Insights from international environmental legislation and protocols for the global plastic treaty

Our analysis yielded three main results: (i) the decision to strictly regulate pollutants is primarily influenced by the availability of viable substitutes rather than the scientific weight of evidence that they cause harm to the environment or humans; (ii) evidence for environmental concerns have been a key driving force behind previous pollution regulations, prioritizing the mitigation of environmental harm over potential risks to human health; and (iii) the Montreal Protocol, designed to protect the stratospheric ozone layer by phasing out CFCs, serves as a noteworthy model for transitioning away from polluting substances that resonate with the current state of process related to plastic pollution.

Environmental harm

Figure 1 presents results from the CA on a matrix combining the environmental harm and socio-economic factors displayed in Table 3 and including 209 out of 217 substances (substances with 0 score on all factors are deleted). While it is possible to display both factors and the substances in the biplot, we concentrate on the factors and depress the regulated substances because including the 209 substances makes the biplot difficult to read. In the case of environmental harm, it takes five dimensions (Dim) to explain 100% of the variation in the data. Each Dim corresponds to the proportion of variance covered, as measured by the Eigenvalue, and the first Dim account for the largest proportion, Dim 2 the second largest, and so on. Figure 1 displays the first four Dim, which cover 95.7% of the total variation in the dataset. These four Dim are displayed along the x- and y-axes in the biplots in Fig. 1. In the supplementary information, Fig. S1, is the biplot of Dim 4 and 5.

Figure 1

Biplots for CA of the factors indicating which environmental harm the substances cause and whether they are banned or not or have close substitutes or not. 209 out of 217 substances included. Dim 1 and 2 in panel (a), Dim 3 and 4 in panel (b). The color bar to the right shows the cosine square (cos2), and higher numbers (blue colors) indicate that the variable is well represented in the factor map.

Panel A of Fig. 1 displays Dim 1 and 2, representing 72.7% of total variation, and the color of the factor names shows that LAND_TOX contributes the most to explain variation along Dim 1 and 2, whereas SUBS and BAN explain the least of variation along the two first Dim. This can be seen from the color bar to the right in the plots giving the relative contribution of the factors to the two Dim. Dim 1 (x-axis) shows that substances contributing to climate change (CC) and ozone layer depletion (OZON), are closely associated with each other and that substances causing land and aquatic toxicity (LAND_TOX and AQUA_TOX, respectively) are closely associated with each other. Further, the fact that (CC & OZON) and (LAN_TOX & AQUA_TOX) are located at opposite ends of the x-axis means that substances vary greatly with respect to scoring on the two former and the two latter. Dim 2 (y-axis) explains variation in the data when variation along Dim 1 is controlled for. Dim 2 shows that LAND_TOX stands out from the other factors, i.e. the substances differ greatly when it comes to whether they are toxic to land ecosystems but not when it comes to the other factors. The colors confirm LAN_TOX is the single factor that explains most of the variation in the biplot, whereas SUB and BAN explain the least variation.

Panel B of Fig. 1 displays Dim 3 and 4, representing 21.3% of the total variation in the data, and from the color bar we can see that it is the regulations stringency (BAN) and presence of substitutes (SUBS) that explain the most of variation along these Dim. Dim 3 (x-axis) shows that substances being banned and having substitutes are closely correlated, and these distinguish from substances that cause ozone depletion, climate change and are toxic to aquatic ecosystems. Dim 4 (y-axis) shows that when variation along the three first Dim are controlled for, it is regulation stringency (BAN) that stands out from the other factors, i.e. the substances differ greatly when it comes to whether they are banned as input in economic activity, but less when it comes to the other factors. In particular, Dim 4 distinguishes between the factors BAN and SUBS as the two extremes, i.e. those substances being banned for use in economic activity are the least associated with substances that have substitutes when used in economic activity. Combining Dim 4 and 5 (see Supplementary information Fig. S1) confirms that Dim 4 distinguishes regulatory stringency (BAN) from the other factors, while Dim 5 mainly distinguishes between harming the ozone layer (OZON) and causing climate change (CC). Dim 5 shows that all factors except for CC and OZON are closely related. The colors confirm that SUBS and BAN are the factors that explain most of the variation in the biplot.

