Later today, the board of the $591bn California Public Employees’ Retirement System will vote on a proposal to ditch its traditional asset allocation framework and adopt a newfangled strategy that proponents have dubbed a “Total Portfolio Approach”. It’s potentially a big deal.
The shift is likely to be approved. After all, Stephen Gilmore, the pension fund’s new-ish chief investment officer, had deployed TPA to good effect in his previous role running New Zealand’s sovereign wealth fund before taking up the worst job in finance. And TPA is now a buzzy if fuzzy phenomenon in the normally pretty staid world of asset allocation.
More than a couple of Nobel Prizes (real ones, not the physics, literature and medicine ones) have been awarded to the folks who made up maths that helps you choose a long-term investment benchmark. This benchmark choice — an investor’s so-called “strategic asset allocation” — is generally understood to be the thing that determines pretty much all of your investment returns.
Iterations in institutional asset allocation like the Yale model (that skews benchmarks towards private assets) and the Canadian model (that takes this skew and internalises the management of private assets) have been seismic in their own ways. But they haven’t really challenged the idea that an asset owner should first set a benchmark, and second should set guardrails around how far the allocation can deviate from this benchmark.
The Total Portfolio Approach is different. How different? And is it actually any good?
Back up, give me ‘Strategic Asset Allocation’ 101
Readers who don’t want an ABC of SAAs can skip this section. But for everyone else, here’s our attempt to boil down what you need to know.
The canonical strategic asset allocation process starts with a bunch of capital market assumptions — 10 or maybe even 20 year return forecasts for a range of asset classes. Let’s keep it simple and look only at US stocks and bonds.
Bond forecasts tend to be pretty straightforward; the return of 10-year bonds tends not to deviate too crazily from starting yields. And so with 10-year US Treasuries yielding 4.1 per cent right now, the average forecast return for Treasuries from a bunch of sophisticated financial firms after a lot of modelling is *checks notes* 4.1 per cent.
Stocks are trickier. Pretty much every forecaster has their own methodology, and these will spit out quite varied numbers. But almost invariably, each framework almost always produces a number for stock returns that’s a bit higher than whatever bond number they came up with. The average 10-year return forecast for US stocks from companies we checked with is 5.7 per cent:
OK — so stocks beat bonds in (almost) everyone’s forecast. Why do institutional investors then bother having any bonds at all in their benchmarks?
One reason is that while stocks are great and everything, they are also quite risky. And by risky we mean volatile. US stocks have rocketed up over the decades, but they have had some serious wiggles along the way.
This shouldn’t matter if you’re immortal, don’t need to divest your portfolio anytime soon, and if you’ve got total confidence in your forecasts. But a lot of asset owners often do need to dip into their portfolios to make pesky things like pension payments or to meet insurance claims.
And who, Dunning-Krugerites aside, have total confidence in their forecasts about future investment returns? Japanese institutional investors who trusted in efficient markets and the power of long-term equity outperformance at the start of 1990 had to wait three decades just to get their money back. While this is extreme, it’s not uncommon for stocks to lose maybe a third of their value once a decade. And when they do crater it’s sometimes hard to see a way back up (which is generally why they’ve cratered).
Bonds can crater too, but unless they default they’ll eventually spit all the money you spent on them back at you — and more — in a specified time period. And so, if there are pensions to pay, insurance claims to dish out, or whatever else it is your portfolio exists to fund, your appetite for investment risk may not be unlimited.
Moreover, because bond prices are meaningfully influenced by central banks — who tend to cut rates when the economy (and the stock market) hits the skids — bonds can do well when stocks do poorly. In other words, their returns are poorly (or even negatively) correlated. Which means that when you throw your capital market return and correlation assumptions into a spreadsheet, you get a beautiful bendy line — aka an “efficient frontier” in finance geek language.
Here’s a historic efficient frontier that takes actual rather than forecast returns from 10-year US Treasuries and the S&P 500 over the last 25 years to show you what we mean:
If the past was a perfect guide to the future (which it isn’t) very nervous investors could have built their benchmarks consisting solely of bonds. Hover your mouse over the 0:100 dot and you’ll see that this allocation returned an average of 3.6 per cent per annum and came with a annualised standard deviation of return of 7.5 per cent — a measure of wobbliness.
