The Ambition Trap: How Overpromising on Climate Action Could Undermine the Paris Agreement Open Access

To be effective, the 2015 Paris Agreement (PA) must secure broad participation, elicit progressively ambitious pledges, and attain (reasonably) high implementation levels (Barrett 2008; Dimitrov et al. 2019). Meeting all three conditions is crucial. Any climate agreement characterized by limited participation, unambitious pledges, or widespread implementation failure will fall short of effectively addressing the climate change problem.

The PA clearly meets the broad participation criterion: only a few years after its adoption, nearly all countries had ratified the agreement. However, a central question remains: can the PA also achieve both the high levels of ambition required by its collective global temperature goals and sufficiently high implementation levels? Recent assessments provide sobering evidence on both fronts. Although the ambition levels of current nationally determined contributions (NDCs) differ significantly, in aggregate they remain inadequate for meeting the PA’s temperature goals (UN Environment Programme 2023). And for many member countries, a significant implementation gap is emerging between their NDCs and their current emissions trajectories (Fransen et al. 2023; Intergovernmental Panel on Climate Change 2023).

These observations raise fundamental questions about international regime effectiveness that have been subject to long-standing scholarly debate (Barrett 1999; Bernauer et al. 2016; Downs et al. 1996; Gilligan 2004; Johns 2014). A central theoretical conjecture posits that more ambitious requirements for behavioral change in international agreements entail reduced implementation likelihood (see, e.g., Johns 2014). This conjecture assumes that more ambitious commitments are also more costly and therefore more challenging to fulfill. Applied to the PA, this argument implies that more ambitious NDCs are less likely to be implemented (ceteris paribus). However, some scholars have recently claimed that the empirical relationship is the opposite: that the most ambitious NDCs are also the most credible (Victor et al. 2022).

We contribute novel empirical evidence to unresolved debates regarding a potential ambition–implementation trade-off under the PA (e.g., Dimitrov et al. 2019; Fransen et al. 2023; Rogelj et al. 2023; Stankovic et al. 2023; Tørstad 2020; Victor et al. 2022). We are motivated by the observation that while many states face strong incentives to pledge ambitiously (e.g., the possibility of international reputational gains and of satisfying domestic interest groups), the historical record suggests that countries failing to meet their NDC pledges are unlikely to face adverse consequences. Devoid of enforcement mechanisms, the PA risks an “ambition trap” whereby policymakers are tempted to pledge ever more ambitious targets without being willing or even able to ensure these targets’ implementation. Thus we draw attention to a possible downside of ambition: unrealistically ambitious pledges might undermine the PA’s long-term credibility.

We consider two questions. First, does a trade-off exist between ambition and implementation likelihood? Second, if such a trade-off exists, how does the public prioritize between ambition and implementation likelihood? We provide empirical evidence, first, that a trade-off does indeed exist between NDCs’ ambition levels and the likelihood of their implementation and, second, that the public is far more concerned with international climate agreements’ implementation prospects than with their ambition levels. An important policy implication is that policymakers could both cater to public opinion and help safeguard the PA’s long-term credibility by keeping their NDCs’ ambition levels realistic.

The remainder of this article is organized as follows. The following section reviews literature on the depth–breadth trade-off in international environmental agreements. The third section discusses the theoretical case for a trade-off between ambition in pledges and implementation prospects, including the role of implementation capacity. Next, we present our data and methods. In the fifth section, we provide cross-country empirical evidence showing that the implementation likelihood of NDCs is negatively associated with their ambition levels. The following section presents results from a randomized conjoint experiment suggesting that the public prioritizes implementation prospects over ambition. The final section concludes by highlighting policy implications for the next round of NDCs (due in 2025) and suggesting avenues for future research.

Effective international environmental agreements must be both broad (i.e., all major countries must participate) and ambitious (i.e., deep/stringent). These two effectiveness requirements are arguably subject to a trade-off: more ambitious obligations tend to limit participation, while more shallow stipulations tend to broaden it. In either case, the result is limited effectiveness in problem solving (Barrett 2008; Barrett and Stavins 2003). The Kyoto Protocol (KP) may serve as an example. Its stringency level was widely considered too high to garner sufficient participation for effective global climate action. Stringency in the case of the KP was interpreted as relating to legal form and what is now called a top-down approach: legally binding, internationally agreed-upon mitigation obligations. The emissions reduction requirements themselves were arguably not very ambitious (5% below 1990 levels on average). Nevertheless, the KP failed to gain support among a broad enough set of states, especially among major emitters (e.g., the United States, Canada, and Japan either refused to participate or withdrew later). This depth–breadth trade-off has been subject to extensive scholarly discussion; it has been nuanced on theoretical grounds (Gilligan 2004) and challenged on empirical grounds (Bernauer et al. 2016; Spilker and Koubi 2016).

