Larry Swedroe: No Point To Timing Factors

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
Random Walker
Posts: 4159
Joined: Fri Feb 23, 2007 8:21 pm

Larry Swedroe: No Point To Timing Factors

Post by Random Walker » Mon Feb 11, 2019 10:05 am

https://www.etf.com/sections/index-inve ... nopaging=1

Diversifying across low to negatively correlated factors with expected premiums appears to be a winning game; trying to time exposure to the factors does not.

Dave

User avatar
hdas
Posts: 1254
Joined: Thu Jun 11, 2015 8:24 am

Re: Larry Swedroe: No Point To Timing Factors

Post by hdas » Mon Feb 11, 2019 10:23 am

Random Walker wrote:
Mon Feb 11, 2019 10:05 am
https://www.etf.com/sections/index-inve ... nopaging=1

Diversifying across low to negatively correlated factors with expected premiums appears to be a winning game; trying to time exposure to the factors does not.

Dave
The ppl of AQR seem to think different: See Factor Momentum
In this article, the authors document robust momentum behavior in a large collection of 65 widely-studied, characteristic-based equity factors around the globe. They show that, in general, individual factors can be reliably timed based on their own recent performance. A time series “factor momentum” portfolio that combines timing strategies of all factors earns an annual Sharpe ratio of 0.84. Factor momentum adds significant incremental performance to investment strategies that employ traditional momentum, industry momentum, value, and other commonly studied factors. Their results demonstrate that the momentum phenomenon is driven in large part by persistence in common return factors and not solely by persistence in idiosyncratic stock performance.
"whenever there is a randomized way of doing something, then there is a nonrandomized way that delivers better performance but requires more thought" ET Jaynes

Topic Author
Random Walker
Posts: 4159
Joined: Fri Feb 23, 2007 8:21 pm

Re: Larry Swedroe: No Point To Timing Factors

Post by Random Walker » Mon Feb 11, 2019 10:34 am

My impression is that AQR people do not espouse factor timing.

https://www.aqr.com/Insights/Perspectiv ... ng-is-Hard

My impression of your quote above is that they are discussing persistence of momentum in some factors, and observing momentum in factors may well be different from timing them. Now, I’ll go read the link you provided. Also, in his article, Larry references an AQR paper : https://jpm.iijournals.com/content/43/5/72

Dave

User avatar
hdas
Posts: 1254
Joined: Thu Jun 11, 2015 8:24 am

Re: Larry Swedroe: No Point To Timing Factors

Post by hdas » Mon Feb 11, 2019 10:41 am

Random Walker wrote:
Mon Feb 11, 2019 10:34 am
My impression is that AQR people do not espouse factor timing.

https://www.aqr.com/Insights/Perspectiv ... ng-is-Hard

My impression of your quote above is that they are discussing persistence of momentum in some factors, and observing momentum in factors may well be different from timing them. Now, I’ll go read the link you provided. Also, in his article, Larry references an AQR paper : https://jpm.iijournals.com/content/43/5/72

Dave
What part of this quote from the abstract is different from timing for you?
They show that, in general, individual factors can be reliably timed based on their own recent performance. A time series “factor momentum” portfolio that combines timing strategies of all factors earns an annual Sharpe ratio of 0.84.
"whenever there is a randomized way of doing something, then there is a nonrandomized way that delivers better performance but requires more thought" ET Jaynes

Beliavsky
Posts: 932
Joined: Sun Jun 29, 2014 10:21 am

Re: Larry Swedroe: No Point To Timing Factors

Post by Beliavsky » Mon Feb 11, 2019 10:44 am

Suppose my value strategy, call it SPValue, is to own the 50% of S&P 500 stocks with P/E ratios below the median. I invest in both the S&P 500 and SPValue. Should I really have the same percentage allocation so SPValue when the P/E ratio of SPValue and S&P 500 is 14.9 and 15.1 as when the P/E ratios are much different, say 10 and 20? If you think value works because valuation measures mean-revert (the P/E ratios of high (low) P/E stocks fall (rise) toward the mean) then a value tilt becomes less attractive when valuation dispersion falls.