But as you can see, super-duper nervy investors could actually allocate a fifth of their portfolio to stocks, and enjoy both lower volatility and higher returns, despite adding a dash of supposedly riskier assets to the mix. 🥳
In fact, a 50-50 stock bond allocation is roughly as volatile as an all-bond portfolio, and boasts nearly twice the return. This uplift in return for comparable volatility is a the “free lunch” attributed to diversification that the economist Harry Markowitz won his Nobel for.
The late Charlie Munger was famously dismissive of the idea of diversification, saying it was “for the know-nothing investor”. And, sure, with perfect foresight and no withdrawal requirements over a 25 year horizon, no investor needed to be nervy. But YOLOing pension funds into single stocks or even asset classes requires a lot of knowledge (or at least a very strong belief) about the future. Maybe more knowledge than a CIO can persuade their board they truly have.
And importantly, the whole SAA edifice produces not only a benchmark for institutional investors but a governance process.
After all, if your capital market assumptions lead you to, say, a 60/40 equity/bond allocation, all deviations from this benchmark can be measured in risk terms.
Sure, you can have 65 or 55 per cent invested in stocks if you think stocks are heading higher or lower, but this will consume some risk budget versus your ‘neutral’ allocation. Or if you want to carve out a large chunk of the risk budget for stock pickers or factor tilts, you might have little left for tactical asset allocation.
So what is this fancy new ‘Total Portfolio Approach’ then?
For a supposed new paradigm, TPA is surprisingly hard to nail down.
Jayne Bok, head of Asian investments at Willis Tower Watson, says that “TPA is not a specific model with a singular destination, but rather a range of approaches”. Chris Hyde, Head of Asset Allocation at NZ Super calls it a “state of mind”. And Jacky Lee, Head of Total Portfolio Group at Healthcare of Ontario Pension Plan, and Redouane Elkamhi, a finance professor at the Rotman School of Management, describe it in a paper earlier this year (their emphasis) “not [as] a single set of rules, but a shift in mindset.”
So is it really just vibes, with a bit of marketing lathered on top, or is there something real here?
To understand what’s going on we really we need to step back into that part of the SAA process where the investment committee looks inwards to crystallise what returns they are really trying to achieve — and how bumpy a ride they are willing to stomach.
So, working with the average 10-year US bond and equity forecasts we looked at earlier, let’s pretend that you need to get whatever the inflation rate is plus 2.5 per cent over the medium term to make your pension payments, and you reckon inflation is going to average 2.5 per cent. So what is the minimum volatility benchmark that gets you a 5 per cent per annum return?
🤖 🧮 0.565 x [stock forecast = 5.7%] + 0.435 x [bond forecast = 4.1%] = 5%
OK, a benchmark of 56.5/43.5 stocks/bonds is the SAA solution. But it’s not the be-all and end-all. It’s only a means to an end — the end being inflation+2.5 per cent — and a means that is based on the accuracy of long-term forecasts. Which is why problems can happen.
While a lot of effort goes into getting these forecasts right, they tend to be wrong. Luckily for asset owners, they’ve tended in recent times to be wrong by being too conservative. But in going through the SAA process the investment committee not only establishes the sort of market index benchmark they might need to adopt to get to their return goal. They also establish the kind of volatility they’re willing to endure. And this is useful.
Going back to our ex-post “efficient frontier” of US stocks and bonds, if the past is a perfect predictor of the future [erm . . . ] then any committee that is relaxed about signing off a 56.5/43.5 bond/equity split will have also decided that they are OK with the bumpiness of the ride this brings. In chart terms:
So this is what TPA basically is. Rather than have a dusty old investment committee set the asset mix that will determine overall returns once every one or three years, the portfolio’s target is handed over to the asset owner’s investment staff — along with an overall expectation for return volatility. They then enjoy more day-to-day freedom to decide how to do this in practice.
Not everyone implements it this way, but we think this captures the gist of TPA.
