The Truth About Fixed Income Factors

Luca Bosatta, our Head of Multi-Asset Class Research, outlines why current risk models lag behind the excitement surrounding fixed income factor investing.

Easy to define, tough to measure

Fixed income factor investing is getting a lot of attention, with publications on the subject appearing regularly. With the increased spotlight there is also more scrutiny and pressure on underlying risk models to live up to more advanced demands. However, traditional fixed income models are not well placed to satisfy these expectations. So what to do?

One could discuss the reasons why they offer limited support (something impossible without resorting to talking about “liquidity” and “data quality”). But this does not change the current state of affairs: in fixed income it is easy to define factors but very difficult to measure factors. How often do we see a chart of the time series of a fixed income “value” style factor return? Or a time series of a portfolio’s exposure to such a factor?

It is relatively straightforward to define investment strategies by a set of rules but more elusive to quantify the drivers of that strategy. For example, a strategy that picks the top 10% bonds ranked by a certain “value” score which could be called the “factor”. Of course, we’re not saying that choosing well behaved, high performing scores and replicating executable strategies is trivial – and it may involve the asset manager’s secret sauce – but, demonstrating the historical performance of such a strategy does not mean the sources of risk and return are well understood. It could be that factors other than the “value” play a large part in explaining the performance or are responsible for risks that weren’t part of the design.

Isn’t this just a copy and paste job from equities?

To understand what drives risk and return we need to be able to measure the underlying factors which capture systematic and specific behaviors of credit returns. Sophisticated techniques exist already to address this problem. These evolved mainly for equities but the fixed income world has found it harder to adapt them. What the industry has learned is that applying an equity cross-sectional risk methodology to fixed income data is certainly not a plug-and-play job. The corporate bond market for example, is particularly notorious at spoiling such attempts.

The crucial ingredients

The underlying challenges that make this such an ordeal have plagued the fixed income industry for decades and it’s not just about blaming it on “data cleansing”. Instead they need to be tackled through a combination of three things: through quantitative modelling (to address noise as well as data consistency of seemingly simple things like classifications), thorough understanding of how practitioners navigate financial markets and finally, through modern cloud technology[1] which allows data to be processed and computed at unprecedented scale and efficiency. Crucially, we set out to leverage not only the methodologies but the actual technology we developed as a leading provider of equity models.

The equities community of systematic investing is used to a suite of acceptance tests to which it submits equity risk models (they include statistical tests, historical backtests, evaluating attribution and validating factor mimicking portfolios). In the past we found that fixed income models struggled to live up to similar expectations. Recently however, things have started to change and our efforts on data are paying off: a new generation of fixed income risk models is emerging that meets acceptance standards which used to be the purview of equities.

Chart: Cumulative factor log returns estimated from a global (IG and HY) universe of corporate issuer spreads to swaps. Decreasing spreads means increasing cumulative portfolio price returns attributable to the factor.  The global market factor shows the three peaks of financial, sovereign and energy crisis. The style factors display a long-term trend.

The truth revealed

Our results confirm the fundamental thesis that factors that earn risk-premia (like e.g. Value, Beta, Momentum) can indeed be extracted from corporate fixed income data if their definition is carefully chosen. A risk system which incorporates these factors can now have a meaningful impact on the investment cycle. To improve risk budgeting, to control exposures in portfolio construction and to help to be able to attribute returns to the desired factors – this is where the added value of a successful risk model lies.

Axioma's cloud-native technology is powered by Microsoft Azure.

Luca Bosatta

Luca is Head of Multi-Asset Class Research and developed Axioma’s next generation fixed-income risk model. Much of his research is dedicated to working with complex data and deriving robust fixed-income curves. This innovative solution opens new avenues into factor-driven FI attribution and portfolio construction. Luca is co-founder of the award-winning UBS Delta portfolio risk management technology and joined Axioma ...