For the PA, any trade-off between participation and ambition is less relevant. Because the PA allows each participating country to determine its own ambition level, a country can simply adjust its ambition level to make participation acceptable. However, the PA’s architecture highlights the importance of a second potential trade-off, that between a country’s individual ambition and the degree to which its pledges are met (i.e., implemented). This demanding balancing act is further complicated by the PA’s progression principle, which mandates the gradual ratcheting up of ambition. The PA’s voluntary, bottom-up nature (pledge and review) thus shifted the interpretation of “depth” toward individual climate policy ambition expressed in NDCs. It also sharpened concerns about credible implementation (e.g., Bang et al. 2016; Dimitrov et al. 2019; Dubash 2020; Tørstad 2020; Victor et al. 2022). According to the “effectiveness trilemma,” a climate agreement must simultaneously attract broad participation, muster high ambition, and achieve reasonably high compliance (or implementation) rates to significantly reduce global emissions (Barrett 2008; Dimitrov et al. 2019).

Many scholars hold that increasing the ambition of climate pledges makes them more difficult to implement (e.g., Barrett 1999; Dimitrov et al. 2019; Downs et al. 1996; Fransen et al. 2023). Early work considering this trade-off includes Downs et al. (1996), who argue that international treaties’ high compliance rates (despite little international enforcement) result from these treaties’ shallowness: implementation of international commitments is often easy because they rarely require member states to do much more than they would have done anyway. Theoretically, the ambition–implementation trade-off in international agreements is relatively straightforward when ambition is understood as deep behavioral change. For example, Johns (2014) shows formally how strengthening the depth of international trade cooperation lowers the likelihood of full implementation. If this trade-off holds for benefit-sharing agreements like those concerning trade, it might be expected to be even more forceful for burden-sharing agreements like those addressing climate change mitigation.

Recent research offers preliminary support for the existence of an ambition–implementation trade-off, including both anecdotal accounts (Stankovic et al. 2023) and more systematic, empirical evidence (Fransen et al. 2023). Fransen et al. argue that countries’ implemented emissions reductions fall short of their pledges for reasons related to policy adoption (e.g., misjudging the domestic ability to adopt necessary policies) as well as for reasons related to policy outcomes (e.g., faulty policy design or lacking enforcement capacity). They show that countries with a higher ambition (e.g., the Philippines, Norway, Ethiopia, Kazakhstan) tend to face a wider implementation gap.

However, other scholars doubt the existence of such a trade-off. Analyzing expert perceptions, Victor et al. (2022) find that the most ambitious national pledges also have the highest chance of implementation. Importantly, these experts assessed their own countries’ ambition relative to the economic strength of the country concerned, rather than based on global scientific measures of what is needed to achieve the PA’s temperature goals. “Ambition relative to national economic circumstances” presents very different demands than “ambition relative to a quantified goal.” The former implicitly takes into account the implementation capacity of the country under consideration. If this capacity is low, a mitigation commitment can be ambitious relative to national circumstances despite being far from sufficient for achieving the 1.5°C/2°C goal. Moreover, the fact that Victor et al.’s experts are predominantly European or American also likely affects the findings concerning a potential trade-off. The EU and the United States presumably have high implementation capacity compared to the rest of the world.

In sum, existing literature is divided on whether an ambition–implementation trade-off exists in NDCs. Moreover, some empirical studies (e.g., Fransen et al. 2023; Victor et al. 2022) indicate that implementation capacity affects whether the trade-off materializes. We argue theoretically that an ambition–implementation trade-off in NDCs exists within countries (for a given implementation capacity level) and may or may not exist across countries (when the implementation capacity level varies). We expect the trade-off across countries to be more strongly apparent when implementation capacity is accounted for than when it is not. We then show empirically that a trade-off also does exist across countries, both when implementation capacity is controlled for and not. Finally, we provide empirical evidence that citizens in five democracies prefer higher implementation likelihood over higher ambition.

We first theorize the case for an ambition–implementation trade-off under the PA. Next, we consider how policymakers must navigate this trade-off when determining the ambition level of their countries’ NDCs.

In this article, we use the term implementation to signify what most previous (international relations) literature calls “compliance.” Compliance can be defined as the fulfillment of a legal obligation, rule, or norm (Raustiala and Slaughter 2002, 539). Strictly speaking, implementation of NDCs does not equal treaty compliance under the PA, as this agreement does not contain a legally binding obligation to fulfill pledges. Nevertheless, the international compliance literature is insightful for the question of NDC implementation because it addresses the same problem for legally binding treaty obligations.