What does this mean for actual investment returns?
The investment consultant’s Willis Towers Watson’s Thinking Ahead Institute reckons that TPA adopters have achieved an average performance edge of 1.3 per cent per annum over SAA adopters over a 10-year period, relative to traditional SAA investors.
This is a killer number. Moreover, it has since been updated to 1.7 per cent for the period ending 2024.
We got in touch with the Thinking Ahead Institute to understand where the number came from. After all, comparing returns across currencies and risk profiles is very tricky.
It turns out that the figure comes from a big asset owner peer study the group puts together. The Thinking Ahead Institute scored each of the participating asset owners on their TPA scale, from 0-5 and then compared 10-year total returns, and 10-year alpha. They then compared the average 10-year alpha of asset owners that WTW considered to be on the SAA end of the scale (0-1 scores) to the average alphas of asset owners at the TPA end of the scale (4-5):
2.3 per cent minus 0.6 per cent is 1.7 per cent. Bingo!
We took a look at three of the asset owners frequently held up as poster children for TPA investing, and who all score 4 or 5 according to the Thinking Ahead Institute. These are the New Zealand Super Fund, Singapore’s GIC and the Canadian Pension Plan. Maybe looking at what they do will help us better understand this phenomenal performance advantage?
Lord of the Investment Returns: New Zealand Super
NZ Super is the country’s sovereign wealth fund, financed in part by debt issuance, and has recorded staggeringly strong performance. As we know, it was led until 2024 by current CalPERs CIO Stephen Gilmore.
In what we hope was a nod to Tolkien, the government christened the committee tasked with handling this money to “the Guardians”. And these Guardians have seen fit to provide the NZSF with a reference portfolio consisting of 80/20 stocks/bonds, hedged to New Zealand dollars (although the allocations they make can deviate quite a lot).
This is not an easy reference portfolio to beat: it has done really really well. But the NZSF has done even better. Here’s how the ex-post efficient frontier looks using their mix of indices over the past decade:
As we wrote in a deep dive on the whole operation last year, their performance kicker has come largely from strategic tilts applied by the CIO. Huge macro punts, like shorting the Kiwi dollar to the tune of 40 per cent of the SWF’s NAV, have generally ended well. 😮💨
And currency bets have been key over the past decade. The reference portfolio is hedged entirely to Kiwi dollars, presumably reflecting the Guardians’ ultimate goal being to pay pensions in NZ dollars. And decisions to maybe not hedge back to Kiwi dollars have been massively well rewarded, as we can see by slinging on a second efficient frontier in which benchmark stock exposure is no longer hedged back to NZ dollars:
To be clear, we’re not belittling NZ Super’s performance. They’ve made tonnes of investment returns by latching on to a megatrend (presumably among other things), and captured it for the public purse.
Moreover, this kind of killer macro punting would not have been easy to do if the CIO’s hands were tied to the kind of strategic asset allocation approach where individual components of the overall fund are farmed out to managers each tasked with either delivering index returns, or outperformance using the tools that won them their mandate (eg, picking better stocks, managing commercial properties more lucratively, etc).
So we’re happy to chalk it down as a win for TPA. And given that this is Gilmore’s experience of TPA, maybe it’s a decent model of what to expect it to look like in California.
Enter the Dragon: Singapore’s GIC
The Singaporean state knows what it wants from GIC — manager of the country’s sovereign wealth, a lot of its foreign reserves, and all the state’s assets backing retirement guarantees it has issued residents. It wants to preserve and enhance the international purchasing power of reserves over the long term. This means beating global inflation, however they choose to measure it.
But just like NZ Super, Singapore’s GIC is given a reference portfolio that provides some sense as to how bumpy a ride the government is willing to endure. And, like NZ Super, the reference portfolio is framed as a portfolio of stocks and bonds. In the case of Singapore, it’s a 65/35 equity/bond split.
GIC is more than a little secretive, but it does disclose some high level asset allocation splits in their annual reports. You can see that it doesn’t actually stick too closely to this 65/35 reference portfolio split:
GIC has allocated slugs to private equity, real estate, and held way more EM stocks than might be expected from a purely market-cap approach to global equity portfolio management.
