from the whole constituency, the sample results will have a confidence level of about 95%, plus or minus 4% points.

Follow us on Telegram for more informative & easy-to-read articles, or download the Goody Feed app for articles you can’t find on Facebook!

In language that normal people can actually understand, this means that 95% of the time the actual election results will be within 4% of the sample count.

So for example, the West Coast candidates will be ultra-nervous after sampling, cause the sample counts were 52% and 48% for the PAP and PSP respectively. This meant that the PAP final count could be anywhere from 48% to 56%, and PSP’s final count could be anywhere from 44% to 52%.

(But we now know the final result and it’s pretty close to the sample lah.)

Reader Bao: Hah! So this means that I only need to survey 100 people the next time I need some data!

Remember, this only works if you choose 100 randomly from the whole constituency. So for example, you have a project that needs you to survey people from Tampines.

Reader Bao: Yah, so I just stand at Tampines MRT and wait for 100 people right? Easy lah!

Wrong.

Reader Bao: Wait, what?

Cause if you do that, then that’s not random anymore, but becomes a selection of 100 people at the time slot you choose to stand at the MRT. The correct way to do a random sampling like this is to take all the Tampines residents, randomly pick 100 of them, then go survey them.

In the case of the sample count, that’s kind of what happened since everybody has already given their vote, which is why it works.

If for example, only the left side residents in West Coast voted, taking a sample count would only be indicative of the left side resident votes and not the entire West Coast. Which is why taking the sample count only works after all votes are in.

And yes, the sample count for Sengkang GRC is accurate. You’ve stayed up late for nothing.

And on a side note, politicians have been talking about NCMP (Non-Constituency Member of Parliament) in recent days. So, what’s an NCMP? Do you know that it’s just like an MP but the allowance is much lower? Watch this video to find out more: