Interactions between concerns for the environment and other sources of concern in 31 European countries

The analysis is based on the public opinion data on personal worries collected by the Eurobarometer between 2012 and 2019. We consider all the 31 countries for which complete data is available. Each European Barometer survey consisted of approximately 1000 face-to-face interviews per country. As we are interested in personal worries, we focus on the question 'Personally, what are the two most important issues you are facing at the moment? (max. 2 answers)'. The data is reported in terms of percentages of people who indicated a given worry.

The possible answers to the question were: (a) crime, (b) the economic situation, (c) rising prices/inflation/cost of living, (d) taxation, (e) unemployment, (f) terrorism, (g) housing, (h) the financial situation of your household, (i) immigration, (j) health and social security, (k) the education system, (l) the environment, climate and energy issues, (m) pensions, (n) working conditions, (o) living conditions, (p) defence/foreign affairs, (q) other, (r) none, and (s) do not know. We consider all possible answers that were included in the years considered (see SI). We aggregate all worries in five categories, and consider the answers other, none and do not know as a residual category (see table 2). To statistically support our grouping into categories we carried out a factor analysis both exploratory (to identify the hidden factors) and confirmatory (to validate the proposed clusterization). However, all the tests we carried out confirmed that the data matrix is not factorizable. For instance, the Kaiser–Meyer–Olkin value for the data aggregated at EU level is 0.393, which is much below the acceptable level (Watkins 2018). Similarly, the dataset fails also the Haitovsky multicollinearity test. This was not surprising as the data from the Eurobarometer gives us only aggregated data, making the number of observations lower than the parameters that need to be estimated.

Let $W_{i,j}(t)$ be the total number of respondents in the jth country that at time t indicated a worry included in the ith category, i.e.:

$$\begin{equation}W_{i,j}\left( t\right) = \displaystyle \sum \limits_{h_i = 1}^{n_i}w_{h_i,j}\left( t\right) ,\quad i = 0,{\ldots},5, \quad j = 1,{\ldots},31,\end{equation} \tag{ 1 }$$

where $w_{h_i,j}$ is the number of respondents indicating the worry h_i belonging to the ith category, and n_i is the number of worries of the ith category for the jth country.

Then, the shares $P_{i,j}(t)$ at time t for the categories environment (i = 1), safety (i = 2), economy (i = 3), immigration (i = 4), social issues (i = 5), and outside option (i = 0) for the jth country are determined as follows:

$$\begin{equation} P_{i,j}\left( t\right) = \frac{W_{i,j}\left( t\right)}{\displaystyle \sum \limits_{h = 0}^{5}W_{h,j}\left( t\right) },\quad i = 0,{\ldots},5,\quad j = 1,{\ldots},31. \end{equation} \tag{ 2 }$$

We identify the shares $P_{i,j}\left( t\right)$ with the probability of choosing the category i from all possible categories via the logit model, i.e.:

$$\begin{equation} P_{i,j}(t) = \dfrac{\exp (f_{i,j}\left( t\right) )}{\displaystyle \sum\limits_{h = 0}^{5}\exp (f_{h,j}\left( t\right) )},\quad i = 0,{\ldots},5,\quad j = 1,{\ldots},31, \end{equation} \tag{ 3 }$$

where $f_{i,j}\left( t\right)$ is the utility function for a respondent of jth country to choose a worry in the ith category at time t. In particular, each utility function $f_{i,j}\left( t\right)$ is defined as a (linear or nonlinear) function of all aspects and attributes impacting the choice among alternative worries. Furthermore, since the category Outside option (i = 0) plays the role of the outside good, then equation (3) becomes:

$$\begin{align} P_{i,j}(t)& = \dfrac{\exp (f_{i,j}\left( t\right) )}{1+\displaystyle\sum\limits_{h = 1}^{5}\exp (f_{h,j}\left( t\right) )},\quad\nonumber \\ &\quad i = 1,{\ldots},5,\quad j = 1,{\ldots},31\nonumber \\ P_{0,j}(t) &= \dfrac{1}{1+\displaystyle\sum\limits_{h = 1}^{5}\exp (f_{h,j}\left( t\right) )}, \end{align} \tag{ 4 }$$

where $P_{0,j}(t) = 1-\displaystyle \sum\limits_{i = 1}^{5}P_{i,j}(t)$ at any time t.