How has this worked out? GIC doesn’t share its monthly investment return series, which might have allowed us to pull together funky efficient frontier charts and do a coulda-shoulda-woulda analysis. But it does publish historic return and volatility numbers for both the actual GIC portfolio and the reference portfolio each year:
Observant readers will note that all the red dots — representing the annualised investment returns and the volatility of these returns for the GIC’s portfolio over 5 years, 10 years and 20 years — are to the left of all the blue dots (which show the annualised returns and volatility for the GIC’s reference portfolio over the same periods). Which is a bit weird given that the whole point of a reference portfolio is to frame the amount of investment risk the state is willing to take.
Two explanations come to mind. First, it could reflect GIC being a bit of a nervous Nelly. Financial markets have been on a tear, but the world has not been short of things to worry about, perhaps tipping them into a more conservative approach than the reference portfolio allows.
Second, this could reflect some “volatility laundering”. We don’t know how GIC has treated their private market holdings, but if it follows the crowd in smoothing valuations, these holdings will depress GIC’s volatility compared to that shown for their reference portfolio.
But really observant readers will notice that almost every one of the lines connecting each year’s red and blue dots in the five and 10 year charts slopes north-east. Which is to say that no matter how GIC have treated the valuation of their private market assets, they’ve failed to capture the abundant returns associated with the risk budget that they’ve been handed.
So, do we chalk this up as a loss for TPA? It’s hard to say. The portfolio has beaten its target of global inflation handsomely. But frankly, pretty much every mix of risky assets would have done so. And an SAA approach would almost certainly have delivered stronger returns if it had anything like the investment risk the Singaporeans signed-off as appropriate for GIC.
Moving to Canada: CPPIB
Canada Pension Plan Investment Board — which manages Canada’s public pension assets — has an interesting investment process: the risk punting is done up front and irregularly by the sort of committee that you’d think would set an SAA.
It makes a bunch of capital market assumptions and then estimates what it terms “the minimum level of market risk that would generate sufficient investment returns to maintain the long-term financial sustainability”. The result is a de facto two-asset reference portfolio consisting of Canadian government bonds and global stocks.
But they then, “at least once every three years”, work out how much risk they really want to take. Since 2019 the answer has been quite a lot. In fact, since 2019 CPPIB has targeted overall risk in line with a portfolio consisting of 15 per cent bonds and 85 per cent global stocks.
The below chart shows not their actual allocation, but rather the original SAA that results from CPPIB’s capital market assumptions process on the left, and on the right the reference portfolio associated with the committee’s evolving risk appetite:
So if we’re going to assess whether CPPIB is a TPA hit or miss, we should probably keep both the reference portfolio and this target risk portfolio in mind. Because CPP:
. . . manage[s] the market risk levels of each of the base CPP and additional CPP Investment Portfolios to closely match their respective targeted level of market risk.
(As an aside, it’s probably worth telling you that CPPIB run a “base CPP” and a smaller “additional CPP” portfolio which is run with a different level of risk. But in the interest of brevity [Ed note: lololololol] we’re just going to ignore the additional CPP portfolio which accounts for only about 8 per cent of the total assets.)
If CPPIB had taken a SAA approach this would involve allocating 15 per cent to a bunch of bond managers and 85 per cent to a bunch of equity managers (to match the middle column in the chart below). Instead it invests in line with the column on the right: big slugs in credit, real estate, infrastructure, public and private equity.
And because the asset mix is — in aggregate — less volatile than a reference portfolio consisting overwhelmingly of listed stocks, CPPIB uses leverage to dial up investment risk and hit its target (that’s why the chart below shows them with negative cash to represent the borrowing):
Moreover, with TPA the investment team can weigh up the merits of credit vs real estate vs infrastructure vs rates etc and have pools of capital compete against each other on an ongoing basis.
But have the Canadians become richer by investing in this different bunch of assets vis-à-vis what they could have done with a cheap and cheerful tracker fund?
