Risk Parity Allocator

True diversification. Traditional portfolios are dominated by equity risk. Risk parity balances risk across asset classes so no single asset drives portfolio performance. This typically results in smoother returns and smaller drawdowns.

Because bonds have lower volatility. To contribute equal risk, bonds need much higher dollar weight than stocks. If stocks have 18% vol and bonds 6%, you need ~3× more bonds to balance the risk contribution.

Yes, but returns may be lower. Without leverage, a risk parity portfolio might have 6-8% volatility vs 10-12% for 60/40. You get smoother returns but potentially lower absolute returns. Some add commodities to boost volatility naturally.

Monthly or quarterly is typical. Rebalancing too frequently increases costs. Many use threshold-based rebalancing - only trade when risk contributions drift by more than 5% from target.

Typically: equities (domestic and international), bonds (government and possibly corporate), commodities (gold, broad commodities). The key is including assets with different risk characteristics and low correlations.

RC_i = w_i × (Σw)_i / σ_p, where w is weight vector, Σ is covariance matrix. For equal risk: solve for weights where all RC_i are equal. Python packages like riskfolio-lib can handle this.

Inverse volatility (1/σ weighting) assumes zero correlation and is a simple approximation. True risk parity accounts for correlations in the covariance matrix. For diversified portfolios with moderate correlations, the difference is often small.

Calculate risk parity portfolio volatility. If target is 10% and portfolio vol is 7%, scale all weights by 10/7 = 1.43×. Total weights exceed 100%, requiring leverage (margin, futures, or leveraged products).

Correlations change over time, especially in crises (tend to spike). Use rolling or EWMA estimation to adapt. Some systems use regime-switching models to detect and adapt to correlation changes.

Historically, risk parity has shown higher risk-adjusted returns (Sharpe ratio) due to better diversification. Absolute returns depend on leverage. In equity bull markets, 60/40 may have higher returns; in diversified environments, risk parity often wins.

Options: Ledoit-Wolf shrinkage (toward diagonal or factor), exponential weighting (EWMA for faster adaptation), DCC-GARCH (dynamic correlations), factor models. Shrinkage methods are popular for stability.

Cyclical coordinate descent is simple and effective. Newton-Raphson is faster but needs good initialization. The Spinu (2013) algorithm is efficient. Python: use scipy.optimize or riskfolio-lib.

Options: 1) Margin borrowing (costly), 2) Futures (efficient, lower cost), 3) Leveraged ETFs (not recommended for long-term due to decay). For Canadian retail, margin or lower vol target without leverage is practical.

Use walk-forward: estimate covariance with past data only, calculate weights, apply to next period. Avoid look-ahead bias. Test across multiple market regimes. Compare to 60/40 and equal weight benchmarks.

1) Heavy bond allocation hurts in rising rate environments, 2) Leverage adds risk and cost, 3) Correlation breakdown in crises, 4) Estimation error in covariances. Mitigate with dynamic allocation and multiple estimation methods.