3. Methodologies and Tools for Integrated Systems Modelling

Alan Kirman
Rob Dellink
Jean Chateau
Sebastian Barnes
Elena Rovenskaya
Ulf Dieckmann
Bas van Ruijven
Keywan Riahi
Fabian Wagner

Integrated quantitative modelling of coupled socio-environmental systems is a useful way to enhance our understanding of complex systems and to generate systems-based policy advice. OECD and IIASA both develop and make extensive use of such models to devise policy options and to provide outlooks in a variety of areas, including economic growth, employment, agriculture, energy, and the environment, as well as their interrelations. Models can also help organise and structure discussions around so-called “wicked” problems1.

Model suites at OECD and IIASA include simulation models, optimisation models, and statistical models spanning wide ranges of geographical and temporal scopes and granularities, from local to global and from annual to centurial. This chapter discusses the role of integrated systems modelling in informing policy decisions and looks at how extending existing models and developing new tools based on systems approaches could yield additional insights.

What is integrated systems modelling? In the context of systems analysis, a model is a simplified representation of reality expressed in equations and/or computer code, focusing on the key mechanisms at play. Any model builds on a conceptual skeleton, which defines the system’s boundaries, external conditions (parameters), internal elements (variables), and the relations between them. Such a qualitative skeleton is turned into quantitative equations or transition rules based on theories and data. Implemented in computer code, a model can be used for analysing outcomes and for gaining a better understanding of a system’s response to specific policy interventions. A systems approach often challenges a system’s traditionally considered boundaries, bringing in elements that may lie outside the immediate area of interest, but have the potential to impact that area of interest. For example, integrating a water model with an energy model enables a more realistic evaluation of future energy production, because water availability may limit production possibilities.

The OECD model suite. At OECD, statistical models and analytical models based on micro-foundations are two major types of models used to support economic policy. Statistical models have the advantage of often having medium complexity, as such models describe the properties of aggregate economic variables directly. This approach fell out of favour in recent decades because it may lead to unreliable results, as highlighted by the Lucas Critique that emphasised the problems caused when inferring future behaviour from the past in the presence of structural breaks (Lucas, 1976). Nevertheless, statistical modelling remains an important approach for devising large-scale macroeconomic models and empirical research on aggregate-level phenomena. Structural models have become the dominant approach to analyse macroeconomic issues, particularly through two types of general equilibrium models called dynamic stochastic general equilibrium (DSGE) models (Cacciatore et al, 2012) and computable general equilibrium (CGE) models, respectively. The use of different model types is dictated by data available and the purpose of modelling – forecasting or policy evaluation.

At OECD, some of the main applications of economic modelling to support policy include:

  • Short-term macroeconomic forecasting and policy simulations are based on reduced-form large-scale macroeconomic models, such as the NIESR (National Institute of Economic and Social Research) NiGEM model (National Institute’s Global Econometric Model) (Barrell and Pain, 1997).

  • Long-term macroeconomic models that provide a globally consistent set of long-run projections for potential output and other variables (Guillemette and Turner, 2018).

  • Trade policy issues are analysed by using static CGE models, such as the METRO model (ModElling TRade at the OECD).

  • Agriculture and food projections are produced by means of the Ag-Link model, which is a dynamic-recursive partial equilibrium model developed in collaboration with the United Nation’s Food and Agriculture Organization (FAO) and used for the annual OECD-FOA Agricultural Outlook series.

  • Projections and policy analysis focused on environmental issues are done using the ENV-Linkages model, which is a large-scale dynamic CGE model with global coverage and more than 50 economic sectors, supported by the ENV-Growth model, a Solow-type macroeconomic growth model.

The IIASA Integrated Assessment Modelling Framework. IIASA has a long tradition of modelling physical systems related to energy, air pollution, land use, and water, and of developing methodologies, such as scenario analyses, to apply these modelling tools. The main large-scale IIASA modelling tools are coupled within an integrated systems framework and include:

  • Global and national energy-systems modelling with a process-based, dynamic, systems-engineering, integrated assessment model of the global energy-economy-environment system.

  • Air-quality modelling based on global and regional emissions and technologies with GAINS (Greenhouse gas-Air pollution INteractions and Synergies).

