The Paris Agreement at Ten Years

3. The added value of the Paris Agreement

Abstract

To understand the perceived effect of the Paris Agreement, the OECD survey harnesses the expert opinion of policymakers and other climate mitigation experts by asking them to imagine what climate action would look like in a world without the Paris Agreement. This chapter describes the results that stem from this exercise. It first focuses on the past, present and future added value of the Paris Agreement, as compared to a hypothetical world without it. It then turns to its added value compared to the Kyoto Protocol. Finally, it outlines the specific characteristics of the Paris Agreement that underpin its value.

3.1. The past, present and future added value of the Paris Agreement

Policymakers believe that the Paris Agreement elevated climate change mitigation as a domestic policy priority. As indicated in the top left panel of Figure 3.1, the average policymaker in the sample reports that climate change mitigation was already somewhat of a policy priority in 2015. The average agreement index1 is 66%, which corresponds to “somewhat agree” (60%). For 2024, the agreement index rises to 84%, which is the value corresponding to the response “agree” (80%). This indicates a clear belief that mitigation is now viewed as a domestic priority under the Paris Agreement. The agreement index rises further when questioning whether climate change mitigation will be a domestic policy priority in the future in the presence of the Paris Agreement. By 2040 it reaches the value 89%, which corresponds to a response lying midway between “agree” (80%) and “strongly agree” (100%). While the Paris Agreement pathway shows a strong and consistent upward trajectory, the counterfactual path is flat and conservative. The 2024 value stays near neutrality (51%) and policymakers’ expectations remain close to a “somewhat agree” response through 2040. These findings suggest that the Paris Agreement has substantially strengthened the perceived importance of climate mitigation in national policy agendas.

Non-government experts believe that the Paris Agreement has safeguarded mitigation as an international policy priority and expect the agreement to elevate the importance of mitigation in the future. As illustrated in the top right panel of Figure 3.1, experts report a noticeable difference between the Paris Agreement pathway and the counterfactual scenario. Under the Paris Agreement pathway, the agreement index steadily increases between 2024 and 2040, reaching the value of 81% by 2040, i.e. above the level corresponding to “agree” (80%). This suggests growing confidence that mitigation will become a more central objective in the international policy agenda of the future. In contrast, under the counterfactual scenario, the index remains below a level corresponding to “somewhat agree” (60%), even by 2040. Notably, the gap between the two trajectories opens between 2015 and 2024 and remains largely stable thereafter. A closer look suggests that the divergence of beliefs in the period 2015-2024 does not stem from increasing optimism with the Paris Agreement, but rather from declining confidence in its absence. Without the Paris Agreement, experts are much less confident that mitigation would constitute a priority in 2024 (45%) than they believe it was in 2015 (63%). With the Paris Agreement, confidence in 2024 slightly exceeds that of 2015 (65% versus 63%). This suggests an alternative belief by climate change experts, notably that the key contribution of Paris Agreement during 2015-2024 was to safeguard, rather than raise the profile of climate change mitigation in international policy priorities.

Figure 3.1. The perceived evolution of mitigation with and without the Paris Agreement

Degree of respondents’ agreement with the statements

Notes: The panels on the left juxtapose the policymakers’ beliefs across the actual pathway (i.e. with the Paris Agreement, in the blue curve) against their respective beliefs across the counterfactual pathway (i.e. without the Paris Agreement, in the red curve); the panels on the right display the respective beliefs of individual experts; Upper panels display beliefs with respect to climate change mitigation being a policy priority (prioritisation). Lower panels display the respective beliefs with respect to climate change mitigation policies being stringent. The numbers corresponding to dotted points represent sample averages of the agreement index at the respective years. The value of 0 indicates strong disagreement, 0.5 (i.e. 50%) indicates a neutral position and a value of 1.0 (100%) strong agreement (see technical notes below). The two samples (policymakers, individual experts) are not directly comparable: government responses in the left panel evaluate the impact of the Paris Agreement domestically, while individual experts evaluate it internationally. The sampling process has been asynchronous, as the collection of responses from government experts was initiated before that of non-government climate change experts. The questionnaire for the latter group was extended to cover the year 2000, which adds insights discussed within the text and is explicitly used in the econometric analysis. Therefore, displayed results in the left and right panels should not be used for comparative analysis, but rather interpreted separately and exclusively as sources of complementary information. Displayed results correspond to parts A.1 and A.2 of the survey (See Chapter 2 and Annex A).

Technical notes: The agreement index of a respondent $i$ (denoted by $s_i$) takes the symmetric values 0, 0.2 (20%), 0.4 (40%), 0.6 (60%), 0.8 (80%), 1.0 (100%) if the respondent (policy making institution or individual expert) respectively “strongly disagrees”, “disagrees”, “somewhat disagrees”, “somewhat agrees”, “agrees” and “strongly agrees” with the statement being evaluated. The sample average agreement index is given by: $S=\left(1/N\right)\bullet \sum _i{s}_i$, where $N$ is the number of responses (sample size). A sample average agreement index of 0.5 indicates an overall neutral stance (see Chapter 2 for a discussion on why neutral positioning is not included as a response option but can emerge as a collective outcome). The sample average values of the agreement index in government responses are (values outside square brackets refer to the first question, and values within square brackets to the second): 0.66 [0.46] (2015), 0.84 [0.70] (2024, with Paris Agreement), 0.87 [0.79] (2030, with Paris Agreement), 0.89 [0.83] (2040, with Paris Agreement); 0.51 [0.41] (2024, without Paris Agreement), 0.57 [0.47] (2030, without Paris Agreement),0.59 [0.51] (2040, without Paris Agreement). In non-government expert responses, the corresponding values are: 0.35 [0.21] (for 2000), 0.63 [0.36] (for 2015), 0.65 [0.52] (for 2024, with Paris Agreement), 0.76 [0.61] (for 2030, with Paris Agreement), 0.81 [0.70] (for 2040, with Paris Agreement); 0.45 [0.32] (for 2024, without Paris Agreement), 0.52 [0.37] (for 2030, without Paris Agreement), 0.58 [0.45] (2040, without Paris Agreement).

