🔬 physics

Can homophily explain public underestimation of climate policy support?

This paper uses computational modeling to demonstrate that while homophily alone cannot fully explain why both Republicans and Democrats underestimate public support for climate policies, the combination of media bias with realistic, symmetric homophily successfully reproduces the observed patterns of misperception.

Original authors: Ekaterina Landgren, Shriya Nagpal, Joshua Garland, Yaw Acquah, Matthew G. Burgess

Published 2026-06-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ekaterina Landgren, Shriya Nagpal, Joshua Garland, Yaw Acquah, Matthew G. Burgess

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Mystery: Why Do We Think Everyone Disagrees?

Imagine a town where 66 out of 100 people actually support building a new community park. However, when you ask people what they think the town thinks, almost everyone guesses that only about 37 people support it.

This is exactly what happens with climate change policies in the U.S. Most Americans support them, yet both Republicans and Democrats think the other side is much bigger than it really is. They are suffering from "pluralistic ignorance": they think their opinion is in the minority, even though it's actually the majority.

The authors of this paper wanted to solve a puzzle: Why does this happen? Specifically, they tested if "homophily" (the tendency for birds of a feather to flock together) is the culprit.

The Main Character: Homophily (The "Echo Chamber" Effect)

Think of homophily like choosing who to sit with at lunch.

Supporters of climate policy tend to sit with other supporters.
Opponents tend to sit with other opponents.

If you only talk to people who agree with you, you get a distorted view of the world. It's like looking at a room full of red balloons and thinking the whole world is red.

The researchers asked: Does this "lunch table" effect explain why everyone underestimates support for climate policy?

The Experiment: Two Digital Towns

To test this, the authors built two different types of digital "towns" (computer models) to see how people form opinions based on who they talk to.

The "Popular Kids" Town (Preferential Attachment): In this town, new people are more likely to become friends with the most popular people (those with the most friends). This creates a few very famous influencers and many regular people.
The "Club" Town (Stochastic Block Model): In this town, people are sorted into strict clubs. You are either in the "Support Club" or the "Oppose Club." You mostly talk to people in your own club, but there is a set chance you might talk to someone from the other club.

The Findings: Why "Lunch Tables" Aren't Enough

The researchers ran their models to see if homophily alone could create the massive underestimation seen in real life. Here is what they found:

1. The "Opponent" Problem:
If climate policy opponents only hang out with other opponents, they will correctly guess that support is low within their group. This makes them underestimate the total support. This part works.

2. The "Supporter" Problem:
If climate policy supporters hang out with other supporters, they should see lots of support and overestimate the total numbers.

The Catch: For the model to match reality (where supporters also underestimate support), the supporters would have to be weirdly un-homophilous. They would have to prefer hanging out with the opponents more than their own kind.
The Reality Check: The authors say this is unrealistic. Supporters don't usually seek out opponents; they usually stick with their own group.

Conclusion so far: Homophily alone cannot explain why everyone underestimates support. It only explains why opponents are wrong, unless we assume supporters are acting strangely.

The Twist: Adding "Brain Glitches" and "Biased News"

Since simple homophily didn't work, the authors added two extra ingredients to their digital towns to see if that fixed the puzzle.

Ingredient A: The "Brain Glitch" (Bayesian Rescaling)

Sometimes, our brains are bad at guessing percentages. If we are unsure, we tend to guess that things are closer to 50/50 than they really are.

The Analogy: Imagine you are guessing the ratio of red to blue marbles in a jar. Even if you see mostly red, your brain might say, "Well, it's probably closer to half-and-half just to be safe."
The Result: Adding this "brain glitch" helped a little, but it still required the supporters to be weirdly un-homophilous (hanging out with opponents) to get the numbers right.

Ingredient B: The "Biased News" (Media Bias)

This is where the model finally matched reality. The authors simulated a scenario where the "influencers" or "news outlets" in the town were biased.

The Analogy: Imagine a town where the loudspeakers (media) mostly play the voices of the minority group (opponents), making them seem much more common and influential than they actually are.
The Result: When they combined realistic homophily (people sticking with their own kind) with media bias (amplifying the opposing view), the model produced the exact pattern seen in real life: both supporters and opponents underestimated the true level of support.

The Final Verdict

The paper concludes that homophily alone is not the whole story.

If people just stick with their own groups, supporters should feel confident, not doubtful.
For everyone to feel like they are in the minority, something else must be amplifying the opposing view.
The most likely culprit in their models is media bias (or a similar mechanism) that makes the opposing opinion seem louder and more central than it really is.

