Model Investigates the Widespread Underestimation of Public Support for Climate Policies

By Jillian Kunze

Overall support in the U.S. for public policies that address climate change is high — for instance, the Yale Program on Climate Change Communication reports that 66 percent of Americans support taxing fossil fuel companies. “This is the rare piece of good news on climate change,” Ekaterina Landgren of the University of Colorado Boulder said. “However, if you ask somebody about the level of support for climate policy, they’ll tell you it’s much lower.” A 2022 survey study found that people generally underestimate support for climate policy by more than 20 percent.

During a presentation in the Women in Network Science minisymposium session at the 2023 SIAM Conference on Applications of Dynamical Systems, which is taking place this week in Portland, Ore., Landgren described joint work with Matt Burgess (University of Colorado Boulder) and Joshua Garland (Arizona State University) to investigate the causes of this widespread misconception. She explored a social network model that depicts public perception of climate policy support, with a foundation in real opinion surveys.

Landgren introduced two terms borrowed from psychology early in her talk: false consensus and pluralistic ignorance. In this context, she took false consensus to mean that the minority opinion holders (who do not support climate policies) think that they are in the majority, and pluralistic ignorance to mean that the majority opinion holders (who do support climate policies) think that they are in the minority. “Both phenomena lead to the underestimation of the majority opinion,” Landgren said.

The 2022 survey asked respondents to estimate what percentage of other people are worried about climate change, as well as their support for four particular climate policies: carbon tax, renewable energy on public land, 100 percent renewable energy by 2035, and the Green New Deal. It found that while over 60 percent of Americans are worried about climate change, survey respondents on average estimated this number to be about 40 percent — a widespread misperception (see Figure 1). Qualitatively, there is a recurring bimodality with two peaks in the data across the different questions; the mean estimated support is also always lower that the true support.

Landgren also looked at the estimated support by political party and found that Republicans skewed more towards an estimation of zero percent support for climate policies. However, media consumption mattered — Democrats who watched Fox News also had a higher misperception of public support. The strongest correlation for underestimation turned out to be watching other conservative outlets besides Fox News. Underestimation became less severe for more neutral or liberal outlets, but looped around to a slightly greater level of misperception for the category “other liberal outlets.”

Based on these survey results, Landgren asked four main questions: Can a social network model explain the widespread underestimation of public support? What features of a network model will replicate this behavior? What can explain the bimodality in the perception distributions? And to what degree does media consumption affect the misperception of public opinion?

To investigate these questions, Landgren presented some modeling work that is currently in progress. “This is a little ad hoc and in flux, but we’re going to ground it in the data,” she said. She described a model for preferential attachment that includes \(m\) node connections, with each node assigned either the minority or majority property. The model assumes a particular homophily (i.e., clinginess of network nodes) for both in-group and out-group connections (see Figure 2).

In this model, a totally heterophilic network exhibits pluralistic ignorance but not false consensus. A homophilic network is the opposite, exhibiting false consensus but not pluralistic ignorance — e.g., the phenomenon of an echo chamber (see Figure 3). "Something that we've noticed is that pluralistic ignorance is achieved through the centrality of minority nodes,” Landgren said. “When it comes to opinions on climate, we don't live in a heterophilic network, but we might live in a world where minority opinions on climate are overrepresented among central nodes.”

This overrepresentation of minority opinions can be caused by media coverage. Data from the Media and Climate Change Observatory at the University of Colorado Boulder, measuring the television coverage of climate change in the U.S. over time, demonstrated that Fox News is a central node that potentially drives minority opinion. So what exactly happens when minority nodes are overrepresented among the central nodes of the network?

Landgren and her collaborators first tried just swapping the highest degree majority node with the lowest degree minority node. "This is the process by which we're going to make some changes to our homophilic network and see what happens,” Landgren said. They then began to represent more minority opinions among the central nodes, finding that the mean opinion estimate dropped drastically even for fairly small network changes. 

For a larger value of \(m\) (i.e., more network connections), qualitative trends begin to arise in the modeled histograms of public perception that are similar to the actual data (see Figure 4). The presence of both pluralistic ignorance and false consensus is also apparent. In addition, the histogram for opinion perception among Republicans from the survey data appears similar to the modeled histogram for the minority nodes (though it is important to remember that in the real world, some Republicans will actually support climate policies and some Democrats will not).

Overall, the homophily in the social network seems to give rise to pluralistic ignorance or false consensus, but not both at the same time. However, adding an oversampling of the minority opinion among the central nodes is able to simultaneously capture both phenomena. "The future work will involve both data and modeling,” Landgren said. “On the modeling front, I think something that could be really interesting is finding a plausible mechanism for generating the oversampling of minority opinion nodes among central nodes." She also intends to further investigate whether a social network model is indeed the right tool to model misperception of climate policy.

Jillian Kunze is the associate editor of SIAM News.