Fixed income attribution in the investment process
PORTFOLIO PERFORMANCE measurement tells us how a fund has done historically and what the level of performance relative to the benchmark has been. It’s the quality control element of the investment process and provides clients with the necessary information to accurately assess the results of that process.
However, performance measurement doesn’t tell you how the returns were generated. Outperformance may have been the fortuitous combination of investments that didn’t form part of a rigorously defined investment process.
If a manager’s performance were random (ie, luck) we wouldn’t expect that performance to be replicated in subsequent years.
Consequently, you’d want to ask if there was a way to decompose a portfolio’s performance in order to isolate the sources of return. In particular, is there a way to analyse the returns of a fixed income portfolio in order to show the manager is losing money from credit spread changes while making money from his duration positioning? There is ‒ and it’s called attribution analysis.
Performance attribution analysis is the exercise of decomposing a portfolio’s performance to determine how the money manager has achieved the calculated returns for a given asset class (for example, bonds) or across asset classes.
Thanks to Gary Brinson and many others, people generally agree on how to calculate equity performance attribution and are comfortable with its interpretation. However, the common debate is whether classic equity attribution models can be adapted for the performance of fixed income portfolios?
That argument is in light of the very different investment processes followed by fixed income and equity teams.
Another area of concern is the performance spread between a bond portfolio and its benchmark, which is generally smaller than that of an equity portfolio and thus a higher level of precision is required for a fixed income attribution model.
But there will never be a “one size fits all” approach to fixed income performance attribution. Ideally, the factors on which a performance attribution model is based should be congruent with the investment process the portfolio manager professes to use.
Thus, instead of looking at performance in terms of externally imposed sectors, performance attribution models ‒ based on the different investment processes ‒ allow you to look at the returns generated by each type of investment decision made by the manager.
One example of such a customised fixed income performance attribution approach is the application of principle component analysis (PCA) on the returns of vanilla fixed coupon bonds and cash. PCA is a statistical technique that extracts the most common types of yield curve shifts over a period.
The three most common types of yield curve shifts are level shifts (that capture the duration view of the portfolio), slope changes and curvature changes.
Subsequent calculations into the returns on the fixed coupon portion of the portfolio are broken down in returns due to coupon payments and pull-to-par, as well as the yield pick-up due to holding corporate bonds and credit spread changes. All other instruments are priced separately and don’t contribute to the factors resultant from the PCA. Their returns are then appended to the return breakdown.
The graph illustrates an example of such an attribution on an active bond fund. That’s an attribution of the outperformance relative to the all-bond index, which was 1,05% over the period.
If we look at the graph we see the bulk of the outperformance came from level shifts in the yield curve. That comes mainly from the modified duration positioning of the fund relative to the benchmark. Modified duration is the measure of change in the portfolio value due to changes in the underlying interest rate.
We also see changes in the shape of the curve added another 11 basis points to outperformance, whereas the big negative impact came from credit spread changes.
An investor can look at that analysis to determine whether the portfolio’s sources of return match the investment process the portfolio manager claims he’s using or if they come from other sources not covered in the investment process.
If the former is true it increases the confidence the investor can have on the quality and potential replication of the alpha delivered by the manager.