constraint
Expected Return
The forecasted average return an asset is projected to produce, used as the optimizer input for upside.
Expected return is the average yearly return a model projects for an asset or a portfolio over a future period. It is an estimate, and actual returns scatter widely around it from year to year. In portfolio optimization it is the reward input that the optimizer weighs against risk.
For a portfolio, expected return is the weighted sum of each holding's expected return: multiply each weight by that holding's estimate and add up the results. The hard part is the estimate for each asset. Common sources include historical averages, valuation-based estimates such as dividend yield plus long-run dividend growth (the Gordon growth model), analyst price targets, factor models and equilibrium returns implied by market prices, as in the Black-Litterman model. Blending several sources and pulling extreme values toward the middle can help, since each source on its own is noisy.
Treat any single figure as a central guess with wide error bands. A stock with 30% volatility can easily miss a 9% expected return by 30 points in a given year, and a long-run average needs decades of data to pin down within a few points. Optimizers are very sensitive to these inputs: raising one asset's estimate by a point or two can shift a large share of capital into it. An expected return that looks unusually high can reflect a recent run-up or an aggressive target, so check the inputs before trusting a result.
Example with illustrative numbers: a portfolio holds 50% in Stock A with an expected return of 8%, 30% in Stock B at 11% and 20% in a Treasury bill fund at 4%. Its expected return is half of 8% (4.0%) plus 30% of 11% (3.3%) plus 20% of 4% (0.8%), or 8.1%. If B's estimate were 14% instead, the portfolio figure would rise to 9.0%, and an optimizer would likely want more of B.
On GNG, the Portfolio Optimizer builds a forward-looking expected return for each name by blending three inputs when all are available: a dividend-based estimate (yield plus five-year dividend growth, with growth capped at 10% and the input used only when the yield is at least 0.5%) at 45%, the return implied by analyst price targets at 35%, and the historical average return, with extreme values capped, at 20%. Turning on the Momentum tilt blends in recent price momentum. Estimates are capped at 50% a year and floored at -30%, and cash-like funds are set to the risk-free rate. The portfolio's Expected return leads the results, and the default Black-Litterman engine blends returns implied by market capitalizations with either your own views or these estimates.
Mean-Variance Optimization turns expected returns into weights, and the Lookback Window sets how much history feeds them.
Related terms
More constraint terms
Get full access to GNG Research
Create a free account to access portfolio tracking, advanced tools, and more.