LDI and the Efficient Market Hypothesis – my fictional pub chat with John Ralfe

A couple of weeks back I wrote a work piece about asset allocation. John Ralfe (@JohnRalfe1) sent a mini-Tweetstorm response and invited me to have public discussion about it on Twitter. I think that this would be better done over a pint in a pub, but since John doesn’t actually live in London and I wasn’t proposing to get on the train to Nottingham I’ve written this blog.

Given that my initial piece was a work thing (as opposed to a personal blog thing) I’m not going to go into it here, although you can easily read it on my employer’s website. It is basically an outline of how strategic asset allocation processes tend to operate, and active asset allocation can prospectively operate under forthcoming Local Government Pension Authority reform. Real Friday night stuff.

But John’s critique of it is not actually dependent upon the piece’s contents, so don’t worry about reading it to make sense of this blog. I understand John as instead wanting to critique the very notion that defined benefit pension funds should make ever asset allocation decisions.

For folks of you not familiar with John, he has quite a reputation in the world of institutional pension management. As former head of corporate finance for Boots (the biggest pharmacy chain in the UK) in the early noughties he was credited for shifting Boots’ pension assets out of equities and into bonds in an effort to match estimated liabilities with cashflows from fixed income securities (I am hazy on the detail, and you should ask him). This made headlines in not only trade press but mainstream press. As a 26 year old fixed income fund manager I followed the whole thing with great interest. Everyone in the industry did. Importantly, I understood his call to shift assets not as a call on the likely returns from the equity or the bond market, but rather a decision that Boots as a sponsor of a large pension fund should not be making calls on the likely returns from the equity and bond markets – even over very long investment horizons. Instead, they should be taking risks in its area of expertise (pharmacy retailing, and associated businesses).  The development of pensions regulation, invention of the Pensions Regulator and the Pension Protection Fund, the development of accountancy rules (think FRS17, IAS19 etc), and the frankly unreliable and volatile risk markets has led to the mass popularisation of this approach in the UK. I have met many a company finance director expressing the desire to want to get pensions out of the boardroom, and some of my day job involves helping them do so. So while it may not be technically true, I think of John as the founding father of LDI – Liability Driven Investment – in the UK.

LDI is a mammoth thing in the UK – so mammoth that anyone with a remote interest in finance or financial markets without a passing understanding of it should probably read up to familiarise yourselves with it. I draw a horribly crude thumbnail sketch in paragraphs below, but this is a very readable, short non-technical overview. In fact, as I’ve argued here the conditions that have led to its success are sufficiently mammoth a thing that they need to be considered when contemplating non-traditional monetary policy in the UK.

Around about the time that John invented LDI (or at least brought into the mainstream), markets were experiencing what we now refer to as the dotcom crash. This was accompanied by accountancy scandals at Enron, Worldcom, et al that ultimately brought about the dissolution of Arthur Andersen – one of what were the big five global accounting firms. Equity markets were all over the place and the vast bulk of UK defined benefit pension schemes were thrown into massive deficits. Why were they thrown into deficit? To answer this we need to step back and ask of what a defined benefit pension actually consists. The answer to this question has (I believe) changed from a legal perspective over the decades, and has ultimately landed as a form of deferred pay that ranks pari passu with senior unsecured creditors (this off-the-top-of-my-head legal definition is complicated by the creation of the PPF a few years back, and I’m happy to insert a snappy more correct definition of the legal status of someone’s defined benefit pension assets not yet in payment).

Firms that have promised to pay their retired ex-staff monthly payments until death don’t know the exact size of the final bill for which they have put themselves on the hook. They employ actuaries to estimate the size and timing of likely future payments that they have promised to make. Changes in the actuaries’ estimates of likely mortality developments will impact their assessment of payment sizes and timings; financial market volatility will not. Reasonable actuarial estimates of these likely payments can be huge for individual companies. Mega-humongous-huge for the UK in aggregate.