Source: OECD survey on the transformative effect of the Paris Agreement, 2025.

Statlink

The Paris Agreement is also seen as a catalyst for more stringent mitigation policies. Perceptions regarding the effect of the Paris Agreement on the stringency of mitigation policies follow a similar pattern to those on mitigation as a policy priority. The lower panels of Figure 3.1 show that both government (left panel) and non-government experts (right panel) associate the Paris Agreement with noticeably more stringent mitigation policies. The agreement index, here reflecting agreement that mitigation policies are stringent in each time point, lies relatively low for initial years (between 21% and 46%). Policymakers slightly disagree that mitigation policies were stringent in 2015 (agreement index at 46%), and individual experts express a stronger disagreement (36%). The gap between the actual and counterfactual scenario widens over time in both groups. As a result, the level of optimism about the stringency of mitigation policies in 2040 is aligned with the level of optimism about climate change mitigation being a policy priority.

Climate change experts do not regard mitigation as a substantial policy priority in 2000, nor do they view mitigation policies as having been stringent in 2000. The agreement index regarding the statement “mitigation was an international policy priority in 2000” lies at 35%, i.e. below the value corresponding to somewhat disagree (40%). Levels of agreement reported by experts from non-OECD countries (45%) differ significantly from those reported by experts from OECD countries (28%).2 With respect to the statement “mitigation policies were stringent in 2000”, the agreement index lies at 21%, which corresponds to the label “disagree”. Here, opinions diverge more between experts from non-OECD (32%) and OECD (13%) countries.3

Climate change experts believe that substantial progress was achieved in the period that precedes the Paris Agreement (2000-2015). Confidence regarding “mitigation being a policy priority” and “mitigation policies being stringent” in 2015 is systematically higher compared to confidence levels for year 2000. While less than 40% of experts would agree that “mitigation was a policy priority in 2000”, over 60% would agree that the same statement would be true in 2015. Regarding mitigation policies being stringent, the respective change in the same period (2000-2015) is also remarkable, from 21% to 36%.

Experts believe that their perceived progress in the period 2000-2015 would not have been possible after 2015 without the Paris Agreement. Confidence of climate change experts regarding “mitigation being a policy priority” and “mitigation policies being stringent” in absence of the Paris Agreement declines between 2015 and 2024. Regarding 2024-2040, the expectations of experts rebound but at a pace that is slower than that observed during 2000-2015. Therefore, experts highlight that the progress they consider that was achieved prior to the agreement could not be replicated in the future without the Paris Agreement.

Figure 3.2. Assertion, agreement, and disagreement on the added value of the Paris Agreement

Distribution of opinions regarding climate change mitigation being a policy priority (left panels) and climate change policies being stringent

Notes: All panels display the percentage of respondents that “Strongly disagree”, “Disagree”, “Somewhat disagree”, “Somewhat agree”, “Agree” or “Strongly agree” with the statements shown (above/below) in the graph. Upper panels juxtapose policymakers’ beliefs across the actual pathway (i.e. with the Paris Agreement) against their respective beliefs across the counterfactual pathway (i.e. without the Paris Agreement); Lower panels display the respective beliefs of individual experts; Left panels display beliefs with respect to climate change mitigation being a policy priority; Right panels display the respective beliefs with respect to climate change mitigation policies being stringent. Differences between curves of the same colour across two neighbouring panels in a given year reflect the overall belief about the contribution of the Paris agreement. For example, by comparing the black curve in upper right panels, it can be seen that 29% of policymakers strongly agree that climate change mitigation policies will be stringent in 2030 with the Paris Agreement, while 7% are equally confident (i.e. “strongly agree”) that climate change mitigation policies would be stringent in 2030 without the Paris Agreement. Differences between curves of the same colour within a panel reflect the distribution of opinion at any given year. For example, 84% of the policymakers “agree” or “strongly agree” that mitigation policies will be stringent in 2040 with the Paris Agreement. However, less than 40% “strongly agree” with the same statement. All responses represent perceived effects of the Paris Agreement and can only be interpreted as such.

Source: OECD survey on the transformative effect of the Paris Agreement, 2025.

Statlink

The Paris Agreement is associated with stronger conviction that climate change mitigation is, and will remain, a policy priority, and that mitigation policies are, and will become, more stringent. To show the difference in strong convictions (i.e. “strongly agree”), Figure 3.2 unpacks the agreement index. This figure displays the entire distribution of opinions, from “strong disagreement” to “strong agreement”. Distributions are shown with respect to the same set of forward- and backward-looking statements on the prioritisation of climate change mitigation and the expected stringency of mitigation policies. The difference between the two scenarios is striking, when focusing on the percentage of respondents “strongly agreeing” with these statements. While 8% of policymakers and 10% of non-government experts strongly agree that “mitigation policies will be stringent in 2040” in the absence of the Agreement, these figures rise to 36% and 28% respectively in the presence of the Paris Agreement. The left panel of Figure 3.2 shows that, under the Paris Agreement, the prevalence of respondents “strongly agreeing” that climate change mitigation is, and will be, a policy priority is at least 2.4 times larger than the respective prevalence of those “strongly agreeing” that this would be the case without the Agreement in 2030 and 2040.

