Risk Parity: Asset Classes versus Risk Factors
Traditional asset allocation varies the asset class weights when optimizing the portfolio for efficiency. A problem arises from this methodology because each asset class can possess exposure to other risk factors. Equities may have commodity and currency exposure, while corporate bonds can contain interest rate and equity factors. Disentangling the factor exposures requires robust statistical analysis. More important is whether this increased complexity matters. The answer is addition by subtraction.
Exhibit 1. Traditional Portfolio Risk Factor Attribution
Source: Capital Risk calculations. The portfolio is 60% Global Equities (ACWI) and US Aggregate Bonds (AGG), which is further subdivided into ten assets classes. The period is June 2010 to June 2020 and weights are as of June 2020.
The standard asset allocation portfolio of 60% equities and 40% bonds clearly shows the factor loadings (exhibit 1). Despite the portfolio allocating over a quarter of the portfolio to international equities, the equity risk factor accounts for nearly 93% of the total. While this number is smaller than the risk contribution shown earlier, it still accounts for almost all the exposure in one factor. The ten asset classes in the portfolio reduce to only four factors, which is a triumph of parsimony over complexity.
Every great recipe is contingent on the ingredients and the ingredient’s proportions. Since return premia are time-varying, the mix of factors must also change over time. An asset classes’ relevant factors will change as the market environment changes, and the managers dynamically adjust their views. Factors are no different. Thus, the number of relevant factors and their weights will vary over time.
Determining how to adjust the factor exposures over time is the domain of financial theory and statistical analysis. The usual techniques to calculate the factor loadings are regression analysis using data known at that time. The range of possible methods is unbounded. Leaning towards simplicity in model selection reduces the required assumptions. Tangible benefits accrue to this decision, including a ready explanation of the linkages between factors and indices, reduced model risk from parsimony in factor selection, and reduced implementation and transaction costs.
More significant to the investor is the selection of the appropriate risk level. All investors have unique objectives for their portfolio due to their differing return requirements and risk appetites. A well-balanced risk factor portfolio may only require a small completion allocation to risk parity to achieve its target portfolio efficiency. Another investor may seek risk parity with a higher volatility level due to the need to improve portfolio efficiency and generate a higher return. Thus, each investor must determine their level of risk parity that benefits their portfolio.
There is not a general rule for an investor on what level of risk parity to target. Even though investors share similar factors exposures, the amount of their exposures varies. Crucially, the varying levels of risk parity, transparency, and a continual communication of the risk factor exposures empower the investor with the means to tailor the exposures to their needs. The potential benefit is increased diversification in their portfolio.