Perceptible climate warming amplifies how education increases climate concern in the US

Perceptible climate warming amplifies how education increases climate concern in the US

Introduction

With a majority of Americans now believing that climate change is caused by human activities, two prominent predictors of concern in the US are the level of education1,2,3 and perceptible climate change that people experience4,4,5,6,7,8,9,10,11. This concern is moderated by political affiliations12,13,14, economic situations15, value and belief systems2,14,16,17,18,19. In the US, where the two main political parties have quite different climate mitigation platforms, climate concern strongly predicts political affiliation20.

In contrast to these factors that correlate with climate concern, education arguably serves as the more causal factor, by equipping people to understand climate science as a means of explaining observable change16. In advanced economies, people with more education are more likely “to say they are willing to adjust their lifestyles in response to the impact of climate change”21.

Here, we propose that the effect of education should increase with local climate warming. Under this hypothesis, the concern increases due to observable climate-related events9,10,11 combined with education5. Education acts like a lever, in that the higher the average education level in a population, the more climate concern will increase in response to local climate change.

We address this hypothesis by combining data, at the US county level, on climate concern, education level, and health risks of increased heat, together with a measure of climate warming in the state over the last decade.

Data for climate concern use the 2023 county-level estimates from the Yale Program on Climate Change Communication (YPCCC). Since the YPCCC climate opinion data, consisting of 73 related opinion estimates, are highly correlated, we reduced the dimensions through principal component analysis. This yields one major principal component, PC 1, which captures 40% of the variance in the county-level data of YPCCC 2023, with loadings that are clearly indicative of concern about climate change (see Supplementary Table S2). We refer to the PC 1 score as the “climate concern” estimated in each county.

For education levels, we use US Census 5-year (2018-2022, county-level) estimates of percentage bachelor’s degrees for adults 25 and over (see Methods and Data Availability). We also test for an effect at other levels of educational attainment, from 9th grade to graduate degrees.

As a measure of warming, we use the aggregated statistic of mean temperature increase, which serves as a proxy for the more episodic, tangible change that people experience21. A rising mean temperature leads, for example, to more frequent hot summer days6, increased wildfires, flooding, and storm activity22,23, along with prolonged droughts and heatwaves24, and shifting seasons11,25. In Wisconsin, an increase by less than 2 C since 1950 has brought more frequent days above 30 C, heat indices exceeding 40 C, and lakes freezing later in winter. Florida has seen increased hurricane activity, tidal flooding and severe heavy rainfall events.

As a parsimonious proxy for perceptible recent warming, we define “state-level warming” as a fractional measure that compares the past decade of warming in the state to a baseline average. For each state, we use the average the past ten years (2014–2023) of temperature anomaly determinations (in Celsius) determined by NOAA26, Ta, which we then divide by (bar{T}), the mean annual baseline (from 1971 to 2000) temperature (C) throughout the state. Since this fraction, (W={T}_{a}/bar{T}), varies geometrically—for a 1 °C increase, it would be 0.2 in Maine (annual ave 5 C) but only 0.05 in Florida (ave 21.5 C)—we take its logarithm, (ln (W)), as human intuition of such statistics tends to be logarithmic27. As a comparison, we also add the Heat and Health Index (HHI) from the US Center for Disease Control and Prevention to capture “communities where people are most likely to feel the effects of heat on their health”28.

Results

First, we confirmed the expected correlation among US counties between education level (% bachelor’s degrees for adults 25 and over) and climate concern (PC 1 from YPCCC 2023), among US counties. Among the 3048 US counties for which data exist on both variables, the regression slope is 0.0390 ± 0.0009 (adjusted r2 = 0.389, F1,3048 = 1944, p < 0.000001). The Pearson correlation is r = 0.624; we label this correlation as the “education effect”, as a proxy for the degree to which concern about climate change increases with education level. This education effect increases modestly with the age of the cohort (Table 1 and Supplement Table S1).