Maybe? The CPP base fund reported an annualised return over the past 10 years of 8.4 per cent (with an unknown level of volatility), which is a bit behind the 8.7 per cent that we calculate an 85/15 portfolio of global stocks and Canadian bonds would have given you in Canadian dollars over the same period. Confusingly for us, the CPP annual report thinks the number for the market risk portfolio is higher over the past decade at 9.1 per cent, but also states that:
. . . we no longer measure relative performance against the Market Risk Targets as they are an expression of our targeted level of market risk rather than prudent investible alternatives to our Investment Portfolios.
Removing the comparator return number you think you’re lagging? Roger that.
Still — 8.4 per cent per annum is a strong number, and almost 200 basis points in excess of the 60:40 minimum risk portfolio. And this was where their initial strategic asset allocation process put them.
Moreover, the CPP investment team was obviously comfortable having a risky portfolio spread across a number of asset classes like infrastructure, property, credit, stocks and bonds. It’s not immediately clear that they would have been equally comfortable squeezing the C$714bn fund into a passive stock index that has become increasingly concentrated in a few massive US tech names (even if doing so has worked out pretty well).
So maybe a TPA was instrumental in giving them the confidence to take more investment risk that has ultimately worked out very well for them.
About that alpha
We know what you’re thinking: if Alphaville are finding it tricky to score wins, losses and draws among the TPA poster children, how was that 1.7 per cent figure from Willis Towers Watson’s Thinking Ahead Institute actually calculated?
They kindly sent us their underlying data. So first up we split the three TPA score buckets into six buckets and compared asset owner alphas:
While superior alpha in bucket zero (the hardest of hardcore SAA adherents) looks a bit out of place, this six-bucket split really shows that there does seem to be something to this TPA thing.
Presumably the asset owners with the highest TPA scores not only outperformed their benchmarks by the most, but also produced the highest total returns?
Oh.
Well, perhaps this was a dumb question. Benchmarks are designed according to the specifics of the institution. If a particular institution has the appetite to take only a low level of investment risk — perhaps because they have regular cash flows to make or lack the stomach for deep drawdowns — it would be plain foolish to compare their total returns. And all these institutions may be denominated in different currencies anyway, so alpha is probably a better measure after all.
Is it really alpha though?
As any finance student will tell you, alpha isn’t just an excess return. It’s an asset or a portfolio’s excess return over the market or benchmark’s return on a risk-adjusted basis. Taking a tonne of broad market investment risk and measuring yourself against a benchmark of cash can give you stonking excess returns, but needn’t deliver any alpha.
Without volatility numbers for the portfolios and their risk budgets we won’t actually know whether the excess returns are true alpha, But we can at least check out the benchmarks against which each of the asset owners in the survey are measured. You know, just to make sure they’re not completely mad:
Some of these (in dark blue) are plain vanilla investable SAAs or reference portfolios. Sure, TPA asset owners won’t be buying the reference portfolio, but they are, by definition, a fair measure of the return to risk they represent. Some are a little more funky. But some are just numbers — and while it’s fine to record an asset owner’s excess returns to this target number as an excess return, they are not (strictly speaking) alpha.
This is not to say that it is in any way a bad idea for an investment committee to hand the problem that they’re trying to solve with their SAA over to the actual investment team. The whole point of an SAA might be to deliver a real inflation-adjusted return of 3 per cent a year and throwing together benchmark components are just a best-guess means to this end.
There are wild uncertainties attached with the whole capital market assumption process, and it is arguably bizarre that an organisation should spend the vast majority of their investment risk budget in a committee populated by some mix of investment consultants, internal staff and senior muggles that meets perhaps once a year.
However, funky uninvestable benchmarks are increasingly common the more we move rightward across the chart. And this is where the asset owners holding higher TPA scores reside. And this means it’s hard to make the case that a TPA actually helps deliver higher alpha. Because generating decent investment returns that happen to be higher than inflation is different to taking roughly market-neutral positions that generate actual outperformance.
Perhaps this is pedantry. Should anyone really care what we call the excess returns delivered by high TPA funds?
Yes. Yes we should care a lot if the motivation for adopting TPA is that it will bring higher returns.
