'The best of two worlds' - alternating between mean variance and risk parity

In our present challenging investment environment, it is important for pension funds and other institutions to make use of the opportunities offered by innovations in asset allocation. In a knowledge-sharing session held in May, Robeco discussed three such new developments: 'mean variance versus risk parity', 'global diversified carry' and 'strategic allocation to commodity factor premiums'. These are innovations that - as the session showed - can all come up trumps by adding value to a portfolio. In this first of three articles we look at the implementation of a market view in a risk parity portfolio.

For a long time, portfolio managers based their asset allocation on 'mean variance' - the method linked to the modern portfolio theory of weighing (expected) risk against return. Roderick Molenaar, portfolio strategist at Robeco, describes this method of portfolio optimization as 'a mathematically satisfactory solution', though one that has fallen in popularity, and not without good reason. 

Starting in the late 1990s, investment professionals increasingly began to favor risk-based solutions in determining their strategic asset allocation. Mean variance solutions had become too sensitive to expected returns: small adjustments led to major portfolio changes. Risk parity solutions (in which expected returns do not play a role) gained in popularity. Triggered by the financial crisis of 2007-2009, this movement took on landslide proportions. Determining the right parameters for portfolio construction turned out to be more difficult than imagined in previous decades. 

"Skepticism about relying fully on a market view increased,” says Molenaar. “Expectations were often uncertain, but they still had a huge impact. You have to be very sure of your view to be able to rely on it completely. Mean variance could also produce a concentrated risk profile in a portfolio. Investment professionals wanted to get away from this." 

‘You have to be very sure of your view to be able to rely on it completely.’ 

This resulted in strategies that partly disregarded expected returns. Risk parity, or as Molenaar prefers, 'equal risk contribution', became popular as a method of setting up portfolios in such a way that the different investment categories held in a portfolio made up an equal risk budget. In practical terms, this often means that exposure to bonds will be much greater relative to equities when compared to a portfolio constructed on the basis of mean variance.

"Choosing a specific strategy is linked to the degree of confidence you have in your data on expected returns, volatility and correlations," says Molenaar. The strategist cites examples of other methods of constructing portfolios, such as maximum diversification, minimum variance, volatility parity (1/σ), risk parity and equal weighting (1/N). The methods differ from each other as regards the extent to which they include data on volatility, correlations and investment expectations in the considerations. In the case of risk parity, for instance, investors will assume that while the available information on volatility and correlations is sufficiently reliable, they cannot forecast expected returns, or can only do so inadequately.

What if an investment professional has a strong view regarding market changes? Risk diversification, a key factor in risk parity, is important, but how can you put your view to use? Risk parity is popular and has been successful for a long time,” says Molenaar. “However, in May 2013 [when the Fed first started talking about 'tapering', i.e. reducing its bond-buying program], the rise in interest rates had a heavy impact on this strategy." 

He expressed a further concern: "The success of risk parity is measured mainly in terms of returns. This seems strange in itself, as portfolio construction under this strategy in fact takes place without considering returns. While with risk parity volatility is low, returns lag. And the Sharpe ratio is high, but you cannot live on that." According to Molenaar, one way of boosting returns here is to use leverage. "But is that what we want? And will the regulator agree to this?” he asks. 

Molenaar thinks that since it seems that interest rates can hardly go any lower, the success of the risk parity method could be in doubt. According to various sources, the chances of rising bond yields in the coming years are considerable. "When confidence in a specific market view is strong, you would do well as an investment professional to implement it,” he says. “But the big question is how to proceed if you have constructed your portfolio on the basis of risk parity. In the strength of risk parity also lies its weakness: it does not use market views, however strong." 

‘In the strength of risk parity also lies its weakness’

According to Molenaar, the Black & Litterman model (which provided a solution in the 1990s to the problems surrounding the assessment of expected returns on investment categories), can also introduce market views into risk parity portfolios. Based on Bayesian techniques, Molenaar and his fellow researchers show how optimized portfolios can be influenced, depending on the amount of confidence there is in such views. The amount of confidence in a view is central to the solution studied. If there is very little or none, the portfolio manager will construct the portfolio around risk parity. If there is substantial confidence, asset allocation will resemble that for a traditional mean variance portfolio. As a function of the degree of confidence, the weights of the investment categories will be determined by a mix of both extremes. 

"Our starting point is a portfolio based on risk parity, and depending on the way confidence in a market view increases, the portfolio will start to resemble the mean-variance portfolio that goes with that projection,” says Molenaar. “This way we can create a bridge between the two strategies. The best of both worlds come together.”

A practical problem that Molenaar highlights is the challenge posed by the quantification of the degree of confidence in a view. Hans de Ruiter, CIO at the TNO pension fund, picks up on this point in his response to Molenaar's research. He points out, for instance, that the findings from the behavioral finance literature show that investment professionals are generally not particularly good at estimating risk parameters. De Ruiter believes that Robeco's research could be supplemented by taking into account in the analysis the relative attractiveness of the different investment categories (e.g. equities performing better than bonds). 

Molenaar responds in a takeaway for pension fund investment professionals that if funds work with risk parity, they would do well to look beyond just risk considerations: "If you have confidence in a view, then put it to use."