Here's a statement of the obvious: The opinions expressed here are those of the participants, not those of the Mutual Fund Observer. We cannot vouch for the accuracy or appropriateness of any of it, though we do encourage civility and good humor.
Support MFO
Donate through PayPal
Multi-Asset Income Funds: Is the Extra Income Worth the Extra Risk?
"Overall, multi-asset income funds have succeeded in generating plump yields, but that’s because they have heavy concentration risks in areas such as high-yield bonds and large-value equities, meaning the multi-asset income funds can fall much more than one would expect when risk becomes reality."
"The fundamental risks that we see in multi-asset income funds, moreover, can hide in plain view for extended periods.That is, when looking at metrics based on trailing returns, funds can look sedate until a crisis."
"The only Gold-rated fund in the multi-asset income group is Vanguard Wellesley Income VWINX, which resides in the allocation—30% to 50% equity category. It should be no surprise that its profile is more conservative than those of its multi-asset income peers. The fund stands apart by emphasizing dividend growth over dividend level and, strikingly, by swearing off high-yield bonds."
It doesn't reference the new M* categorization system using beta-based MPRS (see a post nearby), but provides good info on relative-performance, relative-beta and relative-SD.
The data lag issue may be explained by the use of monthly data for MPT stats by M*, PV, etc. If they switch to weekly data, then things would improve.
The point in the excerpt below cannot be emphasized strongly enough. There is a reason that some bonds or funds pay more. It is called risk premium. That is not a free lunch. A bet on higher yield may pay off nine years or out of ten, or even better, but that just means that when 00 comes up, the impact is likely to be more severe. IOFIX, SEMMX and their brethren were never "cash subs", regardless of how sedate they looked before 2020.
Return-Based Risk Measurements Miss the Mark The fundamental risks that we see in multi-asset income funds, moreover, can hide in plain view for extended periods. That is, when looking at metrics based on trailing returns, funds can look sedate until a crisis. On this front, multi-asset income funds faced a comeuppance in early 2020 during the pandemic panic from Feb. 19 through March 3.
The classical volatility measure is historical standard deviation, which didn’t signal an impending problem heading into the 2020 crisis. From 2012 up through early 2020, it would have been fair to call multi-asset income funds sedate based on it. Over this nine-year period, they had an average three-year standard deviation of 6.6%, roughly two thirds that of the S&P 500 index’s 10.5% mark. At the end of 2019, multi-asset income funds’ average three-year standard deviation was just below that level, at 6.4%.
It is kind of circular argument by M*. BECAUSE M* uses monthly data, its MPT measurements lack sensitivity that weekly (or daily) data may capture. So, M* should quit complaining and instead change its MPT calculations with (at least) weekly data.
msf said: "A bet on higher yield may pay off nine years or out of ten, or even better, but that just means that when 00 comes up, the impact is likely to be more severe. IOFIX, SEMMX and their brethren were never "cash subs", regardless of how sedate they looked before 2020."
IOFIX and SEMMX were touted as "cash subs" by some participants on another investing board. This was before IOFIX returned -36.18% and SEMMX returned -20.85% in 2020 Q1 ¹.
The data lag issue may be explained by the use of monthly data for MPT stats by M*
BECAUSE M* uses monthly data, its MPT measurements lack sensitivity that weekly (or daily) data may capture.
These are two different issues. Lag (or latency) is merely the delay in reporting something. Sensitivity is the ability to detect small changes. M* is not saying that there's a lag in identifying risk - unless one is suggesting that there's a multi-year lag (due to monthly sampling) because some markets have been sedate for years. Rather, M* is saying that because volatility is not the same as risk, and because there can be risk without volatility, volatility may be a poor metric for it.
On a daily basis, the market is close to a random walk with a positive (upward) drift. I'm still working on finding a good citation for that. Think of it as flipping a biased coin. In the long run, you'll get more heads than tails, but toss by toss, the pattern looks random. Virtually all noise. I've played very little with daily returns and volatility, and not for some time, but what I recall is that it was not especially informative data.
Likewise, I'm not clear on what's being said here: It IS a circular argument as yogibearbull says
What is the argument you see M* making, and why is it circular? I don't see what you're getting at.
Regardless, volatility and risk are not one and the same. Let me try an analogy - Russian roulette where one spins the chamber after each shot.
After three, after five, after how every many shots you like, you find that you are still alive. Does that mean that there's a low risk of shooting yourself, verily, getting lower each time you survive? Or does it mean that your survival to date is not a good indicator of the risk inherent in the "game"?
"Circularity" here arises from complaining about something that may be a result of one's own doing. So, M* is complaining about a blunt tool (MPT based on MONTHLY data) of its own design not being able to detect what happens(ed) over days or weeks.
May be a term other than "circularity" can be used - English or statistics professors here? But in engineering, this is very clear - a sampling at intervals of T or more cannot detect events at intervals of less than T. Basic sampling theory.