Implementation can be understood as a three-stage process (see Fransen et al. 2023). First, the PA parties formulate and submit their NDCs. Second, governments establish or adjust existing policies that pursue the targets formulated in their NDCs. This can include laws that formalize national targets for emissions reductions but also other measures related to knowledge generation, technological development, and adoption, adaptation, and resource provisioning. The third, extended stage involves policy execution, monitoring and evaluating impacts, and possibly policy revisions.

Throughout all three stages, the state’s resources and capabilities are essential for successful implementation. Hence, by implementation capacity, we mean a state’s ability to formulate, enact, and execute policies that align with its climate targets, including the capability to monitor, evaluate, and revise these policies as needed to ensure target achievement. Of course, that a state possesses a high implementation capacity does not necessarily mean that its government is (always) willing to use it. Nevertheless, accounting for countries’ varying levels of implementation capacity can help reconcile the seemingly contradictory results of Victor et al. (2022) and Fransen et al. (2023) on whether a trade-off exists. Countries can pledge both ambitious and credible targets if their implementation capacities are sufficiently strong. Figure 1 shows how implementation likelihood (credibility) is related to ambition, depending on implementation capacity. For any given country (A, B, or C) and a given level of capacity, a more ambitious pledge is always less credible than a less ambitious pledge. Thus each of these countries faces a trade-off between ambition and implementation likelihood. Nevertheless, if we compare across countries, a more ambitious pledge might well be more credible than a less ambitious pledge, as Victor et al. (2022) found.

For example, if each country were to pledge the combination of ambition and credibility located at the intersection between its trade-off line and the arrow indicating capacity (red dots), then A’s pledge would be both more ambitious and more credible than B’s, which in turn would be both more ambitious and more credible than C’s. In contrast, if the three countries were to pledge as indicated by the gray dots in Figure 1, the relationship between ambition and implementation likelihood would be negative even across the three countries. Thus it is an empirical question if the association between ambition and implementation likelihood is positive or negative across countries. It also follows that a cross-country trade-off is more likely to manifest if the analysis controls for countries’ varying degrees of implementation capacity.

Figure 1 illustrates that improving a country’s implementation capacity can augment credibility without compromising ambition and also facilitate higher ambition without reducing credibility.

Each PA member self-determines its ambition level but must submit a new and more ambitious NDC every five years (Article 4.3). This submission cycle is complemented by “ambition mechanisms” like the Enhanced Transparency Framework, which reviews the implementation of national pledges, and the Global Stocktake, which monitors collective progress toward the global goals. Country-level information is subject to international review and enables peer pressure, including “naming and shaming” (Dannenberg et al. 2023). However, while not submitting an NDC constitutes a case of treaty noncompliance, implementation failure does not, and no legal consequences follow if actions do not match pledges. Thus pledges’ credibility largely depends on how policymakers balance ambition and implementation prospects in the NDC formulation process.

Given that higher ambition renders target achievement more challenging, policymakers face opposing incentives when determining NDC ambition levels. On one hand, higher ambition helps in achieving the PA’s collective goals. It might also secure private international cooperative benefits, for example, by spurring other countries to reciprocate in kind or on other issues or by enhancing one’s international status (Keohane and Oppenheimer 2016; Weikmans et al. 2020). Domestic political factors might also tempt policymakers to pledge ambitiously. Such factors include public opinion (Bechtel et al. 2022; Schaffer et al. 2022), level of democracy (Tørstad et al. 2020), the relative influence of green and fossil economic stakeholders in the policymaking process (Aklin and Mildenberger 2020), and vulnerability to climate change (Tørstad et al. 2020; Victor et al. 2022).

On the other hand, pledging less ambitiously makes fulfillment easier and hence curtails the risk of naming and shaming triggered by implementation failure (Dannenberg et al. 2023; Stankovic et al. 2023). Clearly some of the domestic factors mentioned, such as public opinion, form of governance, and the influence of green/fossil economic interests, might also play into policymakers’ prioritization of NDC implementation prospects. For example, policymakers in democracies face stronger accountability mechanisms, including public and civil society pressure (e.g., Dannenberg et al. 2023; Victor et al. 2022) that might incentivize them to weigh the achievement of pledged targets more strongly. Yet, because this article focuses primarily on whether an overall trade-off exists between ambition and implementation likelihood in NDCs, we abstract from a fine-grained consideration of all the potential political, economic, and institutional factors that can affect ambition and implementation likelihood at the domestic level. In our empirical analysis, we nonetheless control for several domestic country characteristics to reduce the possibility of spurious effects. These are listed and justified in the Methods section.