However, responses do not show consensus that mitigation policies are sufficiently stringent, or that climate change mitigation is a policy priority at the moment. At the domestic level, 19% of policymakers disagree that mitigation policies are currently stringent in their countries. The share of policymakers who express substantial confidence (i.e. “agree” or “strongly agree”) is 54% and the share of those who strongly agree that domestic mitigation policies are stringent lies at 21%. When looking at opinions from non-government experts, who were asked to focus on the global situation, figures widely differ. Less than 20% of non-government experts express substantial confidence (i.e. “agree” or “strongly agree”) that mitigation policies are stringent at the global level, and less than 10% “strongly agree” that this is the case. Therefore, the robust beliefs about the role of the Paris Agreement in the period 2015-2025 do not translate into a broad belief that prioritising mitigation and scaling up mitigation policies is already complete.

To ensure that the discussion of results provided above is not only meaningful, but also grounded in solid statistical evidence, the responses can be analysed with statistical methods. This can be done relying on the answers from non-government experts, which constitute a sufficiently large sample. The econometric analysis presented in Annex 3.A focuses on the two main statements treated in this part of the survey: (i) climate change mitigation constitutes a policy priority and (ii) climate change mitigation policies are stringent, internationally. In the econometric context, the agreement index, which is presented in percentage terms to facilitate comprehension, can be interpreted as the level of confidence that an expert would place on the two statements being true. The agreement index, as well as its various interpretations, is discussed in Box 3.1. Complementary econometric analysis is used to explore the likelihood of each separate possible answer, from “strongly disagree” to “strongly agree” (Annex 3.A).

Box 3.1. The agreement index and its possible interpretations

The agreement index constitutes the main quantitative tool of the study. The report assigns numerical values between 0% and 100% to categorical responses indicating different levels of agreement, with 0% and 100% assigned to the responses entitled “strongly disagree” and “strongly agree”, respectively. Values of 20%, 40%, 60%, and 80% are respectively assigned to responses “disagree”, “somewhat disagree”, “somewhat agree” and “agree”. The reported values of the agreement index are averages in the sample of respondents. Therefore, they may attain any value between 0% and 100%, including a midway value of 50%. This value corresponds to a neutral response, e.g. “neither agree nor disagree”, although the questionnaire does not allow for such a response at the individual level.

Apart from its main interpretation as a measure of average confidence that a statement is correct, the agreement index may be subject to various alternative interpretations. Such alternative interpretations of the agreement index are question specific. For instance, a response of 0% to the question “is climate mitigation a policy priority in 2024?” can be interpreted as an indication that climate change mitigation is placed very low in the ranking of policy priorities. Similarly, a response of 100% to the same question can be interpreted as an indication that climate change mitigation is placed very high in the ranking of policy priorities. In this sense, the agreement index could also be interpreted as a proxy to a placement index, although the study avoids this interpretation. Another example is the question: “in 2024, are climate change mitigation policies stringent?”. Here, a response of 0% could be interpreted as an indication that mitigation policies are not stringent at all, while a response of 100% as an indication that they are of maximum stringency. Therefore, in the context of this question, the agreement index could possibly be considered as a proxy to a policy intensity index. As it is the case with the placement index, the report abstains from literally interpreting the agreement index as policy intensity index.

3.2. The added value of the Paris Agreement vis-à-vis the Kyoto Protocol

The study provides the first quantitative comparative assessment of the Paris Agreement vis-à-vis its predecessor, the Kyoto Protocol. The survey invited respondents to assess outcomes under a specific counterfactual in which the Kyoto Protocol remained the main international reference from 2015 to 2024. Judgments were confined to the present (2024–2025), as extending the scenario further was considered implausible. Figure 3.3 displays confidence levels in a range of hypothetical statements describing this alternative situation, offering key takeaways from expert and policymaker assessments.

The Paris Agreement is perceived as being more effective compared to its predecessor, the Kyoto Protocol. Overall, the views of country policymakers and non-government experts converge to a belief that the Paris Agreement is more effective in accelerating ambition and in making laws and policies more stringent than the previous status quo. In addition, respondents present substantial agreement (71% mean agreement for country policymakers; 77% for non-government experts) that the Paris Agreement is more effective in guiding responsibility. There is also agreement that it provides more benchmarks to measure progress compared to its predecessor. At the same time, confidence that the Paris Agreement accelerated actual emission cuts is slightly lower for both groups of respondents. These views align with the findings of the first Global Stocktake (UNFCCC, 2023[1]) and of the Emissions Gap reports (United Nations Environment Programme, 2024[2]), both highlighting the need to increase efforts to close the implementation gap.

The Paris Agreement is perceived as a tool to mainstream climate action. Respondents show a high level of consensus that the Paris Agreement is more effective than the Kyoto Protocol in mainstreaming climate action within the government and the political system (agreement index 72-75%, with the range spanning governments and non-government responses, weighted or unweighted). The confidence that securing public support for climate action is easier under the Paris Agreement is comparable for non-government experts (72%) and policymakers (71%).

