What drives national adaptation? A global assessment
That the climate is changing and societies will have to adapt is now unequivocal, with adaptation becoming a core focus of climate policy. Our understanding of the challenges, needs, and opportunities for climate change adaptation has advanced significantly in recent years yet remains limited. Research has identified and theorized key determinants of adaptive capacity and barriers to adaptation, and more recently begun to track adaptation in practice. Despite this, there is negligible research investigating whether and indeed if adaptive capacity is translating into actual adaptation action. Here we test whether theorized determinants of adaptive capacity are associated with adaptation policy outcomes at the national level for 117 nations. We show that institutional capacity, in particular measures of good governance, are the strongest predictors of national adaptation policy. Adaptation at the national level is limited in countries with poor governance, and in the absence of good governance other presumed determinants of adaptive capacity show limited effect on adaptation. Our results highlight the critical importance of institutional good governance as a prerequisite for national adaptation. Other elements of theorized adaptive capacity are unlikely to be sufficient, effective, or present at the national level where national institutions and governance are poor.
That the climate is changing and societies will have to adapt is now unequivocal, with adaptation becoming a core focus of climate policy (Khan and Roberts 2013). The establishment of the Green Climate Fund is expected to become an important source of adaptation finance for low and middle income countries in the coming years. Similarly, climate adaptation has been articulated as an important component of broader sustainable development goals, and has become an emerging priority in industrialized nations (Ford and Berrang-Ford 2011). As adaptation has become a core element of climate policy and funding has been mobilized, our understanding of the challenges, needs, and opportunities for adaptation has advanced but remains limited (Adger et al. 2007; IPCC 2012). There is a long history of research conceptualizing the key determinants of vulnerability and identifying predictors of adaptive capacity (Smit and Wandel 2006; Gupta et al. 2010; Brooks et al. 2005) and more recently, examining barriers to adaptation (Biesbroek et al. 2013; Moser and Ekstrom 2010; Bauer et al. 2012; Clar et al. 2013; Dow et al. 2013) and tracking adaptation planning (Lesnikowski et al. 2011, 2013, 2013; Berrang-Ford et al. 2011; Ford et al. 2011; Tompkins et al. 2010; Biesbroek et al. 2010; Gagnon-Lebrun and Agrawala 2007) and practice (Berrang-Ford et al. 2011; Ford et al. 2013; Preston et al. 2009). This research has advanced our understanding of what factors are likely to contribute to adaptive capacity. Adaptive capacity reflects the potential for adaptation; however, adaptation is neither inevitable nor automatic even where adaptive capacity is high (Eisenack and Stecker 2012; Ford and King 2013).
Despite these developments, there is limited research bridging our conceptual understanding of what constitutes adaptive capacity with measurable and documented adaptation outcomes or action. What elements of adaptive capacity are most significant in determining adaptation action? Given the magnitude of the adaptation challenge and adaptation financing committed, it is critical that theorized predictors of adaptive capacity be empirically examined to evaluate to what extent they are associated with adaptation in practice. In this study we test which factors are statistically associated with national-level adaptation, providing an assessment of the context and comparative contribution of different elements of adaptive capacity in driving adaptation actions. Our national focus reflects the importance of national governments as the central pivot for adaptation planning, determining policy priorities and distributing resources and support, with countries also a feasible unit of comparison for which systematic and standardized data are available (Gagnon-Lebrun and Agrawala 2007; Ford et al. 2013). This is, to our knowledge, the first global-scale analysis of predictors of national adaptation actions.
