Every week there are stories about the president's approval rating. Realmeter comes out on Monday, Gallup Korea on Friday, and NBS and the Korea Social Opinion Institute somewhere in between. Numbers released in the same week often differ by 10 percentage points. From one article alone you cannot tell whether approval moved or whether the pollster is simply different. So I built a page that draws a separate line for each pollster.
Starting from a Wikipedia table
The raw poll data is registered with the National Election Survey Deliberation Commission, but each poll has its own result-table file, which makes it hard to gather everything at once. English Wikipedia has a table listing every poll since the inauguration, one row per poll, and each row links to the registration data or the pollster's release. I downloaded that table and split it into fieldwork dates, sample, pollster, and approve and disapprove values. It holds 418 job-approval polls from 22 pollsters and 425 party-support polls.
The table is written by hand, so the format varied a little. Some rows had no sort date, and one row was missing a quotation mark. For these 10 rows I rebuilt the date by combining text such as ‘14–18 Sep’ in the date cell with the year from the section heading.
The 70% was not an approval rating
The first Gallup Korea row was 70% approval from the June 10-12, 2025 poll. But when I looked for the news reports, Gallup Korea's first job-approval number was 64%, from the fourth week of June. In the second week of June it had asked whether people expected him to do well, and the 70% was the answer to that. Numbers from different questions had been joined on one line. I removed that one row. I did not check the other rows one by one against the original polls.
The chart takes over the first screen
At first I put the table at the top and the chart below it. The numbers were accurate, but you could not see the trend on the first screen. So, like the map on the Beautiful Spaces of the World page, I spread the chart across the whole screen and laid a translucent panel on the left. The panel holds the list of pollsters and their latest values, and clicking a name adds or removes its line. The table and the full list of polls moved below.
Most of the reasons that pollsters disagree come down to method. A phone-interview poll and an ARS (automated phone) poll reach different kinds of respondents. So the page shows the pollster average in large type, but draws a separate line for each pollster. A move within one pollster is the closest thing to real change.
Three weeks when Wikipedia stopped
After I published, I noticed that the data ended on September 30. Whoever fills in the Wikipedia table had not added new polls since October 2. So I went down to the original sources.
Party support was easy. The National Election Survey Deliberation Commission was piling up each week's registered polls on party support into a single Excel file. It even has the method and sample size, and that file is how I learned that Poll Kkot runs separate phone-interview and ARS polls on the same day. Wikipedia had listed only the phone-interview side.
Job approval is not in the Excel file. You have to open the registered result-table PDF for each poll, and the layout differs by pollster. Some tables start with ‘strongly approve’, and some list answers in order of how many people chose them. So instead of column names I looked at how the numbers relate. The approve total is the sum of two sub-columns, and so is the disapprove total, and approve, disapprove and don't know add up to 100. I find the three numbers that satisfy this, then read the total that comes first as approve. Checked against nine September polls that also appear on Wikipedia, every one matched.
Realmeter registers only its two-day party poll with the Commission. Its weekly approval survey, which asks 2,500 people over five days, is posted only on its own site, so I read it from the posts there. Poll Kkot's result tables have broken font data, so the Korean text could not be read. For that reason Poll Kkot's job-approval figures are still stuck in September.