As part of their Paris Agreement obligations, countries outline and communicate their short- to medium-term climate goals in the form of Nationally Determined Contributions (NDCs). At the centre of NDCs are parties’ emission targets, usually to 2030, expressed in various formulations that involve reductions relative to a historical base year, a future business-as-usual (BAU) scenario, or the current emission intensity of the economy. These targets have been the primary focus of the research on NDCs to date, particularly covering their relative ambition1,2, implications for temperature outcomes3,4,5, and conditionality requirements6,7. Although it is not a mandatory part of NDCs, many countries have now started to include an optional adaptation component, which has started to receive some attention in the literature8,9.
However, NDCs often contain other information that has received less systematic attention, including national context and circumstances; measurement, reporting and verification; fairness, equity and ambition; means of implementation, and information necessary for clarity, transparency and understanding (ICTU). At the same time, NDCs have received criticism for lacking transparency not only regarding mitigation targets but also other issues, including implementation, financing, and adaptation components10,11. Nonetheless, there remains much potential to explore the discourse within countries’ NDCs in more depth, both in terms of current content and deficiencies.
Here, we systematically assess and compare the textual content within NDCs using computational linguistics analysis to go beyond the usual emphasis on mitigation targets. Through this, we establish seven broad topic groups, such as development, implementation, and stakeholders, and 21 constituent topics. We analyse how these relate to the characteristics of the associated parties and their mitigation targets.
Basic features of NDC
When we gathered data on the content of NDCs in June 2023, 166 parties had submitted their first NDC to the UNFCCC, counting the joint submission by the EU countries as one party. Of these, 144 presented at least one updated NDC with some parties having produced up to three updates. Two parties, Yemen and Iran, had submitted intended nationally determined contributions (INDCs) but had yet to deliver their first formal NDC.
The NDCs vary considerably in terms of length, from a mere 96 words for Kazakhstan’s first NDC to 76,275 words for Venezuela’s updated NDC. Figure S1 of the Supplementary Information shows the distribution of NDCs by word count, with a mean and median of 7,760 and 4,094 words, respectively. Figure S2 of the Supplementary Information shows the difference in the length of NDCs between countries and NDC version. Generally, we see a movement towards longer documents in updated NDCs but find that some countries submitted very brief updated NDCs that just outlined key changes from the original submissions. While no first NDCs exceeded 10,000 words and the majority were under 5,000, a substantial proportion of updated NDCs were over 10,000 words. Longer NDCs were also more likely to be submitted by low- or middle-income parties while high-income parties like the EU, USA, Iceland, and New Zealand produced some of the shortest NDCs, both for initial and updated submissions.
NDCs also vary considerably with the level of ambition and the conditionality of associated emission targets. While comparing ambition can be difficult when viewing targets at face value due to differences in the way parties express their targets as well as differences between factors such as base years and BAU projections, this can be ascertained by normalising the targets10. Pledges can also be wholly unconditional or conditional (typically on international support and finance), or contain a mixture of unconditional and conditional targets.
Topic modelling of NDC texts
There are considerable differences among NDCs in terms of content and structure. Through computational linguistic analysis, we can map the most common themes addressed in the NDCs. The specific approach we use is topic modelling, a technique on the intersection of natural language processing and machine learning. It has seen various applications in the social sciences12,13, notably to decompose a large amount of textual data14,15,16. The method essentially clusters multiple texts into distinct topics, allowing their fast, transparent and reproducible quantification. This is achieved by probabilistically assigning words from the texts to different topics so that each topic can be characterised by words frequently and predominantly appearing in its discussion. Subsequently, the identified topics can be statistically related to other observable characteristics of the texts17 (e.g., length of NDC and if it was an original or updated submission) or to features of the associated countries (e.g., GHG emissions and GDP per capita). In turn, this can help to explain why certain themes appear (see Methods for more details).
To relate topic prevalences to the parties that submitted the underlying NDCs, we calculated for each party the vector consisting of 21 topic prevalences (in cases of countries having submitted both first and updated NDCs, we took the average of the two), measured the Euclidian distance between all the 167 parties included in our analysis based on these vectors, and then applied a hierarchical clustering. This enabled us to allocate the 167 parties into nine clusters based on similarity of NDC content.