  • Agriculture, food, and land-use modelling with GLOBIOM (GLObal BIOsphere Management Model), a global model to assess competition for land use among agriculture, bioenergy, and forestry.

  • Global water modelling with CWatM (Community Water Model), including the evolution of future water demand and availability.

These tools are complemented by other models, including stylised models, medium-complexity stock-flow-consistent models, and agent-based models (ABMs).

Why is a next generation of modelling tools needed? While the suite of models outlined above has proven to be very useful in providing insights into a wide range of topics, enhancements as well as new methodological developments are needed to improve the policy realism and relevance of these tools. In particular, next-generation systems analysis models need to focus on a better integration of real-world dynamics such as social and behavioural heterogeneity, which will help in representing social dynamics and complex collective decision-making, and thereby facilitate evaluating the effectiveness of policies and their systemic impacts.

Any modelling effort necessarily involves a compromise between the intention to represent the properties of a problem adequately and the need for the resultant model to remain interpretable as well as feasible in terms of its implementation. Mainstream modelling approaches often focus on stylised assumptions that improve tractability, but do so at the expense of sacrificing some of the richness of the behaviour of the modelled systems. With current computing power, rising data availability, and accumulated knowledge on the boundedly rational behaviour of economic agents, it is tempting to relax the simplifying assumptions of mainstream models in order to reflect reality better. The following innovation dimensions seem particularly important, as detailed in the next two sections:

  • Linking and integrating models through so-called nexus studies should give them the appropriate breadth to cover multiple domains together.

  • The focus of modelling should increasingly move to capturing complex dynamic interrelations and interdependences among agents operating within a system’s boundaries, and the resultant models should be able to describe profound changes among these interactions that give rise to phenomena such as emergence and adaptation.

An important aspect of how models meant to inform policy decisions can be developed concerns their co-design in collaboration with stakeholders and decision makers. Such an approach helps ensure that models are well designed from a technical perspective, are based on appropriate intellectual ownership and buy-in, and address relevant policy concerns in a way that effectively communicates to the associated policy communities.

A promising approach to modelling sustainable-development strategies is to integrate existing modelling tools from different fields, such as linking environmental models with economic growth and trade models or linking demographic and economic models. This extends the boundaries of what is modelled and allows for broader ranges of interactions that help reveal indirect vulnerabilities or strengths of policy interventions.

Potential for OECD-IIASA collaboration. Both OECD and IIASA develop and operate wide arrays of modelling tools, which are largely complementary to each other, as summarised above. OECD has a long tradition in more economy-focused models that describe interactions between macroeconomic growth and interlinked sectoral and regional economic activity. The tools at IIASA have a more physical focus and describe long-term dynamics in areas such as population, energy, technology, air pollution, water, natural disasters, ecology, agriculture, and land use. In the following, we highlight several integration dimensions by focusing on examples that showcase the potential for OECD-IIASA collaboration.

Linkages with energy models. Researchers at IIASA and OECD have been deeply involved in developing socioeconomic scenarios for climate-change assessment, the so-called shared socioeconomic pathways (SSPs). First, IIASA developed population and education projections for five long-term scenarios. Second, these served as inputs to an OECD macro-economic growth model (ENV-Growth), which explicitly accounts for the energy revenues accrued by fossil-fuel exporters2. Third, the resultant economic projections were used in IIASA’s Integrated Assessment Modelling Framework to produce projections of energy demand, energy supply, land use, and greenhouse-gas emissions under different levels of mitigation. Several additional efforts have also focused on coupling energy models and macroeconomic tools. Energy-systems models have added macro-economic feedbacks for changes in energy prices, and CGE models have been linked to energy-systems models to incorporate more-detailed information on energy-sector transformations. While methods for these model linkages have been well-established (Klinge Jacobsen, 1998; Messner and Schrattenholzer, 2000; Böhringer and Rutherford, 2008; Böhringer and Rutherford, 2009; Kypreos and Lehtila, 2015), so far they have focused on a limited number of variables that are harmonised and exchanged between the coupled models. This limits the types of analyses that can be performed within these frameworks and leaves room for major mismatches between the energy system and other economic elements.