Scholars disagree on whether states, when formulating their NDCs, find the benefits of high ambition more (or less) appealing than those of easy implementation. Some argue that the potential benefits of ambitious pledging lead states to adopt overly ambitious mitigation targets (Rowan 2019; Victor et al. 2017). Others propose that states are reluctant to make imprudent climate pledges even absent formal enforcement (Tørstad and Wiborg 2024; Victor et al. 2022). We argue that the incentives to pledge ambitiously, combined with low costs of implementation failure, might tempt policymakers to set targets that are unlikely to be met. When determining pledges, policymakers face an intertemporal trade-off (Stankovic et al. 2023). Pledging ambitiously now is cheap, might entail reputational benefits, and might enhance the PA’s perceived effectiveness. Developing countries can also link enhanced ambition to stronger claims for international support. In developed countries, increased ambition might sit well with domestic audiences. However, it also increases the likelihood of implementation failure and subsequent critique. To enhance the chance of reelection, policymakers must appease myopic constituents and voters and therefore tend to suffer from severe short-term bias (e.g., Bonfiglioli and Gancia 2013). Thus they will likely weigh near-term consequences higher than those arising years ahead. The NDCs containing targets for emissions reductions by 2030 were submitted (in updated form) already in 2021, and only a few policymakers in power in 2021 were likely expecting to remain so in 2030. Moreover, even those who did would likely discount future costs and benefits and therefore place more weight on immediate audience costs than on future audience costs. Thus we might expect NDCs to be biased toward high ambition and hence apt to falling significantly short of full implementation (Bang et al. 2016).

If all pledges were credible, irrespective of their ambition levels, a higher ambition would unreservedly enhance treaty effectiveness. However, widespread nonimplementation of pledged actions might severely undermine confidence in the PA both in the international community and in the public. A long-term collaborative project, the PA depends on broad and sustained confidence. To preserve this confidence, satisfactory implementation levels are essential. Thus, unless ambition levels are kept reasonably realistic, other countries, business leaders, and the public may not trust that future pledges will be followed up by adequate policy measures that ensure implementation. Many current pledges are far from satisfying this requirement (e.g., Rogelj et al. 2023). Indeed, even in a country like Norway, fewer than 10 percent of the respondents in a 2023 survey believed that their government’s 2030 climate goals would be met, and most experts agree (Nettavisen 2023). Perhaps even worse, leading Norwegian politicians are equally skeptical (NRK 2023). Pledges distrusted by the public, experts, and leading policymakers alike are unlikely to be helpful for progress on climate action or for confidence in the PA.

So far, we have argued that a trade-off between ambition and implementation prospects exists within countries and possibly also across countries. We now proceed with a twofold empirical analysis. First, based on analysis of current NDCs, we test whether the theorized trade-off manifests at the cross-country level, both with and without controlling for implementation capacity. Second, we study how policymakers responsive to the public’s preferences can navigate this trade-off within countries. Using a conjoint experiment, we examine which factor in the effectiveness trilemma—participation, ambition, or implementation (compliance)—has the greatest effect on public support for a climate agreement.

On the basis of the preceding theorizing, we expect to observe a negative relationship between ambition and implementation likelihood in NDCs. Our theory suggests that this relationship holds true within countries, while it may or may not exist across countries, depending on the selection of countries and their implementation capacities (see Figure 1). Testing the within-country trade-off proposition empirically would ideally require data on countries’ NDC ambition and implementation levels over several pledge cycles. Unfortunately, such data will not become available for many years. Our analysis is therefore limited to investigating whether a trade-off exists across countries. Following the theoretical discussion, we do this both with and without controlling for implementation capacity. Specifically, we conduct multiple regression analysis, where we couple cross-sectional data on NDC ambition from Robiou du Pont and Meinshausen (2018) with data on implementation probability from Liu and Raftery (2021). Although inapt for establishing causal effects, this predictive inference design is useful for testing whether ambition and implementation likelihood are associated in NDCs.

Our dependent variable is Liu and Raftery’s (2021) NDC implementation probability metric. This metric estimates the probability that countries will meet their NDCs using a country-specific version of the Kaya identity (CO₂ emissions = Population × GDP per capita × CO₂ emissions per unit of GDP) to develop a probabilistic forecast of CO₂ emissions. A statistical model of the last two factors of the identity is combined with probabilistic projections of population from the UN (Raftery et al. 2017). The resulting emissions scenarios can be thought of as projections, assuming that the general range of trends of the past thirty years continues. The projected trends do not simply represent Business-as-usual (BAU), because past and current policies are implicitly accounted for. On the other hand, future policy changes are not incorporated. Substantiating the point that the PA might face significant credibility risks concerning NDC implementation, Liu and Raftery (2021) find that the median probability of NDC implementation is only 35 percent, with most major emitters at the lower end of the spectrum. For example, the United States is given only a 2 percent chance of reaching its target, China’s probability is 16 percent, Japan’s 10 percent, Germany’s 13 percent, and France’s 2 percent. For all these countries, major policy changes are needed to achieve NDC targets. For our purposes, absolute probabilities are less important, as only the relative values across countries affect our results.