Figure 3.3. Visualising the world in 2025 with the Kyoto Protocol still in place

Agreement index on various statements

Notes: Agreement index values of 0, 20, 40, 60, 80 and 100% correspond to qualitative responses “Strongly disagree”, “Disagree”, “Somewhat disagree”, “Somewhat agree”, “Agree”, “Strongly agree”. Mean values are presented as points and standard deviation ranges as error bars. For details on the weighting of expert views see Chapter 2. All results displayed here refer to part A.3 of the survey.

Source: Graph generated by the authors based on estimates using data from the OECD survey on the transformative effect of the Paris Agreement, 2025.

Statlink

Figure 3.4. Breaking down the added value of the Paris Agreement

Agreement index on various statements

Notes: Agreement index values of 0, 20, 40, 60, 80 and 100% correspond to qualitative responses “Strongly disagree”, “Disagree”, “Somewhat disagree”, “Somewhat agree”, “Agree”, “Strongly agree”. Mean values are presented as points and standard deviation ranges as error bars. For details on the weighting of expert views see Chapter 2. All results displayed here refer to part A.4 of the survey.

Source: Graph generated by the authors based on estimates using data from the OECD survey on the transformative effect of the Paris Agreement, 2025.

Statlink

The Paris Agreement is believed to have facilitated, rather than hampered, action to tackle other environmental issues. Non-government experts are much more confident that the Paris Agreement increased action to tackle other transboundary environmental problems (e.g. plastic pollution), than they are that the Paris Agreement consumed part of the attention and political capital available for these other environmental issues (57% versus 39%). Policymakers place more confidence in the former, rather than the latter narrative, although the difference is not decisive (51% versus 47%). As shown in Figure 3.3, the weighting of non-government responses, the way it is described in the Annex of Chapter 2, does not qualitatively affect the findings.

3.3. Characteristics of the Paris Agreement that underpin its added value

The responses to the survey highlight the high perceived added value of the Paris Agreement. The potential sources of added value are split into three categories (Figure 3.4). Policymakers consistently display higher confidence than non-government experts in the contribution of most sources of added value. A first set of contributions is related to scale and efficiency considerations. In terms of scale, both country policymakers and non-government experts are confident (72-83%, with the range spanning both groups, weighted or unweighted) that the Paris Agreement increased the number of countries undertaking substantial climate action but less confident (67-73%) that it had the same effect on the participation of private sector in global climate action. In terms of efficiency, participants place similar levels of confidence (67-73%) on the belief that international climate action under the Paris Agreement is more solution oriented. A second set of perceived contributions regards the unique features that the Agreement uses to scale up ambition, i.e. the ratchet mechanism (61-73%) and the peer pressure (63-73%). Finally, respondents value the Paris Agreement as a diplomatic tool and as a blueprint to tackle other global issues.

References

[1] UNFCCC (2023), Decision 1/CMA.5: Outcome of the First Global Stocktake.

[2] United Nations Environment Programme (2024), Emissions Gap Report 2024: No more hot air … please! With a massive gap between rhetoric and reality, countries draft new climate commitments, United Nations Environment Programme, https://doi.org/10.59117/20.500.11822/46404.

Annex 3.A. Econometric analysis of non-government responses (survey parts A1 and A2)

Ordinary Least Squares (OLS) econometric analysis

Annex Table 3.A.1. Econometric analysis of non-government expert responses

	(i)	(ii)
Item in the survey	mitigation was/is/will be/would be a policy priority	mitigation policies were/are/will be/would be stringent
Dependent variable	Agreement index	Agreement index
Predicted initial value of index (1999)	33.8% (***)	19.9% (***)
Predicted initial value of index (1999), experts from OECD countries	32.0% (**)	14.9% (***)
Predicted initial value of index (1999), experts from non-OECD countries	36.2% (***)	26.9% (***)
Estimated annual increase in the index (2000-2015)	1.40 percentage points (***)	0.94 percentage points (***)
Estimated annual increase in the index (2015-2040), under the Paris Agreement	1.07 percentage points ( )	1.54 percentage points (**)
Predicted annual increase in the index (2015-2040), without the Paris Agreement	-0.11 percentage points (***)	0.30 percentage points (***)

Notes: The agreement index reflects how confident experts are that a statement is true. For example, results imply that in a setting with a binary (“agree”/”disagree”) menu of responses, 33.8% of experts would agree (while 66.2% would disagree) with the statement “Was mitigation a policy priority in 1999?” and 19.9% of experts would agree (while 80.1% would disagree) with the statement “Were mitigation policies stringent in 1999?”. All findings are derived with data from parts A.1 and A.2 of the survey.

