This paper analyzes the safety-first portfolio model under two different target assumptions, the fixed target, which is commonly assumed in the literature, and the random target, which has played only a minor role so far. As both targets can be easily motivated, the open question is, which target choice leads to a better performance? We answer this question by comparing optimal expected portfolio returns of the fixed and the random target strategy. Assuming multivariate normal returns the answer is: (1) The random target strategy outperforms the fixed target strategy if the portfolio return and the random target are positively correlated and riskless investing is prohibited, (2) the fixed target strategy outperforms the random target strategy if the portfolio return and the random target are not positively correlated and riskless investing is allowed. The first result is practically most relevant, in particular for institutional portfolio management and the skilled private investor, which is supported by an empirical analysis. Furthermore, we show that these results also hold when relaxing the normal assumption.
safety-first portfolio optimization random target benchmarking