Friday, December 14, 2012

Gretchen Tai: possibly the best fund manager you have never heard of


Please note: much to my surprise these returns are entirely genuine. I was originally skeptical. I am now convinced.


A little while ago I worked through HP's defined benefit pension fund - trying to work out how underfunded it is (and hence work out the true debt levels at Hewlett Packard). Most defined benefit funds are underfunded. They assume 8% returns in their actuarial assumptions and are two thirds are invested in debt. (It is of course impossible to get long term returns in debt substantially above the starting yield.)

Whilst I was looking for funded status what caught my eye were the very fine returns they had on assets - particularly in the US defined benefit fund - and particularly over the past three years. I can't quite work out how they do it.

I tried to compare the fund to other large endowments (Harvard) and also very large similar pension funds (Ford). This comparison is made harder because Harvard and Yale have 30 June balance dates, Ford has a year-end balance date and Hewlett Packard has an end of October balance date. This difference in balance date is particularly important for the 2009 year which includes more of the crisis and less of the bounce for October balancing funds than December balancing funds.

I would love to go further back - when doing this analysis - but it is a little hard. The size of the asset pool roughly doubled in August 2008 when Hewlett Packard acquired EDS and combined the asset-pool for their defined benefit funds.

This post goes through the various HP Form 10Ks and various press articles quoting the Chief Investment Officer (Gretchen Tai) and tries to piece together how they do it.

Returns to October 2009

Here is a table outlining returns of the three main pools in the year to October 2009 (source 2009 form 10K):



This disclosure reveals the US Defined Benefit Plan made $1509 million in profit on starting fair value in assets of $7313 million . That is 20.63 percent. Given the fund paid out $506 million in benefits and settlements the returns for that year were - money weighted - probably nearer to 21 percent.

From October to early March the year was not fun either. Then the market bounced with a ferocity that only happens when the consensus is almost entirely hopeless.

There is a limited disclosure on how the money was invested.



This is a fairly conventional 60-40 split - with 40 percent in equities and 60 percent in public debt securities. There was a small amount of cash. The equities included 10.9 percent allocated to private equity funds.

These numbers are exceptionally good. Here are some comparables:


2009
S&P 500 Total Return (incl. Dividends)9.8%
MSCI Global ex US30.2%
Barcap iShares 20+ Year Treasury Bonds3.2%
Barcap iShares 3-7 Year Treasury Bonds2.1%
Barcap iShares Short Treasury Bonds-0.1%
Barcap Intermediate Investment Grade16.0%
SPDR Barclays High Yield Bonds21.2%
Ishares iBoxx Investment Grade Corporate Bonds20.6%

If you are going to get returns of 20 percent out of a diversified 60 percent bonds, 40 percent equities portfolio in this year it is going to have to be roughly one half in non-US equities (thus benefiting from that 30 plus percent return in non-US equities). The bonds are going to have be exclusively in beaten up junk bonds and beaten up corporates.

The returns however are not impossible. There were a few indices that performed even better. For instance emerging market debt returned just shy of 40 percent in that twelve month period and the MSCI emerging market equity index was up 64 percent. But those things were badly beaten up in the crisis.

A 21 percent return in the year to October 2009 on a huge sum required the calm rationality to buy the really cheap stuff when everyone is panicking. It speaks to Gretchen Tai's calm excellence.

Returns to October 2010

The 2010 returns are not as spectacular (you can't get 21 percent every year) but they are extremely impressive. The key disclosure as to the returns (from the relevant 10K) is here.


The returns were $1224 million on starting capital of $8371 million. That is a more modest 14.6 percent.

The book again targeted a 60 percent bonds, 40 percent equity mix. They had just shy of 40 percent equities but they also held roughly 5 percent in cash at year end.


For the first time there is disclosure on the private equity securities they hold. The fund contains "level 3" assets. It worried me that the good returns were from mark-to-myth accounting - but it seems not. During that year they booked zero gains on the level 3 assets. Indeed there was a detailed disclosure.


Indeed given that we know how many level 3 assets were held its useful to work out the returns on non-level 3 assets.

There were $7440 million in non-level 3 assets at the beginning of the year (8371-931). All the returns - $1224 million - were on these. That is a 16.5 percent return on the non-level 3 assets. As there were net withdrawals from the fund over the year the money-weighted returns were even higher.