Technical notes: The model specification in column (i) is: ${\mathrm{x}}_{\mathrm{t}\mathrm{i}}=\left(a+a_{OECD}\bullet D_{i.OECD}\right)+\beta \bullet \left(t-1999\right)+\left(t-2015\right)\bullet [ \gamma \bullet D_{t>2015}+\delta \bullet D_{t>2015}\bullet D_{PA==false} ]$, where ${\mathrm{x}}_{\mathrm{t}\mathrm{i}}$ is the agreement index of respondent $i$ that “climate change mitigation was/is/will be/would be a policy priority”, obtaining values between 0 (0% for “strongly disagree”) and 1 (100% for “strongly agree”); $D_{i.OECD}$ is a dummy variable that equals 1 if respondent $i$ is from an OECD country (otherwise zero); $t$ is the year to which the question refers (e.g. if the question reads “will climate change mitigation be a policy priority in 2030, under the Paris Agreement?”, then $t=2030$); $D_{t>2015}$ is a dummy variable that equals 1 if $t>2015$, i.e. if the question regards a year after the Paris Agreement (otherwise zero); $D_{PA==false}$ is a dummy variable that equals 0 if the question utilises the factual scenario and 1 under the scenario that the Paris Agreement is absent. Model specification in column (ii) is identical, except for ${\mathrm{x}}_{\mathrm{t}\mathrm{i}}$, which in this case represents the agreement index of respondent $i$ that “climate change mitigation policies were/are/will be/would be stringent”. Estimates are reported with numerical precision on the second decimal digit. Notation for significance: (***) p-value < 0.001; (**) p-value < 0.01; (*) p-value < 0.05; (.) p-value < 0.10. All p-values are obtained via White standard errors.

Source: Estimations performed by the authors.

Annex Table 3.A.2. Econometric analysis of non-government responses with year fixed effects

	(1)	(2)	(3)	(4)
Survey part/ dependent variable	A1	A2	A1	A2
Year fixed effects
D2000	0.352 (***, ***, ***)	0.208 (***, ***, ***)
	(0.021, 0.022, 0.022)	(0.020, 0.018, 0.018)
D2015	0.627 (***, ***, ***)	0.360 (***, ***, ***)	0.275 (***, ***, ***)	0.152 (***, ***, ***)
	(0.021, 0.018, 0.018)	(0.020, 0.019, 0.019)	(0.022, 0.028, 0.021)	(0.019, 0.021, 0.014)
D2024	0.649 (***, ***, ***)	0.517 (***, ***, ***)	0.298 (***, ***, ***)	0.309 (***, ***, ***)
	(0.021, 0.021, 0.021)	(0.020, 0.019, 0.019)	(0.022, 0.027, 0.027)	(0.019, 0.020, 0.019)
D2030	0.756 (***, ***, ***)	0.610 (***, ***, ***)	0.404 (***, ***, ***)	0.402 (***, ***, ***)
	(0.021, 0.019, 0.019)	(0.020, 0.019, 0.019)	(0.022, 0.026, 0.029)	(0.019, 0.019, 0.021)
D2040	0.813 (***, ***, ***)	0.704 (***, ***, ***)	0.462 (***, ***, ***)	0.497 (***, ***, ***)
	(0.021, 0.018, 0.018)	(0.020, 0.019, 0.019)	(0.022, 0.026, 0.029)	(0.019, 0.021, 0.023)
D2024∙DPA==false	-0.197 (***, ***, ***)	-0.193 (***, ***, ***)	-0.197 (***, ***, ***)	-0.193 (***, ***, ***)
	(0.029, 0.030, 0.021)	(0.028, 0.028, 0.020)	(0.022, 0.020, 0.021)	(0.019, 0.018, 0.020)
D2030∙DPA==false	-0.234 (***, ***, ***)	-0.240 (***, ***, ***)	-0.234 (***, ***, ***)	-0.240 (***, ***, ***)
	(0.029, 0.029, 0.019)	(0.028, 0.028, 0.019)	(0.022, 0.018, 0.019)	(0.019, 0.017, 0.019)
D2040∙DPA==false	-0.229 (***, ***, ***)	-0.249 (***, ***, ***)	-0.229 (***, ***, ***)	-0.249 (***, ***, ***)
	(0.029, 0.030, 0.020)	(0.028, 0.030, 0.020)	(0.022, 0.021, 0.020)	(0.019, 0.021, 0.020)
Other fixed effects	no	no	individual	individual
Number of observations	1424	1424	1424	1424
R2	0.833	0.762	0.597	0.687
R2 Adjusted	0.832	0.7606	0.537	0.640
R2 Within			0.345	0.444
R2 Within Adjusted			0.342	0.441

Technical notes: Columns (1) and (2) correspond to the OLS regression model: ${\mathrm{x}}_{\mathrm{t}\mathrm{i}}=\sum _t({\gamma }_t\bullet D_t)+\sum _{t>2015}({\delta }_t{\bullet D}_t\bullet D_{PA==\mathrm{f}\mathrm{a}\mathrm{l}\mathrm{s}\mathrm{e}})+{\epsilon }_{it}$, where ${\mathrm{x}}_{\mathrm{t}\mathrm{i}}$ is the agreement index of respondent $i$ that “climate change mitigation was/is/will be/would be a policy priority” (survey part A1, column 1) or that “climate change mitigation policies were/are/will be/would be stringent” (survey part A2, column 2), with values between 0 (0% for “strongly disagree”) and 1 (100% for “strongly agree”); $t$ is the year to which the question refers (e.g. if the question reads “will climate change mitigation be a policy priority in 2030, under the Paris Agreement?”, then $t=2030$); $D_{PA==false}$ is a dummy variable that equals 0 if the question utilises the factual scenario (under the Paris Agreement) and 1 under the counterfactual scenario, in which the Paris Agreement is absent. Columns (3) and (4) correspond to a model in which respondent fixed effects ($A_i$) are added: ${\mathrm{x}}_{\mathrm{t}\mathrm{i}}=A_i+\sum _t({\gamma }_t\bullet D_t)+\sum _{t>2015}({\delta }_t{\bullet D}_t\bullet D_{PA==\mathrm{f}\mathrm{a}\mathrm{l}\mathrm{s}\mathrm{e}})+{\epsilon }_{it}$. Reported standard errors are presented in the following order: classical (i.i.d.), White (heteroscedasticity robust) and clustered at the individual level. Notation for significance: (***) p-value < 0.001; (**) p-value < 0.01; (*) p-value < 0.05; (.) p-value < 0.10.

