On The Rise Of ETFs As A Driver Of Bond Returns
The seemingly inexorable rise of corporate bond ETFs (most specifically HYG and JNK is the high-yield market, and LQD in investment grade) have been discussed at length here as both a 'new' factor in the underlying bond market's technicals (flow) as well as their correlated impact on equity and volatility markets. Goldman Sachs' credit team delve deep into the impact of these relatively new (and rapidly growing) structures with their greater transparency but considerably higher sensitivities and conclude that not only are they here to stay but the consequences of ETF-inclusion (dramatic outperformance bias relative to non-ETF bonds) are deepening the liquidity divide (and relative-value) of what is already a somewhat sparsely-traded market. Our concern is that, as the divide grows (and liquidity is concentrated in ETF bonds), given the crowding tendency we have witnessed, (even with call constraints at extremes thanks to low interest rates), this is yet another crowded 'hot potato' trade hanging like a sword of Damocles over our markets (courtesy of Bernanke's repression).
Via Goldman Sachs: The "ETF Bid" - A Robust Driver of Bond Returns.
The dramatic growth of credit mutual funds over the past five years has been something of a paradigm shift for the US corporate bond market. According to data from Lipper, IG and HY mutual funds manage roughly $1.3tr and $279bn, respectively, up from $525bn and $126bn at the end of 2007. Even when scaled by the overall size of the US corporate bond market, these figures suggest a bigger ownership share of credit mutual funds. Perhaps more impressive is the growth of corporate bond ETFs, whose size for IG and HY has increased to $105bn and $31bn, respectively, from just $12bn and $288mn at the end 2007.
Despite all-time low yields, the appetite for corporate bonds remains firm, with net inflows to IG and HY mutual funds still running at a robust pace.
Inflows to the HY market have been steady all year with the exception of a brief episode of outflows from mid-May to mid-June, while IG inflows have remained firm, having been positive for all but two weeks this year.
As we often point out, the strength of the inflows is by and large a reflection of sluggish growth, which has maintained a friendly outlook for inflation as well as a challenging outlook for growth expectations, trends we expect to persist.
In the past we have extensively analyzed the price impact of mutual fund flows, showing that fund flows provided a significant boost to bond returns in the early months that followed the crisis and gradually normalized afterwards. In this Credit Line, we focus on the ETF market and use bond-level data to quantify the impact of ETFs on the cross-section of IG bonds returns.
Despite their recent growth, ETFs remain a relatively small subset of the mutual fund complex, but they are of particular interest to us, for two reasons. First, unlike mutual funds, ETFs generally try to track an existing bond index and are thus more transparent in terms of their composition. Second, unlike institutional investors, ETF managers need to be fully invested, and thus need to put money to work “urgently” when inflows increase. Intuitively this should make returns on ETF bond constituents more sensitive to flows.
The ETF factor: Is it really there?
Our primary goal is to determine whether, all else equal, the bid for ETFs has boosted the return on their bond constituents, both in absolute terms and relative to the broad market.
More specifically and using the bond constituents of the iBoxx USD domestic index and its sub-index the LQD IG index, which is tracked by one the largest IG ETFs, we investigate the following questions:
- Controlling for maturities, ratings, sectors and liquidity, how do the bond constituents of the LQD ETF perform relative to those bonds that are not part of the ETF? Put differently, how do two otherwise identical portfolios of bonds, one with ETF constituents and the other with non-ETF constituents, compare in terms of performance.
- To what extent is this relative performance related to ETF flows?
To address the above questions, we use a factor approach and construct hypothetical portfolios, one with only ETF bonds and the other with non-ETF bonds, that have otherwise identical compositions. These portfolios are constructed using the constituents of the iBoxx USD domestic index of which the LQD ETF is a sub-index.
One immediate challenge that such an exercise raises is that our 'ETF' factor might be hard to disentangle from a 'liquidity' factor. In the extreme case where the ETF only include on-the- run bonds, the ETF factor might just be an arte fact for the on-the-run/off-the-run premium. Partly to address this issue, we include an “on-the-run” dummy variable in our regressions.
Our definition of on-the-run bonds uses the following rule. We group our universe of bonds into three buckets in terms of their original maturity: 1- to 7-year, 7-to 15-year and 30-year and longer. Within each of these buckets and for each issuer, we rank the bonds according to their age. The most recently issued bonds are considered on-the-run while the remaining ones are treated as off-the-run.
Exhibits 5 and 6 plot the cumulative total return and spread change since January 2009 for these two otherwise identical portfolios of ETF and non-ETF bonds. The plot shows that after controlling for maturity, rating, sector and the on-the-run/off-the-run premium, ETF bonds outperformed their non-ETF counterparts from the second half of 2009 to the first quarter of 2011. They subsequently underperformed in the second half of 2011 as the European crisis intensified only to resume their outperformance during the LTRO rally and then more recently following Draghi’s speech in July.
Compared to January 2009 and as shown by Exhibit 6, a monthly rebalanced portfolio of ETF bonds trades 268 bps tighter while an identical portfolio of non-ETF bonds trades 200 bps tighter. This suggests that the bid for ETFs has acted as a tailwind for its bond constituents since the economy turned the corner in the second half of 2009.
Exhibit 7 addresses our second question—the relationship between the relative performance of ETF bonds to ETF flows—showing a simple scatter plot of the monthly return on a long ETF vs. non-ETF bonds against the flow 4-week moving average into IG ETFs.
Our estimates indicate that the correlation is decent with every additional $1 bn worth of weekly average inflows to IG ETFs translating into 60 bp of additional monthly return on a strategy that is long/short two otherwise identical portfolios of ETF and non-ETF bonds.
In sum, the above evidence suggests that fund flows to ETFs are likely to remain an
important driver of credit spreads for the foreseeable future.