Why Political Moderates Look Like the Enemy to Both Sides

Eleven policies on abortion, lined up in a row. At one end, never allowed. At the other, always allowed, with the rest spaced out at four-week intervals in between. When 579 Americans were asked to slide each of those policies onto a scale running from definitely pro-life to definitely pro-choice, the two ends of the list caused no trouble whatsoever; it was the middle, the stretch most of us would call a compromise, where the ratings came apart.

That is the first of five preregistered online survey experiments by Giulia Maimone at the University of Florida and Craig McKenzie at UC San Diego, published in the Journal of Experimental Psychology: General. Across 3,272 US participants recruited through online research panels, the two researchers kept turning up the same pattern: people with a stake in an issue do not read political moderates as moderate. They file them with the other guys.

Working out what counts as moderate was the first job, and the authors let their participants settle it. Those who described themselves as neither pro-life nor pro-choice rated the policies allowing abortion up to 4, 8 and 12 weeks as sitting in the middle, statistically indistinguishable from dead center. Pro-life participants looked at those same three policies and rated them pro-choice, the 12-week one at 1.66 on a scale where 3 means definitely pro-choice. Pro-choice participants looked at the identical policies and rated them pro-life.

Both groups can’t be right. Each camp was reading a middling position as a flag planted on the far side of the field, and the more strongly a participant held their own view, the further away the flag appeared. Maimone and McKenzie call this the “moderate as out-group” effect.

Five Days Before the Vote

The pattern is not confined to abortion, though it does not behave identically everywhere. On 31 October 2024, with the presidential election five days out, 810 participants were shown a set of hypothetical people, each expressing a different degree of support for Donald Trump or Kamala Harris, and asked to place each one on a conservative to liberal scale. Liberal participants read the person who supported both candidates equally as conservative (a mean of 0.21 toward the conservative end of a scale that runs to 2 in either direction, small in absolute terms but consistent). Conservative participants tilted the other way, reading the same person as slightly liberal, but that result did not reach statistical significance, landing at p = .074.

Maimone frames the stakes in institutional terms. “Democratic systems rely heavily on compromise, negotiation and coalition building, and moderates are often seen as the ‘glue’ that helps bridge ideological divides,” she said. “If moderates are rejected by both sides, compromise itself may begin to look suspicious or disloyal.”

Not Loving Your Side, or Not Hating Theirs

The fourth study is where the mechanism gets pinned down, and it’s the one place in the paper where randomized assignment does the heavy lifting. Maimone and McKenzie recruited 796 liberal participants and split them into two groups by lottery: one group judged people expressing positions on Trump, the other judged people expressing positions on Harris. A liberal shown someone with mixed opinions about Trump read that person as conservative, firmly, at 0.50 on the same scale. A liberal shown someone with mixed opinions about Harris read that person as, well, moderate, at 0.03 and not distinguishable from the midpoint. The asymmetry is the whole point: what marks you out is not failing to love your own side, it is failing to condemn the other one, which the authors read as the absence of an out-group hate motive rather than an absence of in-group love.

The last of the five, run in November 2023 on positions about the Israel and Palestine conflict, tightened the screw and then offered a way out of it. Participants judged a hypothetical person, Jamie, holding a moderate position framed either as finding both sides acceptable or as finding neither acceptable. The pro-both version produced the effect: pro-Israel participants read Jamie as pro-Palestine, pro-Palestine participants read Jamie as pro-Israel. Under the pro-neither framing, the gap between the two camps closed and the difference stopped being significant. Same middle ground, opposite reception, depending entirely on whether the moderate was heard to be agreeing with everyone or refusing everyone.

Worth being careful about what all this does and does not show. The APA release announcing the work is headlined on moderates being perceived as disloyal, and none of the five studies measured disloyalty, distrust, dislike or betrayal. What they measured was categorization: which box a person gets put in. Whether being filed with the enemy also means being resented is a reasonable guess, and it is a guess.

The samples deserve a squint too. These are online convenience panels rather than anything resembling the electorate, the exclusions are substantial (the last study analyzed 475 participants of the 764 who finished, dropping 289 for unfamiliarity with the conflict or for rating the scale inconsistently), and the authors are explicit that several of the supporting analyses, including the one establishing that moralization and threat are both needed, were exploratory rather than preregistered. Their own constraints section restricts the findings to US participants and US issues; the theory might travel, but the particular issues that carry moral weight and threat will not.

Still, the direction of travel is intriguing. If the middle reads as enemy territory to everyone, then the people best placed to broker anything are the people nobody trusts to broker it, and the incentive for a moderate is to plump for a side and stop being a moderate. “Our findings suggest polarization can become self-perpetuating because distrust spreads not only toward extremists on opposing sides but also toward people attempting to bridge divides in the middle,” Maimone said. The authors reckon the same logic might sit underneath the hostile media effect, where partisans on both sides see identical coverage as slanted against them. Balanced coverage, on this account, is not read as balanced; it is read as receptive to the wrong people.

Which leaves the one finding with an obvious handle on it. Maimone has suggested that reducing moralized language and not demonizing opponents could make cooperation feel more socially acceptable, though no study here tested an intervention of any kind; what that last study did test is narrower and sharper, and it’s about phrasing rather than sentiment. Say you agree with both sides and you get sorted into the other camp. Say you agree with neither, and, for now at least, you are allowed to stay in the middle.

Frequently Asked Questions