In this chapter we continue down the path of returns-based expected utility optimization. We will create the joint return distribution forecast to feed into our optimizer, alongside the client utility function. Our starting point for this process is the full set of historical monthly returns for the assets we have selected. We introduce two techniques that help diagnose whether the historical return series is a reliable source for future expectations. The chapter ends with a review of how to modify our historical monthly return dataset for custom forecasts, manager selection, fees, and taxes.
- In a returns-based asset allocation framework we require a forecast for the full distribution of future joint returns as input to our optimizer.
- We deploy historical return distributions as our baseline forecast, given the simplicity and zero approximation error introduced into the estimation process.
- Bootstrapped standard errors are introduced to measure whether historical returns are providing sufficient estimation accuracy.
- Historical return distributions are only useful forecasts if they are stationary—that is, their properties are stable over long periods, a feature we test with the Kolmogorov–Smirnov test.
- We adjust historical distributions to account for custom market views, manager alpha, and fees via a simple shift of the return distributions.
- Taxes are accounted for via a more nuanced combination of shifting and scaling of the ...