We regress Liu and Raftery’s (2021) implementation probability metric on Robiou du Pont and Meinshausen’s (2018) “pledged warming” ambition metric (hereinafter RdPM), which measures countries’ ambition levels relative to a “fair” distribution of emissions. Like any quantification of NDC ambition based on global goals, RdPM is based on certain normative criteria for how mitigation efforts should be distributed between countries. RdPM incorporates three fairness principles: capability to pay (GDP per capita), historical responsibility (equal cumulative per capita emissions), and equality (equal per capita emissions). For each country, they calculate what the global temperature impact of its NDC would be, assuming that global emissions are distributed according to the fairness principle most lenient to the country concerned (see Robiou du Pont and Meinshausen 2018). The data scores range from 1.2°C warming (most ambitious) to over 5.1°C (least ambitious). It follows from the fairness principles mentioned earlier that countries with low economic capabilities, low historical responsibilities for causing climate change, and low per capita emissions are assigned higher NDC ambition scores ceteris paribus. For example, Ethiopia’s NDC is evaluated as 1.2°C compatible, and India’s is aligned with 2.6°C of warming, while the NDCs of the EU, the United States, and China lead to warmings of 3.2°C, 4°C, and over 5.1°C, respectively. We invert the scale to facilitate interpretation so that higher scores mean higher ambition, with a range from 0 to 3.9.

Unlike most other ambition assessments, the RdPM metric is previously peer reviewed, covers nearly all countries having submitted NDCs, and avoids making counterfactual assumptions about BAU emissions. For our purposes, benchmarking ambition relative to fair distributions, rather than to BAU or current emissions, has both methodological advantages and a stronger normative justification (e.g., Kartha et al. 2018). If measured as deviation from BAU or a baseline historical emissions level, ambition would risk being negatively related to credibility by construction, because large deviations are ceteris paribus less likely than small deviations.

We note that our primary ambition metric (RdPM) is not completely unaffected by baseline emissions (in this case, 2010 levels), because two of the underlying fairness principles (capability to pay and equality) are applied with a thirty-year convergence period from 2010 emissions levels to the fair distributions. However, because we are looking at ambition regarding emissions in 2030, the benchmarks will be two-thirds of the way from baselines to fair distributions, meaning that the influence of baseline levels will be strongly diluted. Furthermore, for two of the fairness principles (capability and responsibility), the eventual fair distributions are negatively related to 2010 distributions, because countries with higher emissions and income in 2010 are assigned smaller future emissions allocations.

As a robustness check, we rerun the analysis with an alternative measure of ambition: pledged per capita emissions reductions for the period 2015–2030 (using data from Meinshausen et al. 2022). While this second measure does not account for “fair” mitigation contributions in a corresponding vein as the RdPM metric, it offers an intuitive and transparent measure of countries’ pledged efforts. Another drawback for our purposes is that unlike Liu and Raftery (2021) and Robiou du Pont and Meinshausen (2018), it incorporates updates in NDCs during 2020 and 2021.¹ To facilitate interpretation, we also invert and standardize this ambition variable so that higher scores mean higher ambition (see details in the Supplementary Material).

Control variables are introduced sequentially to the regression models. First, we control for implementation capacity as per discussions in the Theory section. Our primary operationalization here is Hanson and Sigman’s (2021) state capacity index, which combines twenty-one different indicators related to states’ extractive, coercive, and administrative capacities. Additionally, we include countries’ levels of GDP/capita,² which is a cruder but nonetheless relevant indicator of states’ capacity to implement their NDCs. Subsequently, we introduce two other variables that might affect both NDC ambition and compliance probability, to mitigate omitted variable bias. The first is form of governance. Democracies are commonly theorized to be more compliant with international agreements than are autocracies, owing to factors such as stronger institutional accountability mechanisms, the inclusion of civil society actors, and higher responsiveness to public demand (see, e.g., Baker 2023). Relatedly, existing literature indicates that climate negotiators from democracies have systematically different views on the PA’s “soft” compliance mechanisms (Dannenberg et al. 2023) than do negotiators from autocracies. Moreover, democracies tend to generate stronger climate policy outputs than autocracies on average (Lindvall and Karlsson 2024), which is relevant for the implementation probability of NDCs. We use the V-Dem polyarchy index to measure form of governance in the main analyses,³ but robustness tests (Supplementary Material IV) show unchanged results if we use Polity V scores instead.⁴ Finally, we control for EU membership because EU members have joint climate targets and compliance mechanisms. Supplementary Material I provides operationalizations, data sources, descriptive statistics, and correlations between the independent variables.

