What RP is NOT

To borrow some wisdom from Charlie Munger, sometimes when we want to understand something, what we need to do is invert the question – in this case, what is the dominant idea in portfolio construction that Risk Parity departs from?

The dominant idea in portfolio construction is officially called “Mean-Variance Optimization,” (MVO), though this is actually one of those cases where something is so widespread and taken for granted that many people who follow it do so without ever knowing its official name. To break the term down, “mean” is better known as “returns” and “variance” as “risk,” so basically it means getting the right balance between risk and reward for each investor. As Obregon, Yontar, and Benham (link coming soon) put it:

MVO is the most popular methodology used by institutional investors to build portfolios. This simple, yet powerful tool creates “efficient” portfolios that attempt to achieve objectives, such as maximum return or minimum risk portfolios, by selecting assets based on their expected return, expected risk (as defined by their standard deviation of returns) and correlations with each other.

The idea first started with Harry Markowitz (link coming soon!) and his paper “Portfolio Selection,” which argued (among other things) that combining higher-risk, higher-return equities and lower-risk, lower-return assets like bonds or cash could produce better risk-return profiles than either asset could by itself. This is undoubtedly true, and as for Risk Parity – so far, so good, as this is a key tenet of a Risk Parity approach, too.

From this basic idea advanced by Markowitz came a way of looking at portfolios where you essentially had two elements: one to drive the return, and then something to dampen the ups and downs that inevitably accompany the search for return. If stocks have an average volatility (as measured by standard deviation) of 15, and an investor can handle the ride that comes with that, then a 100% equity portfolio might be the recommendation. If they only wish for half of that volatility, then they would be in a 50/50 portfolio. The safe portion, often in cash and sometimes just as a general bond fund, is included in the portfolio as a kind of thermostat, dialed up when the investor wants/needs stability, and dialed back when the investor is focused on growth.

When put like that, this is not a very controversial concept, and RP adherents would definitely agree with the broad strokes, but the MVO way of thinking about portfolios is missing some things. For one, with one asset class for growth and another that is essentially inert, you wind up still getting all the risk from equities (see Qian 2005). With risk, MVO approaches ask “how much?” there is, and then find the right amount of safer assets to get to the right level of total portfolio risk. Risk Parity approaches, meanwhile, would focus on “what type?” of risk we’re talking about, and would then try to balance the risks out. This, to me, is a better question, since not all risk is the same.

Second, and more importantly, is that once people are locked into a MVO framework, they tend to only move within that framework by sliding up or down on the risk scale. If returns are disappointing, they might move towards more equities, and after facing a huge decline, they may be tempted to dial back the risk by going towards more fixed-income. This will likely ensure that they are buying high and selling low, for one, and also closes off alternative approaches to building their portfolios. With that sliding scale of risk/return in mind, they find themselves unable to think of other ways to actually getting the risk-reward balance correct.

In all, while I don’t think it’s fair to say that RP is directly opposed to MVO, Risk Parity is different. RP has more of a focus on analyzing types of risk and better understanding correlations between asset classes, all the while being more open to alternative asset classes beyond just one for growth and one for stability.

For more on MVO, here is my write-up of the paper by Obregon et al. that explains it.