The Lee Jae-myung Era: A Map of People - National Assembly Ideology Map - 경향신문

Scores are estimates based on documents from the analysis period. Where intervals overlap, the order of members is uncertain. All figures are shown to three decimal places. A credible interval is the range that contains a member's true position with 94% probability under the Bayesian estimate. The dot marks the estimated mean and the line through it marks the interval. A narrow interval means the data were sufficient for a firm estimate. When two members' intervals overlap, we cannot say which of them is more progressive or more conservative.

Where does each of the 319 members sit on the ideology axis (progressive vs. conservative) and the partisanship axis (Democratic bloc vs. People Power bloc)? Ideology measures whether the content of a document is progressive or conservative in philosophical terms. Partisanship measures which party bloc a position aligns with, regardless of ideology. Switch between the two below, and compare the estimate from all sources combined with estimates from Assembly records or news coverage alone. Searching a name highlights that member's position and interval, and the search stays when you change tabs.

The chart below places ideology on the horizontal axis and partisanship on the vertical axis. The two are independent: criticizing the Lee government does not by itself score as ideologically conservative. A member who says the government's rightward industrial policy leans too far toward technology and neglects labor rights is criticizing it from a position further to the left than the government itself.

Each dot is one member (posterior mean). The horizontal and vertical lines are the 94% credible intervals on each axis. Hover to see the name, coordinates and intervals. Click the legend to show or hide a party.

How divided is the People Power Party over former president Yoon Suk-yeol? This axis scores each member's attitude toward Yoon as expressed in news coverage, from −3 (strongly hostile) to +3 (strongly favorable), and estimates it with a separate model. The party splits clearly between a critical side and a favorable side.

Each dot is one member. Only the members at the two extremes are labeled. A pink left-pointing triangle marks a critical direction and a red right-pointing triangle a favorable one. A gray circle means the 94% credible interval includes zero, so the direction cannot be determined. Horizontal lines are 94% credible intervals. The estimate covers People Power members. Of the party's 112 members, those with fewer than three news items that could be scored for attitude toward Yoon were left out of the model: Baek Jong-heon and Kim Ki-woong had two such items each, and Lee So-hee had none.

Among Democratic Party members, the standard deviation of attitude toward Lee Jae-myung was 0.009, far below the 0.849 found for People Power members' attitude toward Yoon Suk-yeol. Because the differences between individuals are so small, the chart below shows instead how the party's attitude toward President Lee moved month by month. The line is the average attitude score at each point in time. The bars show the share of negative assessments among news items in which lawmakers discussed Lee.

Monthly average of the attitude score toward President Lee (−3 strongly hostile to +3 strongly favorable) in news items where a Democratic Party member is the speaker (left axis). The number above each point is the count of valid attitude observations that month. The bars behind show the share of negative scores (−3 to −1) among them (right axis). Neutral (0) counts in the denominator; missing values are excluded. These are raw counts, not model output.

To track change over time, we split the same scored documents into weeks (109 in total) and kept the structure of the model above, but assumed that each member's latent position follows a random walk that drifts a little every week. Intercepts and slopes for each document type were fixed at the estimates from the static model, and six models were fitted: ideology and partisanship, each for all sources, Assembly records only, and news coverage only. Below you can switch the measure and source or search a member's name to compare party averages with individual trajectories. The shaded band around a selected member is a 95% approximate interval; read the movement of the line together with that uncertainty.

We compared party loyalty measured from floor votes with loyalty measured from bills and news coverage. The horizontal axis is the average share of each member's valid votes cast against the majority of their own party; further left means closer agreement with party colleagues. The vertical axis is partisanship estimated from the four document types, standardized within each party; higher means the member defends their own side more than the party average does. When reading the two together, keep "how far from the party average" separate from "in which direction".

The z-score takes the partisanship θ from the model and computes (θ − party mean) ÷ party standard deviation within each party. Signs are flipped for the Democratic Party and the Rebuilding Korea Party; People Power keeps the original sign. The vote measure is a raw count; the speech measure is the combined estimate standardized within the party.