Houweling has extensive experience in managing factor credit strategies, as he is also the portfolio manager of conservative credits mandates, which were launched in 2012. The Conservative Credits strategy exploits the Low-Risk factor and, in line with the Robeco philosophy not to go against other factors, also has exposure to Value, Momentum and Size.
Disciplined investment process
Robeco Global Multi-Factor Credits uses a disciplined, rules-based investment process, which starts with a global universe of all USD, EUR and GBP credit bonds, some 14,500 in total.
Improving the universe
The next step is enhancing the investment universe compared with the traditional investment grade index by removing long-maturity bonds and adding BB-rated bonds.
We filter out the longest maturities, because historically long-maturity bonds had lower returns and higher volatilities than shorter-maturity bonds (see e.g. Ilmanen, Byrne, Gunasekera & Minikin (2004), Frazzini & Pedersen (2014), Houweling, Van Vliet, Wang & Beekhuizen (2015)). We exclude the longest 10% to avoid going against this effect.
BB-rated issuers are added to the investment universe for three reasons. First, it allows us to hold on to bonds after they have been downgraded from investment grade to BB, rather than sell them when rating-restricted investors are selling as well. These ‘fallen angels’ often subsequently outperform. Second, the inclusion of BBs allows us to select ’rising stars’: low-risk BBs with attractive valuations and strong momentum that are more likely to upgrade to investment grade. Last but not least, historically BBs had the highest Sharpe ratios of all investment grade and high yield credit ratings.
Co-Head of Quant Fixed Income and Lead Portfolio Manager
For all factors, we uses enhanced definitions compared with the standard academic definitions.
Quant ranking model
The bonds that are in the investment universe after the previous step are evaluated by our quantitative multi-factor model. The model prefers low-risk companies (Low-Risk), bonds whose market credit spread is above the ’fair spread’ (Value) and recent winners from the equity market (Momentum). Exposure to the Size factor is incorporated via the portfolio construction step rather than in the ranking model. Each bond gets a score on each factor and bonds are ranked on their total model score. The model ranks across ratings, because we find that our risk measures in the Low-Risk factor are superior to ratings in discriminating between low-risk and high-risk companies. The model separately evaluates financials and non-financials, because of differences between their balance sheets and business models.
Like in all Robeco quant strategies, we perform a human check on the model. We check on data quality (e.g. missing guarantees between issuing entity and parent company), bond characteristics (e.g. liquidity) and additional risks that may not be captured by the model, such as mergers & acquisitions, country or sector concentration risks and ESG-related risks.
Finally, in the portfolio construction stage, we build a diversified portfolio of 150-200 names. We target an equal weight per name (subject to a limit of 10% of an issuer’s total debt), so that the fund overweights small companies, and underweights large companies, efficiently capturing the Size factor. We take into account liquidity and transaction costs. Furthermore, we apply various constraints; table 1 gives an overview of the most important constraints. Bonds are held until their model ranking drops to the lowest 20% of the universe, the credit rating drops below BB-, or portfolio managers or analysts identify a strong downside risk that is not captured by the model.
Table 1 | Portfolio construction: constraints
Note: Above mentioned risk limits are internal guidelines. Limits mentioned in the prospectus are leading.
Portfolio simulation results
We tested the actual investment strategy of Robeco Global Multi-Factor Credits as realistically as possible on historical data, taking into account illiquidity and transaction costs of individual bonds. Table 2 shows the portfolio simulation results (net of transaction costs) over the 2002-2014 research period. The strategy generated a superior Sharpe ratio of 0.5 versus 0.1 for the investment grade index. Excess returns vs. government bonds are higher than those of the index at similar volatility. On average, the strategy had a market sensitivity (credit beta) of around 1, which indicates that the strategy earned most of the outperformance from issuer selection rather than from beta allocation. The beta is measured using our ‘risk point’ methodology (see our white paper ‘Smart credit investing: Risk points and their applications’). The strategy has a preference for bonds with shorter durations (benefiting from the Low-Risk factor) and higher credit spreads (benefiting from the Value factor). On average these preferences result in a credit beta of around 1, although periodically it will be above 1 or below 1. Because the interest rate duration of the selected bonds is on average below the index duration, we use interest rate derivatives (swaps and/or futures) to hedge the portfolio duration to the index duration. All FX risk is hedged to the share class currency.
*January 2002 – December 2014. IG Index = Barclays Investment Grade index; Multi-Factor Credits = Investment Grade + BB universe. Excess returns are calculated vs. duration-matched government bonds. Returns, volatilities, Sharpe ratios, turnover and transaction costs are calculated on monthly data and subsequently annualized. Strategy returns are net of transaction costs, but gross of management fees. The value of your investments may fluctuate. Results obtained in the past are no guarantee for the future.