Source: Estimations performed by the authors.

Annex Table 3.A.3. Statistical tests on econometric regression results

item	Null hypothesis	H0	H1	p-value
A1	Stagnation (2015-2024) under the Paris Agreement pathway: mitigation in 2024 is as much of a priority as it was in 2015.	γ2024=γ2015	γ2024>γ2015  or γ2024<γ2015	0.441
A1	Persistent stagnation (2015-2030) under the Paris Agreement pathway: experts expect mitigation in 2030 to be as much of a policy priority it is today	γ2030=γ2024	γ2030>γ2024	1.287∙e-04
A1	Setback in beliefs (2015-2024) under the counterfactual: without the Paris Agreement, mitigation in 2024 would not be more of a policy priority than it was in 2015.	γ2024+δ2024=γ2015	γ2024+δ2024>γ2015	1.00
A1	Persistent mid-term setback in beliefs (2015-2030) under the counterfactual: without the Paris Agreement, mitigation in 2030 would be much less of a policy priority than it was in 2015.	γ2030+δ2030=γ2015	γ2030+δ2030> γ2015	1.000
A1	Persistent long-run setback in beliefs (2015-2040) under the counterfactual: without the Paris Agreement, mitigation in 2040 would be much less of a policy priority than it was in 2015.	γ2040+δ2040=γ2015	γ2040+δ2040> γ2015	0.929
A2	Non-acceleration: The Paris Agreement (2015-2024) has not accelerated the speed at which mitigation policies were becoming gradually more stringent before it (2000-2015).	(1/15)∙(γ2015-γ2000)=(1/9)∙(γ2024-γ2015)	(1/15)∙(γ2015-γ2000)<(1/9)∙(γ2024-γ2015)	0.045
A2	Setback in beliefs (2015-2024) under the counterfactual: without the Paris Agreement, mitigation policies in 2024 would be much looser than they were in 2015.	γ2024+δ2024=γ2015	γ2024+δ2024>γ2015	0.901
A2	Persistent mid-term setback in beliefs (2015-2030) under the counterfactual: without the Paris Agreement, mitigation policies in 2030 will be much looser than they were in 2015.	γ2030+δ2030=γ2015	γ2030+δ2030> γ2015	0.358
A2	Persistent long-run setback in beliefs (2015-2040) under the counterfactual: without the Paris Agreement, mitigation policies in 2040 will be much looser than they were in 2015.	γ2040+δ2040=γ2015	γ2040+δ2040> γ2015	3.15∙e-04

Technical notes: The tests are performed on the estimated coefficients of the models corresponding to the first columns (1) and (2) of Annex Table 3.A.2. The OLS model that gives rise to the estimates is: ${\mathrm{x}}_{\mathrm{t}\mathrm{i}}=\sum _t({\gamma }_t\bullet D_t)+\sum _{t>2015}({\delta }_t{\bullet D}_t\bullet D_{PA==\mathrm{f}\mathrm{a}\mathrm{l}\mathrm{s}\mathrm{e}})+{\epsilon }_{it}$, where ${\mathrm{x}}_{\mathrm{t}\mathrm{i}}$ is the agreement index of respondent $i$ that “climate change mitigation was/is/will be/would be a policy priority” (A1) or that “climate change mitigation policies were/are/will be/would be stringent” (A2), with values between 0 (0% for “strongly disagree”) and 1 (100% for “strongly agree”); $t$ is the year to which the question refers (e.g. if the question reads “will climate change mitigation be a policy priority in 2030, under the Paris Agreement?”, then $t=2030$); $D_{PA==false}$ is a dummy variable that equals 0 if the question utilises the factual scenario (under the Paris Agreement) and 1 under the counterfactual scenario in which the Paris Agreement is absent.

Source: Estimations performed by the authors.

Logistic econometric analysis

The analysis of the agreement index presented within Chapter 3 and in the first part of this Annex is well-suited to predict the overall (i.e. average) level of agreement with a statement. It is also suitable to explore how this overall level of agreement evolves over time.

Additional insights can be obtained with estimations from an ordered logistic regression. Instead of predicting average agreement with a statement, ordinal logit models distribute their focus across every possible response on the Likert-scale, from “strongly disagree” to “strongly agree”. Therefore, they estimate the probability with which a respondent will “strongly disagree”, “disagree”, “somewhat disagree”, “somewhat agree”, “agree” and “strongly agree” with a statement. In contrast to the OLS estimations, ordered logit models do not require an assumption about the “distance” separating these subsequent responses in the Likert scale (e.g. an assumption on how far “agree” lies from “strongly agree”). Their only requirement is that the ranking of response labels is ordinal (e.g. that “somewhat agree” implies a higher level of agreement than the previous label, i.e. “somewhat disagree”, and so on).