Again this is darn impressive. We can compare to the usual indices:



2010
S&P 500 Total Return (incl. Dividends)16.5%
MSCI Global ex US9.8%
Barcap iShares 20+ Year Treasury Bonds4.8%
Barcap iShares 3-7 Year Treasury Bonds5.6%
Barcap iShares Short Treasury Bonds0.0%
Barcap Intermediate Investment Grade5.0%
SPDR Barclays High Yield Bonds7.6%
Ishares iBoxx Investment Grade Corporate Bonds6.3%



Against this 16.5 percent is very impressive. There must have been some mighty nice picking. Again a big overweight position in emerging markets would help. Emerging market debt put on another 17.7 percent and the MSCI emerging market index put on almost 24 percent. Whist the Barclays high yield index was not impressive (7.6 percent) some bond indices did better (there were very high returns for some long dated inflation adjusted bonds for instance).

As a rule persistence in ownership of the things most beaten up in the crisis (emerging markets, deeply discounted junk bonds) was one way of getting these returns.

The 2010 form 10K gives - for the first time - a more detailed breakdown of assets.



The portfolio is more balanced than the returns suggest. They do own some non-US market equities but they do not disclose what proportion are "emerging market" equities. They also own a lot of corporate bonds and do not disclose what proportion are emerging market bonds. Whatever - on this sort of portfolio the returns are very fine indeed.

The 2011 returns

The 2011 returns (whilst lower still) were also impressive. The key disclosure is in the 2011 form 10K:



The returns were $1389 million on $9427 million or 14.7 percent.

However this year some of the returns came from Level 3 assets. We have the following disclosure:



In this we see that 282 million of gains came from Level 3 assets - 155 of that from assets actually disposed of (so they were gains received in cash). The starting level 3 assets were 1040 million.

So net of level 3 assets the fund gained $1107 million (1389 - 282) on $8387 million (9427-1040) of assets. That is a return of 13.2 percent.

Again this compares exceptionally favourably to the relevant indices.


2011
S&P 500 Total Return (incl. Dividends)8.1%
MSCI Global ex US-7.3%
Barcap iShares 20+ Year Treasury Bonds15.4%
Barcap iShares 3-7 Year Treasury Bonds2.2%
Barcap iShares Short Treasury Bonds0.0%
Barcap Intermediate Investment Grade-0.5%
iShares 10+ Year Investment Grade9.4%
SPDR Barclays High Yield Bonds-4.7%
PIMCO Investment Grade Corporate Bonds2.2%
Ishares iBoxx Investment Grade Corporate Bonds2.1%




To get these sort of numbers your portfolio needs to be full of very long-dated bonds. This was after all the year that the long-end yield collapsed. Moreover the best returns from the previous two years were in emerging markets. They were not good in the year to October 2011. The bonds scored 4 percent but the equities were -7.

The asset mix (also disclosed in the 10K) was not much changed.



There is so much alpha generated in these three years I would create insane jealousy amongst my readers if I were to calculate it. But if a masochist wants to work it out (and hence make themselves feel inadequate) go right ahead.

Prior evidence of genius

These returns are startlingly good. And it is not the first time - indeed it is the continuation of a trend. The 2007 form 10K reveals the genius in full flight...

In the beginning of fiscal 2008 (meaning November 2007) the HP fund moved almost entirely to bonds as per this (contemporaneous) quote:
In the beginning of fiscal 2008, we implemented a liability-driven investment strategy for the U.S. defined benefit pension plan, which will be frozen by December 31, 2007 and is currently overfunded. As part of the strategy, we have transitioned our equity allocation to predominantly fixed income assets. The expected return on the plan assets, used in calculating the net benefit cost, has been reduced from 8.3% to 6.3% for fiscal 2008 to reflect the changes in our asset allocation policy. Our medical cost trend assumptions are developed based on historical cost data, the near-term outlook and an assessment of likely long-term trends. Actual results that differ from our assumptions are accumulated and are amortized generally over the estimated future working life of the plan participants.