Table 1 Correlations, among 3048 counties, between bachelors degree % in 2022 and climate concern (PC 1 from YPCCC 2023)
Full size table

Second, we confirm that climate concern (PC1) is well-correlated with politics in the US. Across the 3048 US counties, the correlation between PC1 and percent Democratic vote in the 2020 election (Supplementary Figure S1) is so strong (Pearson’s r = 0.91) that we approach party vote and climate concern as very nearly proxies for each other. Although the residual climate concern after regressing against Democratic vote is relatively small, we test effects on residual climate concern, below.

Next, we determine the education effect in each of 43 US states as the correlation between education level and climate concern, among the counties of each state (seven US states do not have enough counties to derive this statistic: AK, CT, DE, HI, RI, DE). As US states differ considerably in climate, we can compare the state-level education effects to mean annual temperature in the respective states. Figure 1a shows that the education effect decreases with mean annual temperature (slope = − 0.0219 ± 0.0074, F1,43 = 8.8, adjusted r2 = 0.1504, p = 0.0049). We propose that education effect decreases with mean annual temperature because recent climate warming has been more perceptible in colder climates of the US. Consistent with this, state-level warming, a relative measure, predicts education effect (Fig. 1b). For the linear model between state-level warming and education effect, the coefficient is 0.217 ± 0.060 (adjusted r2 = 0.211, F1,43 = 12.8, p < 0.001). Fitting a linear model between W and education effect yields a coefficient of 0.918 ± 0.30 (adjusted r2 = 0.159, F1,43 = 9.31, p < 0.004). These correlations indicate that education increases how state-level warming engenders climate concern. Re-calculating this for different levels of educational attainment (Table 2) suggests the effect manifests at high-school level and above.

Fig. 1: Education effect is stronger in cold states, where relative warming has been greater.
Perceptible climate warming amplifies how education increases climate concern in the US

a Education effect (see text) in 43 states versus mean annual temperature (Celsius) in the state. b Relationship between the education effect and the logarithm of the warming fraction measure, W, for the state. The circles for the 43 states are colored by the 2022 Cook Partisan Voter Index (Note values on x-axis are negative because they are logarithms).

Full size image
Table 2 Linear regressions, in 43 states, between the ‘education effect’ and log(W), when ‘education effect’ is calculated for different education levels
Full size table

Interaction analysis

To further examine the interaction between education and warming, we use mixed effects regression, where counties (level 1) are clustered under states (level 2) and the dependent variable is climate concern (PC1). The results of this model are shown in Table 3. The interaction between education and state-level warming is significant in the mixed effects model (Table 3). Figure 2a shows this significant interaction when grouped by education level, with state-level warming on the x-axis and average climate concern (PC 1) among the grouped counties on the y-axis. The Intraclass Correlation Coefficient of 0.766 indicates a high correlation between county climate concerns within states, with ± 0.339 average variation among counties within a state. Table 3 shows that both education and HHI affect climate concern on their own (both p < 0.001), but they do not significantly interact with each other (p = 0.436). This indicates that education moderates the effect of state-level warming on climate concern (p < 0.001) but it does not moderate the effect of HHI on climate concern (Fig. 2b).

Table 3 Mixed effects regression model, where the dependent variable is Climate Concern (PC1) and counties are clustered under states
Full size table
Fig. 2: Interaction between state level warming on climate concern and (a) percentage with a bachelor’s degree per county, and (b) CDC Heat and Health Index.
figure 2

c Residual climate concern (not predicted by 2020 Democratic vote) vs temperature anomaly. The overall regression slope is 0.00856 ± 0.00044 (adjusted r2 = 0.109, F1,3048 = 373, p < 0.000001) and is slightly steeper for republican-leaning (gray line) than democratic-leaning (black line) counties. d Education effect on residual climate concern vs (log W), with no significant relationship.

