Skip to main content
European Commission logo
English English
CORDIS - EU research results

Scaling up behavior and autonomous adaptation for macro models of climate change damage assessment

Periodic Reporting for period 3 - SCALAR (Scaling up behavior and autonomous adaptation for macro models of climate change damage assessment)

Reporting period: 2021-04-01 to 2022-09-30

Key scientific problem:
The SCALAR project aims to bridge a gap between micro and macro research traditions in climate change damage assessment by modeling behavioral aspects of private adaptation processes of heterogeneous agents, and integrating them into macro climate policy models. The project focuses on sea level rise and flooding, as the costliest climate-induced hazard worldwide. It adds novel contribution to the scientific understanding of feedbacks between the adaptation behavior of heterogeneous actors and macro level damage and resilience assessments, eliciting potential tipping points along the climate change adaptation pathways. As such the project revisits the classic micro-macro aggregation problem in social sciences by means of new computational social science methods and micro-data. Specifically, SCALAR uniquely combines:
1) New behavioral data on climate adaptation decisions collected via longitudinal cross-country household surveys, going beyond a snapshot to uncover evolving decision processes;
2) Advances in agent-based modeling to scale up adaptation decisions of heterogeneous households and firms to a regional economy while including rich behavioral, economic & spatial hazard data;
3) Cutting-edge ways of integrating micro-simulation models with traditional macro models, like Computable General Equilibrium models (CGE) to synergize the two approaches for developing new theory- and data-grounded macro damage assessments.

Importance for society:
SCALAR makes a step change in integrating behavioral aspects of human decision-making into macro climate policy models. It enables quantifying distributional impacts of shocks and of adaptation among various households, businesses and regions. By consolidating micro-data on behavioral adaptation of households with computational agent-based models and macro-economic damage assessment models, SCALAR unlocks the development of a new generation of CGE & integrated assessment models (IAMs). Being grounded in behavioral and regional economic theories and supported by empirical private adaptation studies, such models will explore an interplay of public and private adaptation efforts. Current IAMs – developed to support climate change mitigation – face limitations in informing design of climate change adaptation policies that include adaptation across different scales (stakeholders), from government-led public adaptation to private adaption of households and businesses. It will enable the quantitative exploration of cross-scale damage cascades, the identification of thresholds over which private adaptation impacts the macro level, and the tracing of the emergence of socio-economic resilience as climate change unfolds. The methodological advancements may have impact beyond the domain of climate adaptation.

The objective of SCALAR is to bridge the gap between micro and macro research traditions by modeling behavioral aspects of private adaptation processes among heterogeneous households and firms from the bottom up, and by integrating them into macro level climate change policy models. Towards this end, SCALAR aims to:
1. Synthesize the existing empirical evidence on autonomous climate change adaptation globally to identify generic patterns of relationships between the private adaptation of heterogeneous agents, resilience and damages, and complement it with new detailed behavioral data on the process of individual adaptation decision-making in four different geographic contexts (we have selected two Global North countries – USA and the Netherlands, and two Global South countries – Indonesia and China).
2. Develop innovative simulation tools to aggregate the private adaptation of heterogeneous individuals and firms in a regional economy enabling cross-scale feedbacks between adaptation, resilience and damages.
3. Integrate micro simulations, which will provide solid theoretical and empirical grounds for climate change adaptation and damages, with macro level climate policy models (CGE, IAM) to trace feedbacks between adaptation, damages and non-monetary aspects characterizing socio-economic resilience.
I. Patterns in household CCA behavior
We conducted a meta-analysis of surveys on household flood adaptation. We founds a Global North research bias in the collected surveys, and statistically significant relationships between several key drivers of household adaptation and different cultural aspects - ranked quantitatively using Hostede cultural rankings (Noll et al; 2020). We ran the 1st survey in 04.2020 (N=6400) with 3 subsequent surveys (10.2020; 04.2021; 10.2021) in 4 countries (US, NL, CN, ID). Our surveys elicit household CCA (18 on-site actions, insurance, relocation), their drivers using constructs from few theories, self-assessed resilience, place attachment, socio-economic and damage data.
We published 3 papers analyzing the survey data regarding household perceptions, intentions, and behavior (Noll et al. 2021, Noll et al. 2022, Noll et al. 2023). The 1st article appeared in Nature Climate Change and got on Dutch National Television. An open access survey data batch is on DANS.

II. From private CCA behavior to regional economic resilience
We conducted a systematic literature review of flood risk ABMs and resilience through the lens of complex adaptive systems (Taberna et al; 2020). We developed the Climate-economy Regional Agent-Based (CRAB) model to trace the interplay between climate shocks and agglomeration dynamics of heterogeneous firms and households (Taberna et al 2022A; Taberna et al 2022B). We linked CRAB with our surveys from Florida and analyzed inequality and regional impacts of household CCA along a gradient of rationality – from perfect-optimizers to empirical behavior (invited paper to PNAS). We also calibrated CRAB with our surveys from Shanghai and studied synergies of CCA across scales. We now analyze firms data that tracks CCA investments in 137 sectors in studied coastal cities.

III. Macro-economic damages & private CCA
Climate-damage assessment is traditionally conducted at the level of global regions (IAMs), national or provincial level (CGE). Hazard models (e.g. flood) offer high resolution estimates but usually only for physical damage, with simplified direct economic damage. Considering that CCA has local impacts, we now develop a downscaled economic analysis for coastal urban regions. Further, we develop methods to combine government-led adaptation with private CCA of households and firms in CGE/IAM. We collaborate with CGE/IAM experts from PBL (NL, Environmental Assessment Agency) and the Euro-Mediterranean Center on Climate Change (CMCC, IT).

