LDI - it’s much like the harness for your Sheepadoodle (or child); you need to mind the fit!
Why do assets appear to have performed differently from the liabilities they have been designed to protect?
As I put my 27kg (and rapidly growing) six-month old Sheepadoodle, Wilma (pictured below), in her harness the other day and clipped on her seatbelt, it struck me that the harness, much like an LDI strategy, is designed to keep Wilma (the asset) on the seat (the liabilities), particularly if the car came to a sudden stop (market movement). The looser the fit (or match), the more likely she is to move around.
Using assets to match the liabilities is much the same. The more daylight there is between the assets and liabilities, the more scope there is to see divergence. There are three main sources:
- Constituents of the matching portfolio – schemes sometimes invest in corporate bonds but these can behave differently from government bonds, which are often the reference point for liability valuations
- Lack of availability of long-dated assets – schemes will typically have pension payments beyond 50 years, but there aren’t many assets that provide cashflows beyond 50 years
- The liability discount basis – if matching is done using government bonds but the liabilities are valued using a higher discount rate (e.g. government bond rates + 1.5%), then the liabilities will be expected to grow faster than assets
Let’s expand each of these in turn:
1. Constituents of the matching portfolio
It is common to use corporate bonds as part of the asset portfolio as they contribute to liability matching and are expected to yield an additional return over government bonds.
However, liabilities are typically valued with reference to government bonds. So to the extent that corporate bonds perform differently from government bonds (and they often do), this can contribute to assets and liabilities performing differently.
The following chart compares two returns over a three-year period: the performance of corporate bonds, versus a liability valued using government bond rates plus the extra yield (spread) for corporate bonds as at March 2016. All else being equal, if the spread doesn’t move, both returns are expected to be the same. However, as is often the case, spreads did indeed move and so the corporate bond market generated different returns to the “expected” credit spread relative to government bonds.
2. Lack of availability of long-dated assets to match the liabilities
Liability cashflows (i.e. pension payments) are paid until the last recipient dies, so these can stretch for over 50 years. However, there aren’t many assets of an equivalent maturity that can be bought to match such long-dated liabilities.
In LDI, rather than considering the actual cashflows, we look at the sensitivities of the value of those cashflows to a small (0.01%) change in interest rates. For pragmatic purposes, cashflows are grouped into five yearly “buckets”. The exception is the 50-year bucket, which also contains all the liability cashflows beyond 50 years. An example of divergence over time is shown below.
If we roll forward three years, what was a 50-year bond in 2016 is now (in 2019) a 47-year bond. Although from a cashflow perspective the situation is relatively unchanged, from a sensitivity point of view some of the asset sensitivities are now in the 45-year bucket. But for the liabilities, it is still the case that for the payments due beyond 50 years, the sensitivities are still mostly in the 50-year bucket. In other words, over time, the sensitivities of both the assets and the liabilities shorten, but the liability sensitivities shorten by less. This mismatch is illustrated in the chart below.
Source: LGIM as at 30th September 2019. For illustration only
3. The liability basis: different discount rate effect
A scheme may value the liabilities on a basis which assumes that the scheme assets will earn a higher return in the future.
However, if matching is done using only government bonds, then even though the match is aligned on day one, the value of the liabilities will grow over time by more than the value of the government bonds being used to match them. For example, let’s start with liabilities that are worth £1,000 today (valued using government bonds plus 1.5%). They have exactly the same interest rate sensitivity as £986 of government bonds, so from a sensitivity perspective the match is aligned.
We would expect the liabilities to grow by government bonds plus 1.5%, and the assets to grow by the government bond return (in this example, 0.5%). This means that if the value of the liabilities grows (by government bonds +1.5%) to, say, £1,020, then assets will have only grown by 0.5%, from £986 to £991. As the liabilities have grown by more than the assets, the deficit has increased.
It is also worth noting that the sensitivity of the assets and liabilities to changes in interest rates will now no longer be aligned.
Conclusion – how snug is your harness?
So, next time your liabilities and matching assets move differently, the key questions are:
- Are the matching assets different from those referenced to value the liabilities?
- Are the liabilities longer dated than the assets held?
- Are the liabilities being valued using a higher discount rate?
These can all lead to drift over time, which is why many schemes delegate the responsibility of monitoring and revising the strategy to their investment manager, rather than “set and forget”.
However, it is worth remembering that, just like Wilma’s harness, both your matching portfolio and your liabilities are probably pragmatic approximations, with room for movement.