Full size image

While the interaction of state-level warming with education is clear, HHI does not interact with education to increase climate concern (Fig. 2). This is likely because HHI correlates with poverty28. Lower income tends to predict lower support for climate policies because concern about day-to-day finances is prioritized over the future of climate change1,2,3,15,29. Also, under the psychological burden of poverty30, the responsibility for dealing with global climate change policies tends to be attributed elsewhere15,31.

Other variables will certainly have an effect. The political party acts as a proxy for climate concern and many Republican-leaning states are in the warmer US South. Figure 1 shows the states colored by the 2022 Cook Partisan Voter Index (see Methods), and indeed there is a modest correlation at the state level between education effect and PVI (slope − 0.009 ± 0.003, adjusted r2 = 0.143, p = 0.006). As such, politics help explain the outliers in our analysis, North and South Dakota (Fig. 1b), where education effects are anomalously low given the state-level warming. In North Dakota, for example, a strong partisan divide over climate policy reflects the state’s economic reliance on the oil industry32. Without the Dakotas, the correlation between state-level warming and the education effect is considerably stronger (slope 0.266 ± 0.047, adjusted r2 = 0.429, F1,41 = 32.5, p < 0.000002).

As a further check on the effect of politics, we look at the effects on the residual climate concern left after regressing climate concern (PC1) is regressed against the percentage Democratic vote in the 2020 election. Across 3048 US counties, there is still a correlation between the percentage bachelor’s degrees and residual climate concern (Fig. 2c) but the magnitude of that correlation—education effect—is no longer correlated with the warming measure (Fig. 2d) and the Dakotas are no longer outliers.

Discussion

The education effect is correlated with recent state-level warming, but the residual education effect is not. Where people vote according to their concerns, this is expected—the more concerned about climate change, the more likely to vote for the Democratic platform. Liberals and moderates tend to show greater trust in science as their education levels rise, while conservatives may become more skeptical33. From 2013 to 2023, the fraction of Democrats who considered climate change a major threat to well-being increased from 58 to 78 percent while that concern among Republicans remained at 23 percent.20

We suggest education enables climate concern, whereas politics is more a reflection of concern. They also interact, in that people approach scientific information through their group values34, such that more educated conservatives might use their knowledge to critically dismiss climate science, while those with less education might be less engaged35,36.

In summary, we find that perceptible warming in the last decade increases the effect that education, but not health risk, has on climate concerns. Building on findings that concern about climate change is increased by observable climate change combined with education5, our finding identifies feedback, in that the effect of education is further increased by the warming itself. This suggests that education increases climate concern by enabling people to explain their own experiences of climate change—the more warming they notice, the greater the effect of education.

Methods

Data for climate concern use the 2023 county-level estimates from the Yale Program on Climate Change Communication (YPCCC). Since the YPCCC climate opinion data, consisting of 73 related opinion estimates, are highly correlated, we reduced the dimensions through principal component analysis. This yields one major principal component, PC 1, which captures 40% of the variance in the county-level data of YPCCC 2023, with loadings that are indicative of concern about climate change (see Supplementary Table S2). We refer to the PC 1 score as the “climate concern” estimated in each county.

County-level education levels are drawn from the US Census 5-year (2018-2022) American Community Survey estimates about educational attainment (ACS S1501; see Data Availability). We use percent with bachelor’s degrees to determine this education effect while testing it at other levels of educational attainment, from 9th grade to graduate degrees.