In an idealized case, suppose all stocks had P/E of either 14.9 or 15.1. If all the P/E ratios converge to 15.1, the difference in performance is only 0.2/15.0 = 1.3%. If stocks start with P/E ratios of either 10 or 20 and all converge to a P/E ratio of 15, the performance difference is 50% - (-25%) = 75%.
Last edited by Beliavsky on Mon Feb 11, 2019 1:32 pm, edited 1 time in total.

larryswedroe
Posts: 15989
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Larry Swedroe: No Point To Timing Factors

Post by larryswedroe » Mon Feb 11, 2019 10:57 am

To be helpful
AQR does not attempt to factor time, and Asness himself has advised against it.
While spreads between valuations determine the premium (as they get wider premiums tend to get larger) there is always a premium (it's just larger or smaller) and therefore market timing not likely to work
TS (not CS) MOM is of course a market timing factor as it shifts allocations based on recent price moves
Larry

User avatar
nedsaid
Posts: 12666
Joined: Fri Nov 23, 2012 12:33 pm

Re: Larry Swedroe: No Point To Timing Factors

Post by nedsaid » Mon Feb 11, 2019 11:09 am

Market timing is difficult because the market anticipates or at least tries to anticipate the future. Pretty much you have to outguess the other market participants and you also have the being right too early problem. In the days of Math PhDs and 24-7 computing and number crunching with powerful software and hardware, it is harder and harder to get an edge. We are also in the dawn of artificial intelligence applied to the investment world. Do you think someone like me eyeballing ratios and graphs is going to outsmart all of this?

One example of why this doesn't work is the strategy of buying the so-called defensive stocks whenever you think the economy is approaching recession. Many of these stocks are consumer staples, stuff you buy whether in good or bad times, stuff you need like soap, razors, toilet paper, etc. The idea is that the earnings from these stocks will stay steady during times of economic uncertainty. This is a well-known strategy and everyone and his brother knows about it. So these stocks get a nice bump up in price whenever the market thinks a recession is nigh.
A fool and his money are good for business.

User avatar
hdas
Posts: 1254
Joined: Thu Jun 11, 2015 8:24 am

Re: Larry Swedroe: No Point To Timing Factors

Post by hdas » Mon Feb 11, 2019 11:16 am

larryswedroe wrote:
Mon Feb 11, 2019 10:57 am
To be helpful
AQR does not attempt to factor time, and Asness himself has advised against it.
While spreads between valuations determine the premium (as they get wider premiums tend to get larger) there is always a premium (it's just larger or smaller) and therefore market timing not likely to work
TS (not CS) MOM is of course a market timing factor as it shifts allocations based on recent price moves
Larry
Well, it's interesting that they publish research in favor of timing factors but not use it.......why would they do that?

Here's the conclusion of the paper:
We document robust persistence in the returns of equity factor portfolios. This persistence
is exploitable with a time series momentum trading strategy that scales factor exposures up
and down in proportion to their recent performance.
Factor timing in this manner produces
economically and statistically large excess performance relative to untimed factors. We
aggregate individual factor timing strategies into a combined “time series factor momentum”
strategy that dominates all individual timing strategies. TSFM is complementary with stock
momentum, as both enter optimized multi-factor portfolios with significant positive weights
(particularly when combined with HML-Devil).
An interesting aspect of factor momentum is its stability with respect to the definition of
“recent” performance. Whether the look-back window is as short as one month or as long
as five years, our strategy identifies large positive momentum among factors. This contrasts
sharply with stock momentum, which exhibits reversal with respect to recent one-month
performance, momentum at intermediate horizons of around one year, and again reversal for
windows beyond two years.
"whenever there is a randomized way of doing something, then there is a nonrandomized way that delivers better performance but requires more thought" ET Jaynes

Topic Author
Random Walker
Posts: 4159
Joined: Fri Feb 23, 2007 8:21 pm

Re: Larry Swedroe: No Point To Timing Factors

Post by Random Walker » Mon Feb 11, 2019 11:17 am

larryswedroe wrote:
Mon Feb 11, 2019 10:57 am
To be helpful
AQR does not attempt to factor time, and Asness himself has advised against it.
While spreads between valuations determine the premium (as they get wider premiums tend to get larger) there is always a premium (it's just larger or smaller) and therefore market timing not likely to work
TS (not CS) MOM is of course a market timing factor as it shifts allocations based on recent price moves
Larry
Note the first chart in Larry’s article. Of 6 premia across 4 regimes, the results are negative in only 2 of the 24 combinations.

Dave

larryswedroe
Posts: 15989
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Larry Swedroe: No Point To Timing Factors

Post by larryswedroe » Mon Feb 11, 2019 11:50 am

hdas
You are mixing two things They do have TS momentum strategies and of course that is market timing. But they don't try to time factors in their other funds. So for example, in their multi-style funds (like value + momentum) funds they would not go less long value if the spreads between value and growth narrowed, nor would they increase it in when spreads widened. The research you are referring to is specific to TS MOM and has NOTHING to do with valuations and spreads, only price movements. I hope that is helpful,
Larry

