This is where the whole project lies. Several different approaches lie here. Small experimental baby steps etc.
LambdaSim's main goal is to act as a digital twin of the world from a
macroeconomic perspective and to simulate future scenarios by taking
advantage of agent based modeing and various methods from multivariat
data analysis.
Throughout human history, we have made remarkable progress in solving analytical problems. For example, in 1977, we successfully calculated the precise trajectories for Voyager 1 and 2, ensuring their successful journeys through space. Similarly, we have built increasingly more advanced architecture thanks to the continual progress of material science, CAD software and general physics. However, while these endeavors are intricate and require detailed calculations, they aren't truly complex. -->
Here is a peek at what's driving this project along with a project timeline.
Complexity arises when there are nonlinear relationships, feedback loops, and unstable dynamics. Examples range from predicting traffic jams in a bustling city to understanding how the weather will unfold based solely on initial conditions.
Fortunately, we can make significant progress in predicting complex systems by breaking down the system into smaller components often called agents. These agents follow a simpler set of rules regarding how they act with their environment. This approach allows us to iteratively simulate the behavior of the entire system. By trying different values for the variables that we can change in real-world scenarios within the simulation, we gain valuable insights about how the system as a whole will be affected. As a consequence, this insight enables us to make better, more informed decisions in the real world.
In science, testing a hypothesis is crucial for proving or disproving a theory. However, in economics, especially on large scales, it is both analytically unpredictable and often very costly or impossible to test new government policies, such as those listed here, before analyzing their real-world implementation results and aftereffects. Additionally, in democratic governments, where new politicians are elected every four years, a policy might be reversed before any meaningful impact, positive or negative, can be measured.
This is the problem that Lambdasim aims to tackle. By creating detailed and dynamic models of economic systems, simulations can predict the potential outcomes of various policies without the need for real-world implementation. These simulations can account for numerous variables and their interactions, providing a more comprehensive analysis of possible results. This approach allows policymakers to evaluate the potential impacts of their decisions in a controlled environment, reducing the risks, costs, and ambiguities associated with trial-and-error in real-world policy testing.
Highly relevant paper
The paper applies an Agent Based Model (ABM) model to simulate Austria’s medium-run economic recovery from COVID-19 lockdowns, finding the ABM’s results consistent with official forecasts by Austrian institutions.
Key Takeaways
- The paper argues ABMs can now be realistic, empirically validated tools for forecasting and policy analysis, challenging the idea that the status quo Dynamic stochastic general equilibrium models (DSGEs) are the only serious condender for economic forcasting.
- ABMs naturally incorporate bounded rationality, agent heterogeneity, and financial frictions—areas where DSGEs struggle without heavy extensions.
Why It’s Important
- This approach makes it possible to model the economy WITHOUT assuming that all agens behave perfectly rationally, while still providing forecasts as good as mainstream models.
- It opens the door for using ABMs in practical macroeconomic forecasting, stress testing, and analyzing the impact of policies at both aggregate and sectoral levels.
Read the Full Paper
Download the PDF of "Economic Forecasting with an Agent-Based Model"
Project Timeline
(Click the blue boxes to expand)
With the purpose of sharing the project with friends, interested people, etc.
Since this is a software-based project, there are no upfront costs. The time is well spent even if there isn't much direction yet. In parallel, the plan is to figure out what might be needed in the market and what already exists.
After brainstorming and research, the project landed on building a relational database — ideal for managing structured data and complex relationships. It ensures data integrity, scalability, and powerful SQL querying.
Thanks to supportive professors at NTNU's IE faculty, LambdaSim was granted as a course worth 7.5 study points, ensuring guidance and credit. More info: TFE4205 - Student Defined Development Project.
Who knows what might happen here?
Hopefully, this will get clearer by the end of September.
Saying no is scary.
I declined an offer to work with a satellite communication system to dedicate more time to LambdaSim in summer 2025.
The project is now quite concrete.
The current goal: develop a macroeconomic model that combines real-time data and complex adaptive systems to predict global trade dynamics and identify how different countries’ economic decisions affect each other.
This paper is very useful and informative for me and this project because it provides detailed insights into economic forecasting using agent-based models, helping to shape the project’s approach and giving a solid theoretical and empirical foundation.
Read the paper directly: View the PDF