We systematically reviewed reporting on adaptation in 117 countries using the UNFCCC Fifth National Communications (NC5) as a proxy data source for global adaptation. We included only NCs submitted between 2008 and July 1, 2012 in our analysis in order to capture a sample of updated and temporally comparable data. Only documents available on the UNFCCC website are included, and our dataset contains 117/195 national signatories (Supp Mat). Most large global nations are included, with the notable exception of China, Venezuela, and parts of central and east Africa, and south Asia. While the NCs are an imperfect proxy for calculating the Adaptation Initiatives Index (AII), they are one of the only systematically collected data sources available to compare trends in adaptation among nations, and have been used previously to evaluate adaptation progress and highlight trends among industrialized nations (Lesnikowski et al. 2011, 2013; Gagnon-Lebrun and Agrawala 2007; Ford et al. 2013). The benefit of the NCs is the availability of standardized and comprehensive data for a sufficient number of countries to allow multivariate statistical analysis. Analysis using the NCs necessitates focus on national-level adaptations only, however, and is subject to bias based on reporting effort and ability among participating countries. Our data source should thus not be used to infer adaptation processes and drivers at the regional or municipal levels, for example, nor in the non-governmental or private sectors.
N. of types of groundwork action (ranging from 0 to 5 types of groundwork action)
N. of types of adaptation action (ranging from 0 to7 types of adaptation action, with a maximum value of 14)
The possible range of AII scores is 0–19, with 0 representing no evidence of any response taken to climate change risks and 19 representing demonstrated response to all 12 types of action.
Theorized predictors of national adaptation. Detailed variable descriptions and justification for selection are provided in the Supplemental Materials
aGlobal climate risk index (high scores reflect nations most impacted)
aCoastal exposure (% population, within 100 km of the coast)
Population density (pp/km2)
GDP per capita
aInternet users (per 100 people)
Mobile cellular subscriptions (per 100 people)
Information & skills
aPublic perception of climate change risks
Public awareness of climate change
aLiteracy rate (>15 years)
Public expenditure on education as % GDP
aEnvironmental sustainability index: treaty component (72)
aGood governance (Corruption perceptions index), (lower values, higher corruption)
Voice and accountability
Environmental Performance Index (EPI)
Due to high colinearity among predictor variables, we identified a reduced set of variables for use in multivariate regression using bivariate analyses comparing all predictor variables with our Adaptation Initiative Index (AII). Variables significantly correlated with adaptation at p < 0.10 were considered for retention. Pairs of variables within adaptive capacity categories that were strongly correlated with each other (Pearson correlation: p > 0.7) were considered for elimination where the more significant variable (lowest log likelihood) was retained. All variables were checked for linear relationships with the outcome (AII) and transformed where appropriate. Variables retained for multivariate regression are highlighted in Table 1.
We cumulatively constructed a series of models to test for the relative significance of variables within our 5 categories of adaptive capacity (Models 1–5). We used general linear models (GLM) to compare model performance using the AIC, presenting both AIC and R2 values from ordinary least squares regression. We constructed a best-fit model (Model 6) based on minimization of the AIC. OLS post-estimation procedures were performed to confirm that the best-fit model was correctly specified and modeling assumptions met. Data were stored in MS Access, and all analyses were performed using STATA (StataCorp, SE 11).
Predictors of adaptation capacity, all countries
Generalized linear model outcome: adaptation initiatives index (AII) Coefficient (Std. Error)
Model 1 (Exposure)
Model 2 (Exposure + Economic resources)
Model 3 (Exposure + Econ res + Technology)
Model 4 (Exposure + Econ res + Technology + Info & skills)
Model 5 (Exposure + Econ res + Technology + Info & skills + Institutions
Model 6 (Best fit)
Number of observations
Log likelihood (GLM)
Adjusted R2 (OLS)
Global climate risk index
Coastal exposure (% population, within 100 km of the coast)
Total population (log)
Internet users (per 100 people)
Information & skills
Public perception of climate change risks
Literacy rate (binary)a
Environmental sustainability index (Treaty component)
Good governance (Corruption perceptions index)
Environmental performance index (EPI)
The strongest and most consistent predictor of variation in national adaptation is our measure of good governance, based on Transparency International’s Corruption Perceptions Index (CPI) (see maps, Supp Mat). This variable alone explains 19 % of the variation in adaptation action among countries and the strength and significance of its association with adaptation is unaffected by the inclusion of any of the variables we tested. The CPI is a composite measure ranking countries based on how corrupt their public sector is perceived to be, and thus reflects a generalized measure of the quality and efficacy of governance institutions within a country. The addition of our environmental governance variables, ESI and EPI, increases our explanatory power to 28 %, presumably explaining the environmentally-specific governance commitment of nations. The negative coefficient for EPI in the models reflects confounding between the two environmental indices (EPI and ESI), indicating that while the ESI is capturing institutional commitment towards environmental treaties, ratification alone may result in over-prediction of actual adaptation action in the absence of effective institutional commitment to policy implementation.