Related Articles

National commitments to Aichi Targets and their implications for monitoring the Kunming-Montreal Global Biodiversity Framework

The Convention on Biological Biodiversity (CBD) exists as a major multilateral environmental agreement to safeguard biodiversity and “live in harmony with nature”. To deliver it, strategies and frameworks are set out in regular agreements that are then implemented at the national scale. However, we are not on track to achieve overall goals, and frameworks so far have not been successful. This could be due to unambitious targets, low follow-through on commitments, or desired outcomes for nature not being achieved when action is taken. Here, we focus on national planning and reporting documents from a set of 30% of Parties to the CBD. We found that nearly half of the commitments mentioned in national planning documents did not appear in the Sixth National Reports and that further losses emerged due to measures reported as incomplete or ineffective. There were differences between commitments to each of the Aichi Targets, with more losses in high-profile and “institutionally challenging” Targets. Commitments from Parties in different Human Development Index categories had different outcomes among Targets, and Parties self-identifying as “megadiverse countries” had overall higher rates of reported success. Our results are important for informing the monitoring of commitment implementation in the Kunming-Montreal “global biodiversity package”.

Why do travelers discontinue using integrated ride-hailing platforms? The role of perceived value and perceived risk

Despite integrated ride-hailing platforms have provided many benefits to travelers, there are also various potential risks. This study aims to examine travelers’ discontinuance behavioral intention toward integrated ride-hailing platforms. The research framework was established by extending the theory of planned behavior (TPB) with perceived value and perceived risk. Perceived value was classified into utilitarian, hedonic, and social values, while perceived risk was classified into privacy, performance, security, and financial risks. Additionally, the factors of switch cost and personal innovativeness were included. An empirical analysis was carried out using partial least-squares structural equation modeling (PLS-SEM) based on a survey conducted in Nanjing, China. Furthermore, a multi-group analysis (MGA) was performed to examine behavioral differences across demographic variables. The findings suggest that discontinuous behavioral intention is influenced by subjective norms, perceived behavioral control, and attitude. Among them, perceived behavioral control shows the strongest impact (−0.190). Perceived value, including utilitarian, hedonic, and social dimensions, negatively influences discontinuance intention, whereas the four variables of risk perception positively affect discontinuance intention. Notably, social value, performance risk, and privacy risk act higher total effects on discontinuance intention. Switch cost is negatively associated with attitude (−0.222), and positively affects discontinuance intention (0.189). Personal innovativeness has positive and stronger effects on perceived value (0.237), negative effects on perceived risk (−0.174), and negative effects on discontinuance intention. Regarding MGA results, older travelers demonstrate a stronger impact of social value on perceived value, higher-income groups exhibit greater sensitivity to security risks, and frequent travelers prioritize utilitarian value.

Trust in scientists and their role in society across 68 countries

Science is crucial for evidence-based decision-making. Public trust in scientists can help decision makers act on the basis of the best available evidence, especially during crises. However, in recent years the epistemic authority of science has been challenged, causing concerns about low public trust in scientists. We interrogated these concerns with a preregistered 68-country survey of 71,922 respondents and found that in most countries, most people trust scientists and agree that scientists should engage more in society and policymaking. We found variations between and within countries, which we explain with individual- and country-level variables, including political orientation. While there is no widespread lack of trust in scientists, we cannot discount the concern that lack of trust in scientists by even a small minority may affect considerations of scientific evidence in policymaking. These findings have implications for scientists and policymakers seeking to maintain and increase trust in scientists.

Urban growth strategy in Greater Sydney leads to unintended social and environmental challenges

Cities have advanced in terms of economic and social status over the past five decades, improving the living conditions of hundreds of millions of people. However, population growth and urban expansion have put pressure on social and environmental conditions. This study examines urban policymakers’ perceptions about causal relationships in the urban system as revealed in urban planning reports. Here we analyzed 500 pages from published urban plans of Greater Sydney between 1968 and 2018 and coded the text into causal maps. The findings show that policymakers adopted a dominant urban development strategy over the past 50 years to pursue economic and public infrastructure growth. Over time, this growth strategy resulted in a number of social and environmental challenges that negatively impacted societal well-being. Although policymakers eventually recognized the seriousness of social and environmental challenges, they never attempted to fundamentally change the dominant growth strategy. Instead, policymakers sought to address the challenges (that is, symptoms) by responding to each issue piecemeal.

Responses

Your email address will not be published. Required fields are marked *