Identified topics in NDCs
We identified 21 topics (designated with the prefix “T”) in the set of 309 NDCs. These can be aggregated into seven broader topic groups (designated with the prefix “G”), as indicated in Table 1. More details about the constituent topics are provided in Table 2, namely a ranking by their prevalence and a short description of their content. Figure S3 of the Supplementary Information further illustrates the most common and exclusive words for each topic in the form of a word cloud.
Development (G1), covering 25% of all NDC content, is the most common topic group discussed in the NDCs. It captures topics related to sustainable (T1), economic (T2), and rural (T15) development. Other topic groups are parties’ implementation, plans, and processes (G2) accounting for 21.5% of NDC content, while a further 20.3% is dedicated to parties’ mitigation targets (G3): greenhouse gas accounting (T3), emission scenarios (T5), and targets and reporting (T8).
A further 18.7% of NDC content addresses the execution of mitigation and adaptation plans at the sector level in detail. This includes the topic group related to policies and technologies (G4) ‒ covering green energy policies and technologies (T10, T12), and waste, transport, and energy (T17) ‒ and a more particular topic group for agriculture, forestry, and other land uses (AFOLU) (G6), capturing three individual topics related to this sector.
The impacts of climate change (G5) also receive considerable attention (11% of all NDC content, particularly in terms of the climate system (T9) and the vulnerability of coastal regions (T7). The remaining minor share (4%) of NDC content concerns stakeholder involvement (G7), covering (inter)governmental organisations (T18) and gender and youth (T19).
Table 1. Aggregation of 21 NDC topics into seven topic groups
Topic group
|
Prevalence (%)
|
Constituent topics
|
Prevalence (%)
|
G1: Development
|
25.0
|
T1: Sustainable development
|
12.9
|
T2: Economic development
|
9.0
|
T15: Rural development
|
3.1
|
G2: Implementation, plans and processes
|
21.5
|
T4: International support and funding
|
6.9
|
T6: Adaptation planning
|
6.0
|
T11: Goals and commitments
|
4.2
|
T14: Project financing
|
3.3
|
T20: Action plans
|
1.1
|
G3: Mitigation targets
|
20.3
|
T3: Greenhouse gas accounting
|
8.2
|
T5: Emission scenarios
|
6.7
|
T8: Targets and reporting
|
5.4
|
G4: Policies and technologies
|
11.3
|
T10: Green energy policies
|
4.4
|
T12: Green energy technologies
|
3.9
|
T17: Waste, transport and industry
|
3.0
|
G5: Impacts of climate change
|
10.7
|
T7: Vulnerability of coastal areas
|
5.8
|
T9: Climate system impacts
|
4.9
|
G6: AFOLU sector
|
7.4
|
T13: Forests and ecosystems
|
3.3
|
T16: Agriculture resource management
|
3.1
|
T21: Agriculture mitigation
|
1.0
|
G7: Stakeholders
|
3.8
|
T18: (Inter)governmental organisations
|
2.0
|
T19: Gender and youth
|
1.8
|
Table 2. Descriptions of 21 NDCs topics
Topic
|
Prevalence (%)
|
Description
|
T1: Sustainable development
|
12.9
|
Sustainable development strategies and actions, particularly in the context of Small Island Developing States (SIDS).
|
T2: Economic development
|
9.0
|
Economic development and growth, including technological development and the legal frameworks around this.
|
T3: Greenhouse gas accounting
|
8.2
|
Emission inventory methodologies, including how parties estimate, account, and remove emissions.
|
T4: International support and funding
|
6.9
|
Financing needs to achieve mitigation and adaptation goals, including requirements of international support.
|
T5: Emission scenarios
|
6.7
|
Evaluation of emission scenarios, including conditional and unconditional reductions from baseline (BAU) projections.
|
T6: Adaptation planning
|
6.0
|
Implementation of national adaptation plans, including processes, frameworks, and stakeholder engagement.
|
T7: Vulnerability of coastal areas
|
5.8
|
Coastal vulnerability and climate impacts on water resources, including risks and measures to enhance resilience.
|
T8: Targets and reporting
|
5.4
|
Emission reduction targets and updates, including reporting mechanisms.
|
T9: Climate system impacts
|
4.9
|
Impacts on the climate system, including changes in temperature, precipitation, and their consequences.
