Ridge Regularization or Weight Constraints? Portfolio Optimization in a Small Frontier Market

This paper examines whether ridge regularization improves portfolio performance in the Croatian equity market and whether it adds value when individual portfolio weights are already capped. We compare long-only global minimum variance (GMV) and minimum conditional value-at-risk (CVaR) portfolios using CROBEX constituent returns from 2004 onward, covering 43 index revisions and six estimation and holding-window schemes. The regularization parameter is selected adaptively using information from previous holding periods. Performance is assessed against corresponding unregularized portfolios, an equally weighted portfolio, and the CROBEX index. We also examine turnover, transaction costs, and crisis performance. Under the baseline 30% maximum-weight constraint, adaptive ridge regularization does not produce a statistically significant improvement in Sharpe ratios over the corresponding unregularized portfolios. For CVaR portfolios, the weight constraint accounts for a substantial part of the observed Sharpe-ratio improvement, while the additional benefit of ridge regularization is limited and varies across specifications. Both mechanisms have only small effects on GMV performance. Although selected CVaR specifications outperform CROBEX in cumulative returns, GMV generally provides better realized downside protection. These results suggest that ridge regularization and maximum-weight constraints are partly substitutable in this small frontier market.