Market-cap equity indices have come in for stiff criticism over recent years, but Martin Steward finds their shortcomings are nothing compared with the bond market. A new breed of indices attempts to address their worst failings
“The cap-weighted index is a fair representation of the market opportunity,” says Shane Shepherd of fundamental indexation pioneer Research Affiliates. “But it comes with a whole host of other problems.”
We are now used to hearing these words about the equities universe. The arguments are strong, but every now and again the contrarian inside us all wants to respond: ‘Hold on a second! There are some good things about market-cap, too.’
There is the practical advantage of directly linking portfolio weights to their constituent prices - there is no need to rebalance unless a completely new constituent enters the index or an existing constituent drops out, and that means very low trading costs.
In addition, buying companies whose share price has gone up sits well with the fact that equities exhibit momentum as investors buy more shares in steadily improving companies.
But Shepherd is Research Affiliates’ head of fixed income research - not equities research - and these arguments collapse in the face of standard ‘cap-weighted’ bond indices, which weight constituents according to the value of outstanding debt issuance.
While companies tend not to issue equity very often and countries don’t issue equity at all, both issue bonds regularly, which means the weight of an issuer in the index is a moving target - leading to a lot of rebalancing.
Even if weights were determined by, say, price per bond, would it make any sense to put more money into entities that are becoming more and more indebted? It might do if you thought yields would rise with the stock of debt, thereby compensating you for the extra risk.
That might be the case, but it can be difficult to tell: where public equity prices are available from one millisecond to the next on electronic exchanges, bonds trade over the counter - and many don’t trade for days on end. Outside the biggest, most liquid markets, index weights are determined by ‘stale’ prices. And as for the biggest, most liquid markets, well - a quick look at the US and Japan will disabuse you of the notion that net new issuance pushes up yields.
Of course, that is no different from the ‘buy high, sell low’ dynamic in cap-weighted equity indices - but at least there you can feel there is some reflection of company fundamentals in share prices. The bond markets, by contrast, are systematically rigged by governments and their central banks, and traditional indices are very likely to push you to the issuers who are rigging it most furiously. Construct an OECD outstanding issuance-weighted index and the US and Japan will hog 30% each and the next five spots, accounting for another 22%, will be taken by Italy, France, Germany, the UK and Spain.
Since the credit crisis struck, asset managers such as Lombard Odier and PIMCO, and index providers like Barclays Capital and Research Affiliates, have been working to develop indices that take some account of issuers’ ability to pay back all the debt they take on.
Lombard Odier’s ‘fundamental-weighting’ methodology is arguably the most comprehensive. Pure economic size (GDP adjusted for purchasing power parity) accounts for 30% of a weighting, and GDP growth 10%. Public and private debt-to-GDP ratios account for 15%, and 5% is down to the ratio of foreign-held debt-to-GDP.
These three measures tie the process into the stock of debt that determines weightings in a traditional bond index, but invert the result - if two countries have the same GDP, the one with less debt outstanding will be more heavily-weighted.
A further 10% of a weighting is determined by the current account-to-GDP ratio. Finally, demographic and political risk - in the shape of the dependency ratio and the issuer’s place on the ‘misery index’ of unemployment and inflation - account for 20% of a weighting.
Barclays Capital’s ‘fiscal strength-weighted’ indices are similar in that they use country scores derived from measures of financial solvency, dependence on external financing, and institutional strength - but only to adjust what remains an essentially cap-weighted benchmark.
Research Affiliates’ ‘fundamental-weighting’ process is more like Lombard Odier’s in the sense that it does not start with the cap-weighted index, but differs by focusing purely on economic size factors, eschewing ratios like debt-to-GDP.
For corporate issuers it uses factors familiar from its well-known equity indices: sales, cash flow, dividends paid and book value. For sovereigns it looks at GDP, population, land area and energy consumption as proxies for each country’s capital, labour force, natural resources and technological sophistication.
PIMCO’s Global Advantage Bond index methodology (‘GLADI’) similarly focuses on economic size and ignores stock and flow of debt, but goes a step further in simplification: it is a pure GDP-weighted approach.