In the regression where we use our alternative ambition metric—pledged per capita emissions reductions—as the dependent variable, we include two additional controls: greenhouse gas (GHG) emissions per capita (2015) and historical changes in GHG emissions per capita (1990–2015) (Gütschow et al. 2021). The rationale is to avoid a grandfathering bias in the data, which could arise because countries start from very different emissions levels at the outset of the NDC period. Countries with (much) higher emissions levels in 2015 would in general find it easier to reduce emissions than would countries with lower emissions levels.

To ensure that our results do not depend on our choice of covariates, we run several robustness tests where we introduce additional controls. These tests show substantively unchanged results when we control for fossil fuel rents, vulnerability to climate change, regime corruption, the World Governance Indicators’ Government Effectiveness Index (Kaufmann et al. 2010), countries’ tax revenues as percentage of GDP, and Polity V scores (see Supplementary Material IV). Furthermore, we do not find indications that the relationship between ambition and implementation probability is different across countries with different levels of implementation capacity, which we test for using both an interaction term (Table S10 in Supplementary Material IV) and split-sample analysis (Table S11 in Supplementary Material IV). However, all analyses rely on Liu and Raftery’s implementation probability metric as the dependent variable, as existing literature offers no alternatives.

To test which factor in the effectiveness trilemma garners most support among the public, we conducted a randomized conjoint experiment on a balanced sample of 766 respondents from five major democracies—Germany, Mexico, South Africa, the United Kingdom, and the United States.⁵ Participants were first provided the following information:

The participants were then presented choice cards, each featuring two different hypothetical climate agreements, and asked to select their preferred agreement. The hypothetical climate agreements had three attributes reflecting the elements of the effectiveness trilemma (participation, ambition, and implementation/compliance). For simplicity and clarity, we used the term stringency instead of ambition to refer to the depth of the hypothetical agreements’ mitigation targets.⁶

Each attribute had two levels: whether the subject’s home country participates in the agreement; whether the agreement requires a 20 percent cut or a 40 percent cut in GHG emissions by 2030; and whether the mitigation target of the agreement concerned has a 20 percent or a 50 percent probability of being reached. The stringency levels (20% and 40%) were selected because they are among the most common NDC headline targets: 40 percent emission cuts is the mode of headline NDC targets, whereas 20 percent is the second most common headline target (Rowan 2019).⁷ The implementation likelihood figures (20% and 50%) were selected based on Liu and Raftery’s (2021) finding that the median probability of full NDC implementation is 35 percent. Our two implementation likelihood scenarios are thus 15 percentage points higher and lower, respectively, than Liu and Raftery’s projected median probability of full implementation.

The order of attributes and the information in the attribute levels of the hypothetical agreements were randomized for each choice card, with eight different possible attribute level combinations. The participants were asked to choose between pairs of agreements that contained various combinations of attribute levels, as illustrated in the example choice cards in Figure 2. The randomization of information shown in the choice cards ensured statistical independence across both attributes and attribute levels, preventing any systematic pattern from linking them within the sample. A main benefit of this randomized design is that it allows us to isolate and compare the causal effects of our three attributes of interest simultaneously. While respondents may have held prior beliefs about how the attributes are correlated (e.g., that higher stringency leads to lower implementation likelihood), randomizing the information shown in attribute levels for each choice card allows us to isolate the separate effect of each attribute and estimate unbiased causal effects (see Hainmueller et al. 2014, 26). To enhance the precision of our estimates, we asked all participants to complete the conjoint task twice, resulting in a total of 1,532 observations (there were no statistically significant spillover effects between the tasks).

Participants’ agreement choice—a binary indicator for whether a given agreement was preferred or not preferred—constitutes the dependent variable in the analysis. We fit generalized linear models to calculate average marginal component effects (AMCEs) of each agreement attribute (Hainmueller et al. 2014). The AMCEs represent a measure of the heightened probability that the participants would endorse a particular agreement, brought about by modifications to specific attributes of said agreement (e.g., heightened ambition). When assessing cross-country differences, we report marginal means—which are average outcome values conditional on different levels—instead of AMCEs (see Leeper et al. 2020). Finally, we note that the elicited preferences in the conjoint experiment may have limited generalizability. While balanced in terms of nationality and gender, the respondent sample is not generally representative of the general populations in the respective countries. Supplementary Material III provides demographic information on survey respondents.

Our findings indicate that a trade-off exists between ambition and implementation likelihood at the cross-country level. Figure 3 shows the conditionally adjusted predictions of how NDC ambition is associated with implementation probability, controlling for state capacity, form of governance, GDP per capita, and EU membership. The y-axis shows implementation probability (data from Liu and Raftery 2021), while the x-axis shows NDC ambition (data from Robiou du Pont and Meinshausen 2018). The predicted implementation probability decreases from approximately 65 percent for the least ambitious pledges to approximately 15 percent for the most ambitious pledges. This finding provides strong support for the hypothesis that a trade-off exists: the more ambitious the pledges are, the more challenging they are to implement (ceteris paribus).

