Policy Portfolio Builder

Forward-looking estimates for each sleeve, not historical averages. Expected return is the long-horizon annual return assumed for the asset class over the policy horizon, volatility is its assumed annual standard deviation, and the correlations describe how the sleeves are assumed to move together. Published assumption sets from asset managers are a common starting point. Whatever source is used, all four sleeves need to come from one consistent framework, since mixing sources produces a covariance structure that does not hold together.

It is a simplifying assumption that keeps the covariance matrix tractable while remaining close to reality for a short-dated cash sleeve, whose value barely moves with equity or credit markets. It is not exactly true: cash yields respond to the same policy rate that influences bond prices, so a small relationship exists. The effect on a portfolio result is minor at typical cash weights and grows as the cash allocation grows, which is worth remembering when testing a high-cash mix.

Expected return is a linear function of the weights, so combining assets averages their expected returns exactly, with no interaction term. Variance is quadratic: it depends on every pair of holdings through their covariances, so the combination can be lower than any weighted average of the individual volatilities. That asymmetry is diversification. It is also why the correlation inputs affect the volatility and Sharpe outputs while leaving the expected return figure untouched.

It measures how much portfolio volatility would change for a small increase in that sleeve's weight, which is not the same as the sleeve's own volatility. A holding that moves against the rest of the portfolio can carry high standalone volatility and a low, even negative, marginal contribution. Multiplying each marginal contribution by its weight gives contributions that sum to total portfolio volatility, which is what makes them comparable across sleeves.

It means a higher expected excess return per unit of assumed volatility, given the inputs entered. Whether that describes a better policy depends on things the calculation does not see: whether the mix meets a required spending rate in absolute terms, whether its drawdown behavior is tolerable, whether the sleeves can actually be held and rebalanced, and whether the assumptions themselves are reasonable. A ratio computed from optimistic inputs is a statement about the inputs.

The math assumes a fully allocated portfolio, so the weights are shares of one whole. A set summing to less than 100% implicitly leaves capital unmodeled, and the expected return and volatility figures would then describe only the part that was entered while being read as though they described everything. Holding a deliberate cash position is modeled by putting that share into the cash sleeve rather than by leaving the total short.

The tool works in four aggregate sleeves, so a finer allocation is represented by mapping each holding into the closest sleeve and blending its assumptions by weight. That approximation is reasonable when the components inside a sleeve behave similarly and misleading when they do not, since blending averages away the differences that made them separate allocations. A mix where private assets and listed real assets sit in the same sleeve is the common case where the approximation costs the most.

Portfolio volatility responds directly to them, and the response grows with the weights of the two sleeves involved. Changing the equity to bond correlation across the range it has historically occupied can move portfolio volatility noticeably for a balanced mix, while leaving expected return unchanged. Running the same weights across a range of correlation values, rather than one point estimate, is the practical way to see how much of a result depends on that single assumption.

No. The figures describe a gross portfolio built from the assumptions entered. Fund expenses, advisory fees, taxes on distributions and the cost of rebalancing all reduce what an investor actually receives, and they do so in ways that depend on the vehicles, account types and turnover involved rather than on the allocation alone. Comparing two mixes here compares them before those effects, which can differ materially between the two.