User avatar
hdas
Posts: 1254
Joined: Thu Jun 11, 2015 8:24 am

Re: Larry Swedroe: No Point To Timing Factors

Post by hdas » Mon Feb 11, 2019 12:32 pm

larryswedroe wrote:
Mon Feb 11, 2019 11:50 am
hdas
You are mixing two things They do have TS momentum strategies and of course that is market timing. But they don't try to time factors in their other funds. So for example, in their multi-style funds (like value + momentum) funds they would not go less long value if the spreads between value and growth narrowed, nor would they increase it in when spreads widened. The research you are referring to is specific to TS MOM and has NOTHING to do with valuations and spreads, only price movements. I hope that is helpful,
Larry
Larry,
Kindly don't misrepresent my view. Let's go step by step:

1. Did you read the paper Factor Momentum I referenced?.......or at least the excerpts I posted?
2. Ok, If you did, you realize that they are using price movements (recent performance)...aka as TS momentum ON THE FACTORS THEMSELVES....for timing of course. Nothing wrong with this.
3. You don't necessarily need valuations or spreads to time factors, as the paper shows you can use only the factors own performance for that purpose
4. In the paper they seem to think their approach is solid and there are benefits to it.

Ok, if the 4 points above are clear, my question is:

Why would they publish something like this and not use it in their own funds? (It's a legitimate question, not rhetorical)

hdas
"whenever there is a randomized way of doing something, then there is a nonrandomized way that delivers better performance but requires more thought" ET Jaynes

HippoSir
Posts: 60
Joined: Tue Jul 03, 2018 2:56 pm

Re: Larry Swedroe: No Point To Timing Factors

Post by HippoSir » Mon Feb 11, 2019 12:42 pm

Very interesting article, thank you Larry, I had posted a similar question a few days ago (viewtopic.php?f=10&t=271697) about the new(ish) dynamic multifactor funds and this answers it nicely.

I may be biased however since it makes intuitive sense to me that factor timing would be a losing game (although perhaps possible, I just don't see sufficient data to support it, unlike factors themselves), and I've thus avoided any of these dynamic multifactor strategies.

User avatar
asset_chaos
Posts: 1646
Joined: Tue Feb 27, 2007 6:13 pm
Location: Melbourne

Re: Larry Swedroe: No Point To Timing Factors

Post by asset_chaos » Mon Feb 11, 2019 5:20 pm

Random Walker wrote:
Mon Feb 11, 2019 10:05 am
Diversifying across low to negatively correlated factors with expected premiums appears to be a winning game
Probably to the extent that holding any reasonably diversified stock portfolio and staying the course will likely produce pretty good results in the long term.
trying to time exposure to the factors does not.
Yes, factor timing seems as unlikely to succeed as any other kind of market timing.
Regards, | | Guy

Jiu Jitsu Fighter
Posts: 180
Joined: Mon Nov 21, 2016 10:22 am

Re: Larry Swedroe: No Point To Timing Factors

Post by Jiu Jitsu Fighter » Mon Feb 11, 2019 6:40 pm

This is one of the few times that I disagree with Larry. I agree that purely trying to time market factors is likely a losers game. Where I disagree with him is with regard to the RAFI funds. I don't think that constitutes factor timing since the methodology does not change. It's not like, since growth has been outperforming, those funds decide to purposely shift to value. They continue to weight the stocks in the portfolio using the same metrics which will have the fund naturally drift towards either value or growth. This organic nature is why the RAFI funds are a reasonable approach to tilt your portfolio and reduces the negative momentum and alphas associated with traditional value indices.

I think this has more to do with Larry siding Asness vs Arnott (I never understood that squabble). Even Asness admitted that the RAFI indices are, theoretically, a good way to get factor exposure. What I will agree with Larry on as a criticism of RAFI is the infrequent rebalancing that leads to style drift and inconsistent factor exposure.

Dead Man Walking
Posts: 885
Joined: Wed Nov 07, 2007 6:51 pm

Re: Larry Swedroe: No Point To Timing Factors

Post by Dead Man Walking » Tue Feb 12, 2019 1:22 am

My investing experience has taught me that the most significant market timing is simply time in the market.

DMW

typical.investor
Posts: 1223
Joined: Mon Jun 11, 2018 3:17 am

Re: Larry Swedroe: No Point To Timing Factors

Post by typical.investor » Tue Feb 12, 2019 3:35 am

Jiu Jitsu Fighter wrote:
Mon Feb 11, 2019 6:40 pm
This is one of the few times that I disagree with Larry. I agree that purely trying to time market factors is likely a losers game. Where I disagree with him is with regard to the RAFI funds. I don't think that constitutes factor timing since the methodology does not change. It's not like, since growth has been outperforming, those funds decide to purposely shift to value. They continue to weight the stocks in the portfolio using the same metrics which will have the fund naturally drift towards either value or growth. This organic nature is why the RAFI funds are a reasonable approach to tilt your portfolio and reduces the negative momentum and alphas associated with traditional value indices.

