An Empirical Study on the Construction of A Non-Convex Risk Parity Portfolio Using A Genetic Algorithm

Abstrak

Risk-based portfolio optimization has become increasingly crucial due to the limitations and underperformance of traditional Mean-Variance (MV) portfolios. In the context of the Risk Parity (RP) portfolio, capital is allocated in such a way that each asset contributes equally to the overall risk of the portfolio. However, optimizing a non-convex RP portfolio presents significant challenges. While conventional numerical methods can be applied, they often struggle with inefficiency and fail to deliver optimal results. This study addresses these challenges by proposing the Genetic Algorithm (GA) and comparing its effectiveness against Successive Convex for RIsk Parity (SCRIP) to solve the non-convex RP portfolio. Using stock price data from companies listed on the Jakarta Islamic Index (JII), we demonstrate that the GA, although yielding a solution that slightly deviates from the ideal equal-risk-contribution, offers a more efficient and practical approach than SCRIP. Notably, the RP portfolio outperforms both the Equal Weight (EW) and Global Minimum Variance (GMV) portfolios in terms of lower Value at Risk (VaR) and Turnover (TO) ratios. This research contributes to the field by showcasing the GA as a viable tool for overcoming the limitations of traditional optimization methods, particularly in addressing the complexities of non-convex RP portfolios while considering practical investment constraints.