I did look at events of early-2020 using DAILY volatilities. I found that one had an advance warning of several days, may be a couple of weeks. Whether one could/did act on that is another matter. But some institutional programs based on VaR (value-at-risk), itself controversial, may have auto-triggered.
Anyway, it would be better if the fund industry moved to MPT stats based on weekly or daily data rather than monthly data. That was my main point, and it is a suggestion that I have made to M* several times. I am sure that such proprietary or subscription tools exist now, but I can only afford freebies (-:).
In terms of risk, we may be talking about two different things. You're looking at short term predictions (early 2020 presaging March events) as exemplified by this IBM commercial. It dramatizes how Watson could predict that an elevator might fail without maintenance. Morningstar is talking about intrinsic risk in portfolios, which might, I suppose, be analogized to two different elevators, one built with reliable parts and one with third rate reconditioned parts. Or two different portfolios, one built with bonds (BND) and one built with stocks (VTI).
Certainly a higher sampling frequency would facilitate the detection of short duration events. Worth doing. Standard deviation can be used in identifying changed conditions (see, e.g. Westinghouse Rules, aka WeCo Rules) so long as the sampling rate is high enough. That's different from asking whether volatility (taken over years at any sampling frequency) is a good measure of risk. https://ats-help.com/ATS_Inspect_6_3/Variable_Data_Collect/WECO_Rules.htm
Comments
It doesn't reference the new M* categorization system using beta-based MPRS (see a post nearby), but provides good info on relative-performance, relative-beta and relative-SD.
The data lag issue may be explained by the use of monthly data for MPT stats by M*, PV, etc. If they switch to weekly data, then things would improve.
"A bet on higher yield may pay off nine years or out of ten, or even better, but that just means that when 00 comes up, the impact is likely to be more severe. IOFIX, SEMMX and their brethren were never "cash subs", regardless of how sedate they looked before 2020."
IOFIX and SEMMX were touted as "cash subs" by some participants on another investing board.
This was before IOFIX returned -36.18% and SEMMX returned -20.85% in 2020 Q1 ¹.
¹ returns reported by Morningstar
The data lag issue may be explained by the use of monthly data for MPT stats by M*
BECAUSE M* uses monthly data, its MPT measurements lack sensitivity that weekly (or daily) data may capture.
These are two different issues. Lag (or latency) is merely the delay in reporting something. Sensitivity is the ability to detect small changes. M* is not saying that there's a lag in identifying risk - unless one is suggesting that there's a multi-year lag (due to monthly sampling) because some markets have been sedate for years. Rather, M* is saying that because volatility is not the same as risk, and because there can be risk without volatility, volatility may be a poor metric for it.
On a daily basis, the market is close to a random walk with a positive (upward) drift. I'm still working on finding a good citation for that. Think of it as flipping a biased coin. In the long run, you'll get more heads than tails, but toss by toss, the pattern looks random. Virtually all noise. I've played very little with daily returns and volatility, and not for some time, but what I recall is that it was not especially informative data.
Likewise, I'm not clear on what's being said here:
It IS a circular argument as yogibearbull says
What is the argument you see M* making, and why is it circular? I don't see what you're getting at.
Certainly if one assumes that volatility and risk are one and the same, then a period where volatility is low would of necessity mean that risk is low. That's not circular, it's a tautology.
https://philosophy.stackexchange.com/questions/34409/difference-between-tautology-and-circular-reasoning
Regardless, volatility and risk are not one and the same. Let me try an analogy - Russian roulette where one spins the chamber after each shot.
After three, after five, after how every many shots you like, you find that you are still alive. Does that mean that there's a low risk of shooting yourself, verily, getting lower each time you survive? Or does it mean that your survival to date is not a good indicator of the risk inherent in the "game"?
May be a term other than "circularity" can be used - English or statistics professors here? But in engineering, this is very clear - a sampling at intervals of T or more cannot detect events at intervals of less than T. Basic sampling theory.
I did look at events of early-2020 using DAILY volatilities. I found that one had an advance warning of several days, may be a couple of weeks. Whether one could/did act on that is another matter. But some institutional programs based on VaR (value-at-risk), itself controversial, may have auto-triggered.
Anyway, it would be better if the fund industry moved to MPT stats based on weekly or daily data rather than monthly data. That was my main point, and it is a suggestion that I have made to M* several times. I am sure that such proprietary or subscription tools exist now, but I can only afford freebies (-:).
Morningstar is talking about intrinsic risk in portfolios, which might, I suppose, be analogized to two different elevators, one built with reliable parts and one with third rate reconditioned parts. Or two different portfolios, one built with bonds (BND) and one built with stocks (VTI).
Certainly a higher sampling frequency would facilitate the detection of short duration events. Worth doing. Standard deviation can be used in identifying changed conditions (see, e.g. Westinghouse Rules, aka WeCo Rules) so long as the sampling rate is high enough. That's different from asking whether volatility (taken over years at any sampling frequency) is a good measure of risk.
https://ats-help.com/ATS_Inspect_6_3/Variable_Data_Collect/WECO_Rules.htm