Determinants of South Korea’s grants for adaptation aid

(1) Recipient Needs

① Physical Vulnerability

Following Weiler et al. (2018), this study firstly employs the exposure variable from the Notre Dame Global Adaptation Index (ND-GAIN, 2024) to assess physical vulnerabilities. The measure captures countries’ long-term exposure to climate change through projected changes in temperature, precipitation, and agricultural yields. Because the variable exhibits no temporal variations and is scaled from 0 to 1, it helps to reduce potential collinearity. Consistent with the hypothesis H1a, the study expects a positive relationship, whereby countries with higher exposure value receives greater bilateral grants for adaptation aid by South Korea.

In line with Weiler et al. (2018), the second indicator is the Climate Risk Index (CRI) (Germanwatch, n.d.), which quantifies short-term damages of extreme weather events based on human and economic losses. The measure focuses on short-term climate extremes and is published annually. Lower CRI values denote higher vulnerability, therefore, the variable is inverted for analytical alignment with the ND-GAIN exposure index and prior studies. Under this specification, the study anticipates a positive relationship between CRI and South Korea’s grants for adaptation aid, consistent with the hypothesis H1a.

② Adaptive Capacity

Adaptive capacity in this study is measured using three dummy variables (least developed countries (LDCs), African countries, and small island developing states (SIDS)), GDP per capita, and the ND-GAIN adaptive capacity index. The three group dummies capture countries with limited adaptive capacities: LDCs and African countries imply limited financial resources, whereas SIDS represent heightened exposure to climate-vulnerable risks. Following Weiler et al. (2018), the study treats these categories separately, recognizing that donors often differentiate their interventions across groups. Consistent with this, a positive relationship is expected between three vulnerable dummies and South Korea’s grant-based adaptation aid.

Consistent with Weiler et al. (2018), GDP per capita is a measure for domestic financial capacity to manage climate risks. The variable is retrieved from the World Development Indicators, log-transformed to address skewness and specified with one-year lag. As national income increases, countries can increasingly finance their own adaptation measures, reducing dependence on external support. Accordingly, the hypothesis H1b anticipates a negative linear relationship between GDP per capita and South Korea’s grants for adaptation aid.

Lastly, the adaptive capacity sub-index of ND-GAIN (scaled 0–1) captures countries’ capacities in six sectors—water, food, ecosystem services, health, infrastructure, human habitat —summing up 12 indicators including disaster-preparedness systems, infrastructure quality, access to sanitation and electricity. Because higher values denote lower adaptive capacity, and the variable is specified with one-year lag, the study expects a positive relation between this measure and South Korea’s grants for adaptation aid.

(2) Recipient Merits

In line with Weiler et al. (2018), the World Bank’s Worldwide Governance Indicators (WGIs) (World Bank, 2024b) represent one of the most widely used cross national datasets for assessing governance quality. The variable is composed of six composite dimensions including political stability and absence of violence/terrorism, rule of law and control of corruption, regulatory quality. This study employs WGIs (World Bank, 2024b) to capture recipient governments’ institutional quality. In alignment with Weiler et al. (2018), all the dimensions are summed up with equal weights. The variable is specified with one-year lag in the models. Higher WGI scores suggest stronger governance capacity, therefore the analysis expects a positive relationship between WGIs and grants for adaptation aid by South Korea.

Secondly as for environment governance indicator, Weiler et al. (2018) argue that efficient resource use enhances countries’ capacity to manage climate-related risks, placing it as a key component of good governance. Accordingly, environment governance in this study refers to the institutional and legal arrangements through which states manage environmental resources and climate challenges. Empirically, it is captured through a composite “Environment Governance” variable that sums a country’s ratification of three Rio Conventions (UNFCCC, the Convention to combat desertification and Convention on Biological Diversity) and the establishment of National Adaptation Plan (NAP) (UN, 2024). Each ratified convention and the existence of NAP contribute one point, yielding scores from 0 to 4.

Since 1998, OECD has tracked support for Rio Conventions using Rio markers (OECD, 2011), and UNFCCC (2024) highlights the establishment process under COP16 to support national adaptation planning. The variable is lagged by one year. Higher scores reflect stronger environmental governance, and the analysis anticipates that higher values on this measure correspond to greater allocation of South Korea’s adaptation grants.

(3) Donor Interests

To capture donor interests, this study first employs trade flows (exports and imports) from South Korea to recipient countries using annual trade data from the Korea International Trade Association (KITA, 2025), which differs from Weiler et al. (2018), in order to reflect the context of South Korea where imports as well as exports are equally important in the sense that imports of natural resources of recipient countries play critical roles in manufacturing products and goods to be exported.

