Portfolio construction has re-emerged as an important topic of research in the last few years, as traditional diversification techniques don’t always work when there are strong correlations within the market. The tactic of organizing a portfolio around distinct risk premia provides a possible alternative to sector or geographic diversification, but eliminating unwanted correlations isn’t always an easy task.
Standard approach to gain risk premia
“The standard approach to gaining risk premia exposure is through ranking stocks based on quantifiable characteristics and then buying equal amounts of the top quintile whilst shorting the bottom. This roughly equates to zero market exposure whilst gaining premia specific exposure,” write Citi analysts Chris Montagu and Liz Dinh. But the quintile approach ignores information in the center of the distribution and averages out information contained in the top and bottom quintiles. There’s also a chance that style exposure may be “infected,” with unintended correlations to other risk premia.
Put a simpler way, using the quintile method doesn’t protect you from correlations between risk premia, and you could end up taking a position on one risk premium intentionally while another, unintended position comes along for the ride.
Yarra Square Partners returned 19.5% net in 2020, outperforming its benchmark, the S&P 500, which returned 18.4% throughout the year. According to a copy of the firm's fourth-quarter and full-year letter to investors, which ValueWalk has been able to review, 2020 was a year of two halves for the investment manager. Q1 2021 hedge fund Read More
Montagu and Dinh propose a new approach in a recent Citi paper, constructing a portfolio using what they call pure risk premia.
Pure risk premia indices construction
“Pure risk premia indices are constructed such that they provide exposure to one particular risk premia and zero exposure to other premia. The significant benefit to this is that taking a view on one particular sector, style or macro attribute does not lead to additional unintended views which can significantly increase risk,” they write.
The breakdown starts by deconstructing systematic risk into four components: macroeconomic risk, regional risk, sector risk, and style risk. Macro risk and style risk are somewhat subjective, but in their analysis macro risk is composed of long rates, GDP growth, credit spreads, oil price, commodity prices, and the dollar-euro exchange rate. Style risks include value, growth, quality, low risk, price momentum, estimates momentum, and size, each of which is defined more precisely according to company-specific factors.
Regressions against the universe of stocks
Two separate regressions are run against the universe of stocks that you’re interested in, one dealing with macroeconomic factors, and the other with style risk. “There is a clear distinction between macro factors and style risk premia in that macro sensitivities are market observed variables whereas style risk premia are mispriced anomalies which are thought to generate risk adjusted returns in excess of the market. It is therefore intuitive to keep them separate,” they write.
These regressions can be used to create risk premia indices with independent (orthogonal) style exposures, essentially breaking style risks into eigenvalues. The result is much lower pairwise correlation.
Montagu and Dinh argue that also ridding the style risk premia of macroeconomic correlations, instead of just sector and regional correlations, would result in little more than noise. “It is desirable to have style risk premia with sensitivity to macroeconomic factor exposures, as long as these exposures are not so large that they introduce significant exogenous volatility,” they write.