And so it would be weird if accountants didn’t want to represent within the financial statements of a firm some clues pertaining to these often very sizeable liabilities. But rather than reflecting the terminal (estimated) value of liabilities (eg, I owe Deborah £100 fifty years from now if she doesn’t first die), accountants decided that a more sensible approach would be to reflect the estimate of terminal liabilities discounted using market interest rates, so arriving at the present value of liabilities. So instead of showing that I owed Deborah £100, the accounts would show that I owed Deborah around [£100/(1+50yr bond yield)^50] which, when using a yield of 5% equates to me owing Deborah £8.71. This may sound barmy, but if I were to put £8.71 into a fifty year zero coupon bond at a yield of 5% I would get back £100 in fifty year’s time – just in time to pay off my debt to Deborah. So the accountants thought this a sensible approach, and put this way it does sound pretty reasonable.

What has happened in recent years is that bond yields have fallen (with my understanding of structural drivers of bond yields outlined here). Furthermore, bond yield declines have been most substantial during equity market falls (think flight to quality, anticipation of lower interest rates and lower inflation in forthcoming years due to the disruption associated with financial market crisis, etc). So using the example of my debt to Deborah, let’s say that 50yr bond yields fall from 5% to 3%, the present value of the debt I owe Deborah will have risen from £8.71 to £22.81. It is important to note that the amount I owe Deborah hasn’t changed (it is still £100), but the liability reflected in financial statements will have grown almost three-fold due to changes in market rates used to turn terminal values into present values. By discounting terminal values (estimated by actuaries) back to present values, accountants could present stakeholders with a fair picture of firms’ financial health in a manner consistent with the efficient market hypothesis which won Eugene Fama his Nobel prize.

Getting back to pension schemes being thrown into massive deficit by the dotcom crash, we can see that while the terminal value of pension liabilities did not change (ok, they did go up a bit as actuaries have for many years increased their life-expectancy expectations) the present value of liabilities shot up as bond yields fell, and – for funds that chose to invest in equities rather than cashflow-matching bonds – asset values fell. Falling assets and rising liabilities made for a toxic mix, the result of which has been wholesale closure of defined benefit pension schemes, the diversion of large amounts of corporate income into financial assets in order to begin to plug the deficit (especially in an environment which saw the birth of the Pensions Regulator and Pensions Protection Fund in the UK). Here’s a picture of the ongoing death of DB pensions from the Pension Regulator’s excellent annual Purple Book :

Screen Shot 2016-03-10 at 22.29.00

Meanwhile Boots pension funding levels would have been largely unaffected by falls in yields and falls in equity markets from the point at which it did the asset switch. It had largely closed down the mismatch between accountants’ estimates of the present value of actuaries’ estimates of ultimate cashflows payable to defined benefit pensioners by buying long-dated cashflows (in the form of fixed cashflows associated with bond coupons and principal payments). So while the accountants’ estimates of the present value of liabilities soared, so the value of assets also soared as bond prices rose (both being a function of long-dated yield collapse).

This has been a long preamble, and could have been skipped if John and I had met in the pub. Despite it being long I have skipped any discussion of real rates versus nominal rates (more than a little important), and a whole host of other real life (but complicating from a blog perspective) issues/ factors. But we have to move on to John’s Twitterstorm that he has since asked me to answer (a few times).

Here it is in text, with links to the original tweets embedded:

1 “Strategic Asset Allocation”, as correctly applied to an individuals fund, is a category mistake when applied to a DB pension fund @toby_n

2 A DB pension fund is not a self contained economic entity, but simply part of the capital structure of the sponsor @toby_n

3 Asset/liability mis-match in DB pension fund changes capital structure & increases financial risk for sponsor just like gearing @toby_n

4 If a sponsor wants to increase gearing, then directly borrow – transparent & tax efficient – not indirectly thro DB fund @toby_n

5 If a sponsor believes in “SAA” in DB fund, then borrow long term debt & buy financial assets & explain why to shareholders @toby_n

As you can see, it’s an entirely reasonable position from the founding father of LDI in the UK. And John makes very strong points that it would be silly to refute if one accepts his premises (which I have tried to relate in paragraphs above but may have got wrong, and with no John in pub, hard to know). Furthermore, his premises (if these indeed are his premises) are tricky to theoretically refute, rooted as they are in Nobel prize-winning stuff. The issues I have with John’s position (and the whole LDI edifice) can probably be summed up in a rambling question and a couple of observations.