Moving your equity portfolio almost entirely to bonds at the peak of the market is inspired. Gretchen Tai in the press downplays the genius:

"To be honest, we weren’t timing the market. HP had decided to freeze the U.S. DB plan, which at the time was 106% funded and, as a result, we decided with our senior management that an immunization strategy was the right strategy going forward even with pension expense considerations; so, in November 2007, we decided to move out of public equities and into fixed-income and hedged out interest-rate exposure 100%. Although it wasn’t our main consideration, timing of the strategy was fortunate: We were not only able to maintain our funding status through the financial crisis, but also it improved significantly due to strong returns from fixed-income portfolios as well as the interest-rate hedges.

This is the sort of out-there-genius moves that I would not even believe unless it was stated contemporaneously and backed by great numbers. But it was stated contemporaneously. [The 2008 returns were not so brilliant though - but they included the EDS plan that was acquired just before the peak of the crisis. Gretchen Tai can't be held responsible for that.]

Chief Investment Officer magazine clearly knows the important talent they had when they interviewed Ms Tai. She warrants a flattering cartoon of her dressed as some kind of post-modern superhero leader:



How does Gretchen Tai do it?

I can't find anywhere a detailed disclosure of the assets held by the Hewlett Packard defined benefit plan. However Hewlett Packard has moved most of the management of the defined contribution plan in house. This is mentioned in the above mentioned press article where she compares the returns of defined benefit plans (mostly good) to defined contribution plans (mostly poor) and says that the difference is professional management. To quote:
We also manage the company’s DC plan. HP again is being really progressive here. Marketwide, defined benefit plan returns are so much better than defined contribution returns in historical performance—mostly because you have professionals looking after DB assets, and not looking after DC plans. With $15 billion in U.S. 401(k) assets, however, we really want a fresh eye on the plan—we want it to be world class. To do this, we’ve tried to eliminate mutual funds from the investment options, to lower fees. You won’t see brand name mutual funds in our plans—we want participants to focus much more on asset allocation decisions than picking mutual fund managers. Because our plan is so large, we often can use a custom fund-of-funds approach for each offering—which allows our participants to get the benefits of diversification without doing a lot of work."
There are clearly a bunch of fund managers Gretchen outsources to - but not "brand name mutual funds". Hewlett Packard also bought the defined contribution plans of EDS in house (which saved money). To quote another press article:

HP officials spent the time in between not only analyzing the elements of both DC plans but also developing a strategy to reduce costs, consolidate record keepers and reduce redundant options. 
HP has achieved “significant savings” by consolidating providers and using the combined plan's size to negotiate lower fees, said Ms. Tai, who declined to quantify the savings. For some investment options, Ms. Tai said HP uses the same managers for its DC plan as its defined benefit plan, but she declined to identify the managers or the assets they manage.
It is a pity she "declined to identify the managers or the assets they manage". I am sure many of my readers would love to know how to find asset managers like that.

However the defined contribution plan publishes a form 11-K - which details the portfolio and the returns on it. Unfortunately that has a December balance date and so the numbers are not strictly comparable.

That defined benefit plan - for the year ended December 2011 - had almost $14 billion in assets available for distribution:



 


But the returns were negative as per the following table:



We have $558 million of negative returns on $14195 million of starting assets. That is roughly minus 3.9 percent.

There are a few obvious differences. The defined contribution fund contained $561 million in Hewlett Packard shares which were a very poor performer. They moved from roughly $42 to roughly $27. That alone would account for roughly half the loss - but can't account for anything like the difference in performance.

The difference in balance date clearly accounts for much more. For instance the S&P returns for the year until October 2011 were 8.1 percent. They were only 2.1 percent in the year until December. The 20 year bond was 15.4 percent in the year until October. It was 28.8 percent in the year until December.

Lower equity returns clearly matter as the defined contribution fund was carrying 4 billion in direct equity investments (a very long list indeed) and another few billion in indirect equities (mostly index funds).

The higher long-bond returns clearly don't matter much as the fund is strangely bereft of long bonds - carrying huge amounts of mortgaged backed paper (mostly Fannies, Freddies, Ginnies etc).

In other words the defined contribution fund was not as well positioned as the defined benefit funds and did not perform as well, and I can't use the defined contribution fund to solve the question of how Gretchen Tai does it.