Population size is retained in the best-fit model, suggesting that more populous countries have higher capacity to implement adaptation action or policy. Notably, population size and GDP were highly collinear (Pearson’s Correlation co-efficient = 0.73, p < 0.01, Supp Mat); while population size may not necessarily imply wealth, the two variables co-vary and cannot be modeled together. Thus population size may be acting – at least statistically – as a generalized proxy for national wealth and size of national economy in the model. More populous, wealthier countries may, however, simply have better capacity to report adaptations and produce detailed UNFCCC reports. To test this, we added a proxy for reporting to the model, reporting intensity, which is based on the total number of actions (or rows) recorded in our database for each country. When reporting intensity is included, population size is no longer significant in the model. Given that total number of adaptation actions in the database may in fact reflect real adaptation activity, however, including reporting intensity in the final model risks over-fitting and is therefore not included in Table 2. It thus remains unclear whether population size and wealth contribute to more detailed and extensive reporting of adaptation or whether more populous, wealthier countries are actually engaging in more adaptation. Notably, good governance remains highly significant even when controlling for reporting. The effect of good governance on adaptation, therefore, cannot be attributed solely to more intensive or detailed reporting by nations.
Many of our predictor variables were significantly correlated with adaptation yet highly collinear with institutional variables (particularly CPI). This was particularly true for internet use. This may indicate that good governance is an antecedent variable, with governance affecting adaptation outcomes through supporting and enabling technological development. If institutional capacity is indeed a causal antecedent to other variables, other theorized predictors of adaptive capacity may thus not be insignificant, but rather more proximal drivers of adaptation. We removed institutional variables as antecedents and find that improved technology, population size and biophysical exposure variables are the most likely variables retained (model not shown). Technology arises as a likely causal pathway through which institutional capacity may facilitate adaptation.
We have previously assessed national-level factors associated with adaptation actions for the health sector among 38 Annex I Parties to the UNFCCC (i.e. developed nations) (Lesnikowski et al. 2013). This work highlighted national environmental governance, wealth, population size, and corruption as correlates of national adaptation outcomes among developed nations. Inference from these results, however, was limited by a small sample size that precluded multivariate analysis and comparative exploration of how determinants of adaptive capacity differentially contribute to adaptation outcomes. Here, we tested and validated the results of Lesnikowski et al. (2013) using a global dataset of 117 countries, focusing on adaptation across all sectors, and employing multivariate analysis.
Our results highlight several important implications for national and international adaptation policy and planning. The good news is that there is evidence of national adaptation action even in small nations with low national GDP (St Lucia) or GDP per capita (Rwanda, Malawi). The bad news is that global adaptation progress, financing and investment may be constrained by lack of governance capacity. Our results suggest, for example, that the importance of all other theorized determinants of national adaptive capacity may be predicated on the presence of good national governance (here proxied by low corruption). The significance of good governance held even when controlling for population size, wealth, environmental policy commitment, biophysical exposure, development indices, and reporting intensity. Secondly, elements of adaptive capacity do not act independently. Some elements, such as technology and possibly information and skills, may be intermediate mechanisms through which good governance affects national adaptation action and capacity.
Our analysis focuses on the national level, and therein reflects primarily adaptations led by national governments. What we do not know is whether nations with poor governance are simply not engaging in adaptation action, or if adaptation is occurring within other jurisdictions (e.g. municipal, regional, civil society). Research might fruitfully investigate whether and where adaptation is occurring in nations with poor governance, and whether governance is also associated with adaptation action at different scales and jurisdictions, for example at the regional and municipal levels and in the non-governmental and private sectors. The next challenge of adaptation tracking also includes consideration of adaptation effectiveness, a methodologically and conceptually tricky yet essential endeavour.