The analysis is based on the same explanatory variables used in the OLS regression presented in the first part of this Annex. These explanatory variables include year dummy variables (2000, 2015, 2024, 2030 or 2040) and their interactions with the two scenarios (i.e. the factual scenario including the Paris Agreement and the counterfactual scenario without the Paris Agreement). Models also include an indicator variable that equals 1 if the respondent is from an OECD country and 0 otherwise. This way, the probability that a respondent gives a certain response is differentiated between the two scenarios and between OECD and non-OECD countries. The estimated choice probabilities are displayed in Annex Figure 3.A.1 and reported in Annex Table 3.A.4.

The results show that OECD respondents possess more conservative views than their non-OECD counterparts. A respondent from a non-OECD country is 1.3 times more likely to strongly agree that mitigation is a policy priority and 3.5 times more likely to agree that mitigation policies in her country are stringent. When including a weaker form of agreement, the latter ratio remains notably high. In particular, a respondent from a non-OECD country is 2.9 times more likely to agree or strongly agree that mitigation policies in her country are stringent. Including all forms of agreement maintains the ratio significantly larger than one (1.5).

Both groups expect stronger mitigation policies in future years. Compared to the year 2000, a respondent from an OECD country is 3.3 times more likely to strongly agree that mitigation policies were stringent in 2015. With the Paris Agreement, the ratio increases to 9.3 for 2024, 12.8 for 2030 and 30.3 for 2040. These ratios do not substantially differ when it comes to non-OECD experts, but they do become considerably different when examining the counterfactual scenario in which the Paris Agreement is absent: just 2.5 for 2024, 3.5 for 2030, and 6.3 for 2040.

Annex Figure 3.A.1. Response probabilities predicted by an ordered logit model

Predicted probabilities of agreement levels resulting from the partial proportional odds logistic models

Technical notes: Initial estimates were obtained from the ordered logit model with proportional odds. In this model the probability that respondent $i$ expresses a level of agreement $Y_i$ up to or equal to $j$(that is, $P(Y_i\le j)$), relative to the probability she expresses a level of agreement $Y_i$ higher than $j$ (that is, $P(Y_i>j)$) is modelled as: $\mathrm{l}\mathrm{o}\mathrm{g}\left(\frac{P\left(Y_i\le j\right)}{P\left(Y_i>j\right)}\right)={\theta }_j\mathrm{}-\underset{{\mathrm{x}}_{\mathrm{i}\mathrm{j}}}{\underbrace{[\mathrm{ }\beta \bullet D_{oecd,i}+\sum _t{\gamma }_t{\bullet D}_t+\sum _{t>2015}{\delta }_t{\bullet D}_t\bullet D_{PA}]}}$, where $j\in$ {strongly disagree, disagree, somewhat disagree, somewhat agree, agree, strongly agree} and the remaining notation can be retrieved from Annex Table 3.A.1. The proportional odds specification assumes that $\beta , {\gamma }_t$and ${\delta }_t$ are constant for all levels of agreement $j$. The assumption is tested and rejected with the Brant test. Subsequently, the component ${\mathrm{x}}_{\mathrm{i}\mathrm{j}}$ is remodelled to allow $\beta ,{\gamma }_t,{\delta }_t$ to differ across agreement categories ($j$). For the model shown in Panel A and C, corresponding to question A1, the Brant test indicated a violation of the proportional odds assumption for $(D_{2030}\bullet D_{PA})$ , $(D_{2040}\bullet D_{PA})$ and $D_{2015}$. For the model corresponding to question A2 (Panel B and D), the Brant test indicated a violation for $(D_{2030}\bullet D_{PA}$) and $D_{oecd}$. The proportional odds assumption was then relaxed accordingly. The graphs displayed in the figure are based on the estimates from this updated model. Predicted probabilities used in this graph are reported in Table A.D.4.

Source: Estimations performed by the authors.

