The research addresses a major challenge in investment management: conventional portfolio models can become less reliable when financial markets experience sudden and extreme disruptions. The COVID-19 pandemic created precisely such conditions, with sharp market declines, capital outflows, and rapidly changing relationships between asset prices in Indonesia. The researchers therefore tested whether an adaptive computational approach could help investors identify better portfolio choices under crisis conditions.
Why Traditional Diversification Can Struggle During Crises
For decades, investors have relied on Modern Portfolio Theory, developed by Harry Markowitz, to balance expected returns against investment risk. The basic idea is straightforward: investors can combine assets to seek higher returns without taking unnecessary risk.
However, extreme market conditions can undermine the assumptions behind conventional models. During a financial crisis, assets that normally move differently may suddenly decline together. This reduces the protective effect of diversification and can cause traditional optimization models to produce portfolios that are less suitable for rapidly changing conditions.
Indonesia experienced severe market disruption during the early COVID-19 period. The Indonesia Stock Exchange faced dramatic declines in March 2020, accompanied by large capital outflows and heightened market uncertainty. The LQ45 index, which represents highly liquid and highly capitalized Indonesian companies, also showed substantial differences in performance across sectors.
How the NSGA-II Approach Works
Piero and Yanuarta RE used a quantitative computational approach based on historical stock-price data. The researchers selected companies that remained consistently included in the LQ45 index during the observation period to reduce the risk of survivorship bias.
BBCA and BRPT Dominated the Optimal Portfolio
The results showed substantial differences among individual stocks during the crisis period.
BBCA recorded an annualized expected return of 17.37 percent and the lowest standard deviation in the sample at 28.73 percent. In contrast, BRPT produced the highest expected annualized return, reaching 62.92 percent, but also carried the highest risk, with a standard deviation of 63.83 percent.
The strongest portfolio generated by NSGA-II achieved:
- Sharpe Ratio: 1.034
- Expected annualized return: 42.96 percent
- Standard deviation: 41.52 percent
- Main allocation: 39.03 percent BBCA
- BRPT allocation: 55.06 percent
- EXCL allocation: 5.45 percent
- KLBF allocation: 0.13 percent
The portfolio therefore did not simply divide money equally among available stocks. Instead, the algorithm produced a concentrated combination of assets with different risk characteristics. The allocation chart on page 12 of the journal article shows BBCA and BRPT as the dominant holdings, with much smaller positions in EXCL and KLBF.
The researchers describe this combination as a strategic balance between stability and growth potential. BBCA provided a comparatively lower-risk component, while BRPT contributed substantially higher expected returns. EXCL and KLBF received smaller allocations, reflecting patterns associated with telecommunications and healthcare demand during the pandemic.
At the same time, the algorithm assigned zero or nearly zero allocations to several heavily affected stocks. ITMG, for example, had an expected annualized return of -36.12 percent, while BBTN recorded -29.91 percent.
Potential Applications for Investment Management
The findings suggest that evolutionary algorithms could become useful decision-support tools for institutional investors facing periods of extreme market uncertainty. Instead of relying on a single static portfolio, investors could use an adaptive system to generate multiple risk-return alternatives as market conditions change.
Piero and Yanuarta RE argue that NSGA-II can help portfolio managers respond to severe market disruptions by efficiently exploring complex combinations of assets and investment constraints. The approach could potentially be incorporated into algorithmic investment systems or robo-advisory platforms to support data-driven portfolio adjustments.
“Rather than suggesting a uniform distribution,” the authors’ analysis shows, the algorithm identified strategic concentrations in resilient defensive stocks while allocating substantial capital to higher-growth assets. In the context of Universitas Negeri Padang’s research, the result demonstrates how evolutionary computation can serve as an adaptive alternative for portfolio decision-making during systemic crises.
However, the results should not be interpreted as a guarantee of future investment returns. The researchers acknowledge that their analysis focuses on the 2019–2020 crisis period and uses only two primary objectives: return and variance. Transaction costs and other market frictions were not incorporated. Future research could test the approach across longer economic cycles and add factors such as liquidity and environmental, social and governance (ESG) considerations.
Author Profiles
Diego Armando Piero is affiliated with the Master of Management Program, Universitas Negeri Padang, Padang, Indonesia. His research in this article focuses on portfolio optimization, evolutionary algorithms, computational finance, and investment decision-making under market stress. The article does not specify his academic degree beyond his affiliation with the Master of Management program.
Ramel Yanuarta RE is also affiliated with the Master of Management Program, Universitas Negeri Padang and serves as the corresponding author of the article. His work in this study covers portfolio optimization, evolutionary computation, financial markets, and risk-adjusted investment strategies. The publication does not state his academic degree.
0 Komentar