Risk parity
Weighting holdings by how calm they are, so each adds about the same amount of risk.
5 min readLesson 54 of 57
In short
A stock half as volatile gets twice the weight. Compared with equal weights, it leans toward steadier holdings, so no single wild one dominates the swings. It aims for balanced risk, not the highest return.
Why it matters
Putting equal money into each holding sounds balanced, but the risk isn't. If one stock swings three times as much as another, it drives most of the portfolio's ups and downs, even at the same weight. Equal dollars can mean very unequal risk.
Risk parity sets the weights so that each holding adds about the same amount of risk instead. It's one of the main alternatives to weighting by money or by size, and some of the largest funds in the world use versions of it to spread risk across stocks, bonds and other assets.
How it works
Work out each holding's Volatility, then weight it by one divided by that volatility, scaled so the weights add up to 100%. A stock half as volatile as another gets twice the weight. Since each holding's weight times its volatility comes out the same, each contributes about the same amount of swing.
MarketLab's Risk parity backtest uses the standard deviation of each stock's last 63 daily returns, about three months, and resets the weights at the start of every month or quarter, as you choose. Between resets, the weights drift with prices, as in Rebalancing. It waits in cash until there are 63 returns to measure.
This simple version ignores Correlation: it treats each holding's risk on its own. Two closely related stocks each get their full share, so together they add more risk than an unrelated one. Full risk parity, used by the Optimise tab in the portfolio's Risk analysis, accounts for how holdings move together and sets weights so their risk contributions come out equal, within its cap on any one stock.
Because risk parity leans toward calmer holdings, it usually swings less than equal weights. It doesn't aim for the highest return: calm stocks are often slower growers, so in a strong rally led by volatile stocks it tends to lag.
Optimisers chase noise. The Optimise tab's Least risk and Best Sharpe weights come from an optimiser, which takes estimates of each stock's return, volatility and correlations and finds the best mix for them. Those estimates come from a few years of noisy prices, and the optimiser can't tell a real pattern from luck: it piles into whatever happened to rise most with the least swing. So weights that were optimal in hindsight usually disappoint in the years that follow. Expected returns are the worst offender, being far harder to estimate than risk.
Careful optimisers tone their estimates down. MarketLab's Optimised weights backtest pulls each expected return halfway towards the stocks' average and each correlation a quarter of the way towards theirs, never goes short, and caps any one stock at 40% (80% with only two). Each month or quarter it re-optimises from the last 12 or 36 months of prices only, for least risk or best Sharpe ratio, holding equal weights until 3 months of prices exist. The chart compares it with equal weights rebalanced as often, a bar optimisers often fail to clear.
weight of a holding = (1 ÷ its volatility) ÷ (sum of 1 ÷ volatility over all holdings)
The calmer a holding, the bigger its weight, so each adds about the same risk.
A worked example
You hold two stocks: a steady one with a volatility of 15% and a volatile one with 30%. Compare equal weights with risk parity, then add a third stock.
- Equal weights: 50% × 15% = 7.5% of swing from the steady stock, and 50% × 30% = 15% from the volatile one. The volatile stock contributes twice as much.
- Risk parity: 1 ÷ 0.15 ≈ 6.67 and 1 ÷ 0.30 ≈ 3.33, which add up to 10. The weights are 6.67 ÷ 10 ≈ 66.7% and 3.33 ÷ 10 ≈ 33.3%.
- Check the contributions: 66.7% × 15% = 10% and 33.3% × 30% = 10%. They're equal.
- With three stocks at 10%, 20% and 40% volatility: 1 ÷ volatility gives 10, 5 and 2.5, which add up to 17.5. The weights are 10 ÷ 17.5 ≈ 57.1%, 5 ÷ 17.5 ≈ 28.6% and 2.5 ÷ 17.5 ≈ 14.3%.
Result: Risk parity puts about two thirds in the steady stock, so each holding adds the same 10% of swing. With three stocks, the calmest gets four times the weight of the wildest, because it's a quarter as volatile.
How to read it
- One holding with a very large weight
- It's much calmer than the others. Check that it isn't just a stock that happened to be quiet for three months.
- Weights close to equal
- The holdings have had similar volatility, so risk parity and equal weights give almost the same portfolio.
- Lower volatility than equal weights
- The usual result: leaning toward calm holdings smooths the ride.
- Lower return than equal weights
- Common when the volatile holdings led the market. Balanced risk is the aim, not the top return.
- Optimised weights trailing equal weights
- The usual out-of-sample story: the optimiser's estimates were mostly noise.
In the real world
- Bridgewater's All Weather fund, launched in 1996, popularized the idea across asset classes: rather than the classic 60% stocks and 40% bonds, where stocks drive most of the risk, it balances risk between stocks, bonds and commodities, often using borrowing to lift the calmer bond side.
- Risk parity funds had a hard year in 2022, when stocks and bonds fell together. Balancing risk only spreads it if the holdings don't all fall at once, which is why correlation matters.
- Many equity funds use the simpler inverse-volatility weighting described here, sometimes as part of low-volatility strategies, because it needs only each holding's own volatility, which is far easier to estimate than a full set of correlations.
- A 2009 study by Victor DeMiguel, Lorenzo Garlappi and Raman Uppal tested 14 optimisation methods out of sample and found none consistently beat simply splitting money equally.
Common mistakes
- Thinking equal risk means low risk. A risk parity portfolio of five volatile tech stocks is still very volatile; it just spreads that risk evenly.
- Forgetting correlation. The simple inverse-volatility version can still concentrate risk in a group of related holdings.
- Trusting three months of volatility too much. A stock that was calm lately can turn volatile, and the weights only catch up at the next reset.
- Expecting it to beat the market. It's built to balance risk, and it often trails in strong rallies led by the most volatile stocks.
See it in MarketLab
Backtest: risk parity →
Opens a monthly risk parity backtest of NVIDIA, Apple, Coca-Cola and Johnson & Johnson, compared with holding them in equal amounts.
Portfolio Optimise tab →
Under Risk analysis, the Optimise tab compares your weights with Least risk, Best Sharpe and Risk parity weightings of the same stocks, and plots each on a chart of risk against return. Its Test the same idea over time links open the matching backtest.
Backtest: optimised weights →
Re-optimises the same four stocks for best Sharpe ratio each month from the last 36 months of prices, compared with equal weights rebalanced monthly.
Test yourself
Optional: 5 quick questions on the words, using the idea, and how it plays out in real markets.
Loading your answers…