MULTI-AGENT BASED MODELING APPLIED TO PORTFOLIO SELECTION IN THE DOOM-LOOP OF SOVEREIGN DEBT CONTEXT*

ABSTRACT This study explores the self-fulfilling dynamic between sovereign debt risk and rational choices of neutral, risk-seeking and risk-averse investors, with implications to the systemic risk emergence. The agent-based model parameterization includes investment strategy (randomly selected assets, stock exchange participation, economic segment, and technical analysis), portfolio rebalance period, and stop gain/loss option. We use Brazilian markets data from 2006 to 2017 to simulate stochastic distributions of investments by a set of 3,000 agents in both stages of model verification and validation (robustness check). Using the Capital Asset Pricing Model, we confirmed our proposition that the optimal rational risk attitude (less risk appetite) constitutes a trigger for the self-fulfilling dynamic, having its foundation on government securities yield and in the debt dynamics. This finding is contrary to the equity premium puzzle in the Brazilian case. The findings have implications to policymakers regarding systemic risk issues, among other public policies.

摘要 本研究探讨了主权债务风险与中性、风险追逐型、风险厌恶型投资者的理性选择之间的自我实现动态，并分析其对系统性风险生成的影响。基于智能体的模型（agent-based model）的参数化设置涵盖投资策略（随机选取标的资产、证券交易所参与行为、经济板块配置与技术分析）、投资组合再平衡周期，以及止盈止损机制（stop gain/loss option）。本研究采用2006年至2017年的巴西市场数据，在模型验证与确认（稳健性检验）两个阶段，对3000个智能体（agent）的投资随机分布情境进行模拟。借助资本资产定价模型（Capital Asset Pricing Model），本研究验证了提出的命题：最优理性风险态度（更低的风险偏好）构成了自我实现动态的触发因素，其理论基础植根于政府证券收益率与债务动态。该发现与巴西市场的股权溢价之谜相悖。本研究结果不仅对系统性风险相关议题的政策制定具有参考价值，也可为其他公共政策提供借鉴。