We solve for optimal consumption and portfolio choice in a life-cycle model with short-sales and borrowing constraints; undiversifiable labor income risk; and a predictable, time-varying, equity premium and show that the investor pursues aggressive market timing strategies. Importantly, in the presence of stock market predictability, the model suggests that the conventional financial advice of reducing stock market exposure as retirement approaches is correct on average, but ignoring changing market information can lead to substantial welfare losses. Therefore, enhanced target-date funds (ETDFs) that condition on expected equity premia increase welfare relative to target-date funds (TDFs). Out-of-sample analysis supports these conclusions.

We study quantitatively how uncertainty in expected stock return predictability affects life-cycle portfolio choice and wealth accumulation in the presence of undiversifiable labor income risk. Households filter information about future expected returns from observed predictors and realized stock returns. Therefore, optimal portfolio choice does not only depend on financial wealth and age, as in more traditional life-cycle models. Counterfactuals demonstrate the magnitude of portfolio demand changes that depend on perceptions about underlying expected returns. On average, life-cycle asset allocation becomes more conservative than models with either i.i.d. stock returns, or with clearer signals about expected stock returns.

Using a calibrated, collective life-cycle portfolio choice model for a dual-income couple, we show that an increase in the ability to share risk within the household due to a mean-preserving spread in the partners’ coefficients of relative risk aversion leads to a substantial increase in financial risk-taking. Importantly, we show that risk sharing has a larger economic impact on portfolio choice than risk diversification. While unitary models usually do not fully replicate the optimal portfolio choice of collective models, we propose approximations that work reasonably well for moderate background risk. We provide strong empirical support for our key findings.