Linkages with air-pollution models. Another example of recent collaboration is the use of IIASA’s GAINS model to estimate air-pollution emission factors that were then used in OECD’s ENV-Linkages model to project air-pollution emissions and the damages from air pollution until 2060. This very fruitful collaboration has produced a widely used set of climate-change assessment scenarios, distributed through IIASA’s community databases, as well as a report on air pollution that has been instrumental in increasing policy attention to air pollution. However, these collaborative efforts were also one-directional exercises, which left out several key interactions that will be important to explore in the future. These include feedbacks from changes in energy production and consumption and land use on economic activity; feedbacks from economic development on population and migration; and feedbacks from climate-change impacts on energy and agriculture on economic activity and population.

Expected benefits of model integration. Developing novel methods for linking the economic modelling tools of OECD with the physical system models of IIASA would add great value to both institutions. There is room for improvement of the existing methods, especially when the information exchanged between the models is expanded to cover not only energy supply, but also investments, distributional consequences for different household groups, and minimum energy-consumption requirements (such as those needed for achieving decent living standards; see Rao and Min, 2017). Furthermore, such integration would open up the possibility to link economic activity, ecosystem services, and biodiversity losses (see the chapter by Karousakis et al. elsewhere in this volume)3.

By linking the flagship modelling tools of IIASA and OECD, important new avenues of interdisciplinary policy questions can be addressed. This will allow researchers to treat systems barriers to economic growth more seriously, e.g., through feedbacks from environmental degradation, interactions with energy exports for countries whose national incomes heavily rely on fossil fuels, and feedback effects between demographics, education, and economic activity. It will also allow them to improve the economic backbone of the biophysical systems analysis done by IIASA. Brought together, these advances will help produce more robust policy insights. Among other benefits, the highlighted dimensions of model integration will be important for addressing the challenges discussed in Chapter 2 (promoting different dimensions of human well-being); Chapter 4 (modelling in support of the United Nation’s Sustainable Development Goals and of examining the linkages among them); Chapter 5 (understanding critical interconnections between water and food systems and their implications for biodiversity); Chapter 6 (exploring the air-pollution implications of ecosystem dynamics and energy transformations); and Chapter 10 (understanding the diverse implications of digitalisation).

Opportunities for strengthening the modelling capacities of IIASA and OECD beyond the integration of existing tools may involve pioneering applications and innovative methodologies and tools in several key areas, as outlined below.

Explicit accounting for uncertainty. Stochastic optimisation can be used in decision-support tools to derive so-called robust decisions (informally referred to also as no-regret decisions). Robust decisions allow a satisfactory outcome of a process to be achieved irrespective of the particular realisation of uncertainty that is actually observed. At IIASA, this approach has been applied to designing insurance markets mitigating against natural disasters (Ermolieva et al., 2016a); land-use planning (Ermolieva et al., 2016b); and evaluating energy investment portfolios (Krey and Riahi, 2013).

Multiplicity of agents with strategic interactions. Evolutionary game theory can be used to describe the behaviour of agents pursuing their interests by making strategic decisions based on observing other agents. The resultant strategic interactions among agents often lead to social dilemmas collectively known as the ‘tragedy of the commons’: when agents pursue their selfish interests, collective interests typically get jeopardised. This phenomenon is of universal relevance, being germane to managing key challenges associated with the many common goods on which humankind depends. Examples are as diverse as mitigating climate change, securing clean air, preventing environmental pollution, managing sustainable land use, exploiting renewable resources, achieving prudent urbanisation, ameliorating family planning, protecting social welfare, and governing the internet. At IIASA, this approach has been applied, for instance, to designing innovative incentive systems for protecting common goods (Chen et al., 2015) and analysing the threats of institutional corruption (Lee et al., 2019).