The Ford fund as a comparison

Here are the Ford returns for calendar 2009 (from the 2009 year form 10K):


Pension Benefits
U.S. Plans
Non-U.S. Plans
Worldwide OPEB
2009
2008
2009
2008
2009
2008


















Change in Plan Assets (a)
Fair value of plan assets at January 1
$37,381$45,696$14,707$21,396$2,786$3,875
Actual return on plan assets
4,855(4,480)1,692(2,036)792(1,011)
Company contributions
1361389681,209
Plan participant contributions
272580101
Benefits paid
(3,908)(3,960)(1,456)(1,380)(62)(77)
Settlements
(1)(58)(3,517)
Foreign exchange translation
1,581(4,510)
Divestiture
(3)
Other
(34)(38)(7)(12)1(1)
Fair value of plan assets at December 31



In that year Ford produced $4855 million on starting assets of $37381 million or 13.1 percent. That is a long way below the 20.6 percent produced by Hewlett Packard in their 2009 year.

Moreover Ford had the advantage this year of a December year end (and thus includes less of the crisis and more of the recovery in their numbers). This advantage is offset by the very weak long bond at the end of calendar 2009 as the money-printing inflation fear was peaking.

Ford also gives us an asset-mix disclosure for the US fund. It is actually fairly similar to HP (but carries more detail):


U.S. Plans
2009
Level 1
Level 2
Level 3
Total
Asset Category
Equity
U.S. companies
$8,675$26$15$8,716
International companies
8,41348928,553
Commingled funds
3863389
Derivative financial instruments (a)
(1)(1)
Total equity
17,08746011017,657
Fixed Income
U.S. government
2,3402,340
Government-sponsored enterprises (b)
1,31071,317
Government – non-U.S.
449256705
Corporate bonds (c)
Investment grade
8,403858,488
High yield
1,152151,167
Other credit
332154
Mortgage-backed and other asset-backed
1,4882781,766
Commingled funds
338338
Derivative financial instruments (a)
(8)(149)(42)(199)
Total fixed income
2,33213,02462015,976
Alternatives
Private equity (d)
1,0051,005
Hedge funds (e)
1,9861,986
Real estate (f)
11
Total alternatives
2,9922,992
Cash and cash equivalents (g)
71,8641,871
Other (h)
(62)26(3)(39)
Total assets at fair value
$19,364$15,374$3,719$38,457
_______



That is about 45 percent equities - slightly more than HP - which should be favourable to returns because equities were the better performing asset class.

Here are the returns for the 2009 calendar year of various indices to compare:


2009
S&P 500 Total Return (incl. Dividends)26.5%
MSCI Global ex US37.4%
Barcap iShares 20+ Year Treasury Bonds-24.7%
Barcap iShares 3-7 Year Treasury Bonds-4.7%
Barcap iShares Short Treasury Bonds-0.2%
Barcap Intermediate Investment Grade7.3%
SPDR Barclays High Yield Bonds19.8%
Ishares iBoxx Investment Grade Corporate Bonds2.5%


Equities gave 30ish percent (depending on the percentage that was foreign), junk bonds about 20 percent, corporates about 7 percent and longer treasuries were sharply negative the Ford return of 13 percent seems about right. Nothing special - but certainly not bad.

Ford's 2010 year looks considerably better than benchmark at Ford:


 
Pension Benefits
U.S. Plans
Non-U.S. Plans
Worldwide OPEB
2010
2009
2010
2009
2010
2009


















Change in Plan Assets (a)
Fair value of plan assets at January 1
$38,457$37,381$17,556$14,702$$2,786
Actual return on plan assets
5,1154,8551,4871,695792
Company contributions
1351361,236962
Plan participant contributions
23274780
Benefits paid
(3,704)(3,908)(1,281)(1,456)(62)
Settlements
(1)(3,517)
Foreign exchange translation
(356)1,581
Divestiture
(66)
Other
(66)(34)(8)(7)1
Fair value of plan assets at December 31
$39,960$38,457$18,615$17,556$$


Starting assets were $38457 million and actual returns were $5,115 million. That is 13.3 percent.