The variable is log-transformed to address skewness and specified with one-year lag. Higher trade levels indicate greater economic engagement and, by extension, stronger donor interests. Accordingly, the study expects a positive correlation between trade volume and South Korea’s grants for adaptation aid.

Secondly, similarities in diplomatic positions are measured using voting alignment at UN General Assembly (Weiler et al., 2018). Following Bailey et al. (2017), ideal points are estimated through an Item Response Theory (IRT) model and normalized to a mean of zero and a standard deviation of one. Ideal points are updated annually, specified with one-year lag. The study anticipates a positive relationship between diplomatic similarity between two countries and South Korea’s adaptation-related grants.

Thirdly, geographic distance is used as an indicator of strategic interests. Bilateral distance data are taken from the CEPII Gravity data (CEPII, 2024), log-transformed and defined as the distance between national capitals. Donors typically exhibit stronger strategic interests in geographically closer countries (Betzold and Weiler, 2018; Clist, 2011; Weiler et al., 2018) Consistent with this, the study expects a negative relationship between distance and South Korea’s grants for adaptation aid, implying that closer partners receive larger volumes of assistance.

Lastly, the indicator of overseas visits by heads of state is employed as a measure to represent political interests of donors. Foreign visits by heads of state tend to occur in pursuit of national interests in international relations and involve high-level decisions between political leaders to facilitate diplomatic processes. Diplomatic contacts in person are likely to expedite decision-making processes (Jung, 2022). The indicator sums up all of both inbound and outbound heads of state visits of both countries in a given year, which is specified with one-year lag. The measure is expected to have a positive relationship with South Korea’s adaptation grants.

Consistent with Weiler et al. (2018), this study constructs a dyadic panel dataset of South Korea’s grants for adaptation aid to recipient countries from 2014 to 2022 using the OECD Creditor Reporting System (OECD, 2024b) in constant 2022 USD. The dependent variables are derived from Rio-marker classifications for climate change adaptation, which require donors to identify whether adaptation is the principal or a significant objective (OECD, 2011).

Extensive literature conducts analysis on limitations on the rio markers. Previous literature indicates risks of double counting caused by flexibilities in interpretations of adaptation activities and inconsistencies of reporting across donor agencies. These are exacerbated by self-reporting nature, therefore empirical assessments identify over-reporting of adaptation aid especially with projects in the category of significant adaptation aid (Michaelowa and Michaelowa, 2011; Weikmans et al., 2017).

To address over-reporting and limitations of Rio markers in line with Weiler et al. (2018), Adaptation Watch (2015) and Wingqvist et al. (2011), this study constructs two specifications: (a) per-capita⁵⁾ grants for principal adaptation aid and (b) per-capita grants for principal plus 50% discounted significant adaptation aid, which is a method in line with practices in accounting of climate change for donors. Annual commitments are aggregated for each South Korea–recipient dyad and log-transformed, and zero values are assigned when no adaptation grants are reported.

Per-capita values are calculated by dividing annual commitments by population size retrieved from the World Development Indicators (World Bank, 2024a). Using both specifications enhances robustness and improves the validity of the dependent variables by reducing potential over-reporting in adaptation-aid reporting.

Building on Weiler et al. (2018), this study includes total grants per capita from South Korea as a control variable. It is constructed from OECD CRS data by dividing annual grant commitments by population figures from the World Development Indicators (World Bank, 2024a). This variable is specified with one-year lag and transformed using the logarithm to correct for skewness. Since grants for adaptation aid constitute a subset of total grant aid, the two are expected to correlate strongly.

The number of population for recipient countries serves as a second control variable, also drawn from the World Development Indicators (World Bank, 2024a). As noted by Weiler et al. (2018), more populous countries tend to attract greater strategic attention, yet population size simultaneously reduces per-capita grant levels. The variable is lagged by one year and log-transformed. Accordingly, the variable is expected to be positively associated with the likelihood of receiving South Korea’s grants for adaptation aid at the selection stage but negatively related with per-capita allocations at the allocation stage.

(1) Physical Vulnerability

The exposure sub-index of ND-GAIN shows statistically significant and positive effect in Models 3 of Table 2 at the selection stage; however, it does not retain significance in the allocation stage for each dependent variable. These findings indicate that the indicator fails to support the hypothesis H1a, suggesting that South Korea does not consider higher long-term physical vulnerabilities.

