Peak LDI in 3 years’ time? It could be more like 2033
River and Mercantile Derivatives’ Mark Davies looks at how some minor tweaks to assumptions in the recent Hymans Robertson paper can have a big impact on ‘Peak LDI’ timescales.
I read with interest the recent piece from Hymans Robertson and Nomura on the idea we are entering the age of “peak LDI”
Somewhat controversially, compared to what most trustees may have heard, the paper proposes that the current pace of Liability Driven Investment (LDI) hedge increases can continue for, at most, 3 years.
Using the same calculations as the paper does (PPF Purple book data on asset and liability values as well as asset allocations and the KPMG LDI survey results which give, from the LDI community, the size of hedging adopted) it does indeed appear that LDI hedging levels will hit asset levels in about 3 years’ time. This assumes £100bn of hedging per year, while it would be slightly later (but not significantly so) if hedging levels are half that.
However, the implication in the paper is at this point there will be a drop off in demand for LDI and this is where our views – and our conclusions – therefore differ, for three key reasons.
Firstly, the current level of hedging by UK pension schemes is lower than that used in the paper. The £1.2trn figure quoted is higher than the £0.9trn used in the 2017 KPMG LDI survey, mainly because the research has taken a limited survey from a selection of LDI managers and applied that hedge ratio to all UK pension schemes.
As the peak LDI paper points out, the KPMG survey covers about 1,800 mandates (across all LDI managers) which is only about 30-35% of schemes in the PPF data set.
We think that as a result, the true level of hedging is more like £1trn, and given the liability numbers are so large compared to the market size, that is meaningful. Indeed, that alone could add another 2 years minimum to the peak point.
Secondly, the historical rate of demand for LDI is also not certain and is arguably lower than the £100bn discussed in the paper.
Over the last 3 surveys KPMG have estimated the growth due to new mandates and increases in hedging (rather than simply from market movements), at £32bn in 2015, £103bn in 2016, and £77bn in 2017.
These numbers alone suggest a demand closer to £70bn than £100bn, and indeed under some circumstances – for example, taking the start of LDI hedging as 2005 and using statistics which show total LDI investments of £1trn over the subsequent 11 years – it could be lower still.
With these alternative – and not unrealistic – assumptions, it again adds another two years to the date of peak LDI.
Finally, the analysis potentially overlooks some technicalities which, when taken into account, increase the potential demand for LDI.
Let’s start with assuming “gilt” liabilities are £2.3trn, current hedge levels are £1.15trn (50% of liabilities) and assets are £1.5trn.
Gilt liabilities have already peaked, as the paper mentions, meaning that next year gilt liabilities are projected to fall to £2.2trn. Consequently, £1.15trn of hedging this year is projected to fall to £1.1trn.
As a result, adding £100bn of hedging will only take hedging levels to £1.2trn next year rather than the £1.25trn assumed in the paper (and reached by simply adding £100bn to the current hedge level).
What this tweak to assumptions means is that the effective increase in hedging each year is lower than the £100bn Hymans quotes, and therefore it takes longer to hit the size of the assets. This change adds a further 5 years to the peak LDI date.
One final assumption that changes the picture further still is whether schemes hedge up to the asset value or higher (up to the liability value, for example).
Whether this will happen is the subject for another paper but to take the most extreme example of hedging up to the value of gilt liabilities, this could increase the aggregate demand by approximately £200bn.
Assuming the same pace of demand (c £70bn per year), I estimate this pushes peak LDI out a further 3 years to 2033, 15 years from now.
There are lots of assumptions above, just as there are in the Hymans paper, but one thing looks clear. Peak LDI still seems a long way off, and certainly further out than 2021.