[General Election Poll Verification - Gwangjin-eul] 5 out of 6 April Polls Accurately Predict Ko Min-jeong's Vote Share Within Margin of Error

All 6 Polls Show Oh Se-hoon's Vote Share Outside Margin of Error
All 6 April Polls Correctly Predicted the Winner
Oh Se-hoon's Vote Share Consistently Outside Margin of Error

Graphic by Jin-kyung Lee

[Asia Economy Reporter Jin-Young Hwang] Various opinion polls conducted to gauge local public sentiment ahead of the 21st general election have been mired in controversy. The core of the debate was that opinion polls failed to accurately capture public sentiment and instead distorted public opinion. Asia Economy is publishing a series verifying how accurate the opinion polls were, focusing on constituencies where many polls were conducted.

Seoul Gwangjin-eul, where former Blue House spokesperson for the Moon Jae-in administration, Go Min-jung of the Democratic Party, faced off against Oh Se-hoon of the United Future Party, was one of the fiercest battlegrounds in this general election.

The final score was Go Min-jung 54,210 votes (50.3%), Oh Se-hoon 51,464 votes (47.8%).

It was a nail-biting contest decided by a margin of 2,746 votes (2.5 percentage points).

Since March, a total of 17 opinion polls were conducted to gauge public sentiment in Gwangjin-eul, with six of those conducted after the official election campaign began on the 2nd.

Media	Polling Company	Survey Period	Go Min-jung	Oh Se-hoon
JoongAng Ilbo	Ipsos	7-8	48.1	41.3
SBS	Ipsos	5-8	47.7	43
Seoul Economy	Embrain Public	8	48.1	35.1
MBC	Korea Research	6-7	50.9	40.1
JTBC	Gallup Korea	6	49.6	39.4
Seoul Shinmun	Research & Research	3-4	45.7	37.7

Unit: %

All six opinion polls conducted during the official election campaign period predicted that candidate Go would win, and five of these polls accurately predicted Go’s vote share within the margin of error.

The MBC-Korea Research poll and the JTBC-Gallup Korea poll differed from Go’s actual vote share by only 0.6 percentage points and -0.7 percentage points, respectively. (For more details, refer to the National Election Survey Deliberation Commission website)

However, the difference from Oh’s actual vote share was 7.7 percentage points for the MBC-Korea Research poll and 8.4 percentage points for the JTBC-Gallup Korea poll, exceeding the margin of error.

On the other hand, none of the opinion polls predicted Oh’s vote share within the margin of error.

In particular, two polls?Seoul Economy-Embrain Public and Seoul Shinmun-Research & Research?showed differences from Oh’s vote share exceeding 10 percentage points.

All six polls underestimated Oh’s actual vote share.

Looking at the difference between poll support rates and actual vote shares, Go’s ranged from -4.6 to 0.6 percentage points, while Oh’s ranged from -12.7 to -4.8 percentage points.

The larger discrepancy between poll results and actual vote shares for the United Future Party candidate compared to the Democratic Party candidate was also observed in Jongno.

Analysis of seven opinion polls conducted among Jongno residents during the official campaign period showed that the difference between poll results and actual vote shares was -7.2 to 5.2 percentage points for Democratic candidate Lee Nak-yeon, while it was -13.2 to -2.8 percentage points for United Future Party candidate Hwang Kyo-ahn.

Among the seven polls, two showed Lee’s support rate higher than his actual vote share, whereas none showed Hwang’s support rate higher than his actual vote share. Four of the seven polls underestimated Hwang’s vote share by more than 10 percentage points.

The phenomenon observed in Jongno and Gwangjin-eul may be explained by undecided voters shifting to support Oh or, as the United Future Party claims, by the presence of “shy conservatives”?conservative supporters not captured by polls.

However, as in Jongno, this discrepancy in Gwangjin-eul is not interpreted as enough to overturn the election result.

JoongAng Ilbo	Ipsos	27-28	47.1	38.4
YTN	Realmeter	28-29	47.1	43.3
MBN MaeKyung	R&Search	23-25	44.3	43.9
Sisa Journal	Mono Research	23-24	43.3	39.3
Dong-A	Research & Research	17-18	43.2	40.7
Seokyung	Embrain	20	40.9	36.6
MBC	Korea Research	14-15	41.7	39.8
JoongAng Ilbo	Ipsos	10-11	44.5	36.8
KBS & Hankook Ilbo	Korea Research	12-14	43.3	32.3
NewsPim	Korea Information Research	2-3	38.6	48.2
Hankook Ilbo	Korea Research	1-2	35.9	38.5

Unit: %

Among the 11 opinion polls conducted in March, two accurately predicted the vote shares of candidates Go and Oh within the margin of error, respectively.

No poll accurately predicted both candidates’ vote shares within the margin of error.

The two polls conducted in early March showed Oh leading Go.

The NewsPim poll commissioned to Korea Information Research conducted on March 2-3 predicted Go Min-jung at 38.6% and Oh Se-hoon at 48.2%, indicating Oh’s lead beyond the margin of error (±3.1 percentage points).

The Hankook Ilbo poll commissioned to Korea Research conducted on March 1-2 showed Go Min-jung at 35.9% and Oh Se-hoon at 38.5%, indicating Oh’s lead within the margin of error (±4.4 percentage points).

Although the NewsPim-Korea Information Research poll got the outcome wrong, it produced the closest support rate to Oh’s actual vote share among the 17 polls conducted in March and April.

The difference between this poll’s Oh support rate (48.2%) and his actual vote share was only 0.5 percentage points.

Unlike other polls with sample sizes around 500, the NewsPim-Korea Information Research poll was designed with a sample size of 1,001 and increased the proportion of landline phone respondents to 41%.

Recent polls typically have a landline phone proportion of around 10%.

Increasing the landline phone proportion tends to better reflect the opinions of voters aged 60 and above, which raises support for conservative candidates and lowers support for progressive candidates.

The NewsPim-Korea Information Research poll showed the highest support rate for Oh and the second-lowest support rate for Go among the 17 Gwangjin-eul polls conducted in March and April.

This content was produced with the assistance of AI translation services.