Climate Change and Logistics Performance: Evidence from a Dynamic Panel Analysis

Climate change has emerged as one of the most critical global challenges of the 21st century, threatening the stability and sustainability of economic systems worldwide. Increasingly frequent and severe weather events, including storms, floods, droughts, and sea-level rise, are disrupting both natural ecosystems and financial infrastructure. According to the Intergovernmental Panel on Climate Change [1], global average temperatures have increased by 1.1℃ since the pre-industrial era, contributing to the intensification of extreme climate events. While the consequences of climate change are widely recognized in agriculture and energy sectors, its implications for logistics, a vital enabler of global trade, are gaining urgent attention.

Logistics systems rely heavily on climate-sensitive infrastructure such as seaports, road networks, railways, and storage facilities. As extreme weather becomes more frequent, these systems face heightened risks of physical damage and operational delays. It is estimated that natural disasters affecting port infrastructure alone result in annual global trade losses of up to 81 billion US dollars and a reduction in global economic output of approximately 122 billion US dollars [2]. More than 60 per cent of global trade is handled through ports located in coastal zones that are at significant risk from sea-level rise and storms [3]. Disruptions in transportation routes, rising costs of freight, and diminished reliability in supply chains compromise not only logistics efficiency but also trade competitiveness, especially for developing countries with vulnerable infrastructure [4]. The need to explore the climate-logistics nexus is therefore essential, as logistics lies at the intersection of climate vulnerability and economic performance.

This study investigates the dynamic relationship between climate change and national logistics performance by employing a panel dataset of 45 countries over the period 2009-2023. In response to increasing environmental disruptions affecting supply chains, the study aims to provide timely insights into how climate factors influence logistics capacity and efficiency across countries. To address potential endogeneity in the model, the Generalised Method of Moments (GMM) is employed, ensuring robustness in causal estimation. A major innovation of this research is the construction of a novel Climate Change Index (LnCLI), derived from the frequency of climate-related keywords extracted from global media and institutional sources. This index is further decomposed into thematic subcomponents, including temperature, precipitation, and wind change and cross-validated with objective physical indicators to enhance measurement validity.

The regression model employed in this study to evaluate the impact of renewable energy consumption on environmental pollution is expressed by the following equation:

$L P I_{i, t}=\beta_0+\beta_1 L n C L I_{i, t}+\beta_2 \operatorname{CONTROL} L_{i, t}+t+u_{i, t}$

where,

- $L P I_{i, t}$ is the Logistics Performance Index for country i at time t. The dependent variable LPI_i,trepresents the overall logistics efficiency and is measured as the average score across six key dimensions provided by the World Bank: infrastructure quality (Infras), shipment capability (Ship), logistics service competence (Compe), tracking and tracing (Track), delivery timeliness (Time), and customs clearance efficiency (Custom).

- ${LnCLI}_{i, t}$ is the log-transformed Climate Change Index (CLI). The key independent variable, ${LnCLI}_{i, t}$ captures the intensity of climate-related discourse and concerns. It is derived from a frequency-based analysis of climate-related terms from global data repositories such as GDELT, LexisNexis, and policy-oriented sources including IMF, World Bank, and UN Reports. The selected keywords include: climate change, global warming, extreme weather, and rising sea levels. This approach draws from the literature on media-based economic indicators, which posits that collective discourse and information salience influence behavior, expectations, and policy responses [16]. In the context of climate change, keyword-based indices have been used to capture climate-related uncertainty, public concern, and risk perception. Such indices reflect not only the environmental phenomena themselves but also their visibility and perceived urgency across countries. This aligns with the notion that climate risk is multidimensional, encompassing both physical exposure and the degree to which it is recognized and acted upon by economic agents. To provide a more granular view of climate discourse, the study decomposes ${LnCLI}_{i, t}$ into three thematic sub-indices, each also based on keyword frequency: ${LnTem}_{i, t}$ captures discourse intensity related to temperature changes (e.g., heatwaves, temperature rise), $LnPre _{i, t}$ reflects references to precipitation and flooding (e.g., rainfall, floods, droughts), and $LnWind_{i, t}$ measures concern over wind-related events (e.g., hurricanes, storms, typhoons). To validate the robustness of this novel index, the study includes three additional objective physical climate indicators: average annual temperature (Temper), precipitation (Precipit), and wind speed (WindS) obtained from NASA [5]. The study deliberately includes both subjective (media-based) and objective (sensor-based) indicators of climate risk to capture its multidimensional nature. The media-based indices (LnCLI, LnTem, LnPre, LnWind) reflect public discourse and perceived salience of climate-related events, while the physical indicators (Temper, Precipit, WindS) represent actual environmental conditions.