Assuming that all the utility functions $f_{i,j}\left( t\right)$ are of class $C^{2}\left( \left[ t_{0},+\infty \right) \right)$, it can be proved that equation (4)₁ are the unique (global) solution of the following Cauchy problem:

$$\begin{align} \left\{ \begin{array}{l} \displaystyle\frac{d {P}_{i,j}(t)}{dt} = g_{i,j}\left( t\right) P_{i,j}(t)\left[ 1-P_{i,j}(t)\right] -\displaystyle\sum\limits_{h = 1,h\neq i}^{5}g_{h,j}\left( t\right) P_{h,j}(t)P_{i,j}(t),\quad i = 1,{\ldots},5, \\ P_{i.j}(t_{0}) = \displaystyle \dfrac{\exp (f_{i,j}\left( t_{0}\right) )}{1+\displaystyle\sum \limits_{h = 1}^{5}\exp (f_{h,j}\left( t_{0}\right) )} \qquad\qquad\qquad\qquad\qquad\qquad\qquad t\in \left[ t_{0},+\infty \right), \end{array} \right. \end{align} \tag{ 5 }$$

where $g_{i,j} = df_{i,j}/dt$ and $j = 1,{\ldots},31$.

For each country, the share $P_{i.j}(t)$ of the ith category increases when its utility function $f_{i.j}(t)$ increases, whereas it decreases when the utility function $f_{h.j}(t)$ of any other category increases. Thus, owing to equation (5), the evolution of the share $P_{i.j}(t)$ of the ith category for the jth country is mathematically determined by the intrinsic growth rate function $g_{i.j}(t)$ and the competition functions $g_{h.j}(t)$ between the ith and hth categories. Then, at any time and for any given country, the competitive interactions between any pair of categories—and therefore the presence of AGT and FPW interactions, or both—are determined by the signs of the functions $g_{i.j}(t)$ according to table 3.

Furthermore, owing to table 3, except for when amensalism or commensalism may occur, the AGT and FPW interactions between all worry categories $P_{i,j}$ and a fixed category $P_{h,j}$, for all $i\ne h$, only depend on sign of the interaction coefficient $g_{h,j}$ (see SI).

To determine the utility functions—and hence the interactions coefficients—from the historical data of the categories of worry we first determine a discrete set of values for each of them as follows:

$$\begin{equation} f_{i,j}\left( t\right) = \ln P_{i,j}(t)-\ln P_{0,j}(t),\quad \forall i, j, \end{equation} \tag{ 6 }$$

then we use a Fourier series of order n to obtain an approximate analytical form of these functions.

In figures 1–3, as an example, we present the results of our model for Germany. The results for all the countries can be found in the appendix.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 1 right panel shows the simultaneous interactions among all sources of worries, whereas figure 2 highlights how the category environment interacts with each of the other categories.

In figure 3 we show AGT and FPW interactions of Environment versus Safety and Economy and vice versa for the Germany in the time interval 2012–2019. We assess the accuracy of our model using the mean square error (MSE) and we found that for all countries and categories, the order of magnitude of the MSE is between 10⁻⁶ and 10⁻⁴ (see appendix table 1).

We use the paired t-test to assess whether (a) the averages of the relative frequency distributions of AGT and FPW among the concerns differ; (b) the averages of the relative frequency distributions of AGT between any pairs of concerns, e.g. environment vs economy and environment vs safety, differ.

Let data₁ and data₂ be a paired samples of equal length n = 31. After verifying that the differences distribution $data_1-data_2$ forms a sample from a normal population, we test whether the mean of $data_1-data_2$ is zero using the Student paired t-test. In detail, we test the null hypothesis $H_0: \mu_{12} = \mu_0$ against the alternative hypothesis $H_0: \mu_{12} \ne \mu_0$, where µ₁₂ is the mean of the paired differences of the two data sets $data_1-data_2$ and $\mu_0 = 0$. The test statistic is assumed to follow a Student distribution, and the null hypothesis $H_ 0$ is rejected only if $p\lt\alpha$, where the significance level α is set to 0.05.

We find that the null hypothesis H₀ is rejected in all cases except for the averages of the relative frequency distributions of AGT and FPW for environment vs safety (see appendix table 2–4).

First, we investigate whether FPW and AGT can simultaneously coexist within a given pair of worries. In our framework, this condition is realized when two worries are in a predator–prey interaction, i.e. when the interaction coefficients have opposite signs. In fact—consistently with the FPW hypothesis—an increase in the worry-predator pushes down the worry-prey, while—consistently with the AGT—an increase in the worry-prey favors the growth of the worry-predator.

Figure 4 indicates how often predator–prey relationships emerge between the various categories of worries on average across all countries. We observe that predator–prey interactions represent on average about 36% of the total interactions across all countries in all time periods. Moreover, we note that for all pairs of worries there are at least some instances in which AGT and FPW simultaneously coexist.

