The Effects of Illiquidity and Lock-Ups on Portfolio Weights

Using several recently proposed portfolio policies, we study the effects of a partial lock-up, or transaction costs, on portfolio performance and the weight of the illiquid asset in both in- and out-of-the-sample tests. We use REITs as a proxy for a low liquidity asset. Our first approach follows that of Brandt and Santa-Clara (2006) and de Roon, Guo and ter Horst (2009). In an unconditional setting, we find that the weight for our illiquid asset is in general lower that in prior studies, and is reduced to values close to empirically observed weights for real estate, once a lock-up is introduced, producing a potential answer to an earlier puzzle. Our results also indicate that the Brandt and Santa-Clara (2006) methodology and the Kan and Zhou (2007) shrinkage strategy, suggested as more stable than many traditional asset allocation approaches, still result in extreme and unstable weights and reduced out-of-the-sample performance. The Brandt, Santa-Clara and Valkanov (2009) model with rebalancing costs, which we apply to an asset allocation problem, substantially reduces the noise from conditional signals, and improves performance.