Investing for the endgame at low spreads

In our latest research, we find some merit in underweighting credit exposure when spreads are low. However, we also find that relying on strong overweights at high spreads can harm efficiency. Despite a strong relationship between higher spreads and higher excess returns, we believe it’s challenging to substantially outperform a static policy on a risk-adjusted basis.

Credit spreads have historically been strongly mean reverting in nature, as you can see from the chart below:

Is this something pension schemes can seek to exploit when investing for the endgame?

To investigate, we compared different allocation ‘policies’. These specify how much to allocate between investment grade (IG) credit exposure and LDI as credit spreads change through time. We considered three policies:

• Static: The investor holds a constant 50% proportion in credit
• Proportional weighting: The investor varies the proportion in credit, increasing the allocation as the level of credit spreads rise, up to a limit of 75% (to ensure enough is left in LDI for hedging)
• Proportional weighting at low spreads: The investor varies the proportion in credit, increasing the allocation as the level of credit spreads rise, but once credit spreads reach an average level of 1.5% or above the percentage allocation stops rising[1]

The policies are compared using back tests and simulations. These results should be treated with a healthy degree of scepticism and caution, but we believe they can potentially offer some valuable insights. The policies are illustrated graphically below. They’ve been calibrated so each has the same long-run expected return.

You may notice a strange feature – the third strategy has less in credit than the first strategy at all spread levels, yet (we claim) has the same expected return. As we shall see, this is because sometimes you can seek to improve expected outcomes by waiting for more attractive spreads.

Back tests

Below are the results from our back tests based on US corporate bond data (given its relatively long history). We believe similar patterns can be read across to the UK market.

The top-performing policy in terms of the Sharpe ratio achieved over 10-year periods was (3) Proportional weighting at low spreads. We experimented with different levels at which to cap spreads in the third policy – it’s impossible to be precise, but around 1.5% appears to be the sweet spot.

The next best was (1) Static, with its constant percentage holding in credit. Finally, (2) Proportional weighting did the worst.

As you can see it held a modest amount in credit in normal times (around 10-15%) but when spreads spiked during the global financial crisis (GFC), ramped up the credit allocation. This is when the policy made the bulk of its excess returns. The back test highlights a danger with such a strategy – the opportunity to invest at high spreads may never arrive. Many of the rolling 10-year periods don’t contain the GFC and so don’t present the critical opportunity the policy was waiting for.

Stochastic calculations

As an alternative to back tests, we can use stochastic modelling. These are the results[2]:

The output is broadly consistent with the back test – the second policy damages efficiency relative to a static approach, whereas the third improves it.

What’s driving these results?

In addition to the issue highlighted whereby it can be dangerous to wait for high spreads that may not arrive, there are a couple of points worth noting:

• Efficient strategies often involve taking an even amount of risk through time. Given risk rises with spreads, this suggests holding less credit when spreads are higher. Such a policy doesn’t work well here[3], but it’s still the case that increasing exposure when risk is highest results in concentrated bets that may not pay off. During the GFC huge waves of defaults were avoided but only narrowly
• Avoiding low spreads has some merit. As with equities, high prices may reflect animal spirits rather than strong fundamentals

Other rebalancing strategies and trading costs

Trading costs matter for the real-life viability of strategies. For the third strategy (the green line), we made some sensible assumptions regarding trading costs[4] and found that using a band-based strategy with an absolute tolerance of around 5% is sensible.

We also checked that allowing for trading costs didn’t impact the ranking (the third policy still won). How the exposure is varied matters, for example portfolio trades of bonds could mitigate costs substantially and/or allow for more frequent rebalances.

What about buy-ins and buyouts?

The dynamics are a bit different to run-on:

• Cheaper insurance when spreads are wider comes with (virtually) no strings attached in terms of risk to members
• Buy-ins and buyouts are irreversible
• Credit helps to hedge moves in buy-in prices

A practical way to think about credit exposure prior to buy-ins and buyouts is that it’s similar to run-on, but the average credit allocation is also driven by a desire to hedge buyout pricing. An appropriately scaled version of the third policy may therefore be sensible.

Other practical points

This is only a rough guide to a complex topic. Wider considerations include:

• The extra governance needed for a more complex strategy. The static portfolio was still a good choice, especially after costs
• How you add or subtract credit beta matters. Credit default swaps (CDS) incur lower costs than physical credit but have a lower credit beta. The use of portfolio trades may offer a cheap way to access physical credit. Trustees might also choose to move towards shorter-dated or higher-rated credit when spreads are low. Buyout-focused investors may consider if their credit is matching-adjustment eligible
• Some schemes may prefer one-way triggers i.e. increase credit exposure if spreads go up, but don’t sell if spreads fall back. In theory, if an asset is too expensive to buy then it should also be considered a good sale. Nonetheless we’re sympathetic to such an approach
• The possible inclusion of higher-growth assets such as equities

Watch this space for a deeper dive into these and other practical aspects.

[1] Note that the average spread used for the purposes of policy 3 uses the average spread level since Jan 2000. The increase in the average spread over this period is consistent with a deterioration observed in average IG credit quality as determined by credit rating breakdown during this period.

[2] Eagle-eyed readers will note we used evergreen credit in these calculations. Most schemes want to invest more credit at long horizons, but can’t due to a lack of long-dated credit with attractive spreads. As a result, they tend to maintain their credit duration even as their liability duration shortens.

[3] You tend to get a better return/risk trade-off in stress periods, at least over short horizons whilst spreads mean revert to average levels.

[4] This assumes a sensible model for trading costs of max(0.25%, spread/6) and monthly review. Note that the c.5% tolerance would be scaled if the 3^rd strategy were scaled.