Zoom In Zoom Out Reset image size

Second, we analyze whether the FPW hypothesis and the AGT alternate over time for a given pair of worries. To put it differently, we study if the kind of interaction among worries changes over time. As an example, in figure 5 we show how interactions consistent with FPW and AGT alternate over time in Denmark, Finland, France, Germany and United Kingdom. We observe that in all countries considered AGT and FPW interactions alternate for all pairs of worries. Third, figure 5 also shows that at a given moment there can be a pair of worries standing in an interaction that is consistent with one theory (e.g. AGT), while another pair stands in an interaction that is consistent with the other theory (e.g. FPW).

Zoom In Zoom Out Reset image size

These results suggests that AGT and FPW cannot be portrayed as mutually exclusive when there are multiple sources of worries interacting in people's mind, and the way in which these worries interact changes over time.

A large body of literature has investigated the relationship between the economy and the environment (Tiba and Omri 2017). For instance, an influential strand of literature emphasizes that—among other things—ecological limits place inescapable constrains on future economic growth, and therefore that countries should aim at managing economic degrowth (Kallis et al 2012). This literature reinforces the idea that there is a crucial relationship between the economy and the environment, which is why we focus our attention mostly on this relationship. However, fewer studies investigate how people perceive this relationship. And yet this question is extremely relevant. First, concerns for the environment influence private actions (Bouman et al 2020), which in turn can have a significant impact on climate change (Dietz et al 2009). One key problem is that the effect on pro-environmental behaviors of extrinsic incentives is generally short-lived (Van Der Linden 2015). Instead, if people internalize that being concerned about the environment and acting accordingly is the right thing to do, then pro-environmental behaviors are more likely to be sustained over time (Van Der Linden 2015). Second, it is harder to implement policies to protect the environment and mitigate climate change if people are not concerned about global warming or the environment in general. 'To put it differently, it is people who are the drivers of, are affected by, and have the capacity to respond to global change' (Weaver et al 2014).

Turning to studies investigating the relationship between concerns for the environment and the economy, Whitmarsh (2011) observed that between 2003 and 2008 the perceived severity of climate change sharply declined. She attributed this effect to the looming financial crisis, thus suggesting that increased concerns about the economic situation might have decreased the concerns about climate change. Similarly, Scruggs and Benegal (2012) found that short term economic concerns—and especially unemployment—have a strong chilling effect on climate concerns.

Given the importance of this relationship, and the limited number of studies on the issue, we start by analyzing the impact of changes in concerns for the economy on concerns for the environment. We find that across all countries FPW interactions emerge 60.6% of the time, whereas AGT interactions only emerge 39.4% of the time. Thus, FPW interactions are almost 54% more frequent than AGT interactions (figure 6, right panel). This difference is statistically significant ($t = -3.3977$, p = 0.0019).

Zoom In Zoom Out Reset image size

On the contrary, when focusing on the impact of changes in concerns for the environment on concerns for the economy we observe a prevalence of AGT interactions. On average, across the 31 countries considered AGT interactions emerge 61.7% of the times, whereas evidence for the FPW interactions emerge 38.3% of the times. Thus AGT interactions are approximately 61% more frequent (figure 6, left panel). This difference is statistically significant (paired t-test: t = 3.4392, p = 0.0017). We then turn to the single countries. We observe that when looking at the effect of concerns for the economy on concerns for the environment FPW interactions are more common in 21 countries (approximately 68%) (figure 6, right panel). Vice versa, in 27 out of 31 countries (approximately 87%) AGT interactions are more common than FPW interactions when considering the effect of concerns for the environment on concerns for the economy (figure 6, left panel).

Taken together, these results suggest that the AGT is predominant when focusing on the effect of concerns for the environment on concerns for the economy. Therefore, it seems that people have internalized the economic consequences of environmental issues, and consequently concerns for the environment often favor the growth of concerns for the economy. However, our results also suggest that an increase in concerns for the economy pushes down a less immediate concern like the one for the environment. This result is consistent with the findings of Whitmarsh (2011) and Scruggs and Benegal (2012).

Many studies have investigated the relationship between climate change and migratory dynamics. As the problems caused by climate change worsen, more people are displaced and migratory fluxes increase (Cattaneo et al 2020, Kaczan and Orgill-Meyer 2020). These dynamics suggest that AGT interactions should be predominant when analyzing the effect of concerns for the environment on concerns for immigration. However, we observe that AGT interactions from the environment to immigration are less common than FPW interactions (paired t-test: $t = -2.3917$, p = 0.0232). This pattern holds also when looking at single countries. Looking at the effect of concerns for the environment on concerns for immigration FPW interactions are more frequent in 24 countries (approximately 77%) (figure 7, center right panel).