Additionally, the Climate Risk Index (CRI) does not have statistical importance with two dependent variables in Table 2 at the selection stage, yet the measure shows statistical significance with positive coefficients on the dependent variables in Models 2 and 4 of Table 3 at the allocation stage, suggesting that countries with short-term climate risks are not likely to be prioritized to receive grant-based adaptation aid by South Korea but, once chosen, they tend to receive larger amounts of adaptation grants. Overall, these findings do not support the hypothesis H1a.

(2) Adaptive Capacity

Turning to vulnerable statuses (LDCs, Africa, and SIDS), the LDC dummy is negatively related with both dependent variables and is statistically significant across all models in Tables 2 and 3 at both stages. This result does not validate the hypothesis H1b, indicating that LDC status does not drive South Korea’s allocation of grants for adaptation aid. The Africa dummy is statistically insignificant in all models in Table 2 and 3 at both stages except Models 2 in Table 2 and 3 respectively where coefficients are positive at the selection stage and negative at the allocation stage. These results also do not validate the hypothesis H1b, suggesting that Africa is not prioritized for South Korea’s adaptation grants.

The SIDS dummy shows statistical significance only in Models 1 and 2 of Table 2 at the selection stage, with positive coefficients for principal adaptation grants per capita. However, it loses significance in nearly all allocation-stage models, except Model 3 of Table 3 at allocation stage. This evidence does not support the hypothesis H1b, indicating that SIDS are not considered to be prioritized in South Korea’s grant-based adaptation aid.

The adaptive capacity sub-index of ND-GAIN shows no statistical significance in selection-stage model but becomes significant with positive coefficients in all allocation-stage models except Model 3 of Table 3. These findings do not support the hypothesis H1b, suggesting that the variable does not influence the likelihood of receiving grants for adaptation aid, but among selected countries, those with limited adaptive capacity tend to benefit from larger adaptation grants by South Korea.

Finally, GDP per capita is negatively related with both dependent variables and statistically significant in all selection-stage models in Table 2. However, it loses significance in all allocation-stage models in Table 3. Thus, the evidence does not validate the hypothesis H1b, indicating that GDP per capita affects the likelihood of receiving South Korea’s adaptation grants but does not determine the volume of support once countries are selected.

For recipient merits, WGIs display positive, significant coefficients in Models 3 and 4 of Table 2 at the selection stage, yet the variable loses statistical significance in all Models of Table 3 at the allocation stage. These results do not support the hypothesis H2, indicating that governance quality exerts limited influence on South Korea’s grants for adaptation aid.

Environment governance is statistically insignificant across all models in Tables 2 and 3. Thus, the results fail to validate the hypothesis H2, suggesting that environment governance does not shape South Korea’s allocation of grants for adaptation aid.

Regarding donor interests, total trade to recipient countries show positive and statistically significant effects on both dependent variables across all four selection-stage models in Table 2 and two allocation-stage models (Models 1 and 2) in Table 3. These findings validate the hypothesis H3, indicating that higher levels of South Korean exports increase both the likelihood of being chosen for adaptation grants and the amount received.

Geographic distance consistently displays negative coefficients across all model specifications in Tables 2 and 3, though statistical significance appears only in Models 3 of Table 2 at the selection stage. These patterns do not support the hypothesis H3, suggesting that geographic proximity does not meaningfully influence South Korea’s grant-based adaptation aid allocation.

UN voting alignment shows limited statistical significance only in Models 2 and 4 of Table 2 at the selection stage, where the measure is positively related with the likelihood of allocation of grants for adaptation aid. However, it isn’t statistically important in all allocation-stage models in Table 3. These findings do not support the hypothesis H3, indicating that the variable does not play a substantive role in determining South Korea’s grants for adaptation aid.

Lastly, the measure of overseas visits by heads of state is statistically insignificant across all models in Tables 2 and 3. Thus, the results do not support the hypothesis H3, meaning that the measure does not determine South Korea’s allocation of adaptation grants.

Across all models in Tables 2 and 3, South Korea’s total grants per capita are positively and statistically significantly related with two dependent variables at both stages. This pattern indicates that this variable is the strong predictor of South Korea’s grants for adaptation aid among all explanatory variables.

For the population size, the results show statistical significance in most model specifications—Models 1, 3, and 4 of Table 2 at the selection stage, and all the Models of Table 3 at the allocation stage. Population is positively related with the likelihood of receiving adaptation grants at the selection stage, but negatively associated with per-capita allocations at the allocation stage. These findings imply that population increases the probability of selection for South Korea’s grants for adaptation aid, yet more populous countries typically receive a smaller proportion of adaptation aid per capita.