In this case some of the returns are attributed to gains on Level 3 assets. There is a table of these which shows that at the beginning of 2010 there were $3719 in level 3 assets. The gains on these were 522 of gains on them . Net of this the portfolio produced $4593 million (5115-522) in gains on $34738 million (38457-3719) in assets. That is still a respectable 13.2 percent.

My usual indices show that this is indeed a very fine return:

S&P 500 Total Return (incl. Dividends)15.1%
MSCI Global ex US8.4%
Barcap iShares 20+ Year Treasury Bonds4.7%
Barcap iShares 3-7 Year Treasury Bonds4.1%
Barcap iShares Short Treasury Bonds0.0%
Barcap Intermediate Investment Grade2.4%
iShares 10+ Year Investment Grade4.9%
SPDR Barclays High Yield Bonds2.3%
Ishares iBoxx Investment Grade Corporate Bonds4.1%



However you could still get 20 plus percent returns out of emerging markets during that time.

The 2011 returns for Ford are more modest. The usual table from the 10K shows $2887 million in gains on $39960 in starting assets for the US fund - a fairly modest 7.2 percent.

Here are the index returns for this year:


2011
S&P 500 Total Return (incl. Dividends)2.1%
MSCI Global ex US-16.1%
Barcap iShares 20+ Year Treasury Bonds28.8%
Barcap iShares 3-7 Year Treasury Bonds6.4%
Barcap iShares Short Treasury Bonds0.0%
Barcap Intermediate Investment Grade1.9%
iShares 10+ Year Investment Grade11.4%
SPDR Barclays High Yield Bonds-3.2%
PIMCO Investment Grade Corporate Bonds1.9%
Ishares iBoxx Investment Grade Corporate Bonds4.9%



If you avoided global equities and were long some duration you did well that year. Pension funds are naturally long duration. That said Ford's performance was within the expected range (maybe light if you expected them to hold duration).

Summary

Gretchen Tai and Hewlett Packard perform much better than you would expect given the diversity of their portfolio. Gretchen Tai really is one of the great undiscovered asset managers.

I am looking forward to the filing of another Hewlett Packard form 10K (due within days) to find out what Ms Tai has been up to this year.





John

18 comments:

Simon Forrest aka sinoforestanalyst said...

So what you're really saying is: if it looks, walks, and quacks like a duck it's a... _____.

John Hempton said...

No. HP really did contemporaneously say they were getting out of equities near the peak of the market.

Brilliance I say. Brilliance.

Nemo Incognito said...

The end of this is going to be like a scoobie doo episode:

It's Ray Dalio in a latex mask!

Diogenes said...

I never did wholly subscribe to the view that if something looks to good to be true, it usually is. Though it would be nice to have a clearer account of the assets. Perhaps someone has made off with the information.

Huss said...

John, just to clarify, the index comp return tables you are using for HP are Oct-Oct, correct?

nic said...

Nic barnes (I help trustees manage a UK db and dc plan)
It is red wine and risotto night no might have read this inaccurately but:
Is there confusion about absolute and relative returns? Managing a db scheme is about managing versus the liabilities of the scheme. If hp have hedged the rate risk (and in the us db plans are sensitive to nominal not real rates I.e inflation) the absolute returns will have been highly positive. To see if this has worked well it is important to look at the risk relative to the liabilities.
On the dc side it is about absolute returns as plan sponsors typically look to hedge members' forward annuities rates.
So comparing the relative return of the db and the absolute return of the dc might be a good thing.
Was anyone else concerned / bemused that hp equity features so prominently in the dc plan? Stuff diversification......

Anonymous said...

Allocation decision was likely HP, but they certainly weren't hiring unknowns or managing much internally.

Their ERISA 5500 filings for 2011 indicate they paid health investment advisory fees to some usual suspects -- PIMCO, Blackrock, Lord Abbett, Goldman, Fidelity. There are a few other less well-known managers (Axiom, NISA, Adams Street).

2009 had a much longer list so it looks like they dropped a few known names after 2009 (e.g., Western Asset, Hotchkis and Wiley, AQR and Dodge and Cox were all terminated).

401k offerings are definitely diverse though.

dearieme said...

Something very similar was done by Boots, the British chain of pharmacies, someyears ago - its defined benefit pension fund got out of equities into bonds to offset its liabilities, and did very well as a consequence.