Bounded rationality, including consumption preferences and consumer choices. Behavioural economics and agent-based models (ABMs) emphasise and make use of the fact that individuals act in ways that are not fully rational, including by following simple heuristics rather than optimising their behaviours. In particular, people may make their decisions based on the actual or perceived behaviours of others, rather than by acting independently, as assumed in mainstream economics. For example, decisions about changing to more sustainable behaviours may depend on whether household neighbours are perceived as taking similar measures. If there are thresholds in people’s actions depending on actions taken in their neighbourhood, this can qualitatively alter outcomes at the systems level. ABMs are built on the assumption that numerous agents interact following simple behavioural rules in a well-defined environment. Individual behaviours, their interactions, and changes to the structure of these interactions can give rise to rich systems dynamics, including emergence and high levels of complexity. ABMs can capture heterogeneities among agents that play an important role in how a system evolves. They can also predict implications across such distributions of agents. ABMs can typically only be solved through simulations, rather than analytically, but increased computing power has made this a viable approach even for large-scale models. ABMs can be applied to a vast array of policy problems. OECD and IIASA have used ABMs and other simulation models to suggest instruments to reduce financial systemic risk (Poledna and Thurner, 2016), to analyse transport systems (ITF, 2017, McCollum et al., 2017) and the potential for the development of shared mobility, and are developing ABMs to look at other issues, including interactions between the real economy and the financial system.

Complex interconnections and systemic risks. Network theory demonstrates that a system’s structure, in terms of the linkages between its elements, is important for determining how it responds to exogenous disturbances. Network theory also suggests how to use information about an agent’s position in a network to specify policies or actions that are particularly suitable for that agent. Compared with mainstream models that rely on simple or uniform interactions among agents, network theory yields insights for more realistic relationships, for example, when some agents in a network are much better connected to other agents or are more central to the network as a whole than in simpler configurations of interconnections. With a network’s structure determining how broadly or narrowly its interactions transmit the impacts of shocks (e.g. Kharrazi et al. 2017), policy interventions can be designed to modify this structure to yield desired outcomes.

The integration of existing tools and the application of innovative methods will be critically important for generating new policy-relevant insights. For example, they can help map out causal linkages that were not well articulated in traditional models, such as the feedback between economic activity and natural resources.

These improved modelling approaches become increasingly feasible as computing power keeps growing, making it easier to solve complex models or to simulate the behaviour of a large number of agents. The computational techniques themselves have developed alongside the computational technology. Improved data, including big data, and a growing understanding of the behaviour of key system elements, including human behaviour, can be used to validate, calibrate, and enrich these models.

Detailed models can be complemented by stylised models that provide useful qualitative insights and help enhance the intuitions and insights of researchers and policymakers about how economic, environmental, and social systems behave and interact.

While being essential for next-generation models, the better incorporation of social sciences and of heterogeneities among agents poses new challenges. First, behaviour is almost always context-specific, which means that data requirements for quantifying behaviour are often large and difficult to generalise. Second, system-level properties can be remarkably sensitive to detailed assumptions made about such behaviour, and modellers must hone their comprehension of those sensitivities and their implications. Given the current – still rather incomplete – understanding of human behaviour, the heuristics implemented in models may turn out to be too ad hoc and, given their potentially large impact on model results, need to be subjected to carefully designed robustness checks.

Models need to be formally validated on a routine basis. This may pose challenges due to large uncertainties and since current econometric tools often lack precision. The growing availability of salient data, however, provides a better basis for calibrating complex tools, including network models and ABMs. The development of machine-learning techniques creates opportunities to explore data without committing to rigid assumptions. In particular, machine-learning models are very versatile for detecting and investigating nonlinear features of empirical systems.

Complementing the greater availability of raw data, more empirical studies are needed to estimate causal relationships and determinants of change that can guide quantitative future projections involving human behaviour, e.g. regarding the evolution of diets, new technologies, individual preferences, or social norms.

Modelling tools at IIASA and OECD will continue to be essential for a systems-based assessment of the pending transformation towards sustainable development. A coherent strategy across the institutions requires the extension and integration of existing tools as well as the application of innovative methodologies to better represent real-world complexities in regard to heterogeneity, uncertainty, strategic interaction, bounded rationality, and network structure. There is a clear need for different tools that can answer different questions. It is thus important to build on existing strengths while in parallel pursuing new methodological developments.