- ${CONTROL}_{i, t}$ is a vector of control variables. The vector of control variables includes:

• Nominal GDP (${LnGDP}_{i, t}$) to account for economic scale. Higher income levels typically enable greater investment in logistics infrastructure and service quality. This variable is widely used in logistics performance models [20].
• Globalization index $\left(K O F_{i, t}\right)$ to reflect exposure to global integration. A higher level of global integration often reflects greater cross-border trade activity and exposure to international logistics standards [21].
• Inflation rate $\left(I N F_{i, t}\right)$ to capture macroeconomic stability. High inflation may reflect economic uncertainty and can distort investment decisions related to infrastructure and logistics operations [22].
• Trade openness $\left(L n T R D_{i, t}\right)$ as a share of GDP. This variable captures a country’s involvement in global trade flows, which drives demand for efficient logistics systems [23].
• Mobile phone penetration $\left(\right. LnMobile \left._{i, t}\right)$ as a proxy for digital infrastructure. In contexts where ICT or broadband data are limited or inconsistent across countries, mobile penetration serves as a practical and widely adopted indicator of digital readiness and connectivity [20, 24].
• Per capita CO₂ emissions ${LnCO2}_{i, t}$ as an environmental intensity measure. Countries with higher emissions often have greater transport-related activity and infrastructure exposure [25].
• Population size $\left(L n P O P_{i, t}\right)$ to reflect market size and demographic pressure. Larger populations typically require more extensive logistics coverage and capacity [11].

- t represents time fixed effects,

- u_i,t is the error term.

Detailed definitions and sources for all variables are presented in Table 1.

Table 3 presents the GMM estimation results examining the effect of climate-related discourse on logistics performance (LPI). The lagged dependent variable (L.LPI) is positive and highly significant across all models, with coefficients ranging from 0.565 to 0.599 (significant at the 1% level), confirming the strong persistence of logistics performance and the importance of path dependence in infrastructure development.

The main independent variable LnCLI, constructed through textual frequency analysis of climate-related terms across global datasets, has a positive and statistically significant effect on LPI (coefficient=0.00901, p<0.05). This suggests that increased climate awareness and discourse are associated with improvements in logistics performance, likely through policy shifts and green infrastructure investments.

Further analysis incorporates LnTem, LnPre, and LnWind, which are sub-components of the Climate Change Index (LnCLI). These variables represent the frequency of specific climate-related terms (temperature, precipitation, and wind-related risks) and capture public, media, and institutional concern around particular climate threats. All three components have positive effects on LPI and are marginally significant at the 10% level. This reinforces the interpretation that higher climate discourse intensity - particularly regarding weather extremes - correlates with greater logistics responsiveness and adaptation.

Table 3. GMM estimations of the impact of Climate Change and sub-components on LPI

Note: The numbers in parentheses below indicate the standard errors for each variable. *** denotes significance at the 1% level, ** denotes significance at the 5% level, and * denotes significance at the 10% level.

In terms of control variables, LnGDP and LnTRD maintain positive and significant effects across all models, consistent with expectations that economic scale and trade openness drive logistics performance. The KOF globalization index has a positive but weakly significant effect. Inflation, mobile phone penetration, and population size show negative associations with LPI, reflecting cost pressures and infrastructure limitations. CO₂ emissions per capita (LnCO2) show a negative but statistically insignificant relationship.

In summary, these results demonstrate that not only overall climate discourse (LnCLI) but also specific thematic concerns (LnTem, LnPre, LnWind) influence logistics performance.