I think this has more to do with Larry siding Asness vs Arnott (I never understood that squabble). Even Asness admitted that the RAFI indices are, theoretically, a good way to get factor exposure. What I will agree with Larry on as a criticism of RAFI is the infrequent rebalancing that leads to style drift and inconsistent factor exposure.
The "Optimal Timing and Tilting of Equity Factors" paper cited is interesting. It differentiates between timing and tilting. The Rafi methodology is in the tilting camp and adjustments to weighting are capped.

The paper states:
For factor tilting we rely on six different characteristics embedded in the cross-section of
the factor set: valuation, spread, momentum, volatility and two factors based on the correlation network (note betweenness and distance-to-market).
Using this cross-sectional information the tilting-coefficients
help to slightly improve the risk-return profile of the resulting factor allocation relative to the
equal-weighted factor allocation benchmark and are able to limit the maximum drawdown
remarkably.
The factor tilting approach
delivers an excess return of 0.51 percentage points p.a. over the benchmark, whereas the
volatility is increased by 0.32 percentage points. As a result, the SR (1.25) is comparable to
the the equal-weighted benchmark (1.21). The return distribution is also more extreme, as
the minimum and maximum returns attain higher absolute values. However, the maximum
drawdown is slightly reduced from −5.3% to −4.6%, indicating that the strategy helps to
navigate some of the bad times. The strategy’s IR is 0.61.
It's a very short time (only one year) of real returns but comparing AQR:

QCELX (AQR Large Cap Multi-Style Fund) - value, momentum, quality
MFUS (RAFI Dynamic Multi-Factor) - value, momentum, quality, low volatility
OMFL (Oppenheimer Russell 1000 Dyn Mltfctr ETF)

AQR has worse performance and worse drawdowns.

Fund, CAGR, max drawdown
OMFL, 5.34%, -12.54%
MFUS, 2.00%, -14.68%
QCELX, -2.22%, -16.91%

OMFL - adjusts exposure to economic cycle and hasn't gone though a full cycle so who really knows how it'll work.
FYI, Russell 1000 was 3.60% CAGR and -13.74% max drawdown

typical.investor
Posts: 1223
Joined: Mon Jun 11, 2018 3:17 am

Re: Larry Swedroe: No Point To Timing Factors

Post by typical.investor » Tue Feb 12, 2019 8:48 am

Also, "The Economics of Factor Timing" (Haddad, Kozak, Santosh) found that:
There is another related literature which argues that cross-sectional long-short factors are not very predictable. Ilmanen and Nielsen (2015) argue that though valuation ratios (such as book-to-market) are useful for predicting asset class index returns, “this relation- ship turns out to be weaker for long/short factor portfolios.” Asness (2016) claims that predicting anomaly strategy returns using valuation ratios is fruitless since the resulting timing strategies “aren’t exceptionally strong, and they are way too correlated with the basic value strategy itself to make great impact on a portfolio.” Asness et al. (2017) use book-to-market ratios to construct market timing strategies for the size, value and betting- against-beta strategies and conclude “there is near-unanimous support in the literature for the efficacy of value investing, and thus the value (or value spread) of anything should have some predictive power for that thing. The question is the strength of this relationship. Here we find that strength rather lacking.” In contrast to these papers, we find that valuation ratios are extremely powerful for predicting long-short factor portfolios, once we focus on the common variation in expected returns uncovered through principal components analysis.
So Swedroe is saying Asness is saying it's pointless, but was he event talking about long only portfolios? Also, there is work such as above that says the opposite - that it is effective even in long/short ones. Interestingly, the above cites Asness' "Contrarian factor timing is deceptively difficult" work.
Most of the benefits of timing accrue from timing the anomaly factors rather than the aggregate market return. In addition, the composition of risks affecting the SDF changes strongly over time, at about business-cycle frequency. In other words, changes over time in the pricing of risks are not driven by one common source of variation, in contrast to many theories.5 Rather, the SDF has independently varying loadings on a few weakly correlated sources of risks.
This evidence suggests that the economic risks investors worry about conditionally are often very different from those they worry about on average.
An investor can therefore substantially en- hance her Sharpe ratio by varying exposures to anomaly strategies over time compared to simply passively holding the anomalies and collecting their unconditional risk premia.
and they conclude:
The methods and facts we study in this paper are only the beginning of the economic enterprise of understanding the evolution of drivers of risk premia
In any case, it seems like active management. Some fund will outperform but good luck finding it in advance. Take MOM for instance, there sure seems to be variation in returns for the same factor based on implementation.

Post Reply