Important Caveat: The authors note that while their model needs media bias to work, real-world evidence on whether media actually biases coverage of climate policy is "mixed." Some studies say yes, others say no. So, while the model shows how it could happen, the paper doesn't definitively prove that media bias is the only reason this happens in the real world. It just shows that homophily isn't enough on its own.

Technical Summary: Can Homophily Explain Public Underestimation of Climate Policy Support?

Problem Statement
Public opinion on climate change mitigation in the United States is characterized by a significant misperception: while a large majority (often two-thirds or more) supports climate policies, both supporters and opponents systematically underestimate this level of support. This phenomenon, a form of pluralistic ignorance, is particularly pronounced among Republicans and those with high exposure to right-leaning media. The central puzzle addressed by this paper is identifying the mechanisms driving this underestimation. Specifically, the authors investigate whether homophily—the tendency for individuals to associate with like-minded others—alone can explain why policy opponents underestimate support more than supporters, and why supporters also underestimate support despite holding the majority view.

Methodology
The authors employ analytical and simulation-based agent models to assess how social network structure and cognitive processing interact to produce misperceptions. They utilize two distinct network models:

Preferential Attachment Model: Agents connect based on existing degree (rich-get-richer), incorporating tunable homophily parameters ( $h_{ss}$ for supporters, $h_{oo}$ for opponents) to regulate the probability of within-group versus between-group connections.
Stochastic Block Model: Agents are assigned to blocks (supporters vs. opponents) with connection probabilities governed by within-block ( $H_{in}$ ) and cross-block ( $H_{out}$ ) parameters, creating modular communities.

In both models, agents estimate public support ( $\hat{f}_s$ ) by observing their own opinion and the opinions of their immediate neighbors. Misperception ( $\beta$ ) is defined as the difference between the true population support ( $f_s$ ) and the individual's estimate.

To explore mechanisms beyond pure network structure, the authors extend the models with:

Bayesian Rescaling: A cognitive mechanism where agents adjust their perception of proportions based on prior beliefs ( $\delta$ ) and uncertainty ( $\gamma$ ), simulating the tendency to overestimate small proportions and underestimate large ones under uncertainty.
Media Bias Simulation: In the preferential attachment model, bias is simulated by artificially promoting minority (opposing) nodes to central, high-degree positions, effectively overrepresenting opposing views in the network's influence structure.

The models' outputs are compared against empirical data from Sparkman et al. [3], which surveyed 6,119 Americans regarding their estimates of support for a carbon tax and other climate policies.

Key Results

Homophily Alone: The analysis reveals that homophily alone cannot explain the observed empirical patterns unless the homophily is highly and unrealistically asymmetric.
- For opponents to underestimate support, they must exhibit strong homophily ( $h_{oo} > 0.5$ ).
- However, for supporters (the majority) to also underestimate support, the model requires them to have extremely low homophily ( $h_{ss} < 0.5$ ), meaning they must disproportionately associate with opponents.
- In the stochastic block model, this requirement is even more extreme, with the empirically observed misperception range covering only ~3% of the parameter space.
Bayesian Rescaling: Introducing inaccurate priors (e.g., a belief that the public is split 50:50) and Bayesian rescaling reduces the degree of asymmetry required in homophily to produce realistic misperception. However, even with these cognitive adjustments, explaining aggregate underestimation of >20% still necessitates unrealistic assumptions where supporters preferentially associate with opponents.
Media Bias: The authors find that combining realistic, symmetric homophily (where both groups prefer their own kind) with media bias (simulated as the overrepresentation of opposing views in central network positions) successfully reproduces the observed misperception patterns. In this scenario, both supporters and opponents underestimate public support, with opponents underestimating more.

Significance and Claims
The paper claims to provide theoretical foundations for understanding public opinion misperception on climate change and other polarized issues. Its primary contribution is delineating the specific conditions under which homophily can and cannot generate pluralistic ignorance.

The authors modestly conclude that their results do not resolve the puzzle of climate policy misperception definitively. Instead, they demonstrate that:

Homophily alone is insufficient to explain the data without invoking unrealistic behavioral assumptions (e.g., supporters actively seeking out opponents).
Realistic misperception patterns emerge most plausibly when homophily is combined with other mechanisms, such as media bias that overrepresents minority views or inaccurate priors.
The empirical evidence for the necessary conditions (asymmetric homophily or aggregate media bias against climate policy) is currently mixed or inconclusive.

The study serves as a cautionary theoretical framework, suggesting that future research must look beyond simple network segregation to include information environments and cognitive processing biases to fully explain why the public underestimates support for climate policies.