Acheson, H. (2008) “Strategic Policy Intelligence –setting priorities and evaluating impacts – Ireland”. Mimeo. http://www.oecd.org/science/inno/41379672.pdf

Foray, D. and D. Mowery (2012), “Public R&D and social challenges: What lessons from mission R&D”, Research Policy 41(10), pp. 1697-1702

Gallagher KS, Grubler A, Kuhl L, Nemet GF, & Wilson C (2012). The Energy Technology Innovation System. Annual Review of Environment and Resources 37: 137-162. http://pure.iiasa.ac.at/id/eprint/9882

Grubler A & Fuss S (2014). Technology portfolios: Modelling technological uncertainty and innovation risks. In: Energy Technology Innovation: Learning from Historical Successes and Failures. Eds. Grubler, A & Wilson, C, Cambridge: Cambridge University Press http://pure.iiasa.ac.at/id/eprint/11079

Grubler A & Wilson C (2014). Energy Technology Innovation: Learning from Historical Successes and Failures. Cambridge: Cambridge University Press. ISBN 9781107023222 http://pure.iiasa.ac.at/id/eprint/11142

Grubler A, Wilson C, Bento N, Boza-Kiss B, Krey V, McCollum D, Rao N, Riahi K, et al. (2018). A low energy demand scenario for meeting the 1.5 °C target and sustainable development goals without negative emission technologies. Nature Energy 3 (6): 517-525. http://pure.iiasa.ac.at/id/eprint/15301

Krey V & Riahi K (2013). Risk hedging strategies under energy system and climate policy uncertainties. In: Handbook of Risk Management in Energy Production and Trading. Eds. Kovacevic, RM, Pflug, GC & Vespucci, MT, pp. 435-474 USA: Springer. http://pure.iiasa.ac.at/id/eprint/10557

Mazzucato, M. (2015). A mission-oriented approach to building the entrepreneurial state. Report Commissioned by UK Government, United Kingdom Government

Nilsson M, Chisholm E, Griggs D, Howden-Chapman P, McCollum D, Messerli P, Neumann B, Stevance A-S, et al. (2018). Mapping interactions between the sustainable development goals: lessons learned and ways forward. Sustainability Science 13 (6): 1489-1503 http://pure.iiasa.ac.at/id/eprint/15381

OECD (2016) OECD Science, Technology and Innovation Outlook 2016, OECD Publishing, https://dx.doi.org/10.1787/sti_in_outlook-2016-en

OECD (2018), Meeting Policy Challenges for a Sustainable Bioeconomy, OECD Publishing, Paris, https://doi.org/10.1787/9789264292345-en

OECD-ITF (International Transport Forum) (2016). Shared mobility, Innovation for Liveable Cities. OECD https://www.itf-oecd.org/node/20046

OECD-ITF (International Transport Forum) (2018). The Shared-use City: Managing the Curb. OECD https://www.itf-oecd.org/shared-use-city-managing-curb-0

Parkinson S, Krey V, Huppmann D , Kahil T, McCollum D, Fricko O, Byers E, Gidden M, et al. (2018). Balancing clean water-climate change mitigation tradeoffs. Environmental Research Letters (in press). http://pure.iiasa.ac.at/id/eprint/15591

TWI2050 - The World in 2050 (2018). Transformations to Achieve the Sustainable Development Goals. Report prepared by The World in 2050 Initiative. IIASA Report. International Institute for Applied Systems Analysis (IIASA). Laxenburg, Austria. http://pure.iiasa.ac.at/id/eprint/15347

Wilson C, Grubler A, Gallagher KS, & Nemet GF (2012). Marginalization of end-use technologies in energy innovation for climate protection. Nature Climate Change 2 (11): 780-788 http://pure.iiasa.ac.at/id/eprint/9860


← 1. A wicked problem is a problem that is difficult or impossible to solve because of incomplete, contradictory, and changing requirements that are often difficult to recognise (Rittel & Webber 1973).

← 2. These economic projections were not official OECD forecasts and not approved by OECD governments. Rather, they were unofficial projections made using the OECD modelling framework (Dellink et al. 2017).

← 3. At a technical level, the open database platforms of IIASA are proving to provide an excellent model-linkage framework, in which these new approaches could be implemented.

Metadata, Legal and Rights

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area. Extracts from publications may be subject to additional disclaimers, which are set out in the complete version of the publication, available at the link provided.


The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at http://www.oecd.org/termsandconditions.