This thesis examines how the accumulation of AI-related knowledge has shaped the comparative advantage of countries in global trade since 2010 and assesses the implications for the technological sovereignty goals of the EU. Using an extended Heckscher Ohlin Vanek (HOV) framework with factor-content corrections proposed by Trefler and Zhu (2010), AI and non-AI patent stocks are integrated alongside traditional labour and capital stock endowments to analyse trade patterns across 45 countries from 2010 to 2021. The empirical analysis reveals strong support for the HOV framework, with sign test achieving 90% success rate and patents performing better than traditional factors of labour and capital in predicting trade patterns. The results demonstrate that as of 2021, the US, China, and Korea hold a comparative advantage in AI patents, while the EU does not. Within the EU, only the Netherlands, Sweden, and Finland attempt to maintain a comparative advantage, whereas major economies, including Germany, France, and Italy, lag. The relative factor abundance in AI patents for the EU deteriorated between 2012 and 2019, coinciding with the rise of China and reflecting internal fragmentation among member states. These findings indicate that the lack of comparative advantage in AI for the EU creates strategic vulnerabilities that undermine its technological sovereignty goals, particularly as AI becomes embedded in critical infrastructures. For the EU to achieve genuine strategic autonomy, regulatory leadership must be complemented by strengthened capabilities in AI-enabling hardware, reduced dependence on foreign AI intellectual property, improved translation of research into commercial innovation, and cohesion-oriented policies addressing internal capability divergence. The research contributes methodologically by demonstrating that accumulated knowledge stocks measured through patents constitute formable factors that shape comparative advantage, and empirically by providing the first comprehensive HOV analysis of AI-related patent endowments across major trading economies.
... Read more

Clean energy spending needs to reach $4.5 trillion per year by 2030 to maintain a realistic chance of limiting global warming, yet current investment represents just 37% of what is needed for net zero by 2050. This and other energy transition requirements put enormous pressure on electric utilities to transform their business models while maintaining energy security and attracting investment, creating conditions ripe for potential greenwashing as companies might oversell environmental progress or engage in misleading communication strategies to stay competitive. Three fundamental gaps have limited systematic greenwashing detection: the separation between performance and communication analysis, single-dimensional communication measurement, and the lack of sector-specific methods.
This research addresses the question: What greenwashing risk assessment tool can be developed for European electric utility companies based on environmental performance metrics and multi-dimensional communication analysis? This Master thesis research project developed and tested the first systematic Greenwashing Risk Assessment Tool (GRAT), designed specifically for electric utilities but adaptable to other sectors through modified inputs, weights, and terminology. The GRAT bridges theoretical gaps by integrating quantitative performance measurement with multi-dimensional communication analysis. Constrained ensemble methodology is used to handle weight uncertainty for both performance scoring and greenwashing risk. The greenwashing risk is measured by five different dimensions: performance-communication gap, substantiation weakness, language vagueness, temporal orientation, and reporting consistency.
The GRAT follows a clear three-step process that can be understood and adapted without advanced programming skills. Users gather environmental data from several sources: third-party verified information like Refinitiv Eikon, self-reported metrics like CDP, and English sustainability reports. Once this data is collected, the tool automatically handles performance scoring, runs communication analysis through rule-based NLP methods, and produces integrated risk assessments. Rather than using completely arbitrary weights, the constrained ensemble approach tests thousands of valid weight combinations within available theoretical boundaries.
The tool was tested on 14 European electric utility companies during 2021-2022. Among these companies, the analysis found significant environmental performance differences. Scores ranged from 15.0 to 95.0 points, with year-over-year changes spanning from -35.1 to +63.0 points. Companies used different communication strategies that did not align with their performance scores. Only 36% showed temporal changes where performance and communication moved in the same direction, revealing a pattern where communication strategies often work independently from actual environmental achievements. When combining both aspects including the other communication quality dimensions, risk scores fell between 16.5 and 81.3.
The validation approach is methodologically sound, though statistical limitations arise from the small sample size. External validation provides indicative evidence that the GRAT may distinguish between companies with documented greenwashing accusations and those with clean records, but results could represent random phenomena between 2021-2022. The GRAT enables regulators to screen companies requiring investigation rather than providing conclusive evidence of greenwashing. This helps investors assess risk patterns in sustainability portfolios and offers researchers a tool adaptable to other sectors. ... Read more

