Two-Fund Separation

A wealth manager with a thousand clients, each with a different appetite for risk, seems to face a thousand different portfolio problems. A retiree wants low volatility, a young trader wants to swing for the fences. Building a custom basket of stocks and bonds for each one sounds unavoidable. It turns out to be completely unnecessary.

The analogy before any symbols

A coffee shop could stock a hundred pre-mixed espresso drinks of every strength. Instead it stocks two things: a single concentrated shot and a pitcher of hot water. Every customer's preferred strength, from timid to jet-fuel, is just a different ratio of shot to water. The shop never needs a second concentrate, because blending one strong thing with one neutral thing already spans every strength anyone could want.

Portfolio choice works the same way once a risk-free asset (cash, or a T-bill) is available. The "concentrate" is one specific basket of risky assets. The "water" is cash. Every investor, no matter how risk-averse, ends up holding some mix of just those two.

The mechanics

Without a risk-free asset, the set of best possible risky portfolios is a curved boundary called the efficient frontier: a hyperbola in return-versus-volatility space, and a more risk-averse investor needs a genuinely different mix of risky assets, not just less of the same one, to sit on it.

Add a risk-free asset earning $r_f$. Combining it with any risky portfolio produces a straight line, because expected return and volatility both scale linearly with the risky weight. The straight line that reaches farthest, the one tangent to the risky frontier, dominates every curved point below it. That tangent point is the tangency portfolio, call it $T$. Every investor, regardless of risk aversion, should hold only $T$ and cash, in a weight $w$ chosen to match their own comfort with risk:

$E[R_p] = r_f + w\big(E[R_T] - r_f\big), \qquad \sigma_p = w\,\sigma_T.$

In words: the portfolio's expected return is cash's return plus a slice of the tangency portfolio's extra return, and its volatility is just that same slice applied to the tangency portfolio's own volatility. $w>1$ means borrowing at $r_f$ to hold more than 100% in $T$; $w<0$ means holding more cash than wealth, i.e. shorting $T$.

Drag the correlation slider and watch the curved frontier bend; the straight tangent line from the risk-free point barely moves in direction, because the tangency portfolio's composition is a property of the risky assets alone, not of any individual investor's risk tolerance.

Worked example: two investors, one fund

Tangency portfolio $T$ has expected return 12% and volatility 20%; $r_f = 4\%$. A cautious investor wants portfolio volatility of 10%: $w = 10/20 = 0.5$, so half in $T$, half in cash, giving expected return $4 + 0.5(12-4) = 8\%$. An aggressive investor wants volatility of 30%: $w = 30/20 = 1.5$, meaning borrow 50% of wealth at $r_f$ and put 150% into $T$, giving expected return $4 + 1.5(12-4) = 16\%$. Both investors hold the same risky basket, just scaled differently.

Worked example: reward-to-risk is shared

Because both portfolios lie on the same line, they share one number: the line's slope, $(12-4)/20 = 0.4$ per unit of volatility, which is exactly $T$'s Sharpe ratio. The cautious investor's Sharpe ratio is $(8-4)/10 = 0.4$; the aggressive investor's is $(16-4)/30 = 0.4$. Identical. Separation means risk aversion changes the amount of risk taken, never the efficiency of that risk.

What this means in practice

This is the theoretical justification for target-date funds and robo-advisors: build one well-diversified "growth" fund, then dial risk with a cash weight rather than reformulating the risky basket per client. It also underlies the CAPM logic that, under further assumptions, the tangency portfolio becomes the market portfolio itself.