|
T10: Green energy policies
|
4.4
|
Strategies to energy-sector emissions through government initiatives and policies.
|
T11: Goals and commitments
|
4.2
|
Frameworks, objectives, and management strategies for the fulfilment of climate goals and commitments.
|
T12: Green energy technologies
|
3.9
|
Technologies for low-carbon energy generation and energy efficiency improvements.
|
T13: Forests and ecosystems
|
3.3
|
Forest and ecosystem management, including strategies to reduce deforestation and protect biodiversity.
|
T14: Project financing
|
3.3
|
Funding mechanisms, financial instruments, and investment programs to support climate-related projects.
|
T15: Rural development
|
3.1
|
Strategies, projects, and programs for the promotion of sustainable development in rural communities.
|
T16: Agriculture resource management
|
3.1
|
Strategies to improve the resilience and efficiency of agriculture resources, particularly water management
|
T17: Waste, transport and industry
|
3.0
|
Policies and technologies to reduce emissions in the waste management, industrial, and transportation sectors.
|
T18: Inter(governmental) organisations
|
2.0
|
The institutional framework of organisations, ministries, and departments at both national and international levels.
|
T19: Gender and youth
|
1.8
|
Gender equality and youth empowerment, highlighting the need to involve women and youth in climate action.
|
T20: Action plans
|
1.1
|
Actions, objectives, and strategies to promote environmental protection and sustainable development.
|
T21: Agriculture mitigation
|
1.0
|
Mitigation measures to reduce emissions and promote sustainable production in the agriculture sector.
|
To better understand which combinations of topics countries are likely to address in their NDCs, we assess a network of topics by measuring cosine similarity and testing its significance18. The resulting graph with 21 topics and links (significant at a 5% level) is presented in Figure 1.
Topics on sustainable, economic and rural development (T1, T2 and T15), comprising the first topic group on development (G1), tend to be mentioned by the same parties; “greenhouse gas accounting” is mentioned often in conjunction with “forests and ecosystems” (T3 and T13), “project financing” with “economic development” (T14 and T2), and “rural development” with “green energy policies” (T15 and T10). In contrast, international support (T4), green energy technologies (T12) and action plans (T20) are less likely to be consistently mentioned together with other topics by many parties.
Clusters of parties to the Paris Agreement
Parties can be clustered based on how similar the composition of topics in their NDCs is. Panel A of Figure 2 shows nine clusters of parties resulting from a hierarchical clustering analysis through a dendrogram in radial form (also provided in standard form in Figure S4 of the Supplementary Information), which is presented as a map in Panel B. Panel C shows the average prevalence of a topic group for each of the clusters. Prevalences for each specific topic are provided in Figure S5 in the Supplementary Information. A detailed list of parties constituting the nine clusters is provided in Table S1 of the Supplementary Information.
Inspecting the resulting nine groups closer, we see some parallels between the clusters and various UNFCCC negotiation groups and alliances. Korea and several Umbrella Group countries (Australia, Canada, and the UK) form one cluster (C1), which is the main cluster to go into detail on “policies and technologies” (G4). Some other Umbrella Group countries (Iceland, Kazakhstan, Norway, Japan, and the US) form another cluster with the EU and a few other developed countries (C2). This cluster discusses policies and technologies substantially less and instead focuses on the “mitigation targets” (G3). This is not unexpected as some of the submissions by parties in this cluster were very brief documents primarily outlining their mitigation targets like the first NDCs by the EU and the US. Both clusters containing Umbrella Group countries dedicate relatively little of their NDCs to development (G1), implementation (G2), impacts of climate change (G5), and AFOLU (G6).
Three clusters have a strong focus in their NDCs on the topic group of development (G1). A cluster uniting 24 parties, including Brazil, Russia and some other countries from the Commonwealth of Independent States (CIS) and Like-Minded Developing Countries (LMDCs) groups, dedicate close to 40% of their NDCs to “economic development” (T2). A cluster of mainly India and Small Island Developing States (C5) focuses heavily on “sustainable development” (T1) instead of economic development while also having an above-average focus on the impacts of climate change (G5). The third aspect of development, “rural development” (T15) is primarily discussed by another cluster consisting of nations from the African Group and some other developing countries such as Mongolia (C7).