IV. Scalability & Modularity of ABMs
ABMs are computationally expensive, especially for large number of agents. Liz works on optimization of the current version of CRAB, making ABMs more efficient, improving sensitivity tests and creating possibilities for scaling-up ABMs to large populations.
Our team aspires to make ABMs reusable. This aligns with model integration in SP3 as we sharpen technical aspects of how to make different parts of a complex ABM modular. We developed a template to report ABM modules based on the principles of software engineering, coded examples of open-source shared and transparent modules (e.g. on households CCA decisions grounded in behavioral theories and in data). We plan to launch a website that will serve as an open access international repository for modular ABMs.

V. Education
Brayton won few scholarships and attended several statistical methods courses in EU & US. He was awarded a grant from KNAW for a research visit at LSE (UK) to collaborate on multi-scale household resilience. He has co-supervised 1 BSc and 3 MSc students (2 graduated Cum Laude).
Alessandro took few simulation courses. He won the Young Scientists Summer Program (YSSP, IIASA, AT). He has co-supervised 4 MSc students, 2 from TUD and 2 from Politecnico di Milano (who graduated Cum Laude).
Ignasi took the course on General Algebraic Modeling System needed to run CGE/IAM.
Was done in 3 dimensions:
1) Micro-level data on household CCA:
a.Our meta-analysis revealed the Global North research bias in the empirical survey literature and found that culture could explain some of the variance in the effects in CCA drivers (Noll et al 2020).
b.Using our 4-country survey, we analyzed differences in the statistical effects that various factors had on CCA, and for the first time showed empirically how same behavioral drivers can have different effects on household CCA in different countries (Noll et al 2021). This was not possible before due to the lack of comparable data.
c.The survey data revealed that both past adaptations and additionally intended actions increase individual intention to take a specific CCA. This has implications for estimating the speed and scope of household CCA diffusion (Noll et al, 2022).
d.We showed that survey respondents who select “I don’t know” on hazard probability/damage questions are also most vulnerable (low educated, low income). Previously, they have been dropped or bootstrapped into the analysis, being overlooked as a distinct group that requires other CCA triggers (Noll et al. 2023).
e.We analyze the panel survey data using novel methodologies in this domain (Bayesian panel multinomial regression, structural equation dynamic lagged panel model) to study the intention-behavior gap and consider for the first-time what drives and inhibits individuals to follow through or deviate from their stated intentions.

2) ABM :
a.Our review (Taberna et al, 2020) shows that most flood-ABMs lack primary data & behavioral theories, and omit firms (vital in providing jobs, critical for indirect damages and recovery). The current literature focuses on incremental adaptation, omitting transformational CCA needed for climate resilient societies.
b.We developed the CRAB model - the 1st ABM to explore agglomeration forces, climate shocks and adaptation nexus. The 1st theoretical version (Taberna et al. 2022A and Taberna et al. 2022B) modelled 2 regions (Coastal and Inland). The novel contribution is the inclusion of endogenous migration dynamics of both firms and households that generated self-reinforcing agglomeration forces. CRAB analyzes this under various climate shocks, finding a non-linear relationship between frequency and severity of hazards and economic performance. The 2nd single-region version includes household CCA and is connected to ERC cases using data on flood hazards, regional capital-labor ratios for economic sectors, and our surveys. We parameterized CRAB to Miami (US) and explored different framing regarding household CCA behavior to floods finding that empirical adaptation diffusion among households is below the conventional optimal (under review in PNAS). Secondly, we calibrated the model to Shanghai (CN) and studied the synergies of CCA actions across scales. This is the first ABM that includes direct flood damages that dynamically vary depending on agents' CCA actions, indirect damages, and recovery, which evolve following the interactions among different stakeholders in the aftermath of the disaster.

3) Quantifying macro-economic impacts of private adaptation:
a.We develop methods to quantify indirect economic impacts in CGE beyond the traditional demand-driven effects from disturbed supply chain linkages. Instead, we aim to quantify the impact of widening public deficits due to diminished tax bases and increased spending due to climate-driven losses, and the second-order effects on the borrowing capacity of households, the public and business sectors. Our goal is to examine synergies between CCA measures and the prevention of further economic turmoil due to debt spirals in CGE.
b.The standard CGE model structure is being revised to accommodate households and firms CCA, considering available databases. In addition, we seek to modify the actors’ consumption functions to allow borrowing, particularly for the public sector in the aftermath of a disaster.

Expected results:
1) Unique dataset eliciting autonomous CCA behavior, potentially capturing changes in perceptions and choices over time;
2) Open access regional ABMs that go beyond modeling agglomeration forces by (i) accommodating climate shocks, and by (ii) applying theory- and data-driven climate adaptation decisions of households and firms to the four cases.
3) Advance indirect damage assessment and CCA in macro CGE/IAM, potentially linking them with micro ABMs to quantify cumulative effects of private adaptation, thresholds in regional resilience and structural shifts in damage functions of IAMs.
4) Dissemination: conference presentations, invited talks, TV interviews, practitioner magazines, the website and twitter communication, peer-reviewed articles & 3 PhD theses.
Nature CC Diagram.png
The methodological setup of SCALAR. Source: