With this year’s record flows into Bond ETFs ($172 B ytd), the vast majority being index funds, explicit Fed support for corporate bond ETFs and indexed corporate bond portfolios (link), as well as continued market share gains by index funds in general, I thought I could take the opportunity to educate my loyal readers on the portfolio construction strategies behind these highly popular investments. Notice I hesitate to use the word “passive” in any part of my discussion – this will become clearer later.
First, let’s all agree on the definition of an index fund: An index fund is designed so that its target return matches the performance of a benchmark. What are the various ways this can be accomplished, and what are the different trade-offs?
- Full Replication: Own every component of the index in its exact weight. For a highly liquid asset class with minimal trading costs, this is the most reliable strategy. Large-cap equity funds such as SPY, IVV, and VOO own all the members of the S&P500 in their benchmark weights, and as result, track extremely close to the underlying index. This method holds true even for equity ETFs with a much larger number of holdings, since the constituents are all liquid, for example, those tracking the Russell 1000 (IWB, VONE) and Russell 2000 indices (IWM, VTWO).
- Derivative Exposure: Own a swap, future, or other derivative that is structured to return the performance of the underlying index. While this is more common with ETFs in Europe due to favorable tax treatment, some better-known examples in the U.S. are leveraged or inverse ETFs which use futures to gain exposure to their target benchmarks (QLD, SSO) or ETNs (Exchange Traded Notes) which hold futures or swaps (AMJ, VXX) because the underlying is not available in cash form. Trading costs for most of these derivatives are quite low, but tracking is typically not quite as close as with full replication.
- Security Sampling: No, this is not what you did at the hors d’oeuvre table at your brother’s wedding last summer, this refers to partitioning your benchmark into its main risk factors, and buying a portfolio of securities that will in aggregate approximate the characteristics of your target benchmark. For an asset class such as Fixed Income, where underlying securities can be expensive to trade or may trade infrequently, this allows the portfolio manager to minimize the cost of trading and limiting the number of holdings while targeting benchmark returns. Risk attributes will vary by asset class as well as the degree of sampling, and the risk of tracking is higher if security sampling is not adequate, or if historic correlations change dramatically.
This chart from Vanguard does a nice job illustrating the trade-off between tracking and replication: fund PM can lower transaction costs (buy fewer bonds) but fund tracking may be worse, or she can approach full index replication, but effectively will be “overdoing it” by spending too much on trading, when an optimal result is somewhere in the middle:
WHY is security sampling required in Fixed Income? How prevalent is it?
The most well-known fixed income index, the Bloomberg Barclays U.S. Aggregate has nearly 12,000 constituent securities. Not to get super nerdy about it, but in actuality, it is much higher because all the mortgage securities are combined for index purposes into cohorts. In any case, the corporate portion has nearly 7000 distinct cusips. The Ice BofA corporate index, also a common benchmark, has nearly 9000 bonds. Common High Yield Indices have about 2000 bonds. The vast majority of these bonds do not trade every day, many of them do not trade on a given week. This means that if you try and buy these illiquid bonds, you will end up (1) pushing the price up, or (2) paying a high bid-ask spread, both of which can adversely impact performance and fund tracking.
To illustrate the issue further, AT&T has one stock, but nearly 60 AT&T bonds in a typical benchmark, the same goes for all of the large issuers such BofA, GS, and JPM. Even your favorite stocks such as AMZN (18) and AAPL (44) have myriad bond sizes and maturities to choose from. Corporate bonds tend to trade heavily immediately after issuance, and then gradually dry up. In a 60 bond complex for an issuer such as AT&T, likely 5-10 bonds are trading on a daily basis.
So what’s a sad, overworked bond index PM to do? Using an optimization process, quantitative risk model, or fancy excel pivot tables, a PM will divvy up their target fixed income benchmark into its various key risk factors, which may include: Duration, yield curve, credit rating, industry sector, capital structure, seasoning, size, liquidity, issuer, and others. Parametric put out a nice paper on this topic earlier this year, where they show the standard “Rubik’s cube” of index factors and optimization:
Think of every cell in this multi-dimensional Rubik’s cube as some combination of the characteristics above; for example, one cell is BBB-rated telecom bonds with less than 500mm outstanding issued after 2018. Another cell might be Subordinated Bank bonds over $1B outstanding issued prior to 2016; and so forth. Depending on the type of benchmark (HG, HY, short-maturity, etc.) the determination of the factors and resultant cells will vary. For example, a treasury benchmark will only include duration, on-the-run, and curve factors, while a high yield benchmark will be much more sensitive to issuer, rating, and sector attributes. Once the PM has the index all partitioned nicely, she will buy bonds based on availability, price, and liquidity within each individual cell, and hope that returns proceed to mirror the benchmark going forward. Assuming the modeling and purchasing was done diligently, this usually works quite well.
If we look at the set of broad corporate bond ETFs below, it becomes clear that the degree of sampling can vary widely. For a deeper analysis we would need to evaluate not just the number of bonds owned & missing, but the degree of under and over-weights of the holdings and benchmark bonds.
Factors that can drive the degree of sampling in an index portfolio:
- Fund Size: With bonds trading better in block sizes ($250k+), a large portfolio has wider latitude to purchase a broader choice of bonds without dipping into “odd-lot” territory. Thus, all else equal, a larger fund is likely to hold more of the underlying index bonds.
- Fund Age: A fund has that has been around for many years has had the opportunity to participate in more new issues, received many more flows or creations (for an ETF) , and has had time to rebalance and source pockets of liquidity over a longer time, thus bulking up its diversity of holdings. Thus, all else equal, an older fund is likely to hold more of the underlying index bonds.
- Fund Growth: A fund that has been growing has been receiving cash (or in-kind bonds for ETS) and has had the opportunity to purchase acquire index constituents, or rebalance with new cash rather than having to sell existing holdings (e.g. lower trading costs). Thus, a growing fund is likely to hold more of the underlying index bonds.
- Fund Structure: This needs mentioning mainly in the context of ETFs because of the unique share-class structure that Vanguard has under patent. Basically, Vanguard has been able to comingle its ETF with an underlying mutual fund, thus allowing for a highly diversified fund even when the ETF is quite small or new.
Conclusion: The lessons of this article are that not all Bond Index funds are created equal, that bond index portfolio construction involves a high degree of discretion due to the liquidity limitations of the asset class, and that bond index funds may be more “Active” than you might assume, given the degree of sampling vs. the benchmark. Know-thy-funds.
“Let’s be careful out there” – Sgt. Esterhaus, Hill Street Blues, 1981.
Data Sources: ETFlogic.io, etfdb.com, issuer websites (iShares, SSGA, Vanguard, PIMCO), ICE, Bloomberg.
Elya Schwartzman is the founder and president of ESIC LLC, an independent advisory firm specializing in ETFs, indexing, fixed income portfolio management, and investment infrastructure and technology. ESIC also provides independent research on these topics. Over the past 15 years, Mr. Schwartzman has played a key role in the growth of the ETF industry, having managed teams, portfolios, and investment process initiatives for BlackRock and SSGA.
Disclaimers: ESIC seeks to provide paid advisory services to companies in the ETF ecosystem, including but not limited to issuers, index providers, analytics and data vendors, and market makers. Any views expressed are solely of the author and should not be construed as investment advice.