Nature-based solutions (NbS) such as urban parks, green corridors, and stormwater wetlands are increasingly adopted to enhance climate resilience and livability in cities, yet their social impacts can be uneven. In Cape Town—where apartheid-era planning left certain communities on the under‑resourced Cape Flats while the wealthier people clustered around greenspaces—there is a pressing need to understand whether greening projects might inadvertently displace vulnerable renters. This thesis examines that question by focusing on two contrasting electoral wards: Ward 57, an affluent inner‑city neighbourhood, and Ward 79, the lower‑income Mitchell’s Plain township in the Cape Flats.
To explore how NbS affect housing affordability and demographic composition, a spatially explicit agent‑based model (ABM) was developed for each ward. The city map is discretized into a grid of housing parcels, each with evolving rent levels and ‘livability’ scores that reflect environmental quality, including the presence and maturity of NbS interventions. Household agents—characterised by income, rent‑affordability thresholds, and demographic identifiers—occupy units and face displacement when rising rents exceed their budget or eviction risk thresholds. Vacant units may then be filled by higher‑income in‑migrants. NbS scenarios are introduced exogenously, locally boosting livability scores and triggering rent uplifts that propagate through the spatial grid. Over multi‑year simulations, the model tracks rent burdens, forced displacement events, and shifts in population makeup, enabling side‑by‑side comparison of greening impacts in both wards under a baseline ‘no‑policy’ scenario and alternative housing‑policy regimes.
Results show that, without housing safeguards, NbS raise rents significantly—by roughly 20–25% over ten years—displacing thousands of low‑income households. In Ward 57, this reinforces existing privilege as affluent renters easily absorb price increases, whereas in Ward 79 it displaces predominantly Black African households and Coloured households, substituting them with higher‑income newcomers. Introducing inclusionary housing requirements—which reserve a substantial share of units near NbS sites for affordable rents, dampens rent inflation, retains most incumbent households, and preserves socio‑economic diversity. Time‑limited rent subsidies yield similar short‑term relief but fail to prevent eventual displacement once support expires.
These findings underscore that NbS, while delivering environmental and recreational benefits, can exacerbate urban inequality unless paired with deliberate affordability measures. To achieve just transitions, Cape Town’s green‑infrastructure initiatives should be coupled with inclusionary zoning, dedicated affordability funds, and robust community engagement processes that empower residents and monitor displacement risks. By integrating climate‑adaptation investments with housing‑equity policies, cities can ensure that ecological solutions do not become drivers of social exclusion ... Read more

Climate change intensifies the frequency and severity of flooding in urban areas, such as Jakarta. While government-led adaptation measures are crucial, the limitation of planned adaptations in providing complete protection emphasises household-level strategies in mitigating risks. However, limited adaptive capacity and bounded rationality can lead to maladaptive behaviours. This study investigates the dynamics of household-level climate change adaptations in Jakarta, focusing on their impact on vulnerability and potential maladaptation. Employing a mixed-methods approach, the research analyses micro-level survey data and agent-based modelling to examine adaptation measures like elevation, dry-proofing, and wet-proofing. Findings reveal that socioeconomic factors, such as income and fatalistic attitudes, influence adaptation choices. Low-income households are particularly vulnerable to suboptimal strategies, like wet-proofing, which can exacerbate their risk. The study also highlights a discrepancy between these measures' perceived and actual efficacy, emphasising the need for targeted policies. Recommendations include promoting effective adaptation, providing targeted financial aid, and enhancing community and governmental support. By addressing these factors, policymakers can reduce the likelihood of maladaptation in urban Jakarta. ... Read more

The global biodiversity crisis demands urgent action, with global food production being one of the major drivers of biodiversity loss. Transitioning to less animal-intensive diets is a potential solution to decrease impacts. While it is established that diet change can lead to a decrease in agricultural land use, it remains unclear whether these reductions occur in areas at risk of biodiversity loss. This study uses a spatially-explicit land use indicator, based on co-existence of species richness and agriculture, to analyze the impact on four land use types, ranging from land use with low conservation priority to land use with very high conservation priority. We analyze the impact of a diet change scenario where fifty-four high-income countries adopt the EAT-Lancet diet. Our findings indicate a major decrease of about one third of conservation priority land use across all four land use types considered. This means that the diet change is indeed effective in lowering land use in high conservation priority sites. We recognize that even with the adoption of the EAT-Lancet diet there is a need for additional strategies to further decrease high conservation priority land use, with priority sourcing strategies being the most obvious next step. ... Read more