While the biplots visually describe associations between characteristics of regulated substances, it does not indicate the degree of statistical significance of the correlation. The significance of the correlations among variables in the model can be tested by Pearsons chi-squared test of statistically significant difference between expected frequencies and observed frequencies in one or more categories of a contingency table. For the biplot above the test statistic equal 1278, with a p-value equal to 0.000. Hence, we reject the null hypothesis (H0) about independence, which means that the regulated substances differ significantly when it comes to the distribution of values on the characteristics. In turn, this means that the distance between the factors in the biplots is statistically significant.

Human harm

Figure 2 presents results from CA of data combining human harm and socio-economic factors (see Table 3). The total variation in this data is explained by a total of 7 Dim. While Fig. 2 displays biplots representing the first 4 Dimensions, biplots for Dim 5 and 6, and 6 and 7, can be found in the supplementary Figs. S1 and S2. Panel A of Fig. 2 displays Dim 1 and 2, representing 48.1% of the total variation, and the color bar to the right shows that it is BAN and BIRTH that contribute the most to explain variation along Dim 1 & 2, whereas HA_TOX explain the least of the total variation. Dim 1 (x-axis) shows that being banned (BAN) and causing mutation in DNA (MUTAG) are the factors that discriminate the most among the substances. In other words, for a particular substance, there is little correlation between causing mutation in DNA and being banned. Dim 2 (y-axis) shows that when variation along Dim 1 is controlled for, it is BAN and BIRTH that discriminate the most among the substances, i.e. there is little correlation between substances causing birth defects and the fact that they are banned. The colors confirm that BAN is the single factor explaining most of the variation, and that BIRTH, SUBS and MUTAG also contribute well to explain variation in the biplot.

Figure 2
figure 2

Biplot for CA of the factors indicating which human harm the substances cause and whether they are banned or not or have close substitutes or not. 217 substances included. Dim 1 and 2 in panel (a), Dim 3 and 4 in panel (b). The color bar to the right shows the cosine square (cos2), and higher numbers (blue colors) indicate that the variable is well represented in the factor map.

Figure 3
figure 3

Biplot for CA of factors indicating environmental harm and whether they are banned (BAN) and have close substitutes (SUBS) for substances regulated by the Stockholm Convention (panel (a) and (b), N = 71) and the Montreal Protocol (panel (c), N = 16). All Dim included. The color bar to the right shows the cosine square (cos2), and higher numbers (blue colors) indicate that the variable is well represented in the factor map.

Panel B of Fig. 2 displays Dim 3 and 4, representing 29% of the total variation, and the color bar to the right indicates that it is the factor endocrine disruptive activity (ENDA) that explains most of the variation along these Dim, whereas SUBS contribute the least. Dim 3 (x-axis) shows that substances being toxic to reproduction (REP_TOX) and toxic (HA_TOX) are closely correlated, and these distinguish significantly from substances that cause birth defects (BIRTH) and mutations in DNA (MUTAG). Moreover, substances being toxic to reproduction and to humans in general are also closely correlated with being banned. Dim 4 (y-axis) shows that when variation along the three first three Dim is controlled for, substances that are banned are the least associated with substances being endocrine disruptive (ENDA). Colors show that ENDA is the single factor contributing the most to explaining the variation accounted for in this biplot.

Along all the first 4 Dim the factor SUBS, i.e. whether a substance has close substitutes when used in economic activity, is the factor that is closest associated with BAN, i.e. whether a substance is banned from use in economic activity. This is true although the association along Dims 2, 3 and 4 is not very close. In the supplementary information, Fig. S2 shows that this is no longer the case in Dims 5 and 7. Along these two Dims being banned from use in economic activity (BAN) is closer related to ENDA (interfering with hormones) and with REP_TOX (being toxic to reproduction), CARC (causing cancer) and MUTAG (causing mutation in human DNA). Dim 7 and 5 account for about 15% of the total variation in the data.

The chi-square test statistic for the model equal 2055 (p-value = 0.000). Hence, we reject the H0 about independence, which means that the regulated substances differ significantly when it comes to the distribution of values on the characteristics. In turn, this means that the distance between the factors in the biplots is statistically significant.

