Credit returns: worth waiting for?
How does the rate of return you can expect to earn on corporate bonds depend on the time horizon?
This question matters for how corporate bond portfolios, including those used by defined-benefit pension schemes in their endgames, should be structured. To investigate, we need to delve into the world of “term premia”…
What’s a “term premium”?
A term premium is the difference between what you get for locking up your money in “risk-free” instruments (usually developed-market sovereign bonds) for an extended period versus what you would get if you simply kept rolling over short-term instruments (or cash) for the same length of time. For example, a gilt’s term premium over 10 years would be the 10-year spot yield today minus the return you would expect for sitting in cash for the next decade instead.
And a credit term premium?
This is trickier and needs more care to define! We’re not interested in the risk-free component: that’s the “regular” term premium above. Rather, we’re interested in the excess returns, i.e. returns over risk free, that you expect to earn as compensation for taking on credit risk. How does investing in long-dated corporate credit compare with rolling shorter-dated credit?
When making this comparison, it’s crucial to control for credit quality because the quality of a “buy-and-hold” strategy is likely to degrade over time. We need to assume the long-dated strategy is rebalanced to maintain the same rating throughout the investment horizon to make a fair comparison.
Accordingly, our definition for a credit term premium is the difference in expected returns of the following two numbers:
- The expected excess return (relative to risk-free bonds) on a corporate bond that matures at the end of the horizon and is rebalanced by rating (so if it is downgraded, then the bond is sold and replaced with a higher-quality bond of identical maturity);
- The expected excess return (relative to cash) on rolling short-dated credit of the same rating, over the same horizon.
What’s a reasonable estimate?
Many academic papers and studies now point to there being no term premium on developed-market government bonds. Is the same true for credit? This is a difficult question! Our approach was to examine the following two components:
- On average, what do credit spread curves by rating look like?
- On average, what are the transitions between credit ratings each year?
Given these two ingredients, we can estimate the annual change in market value of a promised cashflow. There’s upward pressure simply due to the cashflow getting closer (it’s discounted by one fewer year) and typically also from spreads being lower at shorter maturities. These generate what are called “carry” and “roll-down” returns respectively. It’s the roll-down aspect that leads to the possibility of a credit term premium, but we must not neglect transitions!
There’s downward pressure on returns from a tendency towards downgrades and higher spreads. And, crucially, the impact of downgrades is bigger the longer dated the cashflow. For example, if spreads jump by 1% due to a downgrade, this has a circa 5% negative impact on market value for a cashflow due in 5 years but a circa 10% impact for a cashflow due in 10 years.
Quantifying these “pressures” is where a model can help. The graph below shows our results for investment-grade (IG) US credit. These calculations are based on:
- US curves from Thomson Reuters covering the period February 2007 to June 2021
- Typical splits of IG credit by rating over the period
- Long-term transition matrices from Moody’s
Source: LGIM calculations as at 30 June 2021.
As you can see, other than a modest upward slope at low terms, expected excess returns are relatively flat. This suggests that there is no credit term premium, or a modest one at best. Higher spreads at longer terms are offset by greater impacts from downgrades.
There are other reasons to suspect the credit term premium should be zero. Conceptually, it would be strange to allow for a credit term premium whilst having no term premium on rates. We also suspect strong institutional demand for long-dated credit (from insurers and pension funds) could be suppressing long-dated credit spreads. These investors are often incentivised to match asset and liability durations using unlevered, physical assets, leading to demand for long-dated credit.
No easy answers
The results above are based on just one model and set of assumptions. Whilst I’ve sensitivity tested it to a reasonable extent, considerable uncertainty remains. It also only represents a “strategic” position: whilst one shouldn’t expect a credit term premium in general, unusual market conditions could lead to a good reason to deviate from this neutral stance.
Nevertheless, our analysis suggests that in general, you shouldn’t expect a much higher return on rebalanced longer-dated credit than on rolling shorter-dated credit of the same rating. As I will explain in future posts and a forthcoming article, this – together with an understanding of risk in a liability-driven context – has important implications for cashflow-driven investing used in defined-benefit schemes.
So just to summarise the risk aspect briefly, this research suggests that rolling short-dated credit incurs greater reinvestment risk but its returns are less sensitive to downgrades. It can also act as a powerful diversifier to the returns on long-dated credit over long horizons: when spreads blow out, long-dated credit is hit the hardest, but this often provides an attractive reinvestment opportunity for short-dated credit. The upshot is that whilst we believe that traditional cashflow matching remains a key component for schemes looking to lock down risk, some degree of diversification into short-dated credit could make sense.
Watch this space for more!
 Note that a buy-and-maintain strategy differs from a buy-and-hold strategy by actively seeking to avoid downgrades and default such that the effects of degradation are reduced.
 See ‘The Term Premium Conundrum,’ Neuberger Berman, March 2019
 The act of rebalancing crystallises these losses in the form of lower promised cashflows.
 We choose to focus on the US for greater consistency with transition matrices and because it is a large market, so curves are likely to be more reliable.
 The key result – that there appears to be no credit term premium – is not sensitive to our exact assumptions on the split of IG credit by rating. Purely for illustration we assumed a 10%/30%/30%/30% split across AAA/AA/A/BBB.