Statlink

Annex Table 3.A.4. Probabilities predicted by generalised logistic regression

Survey question	Scenario	Country	Year	Strongly agree	Agree	Somewhat agree	Somewhat disagree	Disagree	Strongly disagree
A1 Mitigation as a policy priority	Without the Paris Agreement	Non-OECD	2000	4.6%	10.9%	20.0%	17.2%	29.1%	18.2%
			2015	12.5%	30.1%	39.3%	10.7%	3.3%	4.1%
			2024	8.6%	17.6%	25.3%	16.7%	21.4%	10.3%
			2030	12.2%	22.2%	26.7%	15.0%	16.7%	7.3%
			2040	17.4%	26.9%	26.1%	12.4%	12.3%	4.9%
		OECD	2000	3.4%	8.3%	16.6%	16.1%	31.9%	23.8%
			2015	9.3%	25.4%	41.7%	13.6%	4.4%	5.7%
			2024	6.3%	14.0%	22.9%	17.4%	25.5%	13.9%
			2030	9.0%	18.2%	25.6%	16.6%	20.7%	9.9%
			2040	13.1%	23.2%	26.7%	14.5%	15.8%	6.7%
	With the Paris Agreement	Non-OECD	2000	4.6%	10.9%	20.0%	17.2%	29.1%	18.2%
			2015	12.5%	30.1%	39.3%	10.7%	3.3%	4.1%
			2024	23.9%	30.4%	23.7%	9.8%	8.9%	3.3%
			2030	40.6%	30.7%	15.6%	9.0%	0.9%	3.2%
			2040	53.9%	25.4%	12.2%	4.7%	0.9%	2.8%
		OECD	2000	3.4%	8.3%	16.6%	16.1%	31.9%	23.8%
			2015	9.3%	25.4%	41.7%	13.6%	4.4%	5.7%
			2024	18.3%	27.6%	25.8%	12.0%	11.7%	4.6%
			2030	32.8%	31.2%	18.6%	11.8%	1.3%	4.4%
			2040	45.5%	27.8%	15.3%	6.3%	1.3%	3.8%
A2 Mitigation policies as stringent	Without the Paris Agreement	Non-OECD	2000	2.3%	3.7%	10.5%	12.4%	34.6%	36.4%
			2015	7.6%	10.6%	22.4%	17.7%	27.3%	14.3%
			2024	5.9%	8.7%	19.8%	17.4%	30.3%	17.8%
			2030	8.2%	11.3%	23.2%	17.7%	26.3%	13.3%
			2040	13.8%	16.6%	26.9%	16.0%	18.8%	7.8%
		OECD	2000	0.6%	1.0%	5.9%	12.5%	37.2%	42.8%
			2015	2.0%	3.1%	16.6%	24.5%	35.9%	17.9%
			2024	1.5%	2.4%	13.6%	22.2%	38.2%	22.1%
			2030	2.1%	3.4%	17.7%	25.2%	35.0%	16.7%
			2040	3.8%	5.7%	25.7%	27.6%	27.3%	10.0%
	With the Paris Agreement	Non-OECD	2000	2.3%	3.7%	10.5%	12.4%	34.6%	36.4%
			2015	7.6%	10.6%	22.4%	17.7%	27.3%	14.3%
			2024	19.7%	20.3%	27.2%	13.6%	14.0%	5.3%
			2030	25.6%	29.2%	25.1%	13.3%	3.6%	3.2%
			2040	47.7%	23.6%	17.1%	5.6%	4.5%	1.5%
		OECD	2000	0.6%	1.0%	5.9%	12.5%	37.2%	42.8%
			2015	2.0%	3.1%	16.6%	24.5%	35.9%	17.9%
			2024	5.6%	8.2%	31.5%	26.7%	21.2%	6.8%
			2030	7.7%	14.8%	39.1%	27.8%	6.5%	4.1%
			2040	18.2%	19.2%	38.2%	15.0%	7.5%	1.9%

Technical notes: Initial estimates were obtained from the ordered logit model with proportional odds. In this model the probability that respondent $i$ expresses a level of agreement $Y_i$ up to or equal to $j$(that is, $P(Y_i\le j)$), relative to the probability she expresses a level of agreement $Y_i$ higher than $j$ (that is, $P(Y_i>j)$) is modelled as: $\mathrm{l}\mathrm{o}\mathrm{g}\left(\frac{P\left(Y_i\le j\right)}{P\left(Y_i>j\right)}\right)={\theta }_j\mathrm{}-\underset{{\mathrm{x}}_{\mathrm{i}\mathrm{j}}}{\underbrace{[\mathrm{ }\beta \bullet D_{oecd,i}+\sum _t{\gamma }_t{\bullet D}_t+\sum _{t>2015}{\delta }_t{\bullet D}_t\bullet D_{PA}]}}$, where $j\in$ {strongly disagree, disagree, somewhat disagree, somewhat agree, agree, strongly agree} and the remaining notation can be retrieved from Annex Table 3.A.1. The proportional odds specification assumes that $\beta , {\gamma }_t$and ${\delta }_t$ are constant for all levels of agreement $j$. The assumption is tested and rejected with the Brant test. Subsequently, the component ${\mathrm{x}}_{\mathrm{i}\mathrm{j}}$ is remodelled to allow $\beta ,{\gamma }_t,{\delta }_t$ to differ across agreement categories ($j$). For the model shown in Panel A and C, corresponding to question A1, the Brant test indicated a violation of the proportional odds assumption for $(D_{2030}\bullet D_{PA})$ , $(D_{2040}\bullet D_{PA})$ and $D_{2015}$. For the model corresponding to question A2 (Panel B and D), the Brant test indicated a violation for $(D_{2030}\bullet D_{PA}$) and $D_{oecd}$. The proportional odds assumption was then relaxed accordingly. The results of this partial proportional odds model (predicted probabilities) are shown in this table.

Source: Estimations performed by the authors.

Notes

← 1.  To quantify responses on the Likert scale, the agreement index assigns a numerical value to each response (1: strongly agree, 0.8: agree, 0.6: somewhat agree, 0.4: somewhat disagree, 0.2: disagree, 0: strongly disagree). A value of 0.5 can be considered to correspond to a neutral response (e.g. “neither agree nor disagree”). The average agreement index is the mean of these values across all respondents for a given question. It provides a continuous measure of overall agreement, ranging from 0 (strong disagreement) to 1 (strong agreement).

← 2.  Using an alternative interpretation of the agreement index as a placement index (as described in Box 3.1), the estimates could also be viewed to suggest that the average expert would place climate change mitigation somewhere in the 4th quantile (i.e. 60th to 80th percentile) of policy priorities in 2000. However, the study abstains from formally interpreting results this way.

← 3.  Although responses pertain to the average stringency of mitigation policies across the world, the differences between OECD and non-OECD experts highlight that the two groups have a different perception of the global effort made. A possible reason behind these differences is that the perception of global progress is still affected by the local efforts made in different areas of the world, and the relative social costs that these efforts entail.