Another cluster that is more regional in nature is a group of 16 countries mostly from the Independent Alliance of Latin America and the Caribbean (AILAC) group and other Latin American countries (C6) that focuses disproportionally more on “implementation, plans and processes” (G2), the AFOLU sector (G6), and stakeholders (G7). The particular focus on the AFOLU sector is unsurprising considering the importance of the sector to these countries.
The largest cluster of 63 parties along with numerous countries from LMDCs and the Group of 77 (G77) and China (C8) is relatively balanced in terms of topics but does discuss impacts, (G5), AFOLU (G6), and stakeholders (G7) more than average. Venezuela (G4) is an outlier with its own cluster, which can largely be attributed to its updated NDC of 76,275 words being very long and discussing topics particular to its national circumstances, which are largely represented in the “action plans” (T20) topic. Finally, there is a small cluster of seven largely unconnected parties including Serbia and Uruguay but also Yemen and Afghanistan (C9) that devote almost 30% of their NDC texts to discussing the “impacts of climate change” (G5).
Topic relationships with party characteristics
We evaluated how the prevalence of topics and topic groups relates to emission, economic and other characteristics of parties (vulnerability, emissions per capita, emissions per GDP, GDP per capita, total natural resource rent as a percentage of GDP) and their pledges (NDC version number, implied ambition by 2030, conditionality of the pledge, and length of the pledge). Details of the approach and data used for all other variables are provided in the Methods section.
Table 3 presents regression coefficients significant at a 5% significance level for topic groups while Table S2 in the Supplementary Information provides the results for individual topics. We find that development (G1) and implementation (G2) are less often discussed in the updated NDC than in the first submissions, while mitigation targets (G3) and Stakeholders (G7) are more widespread in the updated NDCs. Moving away from discussions of development towards mitigation targets may reflect parties feeling less need to emphasise their broader underlying circumstances and instead provide more concise updates of their level of ambition.
Parties submitting conditional pledges are more likely to stress implementation (G2), stakeholders (G7), and development (G1) ‒ particularly sustainable development (T1). They are also more prone to discuss international support and financing (T4), which is logical given that this topic is particularly relevant to conditional pledges. Conversely, parties with unconditional pledges were more likely to focus on mitigation targets (G3) and policies (G4), particularly the topics of “green energy policies” (T10) and greenhouse gas accounting (T3), suggesting more focussed NDCs on the primary Paris Agreement goal of emission reductions.
Countries with higher emissions per capita are more likely to address development (G1) and policies (G4), while countries with higher emissions per GDP mention mitigation targets (G3) and the AFOLU sector (G6). The latter result likely arises due to our data including emissions from land use change as countries with substantial emissions from this sector tend to have high emissions in relation to their comparatively low GDP.
Countries with a higher share of rents from natural resources in their GDP discuss more development (G1), especially “economic development” (T1), potentially because climate change mitigation presents a trade-off at odds with their resource-based economies. However, countries that currently have a higher GDP per capita speak less about development (G1), instead focusing more on mitigation targets (T3) and policies (G4). This suggests that economic growth may have become less of a priority versus effective climate policy for these countries19.
Countries with a higher vulnerability index unsurprisingly address the impacts of climate change (G5) more, while dedicating less space to mitigation targets (G3). These countries may see NDCs more as an opportunity to voice the unique challenges they face from climate change to the rest of the world. The level of pledge ambition had no significant correlations at the topic level and only a small positive correlation with AFOLU (G6) at the topic group level.
As a separate test, we summed up the prevalence of the topic groups that are less directly relevant to the mitigation targets (G1 Development, G2 Implementation, G5 Impacts of climate change and G7 Stakeholders) and tested its correlation with the level of pledge ambition. The resulting correlation is positive and significant at 5% level (r=0.113, t=2.0015, df=307, p-value=0.04622). In other words, parties that spend a larger share of their NDCs on core mitigation topics such as green energy technologies and policies tend to have more ambitious emission reduction targets. However, this measure of ambition does not factor in wider context and fairness: those with higher percentage emission reductions are more likely to have greater current and historical emissions.
Finally, we find that parties writing shorter pledges typically focus on broader topics like development (G1) and mitigation targets (G3), while parties submitting longer pledges are more likely to address the rarer and specific topics related to the AFOLU sector (G6) and stakeholders (G7).