What does it mean to budget risk instead of money?
Once you can decompose a portfolio into risk contributions, a different way of building one becomes available. Instead of deciding how to allocate capital, you decide how to allocate risk — and let the capital weights fall out of that decision.
This is risk budgeting, and it's how many institutional allocators and multi-strategy funds frame the problem.
The shift
Capital allocation asks: what percentage of the money goes where? The output is weights, and risk is whatever it turns out to be.
Risk budgeting asks: what percentage of the risk should each component be permitted to produce? The output is a target set of risk contributions, and weights are solved backwards to hit them.
From the previous lesson, a 60/40 portfolio delivers a risk split of roughly 92/8. If an allocator wanted risk split more evenly, the capital weights would have to shift dramatically toward the lower-volatility asset — which is where the arithmetic leads, and also where an important caveat lives.
Equal risk contribution
The best-known target is risk parity: set weights so that every component contributes the same amount to total portfolio risk.
In our two-asset example (equities σ = 16%, bonds σ = 6%, ρ = 0.10), equalising contributions puts the bond weight at 72.7%. That is simply inverse-volatility weighting — (1 ÷ 6) against (1 ÷ 16) — and with only two assets the correlation drops out of the condition entirely: equal contributions require w₁σ₁ = w₂σ₂, in which ρ appears on both sides and cancels. With three or more assets it does not cancel, and the weights have to be solved numerically. The capital allocation looks lopsided; the risk allocation is balanced.
Two honest caveats belong immediately next to that description:
The expected return drops with the volatility. Balancing risk contributions says nothing about balancing returns. Practitioners who run this approach typically add leverage to restore a return target — which introduces funding risk, margin risk and the forced-deleveraging dynamics from Unit 2's 2008 lesson. The risk you removed by diversifying is partially reintroduced as a different kind of risk.
It relies entirely on estimated volatilities and correlations. Both are unstable, and both are estimated from a window. A risk-balanced portfolio built on calm-period inputs is balanced only for the calm period.
Risk budgets as governance
Beyond portfolio construction, risk budgets are a management tool. A multi-desk firm allocates a total risk capacity — expressed in VaR, Expected Shortfall or volatility terms — and divides it:
| Desk | Risk budget | Used | Headroom |
|---|---|---|---|
| Equity | 40% | 36% | 4% |
| Rates | 30% | 31% | −1% |
| Credit | 20% | 12% | 8% |
| FX | 10% | 7% | 3% |
This creates a common currency for a hard conversation. A desk asking for more capacity must argue for it against other desks, in the same units. The alternative — every desk sized in notional money — makes an FX book and an equity book impossible to compare.
It also inherits every weakness of the underlying measure. If the budget is expressed in VaR, the gaming incentive from Unit 1 applies directly: a desk can look compliant while holding exposure concentrated past the VaR cutoff. Serious implementations pair the budget with Expected Shortfall limits, stress-loss limits and concentration limits, precisely so no single measure can be optimised against.
The frame
Risk budgeting does not reduce risk. It makes the allocation of risk explicit and deliberate rather than an accident of capital weights. That is a genuine improvement in visibility, and it is not a form of protection. The budget is still denominated in an estimate.
Try it now
- Take the 60/40 portfolio you built in the last lesson from the same two sleeves, below, and compute its current risk contributions. Write them as percentages of total risk.
- Solve for the weights that would make the two contributions roughly equal — trial and error is fine. Note how far the capital weights had to move.
- Now swap the input rather than the weights. Navigate both charts to September 2008 – March 2009, re-estimate the two volatilities and the correlation from that window, and recompute the contributions. Note how far the "balanced" split shifts when only the window changed. Arithmetic and observation — no allocation here is being recommended.