Zoom In Zoom Out Reset image size

Thus, despite the connection between climate change and migratory dynamics identified by the literature (Cattaneo et al 2020, Kaczan and Orgill-Meyer 2020), people do not perceive the existence of a linked faith between these issues.

Moreover, many studies have identified a link between environmental factors and safety. In fact, there is robust evidence that warmer temperatures are associated with higher rates of offending and more police calls for service (McDowall et al 2012, Mares and Moffett 2019), and that warming global temperatures are associated with a variety of crime measures (Hsiang et al 2013). Thus, also in this case it would be reasonable to expect a predominance of AGT interactions when considering the effect of concerns for the environment on concerns for safety. Indeed, we do observe a predominance of AGT interactions overall (52%) and in terms of countries (19% or 61.2% of the countries) (figure 7, center left panel). We note, however, that this difference in the type of interaction is not statistically significant (paired t-test: t = 0.7844, p = 0.4389).

Last, there could be a relationship between concerns for the environment and social issues because people might prefer investing public resources for social issues, instead of supporting climate-friendly policies. This might be especially true when investments in social issues generate immediate benefits (e.g. healthcare) (Andor et al 2018). Against this background, one would expect FPW interactions to be predominant. We observe that FPW interactions arise 58% of the times (figure 7, right panel), while AGT interactions only emerge 42% of the times (paired t-test: $t = -2.4904$ and p = 0.0185). FPW interactions are also predominant at the country level (20 countries, or 64.5%).

We test whether there is a significant difference in how often AGT interactions emerge from the environment to the economy and from the environment to the other categories of worry. We find that AGT interactions emerge more frequently when looking at the effect of the environment on the economy (61.7%), then when looking at the effect of the environment on immigration (paired t-test: t = 5.8655, $p = 2.0316\times10^{-6}$), safety concerns (paired t-test: t = 3.2088, p = 0.0032) and social issues (paired t-test: t = 7.1664, $p = 5.7\times10^{-8}$) (see figure 8 and appendix table 3).

Zoom In Zoom Out Reset image size

Last, we compare how the other categories of worry influence concerns for the environment and for the economy. We find that the influence of other worries on concerns for the economy is dominated by AGT interactions (61.7%), whereas the effect of other worries on the environment is dominated by FPW interactions (60.6%). Hence AGT is more frequent when focusing on the impact of other worries on concerns for the economy than on concerns for the environment (paired t-test: t = 6.4521, $p = 3.97\times10^{-7}$)

This finding is not surprising. It is reasonable that an increase in concerns like immigration or safety might make people more worried about their economic situation. Vice versa, a person concerned about safety might be less focused on concerns on the environment. Thus, this seems to suggest that economic concerns are a concern of a higher order than environmental concerns. This hypothesis is supported by the interactions that characterize other worries. In fact, the only other worry toward which interactions are dominated by AGT is safety (52.1%), while interactions toward social issues and immigration are dominated by FPW (58.2% and 57.9% respectively).

Building on Maslow's famous pyramid of needs (Maslow 1954), one could summarize visually our results using a pyramid a worries (figure 9 and supplementary figure 1). We build the pyramid as follows. At the bottom we place economy because it is the worry that is most often in FPW interactions with other worries. Thus, a growth in concerns for the economy often pushes down the other concerns. We place personal safety just above economy because after economy it is the worry that stands more often in FPW interactions with other worries. We then continue until we reach environment, which sits at the very top of the pyramid because it is the worry that is less often in FPW interactions with other worries. We consider economy and personal safety tier 1 worries, because they are more often in FPW interactions than in AGT interactions toward other worries. Therefore, an increase in the level of concern for these tier 1 worries is likely to push down other concerns. To put it differently, tier 1 worries generally overtake other worries. Instead, at the top of the pyramid there are immigration, social issues and environment, which are worries that are more often in AGT interactions than in FPW interactions toward other worries (tier 2 worries). Thus, an increase in the level of concerns for a tier 2 worry is likely to increase the level of concern also for other worries.

Zoom In Zoom Out Reset image size

Our results refer to the average respondent in the 31 countries considered, and hence it is possible that for subgroups of the population the hierarchy of worries is different. For instance, younger generations might consider environmental concerns a more pressing issue than older generations.