The trajectory · cohesion

Polarization & political violence

Two measures of how divided, and how dangerous, American politics has become. The first is structural: how far apart the two parties sit in Congress, from the standard roll-call measure. The second is physical: how many threats against members of Congress the Capitol Police investigate each year. Both are now at or near modern records, and they have risen together.

How far apart the two parties vote in the U.S. House (distance between the party means, DW-NOMINATE), by Congress.

Threats against members of Congress investigated by the U.S. Capitol Police, per year.

The line, in events

The bars have names attached. A partial, dated record of political violence and attempted violence against officials in the modern era: across both parties, which is the point:

2011	Rep. Gabby Giffords shot in the head at a constituent event; six killed
2017	Rep. Steve Scalise critically wounded at the congressional baseball practice
2021	The January 6 attack on the Capitol (the year threats first crossed 9,000)
2022	Paul Pelosi attacked with a hammer in the Speaker's home
2024	Two assassination attempts on Donald Trump (Butler, PA and West Palm Beach, FL)
2025	Minnesota legislator Melissa Hortman and her husband assassinated; activist Charlie Kirk assassinated at a campus event

The attitudes are moving too. In a September 2025 University of Chicago survey, about 15% of Americans agreed that force is justified to make officials “do the right thing”; a January 2026 PRRI survey found roughly 1 in 5 agreeing that Americans “may have to resort to violence” to save the country. On the polarization side, the gap widened mostly from one direction: since the 1970s the Democratic party mean in the House barely moved, while the Republican mean moved several times further, a measured asymmetry, stated here as data, not blame.

Two different rulers, one caveat each. The polarization measure (DW-NOMINATE) scores how members vote, not what they believe, and is one scaling method among several: the trend is robust across methods, the exact numbers are method-specific. The threat counts depend on Capitol Police reporting, staffing, and how a “threat” is defined, so part of the ~16-fold rise reflects more and better tracking since the 2016 baseline year, not purely more danger; treat the shape of the trend as the story, not the precise multiple. The two panels use different clocks (one per two-year Congress, one per calendar year) and their rising together is suggestive, not proof that one causes the other. This is a measure of the system, not a verdict on either party.