{
  "title": "AI and Macroeconomic Modeling: Deep Reinforcement Learning in an RBC model",
  "publication": "IMF Working Papers, February 24, 2023",
  "sourceUrl": "https://www.imf.org/en/publications/wp/issues/2023/02/24/ai-and-macroeconomic-modeling-deep-reinforcement-learning-in-an-rbc-model-530084",
  "canonical": "https://www.imf.org/en/publications/wp/issues/2023/02/24/ai-and-macroeconomic-modeling-deep-reinforcement-learning-in-an-rbc-model-530084",
  "overlayPath": "/en/publications/wp/issues/2023/02/24/ai-and-macroeconomic-modeling-deep-reinforcement-learning-in-an-rbc-model-530084/index.md",
  "summary": "This study seeks to construct a basic reinforcement learning-based AI-macroeconomic simulator. We use a deep RL (DRL) approach (DDPG) in an RBC macroeconomic model. We set up two learning scenarios, one of which is deterministic without the technological shock and the other is stochastic.",
  "sections": [
    {
      "heading": "Study objective and methodological approach",
      "content": "- Objective: construct a basic reinforcement learning-based AI-macroeconomic simulator.\n- Method: use a deep RL (DRL) approach (DDPG) in an RBC macroeconomic model.\n- Learning algorithm highlighted: DDPG (Deep Deterministic Policy Gradient), an actor-critic algorithm."
    },
    {
      "heading": "Experimental design: scenarios and comparison benchmark",
      "content": "- Two learning scenarios:\n  - Deterministic scenario: without the technological shock; objective is to compare the learning agent's behavior to a deterministic steady-state scenario.\n  - Stochastic scenario: includes stochastic technological shocks.\n- Benchmark: deterministic steady-state outcomes used for comparison with the agent's learned behavior."
    },
    {
      "heading": "Main findings and results",
      "content": "- Agent behavior:\n  - In both deterministic and stochastic scenarios, the agent's choices are close to their optimal value.\n  - The study also presents cases of unstable learning behaviours.\n- Implication: DRL (DDPG) can approximate optimal policy choices in an RBC setting under both deterministic and stochastic environments, but learning stability can be an issue in some cases."
    },
    {
      "heading": "Policy and research implications / recommendations",
      "content": "- Model enhancements suggested for future research:\n  - Add additional variables or sectors to the model.\n  - Incorporate different DRL algorithms beyond DDPG.\n- Research role: this AI-macro model serves as a foundation for exploring AI-driven macroeconomic simulation and can be extended to analyze richer environments or alternative learning algorithms."
    },
    {
      "heading": "Subjects and keywords",
      "content": "- Subjects: Artificial intelligence, Asset and liability management, Debt relief, Economic theory, Labor, Machine learning, Rational expectations, Technology\n- Keywords: Actor-critic algorithms, AI experiment, AI-macroeconomic simulator, Artificial intelligence, DDPG, Debt relief, Deep deterministic policy gradient, Deep reinforcement learning, DRL, Global, learning algorithm, Learning algorithms, Machine learning, Macro modeling, Rational expectations, RBC, RBC model, Real business cycles, Reinforcement learning, RL\n\nIMF Working Papers — \"AI and Macroeconomic Modeling: Deep Reinforcement Learning in an RBC model\", Tohid Atashbar and Rui Aruhan Shi, February 24, 2023.\n\n---\n\n Content in this bundle\n\n- Working Paper\n  - Working Paper (Markdown version){rel=\"alternate\" type=\"text/markdown\"}\n  - Working Paper (PDF){rel=\"external\" type=\"application/pdf\"}\n\n---\n\nSource: https://www.imf.org/en/publications/wp/issues/2023/02/24/ai-and-macroeconomic-modeling-deep-reinforcement-learning-in-an-rbc-model-530084"
    }
  ],
  "bullets": [
    "[Markdown version](/en/publications/wp/issues/2023/02/24/ai-and-macroeconomic-modeling-deep-reinforcement-learning-in-an-rbc-model-530084/index.md)",
    "[Structured JSON version](/en/publications/wp/issues/2023/02/24/ai-and-macroeconomic-modeling-deep-reinforcement-learning-in-an-rbc-model-530084/index.json)",
    "[Bundle manifest](/en/publications/wp/issues/2023/02/24/ai-and-macroeconomic-modeling-deep-reinforcement-learning-in-an-rbc-model-530084/bundle-manifest.json)",
    "Authors: Tohid Atashbar, Rui Aruhan Shi",
    "Published: February 24, 2023",
    "Series: IMF Working Papers",
    "DOI: https://doi.org/10.5089/9798400235252.001",
    "Objective: construct a basic reinforcement learning-based AI-macroeconomic simulator.",
    "Method: use a deep RL (DRL) approach (DDPG) in an RBC macroeconomic model.",
    "Learning algorithm highlighted: DDPG (Deep Deterministic Policy Gradient), an actor-critic algorithm.",
    "Two learning scenarios:",
    "Benchmark: deterministic steady-state outcomes used for comparison with the agent's learned behavior.",
    "Agent behavior:",
    "Implication: DRL (DDPG) can approximate optimal policy choices in an RBC setting under both deterministic and stochastic environments, but learning stability can be an issue in some cases.",
    "Model enhancements suggested for future research:",
    "Research role: this AI-macro model serves as a foundation for exploring AI-driven macroeconomic simulation and can be extended to analyze richer environments or alternative learning algorithms.",
    "Subjects: Artificial intelligence, Asset and liability management, Debt relief, Economic theory, Labor, Machine learning, Rational expectations, Technology",
    "Keywords: Actor-critic algorithms, AI experiment, AI-macroeconomic simulator, Artificial intelligence, DDPG, Debt relief, Deep deterministic policy gradient, Deep reinforcement learning, DRL, Global, learning algorithm, Learning algorithms, Machine learning, Macro modeling, Rational expectations, RBC, RBC model, Real business cycles, Reinforcement learning, RL",
    "**Working Paper**"
  ],
  "related": [
    {
      "title": "Working Paper",
      "role": "paper",
      "sourceUrl": "https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023040-print-pdf.pdf",
      "summary": {
        "path": "/-/media/files/publications/wp/2023/english/wpiea2023040-print-pdf.pdf.md",
        "mime": "text/markdown"
      },
      "binary": {
        "path": "/-/media/files/publications/wp/2023/english/wpiea2023040-print-pdf.pdf",
        "mime": "application/pdf"
      }
    }
  ],
  "alternates": {
    "markdown": "/en/publications/wp/issues/2023/02/24/ai-and-macroeconomic-modeling-deep-reinforcement-learning-in-an-rbc-model-530084/index.md",
    "json": "/en/publications/wp/issues/2023/02/24/ai-and-macroeconomic-modeling-deep-reinforcement-learning-in-an-rbc-model-530084/index.json",
    "bundleManifest": "/en/publications/wp/issues/2023/02/24/ai-and-macroeconomic-modeling-deep-reinforcement-learning-in-an-rbc-model-530084/bundle-manifest.json"
  },
  "generatedAtUtc": "2026-09-16T22:14:37.759Z"
}