Table 1 provides the regression results underlying Figure 3. The association between ambition and implementation probability is negative and statistically significant at the 1 percent level across the five different specifications of different control variables. A particularly strong result is that the trade-off between ambition and implementation probability is detected even when not controlling for implementation capacity. As expected from our theoretical discussion, the trade-off is strengthened when the proxies for implementation are introduced (models 2 and 3). However, contrary to theoretical expectations, these variables themselves do not have a positive effect on implementation probability. To the contrary, GDP per capita has a consistently negative effect, perhaps due to mechanisms other than implementation capacity. The introduction of additional controls (EU membership and level of democracy) does not substantially affect results.

With the alternative ambition metric (Table 2), the trade-off remains significant across specifications. Here it is also unaffected when controlling for state capacity. The predicted implementation probability decreases from approximately 80 percent for the least ambitious pledges to approximately 20 percent for the most ambitious pledges (Figure 4). No other variables are significant in these models. Overall, our statistical analyses indicate the existence of a robust, negative, and substantively strong relationship between NDC ambition and implementation probability.

In the previous section, we provided robust empirical evidence that a trade-off between ambition and implementation probability exists in states’ climate pledges. The existence of such a trade-off means that policymakers face a difficult choice, namely, whether to prioritize high ambition or to give precedence to ensuring that pledges will be fulfilled. Citizens’ support is central to implementing climate policies (Huber et al. 2020a). Thus public opinion contributes significantly to shaping policy choices, particularly in democratic countries (Anderson et al. 2017) and especially regarding climate policy, which often has direct and easily observable impacts for citizens (Huber et al. 2020b). Policymakers will therefore likely take into consideration which factor the public (i.e., the electorate) prefers them to prioritize (see, e.g., Schaffer et al. 2022). Hence we now proceed to testing what importance the public attaches to ambition, relative to implementation prospects.

To the best of our knowledge, no one has previously tested whether members of the public prefer high climate ambition over high implementation likelihood, or vice versa. If the public values implementation likelihood more than it values ambition, decision makers could potentially benefit politically by pledging realistically, rather than by making very ambitious pledges that are unlikely to be met.

Figure 5 presents AMCEs of our participation, stringency, and implementation likelihood attributes. In this context, an AMCE represents the average effect of an attribute (participation, stringency, implementation likelihood) on the probability that a given hypothetical climate agreement will be selected, with the average defined over the distribution of the remaining attributes across repeated samples (see Hainmueller et al. 2014). For example, in Figure 5, the AMCE for participation is 0.2, which means that the probability that a subject prefers an agreement in which the subject’s own country participates increases by 20 percentage points compared to an agreement in which the subject’s own country does not participate (with stringency and implementation likelihood held at their average values).

Figure 5 shows that all three attributes have a causal effect on people’s selection of hypothetical climate agreements. The data indicate a preference for participation, for higher stringency, and for higher implementation probability. Thus, as expected, the participants in the experiment value each component in the effectiveness trilemma. More interestingly, implementation likelihood has the strongest effect of the three and, importantly, a considerably stronger effect than stringency has. Increasing the implementation likelihood from low (20%) to medium (50%) raises the probability that a respondent prefers a given climate agreement by 25 percentage points. In contrast, the corresponding effect of increasing stringency from low (a 20% cut in emissions) to medium (a 40% cut) is only 6 percentage points. It should be noted that we elicited preferences over only a limited range of stringency and implementation likelihood levels, grounded in estimates of current real-world values (see the Methods section). Because preferences are not necessarily linear functions of these variables, the relative valuation of stringency and implementation likelihood could be different in scenarios in which stringency and implementation likelihood take other values. Future research should therefore present respondents with a wider range of scenarios.

Figure 6 disaggregates the experimental results by nationality, using marginal means instead of AMCEs (in line with Leeper et al. 2020). In our analysis, a marginal mean can be understood as the level of favorability toward a hypothetical climate agreement that has a particular attribute level (e.g., participation), marginalizing across all other attributes (Leeper et al. 2020). These estimated means can be interpreted as probabilities. For example, the marginal mean for the attribute level “20 percent implementation likelihood” among Mexican respondents is 0.29, which indicates that the probability that Mexican respondents in our experiment select a hypothetical climate agreement with this particular attribute level is 0.29 on average. The vertical line (at 0.5) in the figure represents the average value of all the marginal means estimates, which implies that values above 0.5 increase the predicted probability of choosing a particular level, while values below 0.5 indicate decreased probability.⁸ Overall, the effects discussed are largely consistent across respondents’ nationalities, with only minor disparities. The Mexican sample exhibits stronger participation and implementation preferences than the others do, and the positive effect of ambition is statistically significant only in the UK and US samples.