There are some interesting lessons from the CA results above. Comparing results from Figs. 1 and 2 we can see that while being banned or having close substitutes when used in economic activity are not important factors in discriminating between substances causing environmental harm (Fig. 1), these two factors do discriminate between substances causing human harm (Fig. 2). In other words, substances causing environmental harm tend to be similar when it comes to being banned and having substitutes, and closer inspection shows that they are likely to score positive on both. Substances causing human harm, on the other hand, tend to differ substantially when it comes to being banned and having substitutes. Hence, while being banned and/or having close substitutes is a unifying factor for substances that cause environmental harm, this is not the case for substances that cause human harm. Furthermore, the factors BAN and SUBS are the two factors closest associated along all Dims except for Dim 4, accounting for 8% of total variation when they are combined with environmental harm factors. The corresponding numbers when combined with human harm factors are 2 (Dim 5 and 7), accounting for 15% of total variation. Hence, there is a closer association between being banned from use in economic activity and having substitutes when used in economic activity for substances causing environmental harm compared to substances causing human harm.

The global agreements

Next, we focus on the Stockholm Convention and the Montreal Protocol that were aimed at regulating substances causing climate or environmental harm. Figure 3 displays CA biplots capturing the relevant environmental harm and socio-economic factors of the substances included in each of the two agreements, respectively. Panels A and B in Fig. 3 show results for the substances regulated under the Stockholm Convention and Panel C relates to substances within the the Montreal Protocol. Panels A and B display 100% of the variance in the data including the Stockholm Convention substances, whereas the Panel C biplot displays 100% of the variance in the data included within the Montreal Protocol substances.

The chi-square test statistic for the model of the Stockholm Convention substances equal 206 (p-value = 0.56). Hence, we cannot reject the H0 about independence, which means that the regulated substances do not differ significantly when it comes to the distribution of values on the characteristics. In turn, this means that the distance between the factors in the biplots is not statistically significant. For the model of the Montreal Protocol substances the chi-square test statistic equal 72.66 (p = 0.55), and the same conclusions as for the Stockholm Convention substancs apply.

The results in panel C of Fig. 3, representing results from the CA of substances included in the Montreal Protocol (rows) and environmental harm and socio-economic factors (columns) are interesting. First, from the color bar to the right we can see that all factors contribute equally in explaining the variation in the data, and total variation is explained by only two Dim. The fact that the factors SUBS and BAN coincide means that they take the same value for all substances, i.e., if a substance has substitutes when used in economic activity it is also banned, and vice versa. Dim 1 (x-axis) demonstrates that the most important variation is between substances causing depletion of the ozone layer and substances causing climate change. Substances causing depletion of the ozone layer are closer to the factors SUBS and BAN, indicating that these substances are more likely to have substitutes and be banned relative to substances causing climate change. This makes sense since it was the ozone-depleting substances the Montreal Protocol was designed to regulate. Dim 2 (y-axis), accounting for only 15% of total variation, demonstrates some variation between having substitutes and being banned on the one hand, and causing depletion of the ozone layer and climate change on the other hand. This variance accounts for the fact some substances causing climate change do not have substitutes and/or are banned.

Panels A and B in Fig. 3 display results from the CA on data including substances regulated under the Stockholm Convention (rows) and environmental harm and socio-economic factors (columns). Different from the Montreal Protocol substances, there is no overlap between being banned (BAN) and having substitutes when used in economic activity (SUBS), which means that substances that have substitutes are not necessarily banned, and vice versa. In other respects the results are similar to those for the Montreal Protocol substances. The Stockholm Convention substances differ mainly when it comes to the factor LAND_TOX, whereas they are relatively similar when it comes to the factors AQUA_TOX, SUBS and BAN. The latter means that causing harm to aquatic ecosystems (AQUA_TOX) is highly correlated with being banned and having substitutes. However, Dim 2 and 3, accounting for about 45% of the total variation, demonstrate that there are distinctions between being banned and having substitutes on one hand, and being toxic to aquatic environments on the other, and between having substitutes and being banned. Hence, although the Stockholm Convention is also effective in addressing the substances it has set out to regulate there is not an equally strong correlation between having substitutes, being banned and being toxic to the aquatic environment.

Previous post UM Nowadays | College of Regulation
Next post Charlottetown’s 10-calendar year system aims to make general public transit ‘simple, direct and frequent’