We have the audacious goal of creating a digital twin of the global economy. The concept essentially revolves around creating a highly detailed, dynamic simulation of the interaction between individual households, firms, banks, governments and central banks. This “digital twin” can then be used as a forecasting tool as well as a sandbox to test hypotheses, policies and financial plans and to predict the potential outcomes of those changes in the real world [1].

How a model of the economy can be used for capital allocation decisions

Modelling of the markets vs modelling of the economy

The market value of assets and firms in financial markets highly correlates with the fundamental value of such assets. This fundamental value itself is a direct function of economic factors such as firms output, the overall interest rate value, inflation, etc. As a result such an accurate model of the economy is a powerful tool for making asset allocation decisions especially in systemic and medium/long term funds where the market value and fundamental value of assets converge.

The current economic forecasting methods are rudimentary for investment management professionals.

The following is a list of differences between ABMs and the traditional economic models:

1. Rich, Micro-level Insights: Unlike conventional models that offer broad strokes, ABMs dive deep into the micro-level, simulating interactions among individual agents like firms, households, banks, and governments. This granularity provides a comprehensive view of the economic landscape, essential for investment strategies in systemic and medium to long-term funds.
2. Dynamic, Real-time Analysis: Traditional quarterly economic outlook reports often become outdated quickly. A digital twin ABM, however, enable real-time analysis of economic trends, adapting to new data and evolving scenarios. This dynamic approach ensures that investment decisions are based on the most current and relevant information since our model receives updates from the knowledge graph that is updated on a daily basis.
3. Understanding Non-linearities and Feedback Loops: The economic environment is characterized by non-linearities and feedback loops, where actions of one agent influence others in unpredictable ways. ABMs capture these complexities, providing a more realistic picture of economic dynamics, crucial for risk assessment and mitigation in investment portfolios.
4. Scenario Analysis and Stress Testing: ABMs excel in scenario analysis, allowing managers to simulate a range of economic conditions and stress test their portfolios against various potential future states of the world. This is invaluable in developing robust investment strategies that can withstand economic volatilities.

Figure 1. A market outlook document that soon becomes outdated after publication compared to a live agent based model of the economy that gets updated and is available through an API.

How we model firms

Our ABM implements a data-driven approach to model each firm (from small private ones to large public companies) similar to approach by Poledna et al. [2,3] :

• Heterogeneous Agents Representing Firms: The firm sector consists of heterogeneous agents representing all firms across various industries in the euro area. Each firm will have its own balance sheet and income statement. The firms interacts through an intricate network of supply chain and input/output connectivity.

Figure 2. Connectivity of firm’s output from various industrial sectors in the EU area. From Poledna et. al. [4]

• Firm Population and Size Distribution: The population and size distribution of firms within each industry are derived from structural business statistics for smaller private companies. For larger ones we use data extracted from filings, news and proprietary database that exists in our knowledge graph.
• Production Function: Each firm produces industry-specific goods using labor, capital, and intermediate inputs from other industries. The production capacity is limited by the availability of labor, intermediate goods, and capital stock.
• Expectations and Adaptive Learning: Firms face fundamental uncertainty regarding future sales, market prices, and cash flow. They form expectations about the future using adaptive learning. In this approach, firms estimate the parameters of their models and make forecasts based on these estimates, including variables like economic growth, inflation, and policy rates.