Where Senators Sit Can Shape How They Vote

With US politics so deeply polarized, it’s only natural to wonder why governing needs to be so combative. It turns out that changing one simple thing could help members of Congress collaborate: where they sit.

After analyzing more than 30 years of roll-call data and 1 million individual votes, Harvard Business School Professor Lauren H. Cohen finds that senators’ seating neighbors influence how they vote. In fact, sitting closer to members of the opposing party increases the odds of voting with them.

“Physical distance is the largest and most consistent of any characteristic outside of party or state in impacting voting behavior,” write Cohen and Peking University professor Bo Li in “The Micro-Geography of Persuasion: Randomized Peer Exposure and Legislative Outcomes.”

When senators reach across the aisle

After analyzing more than 30 congressional sessions, researchers estimate that the probability of a senator voting “yes” on a bill increases by 10 percentage points when all nearby senators switch from “no” to “yes.”

Here is an example of how personal interactions can influence senators’ voting behavior:

As part of their analysis, published as a working paper by the National Bureau of Economic Research in March, Cohen and Li needed to isolate the effect of proximity from other influences, such as party and state affiliation. The Senate’s seating approach gave them an opportunity to test what happens when a senator is randomly assigned to a desk.

Senators choose their desks based on seniority–defined as length of continuous service. Since freshmen are usually elected on the same day, the Senate follows a hierarchy based on previous government experience, ranked by service as:

1. A senator

2. Vice president of the United States

3. A member of the US House of Representatives

4. A cabinet secretary

5. A governor

Senators without federal government experience are ranked by the population of their home state, according to the most recent census.

Based on this, Cohen and Li identified 35 senators who, at one point, were the most junior freshmen and, like in a “game of musical chairs,” couldn’t choose their desks and were assigned to the last seat left in the chamber.

The experiment revealed that when neighboring lawmakers switch from voting “no” to “yes” on a bill, the probability that the most junior senator also votes “yes” increases by almost 12 percentage points. On average, these senators—the ones who can’t choose their seats—voted “yes” 63% of the time, so this increase represents a 19% jump in "yes" votes.

Other findings of their analysis, which spanned from 1989 to 2022:

• Neighboring senators’ influence builds with interaction and decays with time apart. The impact is about 30% larger in a senator’s second year. After the neighbors move to a different desk, their effect lasts for about six years, fading to zero after 7-12 years.

• Extreme positions hold greater influence than moderate ones. Senators with more ideologically extreme voting records have about four times the impact on a freshman’s vote compared to moderate legislators.

• Minorities, immigration, and crime were among the most persuadable topics across party lines.

As a robustness check, researchers also examined the House of Representatives’ independent office lottery. They found that representatives with more opposing-party neighbors are more likely to cross party lines, with up to a 25% increase in voting against their own party on close votes.

Random seating and flipped bills

Using artificial intelligence, researchers also evaluated 3,400 hours of video to observe senator-to-senator interactions and how they voted afterward. They found that face-to-face exchanges have a “strong association with vote-alignment between senators.”

Senators spent 91% of their floor time engaging with members of their own party, and those interactions tended to be longer. However, after party affiliation and state of origin, distance was the largest “driver of observed one-on-one interactions.” Aisle senators make up 8% of the Senate floor, but account for 60% of the top cross-party interactions.

What if seating was random?

As part of their analysis, researchers conducted 100 simulations, randomly assigning desks in the Senate chamber and recalculating the voting record based on their estimated influence on senators’ predicted votes.

The authors further extended their research back to 1790. “Using over 230 years of Congressional voting, the paper finds the same robust patterns of exogenously placed neighboring senators persistently moving the views and voting patterns of senators,” notes Cohen.

The researchers highlight that the analysis shows how a relatively simple, low-cost institutional change can help reduce political divisions.