In Supplementary Material II, we report the AMCEs of stringency and implementation likelihood conditional on country participation. Whether a respondent’s home country participates in the agreement does not significantly alter the effects of stringency and implementation likelihood on climate agreement preferences. In summary, our experimental results indicate that implementation likelihood has a considerably stronger causal effect on citizens’ preferences concerning climate cooperation than ambition has. Although the experimental setting is highly stylized, this finding suggests that policymakers might expect more public support for their climate policies if they pick realistic targets for emissions reductions. More public support might strengthen the long-term project of climate cooperation through the PA, which is unlikely to succeed unless the NDCs are considered reasonably credible by the public and business leaders alike.

This article contributes to the growing body of scholarship on the effectiveness of the PA’s ambition mechanisms by addressing the unresolved question of whether a trade-off exists between ambition and implementation probability in countries’ NDCs (Dimitrov et al. 2019; Fransen et al. 2023; Rogelj et al. 2023; Stankovic et al. 2023; Tørstad 2020; Victor et al. 2022). We identify an ambition–implementation trade-off both when controlling for implementation capacity and when not. The latter finding is particularly noteworthy, as our theory suggested that variations in implementation capacity might eliminate the trade-off across countries. Moreover, our findings contradict the conclusion from Victor et al.’s (2022, 1) expert survey that “countries making the boldest pledges are also making the most credible pledges.” Our results are robust across various model specifications. Based on our preferred metrics, the implementation likelihood is only 65 percent even for the least ambitious pledges and as low as 15 percent for the most ambitious pledges. Consequently, several current NDCs may risk severe implementation failure in 2030, despite an existing ambition gap in many countries. These findings lend support to some scholars’ concerns about constant calls for ever-increasing ambition in international climate cooperation (Dubash 2020).

Using a conjoint experiment, we also find that although citizens value all elements of the “effectiveness trilemma,” they are far more concerned about implementation likelihood than about ambition. Thus the current preoccupation with ambition, coupled with insufficient focus on implementation, could be misguided. Unrealistic pledges may perpetuate rather than mitigate the problem of climate change (Finnemore and Jurkovich 2020) and could have serious adverse consequences for international and domestic climate policy in the long term (Stankovic et al. 2023).

At the international level, widespread nonimplementation resulting from unrealistically ambitious pledges could undermine the PA’s credibility and its ability to spur reciprocity-based climate cooperation between states. At the domestic level, overly ambitious NDCs and long-term goals (e.g., net-zero targets) might serve as substitutes for immediate climate action. As former Norwegian prime minister Jens Stoltenberg remarked, “when we approach the deadline and see that things become difficult, we stop talking about the current target and instead agree on new targets even further ahead. In this way, we have deceived ourselves repeatedly” (NRK 2018). Such use of ambitious pledging to distract from inadequate implementation might produce “aspirational fatigue,” causing citizens to become apathetic toward climate politics and undercutting the logic of the PA by shifting attention from current action to distant targets (Dubash 2020; Finnemore and Jurkovich 2020).

The encouraging news from our conjoint experiment is that politicians could likely enhance public acceptance of climate policies by prioritizing concrete climate action over aspirational rhetoric. For example, in NDC processes, they could emphasize implementability (along with ambition), add implementation capacity assessments, include monitoring programs in domestic policies to track implementation challenges, and frame political messages around implementation plans.

Future research should seek to address shortcomings in our research design. For example, our first study is correlational, and its results depend on the validity of the ambition and implementation probability metrics. In addition, it is worth looking into how implementation likelihood depends on the extent to which implementation entails cobenefits. Substantial cobenefits may be expected to lower the (net) implementation costs and therefore enhance the political support for ambitious targets (e.g., Mayrhofer and Gupta 2016). Moreover, our conjoint experiment was based on specific wordings and incorporated a limited range of scenarios. Future research should evaluate different sets of information treatments, including with implementation probabilities above 50 percent, which likely represents a crucial threshold in beliefs regarding the likelihood of success or failure. It would also be interesting to investigate a larger range of climate targets (e.g., net-zero targets), people’s conceptions of ambition, and to what extent people’s beliefs about ambition and implementation affect voting behavior. Finally, future work should consider the possibility that more ambitious pledges might have the effect of boosting climate action (domestic effort) more than less ambitious ones, even if a country fails to fully implement its NDC. In other words, an ambitious, unfulfilled pledge might lead to more impacts on the ground (e.g., emissions avoided) than a less ambitious pledge that is fully achieved.

This work was supported by the Research Council of Norway (project 324468) and the BI Norwegian Business School. We are grateful to Yves Steinebach, David Victor, three anonymous reviewers, and the GEP editors for helpful comments. We also thank Peiran Liu for sharing data.