These results are consistent with a body of literature characterizing the mechanisms through which governance and institutions can enable the development of adaptation action (Gupta et al. 2010; Moser and Ekstrom 2010; Pahl-Wostl 2009; Vogel et al. 2007; Dovers and Hezri 2010; McDonald 2011). Moser (2010), for example, describes institutional infrastructure and good governance as necessary for effective problem detection, information gathering, assessment of adaptation options, transmission and communication of information to increase awareness and understanding, and engagement of key stakeholders. Gupta et al. (2010) similarly identify dimensions of adaptive capacity, many of which – trust, access to information, legitimacy, and accountability, for example – are embedded within institutional governance structures that can promote or impede adaption. The surprise here is not that good governance promotes adaptation – this has been well-recognized and described in the qualitative literature for some time – but rather the magnitude and dominance of governance as a predictor of national adaptation action. Our results suggest that governance may be acting as a more causally distal determinant of other more proximal drivers of adaptation, notably technology and possibly also information and skills.
Where governance is poor, yet vulnerability is high, adaptation planning and financing should focus on promotion and development of good governance. Notably, many of the world’s most vulnerable nations also have poor governance. Given financing commitments to support adaptation in low-income nations through the UNFCCC, our results have significant implications regarding the importance of the readiness of nations to effectively implement adaptation, and highlight the critical value of focusing national adaptation planning on building governance and adaptive capacity.
This work was supported by the Social Sciences and Humanities Research Council of Canada. We thank two anonymous reviewers who provided constructive comments.
- Adger WN et al (2007) In: Parry ML et al. (eds) Climate change 2007: impacts, adaptation and vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change 717–743. Cambridge University PressGoogle Scholar
- Bauer A, Feichtinger J, Steurer R (2012) The governance of climate change adaptation in 10 OECD countries: challenges and approaches. J Environ Policy Plann 14. doi:10.1080/1523908x.2012.707406
- Biesbroek GR, Klostermann JEM, Termeer CJAM, Kabat P (2013) On the nature of barriers to climate change adaptation. Regional Environmental Change Published on-line ahead of printGoogle Scholar
- Ford, JD, King D (2013) A framework for examining adaptation readiness. Mitigation and Adaptation Strategies for Global Change. doi:10.1007/s11027-013-9505-8
- Ford JD, Berrang-Ford L, Lesnikowski A, Barrera M, Heymann SJ. (2013) How to track adaptation to climate change: a typology of approaches for national-level application. Ecology and Society 18(3):40Google Scholar
- Ford JD, Berrang-Ford L (2011) Climate change adaptation in developed nations: from theory to practice. SpringerGoogle Scholar
- IPCC (2012) Managing the risks of extreme events and disasters to advance climate change adaptation. A special report of working groups I and II of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, 582 ppGoogle Scholar
- Lesnikowski AC et al. (2011) Adapting to the health impacts of climate change: a study of UNFCCC Annex I parties. Environ Res Lett 6. doi:10.1088/1748-9326/6/4/044009
- Lesnikowski A et al. (2013) National-level factors affecting planned, public adaptation to health impacts of climate change. Glob Environ Chang 23:1153–1163.Google Scholar
- Lesnikowskiw AC, Ford JD, Berrang-Ford L, Barrera M, Heymann J (2013) How are we adapting to climate change? A global assessment. Mitigation and Adaptation Strategies for Global Change. doi:10.1007/s11027-013-9491-x
- Preston B, Westaway R, Dessai S, Smith T (2009) CSIRO PublishingGoogle Scholar
- Smit B, Pilifosova O (2001) Adaptation to climate change in the context of sustainable development and equity. In: McCarthy JJ, Canzianni OF, Leary NA, Dokken DJ, White KS (eds) Climate Change 2001: Impacts, Adaptation, and Vulnerability - Contribution of Working Group II to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press.Google Scholar
- Tompkins EL et al. (2010) Observed adaptation to climate change: UK evidence of transition to a well-adapting society. Glob Environ Chang Hum Policy Dimens 20. doi:10.1016/j.gloenvcha.2010.05.001
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.