Back in 2020, a survey of asset owners by Willis Towers Watson found that half thought TPA should produce a performance advantage of at least 50-100 basis points per annum versus a SAA approach. It’s not clear from the data that this has actually been achieved.
Anyone can boost their excess returns by lowering their hurdle rate. And while this is not what high-TPA asset owners meant to do — after all, total returns will be far more important to them than excess returns — we need to be really very careful drawing conclusions from data from two dozen funds’ excess return experience. And especially careful when the targets chosen by high TPA funds turn out to have been easily eclipsed just by putting money into risky assets.
As Andrea Calosi, associate director of the Thinking Ahead Institute, told Alphaville:
[T]here are numerous caveats to this type of analysis — for example, the timing of TPA adoption or implementation varies by fund, performance comparability across different currencies, and performance assessment methodologies may differ among funds, among other potential unknown factors.
Quite.
Other reasons why TPA is winning
But just because we don’t find the historical performance data compelling, this doesn’t mean that there aren’t good reasons why institutional investors shouldn’t make the move.
As Jan Loeys and Alexander Wise at JPMorgan wrote in a note for clients earlier this year, there are a bunch of problems with the traditional SAA approach:
The main one, one we have struggled with ourselves, is that your strategic allocation as your long-term anchor should only be changed every few years or so, but the main inputs into your SAA, expected long-term returns across asset classes, are highly dependent on their starting IRR, which can change from day to day.
Yep. If all your capital market assumption analysis points to stocks producing an average 5.7 per cent return over the next 10 years and the next day prices halve, isn’t that a development that should be acted on by a CIO?
In other words, we can see why it might not make sense for large sophisticated investors to prioritise decisions made by a committee with eight-month old information over those being made by a large staff of investment professional working with up-to-date data.
Willis Towers Watson’s 2025 asset owner survey found “all the peers recognised the value of TPA to provide more flexibility and accuracy in increasingly tricky VUCA [volatility, uncertainty, complexity and ambiguity] conditions”. And they also found that the overwhelming majority want to change their organisations in a TPA direction.
It’s not exactly riskless
However, ditching the guardrails of a SAA poses new problems to boards.
Firstly, it transfers a lot of power into the hands of the CIO. But the boards will be on the hook if their CIO’s macro punts lose a tonne of money, and probably asked why they didn’t have tighter controls. At least with SAAs, everyone sort of understands that poor returns are just because “markets”. Like Keynes wrote, “worldly wisdom teaches that it is better for reputation to fail conventionally than to succeed unconventionally”.
Second, SAAs protect boards from FOMO. Or rather, from the question about what to do when markets rally so hard they bring prospective returns below target. As JPMorgan’s analysts note:
The CIO has two options: they can decide that markets are expensive and go defensive into a lower-risk allocation and thus lower-return assets on the expectation that markets are due to fall, and their IRRs will rise in due time. Or the CIO can decide that in order to meet their return target, they need to raise the weight of higher-risk assets.
The first option leaves the CIO telling the board to trust them and be patient. If it turns out the market has moved indefinitely moved to a lower risk premium (say, the market P/E has moved to a permanently higher plateau, credit spreads or real bond yields to a new lower normal) the whole fund gets marooned in low-risk low-return assets.
And the second option? Having missed the upside, this approach leverages their portfolio into assets just when they’re at their richest.
Wrapping up
This has been an exceptionally long post, and we congratulate you both for making it to the end.
You probably invest any savings you have using a rudimentary Total Portfolio Approach: watch the news, check out prices, do a bit of research, place your bets, and when the facts change, change your portfolio without fixating on a bunch of low-confidence assumptions you made last year. Moreover, your investment goals are probably something like paying off your mortgage, putting your kids through college, building enough savings to retire comfortably, or some such thing rather than beating a composite multi-asset market-cap-weighted benchmark by a certain number of basis points.
Large and sophisticated investors seem increasingly to be deciding that they aren’t so different to you. With Wilshire — Calpers’ investment consultant — telling the board they are “comfortable” with a shift to TPA, another big beast looks likely to abandon strategic asset allocation by the end of the day.