5. Conclusions
We proposed a two-step hybrid investment strategy for pension funds, in which the asset allocation is decided via optimization (the active component). And within each asset class, we follow a passive investment approach based on a representative index. Rebalancing is done once a year based on information from the preceding three years. Also, we tested two metrics (Variance (R) and CVaR(R)) to control risk in the optimization problem. Part of the motivation for our strategy is the overwhelming evidence in favor of passive versus active investment, but also the recognition that asset allocation plays a fundamental role in selecting the portfolio composition. The proposed strategy is an attempt to combine these two key concepts.
The performance of our strategy was compared to the performance of each of the five funds that constitute the backbone of the Chilean pension system, a fairly well established (and highly regarded) defined-contributions scheme. The results indicated that our strategy did remarkably well. In three of the five funds (the three riskiest funds), our approach produced better absolute returns (compared to the returns experienced by the Chilean pension managers), higher Sharpe ratios, and lower volatilities. In the two cases in which our strategy was outperformed by that of the Chilean AFPs, we showed that portfolio allocations limits set by the regulator played a significant role. In fact, the large value associated with the Lagrange multipliers corresponding to those constraints suggested that a small relaxation of the constraints limits would result in a much better performing portfolio. And that was indeed the case. This conclusion is important for it implies that regulators should exercise extreme care when setting the portfolio allocation limits.