Stéphane Monier, global head of fixed income and currencies at Lombard Odier Investment Managers, makes a defence of his firm’s nine-factor approach that seems intuitive - “You can’t truly summarise a country in nine ratios, but at least we have moved from trying to do so with just one.”
His argument that forward-looking demographic and political factors improve on the backward-looking bias of the traditional indices and GDP and debt ratio factors, also makes sense.
Figure 1 shows not only how much difference the process makes compared with standard indexing methodology, but also the granularity with which each constituent is differentiated.
For example, compare how important GDP is for the weightings of Germany and Finland, or the influence of private debt-to-GDP on the weightings of Germany and Belgium.
But that can be the grounds for criticism of over-complexity. Xiaowei Kang is director of index research and design at S&P Indices, which constructs a number of smart-beta equity benchmarks but has not entered the alternatively-weighted bond index market. He observes that the asset management industry is tending towards combining strategies based on simple parameters, rather than looking for individual strategies based on multiple parameters.
“Things get really complicated if you have a dozen factors, and it becomes difficult to determine which ones are really driving performance,” he says. “And the more factors you deploy, the more difficult it is to find investors who agree with every single one.”
David Fisher, account manager for the GLADI strategies at PIMCO, says that his firm focused on GDP alone so that its process was “transparent” and “easy to understand”.
Every factor that goes in requires decisions to be made about how it will be translated into a constituent weighting, he observes, and PIMCO was anxious to minimise those decisions to avoid charges that it was simply building a systematic trading model.
“Other approaches that use a lot of variables veer into the realm of model-based active management,” he suggests.
Shepherd at Research Affiliates says adding more factors results in greater stability of weightings via the diversification effect, but that the benefit becomes negligible beyond four or five. Moreover, the desirability of stable weightings brings him into agreement with Fisher on a preference for absolute measures such as GDP and population, over the ratios like debt-to-GDP that Lombard Odier uses.
“Our measures are more stable than ratios, and we even take five-year averages of those to take out the economic cycle,” he says.
Fisher agrees, pointing out that weighting by GDP will implicitly lead to overweights in issuers whose debt-to-GDP ratios are lower than the average, and vice versa. The same inversion of the outstanding issuance-weighted index’s relationship between weighting and debt value is maintained, but via a more stable parameter that incurs less rebalancing.
Whichever factors are used, alternative-weighting schemes inevitably move indices away from big concentrations in the world’s two or three big debtors. The difference is most pronounced in global indices because emerging economies, which still have small investable debt markets, now contribute a fair proportion of global GDP.
But even in developed-market indices, there can be marked deviations from the traditional weightings. Japan, which has been stacking up debt on a stagnant economy for 20 years, would have represented 15% of a pure GDP-weighted developed world index in 1990, but represents just 8% today. It takes up as much as 30% in some traditional indices.
So moving to an alternative-weighting process has consequences. Those extra 20 percentage points of Japan’s weighting will be distributed among the other constituents in a sort of cascade effect. Even if this is done strictly pro-rata, without any caps on individual issuers (as it is by all the practitioners in this article), that can bring an index perilously close to being equally-weighted.
A pure outstanding issuance-weighted index of OECD sovereign bonds ranges from a 29.3% weighting in the US to less than 0.2% in, for example, New Zealand and Hungary.
Lombard Odier’s Global Government OECD benchmark cuts the US down to 16.8% and has Germany, Korea, Mexico, Turkey and France weighted between 5.0% and 6.5%; the other 14 names are all weighted above 3%.
“The fundamentally-weighted index has a lot more countries with smaller weights, so it does make more sense when you have a large universe of many countries,” says Monier. “In our global index, Portugal and Hungary are almost the same weight as the Czech Republic, which might seem counter-intuitive. But that is why we would always recommend taking an active approach to managing against this benchmark, rather than systematically replicating it.”
What’s the point of an index that you can’t replicate? It’s a fair question - but one that applies just as much to traditional indices. Because these indices are concentrated in one or two big debtors, they are often constructed with caps on single-issuer weightings.
