Asset and Liability Management (ALM) is a critical framework for pension funds, ensuring they have sufficient assets to meet future liabilities (pension payments) while managing investment risks effectively. This paper utilizes Brazilian data to develop an ALM model specifically for pension funds in the country. The model employs an optimization strategy that minimizes expected contributions made by individuals throughout their working lives. This optimization adheres to cash flow limitations and regulatory restrictions. The objective function leverages a min–max robust optimization approach based on a three-scenario planning scheme inspired by Brazil’s Interbank Rate. We incorporate a machine learning approach based on CMARS to predict confidence intervals for the key stochastic model parameters, particularly those related to the real returns of Brazilian investment classes. The findings empower pension fund managers to formulate well-informed investment strategies. We highlight allocation strategies that can reduce contribution rates without jeopardizing fund solvency, even for managers with a more aggressive risk profile favoring higher stock market allocations. Additionally, the study is enriched by an empirical analysis using data from a Brazilian pension fund, demonstrating the model’s practical application. In short, this model offers valuable insights that can benefit a wide range of pension funds in the Brazilian market, and it could also be applied to similar situations globally.

中文翻译：

---

适用于巴西养老基金的资产和负债管理研究


资产负债管理 （ALM） 是养老基金的关键框架，确保它们有足够的资产来支付未来的负债（养老金支付），同时有效管理投资风险。本文利用巴西数据开发了一个专门针对该国养老基金的 ALM 模型。该模型采用一种优化策略，可最大限度地减少个人在其整个工作生涯中的预期贡献。这种优化遵守现金流限制和监管限制。目标函数利用了基于受巴西银行同业拆借利率启发的三情景规划方案的最小-最大稳健优化方法。我们采用了一种基于 CMARS 的机器学习方法来预测关键随机模型参数的置信区间，特别是与巴西投资类别的实际回报相关的参数。这些发现使养老基金经理能够制定明智的投资策略。我们重点介绍了可以在不损害基金偿付能力的情况下降低缴费率的配置策略，即使对于风险状况更激进、偏好更高股票市场配置的管理人也是如此。此外，该研究还通过使用来自巴西养老基金的数据进行实证分析而丰富，证明了该模型的实际应用。简而言之，该模型提供了有价值的见解，可以使巴西市场上的广泛养老基金受益，也可以应用于全球的类似情况。