Table 4 presents the GMM estimation results analyzing how climate-related discourse (LnCLI) influences six key components of the Logistics Performance Index (LPI). Across all models, the lagged dependent variables are positive and highly significant, confirming the persistence and inertia in each logistics dimension. This underscores the long-term nature of logistics system development.

Regarding the main independent variable, LnCLI, the results indicate heterogeneous effects across LPI components. The influence of climate discourse is statistically significant and positive for two components: logistics competence (Compe) and tracking and tracing (Track), with coefficients of 0.00352 (p<0.1) and 0.00326 (p<0.05), respectively. This suggests that heightened attention to climate change is driving improvements in operational efficiency and supply chain visibility—likely through digital innovations, staff training, and climate adaptation efforts in monitoring systems.

In contrast, LnCLI has no statistically significant effect on four other subcomponents: infrastructure (Infras), customs (Custom), international shipment capability (Ship), and timeliness (Time). Among these, the lack of significance for infrastructure is particularly notable. One possible explanation lies in the structural rigidity and time-intensive nature of infrastructure development, which is typically driven by long-term investment cycles and constrained by fiscal and institutional factors.

Table 4. Estimation results of the impact of climate change on LPI components

Note: The numbers in parentheses below indicate the standard errors for each variable. *** denotes significance at the 1% level, ** denotes significance at the 5% level, and * denotes significance at the 10% level.

As such, short-term fluctuations in climate-related discourse may not translate into immediate changes in physical infrastructure [16]. Similarly, customs procedures and international shipment capability may require regulatory or diplomatic coordination across multiple agencies and borders, which tends to evolve slowly. The slightly negative but insignificant effect on timeliness suggests that even as climate discourse rises, immediate performance on schedule adherence may remain vulnerable to extreme weather disruptions, especially in low-resilience settings [3].

Table 5 presents the regression results of models evaluating the impact of climate change on LPI for two groups of countries: developed and developing countries. In this table, columns (1), (2), (3), and (4) show the regression results for the developed country group, while the remaining columns display the corresponding regression results for the developing country group. The results show that the persistence of LPI (L.LPI) is much stronger in developing countries, highlighting their reliance on past performance improvements. Climate-related variables (LnCLI, LnTem, LnPre, LnWind) generally have positive and significant impacts in both groups, though effects are more robust in developing countries, suggesting greater climate vulnerability. Overall, climate change appears to have a stronger influence on logistics systems in countries with weaker infrastructure and lower adaptive capacity.

Table 5. Estimation results of the impact of climate change on LPI for developed and developing countries

Note: The numbers in parentheses below indicate the standard errors for each variable. *** denotes significance at the 1% level, ** denotes significance at the 5% level, and * denotes significance at the 10% level.

Table 6. Estimation results of the impact of substitution measurement of climate change on LPI

Note: The numbers in parentheses below indicate the standard errors for each variable. *** denotes significance at the 1% level, ** denotes significance at the 5% level, and * denotes significance at the 10% level. The S-test is the Sargan test using the J-statistic to test the null hypothesis H0. AR(1) and AR(2) check the correlation of the instrumental variables.

Table 6 presents robustness checks using objective physical climate indicators that are temperature (Temper), precipitation (Precipit), and wind speed (WindS) as substitutes for the text-based climate index. The results are broadly consistent with the main findings. All three indicators show positive coefficients, with temperature (0.014) and precipitation (0.017) being statistically significant at the 5% and 10% levels, respectively, confirming that climate variability influences logistics performance. Overall, the consistency across these models reinforces the reliability of the main results and validates the use of the Climate Language Index.

Table 7 provides robustness checks using OLS, FEM, and REM estimations to validate the core results of the GMM model. The estimated signs and statistical significance of key variables including LnCLI, its components (LnTem, LnPre, LnWind), and control variables are largely consistent across models. Climate-related variables remain positively associated with logistics performance, though with slightly varying magnitudes. These results confirm the stability and reliability of the main findings under alternative estimation techniques.