This can result in a marked cascade effect: The JPMorgan GBI-EM index, for example, would have 27% in Brazil were it purely outstanding issuance-weighted. Instead it has a cap of 10% per issuer. That means 17 percentage points of weighting cascades down to the next biggest constituent - Mexico. But Mexico is already at 13%, which means that 20 percentage points has to cascade down to the next biggest constituent - and so on, until the index ends up with six of its 15 constituents weighted at 10% (the level of the cap), and another four weighted above 7%.
These issues are relative, however. What is absolutely true is that, by moving out of larger debtors and into smaller ones, you move from more liquid to less liquid markets.
Monier protests that liquidity is not simply a function of size, pointing to the fact that, while Italy’s BTP market might have been more liquid than Germany’s Bund market five years ago, it certainly isn’t today.
“Improving liquidity during times of crisis involves adding quality,” he insists. But most, like Fisher, are more circumspect: “That’s an exception to a general rule,” he suggests. “Around the world, the fact is that the most liquid markets are Treasuries, JGBs, Bunds, Gilts.”
As an index provider, Research Affiliates feels obliged to approach the illiquidity problem by creating ‘master’ indices that follow the pure fundamental-weighted process, but also ‘investable’ versions to be licensed by asset managers, which have liquidity constraints and aim to replicate individual yields curves via representative issues.
Shepherd estimates that turnover is about 6% higher in fundamentally-weighted bond indices than traditional ones, and because liquidity in smaller issuers’ bonds can be “terrible”, that adds about 1.5 basis points to its annual costs.
PIMCO, as an asset manager, takes an approach more like Lombard Odier’s - the priority was to create something practical as a benchmark for active management.
“We didn’t want to conduct a great ivory-tower exercise that made a nice academic point but which, on practice, you couldn’t actually buy,” says Fisher. That means constraints on illiquid markets, which, because many are emerging market domestic bonds, are sometimes proxied with currency forwards.
Another issue with moving away from the big debtors is that larger weights go to issuers whose term structures are both patchier and, generally, shorter, than those of the biggest debtors. Even Australia doesn’t issue much beyond 10 years.
That’s not such a big problem, but it necessitates a decision. Do you want to maintain the same duration as the traditional bond indices, or do you consider the shorter-duration bet to be part and parcel of the alternative-weighting process?
Lombard Odier looks to maintain the duration of traditional indices with a simple interest-rate swaps overlay. But both Shepherd at Research Affiliates and Fisher at PIMCO accept the duration mismatch as an inevitable result of the alternative weighting - and even another benefit of it.
“We are also putting other risks into the portfolio, after all,” says Fisher. “The currency risks of overweights to emerging markets or Australia, for instance. “And we do think that, over time, this will become less of a distinction. If someone told you in 1998 that you would see Mexico issuing 30-year nominals in pesos and people buying them at fairly low yields within 12 years, you’d have thought they were crazy.”
And here, perhaps, we have one of the most compelling arguments in favour of breaking free of the outstanding issuance-weighted bond indices. As developed economies have increased their debt loads, they have simultaneously seen their share of global GDP flowing towards fast-growing emerging economies.
The weighting systems discussed here favour issuers that have experienced economic growth in the most recent years - probably because capital market development lags economic development. “GDP weighting gives you first-mover advantage against those who are weighting according to the size of the capital market,” says Fisher.
That probably cannot account for the outperformance of smart-beta bond indices against traditional indices in their back-tests - Research Affiliates claims between 71 and 182 basis points annualised outperformance since 1997 from most its fundamentally-weighted indices, while Lombard Odier claims similar results for its indices since 2001. However, it may well be the driver of tomorrow’s outperformance. That, in turn, might account for the assets in excess of $20bn allocated to these strategies in the months since they were launched.
Frankly, the business of bond indices has always been a mess. The new breed of indices does little to clean up that mess - in some ways it adds to it. But there is no doubt the new approach addresses some of the most egregious biases of traditional indices - biases that have increased investors’ risks and depressed their returns for many, many years.