Men deny more than they believe about climate change on Twitter (X)

The signs are vital in this model output as the opposite signs are indication of significant difference in comparative stances of climate change believer and denier groups in comparison to the referent category of neutral. The variables showing similar sign of coefficients may differ in magnitude but will move in the same direction i.e. the (log) odds will either increase (for positive sign) or decrease (for negative sign) but will not show a comparative difference of opinions. While the signs of coefficients of “seriousness of gas emissions”, “weather extremes”, “aggression” remains same, the other variables show difference in stances and these include the key variable of interest “gender” for this paper. Put simply, in the terms of log odds, the variable gender can be interpreted as follows. One unit increase in gender will reduce the log odds of being a climate believer vs neutral by 0.14 and one unit increase in gender will increase the log odds of being a denier vs neu- tral by 0.27. I have further taken exponential values of all the coefficients to calculate the associated rela- tive risk ratio for the interpretation from a different perspective. The relative risk ratio for a one-unit increase in the variable gender is .87 for being a believer vs. neutral whereas, the same goes as high as 1.31 for being a denier vs. neutral. Similarly, for other variables, we find that relative risk ratio for a unit increase in tempera- ture is 1.02 for believer and 0.96 for denier in comparison to neutral group. Likewise, these val-

ues for remaining variable are 1.17 and 0.43 (resource over consumption), 2.59 and 0.40 (human intervention), 1.73 and 1.05 (seriousness of gas emission), 0.74 and 0.73 (weather extremes), 0.95 and 0.60 (non-aggressive individuals). Thus, except weather extremes, we see moderate to high differences in relative risk of being in the believer and denial groups in the data set. This suggests the posts about weather extremes are produced almost equally by the cli- mate change denial and believer groups across the globe but they have fundamental differences about the cause of the same. We see a consistent trend of high risk ratio among believing cli- mate change groups for all the variables that are found positively related to climate change by the studies applying text analysis using the same data set [41]. The key concern of the paper is the gender effect in the expression on twitter about climate change. Is the similar gender affect detected by other scholars in case of climate change? Based on the review of other scientific studies, the findings on gender extend the findings of [35] that indicates the males take skeptical stance for climate change impact. In a recent empirical study by [42] in the USA, it is found that climate change concerns are affect gender though, eco- nomic outlook moderate the effect of gender. To extend these findings, in a relevant context, a global survey conducted by [43] asserts that both men and women are less concerned about climate change in wealthier countries than in poorer countries but that men’s concern will

decline more rapidly, creating a gender gap in climate concern. The study sample is not as wide as the twitter data set used in this study so probably the results are not in conformity with the results of this study in entirety but underscore the importance of gendered attitude in fight against climate change. Limitations The study is not without limitation. One of the most widely accepted critique of twitter data is the identity of tweeting individuals. We are not 100% certain about the correct information about a bot, or an individual on twitter. So, we cannot be sure if the gender data is correctly provided to the twitter by the users. However, given the large scale data context, we assume this would not create a big difference on results. Therefore, we may hope that the conclusions are generalizable. The data only contained binary gender. So, the results are limited to male and female only. The other limitation was not using the location data as it would generate a PLOS ONE Men deny climate change on Twitter PLOS ONE | https://doi.org/10.1371/journal.pone.0303007 February 13, 2025 6 / 9