My question:

Given that LDI is a hedge against changes in an accountant’s present value calculation of an actuary’s estimate of terminal cash flows, what is the best way to judge the correct price of this hedge; or rather, is it ever possible to overpay for this hedge, and if so, what tools would I need to employ to understand whether I was overpaying?

I have written about this before in relation to whether house prices (as an effective hedge) can ever be too expensive. I don’t have a definitive answer, but I suspect that hedges can be too expensive and asset prices can be wrong (and have bet my career on this suspicion). But I don’t observe anything in the LDI approach to help frame this judgement. Instead I understand it placing what looks to be an intolerable burden on the efficient market hypothesis to determine that a hedge must always be appropriately priced. In other words, this approach lends itself to the conclusion that it is impossible to overpay if you are paying the market price, no matter what the market price.

This leads on to my two observations:

  1. The size of UK defined benefit liabilities is larger than the market of physical hedging instruments on a duration-weighted basis.* Simply put, while it might be theoretically possible for  DB funds to collectively go down the LDI route, (an practically possible for some of them to go down the LDI route), it is not practically possible for them to collectively so do. This will be of surprise to no one working in UK fixed income markets, but surely throws doubt on the idea that pension funds seeking to hedge liabilities can do so without distorting the price of that hedge (seeming to undermine a central tenet behind the use of the efficient market hypothesis by accountants and actuaries).
  1. While fixed rate assets can be used to match accountants’ estimates of actuaries’ estimated liabilities, this is true because of the tautological construct. And as such while I accept that the LDI approach will be the very best possible hedge against my accountants’ estimate of my actuaries’ estimates of my liabilities, I am not convinced that the reality of the tautology should define the investment structure of pension fund assets in the UK (especially given observation 1 above, and my unwillingness to believe that it is impossible to overpay for a hedge). Fixed rate corporate bonds, equities, and commercial property are all just claims on corporate profits. Despite all being but claims on future profitability/ future economic activity, only one of these claims has any place in a full LDI construct. That just seems odd to me.

If I was having my pint with John rather than answering his Tweetstorm in this overly-long blog, I would expect he would parry by repeating point 2-4 (which are super-strong points). And I would probably repeat my question and observations. And I genuinely don’t know what would happen after that. I’m fairly sure that neither of us would change the other’s mind. But I also hope that we would emerge the wiser about the other’s position (and perhaps even with enhanced respect for each other’s position). Hence the interest in the pint.



* I should share my extremely back-of-envelope workings so that they can be properly challenged.

The Pension Regulator’s annual Purple Book indicates that UK private sector DB pensions have combined liabilities on a buy-out basis of £2.1 trillion. They also say that a 10bps fall in Gilt yields increases these liabilities by 2%, inferring a duration of around 20yrs.

Let’s look at how much UK fixed income duration exists that might be used to match these liabilities. (Yes, I know that most LDI is done synthetically, but can’t get away form the idea that the duration has to reside somewhere. This may be my fatal gap in understanding.)

First, let’s take the entire stock of UK fixed income securities (as represented in the BAML Broad Sterling Market UK00 index). This has a face value of £1.47 trillion, but a market value of £1.78tn and a duration of 10.2yrs. Leaving aside curves in our hunger for UK duration *of any description* let’s call this £908 billion of 20yr equivalent duration paper.

And let’s add on to this the entire stock of inflation-linked Gilts with a market value of £540bn and a duration of 12.4yrs and call this another £335bn of 20yr equivalent duration paper. Summing these together we get to £1.24 trillion of 20yr duration equivalent paper.

What about the missing £857bn of 20yr duration securities that needs to be bought so that DB pension funds can cashflow match? It doesn’t look like it exists.

John Ralfe has written a reply here that you should have a read of. 


3 thoughts on “LDI and the Efficient Market Hypothesis – my fictional pub chat with John Ralfe

  1. If I read this right on my phone you say that the size of defined benefit pension liabilities (undiscounted) in the uk is bigger than the entire uk gilts market?

    The effect of some companies putting so much more into their pension schemes, even spread over 12-20 years, must have a chilling effect on wage growth and investment at those companies

  2. Pingback: Reply to Toby Nangle | johnralfesite

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s