Eventually the finance bloke who pulled off the coup left the company - I believe the pension fund went back into equities later.


Anyway, it's a natural comparison for you.

Anonymous said...

Funny, I've met some ex-HP investment folks (though not Gretchen) and was somewhat underwhelmed to say the least.

Anonymous said...

Gerchen Tai only took over the responsibility in Sep 2010, so isn't the credit (?) for the previous performance should be to Kenneth Frier who is now managing Sanford Endowment? Maybe he put in a system that allows these super normal returns?

bipolarboyboston said...

John, I think you gave her too much credit.

1) Gretchen reports to the financial/investment/benefits committee (CFO/Chief Risk /Audit/etc officers). These committee are the folks who responsible for the plan and she will not be able to implement strategies without their approval. Credit may be due to these folks as well.

2) FAS 158 came out in 2006? Which took out the smoothing of db assets on the balance sheet. So many companies start to implement LDI in 2006/2007/2008 and timed the market perfectly. Probably thank their investment consultants.

3) Because they implemented LDI, there performance was largely driven by long duration. They were around 100% funded in Nov 2007 and assuming plan/liability assets of 4000mil a piece and a 12 yr duration on liabilities. There net duration exposure should match 1:1 and since they didn't move 100% into fixed income they must of used derivatives (hence the presence of NISA on there manager list). 60% fixed income investment in the Lehman Brothers Aggregate at 5yr duration isn't that much duration exposure so they probably used leverage treasury futures so thank flight to quality, positive convexity, and 5% 10yr treasury moving down to 1.70% today.

So her performance is largely due to a very well timed move to long duration.

Jake said...

Not sure where to begin here. For one, corporate pension funded status is calculated by assets less liabilities discounted at a corporate bond yield (with a duration that matches the liabilities), not expected return like public plans.

Second.... did you even compare HP's performance relative to corporate plan's as a whole? The performance you outline isn't surprising as MOST corporate plans switched to long duration bonds from equity following 2007's PPA (pension protection act) which required corporate pension plans to put their funded status on their financial statements (i.e. it became an accounting issue). Hence your "remarkable" timing mentioned for HP, was pretty standard in the industry.

Sorry, but this post was extraordinarily sloppy from someone whom I thought got into the weeds.

Anonymous said...

She's daytrading S&P emini.

Mr. S. said...

Is the option selling strategy on top of their equities and even selling insurance something that can explain the extra return, at least partially? I put a comment here last night on this and it disappeared. Weird.

NatScientist said...

It's always cheery to note, there's a magic someone who we can have Faith in who will provide for us as a result of their solitary and generous suffering: a nice bedtime story for the Winter Solstice while start our seeds for Spring.

Anonymous said...

I don't think Gretchen was in charge in 2007. She was there but not the chief.

C.S. said...

Wow I could not agree more with the comments from Jake. You clearly do not know much about the U.S. pension market, and your smug attitude makes you look like the fool here. Moving to an LDI strategy after 2006 was very common for corporate plans, and since most plans estimate their liabilities as having effective duration of 12+ years, they way they implement it is frequently to buy lots of long duration treasuries or the equivalent. Obviously they have benefited from that decision recently and will pay the price when yields rise.

Anonymous said...

all this hard work and then they peeed it up the wall...

or all this hard work so they peeed it up the wall

General disclaimer

The content contained in this blog represents the opinions of Mr. Hempton. You should assume Mr. Hempton and his affiliates have positions in the securities discussed in this blog, and such beneficial ownership can create a conflict of interest regarding the objectivity of this blog. Statements in the blog are not guarantees of future performance and are subject to certain risks, uncertainties and other factors. Certain information in this blog concerning economic trends and performance is based on or derived from information provided by third-party sources. Mr. Hempton does not guarantee the accuracy of such information and has not independently verified the accuracy or completeness of such information or the assumptions on which such information is based. Such information may change after it is posted and Mr. Hempton is not obligated to, and may not, update it. The commentary in this blog in no way constitutes a solicitation of business, an offer of a security or a solicitation to purchase a security, or investment advice. In fact, it should not be relied upon in making investment decisions, ever. It is intended solely for the entertainment of the reader, and the author. In particular this blog is not directed for investment purposes at US Persons.